<?page no="0"?> This volume examines interactions between second/ foreign language acquisition and the development of cognitive abilities in learners who acquire an additional language in pre-schools, primary or secondary schools. The chapters explore possible links between cognitive and linguistic skills displayed by multilingual learners. This book will appeal to different kinds of readers such as language acquisition researchers, linguists, psychologists and language teachers. ISBN 978-3-8233-8194-5 Multilingualism and Language Teaching 4 MLT 4 Piske / Steinlen (eds.) Cognition and Second Language Acquisition Multilingualism and Language Teaching 4 Thorsten Piske / Anja Steinlen (eds.) Cognition and Second Language Acquisition Studies on Pre-School, Primary School and Secondary School Children <?page no="1"?> Cognition and Second Language Acquisition <?page no="2"?> Multilingualism and Language Teaching Herausgegeben von Thorsten Piske (Erlangen), Silke Jansen (Erlangen) und Martha Young-Scholten (Newcastle) Band 4 <?page no="3"?> Thorsten Piske, Anja Steinlen (eds.) Cognition and Second Language Acquisition Studies on Pre-School, Primary School and Secondary School Children <?page no="4"?> DOI: https: / / doi.org/ 10.24053/ 9783823391944 © 2022 · Narr Francke Attempto Verlag GmbH + Co. KG Dischingerweg 5 · D-72070 Tübingen Das Werk einschließlich aller seiner Teile ist urheberrechtlich geschützt. Jede Verwertung außerhalb der engen Grenzen des Urheberrechtsgesetzes ist ohne Zustimmung des Verlages unzulässig und strafbar. Das gilt insbesondere für Vervielfältigungen, Überset‐ zungen, Mikroverfilmungen und die Einspeicherung und Verarbeitung in elektronischen Systemen. Alle Informationen in diesem Buch wurden mit großer Sorgfalt erstellt. Fehler können dennoch nicht völlig ausgeschlossen werden. Weder Verlag noch Autor: innen oder Herausgeber: innen übernehmen deshalb eine Gewährleistung für die Korrektheit des Inhaltes und haften nicht für fehlerhafte Angaben und deren Folgen. Diese Publikation enthält gegebenenfalls Links zu externen Inhalten Dritter, auf die weder Verlag noch Autor: innen oder Herausgeber: innen Einfluss haben. Für die Inhalte der verlinkten Seiten sind stets die jeweiligen Anbieter oder Betreibenden der Seiten verantwortlich. Internet: www.narr.de eMail: info@narr.de CPI books GmbH, Leck ISSN 2199-1340 ISBN 978-3-8233-8194-5 (Print) ISBN 978-3-8233-9194-4 (ePDF) ISBN 978-3-8233-0161-5 (ePub) Bibliografische Information der Deutschen Nationalbibliothek Die Deutsche Nationalbibliothek verzeichnet diese Publikation in der Deutschen Nationalbibliografie; detaillierte bibliografische Daten sind im Internet über http: / / dnb.dnb.de abrufbar. www.fsc.org MIX Papier aus verantwortungsvollen Quellen FSC ® C083411 ® www.fsc.org MIX Papier aus verantwortungsvollen Quellen FSC ® C083411 ® <?page no="5"?> 7 21 37 59 89 107 141 161 Contents Thorsten Piske / Anja Steinlen Introduction. Examining possible relations between linguistic and cognitive skills in pe-school, primary school and secondary school children . . . . . . . Part 1: Linguistic and cognitive abilities in pre-school children Nicole Biedinger Ethnic differences in early cognitive and language skills: Children from age three to ten years . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Katja Schmidt / Yvonne Blumenthal Early immersion education: L2 vocabulary acquisition and the role of non-verbal intelligence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Claire Goriot Developmental differences in executive functioning in bilingual, early-English and monolingual children: Group differences and individual differences both matter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Part 2: Linguistic and cognitive abilities in primary school children Gerda Videsott / Rita Franceschini The link between multilingualism and attention in children with and without a migrant background - Practical implications . . . . . . . . . . . . . . . . Anja Steinlen / Thorsten Piske “Does speaking a foreign language at school make your kids smarter? ” The development of non-verbal intelligence in a primary school offering an immersion and a mainstream foreign language programme . . . . . . . . . . . . . Jana Chudaske (translated by Nina Rogotzki) Competence development in multilingual primary school classes in Germany: Linguistic competence and basic cognitive skills . . . . . . . . . . . . . Holger Hopp / Teresa Kieseier / Markus Vogelbacher / Dieter Thoma Cognitive and linguistic profiles in early foreign language vocabulary and grammar . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . <?page no="6"?> 193 219 245 265 285 309 335 365 370 Part 3: Linguistic and cognitive abilities in secondary school children Tanja Angelovska Cognitive aspects of Processing Instruction . . . . . . . . . . . . . . . . . . . . . . . . . . . Dominik Rumlich Exploring the importance of prior knowledge and verbal cognitive abilities for foreign language learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Sara Dallinger (translated by Nina Rogotzki) The role of languages in bilingual History lessons and its effects on English and History achievement: Use of first language and correctness of foreign language . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Nils Jaekel Does a positive selection bias into CLIL streams explain higher language proficiency? The impact of cognitive abilities and SES on the selection process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Ewa Dąbrowska / Ashley Blake Speed of automatisation predicts performance on “decorative” grammar in second language learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Cordula Glass Collocational proficiency: The effects of ‘target language input’ and ‘age’ . Part 4: Metaphors used in the context of a putative ‘bilingual advantage’ Silke Jansen The path to bilingualism, a road to better cognitive performance? Metaphors of language learning in Cognitive Science . . . . . . . . . . . . . . . . . . . . . . . . . . . . Contributors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Subject index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 Contents <?page no="7"?> Introduction Examining possible relations between linguistic and cognitive skills in pe-school, primary school and secondary school children Thorsten Piske / Anja Steinlen The chapters in this book examine possible links between linguistic and cognitive abilities displayed by learners who learn an additional language, i. e. a second (L2) or third (L3) language in pre-schools, primary schools and secondary schools. Since the beginning of the 20 th century, a large number of studies have examined cognitive abilities such as metalinguistic awareness, intelligence, mental flexibility and cognitive/ executive control shown by learners learning more than one language at home or in different educational institutions. The results of many of these studies suggest that bior multilingual children may show advantages over monolingual children in terms of learning and memo‐ risation skills (e.g. Kuska, Zaunbauer & Möller, 2010). Multilingual children have also been reported to show advantages over monolingual children in studies, for example, examining executive functions, concentration, attention and metalinguistic awareness (e.g. Bialystok, 1999; Bialystok & Martin 2004). Furthermore, the results of several studies examining adults suggest that bior multlingualism may have long-term effects on adults’ cognitive skills and that it can even be associated with a substantial delay in Alzheimer’s disease and Mild Cognitive Impairment (e.g. Bialystok, Craik, Binns, Ossher & Freedman, 2014). The extent to which multilinguals show advantages over monolinguals in cognitive skilks appears to be dependent on different factors, e. g. the amount of exposure to different languages, the level of competence reached in the languages learnt and the amout of actual language use (e.g. Festman & Kersten, 2010). The assessment of possible links between multilingualism and cognitive ability has changed quite drastically from the early 19 th century until today <?page no="8"?> (e.g. Baker, 2001; Jansen, Higuera del Moral, Barzen, Reimann & Opolka, 2021). During a first phase, which lasted until the 1960s, it was often claimed that multilingualism would be a risk for children’s mental and emotional develop‐ ment. According to some authors, multilingualism was a burden on the brain, would inhibit the acquisition of the majority language spoken in a country, and could lead to mental confusion, identity conflicts, split loyalties and even schizophrenia (see e. g. the review in Baker, 2001 or Bialystok, 2015). During the second phase, which began in the 1960s, attitudes towards the interaction between multilingualism and cognitive skills changed completely. An article by Peal & Lambert (1962) inspired a lot of research suggesting that the co-existence of two language systems in a human mind would not lead to mental confusion, but rather to increased mental flexibility. Peal & Lambert (1962) had examined 10-year-old monolingual French and bilingual French/ English children in Canada in several verbal and non-verbal intelligence tests. They found that bilingual French/ English children, who were more or less equally proficient in both languages, achieved higher scores than monolingual children in a number of tasks, in particular in those that required a high degree of mental manipulation and reorganisation of visual stimuli, concept formation and symbolic flexibility. According to the authors, the results of their study suggested that bilingual children “have a language asset, are more facile at concept formation, and have a greater mental flexibility” (Peal & Lambert, 1962, p. 1). In a third phase starting around the end of the 1990s (see e. g. Bialystok, 1999; Miyake, Friedman, Emerson, Witzki, Howerter & Wager, 2000), a large number of increasingly detailed experimental studies focused on individual aspects of multilingualism and examined in which way monolingual and bilingual individuals differ in the development of executive functions. The term executive functions is used to refer to basic cognitive processes such as attentional control, cognitive inhibition, inhibitory control, working memory, and cognitive flexibility. In addition, there are higher-order executive functions such as planning, reasoning and problem-solving (e.g. García-Madruga, Gómez-Veiga & Vila, 2016). Many studies have reported that multilinguals show advantages over monolinguals in the development of executive functions, and that this ‘bilingual advantage’ in executive function is related to degree of bilingual experience, i. e. to increased language proficiency in the different languages spoken by a multilingual (for a review, see e. g. Bialystok, 2015). The advantages shown by multilinguals in executive function tasks are often explained by referring to the simultaneous activation of more than one language in multilinguals. According to Bialystok (2015, p. 118), “bilingualism trains executive function through its 8 Thorsten Piske / Anja Steinlen <?page no="9"?> constant recruitment for language selection.” Bialystok (2015) also points out that an enhancement of executive function is not trivial because some studies have identified it as a major predictor of academic success (e.g. Best, Miller & Naglieri, 2011) and that academic success may predict long-term health and well-being (e.g. Duncan, Ziol-Guest & Kalil, 2010). Finally, it has to be noted that more recently, an increasing number of researchers have also questioned possible cognitive benefits of bior mul‐ tilingualism, because different studies have produced conflicting evidence regarding the relationship between multilingualism and cognitive skills (see e. g. Duñabeitia, Hernández, Antón, Macizo, Estévez, Fuentes & Carreiras, 2014). According to Laine & Lehtonen (2018, no page number), mixed results obtained in studies examining the possible enhancement of executive function through multilingualism may be due to three major problem areas: “the use of research designs that are weak for the task at hand, the lack of a detailed theory on how bilingual experience would modify cognition, and the employment of measures of bilingual behavior and executive functions that are troublesome”. The two authors emphasise that for a long time, research had focused too much on a possible bilingual advantage and that more attention should be given to individual features of bilingual experience likely to be responsible for putative cognitive effects. Moreover, Cox, Bak, Allerhand, Redmond, Starr, Deary & MacPherson (2016) point out that it is still unclear whether learning other languages indeed improves executive functions or whether people with better executive abilities are simply more likely to become bior multilingual. The chapters in this volume do not only address possible links between linguistic and cognitive development, but they also examine effects of other var‐ iables such as linguistic processing, socio-economic status, ethnic background, minority vs. majority language status, type of school programme and prior L2 knowledge on research outcomes. Altogether, this volume consists of four parts: Each of the first three parts focuses on learners of a different age group, i. e. pre-school, primary and secondary school children. The fourth part consists of only one chapter that takes a critical look at the way in which certain metaphors are used in a large number of scientific papers in order to construct an appealing story of a putative ‘bilingual advantage’, which is sometimes hastily interpreted in terms of its possible relevance for educational, cultural and even medical policies. Part 1: Linguistic and cognitive abilities in pre-school children Part 1 starts with a chapter that addresses ethnic differences in early cognitive and language skills in migrant children living in Germany. Nicole Biedinger presents the results of a study examining children with a Turkish background 9 Introduction <?page no="10"?> whose cognitive and language development was followed from the beginning of pre-school until the end of primary school. The main finding of the study, i. e. that ethnic differences in cognitive skills (in this case, sequential and simul‐ taneous processing skills) decrease with age whereas differences in language skills (in this case, German language skills) persist, are discussed by referring to the children’s social background, the early pre-school education they received and their home environment. In chapter 2, Katja Schmidt and Yvonne Blumenthal turn to German-English bilingual pre-schools and examine a possible connection between non-verbal intelligence and receptive L2 vocabulary knowledge. The results of their study suggest that in bilingual pre-school programmes, children with a high (i.e. an above-average) level of non-verbal intelligence may progress much faster in receptive L2 vocabulary acquisition than children with low and average non-verbal intelligence. According to the authors, the children with a high level of non-verbal intelligence may have shown advantages in receptive L2 vocabulary acquisition in their study because they appeared to demonstrate a relatively high level of metalinguistic awareness and seemed to use specific language learning strategies that the children with low and average non-verbal intelligence did not seem to use. Claire Goriot closes Part 1 by examining possible relations between lan‐ guage development (Dutch and English receptive vocabulary) and executive functioning development (switching and working memory). She presents data obtained from children of three age groups (4-5, 8-9, and 11-12 years), who are Dutch-English bilinguals, Dutch children learning English from kindergarten onwards or Dutch monolinguls. The bilingual children are reported to have developed greater knowledge of English vocabulary than the children in the other two groups, whereas no group differences were shown with regard to their Dutch vocabulary knowledge. Moreover, the bilingual and the monolingual children are reported to have shown more advanced executive functioning skills than the children learning English from kindergarten onwards. According to Claire Goriot, the results obtained in her study suggest that both early bilinguals’ language and excutive functioning development is different from that of monolingual children and from children who have only little exposure to a language other than their first language (L1). However, she also points out that her study in general produced relatively mixed results, which supports the assumption that it is rather difficult to operationalise bilingualism. Part 2: Linguistic and cognitive abilities in primary school children The second part of this volume examines different factors that may have an influence on primary school children’s linguistic development. In chapter 10 Thorsten Piske / Anja Steinlen <?page no="11"?> 4, Gerda Videsott and Rita Franceschini compare attention mechanisms in six-year-old children growing up in the multilingual (German, Italian, Ladin) region of South Tyrol in northern Italy. They asked children without a migrant background, whose L1 was either German or Italian, and children with a migrant background, whose L1 was a language other than German or Italian, to answer questions of a short language background questionnaire and to take the Attentional Network Test (ANT). This test examines alerting, orienting, and executive control/ conflict. Independent of the ANT component tested (i.e. alerting orienting, conflict), children without a migrant background were found to respond more accurately but slower than their peers with a migrant back‐ ground, who responded faster but less accurately. According to Videsott and Franceschini, the results of their study support the assumption that experience with multilingual surroundings has an influence on neurocognitive processes and, in particular, on the mechanism of attention. The two authors also point out that the two types of reactions shown by children with and without migrant backgrounds may have both advantages and disadvantages and that in the classroom, different didactic interventions may help children to learn to answer more quickly or to respond more accurately. Whereas Gerda Videsott and Rita Franceschini focus exclusively on the first year of primary school in chapter 4, chapter 5 examines children in year 1 and in year 4, which is the last year of primary school in most federal states in Germany. In chapter 5, Anja Steinlen and Thorsten Piske explore possible effects of non-verbal intelligence on German and English reading comprehension skills shown by children attending either a partial immersion (IM) programme or a regular foreign language (FL) programme. With regard to non-verbal intelligence, significant differences between the children in the two teaching programmes were found for children in year 4, but not for those in year 1. Moreover, teaching programme, parental background (education and relative wealth), gender and language background (majority vs. minority language children) were not identified as significant predictors of non-verbal intelligence in year 4. Finally, whereas no significant differences were found between the IM and FL children in the German reading test, the IM children outperformed the FL children in the English reading test. According to the authors, the results of their study support the findings of previous research that a) irrespective of language background and gender IM students perform better in tests examining foreign language skills than FL students and b) that there may be a positive relationship between non-verbal intelligence and L1 as well as L2 reading comprehension. Like Videsott & Franceschini and Steinlen & Piske, Jana Chudaske also exam‐ ines non-migrant children whose L1 is a majority language and migrant children 11 Introduction <?page no="12"?> growing up with at least one minority language. In Germany, children from mi‐ nority language backgrounds have repeatedly been found to score significantly lower in tests examining their academic achievements and their achievements in the majority language German. In chapter 6, Chudaske examines possible effects of language background and fluid intelligence relating to children’s ability to think abstractly, reason quickly and problem solve independent of any previously acquired knowledge in the majority language German. As in previous research, the migrant children examined by Chudaske are reported to have scored lower in five out of six subtests of the German language test. However, the effect of migrant status on the results obtained in the German test and on teachers’ assessments of the migrant and non-migrant children’s competence in the German language was considerably reduced when basic cognitive skills were controlled for. This is why Chudaske concludes that studies examining the relationship between migrant status/ language background and linguistic achievements should always also examine the relative contribution of cognitive variables to the results obtained in tests examining linguistic skills. In contrast to chapters 4, 5 and 6, chapter 7, as the concluding chapter of part 2, focuses not only on one cognitive variable such as attention, non-verbal intelligence or fluid intelligence, but it examines the role different cognitive factors (i.e. non-verbal intelligence, working memory, phonological awareness and executive control) may play in primary school children’s multilingual development. Holger Hopp, Teresa Kieseier, Markus Vogelbacher and Dieter Thoma examine the relative impacts of cognitive and linguistic factors on the English vocabulary and grammar skills acquired by minority and majority language students learning English as an L2 or an L3 in grades 3 and 4 of primary school. The results clearly suggest that both cognitive and linguistic factors contribute to the development of primary school children’s English skills, but that their relative impacts may differ according to linguistic domain (e.g. vocabulary or grammar), mode (i.e. productive vs. receptive skills), stages in development and group (i.e. majority language students vs. minority language students). Among many other things, the results presented by Hopp et al. suggest that majority and minority language children may differ in the way cognitive and linguistic factors affect foreign language achievement and that for minority language children proficiency in the majority language becomes more important over time than L1 proficiency in foreign language learning. Part 3: Linguistic and cognitive abilities in secondary school children The third part of this volume comprises of six chapters that examine linguistic and cognitive abilities in secondary school students. In the initial chapter of this part, Tanja Angelovska reviews theoretical and empirical research on 12 Thorsten Piske / Anja Steinlen <?page no="13"?> Input Processing in order to explore the role of cognitive aspects in Processing Instruction as the pedagogical application of Input Processing. She also presents the results of an exemplary study that examines in which way Processing Instruction may affect both school-age (M age ≈ 10.5 years) and adult (M age ≈ 26 years) native German learners’ acquisition of the English simple past tense marking -ed. According to the results obtained in this study, Processing Instruction has positive effects on both both school-age and adult learners’ acquisition of L2 grammar, but older learners may show greater improvements than younger learners, which may, among other things, be due to differences in the processing capacity of older and younger learners. In chapter 9, Dominik Rumlich explores the role of prior L2 knowledge, cognitive abilities and demographic factors in the development of L2 proficiency in a formal classroom setting. He presents the results of a pre-post study investigating students’ ‘prior’ knowledge of English as a foreign language in a pre-test in year six (M age ≈ 12 years) and their ‘final’ proficiency in English as a foreign language as measured in a post-test one year later. The findings obtained in the study suggest that whereas prior knowledge of the foreign language and verbal cognitive abilities are both highly relevant for the development of students’ general English as a foreign language proficiency, demographic factors such as sex are less important predictors of students’ L2 proficieny. Based on the results of his study, Dominik Rumlich calls for an incorporation of factors such as prior L2 knowledge and verbal cognitive abilities into studies on the development of classroom students’ general foreign language proficiency. Sara Dallinger also examines students learning an L2 in a bilingual institution, and in particular, in secondary schools following a CLIL (Content and Language Integrated Learning) approach. Although students’ figural and verbal cognitive skills have been assessed in the study, her major focus is not on cognitive factors but on the role language use factors play in CLIL students’ development of L2 proficiency in English and of their subject knowledge in history. Particular at‐ tention is paid to teachers’ use of the L1 German in different teaching situations (e.g. when new technical terms are introduced) and to teachers’expectations of students’ correct use of the foreign language English. The results of Dallinger’s study indicate that in a CLIL context L1 use may have positive effects on learning technical terms and that higher expectations of students’ correct use of the L2 may positively affect both their L2 proficiency and their competence in a subject such as history. As the author points out, her results may have important implications for the development of a specific CLIL methodology, which still appears to be largely non-existent. 13 Introduction <?page no="14"?> Chapter 11 continues to examine factors that may have an influence on CLIL students’ achievements. Nils Jaekel points out that in countries such as Germany, schools offering CLIL programmes often select their students on the basis of a variety of high achievement criteria, which may be one of the reasons why CLIL students have in most studies been found to achieve higher levels of L2 English proficiency than students in mainstream English as a foreign language (EFL) classes. In his chapter, Jaekel examines to what extent 9 th grade CLIL students’ above average L2 proficieny is indeed due to positive selection biases into CLIL streams. He compares the composition of CLIL versus EFL classes with regard to variables such as gender, age, socioeconomic status, home language and cognitive abilities, and he measures the students’ English language proficiency by considering their last grade in English and by using a C-Test battery. The CLIL and EFL streams examined by Jaekel were indeed found to differ significantly with regard to age, socio-economic status and cognitive abilities, and the CLIL students performed significantly better on the C-Tests than their EFL peers. However, controlling for the positive selection bias into CLIL stream classes did not change the strong positive effect of CLIL on L2 proficiency. According to Jaekel, this finding clearly suggests that the specific characteristics of the CLIL approach, i. e. extensive L2 exposure and authentic content-related communication, are a ‘key factor’ in predicting high L2 proficiency scores. In chapter 12, Ewa Dąbrowska and Ashley Blake explore the relationship between the acquisition of two types of grammar, i. e. ‘decorative grammar’ (aspects of grammar which have abstract and largely redundant meanings) and ‘functional grammar’ (aspects of grammar that provide a clear contribution to meaning), and two cognitive abilities, i. e. the ability to automatise a complex cognitive procedure and explicit language aptitude. They report on the results of a study examining 36 10 th grade students (M age ≈ 15.5 years) from a Grammar School in Germany who were learning English as a foreign language and who were asked to complete a) a grammaticality judgment task used to assess performance on decorative grammar, b) a picture selection task used to assess performance on functional grammar, c) a language analysis task used to test explicit language aptitude and d) the Multiple Tower-trial of Hanoi task used to assess the ability to automatise a complex cognitive procedure. According to Dąbrowska and Blake, the results of their study suggest that speed of automatisation is more strongly associated with decorative grammar, whereas explicit language aptitude appears to play a role in the acquisition of both decorative and functional grammar. The two authors also provide an interesting discussion of the validity of their different measures. 14 Thorsten Piske / Anja Steinlen <?page no="15"?> Part 3 finishes with a chapter in which Cordula Glass shares insights into her research on adult (M age ≈ around 21 years) and adolescent (M age ≈ around 15 years) native and non-native speakers’ performance in a test on collocational competence. In Glass’ study, highly educated adult native speakers of English performed close to ceiling in the test on collocational competence in English. Adolescent native speakers of English, on the other hand, scored significantly lower than adult native speakers of English, which according to Glass suggests that even at the age of 15 years, the L1 acqusition of collocations is not complete. Moreover, L2 speakers with a high amount of exposure to the L2 English were found to perform equally well in the test on collocations as L1 speakers of the same age group. According to the author, this finding may indicate that a speaker’s L1 background is a less influencing factor than the amount and quality of exposure to the L2. Part 4: Metaphors used in the context of a putative ‘bilingual advantage’ Part 4 of this volume consists of only one chapter. This chapter is different from all the other chapters in this book, because it does not include data from different groups of L1, L2 or L3 speakers, but it provides an analysis of metaphorical language used in studies examining the so-called ‘bilingual advantage’. Silke Jansen points out that scientific papers focusing on possible cognitive differences between monolinguals and bilinguals largely rely on two conceptual metaphors, i. e. ‘language learning is moving forward on a predetermined path’ and ‘language learning is being exposed to an active substance’. She describes how authors who use a ‘pathforce schema’ of linguistic and cognitive development often adopt a rather ideological perspective and construe bilinguals as high performing, healthy individuals who act as ‘ideal economic subjects’. Silke Jansen’s chapter represents a reasonable and adequate warning that we should not overinterpret the findings of research examining the relationship between linguistic and cognitive development and leap to hasty conclusions regarding the implications these findings may supposedly have for educational, cultural and maybe even medical policies. This book could not have been completed without the support we received from different colleagues. We would like to thank Nina Rogotzki for translating two of the chapters included here. Special thanks also go to our editorial assistants Jessica Schmidt, Anna-Christin Schrötter and, in particular, Jessica Westhues. Unfortunately, it took us much longer than we had originally planned to finish working on this volume and some of the contributors have in the meantime published additional interesting studies on the relationship between multilingualism and cognitive skills. It is mainly due to Jessica Westhues’ tireless support that it was finally possible to publish the studies included in this book. 15 Introduction <?page no="16"?> References Baker, C. 2001. Foundations of Bilingual Education and Bilingualism. 3 rd ed. Clevedon: Multilingual Matters. Best, J.R., Miller, P.H. & Naglieri, J.A. 2011. Relations between executive function and academic achviement from ages 5 to 17 in a large, representative national sample. Learning and Individual Differences, 21 (4), 327-336. Bialystok, E. 1999. Cognitive complexity and attentional control in the bilingual mind. Child Development, 70 (3), 636-644. Bialystok, E. 2015. Bilingualism and the development of executive function: The role of attention. Child Bilingual Perspectives, 9 (2), 117-121. Bialystok, E. & Martin, M.M. 2004. Attention and inhibition in bilingual children: Evidence from the dimensional change card sort task. Developmental Science, 7 (3), 325-339. Bialystok E., Craik F.I.M., Binns M. A., Ossher L. & Freedman M. 2014. Effects of bilingualism on the age of onset and progression of MCI and AD: Evidence from executive function tests. Neuropsychology, 28 (2), 290-304. Cox, S.R., Bak, T.H., Allerhand, M., Redmond, P., Starr, J.M., Deary, I.J. & MacPherson, S.E. 2016. Bilingualism, social cognition and executive functions: A tale of chickens and eggs. Neuropsychologia, 91, 299-306. Duñabeitia, J.A., Hernández, J.A., Antón, E., Macizo, P. Estévez, A., Fuentes, L.J. & Carreiras, M. 2014. The inhibitory advantage in bilingual children revisited: Myth or reality? Experimental Psychology, 61 (3), 234-251. Duncan, G.J., Ziol-Guest, K.M. & Kalil, A. 2010. Early childhood poverty and adult attainment, behavior and health. Child Development, 81 (1), 306-325. Festman, J. & Kersten, K. 2010. Kognitive Auswirkungen von Zweisprachigkeit. In: Massler, U. & Burmeister, P. (Eds.), CLIL und Immersion: Fremdsprachlicher Sachfa‐ chunterricht in der Grundschule. Braunschweig: Westermann. 38-52. García-Madruga, J.A., Gómez-Veiga, I. & Vila, J.Ó. 2016. Executive functions and the improvement of thinking abilities: The intervention in reading comprehension. Frontiers in Psychology, 7 (58). doi: 10.3389/ fpsyg.2016.00058. Jansen, S., Higuera del Moral, S., Barzen, J.S., Reimann, P. & Opolka, M. 2021. Demystifying Bilingualism. How Metaphor Guides Research Towards Mystification. Cham: Palgrave Macmillan. Kuska, S., Zaunbauer, A. C.M. & Möller, J. 2010. Sind Immersionsschüler wirklich leistungsstärker? Ein Lernexperiment. Zeitschrift für Entwicklungspsychologie und Pädagogische Psychologie, 42 (3), 143-153. 16 Thorsten Piske / Anja Steinlen <?page no="17"?> Laine, M. & Lehtonen, M. 2018. Cognitive consequences of bilingualism: Where to go from here? Language, Cognition and Neuroscience, 33 (9). doi: 10.1080/ 23273798.2018.1462498. Miyake, A., Friedman, N.P., Emerson, M.J., Witzki, A.H., Howerter, A. & Wager, T.D. 2000. The unity and diversity of executive functions and their contributions to complex “frontal lobe” tasks: A latent variable analysis. Cognitive Psychology, 41 (1), 49-100. Peal, E. & Lambert, W.E. 1962. The relation of bilingualism to intelligence. Psychological Monographs: General and Applied, 76 (27), 1-23. 17 Introduction <?page no="19"?> Part 1: Linguistic and cognitive abilities in pre-school children <?page no="21"?> Ethnic differences in early cognitive and language skills: Children from age three to ten years Nicole Biedinger Abstract Profound ethnic differences in cognitive and language skills have been reported in many studies. However, only a few studies have dealt with the early stages of such developments. This article introduces the project “Preschool Education and Educational Careers among Migrant Children” (ESKOM-V), which involved 1250 families in Germany with a Turkish migration background and without any migration background. It is the first study on early ethnic educational inequality in Germany and also the first longitudinal study which takes into consideration the period from pre-school to the end of primary school, including the choice for the secondary school track. The main goal of this project is to account for ethnic differences regarding pre-schoolers’ language and cognitive skills in Germany and to determine how these skills affect school achievements and educational decisions later in life. The present study focuses on children with a Turkish background in Germany, whose cognitive and language development was followed from age three to ten. The results indicate that ethnic differences in cognitive skills decrease with age, whereas differences in language skills persist. These findings are discussed in the light of the children’s social background, early pre-school education, and a stimulating home environment. This chapter will conclude with some practical implications and further challenges for future research. 1 Introduction In most Western countries ethnic educational inequality is a well-established phenomenon (Heath & Brinbaum, 2007). As a major cause for this discrepancy, it is often pointed out that immigrant children’s language skills in the school language (in this case: German in Germany) are not age-appropriately developed <?page no="22"?> (Kristen, Edele, Kalter, Kogan, Schulz, Stanat & Will, 2011). This disadvantage in host country language abilities already exists in early childhood even before these immigrant children attend primary school (Niklas, Schmiedeler, Pröstler & Schneider, 2011). Due to processes of cumulation of skills over time (Heckman, 2006), these early differences can lead to disadvantageous positions for children in their later lives, for example regarding their educational or occupational careers. Germany is one of the countries in which rather high performance discrep‐ ancies between children with and without an immigration background have been observed (Stanat & Christensen, 2006). At the beginning of the 21 st century, about one-quarter of all fourth graders in Germany grew up in families with an immigration background (Kristen et al., 2011), but as of 2014, this number rose to 31 % (Statistisches Bundesamt, 2017). Given this growing number, it is a pressing political and societal task to reduce any educational discrepancies due to ethnic background. Adequate educational and occupational prospects among all groups of society are not only essential to guarantee individuals the same quality and quantity of educational opportunities but are also important on a societal level, e. g. with regard to the future economic potential of Germany (Hinte, Rinne & Zimmermann, 2012). Thus, to enhance a smooth integration of children from families with an immigration background, it is crucial to examine conditions and activities that are likely to improve their receiving country language proficiency and consequently to promote their future life prospects. For example, within the family context, engaged parent-child interactions during early childhood, which create a stimulating home environment (e.g. reading aloud to children, telling stories, or playing), positively influence children’s cognitive and language development (e.g. Crosnoe, Leventhal, Wirth, Pierce, Pianta & NICHD Early Child Care Research Network, 2010; Forget-Dubois, Dionne, Lemelin, Pérusse, Tremblay & Boivin, 2009; Melhuish, Phan, Sylva, Sammons, Siraj-Blatchford & Taggart, 2008; Raviv, Kessenich & Morrison, 2004). Recent research on ethnic differences in cognitive and linguistic abilities of young children is quite rare in Germany, though. Especially studies on achievement within the pre-school sector, particularly comparisons of achievements by different ethnic groups, are lacking (Dubowy, Ebert, von Maurice & Weinert, 2008). However, in the last years, some large-scale studies have been set out to fill this gap. This chapter will introduce such a project conducted in Germany, and it will describe the development of cognitive and language skills by children between the age of three and ten years. Following the presentation of the project (section 2) and its results (section 3), the findings will be discussed in relation 22 Nicole Biedinger <?page no="23"?> to other large-scale national and international datasets (section 4), pointing to possible explanations for the ethnic differences. This paper concludes with some suggestions for future research in section 5. 2 Method The project “Preschool Education and Educational Careers among Migrant Children” (Erwerb von sprachlichen und kulturellen Kompetenzen von Migran‐ tenkindern, ESKOM-V) deals with German and immigrant (in this case: Turkish) children’s development of cognitive and language abilities. It was carried out at the Mannheim Centre for European Social Research at the University of Mannheim (Germany) and funded by the German Research Foundation (Deutsche Forschungsgemeinschaft, DFG). German families and families of Turkish origin with a threeto four-year-old child were randomly selected from registration offices in thirty cities and communities of a local region in South-West Germany. Within the context of the study, all families with at least one of the child’s parents or grandparents being born in Turkey were considered as having a Turkish immigration background. This rather broad definition of ‘immigrant background’ was used because studies have shown that in Germany, even third-generation immigrant children at pre-school age have lower language skills than native German children (e.g. Becker, 2011). A computer-assisted personal interview was conducted with the child’s pri‐ mary caregiver, 95 % of them were mothers. The interviews comprised questions concerning the child’s family activities, pre-school attendance, the social and cultural capital of the parents, the demography of the family members, as well as specific questions concerning the immigrant families’ migration history and background. After the interview, the German version of the standardised developmental test “Kaufman Assessment Battery for Children” (K-ABC) was conducted with the child (Melchers & Preuß, 2001) to assess children’s cognitive skills as well as their German language skills. Because the interviewers were fluent in both languages, it was possible to administer both the interview and the K-ABC instructions in either German or Turkish, depending on the preferred language of the families. The project was designed as a longitudinal study, following children and their families from age three to ten. Overall, 1283 children with their families participated at the first time of testing (wave 1) in 2007. The response rates were 69 % in families of Turkish origin and 63 % in non-immigrant (aka German) families, respectively. Because of local limitation, the sample is not representa‐ tive of the population in Germany of Turkish origin or non-immigrant families 23 Ethnic differences in early cognitive and language skills <?page no="24"?> with a three-year-old child. The project continued until all children were about ten years old and left for secondary school (wave 6). The children were tested five times, and about 1000 children participated each time (for details see Becker, Biedinger, Klein & Koch, 2017). In this study, however, only the data of the first five waves (children between three and eight years) are reported. German language abilities The K-ABC includes a sub-test on ‘expressive vocabulary’ which is an adap‐ tation of the ‘picture vocabulary’ subtest of the Stanford-BINET (Melchers & Preuß, 2001), consisting of 24 tasks. The children were shown pictures of objects and were asked to name them. Although Turkish could be used for test instructions, all children had to provide the answers in German. One point was given for each correct answer, and the raw score of correct answers was used as the dependent variable in the analyses. Cognitive abilities The K-ABC includes several subtests for measuring cognitive abilities. This study focussed on gestalt closure, number recall and numeracy, which was assessed in each of the testing times. These subtests measured the children’s sequential processing and their simultaneous processing skills (for more details, see Melchers & Preuß, 2001). In the task for gestalt closure, an inkblot drawing was shown to the child and the depicted object (e.g. a butterfly) had to be identified and named, with 24 items. In the number recall task, the child repeated a series of digits (e.g. 5-1-4-2), which were read aloud by the tester. Only such digits were used that are monosyllabic in German and in Turkish. A maximum of 19 points could be reached. With respect to the numeracy task, the child answered 37 questions with very basic mathematical content, e. g. “How many elephants do you see? ” or “Are there more elephants or tigers? ”. Early school abilities In grade 3 (wave 5), two tests were added to assess the children’s German reading skills (i.e. Würzburger Leise Leseprobe Revidiert, WLLP, Schneider, Blanke, Faust & Küspert, 2011) and their math abilities (i.e. Deutscher Mathematiktest für dritte Klassen, DEMAT 3+, Roick, Gölitz & Hasselhorn, 2004): The WLLP is a German test for reading fluency at word level. The children silently read a word which they matched with one of four given pictures. Within three minutes the children had to identify as many words as possible out of a total of 140 items. The DEMAT-3+ includes mathematical tasks specific to grade 3. Because of time restrictions and curricular differences between schools and the different Federal States of Germany, only the subtest ‘arithmetics’ was used to assess the 24 Nicole Biedinger <?page no="25"?> 1 The vocabulary test within the K-ABC includes 24 different words and is designed for children between the age of two and a half and five years. Because the focus of this project was on longitudinal language and cognitive development (especially for the Turkish children), this subtest was also carried out with older children. Therefore, a substantial number of children answered all questions correctly. Because of these ceiling effects, it is not possible to differentiate between good or extremely well performers. children’s abilities to add, subtract, and multiply. The children had ten minutes to complete the tasks. 3 Results This section presents the main results relating to ethnic differences in the development of cognitive and language skills of children between three and eight years. Table 1 provides an overview of the mean values as obtained for the tests for the children with a Turkish and a German background for each of the five waves. As Table 1 shows, significant differences between Turkish and German children regarding their German language abilities (vocabulary) are found for the waves 1-4. This group difference becomes smaller in wave 5 but remains significant, with some children obtaining ceiling effects in the K-ABC. 1 In general, the Turkish children start out with poorer results but eventually catch up with the German group. In contrast, the German children already have good results at the beginning of testing and continue to perform at this high level. With regard to cognitive abilities, the results, as shown in Table 1, are less con‐ sistent because they differ from subtest to subtest. The German group outper‐ formed the Turkish group in the subtest gestalt closure, with significant group differences in wave 1, 2, 4, and 5 but not in wave 3. Independent of their language background, the children’s cognitive abilities develop as a function of age, and group differences also change over time: The differences between Turkish and German children are more pronounced at the age of four years (wave 1) than at eight years of age (wave 5). For the subtest number recall, no significant differ‐ ences between Turkish and German children are noted for waves 1-4. However, Turkish children outperform German children at the age of eight years (wave 5), with significant differences between the groups. Finally, the results for the test on numeracy are consistent. Throughout waves 1-5, the German children outperform the Turkish children, with group differences being significant, in‐ dependent of the children’s age. 25 Ethnic differences in early cognitive and language skills <?page no="26"?> German Turkish Mean SD Mean SD Wave 1 Gestalt closure 6.16 3.03 4.52* 2.89 (4 years) Number recall 4.96 2.48 4.81 2.37 Numeracy 5.87 3.49 4.04* 3.42 Vocabulary 13.72 2.99 4.24* 4.38 Age 42.05 3.73 42.06 3.90 N 556 544 Wave 2 Gestalt closure 10.33 3.46 9.74* 3.90 (5 years) Number recall 7.23 2.24 7.53 2.59 Numeracy 11.61 4.04 9.58* 4.80 Vocabulary 18.03 3.07 10.50* 5.28 Age 54.61 3.95 54.79 3.97 N 580 525 Wave 3 Gestalt closure 13.97 2.86 13.53 3.23 (6 years) Number recall 9.11 2.34 9.40 2.80 Numeracy 17.35 4.18 16.24* 4.35 Vocabulary 20.49 2.26 15.50* 4.51 Age 72.29 4.16 73.58* 4.30 N 564 481 Wave 4 Gestalt closure 16.38 2.36 15.81* 2.60 (7 years) Number recall 10.38 2.40 10.43 2.56 Numeracy 22.24 3.82 21.10* 4.01 Vocabulary 21.58 1.73 17.56* 3.58 Age 85.06 4.09 86.04* 4.27 N 551 435 Wave 5 Gestalt closure 18.99 2.11 18.55* 2.48 (8 years) Number recall 11.63 2.28 12.11* 2.52 Numeracy 28.59 2.55 27.70* 2.64 Vocabulary 22.82 1.32 20.73* 2.36 Age 109.55 4.71 111.06* 5.18 N 510 395 Table 1: Raw scores of the tests for cognitive abilities (gestalt, recall, numeracy) and language abilities (vocabulary). SD refers to standard deviations, N to sample size, *indicates that mean group differences are statistically significant at the 1 %-level (t-test). In wave 5 the children in grade 3 also completed the reading test WLLP and the math test DEMAT to assess their academic success in school. The results are presented in Table 2. 26 Nicole Biedinger <?page no="27"?> German children Turkish children Mean SD Mean SD WLLP 87.44 22.35 79.12* 21.35 DEMAT (add) 2.36 1.35 2.15 1.41 DEMAT (sub.) 1.53 1.52 1.26* 1.43 DEMAT (mul.) 2.28 1.35 2.09 1.32 DEMAT (total) 6.17 3.04 5.50* 3.16 N 510 395 Table 2: Raw scores of the tests for the school tests for reading (WLLP) and maths (DEMAT, subdivided for addition, subtraction and multiplication). * indicates that the mean differences are statistically significant at the 1 %-level (t-test). As Table 2 shows, a comparison of the results revealed significant differences between German and Turkish children, with the German group outperforming the Turkish group in the WLLP and the DEMAT. However, in the subtests on addition and multiplication, the group differences were not significant. 4 Discussion In the following, several factors are discussed which have been shown to account for the differences between children with a German and a Turkish background regarding their development of cognitive and language skills between three and eight years of age. Cognitive abilities Many studies have shown that children of immigrants perform worse in cogni‐ tive tests than native children (e.g. Glick, Baes & Yabiku, 2009). These differences extend to the first as well as to the second generation (see also Biedinger, 2007; de Feyter & Winsler, 2009). In studies focusing on Germany, three-year-old children with a Turkish background achieved lower scores in cognitive tests, whereas only one year later they performed even slightly better than comparable native German children (Becker, 2010a, 2010b). A similar result was found in the present study, namely that German children generally outperformed Turkish children in cognitive tests on gestalt closure and numeracy, independent of the children’s age. However, no such group differences were noted for the test on number recall. 27 Ethnic differences in early cognitive and language skills <?page no="28"?> 2 The home environment includes factors such as a safe and well-organised physical environment, opportunities for children to play, explore and discover, and the presence of developmentally appropriate objects, toys and books. This variable is often measured regarding joint activities between parents and child and parental economic investments in activities. It has often been argued that ethnic differences are not caused by subjects’ ethnic origin itself but rather by their social background. For example, Hansen & Jones (2010) reported significant differences between native and immigrant five-year-old children regarding their cognitive skills even when their social background (e.g. birth weight, parental education, income and living in social housing) was controlled for. Glick et al. (2009) confirmed such a finding with children under the age of seven when their socio-economic background was controlled for. However, until recently there has been a lack of large-scale da‐ tasets (in Germany and elsewhere) to verify the influence of social background, at least with respect to children’s cognitive development during pre-school and early primary school years. The results of the ESKOM-V project (and some additional data collected in the city of Osnabrück) indicated that even when the social background is controlled for, there are still significant differences between German and Turkish children, with the latter obtaining lower scores in cognitive tests (Biedinger, 2009; Biedinger & Becker, 2006). However, when the home environment 2 is additionally controlled for, these differences disappear (Biedinger, 2010). A more detailed analysis revealed that Turkish children even perform slightly better than native German children if additional factors like the acculturation of the parents are taken into account (Becker, Klein & Biedinger, 2013). For the city of Osnabrück, where data of school entrance tests were re-examined and controlled for social background and pre-school education, Biedinger & Becker (2006) reported that in contrast to German children, Turkish children’s cognitive, linguistic and social tests were considerably lower. However, other groups (like children from the former Soviet Union) performed as well as German children. More research is needed to examine such effects of ethnic groups in more detail. Early cognitive differences between immigrant and non-immigrant children may also be caused by factors influenced by the migration process, for example, the time of maternal arrival in the host country as well as maternal competence in and her usage of the language of the host country (Glick, Walker & Luz, 2013; Glick et al, 2009). These variables, unfortunately, were not assessed in the present study but could, for example, also account for the differences between Turkish and German children in their performance on the cognitive tests on gestalt closure and numeracy. 28 Nicole Biedinger <?page no="29"?> Language skills Regarding German language skills (with a focus on vocabulary), the findings of the present study are consistent with other studies: Immigrant children displayed poorer language abilities than non-immigrant children (e.g. de Feyter & Winsler, 2009; Magnuson, Lahaie & Waldfogel, 2006; Washbrook, Waldfogel, Bradbury, Corak & Ghanghro, 2012). Dubowy et al. (2008) also found significant differences between native German and immigrant children regarding German grammar, vocabulary, verbal memory, and early school language abilities. Moreover, Becker & Biedinger (2006) reported that in Germany 50 % of children with a Turkish background, 25 % of children from the former Soviet Union and 32 % of children from other nationalities require a German language course before entering school. In contrast, only 1 % of children with a German background are in need of such a German course (cf. Becker, 2006; Biedinger & Becker, 2010). Additionally, children whose parents are both immigrants (first generation) displayed poorer German abilities than, for example, children of mixed couples (Becker, 2011; Biedinger, 2007). Finally, third-generation immigrant children obtained significantly better scores in German vocabulary tests than second-generation immigrant children (Becker, Klein & Biedinger, 2013). Other studies also reported that including the parental social background in the statistical analyses did not affect the results to a great extent (Becker, 2011; Biedinger, 2009). Even when the data were controlled for the children’s home environment, Turkish children still obtained lower scores in the test on German expressive vocabulary than German children (Biedinger, 2009). This effect remained stable even when the family’s migration history (e.g. first, second, third generation) was taken into account (Becker, 2011; see also Niklas & Schneider, 2010). Similar results have been reported in studies from other countries. For example, Washbrook et al. (2012) compared immigrant and non-immigrant children in Australia, Canada, the United Kingdom and the United States and reported that some ethnic groups (especially children from Asia) showed even better results in English language tests than English native children (see also Han, 2006). Early math and reading abilities Many studies examined school-related abilities during the early primary school years. In general, children with an immigration background start school later than German children, and they show lower cognitive and social skills in school entry screenings (Biedinger et al., 2008; Tuppat & Becker, 2014). Regarding early math abilities in Germany, early ethnic differences have been reported for example by Anders, Rossbach, Weinert, Ebert, Kuger, Lehrl & von Maurice 29 Ethnic differences in early cognitive and language skills <?page no="30"?> (2012), Becker & Schmidt (2013) and Biedinger, Becker & Rohling (2008). In addition, for the United States, Magnuson et al. (2006) also found ethnic differences with regard to early mathematic abilities but not regarding reading abilities (also see Han, 2006). In fact, the ethnic origin of the parents seems to play a crucial role (Koury & Votruba-Drzal, 2014): Children from Vietnam, Thailand, Cambodia, or Laos performed better in tests on reading and math than comparable English children (Han, 2006). Such ethnic differences may also depend on the generation: Even children of the third generation seem to have poorer reading abilities than native-language peers (Palacios, Guttmannova & Chase-Lansdale, 2008). Whether such effects also hold true for the present sample will be examined in a future study. Similar to the results for cognitive skills, ethnic differences in early school abilities for reading and math are less pronounced when the children’s social background is controlled for, although they do not completely disappear (Becker & Biedinger, 2006; Biedinger et al., 2008). However, these ethnic differences disappear once the duration of children’s pre-school attendance and their cognitive abilities have been controlled for (Becker & Biedinger, 2006). With the ESKOM-V-data the differences in early abilities persist after controlling for the social background but disappear after additionally controlling for the home environment (Becker & Schmidt, 2013). Tuppat & Becker (2014) showed that after controlling for social background there are still ethnic differences with regard to the age at which children enter school. When Niklas & Schneider (2012) controlled for social background, they did not find any ethnic differences for early math abilities. For the international context, the results are rather inconsistent: On the one hand, many studies from the U.S. reported that ethnic differences in early reading abilities persisted after controlling for social background, which was not the case for all studies on early math abilities (Han, 2006; Lahaie, 2008; Lee & Kao, 2009). On the other hand, studies from Great Britain found immigrant children to perform considerably worse than non-immigrant children in both tests on reading and math (Dustmann & Trentini, 2008; Sammons, Elliot, Sylva, Melhuish, Siraj-Blatchford & Taggart, 2004). 5 Future research The results of the present study and other studies clearly indicate that further research is warranted in Germany and elsewhere. First, even though young children even at age three are able to take ability tests of various kinds, only 30 Nicole Biedinger <?page no="31"?> very few studies have so far indeed examined the development of cognitive and language skills of this age group (i.e. for very young children) quantitatively. Second, national and international results indicate that early cognitive abil‐ ities may be affected by various factors, for example by early stimulation in pre-school or by a stimulating home environment. This also holds true for child‐ ren’s language abilities which depend on the stimulation at home and which is ideally provided in the language in which parents are most proficient (which is not necessarily the language of the host country). However, studies which examine the influence of pre-school and home environment in more detail are still needed because it is not clear whether, for example, a stimulating home environment is more important than a stimulating pre-school environment and how such factors may interact in immigrant children’s language development. Third, international studies (e.g. Han, 2006; Koury & Votruba-Drzal, 2014) reported that different ethnic groups (e.g. groups with an Asian or Russian background vs. groups with a Turkish background) perform differently in language tests. Therefore, studies for the educational contexts in Germany should also take into consideration different ethnic groups in more detail. ESKOM-V only focused on Turkish children, but further studies on immigrant pre-schoolers’ language skills should differentiate groups of different ethnic origin, language background and recency of immigration. Fourth, especially longitudinal studies are needed to examine children’s language biography throughout the educational system. It is, for example, possible that some of the ethnic differences reported here will level out with age as the Turkish children’s German language skills improve over time. Other group differences may change in rather unexpected ways (e.g. due to puberty, peer-group pressure or related issues) which is why it would be worthwhile to examine immigrant students’ skills until they leave school and beyond. In sum, this longitudinal study analysed Turkish and German children’s cognitive, linguistic and academic development from three to eight years of age and found ethnic differences in language skills to persist whereas cognitive skills decreased with age. Although many of these findings could be attributed to the children’s social background, early pre-school education as well as a stimulating home environment, many questions need to be addressed in future studies in order to strengthen children’s educational participation in society, independent of their (migration) background. 31 Ethnic differences in early cognitive and language skills <?page no="32"?> References Anders, Y., Rossbach, H.G., Weinert, S., Ebert, S., Kuger, S., Lehrl, S. & von Maurice, J. 2012. Home and preschool learning environments and their relations to the development of early numeracy skills. Early Childhood Research Quarterly, 27 (2), 231-244. Becker, B. 2006. Der Einfluss des Kindergartens als Kontext zum Erwerb der deutschen Sprache bei Migrantenkindern. Zeitschrift für Soziologie, 35 (6), 449-464. Becker, B. 2010a. Wer profitiert mehr vom Kindergarten? Die Wirkung der Kinder‐ gartenbesuchsdauer und Ausstattungsqualität auf die Entwicklung des deutschen Wortschatzes bei deutschen und türkischen Kindern. Kölner Zeitschrift für Soziologie und Sozialpsychologie, 62 (1), 139-163. Becker, B. 2010b. Who profits most from early parental investments? The effects of activities inside and outside the family on German and Turkish children’s language development. Child Indicators Research, 3 (1), 29-46. Becker, B. 2011. Cognitive and language skills of Turkish children in Germany: A comparison of the second and third generation and mixed generational groups. International Migration Review, 2, 426-459. Becker, B. & Biedinger, N. 2006. Ethnische Bildungsungleichheit zu Beginn der Schulzeit. Kölner Zeitschrift für Soziologie und Sozialpsychologie, 58 (4), 660-684. Becker, B. & Schmidt, F. 2013. Ungleiche Startvoraussetzungen zu Beginn der Schul‐ laufbahn? Unterschiede in den mathematischen und sprachlichen Fähigkeiten von sechsjährigen Kindern nach Geschlecht und Migrationshintergrund. In: Hadjar, A. & Hupka-Brunne, S. (Eds.), Geschlecht, Migrationshintergrund und Bildungserfolg. Weinheim: Beltz Juventa. 52-76. Becker, B., Klein, O. & Biedinger, N. 2013. The development of cognitive, language, and cultural skills from age 3 to 6: A comparison between children of Turkish origin and children of native-born German parents and the role of immigrant parents. Acculturation to the receiving society. American Educational Research Journal, 50 (3), 616-649. Becker, B., Biedinger, N., Klein, O. & Koch, F. 2017. Erwerb von sprachlichen und kultur‐ ellen Kompetenzen von Migrantenkindern: Methodenbericht. Mannheim: Mannheimer Zentrum für Europäische Sozialforschung (MZES). Biedinger, N. 2007. Entwicklung und Lebensumfeld von Vorschulkindern: Zur Heterogenität von Familien mit türkischem Migrationshintergrund. Zeitschrift für Türkeistudien, 20 (1), 7-24. Biedinger, N. 2009. Der Einfluss von elterlichen Investitionen auf die Entwicklung von deutschen und türkischen Kindern. Berliner Journal für Soziologie, 19 (2), 268-294. 32 Nicole Biedinger <?page no="33"?> Biedinger, N. 2010. Early ethnic inequality: The influence of social background and parental involvement on preschool children’s cognitive ability in Germany. Child Indicators Research, 3 (1), 11-28. Biedinger, N. & Becker, B. 2006. Der Einfluss des Vorschulbesuchs auf die Entwicklung und den langfristigen Bildungserfolg von Kindern. Ein Überblick über internationale Studien im Vorschulbereich. Mannheim: Mannheimer Zentrum für Europäische Sozial‐ forschung (MZES). Biedinger, N. & Becker, B. 2010. Frühe ethnische Bildungsungleichheit. Der Einfluss des Kindergartenbesuchs auf die Schulfähigkeit. In: Becker, B. & Reimer, D. (Eds.), Vom Kindergarten bis zur Hochschule. Die Generierung von ethnischen und sozialen Disparitäten in der Bildungsbiographie. Wiesbaden: VS Verlag. 49-80. Biedinger, N., Becker, B. & Rohling, I. 2008. Early ethnic educational inequality: The influence of duration of preschool attendance and social composition. European Sociological Review, 24 (2), 243-256. Crosnoe, R., Leventhal, T., Wirth, R.J., Pierce, K.M., Pianta, R.C. & NICHD Early Child Care Research Network. 2010. Family socioeconomic status and consistent environmental stimulation in early childhood. Child Development, 81 (3), 972-987. De Feyter, J.J. & Winsler, A. 2009. The early developmental competencies and school readiness of low-income, immigrant children: Influences of generation, race/ ethnicity, and national origins. Early Childhood Research Quarterly, 24 (4), 411-431. Dubowy, M., Ebert, S., von Maurice, J. & Weinert, S. 2008. Sprachlich-kognitive Kompe‐ tenzen beim Eintritt in den Kindergarten. Ein Vergleich von Kindern mit und ohne Migrationshintergrund. Zeitschrift für Entwicklungspsychologie und Pädagogische Psy‐ chologie, 40 (3), 124-134. Dustmann, C. & Trentini, C. 2008. The Ethnic Test Score Gaps and Pre-School Experience. Mimeo: University College London. Forget-Dubois, N., Dionne, G., Lemelin, J-P., Pérusse, D., Tremblay, R.E. & Boivin, M. 2009. Early child language mediates the relation between home environment and school readiness. Child Development, 80 (3), 736-749. Glick, J.E., Baes, L. & Yabiku, S. T. 2009. Mother’s age at arrival in the United States and early cognitive development. Early Childhood Research Quarterly, 24 (4), 367-380. Glick, J.E., Walker, L. & Luz, L. 2013. Linguistic isolation in the home and community: Protection or risk for young children? Social Science Research, 42 (1), 140-154. Han, W-J. 2006. Academic achievements of children in immigrant families. Educational Research and Review, 1 (8), 286-318. Hansen, K. & Jones, E.M. 2010. Age 5 cognitive development in England. Child Indicators Research, 3 (1), 105-126. Heath, A. & Brinbaum, Y. 2007. Explaining ethnic inequalities in educational attainment. Ethnicities, 7 (3), 291-305. 33 Ethnic differences in early cognitive and language skills <?page no="34"?> Heckman, J.J. 2006. Skill formation and the economics of investing in disadvantaged children. Science, 312 (5782), 1900-1902. Hinte, H., Rinne, U. & Zimmermann, K.F. 2012. Zuwanderung, Demografie und Arbeits‐ markt - Fakten statt Vorbehalte. In: Heinz, A. & Kluge, U. (Eds.), Einwanderung - Bedrohung oder Zukunft. Mythen und Fakten zur Integration. Frankfurt/ M.: Campus, 263-278. Koury, A.S. & Votruba-Drzal, E. 2014. School readiness of children of immigrants: Contributions of early environments. Journal of Educational Psychology, 106 (1), 268-288. Kristen, C., Edele, A., Kalter, F., Kogan, I., Schulz, B., Stanat, P. & Will, G. 2011. The education of migrants and their children across the life course. Zeitschrift für Erziehungswissenschaft, 14 (2), 121-137. Lahaie, C. 2008. School readiness of children of immigrants: Does parental involvement play a role? Social Science Quarterly, 89 (3), 684-705. Lee, E.M. & Kao, G. 2009. Less bang for the buck? Cultural capital and immigrant status effects on kindergarten academic outcomes. Poetics, 37 (3), 201-226. Magnuson, K., Lahaie, C. & Waldfogel, J. 2006. Preschool and school readiness of children of immigrants. Social Science Quarterly, 87 (5), 1241-1262. Melchers, P. & Preuß, U. 2001. Kaufman Assessment Battery for Children: K-ABC. Deutschsprachige Fassung. Leiden: PITS. Melhuish, E.C., Phan, M.B., Sylva, K., Sammons, P., Siraj-Blatchford, I. & Taggart, B. 2008. Effects of the home learning environment and preschool center experience upon literacy and numeracy development in early primary school. Journal of Social Issues, 64 (1), 95-114. Niklas, F. & Schneider, W. 2010. Der Zusammenhang von familiärer Lernumwelt mit schulrelevanten Kompetenzen im Vorschulalter. Zeitschrift für Soziologie der Erzie‐ hung und Sozialisation, 30 (2), 149-165. Niklas, F. & Schneider, W. 2012. Einfluss von “Home Numeracy Environment” auf die mathematische Kompetenzentwicklung vom Vorschulalter bis Ende des 1. Schul‐ jahres. Zeitschrift für Familienforschung, 24 (2), 134-147. Niklas, F., Schmiedeler, S., Pröstler, N. & Schneider, W. 2011. Die Bedeutung des Mi‐ grationshintergrunds, des Kindergartenbesuchs sowie der Zusammensetzung der Kindergartengruppe für sprachliche Leistungen von Vorschulkindern. Zeitschrift für Pädagogische Psychologie, 25 (2), 115-130. Palacios, N., Guttmannova, K. & Chase-Lansdale, P.L. 2008. Early reading achievement of children in immigrant families: Is there an immigrant paradox? Developmental Psychology, 44 (5), 1381-1395. 34 Nicole Biedinger <?page no="35"?> Raviv, T., Kessenich, M. & Morrison, F.J. 2004. A mediational model of the association between socioeconomic status and three-year-old language abilities: The role of parenting factors. Early Childhood Research Quarterly, 19 (4), 528-547. Roick, T., Gölitz, D. & Hasselhorn, M. (2004). Deutscher Mathematiktest für dritte Klassen (DEMAT 3+). Göttingen: Beltz Test. Sammons, P., Elliot, K., Sylva, K., Melhuish, E., Siraj-Blatchford, I. & Taggart, B. 2004. The impact of pre-school on young children’s cognitive attainments at entry to reception. British Educational Research Journal, 30 (5), 691-712. Schneider, W., Blanke, I., Faust, V. & Küspert, P. (2011). WLLP-R Würzburger Leise-Lesep‐ robe - Revision. Ein Gruppentest für die Grundschule. Göttingen: Hogrefe. Stanat, P. & Christensen, G. 2006. Schulerfolg von Jugendlichen mit Migrationshintergrund im internationalen Vergleich. Eine Analyse von Voraussetzungen und Erträgen schuli‐ schen Lernens im Rahmen von PISA 2003. Berlin: BMBF. Statistisches Bundesamt Deutschland. 2017. Bevölkerung und Erwerbstätigkeit 2016. Aus‐ ländische Bevölkerung. Ergebnisse des Ausländerzentralregisters. Wiesbaden: Destatis. Tuppat, J. & Becker, B. 2014. Sind türkischstämmige Kinder beim Schulstart im Nachteil? Kölner Zeitschrift für Soziologie und Sozialpsychologie, 66 (2), 219-241. Washbrook, E., Waldfogel, J., Bradbury, B., Corak, M. & Ghanghro, A.A. 2012. The development of young children of immigrants in Australia, Canada, the United Kingdom, and the United States. Child Development, 83 (5), 1591-1607. 35 Ethnic differences in early cognitive and language skills <?page no="37"?> Early immersion education: L2 vocabulary acquisition and the role of non-verbal intelligence Katja Schmidt / Yvonne Blumenthal Abstract This paper focuses on the lexical abilities of German learners of English in bilingual German-English pre-schools. It examines a possible connection between the children’s cognitive abilities (as operationalised by non-verbal intelligence) and the development of their second language English vocabulary knowledge. Data from a longitudinal study are presented, including 29 German threeto five-year old pre-schoolers who took part in an English immersion programme. The children were tested on their receptive English lexical knowledge twice at a six-month interval. The results were compared to their scores in a non-verbal intelligence (NVI) test. Surprisingly, at the first test time (T1) the lexical abilities of children with a below-average non-verbal intelligence score did not significantly differ from those of children with an average or above-average NVI. However, a repeated measures ANOVA showed that the improvement rates from T1 to T2 differed among the groups, showing large effect sizes of time, non-verbal intelligence group and the interaction time x NVI group. At the second test time (T2), children with an above-average non-verbal intelligence score performed significantly better than their peers with an average or below-average non-verbal intelligence score. 1 Introduction Learning a second language from a very early age is widely seen as advan‐ tageous to language proficiency over time (e.g. Birdsong, 1999; Ellis, 2008; Nikolov, 2009; Paradis, 2004; Singleton & Ryan, 2004). Young children seem to pick up the L2 with ease, especially in naturalistic communicative situations (Ellis, 2008). A successful option in utilising this potential is early immersion education, an intensive form of bilingual education which fosters L2 acquis‐ <?page no="38"?> ition by using authentic settings of communication. The term “immersion” is derived from Latin immersio, which, in a linguistic context, means that children are - metaphorically speaking - plunged into the L2. Stern defines immersion as the exposure of children to the second language in real-life situations […]. The pupil is plunged into a ‘language bath’ in the same way as he found himself immersed in the linguistic environment of his mother tongue (Stern, 1963: 66). In immersion education the target language is not “taught” as in conventional foreign language classes. Instead, it is used as a means of communication in everyday situations and learning activities. Metalinguistic aspects, such as grammar, are usually not dealt with explicitly. Thus, the language is not the object of instruction but the medium ( Jalkanen, 2009). The term immersion originates from the Canadian school context when the L2 is the medium of instruction in at least 50 % of the teaching time (e.g. Wode, 1995, Zydatiß, 2000). Over time, the term has also been adapted to pre-school contexts, where it is now used in the sense that at least 50 % of the input per day is provided in the L2 (Kersten, Drewing, Granados, Leloux, Lommel, Schneider & Taylor, 2010: 82). To guarantee the amount of L2 input, two teachers are usually responsible for one group of children, one of them using the children’s L1 and the other speaking the target language (L2) in contact with the children. This one-person-one-language principle “ensures that the children have regular exposure to and have to make use of each language” (Saunders, 1988: 49). Research on early immersion education has shown positive effects on L2 ac‐ quisition (see e. g. Piske, 2013, for an overview). Children in bilingual pre-schools develop their abilities in the target language rapidly, and this is especially true for receptive skills (e.g. Häckel, 2013; Rohde, 2010; Steinlen, 2013, 2016a, b). With regard to the L2 vocabulary knowledge of bilingual pre-school children, research has shown that a number of internal and external factors influence the rate of acquisition. For example, the duration of L2 contact, as well as the learner’s age, have a significant impact on word knowledge (cf. Couve de Murville, Kersten, Maier, Ponto & Weitz, 2016). Other factors, such as migration background, have only rarely been addressed in the context of early immersion education (but see Häckel & Piske, 2011, 2016; or Steinlen, 2013, 2016a). This article deals with the question of whether non-verbal intelligence impacts receptive L2 lexical learning. In the first part, a short overview of the theoretical background is provided in order to place the study within the framework of vocabulary research. In the second part, the first results of a 38 Katja Schmidt / Yvonne Blumenthal <?page no="39"?> longitudinal study conducted in a German-English bilingual pre-school are presented and discussed. 2 Background 2.1 L2 vocabulary acquisition “Without grammar, little can be conveyed, without vocabulary, nothing can be conveyed” (Wilkins, 1972: 111). This quotation illustrates the fundamental importance of the lexicon within L2 acquisition. Vocabulary knowledge is highly significant for the development of all language skills (Targonska & Stork, 2013: 71). It is not only essential for language production (Levelt, 1989: 181) but also important in oral comprehension: If words cannot be isolated from the speech stream, and if lexical information cannot be used to interpret the utterances, the input will not be comprehended. (Gass, Behney & Plonsky, 2013: 196) The distinction between production and comprehension implies that there are different levels of word knowledge. Receptive vocabulary knowledge, usually associated with reading and listening, refers to the ability of understanding words (Laufer & Goldstein, 2004). Productive vocabulary knowledge, on the other hand, is more commonly linked with speaking and writing and refers to the ability to produce words within an appropriate context (ibid.). L2 research has shown that productive lexical knowledge lags behind receptive knowledge (Rohde, 2010: 51). This is also true for the bilingual pre-school context, where children’s receptive skills in the L2 are far superior to their L2 production skills (e.g. Rohde & Tiefenthal, 2000). Still, the question remains as to whether vocabulary knowledge can be explained by such a simple dichotomy. Teichroew (1982) believes that lexical knowledge can be best described as a continuum, the initial stage being recognition, and the final being production. According to her, there is a gradual increase from receptive to productive word knowledge, based on the learner’s increasing knowledge of the lexical item (ibid.). Bleyhl (1995) adopts this idea and describes the construction of word knowledge as a chronological process, consisting of four consecutive steps: (1) symbol/ form recognition (2) attribution of meaning and assignment to a semantic field (leads to receptive knowledge of the lexical item) (3) cross-linking in the mental lexicon (4) productive knowledge of the lexical item (Bleyhl, 1995: 26). 39 Early immersion education <?page no="40"?> Although Bleyhl’s idea of a chronological acquisition process sounds promising, a lexical item is probably not fully acquired even after all four steps have been mastered. As Cameron (2001: 74: 84) notes, word knowledge is a complex phenomenon: it extends and deepens whenever the learner is confronted with a new word. Thus, the construction of word knowledge is a lifelong process (Read, 2011: 1). But what does it actually mean to “know a word”? There have been several attempts to define word knowledge (cf. e. g. Cameron, 2001; Chapelle, 1998; Thornbury, 2002). For Nation (2001) word knowledge includes (1) the knowledge of form (2) the knowledge of meaning (3) the knowledge of use. (1) refers to knowledge of pronunciation and spelling; (2) includes knowledge of concepts, referents, associations as well as sense relations and lexical networks; (3) refers to knowledge of grammatical functions, collocations and register (Nation, 2001: 27). There is a quantitative as well as a qualitative dimension to word knowledge. Breadth of knowledge refers to the size of the learner’s L2 lexicon, i. e. to the number of words that a learner - perhaps only partially - knows (ibid.). Depth, on the other hand, relates to the qualitative dimension of lexical knowledge and is defined by the degree of knowledge of form, meaning and use (Daniel, 2001: 44). In order to reach a greater depth of knowledge, the learner has to be confronted with the lexical item repeatedly and in different contexts. In this way, the learner has the opportunity to determine relevant semantic and syntactic information (Gass et al., 2013: 212). In sum, the acquisition of the lexicon is a highly complex aspect of L2 acquisition. It is a recursive process that leads to different degrees of word knowledge. Learners in early immersion programmes usually develop receptive lexical knowledge quickly, whereas their productive lexical skills develop at a slower rate. 2.2 Context factors in L2 vocabulary acquisition Although all children have the ability to learn a second language (Chilla, Rothweiler & Babur, 2010), there are considerable individual differences in children’s L2 proficiency. These differences have been related to various internal and external factors (see Kersten, 2019, for a review). In this context, Couve et al. (2016: 91 f.) for example, point to studies that focus on learner-inherent variables, such as gender (Klieme, 2006), motivation (Ushioda & Dörnyei, 2012), 40 Katja Schmidt / Yvonne Blumenthal <?page no="41"?> language aptitude (Shekan, 2012), and age (Spadaro, 2013). Furthermore, Couve de Murville et al. (2016) list a number of external factors which influence learners’ L2 acquisition, for example, the opportunity for language use (Duff & Polio, 1990) or the typological distance between the L1 and L2 (Ringbom & Jarvis, 2009). When examining the effects of early immersion education in Germany, re‐ searchers have identified a number of variables as being beneficial for language learning: The age of the learner, a long exposure to the L2, a high intensity of the language programme, the active use of the L2 and also the specific use of pedagogic strategies used in bilingual programmes have been found to advance the children’s language attainment (Kersten et al., 2010: 79). With regard to L2 vocabulary acquisition in early immersion contexts, studies primarily emphasise the importance of duration as well as intensity of the L2 contact (Couve de Murville et al., 2016: 92; see also Steinlen, 2016b). The age of the learner has also been identified as a significant variable in L2 vocabulary acquisition (Schelletter & Ramsey, 2010). In addition, Weitz (2015) noted that the quality of L2 input has a strong effect on children’s lexical knowledge, i. e. children demonstrate the best results when educators provide high-quality language input. The children’s gender, on the other hand, did not appear to be a relevant context factor for word learning in bilingual pre-schools (Rohde, 2010; Steinlen, Kersten & Piske, 2019). 2.3 Non-verbal intelligence as a context factor in L2 vocabulary acquisition One factor that has only rarely been addressed is the role of intelligence in L2 lexical learning. The term “intelligence” can be defined as a “general set of cognitive abilities involved in performing a wide range of tasks” (Ellis, 2008: 649). According to Cattell (1971), these cognitive abilities can be separated into broadly two different types: “fluid” and “crystallised” abilities (ibid). “Fluid intelligence” involves the capability of solving novel problems as well as of thinking flexibly and quickly. It is believed to have a physiological basis and thus to be independent of education and culturally acquired skills (Horn, 1967). “Crystallised intelligence”, on the other hand, is associated with acculturated knowledge and learnt skills (Horn & Cattell, 1967). It is the result of life-long learning and hence highly dependent on educational and cultural opportunities. In addition to this differentiation, modern approaches point out a variety of cognitive factors which intelligence is composed of. These include, for example, 41 Early immersion education <?page no="42"?> visual-spatial processing, quantitative reasoning, or working memory (for an overview see Coon & Mitterer, 2010). Gardner (1993) proposes a set of “multiple intelligences”, consisting of linguistic, logical-mathematical, spatial, musical, bodily/ kinaesthetic, interpersonal, intrapersonal, and naturalistic intelligence. Musical intelligence, for example, could explain why some learners easily produce intonation patterns of a language (Salehi & Sadighi, 2012). For L2 acquisition, the relationship between L2 development and intelligence has been studied by Genesee (1976, 1987). He investigated the language out‐ comes of school children in French immersion programmes in Canada in relation to their intellectual ability. His studies revealed that children with an above-average level of intelligence tend to score higher in tests of reading and writing in the L2 than children with a below-average level of intelligence. At the same time, children with a low level of intelligence showed better results in the L2 than comparable students receiving conventional foreign language instruction. In other words, students with a below-average level of intelligence benefit from immersion education in the form of enhanced L2 proficiency (Genesee, 1987). The relationship between intelligence and L2 vocabulary knowledge has been studied by Salehi & Sadighi (2012), who investigated the English lexical knowledge of Iranian high school students in relation to their intelligence. The results showed a positive relationship between intelligence and L2 word knowledge. At the same time, students with equal intelligence achieved better results when they were given extra instruction beyond the regular school lessons. This leads to the conclusion that intelligence is a factor in L2 vocabulary acquisition, but that other factors, such as the amount of language practice, also play an important role. Paradis, Genesee & Crago (2011: 46) noted that the link between intelligence and dual language learning has primarily been investigated in the case of L2 learners of school age. To date, this relationship has not been examined for younger children attending a bilingual pre-school. One reason for this circumstance may be that it is relatively difficult to test general intelligence in pre-schoolers. Factors such as distraction, anxiety, or stress can drastically influence a child’s performance. In addition, intelligence scores are not stable during childhood; they fluctuate during the course of childhood development. Thus, there are very few reliable intelligence tests for young children. This is why we decided on a standardised non-verbal intelligence test, namely the Snijders-Oomen Non-Verbal Intelligence Test (SON-R 2 ½ - 7; Tellegen, Laros & Petermann, 2007), which measures ‘fluid intelligence’ (Cattell, 1971), i. e. problem-solving abilities. For the bilingual pre-school context, the 42 Katja Schmidt / Yvonne Blumenthal <?page no="43"?> relationship between L2 learning and non-verbal intelligence has not yet been investigated, although for other learning contexts a link between the two has been shown (see i. e. Brooks & Kempe, 2013, for adults’ L2 English grammar; and Kievit, Lindenberger, Goodyer, Jones, Fonagy, Bullmore, The Neuroscience in Psychiatry Network & Dolan, 2017, for children’s L1 vocabulary skills; see also Steinlen & Piske, this volume, for effects of non-verbal intelligence on children’s L1 and L2 English reading skills). 3 The study 3.1 Research questions The study presented here examines the relationship between L2 vocabulary acquisition and non-verbal intelligence in the context of early immersion education. The following questions will be addressed: (1) What is the level of receptive L2 vocabulary knowledge of children in German-English bilingual pre-schools in relation to their level of non-verbal intelligence after 1.5 years of contact to L2 English? (2) What is the impact of non-verbal intelligence on the amount of progress made in receptive L2 vocabulary for children in German-English bilin‐ gual pre-schools 6 months later, i. e. after two years of contact to L2 English? 3.2 Pre-schools The study was conducted in two bilingual pre-schools in northern Germany, one in Rostock and one in Schwerin. Both institutions offer a partial immer‐ sion programme, i. e. one caretaker only speaks German and the other only English, adhering to the principle “one-person-one-language” (Döpke, 1992). In addition to English immersion, the pre-school in Rostock also provides German-French bilingual groups, which are not the focus of the present study. The German-speaking caretakers at both pre-schools are all native speakers. The English-speaking caretakers are usually Germans with a very good command of English. At both pre-schools, one of the English-speaking caretakers is a native speaker. The pre-schools differ in terms of the number of children as well as their socio-cultural background. In Rostock, approximately 300 children from 20 nations attend pre-school. Most children come from middle-class families who place great value on early education and intercultural learning. The pre-school in Schwerin is located in an area with a high percentage of socially disadvan‐ 43 Early immersion education <?page no="44"?> taged groups. Therefore, most of the 180 children come from families with deprived backgrounds. The percentage of children with a migration background is lower in the pre-school in Schwerin than in the one in Rostock. The amount of exposure to English is difficult to specify because the peda‐ gogical concepts in these two pre-schools differ. In Rostock, two caretakers are responsible for one group - one of them speaks German and the other English. Thus, the children are surrounded by both languages for most of the day. The pre-school in Schwerin adheres to the “open group concept”, i. e. the children are free to choose activities offered by different caretakers in different rooms. This may reduce or increase the amount of exposure to English, depending on whether the activity is led by a Germanor an English-speaking caretaker (see also Steinlen & Rogotzki, 2009; Wode, 2001). Another characteristic of the pre-school in Schwerin is the age at which the immersion programme starts. While in Rostock children start attending the bilingual groups at approximately three years of age, the pre-school in Schwerin offers immersion education for children as early as one year of age. However, not all parents choose this alternative and most children also start with bilingual education at the age of three. 3.3 Participants Altogether, 29 children (52 % female, 48 % male) in the two German-English pre-schools took part in this study (18 from Schwerin and eleven from Rostock). In all cases, the children’s first language was German. The children were tested twice on the British Picture Vocabulary Scale III (BPVS III, Dunn & Dunn, 2009) at an interval of approximately six months. At the first testing time (T1), the children’s age range was between 3; 2 and 4; 8 years (mean: 3; 6 years, standard deviation, SD = 4.9 months), at the second time (T2), their age range was between 3; 9 and 5; 3 (mean: 4; 9 SD = 6.0 months). The children’s contact time with English was between six and 33 months (mean: 17.5 months, SD = 7.9 months) at T1 and between twelve and 37 months (mean: 23.9 months, SD = 7.6 months) at T2. Of the 29 children, 23 had been exposed to English from the age of three years, five from the age of 1 year, and one from the age of two years. Thus, there is no relationship between the children’s age and their exposure to English. The children were all tested individually in a separate room with which they were familiar, and which prevented outside interruptions. Before the test commenced, it was ensured that the children felt comfortable with the situation (see Crain & Thornton, 1998 on the importance of a child-friendly test environment). At T1, the children were tested first with the non-verbal intelligence test SON-R 2 ½ - 7 (Tellegen et al., 2007). The BPVS III was 44 Katja Schmidt / Yvonne Blumenthal <?page no="45"?> conducted a few days later in order not to place the children under any undue pressure. 3.4 Test materials The Snijders-Oomen Non-Verbal Intelligence Test (SON-R 2½ - 7, Tellegen et al., 2007) is a standardised testing instrument for children between two-and-a-half and seven years of age. It provides a broad assessment of non-verbal intelligence without being dependent on language skills. This is true for both the test materials and test administration: all tasks have a non-verbal form, and the entire test can (but does not necessarily have to) be administered without any verbal instruction. In contrast to other intelligence tests, the SON-R 2½ - 7 measures non-verbal intelligence on the basis of a number of diverse tasks. This is why verbal test items - which often depend on knowledge and experience - are not included. Tellegen, Winkel, Wijnberg-Williams & Laros (2009: 13) point out that the SON-R 2½ - 7 is “focused more on the measurement of ‘fluid intelligence’ and less on the measurement of ‘crystallised intelligence’ (Cattell, 1971) than are the other tests”. The SON-R 2½ - 7 comprises six different subtests: mosaics, categories, puzzles, analogies, situations and patterns. In the subtest ‘mosaics’, the children have to copy different mosaic patterns in a frame. The subtest ‘categories’ consists of sorting cards into two groups according to the category they belong to. The ‘puzzle’ subtest requires putting puzzle pieces into a frame so that they resemble an example. In the subtest ‘analogies’, children have to sort disks into two compartments on the basis of form, colour and size. The subtest ‘categories’ comprises completing pictures by filling in the correct piece from a number of alternatives. Finally, the subtest ‘patterns’ includes drawing simple patterns to resemble a specific example. The items in the different subtests are arranged in order of increasing difficulty. After the children have completed all six subtests, partial scores are combined to form an intelligence score that represents the children’s ability relative to their age group. Although intelligence tests are generally less reliable for pre-schoolers than for school-age children, the SON-R 2½ - 7 can be considered a valuable test instrument, with a reliability rate of 0.86 and high correlation rates with other tests on general intelligence, e. g. the Wechsler Intelligence Scale for Children-Revised or the British Ability Scales (Tellegen et al., 2009: 122). In the following, the raw scores of the SON-R 2½ - 7 were converted to z-scores. The British Picture Vocabulary Scale III (BPVS III, Dunn & Dunn, 2009) is a norm-referenced test of receptive vocabulary for British English, designed for children between three years and 16 years eleven months whose L1 is English. 45 Early immersion education <?page no="46"?> However, it is also suitable for children learning English in Great Britain as their L2. When administering the test with German children learning English in a German-English bilingual pre-school in Germany, the environment in which English is acquired differs significantly from the acquisition context in Great Britain, where English is the ambient language (cf. Rohde & Tiefenthal, 2002). Thus, it can be expected that German pre-schoolers will score lower than their peers in Great Britain. Nevertheless, the BPVS III can be considered a valuable test instrument, revealing “information about a universal development of children’s receptive lexicon that is comparable to children’s data in other kindergarten programmes” (Weitz, n.d.: 1). The BPVS III consists of 14 sets of twelve items each, yielding a total of 168 stimulus words. The items belong to different semantic groups, e. g. animals, body parts, food, geographic scenery, emotions, plants, shapes, toys, vehicles (Dunn & Dunn, 2009: 23). Furthermore, the target words cover different word classes, i. e. nouns, verbs and adjectives. The BPVS III is designed as a multiple-choice picture-pointing test. The children have to select one picture from four illustrations, i. e. they have to point to the picture that best illustrates the meaning of a stimulus word spoken by the administrator. The test items of each set are arranged in order of increasing difficulty. When a child makes eight or more errors in a set of twelve items, testing is discontinued. All correct answers form the raw score, which can later be converted to a standardised score. On average, children learning English in a German-English bilingual pre-school do not progress past the third set (see also Weitz, n.d.: 1). The focus of this study is on the raw scores of the BPVS III at T1 and T2. 4 Results 4.1 Non-verbal intelligence At T1, the children were tested on the SON-R 2½ - 7 in order to measure their general level of non-verbal intelligence. Following Resing & Blok’s (2002) classification of intelligence scores, three groups were established, which constituted of group I (NVI<90, i. e. below average), group II (NVI 90-110, i. e. average) and group III (NVI>110, i. e. above average). According to their test results, the children were matched to one of these three groups. Twelve children scored below 90, i. e. below average. Thirteen children scored between 90 and 110, which can be considered an average level of intelligence; while four scored above 110, i. e. above average (see Table 1 and Figure 1). 46 Katja Schmidt / Yvonne Blumenthal <?page no="47"?> NVI group I NVI<90, n = 12 NVI group II NVI 90-110, n = 13 NVI group III NVI>110, n = 4 NVI scores (SON-R 2 ½ - 7) M = 77.8 (SD = 5.3) M = 99.4 (SD = 6.5) M = 118.5 (SD = 6.5) Table 1: Descriptive statistics: Scores of the tests on non-verbal intelligence (SON-R 2 ½ - 7) for the three NVI groups. The results for the three groups are illustrated in Figure 1: IQ groups IQ > 110, n= 4 IQ 90 -110, n = 13 IQ < 90, n = 12 IQ (SON-R 2,5-7) 120 100 80 60 1 Figure 1: NVI scores in three groups (n = 29). As Table 1 and Figure 1 also show, the NVI level in these three groups differs considerably (M NVI<90 = 77.8/ SD NVI<90 = 5.3, M NVI 90-110 = 99.4/ SD NVI 90-110 = 6.5, M NVI<110 = 118.5/ SD NVI<110 = 6.5). Bonferroni-adjusted post-hoc analysis revealed significant differences (p<.001) in the NVI scores of all groups as shown in Table 2. 47 Early immersion education <?page no="48"?> NVI group NVI group Mean Differ‐ ence Std. Error Sig. 95 %-Confid. Interval Lower Bound Upper Bound NVI<90 NVI 90-110 -21.63462 * 2.42024 .000 -27.8279 -15.4414 NVI<90 NVI > 110 -40.75000 * 3.49051 .000 -49.6820 -31.8180 NVI 90-110 NVI > 110 -19.11538 * 3.45679 .000 -27.9611 -10.2697 Dependent variable: NVI Score (SON-R 2½ - 7) Table 2: Results of Bonferroni-adjusted post-hoc analysis for the NVI differences in three groups. 4.2 L2 receptive vocabulary and non-verbal intelligence L2 receptive lexical knowledge was measured with the BPVS III at two test times. Table 3 provides information on the mean scores of the test on English receptive vocabulary at T1 and T2 as well as some background information on the subjects (i.e. their age and contact duration to L2 English). NVI group I NVI<90, n = 12 NVI group II NVI 90-110, n = 13 NVI group III NVI>110, n = 4 Age at T1 M = 3; 6 (SD = 0; 5) M = 3; 5 (SD = 0; 5) M = 3; 6 (SD = 0; 6) L2 contact time at T1 in months M = 21.2 (SD = 8.9) M = 14.7 (SD = 6.7) M = 15.8 (SD = 4.7) NVI scores (SON-R 2½ - 7) M = 77.8 (SD = 5.3) M = 99.4 (SD = 6.5) M = 118.5 (SD = 6.5) BVPS III raw scores at T1 M = 12.1 (SD = 5.2) M = 11.5 (SD = 5.6) M = 20.3 (SD = 10.9) BVPS III raw scores at T2 M = 16.0 (SD = 8.1) M = 14.8 (SD = 6.8) M = 38.5 (SD = 7.3) Table 3: Descriptive statistics: Age, contact to L2 English and scores of the tests on English receptive vocabulary (BPVS III). N = number of subjects, M = mean raw scores, SD = standard deviation. 48 Katja Schmidt / Yvonne Blumenthal <?page no="49"?> 1 We used the raw scores rather than the standardised scores for analysis, because in most cases a transfer from raw scores to standardised scores would not have been accurate enough. The norm tables include standardised scores between 70 and 140. Most children achieved raw scores that were equivalent to standardised scores below 70, which means that transferring these to standardised scores would not have been possible to an appropriate degree of accuracy. Regarding the relationship between the BPVS and the three groups of children with different NVI scores at T1, the results of children with a low NVI (group I) and children with an average NVI (group II) were comparable to one another, the first even scoring slightly higher (M = 12.1/ SD = 5.2) than the latter (M = 11.5/ SD = 5.6) 1 . Although children with an above-average NVI (group III) scored ap‐ proximately eight raw points higher than the other groups (M = 20.3/ SD = 10.9), Bonferroni-adjusted post-hoc analysis revealed no significant difference (p>.05) between these three groups. With regard to T2, the children in all three NVI groups received higher BPVS III scores than at T1. The development from T1 to T2 is illustrated in Figure 2. The results of children with a low NVI (group I: M = 16.1/ SD = 8.1) and children with an average NVI (group II: M = 14.8/ SD = 6.8) were again comparable, the first even scoring one raw score higher than the latter. Although both groups demonstrated a growth rate of about four raw points from T1, this improvement is not significant. Children with an above-average NVI (group III; M = 38.5/ SD = 7.3), however, scored significantly higher than the two other groups (p=.000, ANOVA, post hoc Bonferroni). In a second analysis, the development of receptive L2 vocabulary knowledge was calculated by means of an analysis of variance (ANOVA) with repeated measures and the covariates age, gender and L2 contact time. Significant main effects were found for time (F[1,49] = 18.37, p<.001, partial η2 = .273), for NVI group (F[2,49] = 21.35, p<.001, partial η2 = .466), and the interaction time x NVI group (F[2,49] = 4.37, p<.05, partial η2 = .151). This represents large effects (Cohen, 1988) of time, NVI-group, and the interaction time x NVI group on the receptive L2 lexical knowledge. The effect of L2 contact time on receptive vocabulary knowledge at T2 was significant (F[1,49] = 4.65, p<.05, partial η2 = .087). 5 Discussion This study examined the relation between receptive L2 vocabulary knowledge and level of non-verbal intelligence (NVI) in children at age three to four years who learnt English in one of two German-English bilingual pre-schools. The general level of the children’s non-verbal intelligence was tested with the SON-R 49 Early immersion education <?page no="50"?> test times T 2 T 1 mean raw scores BVPS III 40 35 30 25 20 15 10 IQ > 90, n = 4 IQ 90 - 110, n = 13 IQ < 90, n = 12 IQ groups Figure 2: BPVS III raw scores as obtained at T1 and T2, in relation to NVI scores (n = 29). 2½ - 7. Twelve of the 29 children turned out to have an NVI that is below average (group I). Almost the same number of children (n = 13) had an average NVI (group II). Four children scored higher than average (group III). The high number of children with a below-average NVI score may be surprising. However, this could be due to the fact that a relatively high percentage of the children had little access to education in their homes, i. e. they often came from an underprivileged socio-economic and educational background. The correlation between low scores in NVI tests and low socio-economic background has also been reported in other studies (see e. g. Lervåg, Dolean, Tincas & Melby‐Lervåg, 2019, for a review). The numbers of children with an average and above-average NVI, though, correspond to the standard distribution (see Resing & Blok 2002). In the English vocabulary test BPVS III, the children generally obtained higher scores at T2 than at T1, thus increasing their receptive L2 vocabulary as a function of time. These results correspond with findings of other studies on the development of L2 English receptive vocabulary in the bilingual pre-school context (cf. Couve de Murville et al., 2016; Häckel, 2013; Häckel & Piske, 2016; Rohde, 2010; Steinlen, 2016a, b; Weitz, 2015). With regard to the relationship between non-verbal intelligence and L2 re‐ ceptive vocabulary knowledge, the results of ANOVAs with repeated measures 50 Katja Schmidt / Yvonne Blumenthal <?page no="51"?> showed that the amount of progress (i.e. the scores at T1 to T2) differed signif‐ icantly among the three groups, yielding large effect sizes of time, NVI-group, and the interaction time x NVI group. On the one hand, the growth rates of groups I and II (i.e. children with NVI scores below-average and on average) are comparable to one another, and no significant group differences were noted, achieving similar results for the BPVS III at T1 and T2. It seems, therefore, that receptive L2 vocabulary acquisition evolves similarly in children with a low NVI and children with an average NVI. On the other hand, significant differences were noted for group III at both test times. These children with above-average NVI scores nearly doubled their BPVS scores from T1 to T2, and their scores at T1 and T2 were considerably higher than the T2 scores of the other two groups. Their levels of receptive L2 vocabulary knowledge lies approximately three times higher than those of their peers with an average or below-average NVI. In addition, the L2 contact time also had a significant impact, with a medium effect size on the development of vocabulary skills from T1 to T2. Considering the fact that the children’s average length of exposure to English was only 21.8 months at T2, this increase is amazing. Thus, Rohde’s (2010) premise that a significant improvement of receptive vocabulary can only be measured after an extended period of time (25-72 months) does not seem to hold true for children with a high level of non-verbal intelligence. Their use of language learning strategies may also account for these results. As informal observations in the bilingual pre-schools indicated, all four children with an above-average NVI applied some sort of learning strategies, especially during testing. For example, one child always repeated the word mentioned by the test administrator without being asked to do so. Another child regularly created links between the sound of the English word and the German equivalent, e. g. for astronaut she said “oh das klingt wie auf Deutsch. Das ist das Bild” (this sounds like in German, it is this picture). Through her awareness of sound similarities between English and German, she was able to deduce the meaning of a particular English word. Thus, it seems that children with higher scores in the NVI test, even at pre-school age, demonstrate a kind of metalinguistic awareness and make use of specific language learning strategies that the children of the other two groups had not been observed to use. This may have helped the children of group III to improve their receptive vocabulary knowledge to a higher extent than the children of groups I and II were able to. While the study provided first findings on the relationship between non-verbal intelligence and L2 English receptive vocabulary in two bilingual pre-schools in Germany, there are some limitations. First, the results depend on the instrument and its operationalisation. An analysis with a different 51 Early immersion education <?page no="52"?> instrument (for example Raven’s Progressive Matrices for non-verbal intelligence or other tests on general intelligence, e. g. Wechsler Intelligence Scale for Children-Revised or British Ability Scales) may have led to different results. Second, this study presented a small sample with only 29 pre-schoolers, and the subgroup of children with NVI scores above-average comprised only four children, which further limits the generalisability of the results. The small sample size also negatively affects the power of any statistical analysis (Döring & Bortz, 2016), i. e. any result of this study must be interpreted with caution. Third, additional language measures should be employed in future studies to further corroborate the findings of the present study on L2 receptive vocabulary, for example by also assessing L1 receptive vocabulary (e.g. Miralpeix, 2019) to examine the link between vocabulary in L1 and L2 acquisition, and/ or L2 expressive vocabulary (see e. g. Gibson, Oller, Jarmulowicz & Ethington, 2012) to explore this relationship for the bilingual pre-school context. Finally, the role of non-verbal intelligence in vocabulary learning is not yet well understood, and additional studies are needed to explore whether such a link actually exists (e.g. Santos, de Andrade Varanda, Barbosa, Ya I Sun, de la Higuera Amato & Fernandes, 2016). Thus, the literature review and the results of the present study raise some important issues and indications for future research. 6 Conclusion The results of the present study indicate a relationship between L2 vocabulary acquisition and the level of non-verbal intelligence, particularly for children with above-average NVI, who scored significantly higher on receptive L2 vocabulary than children with low and average NVI. This result suggests that children with an NVI higher than average acquire (receptively) more L2 words and at a faster rate in early immersion pre-school programmes than their peers. Furthermore, there was no significant difference for L2 receptive vocabulary by children with low NVI and with average NVIs, indicating that both groups acquired English words at the same rate and to the same extent. Thus, L2 acquisition (of words in particular) in early immersion education seems to be possible for all children, regardless of their cognitive abilities (as operationalised by non-verbal intelligence), albeit at different rates. References Birdsong, D. 1999. Second Language Acquisition and the Critical Period Hypothesis. London: Lawrence Erlbaum. 52 Katja Schmidt / Yvonne Blumenthal <?page no="53"?> Bleyhl, W. 1995. Wortschatz und Fremdsprachenunterricht oder: Das Problem sind nicht die Lerner. In: Bausch, K.-R., Christ H., Königs F. G. & Krumm H.-J. (Eds.), Erwerb und Vermittlung von Wortschatz im Fremdsprachenunterricht. Arbeitspapiere der 15. Frühjahrskonferenz zur Erforschung des Fremdsprachenunterrichts. Tübingen: Narr. 20-31. Brooks, P.J. & Kempe, V. 2013. Individual differences in adult foreign language learning: The mediating effect of metalinguistic awareness. Memory and Cognition, 41, 281-296. Cattell, R.B. 1971. Abilities: Their Structure, Growth and Action. Boston, MA: Houghton-Mifflin. Cameron, L. 2001. Teaching Languages to Young Learners. Cambridge: Cambridge Uni‐ versity Press. Chapelle, C. 1998. Construct definition and validity inquiry in SLA research. In: Bach‐ mann, L.F. & Cohen, A. D. (Eds.), Interface between Second Language Acquisition and Language Testing Research. Cambridge: Cambridge University Press. 32-70. Chilla, S., Rothweiler, M. & Babur, E. 2010. Kindliche Mehrsprachigkeit. Grundlagen - Störungen - Diagnostik. München: Reinhardt. Cohen. J. 1988. Statistical Power Analysis for the Behavioral Sciences. 2 nd ed. Hillsdale, NJ: Lawrence Erlbaum. Coon, D. & Mitterer, J.O. 2010. Introduction to Psychology: Gateways to Mind and Behavior With Concept Maps. Belmont, CA: Wadsworth. Couve de Murville, S., Kersten, K., Maier, E., Ponto, K. & Weitz, M. 2016. Rezeptiver L2 Wortschatz in der Grundschule. In: Steinlen, A. & Piske, T. (Eds.), Wortschatzlernen in bilingualen Schulen und Kindertagesstätten. Frankfurt am Main: Peter Lang. 85-121. Crain, S. & Thornton, R. 1998. Investigations in Universal Grammar: A Guide to Experi‐ ments in the Acquisition of Syntax and Semantics. Cambridge: The MIT Press. Daniel, A. 2001. Lernerwortschatz und Wortschatzlernen im bilingualen Unterricht. Frank‐ furt am Main: Peter Lang. Döpke, S. 1992. One Parent One Language. An Interactional Approach. Amsterdam: Benjamins. Döring, N. & Bortz, J. 2016. Forschungsmethoden und Evaluation. 5 th ed. Heidelberg: Springer. Duff, P. & Polio, C. 1990. How much foreign language is there in the foreign language classroom? The Modern Language Journal, 74 (2). 154-166. Dunn, L. & Dunn, D. 2009. The British Picture Vocabulary Scale. 3 rd ed. London: GL Assessment. Ellis, R. 2008. The Study of Second Language Acquisition. 2 nd ed. Oxford: Oxford University Press. Gardner, H. 1993. Multiple Intelligences. The Theory in Practice. New York, NY: Basic Books. 53 Early immersion education <?page no="54"?> Gass, S., Behney, J. & Plonsky, L. 2013. Second Language Acquisition. An Introductory Course. 4 th ed. New York, NY: Routledge. Genesee, F. 1976. The role of intelligence in second language learning. Language Learning, 26 (2), 267-280. Genesee, F. 1987. Learning Through Two Languages: Studies of Immersion and Bilingual Education. Rowley, MA: Newbury House. Gibson, T.A., Oller, D.K., Jarmulowicz, L. & Ethington, C.A. 2012. The receptive-expres‐ sive gap in the vocabulary of young second-language learners: Robustness and possible mechanisms. Bilingualism: Language and Cognition, 15 (1), 102-116. Häckel, A. 2013. Bilinguale Kita-Einrichtungen: Untersuchungen zur sprachlichen Entwick‐ lung deutsch-englisch bilingual betreuter Kita-Kinder. Baltmannsweiler: Schneider Verlag Hohengehren. Häckel, A. & Piske, T. 2011. Zur Entwicklung der sprachlichen Fähigkeiten bei deutsch-englisch bilingual betreuten Kindergartenkindern mit und ohne Migrationshintergrund. In: Krafft, A. & Spiegel, C. (Eds.), Sprachliche Förderung und Weiterbildung - transdisziplinär. Frankfurt am Main: Peter Lang. 11-31. Häckel, A. & Piske, T. 2016. Welche Rolle spielen Faktoren wie fremdsprachliche Kontaktzeit, Alter, Geschlecht und Migrationshintergrund beim frühen Fremdspra‐ chenlernen in bilingualen Kindertageseinrichtungen? In: Steinlen, A. & Piske, T. (Eds.), Bilinguale Programme in Kindertageseinrichtungen. Umsetzungsbeispiele und Forschungsergebnisse. Tübingen: Narr. 161-193. Horn, J.L. 1967. Intelligence - Why it grows, why it declines. Trans-Action, 5, 23-31. Horn, J.L. & Cattell, R.B. 1967. Age differences in fluid and crystallized intelligence. Acta Psychologica, 26 (2), 107-129. Jalkanen, J. 2009. Language proficiency and age-appropriate methodology at the English kindergarten of Kuopio. In: Nikoloy, M. (Ed.), The Age Factor and Early Language Learning. Berlin: Mouton De Gruyter. 97-117. Kersten, K. 2019. Einflussfaktoren im bilingualen Fremdsprachenerwerb. In: Rohde, A. & Steinlen, A.K. (Eds.), Sprachenvielfalt als Ressource begreifen: Mehrsprachigkeit in bilingualen Kindertagesstätten und Schulen (Band II). Berlin: Dohrmann Verlag. 35-70. Kersten, K., Drewing, M., Granados, J., Leloux, B., Lommel, A., Schneider, A. & Taylor, S. 2010. How to start a bilingual preschool: Practical guidelines. In: Kersten, K., Rohde, A., Schelletter, C. & Steinlen, A. (Eds.), Bilingual Preschools Volume 2: Best Practices. Trier: Wissenschaftlicher Verlag Trier. 77-101. Kievit, R.A., Lindenberger, U., Goodyer, I.M., Jones, P.B., Fonagy, P., Bullmore, E.T., The Neuroscience in Psychiatry Network & Dolan, R.J. 2017. Mutualistic coupling between vocabulary and reasoning supports cognitive development during late adolescence and early adulthood. Psychological Science, 28 (10), 1419-1431. 54 Katja Schmidt / Yvonne Blumenthal <?page no="55"?> Klieme, E. 2006. Zusammenfassung zentraler Ergebnisse der DESI-Studie. https: / / ww w.kmk.org/ fileadmin/ veroeffentlichungen_beschluesse/ 2006/ 2006_03_01-DESI-Ausg ewaehlte-Ergebnisse.pdf. May 31, 2021. Laufer, B. & Goldstein, Z. 2004. Testing vocabulary knowledge: Size, strength, and computer adaptiveness. Language Learning, 54 (3), 399-436. Lervåg, A., Dolean, D., Tincas, I. & Melby‐Lervåg, M. 2019. Socioeconomic background, nonverbal IQ and school absence affects the development of vocabulary and reading comprehension in children living in severe poverty. Developmental Science, doi: 10.1111/ desc.12858. Levelt, W. 1989. Speaking: From Intention to Articulation. Cambridge: MIT Press. Miralpeix, I. 2019. L1 and L2 vocabulary size and growth. In: Webb, S. (Ed.), The Routledge Handbook of Vocabulary Studies. Abingdon: Routledge. 189-206. Nation, I.S.P. 2001. Learning Vocabulary in Another Language. Cambridge: Cambridge University Press. Nikolov, M. 2009. The age factor in context. In: Nikoloy, M. (Ed.), The Age Factor and Early Language Learning. Berlin: Mouton De Gruyter. 1-37. Paradis, M. 2004. A Neurolinguistic Theory of Bilingualism. Amsterdam: John Benjamins. Paradis, J., Genesee, F. & Crago, M. 2011. Dual Language Development & Disorders. A Handbook on Bilingualism & Second Language Learning. 2 nd ed. Baltimore, MD: Paul Brookes Publishing. Piske, T. 2013. Bilingual education: Chances and challenges. In: Elsner, D. & Keßler, J.-U. (Eds.), Bilingual Education in Primary School. Aspects of Immersion, CLIL, and Bilingual Modules. Tübingen: Narr. 28-40. Read, J. 2011. Assessing Vocabulary. Cambridge: Cambridge University Press. Resing, W.C.M. & Blok, J.B. 2002. The classification of intelligence scores. Proposal for an unambiguous system. The Psychologist, 37, 244-249. Ringbom, H. & Jarvis, S. 2009. The importance of cross-linguistic similarity in foreign language learning. In: Long, M. & Doughty, C. (Eds.), The Handbook of Language Teaching. Oxford: Blackwell. 106-116. Rohde, A. & Tiefenthal, C. 2000. Fast mapping in naturalistic L2 acquisition. Studia Linguistica, 54 (2), 167-174. Rohde, A. & Tiefenthal, C. 2002. On L2 lexical learning abilities. In: Burmeister, P., Piske, T. & Rohde, A. (Eds.), An Integrated View of Language Development. Papers in Honor of Henning Wode. Trier: Wissenschaftlicher Verlag Trier. 449-471. Rohde, A. 2010. Receptive L2 lexical knowledge in bilingual preschool children. In: Kersten, K., Rohde, A., Schelletter, C. & Steinlen, A. (Eds.), Bilingual Preschools Volume 1: Learning and Development. Trier: Wissenschaftlicher Verlag Trier. 45-68. 55 Early immersion education <?page no="56"?> Salehi, M. & Sadighi, F. 2012. The relationship between intelligence and foreign language learning, and the role of practice. Journal of Language, Culture, and Translation (LCT), 1 (1), 33-48. Santos, T.H.F., de Andrade Varanda, C., Barbosa, M.R.P., Ya I Sun, I., de la Higuera Amato, C.A. & Fernandes, F.D.M. 2016. Correlations between nonverbal intelligence and communicative abilities of Brazilian children with autism spectrum disorders. Austin Journal of Autism & Related Disabilities, 2 (1), 1013. Saunders, G. 1988. Bilingual Children: From Birth to Teens. Philadelphia, PA: Clevedon Avon. Schelletter, C. & Ramsey, R. 2010. Lexical and grammatical comprehension in monolin‐ gual and bilingual children. In: Kersten, K., Rohde, A., Schelletter, C. & Steinlen, A. (Eds.), Bilingual Preschools Volume 1: Learning and Development. Trier: Wissenschaft‐ licher Verlag Trier. 101-118. Shekan, P. 2012. Language aptitude. In: Gass, S. & Mackey, A. (Eds.), The Routledge Handbook of Second Language Acquisition. London: Routledge. 381-395. Singleton, D. & Ryan, L. 2004. Language Acquisition: The Age Factor. Clevedon: Multilin‐ gual Matters. Spadaro, K. 2013. Maturational constraints on lexical acquisition in a second language. In: Granena, G. & Long, M. (Eds.), Sensitive Periods, Language Aptitude, and Ultimate Attainment. Amsterdam: John Benjamins. 43-68. Steinlen, A.K. 2013. “Flera språk - fler möjligheter” - Immigrant children’s acquisition of English in bilingual preschools. In: Flyman Mattsson, A. & Norrby, C. (Eds.), Language Acquisition and Use in Multilingual Contexts. Lund: Media Tryck. 170-184. Steinlen, A.K. 2016a. Die Entwicklung des Englischen und des Deutschen von Kindern mit Migrationshintergrund in bilingualen Kindergärten. In: Steinlen, A. & Piske, T. (Eds.), Bilinguale Programme in Kindertageseinrichtungen. Umsetzungsbeispiele und Forschungsergebnisse. Tübingen: Narr. 197-218. Steinlen, A.K. 2016b. Das Hörverständnis der englischen Grammatik und des englischen Wortschatzes in einem bilingualen Kindergarten. In: Steinlen, A.K. & Piske, T. (Eds.), Bilinguale Programme in Kindertageseinrichtungen: Umsetzungsbeispiele und Forschungsergebnisse. Tübingen: Narr. 139-160. Steinlen, A.K. & Rogotzki, N. 2009. Comprehension of L2 grammar in a bilingual preschool: A developmental perspective. In: Marinis, T., Papangeli, A. & Stojanovik, V. (Eds.), Proceedings of the Child Language Seminar 2007 - 30th Anniversary. Reading: University of Reading. 163-173. Steinlen, A.K., Kersten, K. & Piske, T. 2019. Mehrsprachige Jungen als Problemfall in bi‐ lingualen Kitas? Zur Rolle von Geschlecht und sprachlichem Hintergrund. In: Rohde, A. & Steinlen, A. (Eds.), Sprachenvielfalt als Ressource begreifen. Mehrsprachigkeit in bilingualen Kindertagesstätten und Schulen Band II. Berlin: Dohrmann. 71-88. 56 Katja Schmidt / Yvonne Blumenthal <?page no="57"?> Stern, H.H. 1963. Foreign Languages in Primary Education. Hamburg: Unesco Institute for Education. Targonska, J. & Stork, A. 2013. Vorschläge für ein neues Modell zur Beschreibung und Analyse lexikalischer Kompetenz. Zeitschrift für Fremdsprachenforschung, 24 (1), 71-108. Teichroew, F.J. 1982. Receptive vs. productive vocabulary: A survey. Interlanguage Studies Bulletin, 6, 5-33. Tellegen, P.J., Laros, J.A. & Petermann, F. 2007. SON-R 2 ½-7. Non-Verbaler Intelligenztest. 2 nd ed. Göttingen: Hogrefe. Tellegen, P.J., Winkel, M., Wijnberg-Williams, B.J. & Laros, J.A. 2009. SON-R 2 ½-7. Manual and Research Report. Göttingen: Hogrefe. Thornbury, S. 2002. How to Teach Vocabulary. Harlow: Longman. Ushioda, E. & Dörnyei, Z. 2012. Motivation. In: Gass, S. & Mackey, A. (Eds.), The Routledge Handbook of Second Language Acquisition. London: Routledge. 396-409. Weitz, M. n.d. ELIAS British Picture Vocabulary Scale II (BPVS II). Information & Guidelines. www.elias.bilikita.org/ docs/ elias_lexicon_test_bpvs_ii_information_and _guidelines.pdf. May 31, 2021. Weitz, M. 2015. Die Rolle des L2-Inputs in bilingualen Kindergärten. Frankfurt am Main: Peter Lang. Wilkins, D. A. 1972. Linguistics in Language Teaching. London: Arnold. Wode, H. 1995. Lernen in der Fremdsprache. Grundzüge von Immersion und bilingualem Unterricht. Ismaning: Huber. Wode, H. 2001. Multilingual education in Europe. What can preschools contribute? In: Björklund, S. (Ed.), Language as a Tool. Immersion Research and Practice. Vaasa: University of Vaasa. 224-246. Zydatiß, W. 2000. Bilingualer Unterricht in der Grundschule: Entwurf eines Spracherwerbskonzepts für zweisprachige Immersionsprogramme. Ismaning: Huber. 57 Early immersion education <?page no="59"?> Developmental differences in executive functioning in bilingual, early-English and monolingual children: Group differences and individual differences both matter Claire Goriot Abstract In this study, it was investigated whether child Dutch-English bilinguals, Dutch children learning English as a second language in schools from kindergarten onwards (early-English education), and Dutch (functionally) monolinguals differ in their Dutch or English language development and/ or their executive functioning development (switching and working memory). In addition, the relation between language balance and executive functioning performance in bilingual children was examined. Participants were children of three age groups: 4-5-year-olds, 8-9-year-olds, and 11-12-year-olds. Executive functions (switching and working memory) and receptive Dutch and English vocabulary were measured. Bilingual children had larger English vocabularies and more balanced Dutch-English lexicons than early-English pupils. Early-English pupils did not have greater Eng‐ lish vocabulary (except for those in the 11-12-year-old group) or more balanced lexicons (except for 4-5-year-olds) than their monolingual peers. Bilingual and monolingual children of all three age groups outperformed their age-matched peers from early-English on verbal working memory. For switching, no differences between the groups were found. For the bilingual group, lexical balance, which was computed by dividing the English vocabulary scores by the Dutch vocabulary scores, was related to verbal working memory. The results of this research provide evidence that not only group differences in bilingual status may play a role in executive functioning development, but that individual differences in language development do, too. <?page no="60"?> 1 Introduction More and more countries in Europe start providing early foreign language ed‐ ucation to primary-school pupils (Enever, 2011; Enever, 2013). Such programmes are initially meant to increase pupils’ proficiency in the foreign language (Enever, 2011), and sometimes also their international awareness (Nuffic, 2018). From research with early bilingual children, who grew up with two languages, it is, however, known that learning an additional language may not only result in the mastery of two languages but it may also influence cognitive development. It should be noted that this claim is controversial. Some studies have failed to find differences between bilinguals and monolinguals in executive functioning (Paap & Greenberg, 2013; Paap, Myuz, Anders, Bockelman, Mikulinsky & Sawi, 2017). Other studies, though, have found that children who mastered two languages outperform their monolingual peers on cognitive flexibility and inhibition tasks (Barac, Moreno & Bialystok, 2016; Poarch & van Hell, 2012). Such studies have not only incorporated early bilinguals who grew up with two languages but have also investigated the relation between (emerging) bilingualism and cognitive development in early foreign language learners. This chapter provides an overview of research on executive functioning development in early foreign language learners. In addition, the results of a study on the relation between bilingualism and executive functioning are presented. First, group differences in language development and executive functioning development between functionally monolinguals, early foreign language learners, and early bilinguals are investigated. Second, the relation between lexical balance and executive functioning in early bilinguals is examined. 2 The relation between bilingualism and executive functioning Researchers have claimed that bilingual children show advanced cognitive control when compared to their monolingual peers (Adesope, Lavin, Thompson & Ungerleider, 2010). One of the most accepted hypotheses behind this claim is that bilinguals constantly have to suppress one of their languages because both are always active and competing with each other. Even though the suppression of this competition is mostly an unconscious process, this monitoring of two languages calls on processes that are related to cognitive control, in particular executive functioning processes. This would in turn enhance bilinguals’ execu‐ tive functioning (Green, 1998; Green & Abutalebi, 2013). Executive functions are multiple inter-related processes that are involved in behavioural control (Diamond, 2013). A commonly accepted division into 60 Claire Goriot <?page no="61"?> sub-processes is that of Miyake and colleagues who distinguish three executive functioning processes (Miyake, Friedman, Emerson, Witzki, Howerter & Wager, 2000), i. e. switching, inhibition, and working memory. Switching is the ability to alternate between switching rules. Inhibition can be described as the process of overcoming irrelevant responses. Working memory is the ability to remember, manipulate and retrieve information, and can be subdivided into verbal and non-verbal working memory. Since monitoring two languages requires executive control, it may be as‐ sumed that competition between languages is positively related to executive functioning. If that is true, it may then be assumed that the greater competition children experience, the further their executive functioning is developed. If those with more equally developed language proficiency in both languages experience the greatest competition, executive functioning development would consequently be most developed in bilingual children. If these assumptions hold, a relation should be observed between language balance (or the proficiency in two languages) and executive functioning. Following this reasoning, foreign language learners should have more developed executive functions than mon‐ olinguals, but not as developed as early bilinguals, given that early bilinguals are probably more balanced in language proficiency than foreign language learners. Indeed, previous researchers have tested this hypothesis by comparing groups with different proficiency levels in the second language, assuming that those with more experience in the L2 can be considered more balanced bilinguals. In a study by Carlson & Meltzoff (2008), children enrolled in immer‐ sion education did not perform significantly better than English monolingual children on an attention conflict task, whereas Spanish-English bilingual chil‐ dren outperformed both other groups. Poarch & van Hell (2012) found that German children enrolled in immersion education did not perform significantly different from monolingual or bilingual children, whereas the German-English bilingual children outperformed the monolingual children. According to the authors, the bilingual children had more experience with lexical control than the children in immersion education, who had only attended this type of education for less than two years. This would explain the bilinguals’ superior performance. In yet another study in which two groups of immersion education pupils were compared to a group of monolingual children (Purić, Vuksanović & Chondrogianni, 2017), the so-called high exposure group (who was enrolled in the foreign language for five hours a day) performed significantly better on working memory tasks than the two other groups, whereas the low exposure group (that received one-and-a-half hours of foreign language education a day) did not significantly differ from the monolingual group. In these studies, groups 61 Developmental differences in executive functioning <?page no="62"?> of children who differed in their experience with the second language were compared. The underlying assumption seems to be that children who have more experience with the second language are automatically more balanced in their language proficiency (thus in both the L1 and L2) than children who have less experience with the second language. Language balance, however, is dependent on the development of both lan‐ guages. Some researchers have indeed taken into account language balance in relation to children’s cognitive abilities, mainly with early bilingual populations (see for example Gathercole, Thomas, Kennedy, Prys, Young, Guasch, Roberts, Hughes & Jones, 2014), although some studies on second language learners exist, too. For example, Prior, Goldwasser, Ravet-Hirsh & Schwartz (2016) examined executive functioning development in children who had Russian as their home language and learnt Hebrew at school. They investigated children’s vocabulary in Russian and Hebrew and classified children either as balanced or unbalanced. The balanced and unbalanced groups’ performance on an inhibition and a switching task was then compared to that of a Hebrew monolingual group. Bilingual children of 4-5 years old - either balanced or unbalanced in language proficiency - did not perform significantly different from their peers on any of the two tasks. Balanced bilingual children of approximately ten years old, however, showed (marginally) significant advantages in the inhibition task, although only in their reaction times (they were faster) and in congruency effects (they had less difficulty overcoming the incongruent trials). There were no group differences in accuracy on the task. Other researchers have treated balance as a continuum. It was, for example, found that Dutch-Turkish bilingual pre-schoolers who were more equally proficient in both Dutch and Turkish, showed greater verbal and non-verbal working memory skills than children with less balanced proficiency (Blom, Küntay, Messer, Verhagen & Leseman, 2014). In a longitudinal study on Fri‐ sian-Dutch bilingual children (Bosma, Hoekstra, Versloot & Blom, 2017), it was found that at age 5-6 language balance was positively related to selective attention. When children were 6-7 and 7-8 years old, however, such an effect did not exist anymore. In yet another study, the relation between language balance and executive functions was examined in 8-9-year-old Spanish-English bilingual children. The children were either following English monolingual or Spanish-English immersion education. It was found that language balance was related to cognitive control and working memory (Thomas-Sunesson, Hakuta & Bialystok, 2018). The aforementioned studies mainly focused on bilingual children who learnt a minority language at home and the majority language at school. Goriot, 62 Claire Goriot <?page no="63"?> Broersma, McQueen, Unsworth & van Hout (2018a) investigated the role between language balance and executive functioning in Dutch children who were either attending mainstream Dutch primary education or who received English lessons from the age of four for approximately one hour per week. The results showed that language balance was related to switching performance, but not to inhibition or working memory performance. The authors concluded that being exposed to a foreign language for a short amount of time not only enhances pupils’ development in that language, which may in turn positively influence switching abilities - although a causal relation cannot be ensured. These studies thus showed that degree of bilingualism in terms of language balance is positively related to executive functioning performance, although it should be noted that results are not always straightforward (Prior et al., 2016) because sometimes researchers did find a relation between language balance and executive functioning and sometimes they did not (Blom et al., 2014; Bosma et al., 2017; Goriot et al., 2018a; Prior et al., 2016). Studies that treated language balance as a continuum focused mainly on chil‐ dren who had relatively little experience with the additional language, coming either from minority language backgrounds (e.g. Blom et al., 2014; Bosma et al., 2017) or having very little exposure to the second language (Goriot et al., 2018a). In addition, children in all studies were relatively young, being mostly in pre-school or midst of primary school. In contrast to the previous studies, this study investigates whether fully bilingual children who grew up with two languages show more advanced executive functioning performance than functionally monolingual and/ or early foreign language learners. In addition, it is investigated what the relation between language balance and executive functioning performance is in the group of fully bilingual children. Data come from three groups of children of three different age groups (4-5 [start of primary school], 8-9 [middle of primary school], and 11-12-year-olds [end of primary school]). The first group are functionally monolinguals (henceforth ‘monolinguals’), i. e., Dutch-speaking children who attend Dutch primary school and who are not formally exposed to English lessons until the penultimate grade (when they are approximately 11 years old). Given that English is very present in Dutch society (Kuppens, 2010), they may have had some contact with English through media such as television or video games. The second group are early-English learners, i. e., Dutch-speaking children who attend a primary school in which English lessons are provided from the start (at age four) onwards (Nuffic, 2018). The data from these two groups have already been compared previously (Goriot et al., 2018a). Given that the bilingual experience of both groups was very limited, it is investigated in this study whether children with 63 Developmental differences in executive functioning <?page no="64"?> more bilingual experience show more developed executive functions than the groups with limited bilingual experience. These simultaneous bilingual children were exposed to English because at least one of their parents was a native speaker of English. They learnt Dutch before the age of four. The first question is whether the three groups (monolingual, early-English early bilingual) differ from each other in language development. The hypothesis is that the bilinguals will not differ significantly from monolingual and early-English children in Dutch vocabulary, but will have greater English vocabulary and will, consequently, be more balanced in their lexicons. Mirroring the findings of the previous study (Goriot et al., 2018a), it is expected that the early-English and monolingual pupils do not differ from each other in the size of their receptive Dutch vocabulary. Due to the little exposure (one hour per week) to English, it may take time before early-English pupils are distinctively more experienced in English than functionally monolingual children. Therefore, differences between these two groups in English vocabulary are expected to emerge in the eldest group only. The second question is whether the three groups will differ in terms of their executive functioning performance. It is expected that bilingual children will perform better on executive functioning tasks than monolinguals and early-English children, who will not differ from each other. The third and final question is whether a relation between language balance and executive function performance exists in the bilingual group. It is hypothesised that, like in previous research, language balance will be positively related to bilinguals’ executive functioning skills. In this study, bilingualism will not only be operationalised as a categorical variable but also as gradient by investigating individual differences in L1 and L2 balance, which may range from completely unbalanced to perfectly balanced. The results help to specify the complex variable that bilingualism is, as well as to provide insight into the relation between language development and executive functions in children with varying degrees of bilingualism. 3 Method 3.1 Participants Participants were three groups of pupils, i. e. (functionally) Dutch pupils at‐ tending mainstream Dutch primary-schools (n = 98), Dutch pupils attending primary schools where early English lessons were provided (n = 106), and Dutch-English bilingual pupils who grew up with both English and Dutch as their home language (n = 63). Pupils belonged to one of three age groups: 4-5-year-olds (i.e. kindergarten; n = 99), 8-9-year-olds (i.e. middle of primary 64 Claire Goriot <?page no="65"?> school; n = 95), and 11-12-year-olds (i.e. final year of primary school; n = 73). The Dutch and early-English pupils came from four early-English and four mainstream schools. Pupils from mainstream schools are called functionally monolingual, since they may encounter some English in daily life (for example through media) and may strictly speaking not be called monolingual, but they only use the Dutch language in practice, and their knowledge of English is very limited. They did not receive any English lessons until the penultimate grade. At that moment, when pupils are around eleven years old, they receive compulsory English lessons for 45-60 minutes per week. The early-English schools had eight or more years of experience with providing English lessons. From the moment they entered kindergarten at age four, pupils received approximately one hour of English lessons per week. All other lessons are in Dutch. The four mainstream schools were matched to the early-English schools in the urbanised/ rural area, the average income in the neighbourhood, whether the school followed a certain educational philosophy, and religious denomination. Head teachers of the schools were asked to select ten children per age group for participation. The bilingual children were recruited via online advertisements on webpages targeted at bilingual families. Children were eligible for partici‐ pation if they had no speech, language or developmental disorders, had no sight or hearing impairments, and in the case of the bilingual children, had at least one parent who was a native speaker of English. Informed consent for participation was given by the parents of all children. Parents were asked to fill in a questionnaire about out-of-school exposure to English, but unfortunately, the response rate for monolingual and early-English pupils was below 50 %. Therefore, no reliable statistical analyses could be performed on these data in order to investigate whether differences between groups existed in exposure to English. In general, bilinguals were exposed more frequently to English media than early-English and monolingual children. 3.2 Design The research had a cross-sectional design, in which three groups of children who differed in the extent to which they had experience with two languages were compared, namely functionally monolinguals, early-English learners, and bilingual children. The children came from three age groups (4-5, 8-9, and 11-12-year-olds), to investigate whether differences between groups would show once differences in experience got larger due to longer exposure to two languages. In addition, individual differences in language balance were related to individual differences in executive functioning. 65 Developmental differences in executive functioning <?page no="66"?> 3.3 Instruments 3.3.1 Executive functioning Switching Children performed the Dimensional Change Card Sort task (DCCS; Zelazo, 2006), which consisted of three phases. First, one card of a red boat and one card of a blue rabbit were presented. Those cards remained visible during the remainder of the task. The 24 experimental cards either had an image of a blue boat or an image of a red rabbit. The number of blue boats and red rabbits was equally divided. During the pre-switch phase, children were asked to sort six cards on colours. In the post-switch phase, they were asked to sort six cards on shape. If they performed correctly on at least five of these last six trials, they were promoted to the border phase. In the border phase, which was the final phase, they were asked to sort cards on colour if they had a black border (n = 6), and on shape, if there was no border (n = 6). As a pilot study showed that children often forgot the rules while doing the task, it was decided to repeat the instructions halfway through the border phase. In all phases, the order of the cards was pre-determined, therefore being the same for all participants. A card with a border or with the same picture never followed a similar card more than once. Cards with blue boats and red rabbits were equally distributed in all three phases. The final score for switching ability was determined by computing the total number of all trials on which participants performed correctly, and thus ranged between zero and 24. The test took five minutes to administer. Verbal and non-verbal working memory The Automated Working Memory Assessment (AWMA; Alloway, Gathercole, Kirkwood & Elliot, 2008) is a computerised test for working memory skills. The subtest Backward Digit Recall was used to test participants’ verbal working memory. In this test, children were asked to repeat an auditorily presented string of digits in reverse order. Participants started with strings of two digits, and after correct performance of two subsequent trials were presented with strings of one digit more up to a maximum of nine digits. The Odd One Out subtest was used to assess participants’ non-verbal working memory. In this test, children saw three shapes on a row and were asked to indicate the one shape that looked different from the other two. Thereafter, the shapes disappeared, and empty boxes remained. Participants were asked to remember the location of the odd shape. Participants started with one trio of shapes. After every two consecutive items that were performed correctly, an additional trio of shapes was presented. In that case, children had first to indicate the odd one out in the first trio, and then in the second. After having indicated all odd shapes, they were asked to indicate 66 Claire Goriot <?page no="67"?> the location of the odd shapes in the right order. A maximum of seven trios of shapes was presented. Both subtests started with four practice trials and ended after three incorrectly performed test trials. Test-retest coefficients for reliability are between .86 (Backwards Digit Recall) and .88 (Odd One Out) (Alloway et al., 2008). The maximum score for the Backwards Digit Recall was 162, and for the Odd One Out 36. The tasks took approximately five minutes each. English and Dutch vocabularies To assess receptive English vocabulary, the Peabody Picture Vocabulary Task (4 th edition) was used (PPVT-4; Dunn & Dunn, 2007). For Dutch receptive vocabulary, we used the PPVT-III Dutch (Dunn, Dunn & Schlichting, 2005). In both tests, children are presented with four pictures and an auditorily presented word which they have to match to the according picture. The English version consists of 228 words, the Dutch one of 204. In both versions, words are grouped in sets of twelve, and children start with a set that is appropriate for their age. Testing ends when eight (for the English version) or nine errors (for the Dutch version) are made in one set. The test-retest reliability coefficients for children aged between four and 13 range between .91 and .94 for the English version (Dunn & Dunn, 2007) and between .91 and .96 for the Dutch version (Dunn et al., 2005). The raw score was calculated as the number of correctly performed items. Norm scores were not calculated since these are based on native speakers and, therefore, not suitable for the use with L2 learners. Both the Dutch and English PPVT took approximately 15 minutes to administer. Balance Based on Blom et al. (2014), lexical balance was calculated by taking the natural logarithm of the proportion correct items on the PPVT-4 divided by the proportion correct on the PPVT-III Dutch (see Goriot et al., 2018a for a detailed description of the calculation). The outcome was then turned into an absolute number, such that participants with more positive values were either more proficient in English or Dutch, whereas those with a value closer to zero were more balanced. The absolute number instead of the original outcome was taken because in the original outcome the value could either be negative or positive. Bilingual children would likely more often obtain a positive value (more proficient in English than in Dutch) and (functionally) monolingual children a negative value (more proficient in Dutch than in English), but both would be unbalanced in their language proficiency. Intelligence Participants were presented with the (standard) shortened version of the Wechsler Scale of Ability (WNV; Wechsler & Naglieri, 2008), which consists 67 Developmental differences in executive functioning <?page no="68"?> of two subtests. All participants were presented with the subtest Matrices, in which they had to find the correct piece (among four or five alternatives) to complete a figural matrix. The total score ranged between zero and 41. This subtest measures fluid intelligence. Children of four or five years old performed the Recognition subtest, which measures short-term memory. First, a geometric design was presented for three seconds. Then, children had to recognise that design out of four or five alternatives that were presented on a new page. The total score ranged between zero and 21. In both Recognition and Matrices, testing ended when four out of five consecutive trials were performed incorrectly. Instead of Recognition, older children did the Spatial Span subtest that measures spatial orientation. They were presented with ten blocks that were placed irregularly on a wooden platform. The examiner tapped a pattern on the blocks, and children had to repeat that pattern. There was a forward phase in which they had to tap the same sequence and a backward phase in which they had to tap the pattern in reverse order. After two subsequently failed attempts testing ended. The score ranged between zero and 32. The total score of the two subtests was calculated as the sum of the scores on the two subtests. Internal consistency coefficients range between α = .63 to α = .78 (Wechsler & Naglieri, 2008). Since norm scores could not be calculated for the English PPVT, it was decided to work with the raw scores in all tests. 4 Procedure All children were tested individually. Bilingual children were tested in a quiet room at their home, monolingual and early-English pupils at school. Testing was conducted in two sessions of approximately 30 minutes. In the first session, the following tasks were administered: The DCCS, the AWMA (first Odd One Out and then Backward Digit Recall), and the PPVT-4. In the second session, the WNV (first Matrix Reasoning and then Recognition or Spatial Span) were administered first, followed by the PPVT-III Dutch. Monolingual and Early-English children also performed the Simon task (Simon & Small, 1967; see Goriot et al., 2018a for a more elaborate description), but due to technical issues that appeared in the phase of collecting data from bilingual children, it was decided to stop administering this test. For all tests, the task manuals were followed, except for the DCCS, but the general descriptions as outlined by Zelazo (2006) were followed for this test. There was at least one and maximally 35 days between two sessions, except for two children who did session one and two on the same day but with multiple hours in between. 68 Claire Goriot <?page no="69"?> 1 Since monolingual and early-English children were recruited at schools, the data have a multilevel structure. As bilingual children almost all attended different schools, the variable School could not be defined for these children. Multilevel analyses with School as a random factor (set to the same value for the bilinguals) showed that the effect of schools was very minimal (always < 5 %). The choice was made to report analyses with the most simple structure (Tabachnick & Fidell, 2007). 5 Analyses Linear models 1 were performed in R version 3.6.2. First, (language and age) group differences in English, Dutch and language balance were investigated. Thereafter, (language and age) group differences in switching, verbal, and non-verbal working memory were examined. To investigate whether language balance was related to bilingual children’s executive functioning performance, different linear models were compared. The basic model contained age group (4-5, 8-9, 11-12-year-olds) only. The first variable to add was always lexical balance. If the model fit did not improve (as indicated by a smaller AIC value [the Akaike Information Criterion is an indicator for the quality of statistical models, with lower values indicating better quality]), balance was removed and another variable was added. If the model fit did improve, it was checked whether adding intelligence could further improve the model. It was also investigated whether balance could be replaced by either Dutch or English vocabulary. Partial eta-squared was used as the measure for effect size. Values of between .02 and .13 were considered small, between .14 and .26 medium, and higher than .26 large (Cohen, 1988). 6 Results 6.1 Data cleaning Following the procedure described in Goriot et al. (2018a) data were removed from all children that were not able to correctly sort the cards in the shape-sorting phase of the DCCS (n = 27). These data were removed since not being able to make the switch from colour-sorting to shape-sorting indicates that the task is too difficult for a child (Zelazo, 2006). 6.2 Descriptive statistics The descriptive statistics of each group are presented in Table 1. This table shows that, in general, the bilingual group obtained the highest scores on the executive functioning (EF) measures (switching and working memory), except for the 4-5-year-old group. The bilinguals have larger English vocabularies than 69 Developmental differences in executive functioning <?page no="70"?> the two other groups. The 4-5-year-old bilingual children seem to have smaller receptive vocabularies in Dutch than their monolingual and early-English peers. Bilinguals have, in general, balanced vocabularies, with mean scores around zero. Monolingual and early-English pupils’ vocabularies are less balanced, due to the fact that their vocabularies in Dutch are usually larger than their vocabularies in English. Table 1: Descriptive statistics for the language measures and executive functioning tasks. Two Analyses of Variance were performed, with respectively age (in months) and intelligence as dependent variables. These analyses, with age group, bilin‐ gual category, and the interaction between age group and bilingual category, 70 Claire Goriot <?page no="71"?> revealed that there was a statistically significant difference in age between the three bilingual groups (F (2,258) = 13.25, p < .001, Np = .093). Tukey HSD post-hoc tests revealed that the bilingual children were younger than the monolingual (M dif = -4.84, p < .001) and the early-English pupils (M dif = -3.33, p = .001). For intelligence, there was a significant effect of age group (F (2,258) = 199.12, p < .001, Np = .606). The 8-9-year-old group obtained higher scores than the 4-5-year-old group (M dif = 10.21, p < .001), and the 11-12-year-old group obtained higher scores than the 8-9-year-old group (M dif = 5.09). This is a logical result, given that the raw scores on the intelligence test were used, which were not corrected for age as outlined in the method section. 6.3 Correlations between variables Table 2 shows that intelligence scores, English vocabulary, and Dutch vocabu‐ lary correlate significantly and positively with the three EF measures (switching, non-verbal working memory, and verbal working memory). Balance also sig‐ nificantly correlates with the three EF measures, indicating that those with a higher balance score (and thus less balanced lexicons) have lower scores on the EF measures. Age does not correlate with any of the other measures except for Dutch vocabulary, indicating that older children have higher Dutch vocabulary scores. 1. 2. 3. 4. 5. 6. 7. 8. 1. Switching 1 2. Non-verbal working memory .436 ** 1 3. Verbal working memory .405 ** .747 ** 1 4. Intelligence .432 ** .658 ** .744 ** 1 5. English vo‐ cabulary .438 ** .627 ** .623 ** .568 ** 1 6. Dutch vocabu‐ lary .488 ** .798 ** .773 ** .788 ** .701 ** 1 7. Balance -.357 ** -.542 ** -.601 ** -.547 ** -.841 ** -.623 ** 1 8. Age .057 .048 -.028 .078 -.054 .127 * -.052 1 **p < .001, *p < .05 Table 2: Correlations between EF, intelligence, vocabulary, balance and age. 71 Developmental differences in executive functioning <?page no="72"?> Figure 1 shows the correlations between English and Dutch vocabulary for the different age and bilingual groups. In general, both English and Dutch vo‐ cabulary are higher for older participants. Furthermore, the difference between Dutch and English vocabulary is, in general, lower for the older age groups. Englishvocabulary Dutchvocabulary Englishvocabulary Englishvocabulary agegroup 4-5year-olds 8-9year-olds 11-12year-olds bilingual status control early-English bilingual Figure 1: Correlations between English vocabulary and Dutch vocabulary for the dif‐ ferent age groups and the different bilingual groups (black circle = monolingual, grey square = early-English, star = bilingual). To investigate whether the groups differed in Dutch vocabulary, English vocabulary, or balance scores, three ANOVAs with these measures as dependent variables and age group, bilingual category and the interaction between group and category were performed. The results are presented in Table 3. For Dutch vocabulary, there were significant differences in the youngest age group: bilinguals had smaller Dutch vocabularies than early-English and functionally monolingual pupils. For English vocabulary, bilingual pupils in all age groups had significantly larger vocabularies than early-English and monolingual pupils. Only in the oldest age group, early-English pupils had significantly larger Eng‐ lish vocabularies than functionally monolingual pupils. With respect to balance, bilinguals in the 4-5 and 8-9-year-old age groups had smaller balance scores (indicating more balanced vocabularies) than early-English and functionally monolingual pupils. Early-English and monolingual pupils did not differ in 72 Claire Goriot <?page no="73"?> their balance scores, except for pupils in the youngest age group. In this group, early-English pupils had smaller balance scores than functional monolinguals. In the oldest age group there were no significant differences between the language groups at all. Figure 2 provides a visual overview of how the three language groups differed from each other. Dutch vocabulary English vocabu‐ lary Balance (abso‐ lute) F(df), η p2 F(df), η p2 F(df), η p2 Age group F(2,258) = 851.54 *** , Np2 = .87 F(2,258) = 406.82 *** , Np2 = .76 F(2,258) = 198.11 *** , Np2 = .61 Bilingual cate‐ gory F(2,258) = 0.11, Np2 = .00 F(2,258) = 274.29 *** , Np2 = .68 F(2,258) = 89.53 *** , Np2 = .41 Age group * bi‐ lingual category F(4,258) = 5.49 *** , Np2 = .08 F(4,258) = 4.52 ** , Np2 = .07 F(4,258) = 17.42 ** , Np2 = .21 R2 .866 .838 .705 Post-hoc 4-5: b < eE, m 8-9: b, eE, m 11-12: b, eE, m 4-5: b > eE, m 8-9: b > eE, m 11-12: b > eE > m 4-5: b < eE < m 8-9: b < eE, m 11-12: b, eE, m *** p < .001; ** p < .01; *p < .05 Table 3: The relation between age group, bilingual category and vocabulary measures. Figure 2: Differences between age and language groups in Dutch vocabulary, English vocabulary and language balance. 73 Developmental differences in executive functioning <?page no="74"?> 2 Mixed-effects model analyses largely reflected the outcomes of the analyses presented here, except for the effects of bilingualism, which were significant in the linear models but not in the linear mixed-effects models. 6.4 Executive functioning measures To investigate whether there were differences in executive functioning scores between children from the different age and bilingual groups, analyses of variance were performed. In these analyses, age group, bilingual category, and the interaction between age group and category were the independent variables, and the executive functioning measure (switching, verbal working memory or non-verbal working memory), the dependent variable. For switching, there was no significant interaction between age group and bilingual category, nor a significant main effect of bilingual category. There was a significant main effect of age group (F (2,227) = 30.31, p < .001). Post hoc analyses revealed that there were differences between age groups: 11-12-year-olds performed better than 8-9-year-olds, who in turn performed better than 4-5-year-olds. For verbal working memory, the interaction between age group and bilingual category was not significant. There were significant main effects for age group and bilingual status. Post hoc analyses showed that bilinguals and monolin‐ guals outperformed the early-English pupils. No differences existed between the bilingual and monolingual group. Age group differences also existed: 11-12-year-olds performed better than 8-9-year-olds, who in turn performed better than 4-5-year-olds. For non-verbal working memory, there was a marginally significant interac‐ tion effect of age group and bilingual category (p = .054). The main effects for age group and bilingual category were both significant. Post hoc analyses revealed that there were again differences between age groups: the oldest group outperformed the younger groups, which also significantly differed from each other. The bilingual and monolingual group outperformed the early-English group, but the bilinguals and monolinguals did not differ significantly from each other. In summary, the results of these analyses showed that there were differences in executive functioning scores between children of different age groups and, although for working memory measures only, differences between children with different bilingual experiences. In general, older children obtained higher scores on switching, verbal and non-verbal working memory tests, and bilingual children had higher scores on working memory measures than monolingual and early-English children. Early-English children’s scores on any of the three measures did not significantly differ from those of functional monolinguals 2 . 74 Claire Goriot <?page no="75"?> Table 4 shows an overview of the final model for each measure. Figure 3 gives a visual overview of how the bilingual groups and the age groups differ from each other. Switching Verbal working memory Non-verbal working memory Age group F(2,227) = 30.31 *** , η p2 = .21 F(2,252) = 220.33 *** , η p2 = 64. F(2,253) = 255.26 *** η p2 = . 67 Bilingual status F(2,227) = 2.55, η p2 = .02 F(2,252) = 6.65 ** , η p2 = .05 F(2,253) = 3.28 * , η p2 = .03 Age group * Bilingual status F(2,227) = 1.36, η p2 = .02 F(2,252) = 1.63, η p2 = .03 F(4,253) = 2.36, η p2 = .04 R2 0.212 .635 .665 AIC 1116.49 1434.76 1517.61 Post hoc 11-12 > 8-9 > 4-5 11-12 > 8-9 > 4-5 b, m > eE 11-12 > 8-9 > 4-5 b, m > eE *** p < .001; * *p < .01, * p < .05 Table 4: Outcomes of the linear models for the relation between vocabulary development, intelligence and executive functioning. Figure 3: Group differences in executive functioning development 6.5 Balance In previous research (Goriot et al., 2018a), we showed that in functionally monolingual and early-English pupils, balance scores were positively related to switching performance. In other words, children whose vocabulary in Dutch 75 Developmental differences in executive functioning <?page no="76"?> 3 English and Dutch vocabulary were centred, because of their very high correlation with age group (r = .862, and r = .911, p < .001, for English and Dutch vocabulary, respectively). and English was more equally developed had higher switching scores. In order to investigate if this same relation exists in bilingual children, linear models with the executive functioning measures (separately), as a dependent variable, and age group, as an independent variable, were performed. Thereafter, individual difference scores in lexical balance, intelligence, Dutch 3 and English vocabulary were added as covariates to this basic model, in order to investigate whether individual differences could explain differences in the variance in executive functioning scores. The first variable to add was always lexical balance. If the model fit did not improve (as indicated by a smaller AIC value), balance was removed and intelligence was added. If the model fit did improve, it was checked whether adding intelligence could further improve the model. Finally, it was checked whether balance could be replaced by either Dutch or English vocabulary. The basic model for switching (AIC = 251.46, df = 50, R 2 = .350) revealed a significant effect of age group. Adding balance improved the model fit (AIC = 249.54, df = 49, R 2 = .384). The interaction between balance and age group was checked, but not significant (AIC = 249.86, R 2 = .401, df = 47), and this model was discarded. Adding intelligence to the model that also included balance improved the model fit even more (AIC = 249.03, df = 48, R 2 = .400). Changing balance for English vocabulary improved the model fit (AIC = 248.63, df = 48, R 2 = .404). Changing balance for Dutch vocabulary did not improve the model fit (AIC = 249.87, df = 48, R 2 = .390). Adding both Dutch and English vocabulary to the model instead of balance resulted in the best model fit (AIC = 244.26, df = 47, R 2 = .461). In this model, age group and Dutch vocabulary showed a significant relation with switching scores. English vocabulary showed a marginally significant relation (p = .052), whereas the main effect of intelligence was not significant. Table 4 shows the final model. For verbal working memory, the basic model (AIC = 363.17, df = 59, R 2 = .384) showed a significant main effect of age group. Adding balance to the model slightly improved the model (AIC = 363.04, df = 58, R 2 = .467), but balance was not significant. The interaction between age group and balance was checked but appeared not to be significant. Adding intelligence to the model with the main effect of balance resulted in a model fit that was worse than the model with only balance (AIC = 363.94, df = 57, R 2 = .467), thus the model was discarded. Adding English vocabulary instead of balance did not improve the model (AIC = 365.15, df = 58, R 2 = .449). The same was true for adding Dutch vocabulary instead of 76 Claire Goriot <?page no="77"?> balance (AIC = 363.56, df = 58, R 2 = .463), and for adding both Dutch and English vocabulary (AIC = 365.48, df = 57, R 2 = .454). The model with the best fit thus included age group and balance and is shown in Table 5. Figure 4 provides a visual view of the relation between balance and verbal working memory. For non-verbal working memory, the basic model also showed a significant effect of age group (AIC = 396.61, df = 60, R 2 = .632). Adding balance to the basic model did not improve the model fit (AIC = .398.22, df = 59, R 2 = .628). Adding intelligence to the basic model did not improve the model fit either (AIC = .397.56, df = 59, R 2 = .632). Adding English vocabulary also resulted in a worse model fit (AIC = 396.79, df = 59, R 2 = .636). Adding Dutch vocabulary, however, resulted in an improved model fit (AIC = 395.99, df = 59, R 2 = .641). The model with only Dutch vocabulary next to age group had also a better fit than adding both Dutch and English vocabulary (AIC = 397.19, df = 58, R 2 = .639). The final model is presented in Table 5. Switching Verbal working memory Non-verbal working memory Age group F(2,47) = 18.06 *** , = .43 F(2,58) = 27.24 *** , = .48 F(2,59) = 55.59 *** , = . 65 Balance F(1,58) = 2.02, = .03 - Intelligence F(1,47) = 2.46, = .05 - - Dutch vocabu‐ lary (centered) F(1,47) = 6.88 * = .13 - F(1, 59) = 2.51, = .04 (English vo‐ cabulary) F(1,47) = 3.96( * ), = .05 - R2 0.461 .454 .641 AIC 244.26 365.48 395.99 Tukey post hoc tests 11-12 > 8-9, 4-5 11-12, 8-9 > 4-5 11-12, 8-9 > 4-5 ***p < .001; **p < .01, *p < .05, (*) p < .06 Table 5: Outcomes of the linear models for the relation between vocabulary development and EF. 77 Developmental differences in executive functioning <?page no="78"?> Figure 4: Relation between language balance and verbal working memory in bilingual pupils. 7 Discussion This chapter focused on the role between language development and executive functioning development. Data about language development and executive functioning development in Dutch monolingual, early-English, and Dutch-Eng‐ lish bilingual pupils of three different age groups were presented. It was investigated whether the groups differed in their development of Dutch, English, switching abilities or working memory performance. Finally, the role of lexical balance (operationalised by receptive vocabulary) in relation to performance on executive functioning tasks was investigated in early bilinguals. It was found that bilingual children have greater knowledge of English vocabulary than the two other groups, who did not differ in English knowledge. In general, the three groups did not differ in Dutch vocabulary knowledge. Bilinguals showed more balanced lexicons than early-English and monolingual pupils, whose lexical balance did not differ. Bilingual and monolingual children showed more advanced executive functioning skills than early-English pupils. The development of both of a bilingual’s languages play a role in switching and verbal working memory performance, but not in non-verbal working memory performance. The first focus of this chapter was on (emerging) bilinguals’ language development of receptive vocabulary. It was hypothesised that the early bilin‐ 78 Claire Goriot <?page no="79"?> gual group would outperform the monolingual and early-English group on the English vocabulary task, but that the three groups would not differ in Dutch vocabulary knowledge. Consequently, early bilinguals would have more balanced lexicons. This hypothesis was confirmed. Bilingual children of all ages had greater English vocabulary knowledge than children from the two other groups. The early-English children did not differ from the monolingual group, except for the 11-12-year-olds who had larger English vocabularies than their monolingual peers. Bilingual children had similar Dutch vocabulary scores as monolingual and early-English pupils, except for those in the 4-5-year-old group. In this group, bilingual children had smaller Dutch vocabularies than the other two groups. Bilingual children in the 4-5 and 8-9-year-old age groups had more balanced vocabulary scores than early-English and monolingual pupils. In the 11-12-year-old age group there were no differences between the three groups. The early-English and monolingual group did only differ from each other in the 4-5-year-old group, in which the early-English pupils had more balanced vocabulary scores than the monolingual groups. That the youngest bilingual children had smaller Dutch vocabularies than their peers is a result that was also found in previous studies (Blom et al., 2014; Prior et al., 2016), possibly because these young children have less experience with Dutch than those growing up with two parents who speak Dutch. Despite the cross-sectional nature of this study, it seems that the bilingual children catch up on Dutch vocabulary when they grow older, as found previously (Blom et al., 2014). Regardless of the fact that the oldest bilinguals had greater English vocabulary knowledge than their monolingual and early-English peers, these 11-12-year-old bilinguals did not have significantly more balanced lexicons than their peers. The difference between English and Dutch vocabulary between the different bilinguals was smaller in the older than in the younger age groups. Since children derive part of their knowledge of the second language from similar-sounding translations in their native language (cognates) (Goriot, van Hout, Broersma, Lobo, McQueen & Unsworth, 2018b), it may be the case that, as the children’s knowledge of Dutch grew, their knowledge of English cognates grew as well. Given that there were no group differences in Dutch vocabulary, it is unlikely that one group profited more from cognate knowledge than another. It is thus likely that the monolingual and early-English pupils partly catch up on the knowledge of English, compared to the bilinguals. Probably, the bilingual advantage in English vocabulary was not large enough anymore in the oldest group, and all three groups may have had relatively balanced Dutch and English vocabularies. Therefore, a bilingual difference in language balance may not have emerged. 79 Developmental differences in executive functioning <?page no="80"?> Mirroring the results of a previous study on the early-English and monolin‐ gual group only (Goriot et al., 2018a), Early-English pupils did neither differ in their knowledge of Dutch nor in their knowledge of English (except for the 11-12-year-olds) or their lexical balance (except for the 4-5-year-olds). The English vocabulary advantage that was found in this study for the 11-12-year-old early-English pupils over their monolingual peers ties in with the results of a previous study on pupils of that age (De Graaff, 2015). Unlike previous studies (Goorhuis-Brouwer & de Bot, 2010; Unsworth, Persson, Prins & de Bot, 2015), 4-5-year-old pupils did not differ from monolinguals in their knowledge of English. One reason may be that, despite the fact that schools are supposed to teach English for at least 60 minutes per week, actual in-class time devoted to English may have been less than an hour. This may have been the case especially in the younger grades, where teachers often reported to regularly switch to Dutch during English lessons. It seems that less than one hour of English per week is not enough for lessons to have an effect on pupils’ knowledge of English (Unsworth et al., 2015), which may account for the fact that no significant differences in English vocabulary were found in the younger age groups. The second focus of this chapter was on differences in children’s cognitive skills, in particular executive functioning. The hypothesis was that bilinguals would outperform the two other groups on switching and working memory tasks. Group comparisons showed that this hypothesis was partly true for working memory but did not hold for the switching task. Although the effect was small, on both the verbal and non-verbal working memory measure there was a general advantage of the bilinguals over the early-English pupils (regardless of the age group). The monolinguals, however, did not significantly differ in performance from the bilingual group and also outperformed the early-English group. For switching, there were no significant differences between the lan‐ guage groups. The question is why the bilinguals outperformed the early-English pupils but not the monolingual group. Previous findings on bilingual executive functioning performance have been mixed. For example, in a longitudinal study, Blom et al. (2014), compared Turkish-Dutch bilingual children growing up in lower socio-economic families in the Netherlands, with Dutch monolinguals. They found an advantage for six-year-old Turkish-Dutch bilingual children over Dutch monolingual children on a backwards digit recall task for verbal working memory. Such advantage did not appear at the age of five. On the other hand, Prior et al. (2016) showed that ten-year-old monolingual children reacted faster than unbalanced Russian-Hebrew bilinguals of the same age on both an inhibition and a switching task, whereas balanced Russian-Hebrew bilinguals 80 Claire Goriot <?page no="81"?> did not differ significantly in their reaction times from either the monolingual or unbalanced bilingual group. The authors reasoned that because of large individual variation in the development of executive functioning in school-aged children, group differences in executive functioning may be hard to detect (Prior et al., 2016). This may be the case in our sample, too: even though the three groups were clearly distinct in their experience with English, there was large individual variation within groups. The groups may not have been homogeneous enough for group differences to appear. In addition, previous analyses on only the exact same monolingual and early-English group as reported about in this chapter, showed that there were no differences between the two groups in executive functioning (Goriot et al., 2018a). In these analyses, we made use of linear mixed models to account for the effect of school. In the present study, that was hardly possible given the fact that (almost) all bilingual children attended a different school. When the current analyses were repeated with linear mixed models with the factor ‘school’ set to the same value for all the bilinguals, the results were similar to the ANOVAs except for the effect of bilingual category. Despite the fact that the effect of school was very small (always <5 %), it may be that the monolingual advantage over the early-English group actually reflects school effects. Third, the relation between language balance and executive functioning development in bilingual children was investigated. The hypothesis was that there would be a positive relation between lexical development in terms of language balance and executive functioning performance. This hypothesis was partly confirmed. There was a positive relation between lexical balance and verbal working memory. This relation was previously found in six-year-old Turkish-Dutch bilingual children (Blom et al., 2014). As such relation was not found in research with the monolingual and early-English group only (Goriot et al., 2018a) but seems to appear in bilingual groups who grow up with two languages, it may be that a relation between lexical balance and verbal memory only exists in more balanced bilingual children. For non-verbal working memory and switching, no relation with language balance was found in this study. The absence of a relation between lexical balance and non-verbal working memory, again, mirror the findings of Blom et al. (2014), who also failed to find such a relation in their sample of Turkish-Dutch bilingual children from minority backgrounds. The authors did not elaborate on why this would be the case. Previously, a relation between language balance and switching performance was found for the monolingual and early-English group (Goriot et al., 2018a), but not in research with balanced Welsh-English bilingual children (Gathercole et al., 2014). It may thus be that, contrary to verbal working memory, a relation 81 Developmental differences in executive functioning <?page no="82"?> between language balance and switching abilities exists only in emerging L2 learners. Although it is merely speculation, it might be that emerging L2 learners are less experienced in monitoring control over two languages and have to work harder to switch from one language to another. This may in turn influence their behavioural switching abilities. Such advantages might disappear when switching becomes more automatised. For both switching and non-verbal working memory, Dutch vocabulary was significantly related to bilingual children’s scores on the switching and non-verbal working memory task. As the instructions for the tasks were given in Dutch, children’s understanding of Dutch may have played a role in their performance on the task. Especially in the case of non-verbal working memory, the youngest children in all three language groups sometimes had difficulty understanding the instructions despite the attempt to formulate them as clear and as straightforward as possible. The AWMA was a computerised task that came with brief instructions. Additional instructions were added after a pilot study revealed that the original instructions were too brief for children to understand the task. Nevertheless, these instructions may still not have been clear enough. Given that the bilingual children in the youngest age group had somewhat smaller vocabularies in Dutch than their peers who were raised with Dutch as the only language, this may have played a role in their understanding of the task instructions. As all instructions were given in Dutch to keep the testing procedure the same for all children, it is unknown whether the bilingual children would have understood the task better if instructions were given in English or what role English vocabulary plays. The gap in language balance scores of the monolinguals and early-English pupils on the one side and that of the early bilinguals on the other side made that the relation between language balance and executive functioning was analysed in the bilingual group only, and the early-English and monolingual pupils’ scores were not included. Since the relation between balance and executive functioning had already been analysed in the monolingual and early-English group (Goriot et al., 2018a), the outcomes of the analyses in the bilingual group could be compared to these previous results. Nevertheless, future research should include an additional group of intermediate proficient bilingual children who are more proficient than the early-English pupils but less proficient than the bilingual children in this study. This would allow for the treatment of bilingualism as a continuous variable, ranging from (almost) monolingual to (almost) perfectly balanced bilingual. The results of this study show that both early bilinguals’ language and executive functioning development is different from that of children who are 82 Claire Goriot <?page no="83"?> raised with only one language as their mother tongue and who have only little exposure to another language. In addition, language development and executive functioning development seem to be related to each other: individual differences in language development are positively related to individual differences in executive functioning development. However, the cross-sectional design of this study prohibits the formulation of causal relations between the two. Since executive functions are involved in learning processes (Diamond, 2013), it may as well be that children with well-developed executive functions are better at learning languages. The mixed findings of this study mirror those of previous research, in which sometimes bilinguals outperform monolinguals or less balanced bilingual peers on executive functioning tasks (e.g. Blom et al., 2014; Poarch & van Hell, 2012) and sometimes they do not (Gathercole et al., 2014; Prior et al., 2016). For more firm conclusions to be drawn, more research is needed. This study shows once again how difficult it is to operationalise bilingualism. It is especially difficult to characterise bilingualism given the complex context of environmental and individual differences that influence the characteristic of being bilingual. At the same time, this study shows that both a minimal amount of language exposure (i.e. only one hour of English lessons per week at school) and a more extensive form of language exposure (i.e. growing up bilingually) enhance pupils’ English vocabulary knowledge while having no negative effects for Dutch vocabulary development. References Adesope, O. O., Lavin, T., Thompson, T. & Ungerleider, C. 2010. A systematic review and meta-analysis of the cognitive correlates of bilingualism. Review of Educational Research, 80 (2), 207-245. Alloway, T.P., Gathercole, S.E., Kirkwood, H. & Elliott, J. 2008. Evaluating the validity of the automated working memory assessment. Educational Psychology, 28 (7), 725-734. Barac, R., Moreno, S. & Bialystok, E. 2016. Behavioral and electrophysiological differences in executive control between monolingual and bilingual children. Child Development, 87 (4), 1277-1290. Blom, E., Küntay, A. C., Messer, M., Verhagen, J. & Leseman, P. 2014. The benefits of being bilingual: Working memory in bilingual Turkish-Dutch children. Journal of Experimental Child Psychology, 128, 105-119. Bosma, E., Hoekstra, E., Versloot, A. & Blom, E. 2017. The minimal and short-lived effects of minority language exposure on the executive functions of Frisian-Dutch bilingual children. Frontiers in Psychology, doi: 10.3389/ fpsyg.2017.01453. 83 Developmental differences in executive functioning <?page no="84"?> Carlson, S. M. & Meltzoff, A.N. 2008. Bilingual experience and executive functioning in young children. Developmental Science, 11 (2), 282-298. Cohen, J. 1988. Statistical Power Analysis for the Behavioral Sciences. 2 nd ed. Hillsdale, NJ: Lawrence Erlbaum. De Graaff, R. 2015. Vroeg of laat Engels in het basisonderwijs: Wat levert het op? Levende Talen Tijdschrift, 16 (2), 3-15. Diamond, A. 2013. Executive functions. Annual Review of Psychology, 64, 135-168. Dunn, L.M. & Dunn, D.M. 2007. PPVT-4 Manual. Bloomington, MN: NCS Pearson. Dunn, L.M., Dunn, D.M. & Schlichting, L. 2005. Peabody Picture Vocabulary Test-III-NL. 2 nd ed. Amsterdam: Pearson Assessment and Information. Enever, J. (Ed.). 2011. ELLiE: Early Language Learning in Europe. London: British Council. Enever, J. 2013. Contemporary Societal Perspectives on English in Europe. Chile: British Council. Gathercole, V.C.M., Thomas, E.M., Kennedy, I., Prys, C., Young, N., Guasch, N.V., Roberts, E.J., Hughes, E.K. & Jones, L. 2014. Does language dominance affect cognitive performance in bilinguals? Lifespan evidence from preschoolers through older adults on card sorting, Simon, and metalinguistic tasks. Frontiers in Psychology, 5, doi: 10.3389/ fpsyg.2014.00011. Goorhuis-Brouwer, S. & de Bot, K. 2010. Impact of early English language teaching on L1 and L2 development in children in Dutch schools. International Journal of Bilingualism, 14 (3), 289-302. Goriot, C., Broersma, M., McQueen, J.M., Unsworth, S. & van Hout, R. 2018a. Language balance and switching ability in children acquiring English as a second language. Journal of Experimental Child Psychology, 173, 168-186. Goriot, C., van Hout, R., Broersma, M., Lobo, V., McQueen, J.M. & Unsworth, S. 2018b. Using the Peabody Picture Vocabulary Test in L2 children and adolescents: Effects of L1. International Journal of Bilingual Education and Bilingualism. Advance Online Publication, doi: 10.1080/ 13670050.2018.1494131. Green, D.W. 1998. Mental control of the bilingual lexico-semantic system. Bilingualism: Language and Cognition, 1 (2), 67-81. Green, D.W. & Abutalebi, J. 2013. Language control in bilinguals: The adaptive control hypothesis. Journal of Cognitive Psychology, 25 (5), 515-530. Kuppens, A.H. 2010. Incidental foreign language acquisition from media exposure. Learning, Media and Technology, 35 (1), 65-85. Miyake, A., Friedman, N.P., Emerson, M.J., Witzki, A.H., Howerter, A. & Wager, T.D. 2000. The unity and diversity of executive functions and their contributions to complex “Frontal Lobe” tasks: A latent variable analysis. Cognitive Psychology, 41 (1), 49-100. Nuffic. 2018. Wat is vroeg vreemdetalenonderwijs? www.nuffic.nl/ onderwerpen/ vroeg -vreemdetalenonderwijs/ vroeg-vreemdetalenonderwijs-vvto. May 31, 2021. 84 Claire Goriot <?page no="85"?> Paap, K.R. & Greenberg, Z.I. 2013. There is no coherent evidence for a bilingual advantage in executive processing. Cognitive Psychology, 66 (2), 232-258. Paap, K.R., Myuz, H.A., Anders, R.T., Bockelman, M.F., Mikulinsky, R. & Sawi, O.M. 2017. No compelling evidence for a bilingual advantage in switching or that frequent language switching reduces switch cost. Journal of Cognitive Psychology, 29 (2), 89-112. Poarch, G.J. & van Hell, J.G. 2012. Executive functions and inhibitory control in multi‐ lingual children: Evidence from second-language learners, bilinguals, and trilinguals. Journal of Experimental Child Psychology, 113 (4), 535-551. Prior, A., Goldwasser, N., Ravet-Hirsh, R. & Schwartz, M. 2016. Executive functions in bilingual children. Is there a role for language balance? In: Schwieter, J. W. (Ed.), Cognitive Control and Consequences of Multilingualism. Amsterdam: John Benjamins. 323-350. Purić, D., Vuksanović, J. & Chondrogianni, V. 2017. Cognitive advantages of immersion education after one year: Effects of amount of exposure. Journal of Experimental Child Psychology, 159, 296-309. Simon, J.R. & Small, A.M. 1967. Processing auditory information: Interference. Journal of Applied Psychology, 53 (5), 433-435. Tabachnick, B.G. & Fidell, L.S. 2007. Using Multivariate Statistics. 5 th ed. New York, NY: Allyn & Bacon/ Pearson Education. Thomas-Sunesson, D., Hakuta, K. & Bialystok, E. 2018. Degree of bilingualism modifies executive control in Hispanic children in the USA. International Journal of Bilingual Education and Bilingualism, 21 (2), 197-206. Unsworth, S., Persson, L., Prins, T. & de Bot, K. 2015. An investigation of factors affecting early foreign language learning in the Netherlands. Applied Linguistics, 36 (5), 527-548. Wechsler, D. & Naglieri, J.A. 2008. WNV NL. Wechsler Nonverbal Scale of Ability. Nederlandstalige Bewerking. Amsterdam: Pearson. Zelazo, P.D. 2006. The Dimensional Change Card Sort (DCCS): A method of assessing executive function in children. Nature Protocols, 1 (1), 297-301. 85 Developmental differences in executive functioning <?page no="87"?> Part 2: Linguistic and cognitive abilities in primary school children <?page no="89"?> 1 Although this is a joint paper, author 1 takes responsibility for sections 1, 3, 4, while author 2 takes responsibility for sections 2, 5, 6 (this declaration is important to satisfy the requirements for the Italian Ministry). The link between multilingualism and attention in children with and without a migrant background - Practical implications 1 Gerda Videsott / Rita Franceschini Abstract The aim of this study in the field of multilingual neuroscience is to examine the relationship between multilingualism and attentional mechanisms. We compared the attention mechanisms in children with and without a migrant background. The students attended a primary school in South Tyrol in year 1. The Attentional Network Test (ANT), a standardised test for measuring attentional mechanisms, was adopted for this study, which measures alerting, orienting and executive control (see Eriksen & Eriksen, 1974). The results showed a clear difference between the two groups. There was a significant difference in the ‘speed-accuracy trade-off ’, i. e. the relationship between reaction times and the accuracy of the answers: children without a migrant background achieved a higher number of correct answers with a slower response rate, whereas children with a migrant background responded more quickly, yet less accurately. Both types of reactions have advantages and disadvantages, and the implications of the test results have implications for practical language teaching. 1 Introduction In discussions about migrant children’s achievements in schools, the perspective is often a deficit-oriented one, even though teacher training programmes nowadays place more emphasis on negative attitudes towards migration, dis‐ mantling such attitudes as stereotypes and clichés. However, additional facts <?page no="90"?> and evidence-based research on migrant children’s achievements are needed: On the one hand this may help to dismantle the so-called ‘monolingual habitus’ (see Gogolin, 1994), which may still be found in schools. On the other hand, research studies may help teachers to become more informed about advantages and disadvantages that children face when they grow up with two or more languages. These multilingual children are often described as ‘children with a migration background’. Any research and teaching with such a group of subjects should take into consideration the circumstances of bilingual experience, the social conditions of the family, the attitudes of the surrounding society with respect to migration, and other social factors as well because neither praising only the positive aspects of bilingual experience nor negative attitudes towards migration and multilingualism will be beneficial in the long run. Therefore, we started a project at the Free University of Bozen-Bolzano (cf. Videsott, Della Rosa & Franceschini, 2015) with the aim to compare attention mechanisms in monolingual and mul‐ tilingual children. A standardised test for measuring attentional mechanisms, the Attentional Network Test (ANT), was adopted for the study (sometimes also labelled ‘Flanker-test’, see Eriksen & Eriksen, 1974). The following questions will be addressed in this study: a) What are methodological and practical difficulties in operationalising multilingualism and multilingual competence? b) Is there a relationship between multilingualism and attentional mecha‐ nisms? c) What are the didactic implications regarding such a relationship? First, we define ‘multilingualism’ and ‘multilingual competence’, and subse‐ quently, we provide an overview of studies examining the relationship between multilingualism and attentional mechanisms. We then describe the subjects, the test instruments and procedures, followed by the results, which particularly focus on correlational analyses. The paper concludes with a discussion taking into consideration some practical implications regarding the relationship be‐ tween multilingualism and attentional mechanisms in the language classroom. 2 Definition of multilingualism and multilingual competence When examining multilingualism, a definition usually takes into account dif‐ ferent dimensions, which include, for example, the subjects’ language biography (L1-L2-L3, etc., parallel or successive acquisition), the family context (monolin‐ gual, bilingual, and whether the family language/ s correspond/ s to the ambient 90 Gerda Videsott / Rita Franceschini <?page no="91"?> language/ s spoken in the social context), the socio-economic context and the cultural prestige associated with the languages of a multilingual individual (i.e. the multilingual language repertoire, Franceschini, 2016). A multilingual speaker’s competence in each of the languages is yet another dimension - it may be more or less balanced or functionally restricted to certain domains (literature, work, tourism, etc.). Furthermore, language dominance - i. e. a better proficiency in one of the languages - is an important dimension that has to be regarded in a dynamic way, i. e. as an individual’s repertoire. The dominance of a language may, however, change during an individual’s lifetime. This is particularly true in migration contexts, in which L1 attrition may occur. Another change often takes place in early or later childhood, mainly when children with a migration background start to enter kindergarten or school, as it is often the case for second or third generations of migrants. Multilingualism may not only refer to an individual but also to other levels, such as multilingual societies, institutions or groups. In 2009, Franceschini defined the term multilingualism as follows: “The term/ concept of multilingualism is to be understood as the capacity of societies, institutions, groups and individuals to engage on a regular basis in space and time with more than two languages in everyday life. Multilingualism is a product of the fundamental human ability to communicate in a number of languages. Operational distinctions may then be drawn between social, institutional, discursive and individual multilingualism. The term multilingualism is used to designate a phenomenon em‐ bedded in the cultural habits of a specific group, which are characterised by significant interand intra-cultural sensitivity.” (Franceschini, 2009: 34) Thus, a description of an individual’s multilingual competence should take into account the functionally different usage of languages and the flexible use of languages, for example, in the same conversation (changing the language depending on the respective interlocutor, the topic discussed or other reasons, i. e. code-switching). These multilingual practices, and the capacity of transla‐ tion and mediation, characterise only bilingual and multilingual people who are constantly confronted with the demands of a change in language in their linguistic environment. In recent years, the question has been raised whether these special multilingual requirements also have repercussions on other levels of cognition. In the last decade, especially the relationship between multilingual experience and cognitive functions, and, in particular executive functions, has been examined in a large number of studies (for an overview see Bialystok, 2017; Franceschini, 2017). 91 The link between multilingualism multilingualism and attention in children <?page no="92"?> 3 The relationship between multilingualism and attention The relationship between multilingualism and cognitive functions has only recently received attention in studies on multilingual neuroscience. The field of ‘cognitive neuroscience’ first emerged at the end of the 1970s, following the need to outline combined findings from psychology and the neurosciences (see also Gazzaniga, Ivry & Mangun, 2002). Back then, the challenge was to understand the relationship between mental and cognitive processes and their biological foundations (the neural pathways of the brain), and to investigate processes that are necessary to generate specific functions in the brain and the mind. The advance of neuroimaging techniques was fundamental for the development of this discipline. Techniques such as MRI (Magnetic Resonance Imaging) enabled the observation of a healthy brain at work. In the specific case of multilingual neuroscience, researchers wanted to investigate how and where different languages are located and processed in the brain. Thus, the brain was pictured as having a system similar to a desk drawer, with different languages being assigned to different areas of the brain. Such assumptions were, for example, based on observations of aphasic patients, who retrieved the last language they had used first (Papathanasiou & De Bleser, 2003; for more detailed information cf. Paradis, 2004). In this context, Videsott, Herrnberger, Hoenig, Schilly, Grothe, Wiater, Spitzer & Kiefer (2010) examined a quadrilingual group of participants from the Ladin valleys (South Tyrol), using neuroimaging. Their results indicated that the use of English, which was the language that their subjects had acquired last and had least proficiency in, led to increased activities in the region of the cerebellum, which plays an important role in articulation (e.g. Silveri & Misciagna, 2000). It seems that the participants required more effort when pronouncing English words as compared to their other languages, in particular their L1s Italian, Ladin, and German, respectively. As it became clear that different languages are indeed not simply located in specific parts of the brain, researchers started to investigate the cognitive processes and neural pathways that are activated when using language or, in the case of bilingual and multilingual subjects, when using two or more languages. For example, monolinguals only have access to one lexical level when they want to express a semantic concept (e.g. Italian “cane”), whereas bilingual or multilingual persons have many lexical levels they can access (e.g. Italian “cane”, German “Hund” and English “dog”). Such a decision may vary depending on the situation, the context and the interlocutors (cf. Costa & Sebastián-Gallés, 2014). At a neurocognitive level, the major difference between the linguistic processes of a monolingual speaker and that of a bilingual or multilingual speaker lies in 92 Gerda Videsott / Rita Franceschini <?page no="93"?> the more effective use of control and inhibition mechanisms. Thus, the currently activated language at any given point in time inhibits the other language(s) from emerging (see Abutalebi & Green, 2007; 2008). Such decision processes also have implications on the neurocognitive levels, particularly those that involve attentional mechanisms, and, following Posner & Peterson (1990) our focus was on the following three major components: a) Alerting: achieving and maintaining a state of alertness; b) Orienting: selecting information from sensory input; c) Conflict: monitoring and resolving conflicts (executive control). During linguistic processing, control mechanisms constantly adapt to the recurring demands of specific interactions. The ‘Adaptive Control Hypothesis’ (Green & Abutalebi, 2013) describes contexts which occur in a monolingual, bilingual, or switched mode. For example, in a context where both interlocutors speak two or more languages, one language dominates, and it is then the only language in use. Here we observe a case of early selection, because the language is established from the beginning, and the interlocutors are speaking in one specific linguistic code. However, the interlocutors may also switch (spontaneously) between languages. In this case, we observe a case of late selection because this kind of conversation allows interlocutors to switch to another language or to realise in which language the other one is talking even at the very last moment. For the switched mode, Green (2011) has shown that late selection, in contrast to early selection, activates the nucleus caudatus in the basal ganglia, which appears particularly sensitive to language switching because it is not observed when the context does not require switching. Of special interest in this study are the interactions between multilingualism and attentional mechanisms. Because some studies have not found any differences between monolingual and bilingual subjects (cf. for example Duñabeitia, Her‐ nández, Antón, Macizo, Estévez, Fuentes & Carreiras, 2014; Gathercole, Thomas, Kennedy, Prys, Young, Viñas Guasch, Roberts, Hughes & Jones, 2014; Paap & Greenberg, 2013), it seems reasonable to speak of multilingualism as a continuum (instead of categorically differentiating monolinguals and bilinguals). In such a continuum, an individual speaks two or more languages and may or may not differ in terms of proficiency and amount of use (Franceschini, 2016; Poarch & van Hell, 2012). In this respect, it is necessary to carefully monitor the complex language situation of the participants and to analyse and define different variables (such as proficiency, age of acquisition, language use, etc.) in more detail. Such an approach has also been used in our study with primary schoolers in year 1, who were either bilingual (without migration background, speaking German or Italian at home) or 93 The link between multilingualism multilingualism and attention in children <?page no="94"?> multilingual (with a migration background, whose family language was neither German nor Italian) in a language setting where both Italian and German are used as ambient languages. 4 Method 4.1 Subjects The present study is part of a greater research project with the aim of advising teachers on how to promote their pupils’ linguistic skills (see also Videsott et al., 2015). The Language Study Unit of the Free University was commissioned to carry out this project. The current study took place in a school in the Bolzano district in South Tyrol, a multilingual region in northern Italy. In this school, German is the official language, and the majority of the children are from German-speaking families. Italian is learnt as an L2 from grade 1 onwards; English is introduced in grade 4. 57 children (28 girls and 29 boys) participated in the study, 17 of them had a migrant background, corresponding to 29 % (see section 4.2.2 for additional information). The children at age six were in grade 1 and attended one of four parallel classes at the same primary school (cf. Videsott et al., 2015). 4.2 Test instruments In this study, we used the Attentional Network Test as well as a questionnaire, which are described in sections 4.2.1 and 4.2.2 (see also Videsott et al., 2015). 4.2.1 The Attentional Network Test The Attentional Network Test (ANT; first version by Eriksen & Eriksen, 1974) is a standardised procedure to test three attentional networks in children and adults, namely alerting, orienting, and executive control/ conflict (cf. Posner & Peterson, 1990). In short, alerting is examined by changes in reaction time (RT) resulting from a warning signal. Orienting is determined by changes in the reaction time that accompany cues indicating where the target will occur, and finally, executive control is examined by requiring the participant to respond by pressing two keys indicating the direction (left or right) of a central arrow surrounded by congruent, incongruent or neutral flankers. The test runs on a computer for approximately seven minutes. Test instruc‐ tions were first given to all children in the classroom at the same time, they were then asked to perform the test individually on a laptop in a separate room. The ANT software recorded the reaction times (RT) as well as the accuracy of the answers. 94 Gerda Videsott / Rita Franceschini <?page no="95"?> The ANT displays a combination of possible arrow directions and positions which can lead to twelve answers. The participants’ task is to click the right or left mouse button as quickly as possible, based on the direction of the middle arrow out of a series of five. The arrows are presented either above or below a fixation cross. The arrows can point left or right (which explains the alternative name ‘Flanker Task’), as shown below: a) neutral: −−→−− or −−←−− b) congruent: →→→→→ or ←←←←← c) incongruent: →→←→→ or ←←→←← Furthermore, there are four options for displaying the test items, i. e. either a single (central or spatial), double, or no cue are presented on the screen before the arrows appear (the cross is the stable fixation point): Figure 1: Attentional Network Test (ANT) or “Flanker Task”. As shown in Figure 1, these images have different positions of the fixation cross (+) and the symbol (*), which acts as a cue (as a sort of alarm signal) indicating where the arrows will appear next (see Table 1): no cue only fixation cross (no alarm signal) central cue alarm signal on the middle of the screen double cue alarm signals above and below the fixa‐ tion cross spatial cue alarm signal either above or below the fixation cross Table 1: Example of a test sequence for the ANT or “Flanker Task”. 95 The link between multilingualism multilingualism and attention in children <?page no="96"?> The following figure illustrates an example of a test sequence: the fixation cross is presented for 400 milliseconds, followed by the upper spatial cue for 10 milliseconds; the fixation cross appears again for 400 milliseconds and, ultimately, the test items are displayed for 1700 milliseconds (in this case, the incongruent sequence). At this point in time, participants should react with a mouse click corresponding to the direction of the middle arrow (in this case: right). * Figure 2: Example of a test sequence of the ANT as shown on the screen, picture after picture. As expected, participants responded quickest when the alarm signal was displayed in the centre of the computer screen, followed by the neutral sequence (four simple lines and the arrow placed in the middle). The cue in the centre of the screen leads participants to shift their gaze to the direction of the successive arrow and the lines of the neutral sequences. This makes it easier for the participants to identify the direction (left or right) of the arrow. The most difficult sequence proved to be the incongruent one when the presentation of the arrows is not preceded by any cue. The test is designed in such a way that it allows to measure attentional mechanisms by comparing the following conditions: 96 Gerda Videsott / Rita Franceschini <?page no="97"?> • ALERTING EFFECT = Trials No Cue vs. Double Cue; • ORIENTING EFFECT = Trials Central Cue vs. Spatial Cue; • CONFLICT EFFECT = Trials Incongruent vs. Congruent. 4.2.2 The questionnaire After taking the ANT, the children were asked to fill out a short questionnaire, in order to obtain information on their linguistic background. Among other things, the children were asked about the language/ s being spoken at home and their school grades. Based on the questionnaire, language dominance and use in the respective families could be assessed. To differentiate between children with and without a migrant background, a composite value was extracted from the following information: the child’s first and dominant language, the first language of his/ her parents, the effective use of the different languages at home and in leisure time, and the language dominance in different contexts. Based on these individual language profiles, the children were divided into three groups: a) children without a migrant background whose first language is one of the two local languages of the territory (i.e. German); b) children without a migrant background whose first language is one of the two local languages of the territory (i.e. Italian); c) children with a migrant background whose first language is not one of the two local languages of the territory (i.e. neither Italian nor German). School grades were also used to correlate the values of the ANT with the following parameters: average of all school grades, average of school grades in the two language subjects (German and Italian), individual grades in Italian and grades in German, respectively. 5 Results of the study The statistical analyses include ANOVAs, correlations, and regression analyses. Additional information on the data is provided by Videsott et al. (2015). 5.1 General results By means of a 4x3 repeated measures ANOVA model, with the factor ‘cue’ (4 conditions, school grades) and ‘congruency’ (i.e. 3 conditions, 3 subject groups with L1 Italian, L1 German or L1 other), the data relating to the performance of the ANT were analysed in terms of both reaction time and accuracy. Subsequently, in order to assess potential differences between the three groups 97 The link between multilingualism multilingualism and attention in children <?page no="98"?> for the three attentional networks, a repeated-measures ANOVA was conducted on the three attentional effects calculated from the individual ANT conditions. The results of the present study show a clear difference between children with and without a migrant background with respect to the Attentional Network Test. This difference is evident in the ‘speed-accuracy trade-off ’: In general, children with a migrant background responded faster (albeit with more errors) than children without a migrant background who performed slower, but at the same time more correctly. With regard to reaction times, a significant main effect of group was found (F(2,54) = 4.310; p = 0.018): children with a migrant background responded faster than both children with L1 German (141 ms; p = 0.014) and children with L1 Italian (70 ms; p = 0.414), although not significantly faster than the latter. Migrant children’s responses were closer to the children with L1 Italian, who were also in a bilingual situation in this type of officially German-speaking school and whose situation was, thus, similar to that of the children with a migrant background. Interactions between the group factor and the cue condition (p = 0.445) or the congruency condition (p = 0.145) were not significant. In terms of overall accuracy, the accuracy rate for the test was 81 % (mean 6.507/ 8 per condition). No main effect of the group factor was found (p = 0.116), but there was a significant interaction between the group factor and the congruence factor (F(4,108) = 2.986; p = 0.042). Thus, children with a migrant background were less accurate than children with German L1 (mean: 0.518/ 8) and, with a significant difference, than children with Italian L1 (mean: 1.518/ 8; p = 0.069) in the incongruence condition. In terms of ANT effects there was no significant difference between the groups neither with respect to alertness (p = 0.692), orientation (p = 0.239) and control (p = 0.597), nor between children from a migrant background and the non-migrant groups (with German L1 and Italian L1). 5.2 Correlation analyses The correlations between the measures of the attentional effects of alertness, orientation and conflict, calculated from the individual conditions, and the school grades were analysed, distinguishing between school grades, i. e. average of all school grades, average of school grades in the two linguistic subjects (German and Italian), grades in Italian and grades in German. The three groups with L1 German, L1 Italian or L1 others (migrants) are considered separately, with the results of these three groups (all groups) being pooled at the top. Table 2 presents the results. 98 Gerda Videsott / Rita Franceschini <?page no="99"?> Table 2: ANOVAs, correlations and regression analyses. As Table 2 shows, correlational analyses based on school grades did not show any significant correlation between any of the three effects (alertness, orientation and conflict). In addition, insignificant results were noted for school grades (i.e. mean of all grades, mean of grades in the two language subjects German and Italian, Italian grades and German grades), irrespective of the children’s language background. Finally, no significant correlation was found 99 The link between multilingualism multilingualism and attention in children <?page no="100"?> between the three effects and the above-mentioned parameters even when dividing the correlations for each group according to the context of origin and the different L1 of the three groups. 5.3 Regression analyses To verify whether the children’s language background and/ or their school grades may explain the variability observed for the three effects of the ANT, three hierarchical regression analyses were conducted, with alertness, orienta‐ tion and conflict as dependent variables. In each regression analysis, the average of all school grades was included as the first predictor, the second was the average variable of school grades in the two language subjects (German and Italian) and the third the variable L1 background (i.e. German, Italian or other). The results of the hierarchical regression analysis divided by the three effects of ANT showed that the three variables did not predict any variance in the alert condition. As far as orientation is concerned, neither the average of all the school grades nor the average of the school grades in the two language subjects predict a significant portion of variability. However, when the type of context is added, the result was almost significant (R2 = 0.057; F(1,53) = 3.202; t= -1.79; p = 0.079). In other words, a reduction of the orientation effect may be observed in the group of children with a migrant background compared to children without a migrant background, although the regression model is not significant (F = 1.180; p = 0.329). Finally, the three variables do not predict any variance with regard to the effect of conflict. 6 Interpretations and implications The focus of this chapter was on the relationship between multilingualism and attentional mechanisms. To this end, we examined three groups of children, namely children with L1 German, children with L1 Italian, and children with a migrant background who do not speak either German or Italian at home. All subjects attended the first year in a German-speaking primary school in South Tyrol, where German was the dominant language and Italian-as-a-subject being introduced in year 1. The children were tested individually with the At‐ tentional Network Test (ANT), which measures alerting, orienting, and executive control/ conflict. The children’s school grades, in particular their grade in the subjects Italian and German, were also included in the analyses. In terms of reaction time, the results of this study show significant differences between children with and without a migrant background regarding their performance in the test. Independent of the condition (i.e. alerting, orientating, 100 Gerda Videsott / Rita Franceschini <?page no="101"?> conflict), children with a migrant background responded faster than children without a migrant background who performed slower. This result is interesting because the ANT is not a linguistic but a cognitive task, and language back‐ ground should not matter. Nevertheless, the test is evidently sensitive to language experience. It seems as if children with a migration background are more accustomed to reacting rapidly to changing attention. In the language situation they live in, it seems to be natural that they constantly change from one language to the others, and they may, thus, be more trained when it comes to quick decisions. This seems to be a clear advantage and an outcome of everyday experience with the linguistically rich environment in South Tyrol. In this respect, children from a migrant background can be seen as outperformers (see also Videsott et al., 2015). But when we consider the accuracy with which a task is performed, rapidity appears to influence accuracy in a negative way: although faster, the children with a migration background are not as accurate in their responses as the other two groups. Thus, speed and accuracy compete with each other - this effect is called ‘speed-accuracy trade-off ’. These results are particularly prominent in the condition of executive control / conflict where children with a migration background do not seem to inhibit a wrong answer as the children with L1 German or L1 Italian do. In those cases, they behave as if they would prefer speed over accuracy. In children growing up with German and Italian, in contrast, the attentional system seems to be better able to regulate the challenge of any speed-accuracy trade-off. In other words, for them the overlap between the linguistic reference context and the languages of the school context could better regulate the attentional executive system in terms of speed-accuracy trade-off, especially in the condition of greater interference, where it is necessary to suppress information from the distractors for the benefit of accurate response (see also Videsott et al., 2015). This study is not without limitations: The first one relates to the small sample size. Moreover, the languages that the children with a migrant background speak at home, and, more importantly, their proficiency level in their family language/ s should be studied in more detail. Furthermore, this study did not consider possible interrelationships between the components of the ANT (i.e. alerting, orienting and executive control) in more detail. This is why some researchers suggest supplementing any behavioural measures with event re‐ lated potentials (e.g. Galvao-Carmona, González-Rosa, Hidalgo-Muñoz, Páramo, Benítez, Izquierdo & Vázquez-Marrufo, 2014). 101 The link between multilingualism multilingualism and attention in children <?page no="102"?> 7 Conclusion Although this study included a relatively small sample, the results nevertheless support the assumption that multilingualism, understood as the active use of a language or dialect, has an impact on brain structures and on neurocognitive processes and, in particular, on the effectiveness of use of attentional mecha‐ nisms, which can be found as early as childhood. This research specifically demonstrated differences between the children with and without a migrant background regarding speed-accuracy trade-offs, i. e. the compromise between speed of reaction and accuracy of response. In general, children without a migrant background, who speak at home one of the two languages of the school and socio-territorial context (i.e. Italian or German), respond more accurately but slower than their peers with a migration background (whose L1 was neither Italian nor German), who respond faster but less accurately. Apparently, the experience with multilingual surroundings shapes the brain and influences the neurocognitive processes, and, in particular - as we have shown - the mechanism of attention. These results may also have practical implications for the language class‐ room: If the teachers want more accurate responses, they may want to down‐ grade the time component in order to rebalance the ‘speed-accuracy trade-off ’. However, pupils also need to reflect on the fact that fast answers alone do not necessarily lead to the desired aim of providing correct answers. Different didactic interventions may be carried out that, according to different needs, either focus more on the speed or on the accuracy of the response: It may be helpful in this context to adopt “Montessorian” materials, which often require only one possible way to solve a task. In addition, the use of an hourglass may help children to visualise the amount of time they have available to conclude a task. Starting with these more practical exercises, even young learners will improve regarding the accuracy of their responses. It seems that the attentional system needs to be tuned to the different requirements of the various multilingual situations a child is confronted with, which sometimes include a code-switch and sometimes not. Apparently, the differences between the ambient languages at school and the family language constitute a challenge for the attentional system. Perhaps being in grade 1 in primary school is particularly demanding for multilingual children, because they are also confronted with a new learning environment, new teachers and peers, new task demands in ever-changing linguistic contexts which do not correspond to their family language. 102 Gerda Videsott / Rita Franceschini <?page no="103"?> In addition to examining a larger number of children, a longitudinal study could examine the subjects of this study over many years in order to determine whether the attentional system is better able to coordinate any speed-accuracy trade-off as a function of age and experience. We assume that the children in grade 1 are at the beginning of a development which, in the end, will lead to a more effective functioning of the attentional mechanisms. In fact, in another study in South Tyrol with bilingual 9-year-old children (who are 3 years older than those in this study), the attentional system seems to be more adapted. At this age, the children showed a density of grey matter in specific parts of the brain that parallels that of adults (see Della Rosa, Videsott, Borsa, Canini, Weekes, Franceschini & Abutalebi, 2012). Based on these results, we believe that the children in grade 1 of primary school are at an early stage of aligning their attentional mechanisms with the languages they are confronted with, but that their attentional system will adapt to the affordances in the next years. Teachers can rest assured that the children can master the challenging situation of multilingualism. It does help them, though, to involve them in tasks that train both accuracy and speed at the same time. References Abutalebi, J. & Green, D.W. 2007. Bilingual language production: The neurocognition of language representation and control. Journal of Neurolinguistics, 20 (3), 242-275. Abutalebi, J. & Green, D.W. 2008. Control mechanisms in bilingual language production: Neural evidence from language switching studies. Language and Cognitive Processes, 23, 557-582. Bialystok, E. 2017. The bilingual adaptation: How minds accommodate experience. Psychological Bulletin, 143 (3), 233-262. Costa, A. & Sebastián-Gallés, N. 2014. How does the bilingual experience sculpt the brain? Nature Reviews Neuroscience, 15 (5), 336-345. Della Rosa, P. A., Videsott, G., Borsa, V., Canini, M., Weekes, B.S., Franceschini, R. & Abutalebi, J. 2012. A neuronal interactive location for multilingual talent. Cortex, 49, 605-608. Duñabeitia, J.A., Hernández, J.A., Antón, E., Macizo, P., Estévez, A., Fuentes, L.J. & Carreiras, M. 2014. The inhibitory advantage in bilingual children revisited: myth or reality? Experimental Psychology, 61 (3), 234-251. Eriksen, B.A. & Eriksen, C.W. 1974. Effects of noise letters upon the identification of a target letter in a nonsearch task. Perception & Psychophysics, 16, 143-149. 103 The link between multilingualism multilingualism and attention in children <?page no="104"?> Franceschini, R. 2009. The genesis and development of research in multilingualism: perspectives for future research”. In: Aronin, L. & Hufeisen, B. (Eds.), The Exploration of Multilingualism. Development of Research on L3, Multilingualism and Multiple Language Acquisition. Amsterdam: Benjamins. 27-61. Franceschini, R. 2016. Multilingualism and multicompetence. In: Cook, V. & Wei, L. (Eds.), The Cambridge Handbook of Linguistic Multi-Competence. Cambridge: Cambridge University Press. 97-124. Franceschini, R. 2017. Vor- und Nachteile von Mehrsprachigkeit. Eine ergebnisoffene Darstellung. In: Krause, A., Lehmann, G., Thielmann, W. & Trautmann, C. (Eds.), Form und Funktion. Festschrift für Angelika Redder zum 65. Geburtstag. Tübingen: Stauffenburg. 463-474. Galvao-Carmona, A., González-Rosa J.J., Hidalgo-Muñoz A.R., Páramo D., Benítez M.L., Izquierdo G. & Vázquez-Marrufo M. 2014. Disentangling the attention network test: behavioral, event related potentials, and neural source analyses. Frontiers in Human Neuroscience, 8 (813). doi: 10.3389/ fnhum.2014.00813. Gathercole, V.C., Thomas, E.M., Kennedy, I., Prys, C., Young, N., Viñas Guasch, N., Roberts, E.J., Hughes, E.K. & Jones L. 2014. Does language dominance affect cognitive performance in bilinguals? Lifespan evidence from preschoolers through older adults on card sorting, Simon, and metalinguistic tasks. Frontiers in Psychology, (5) 11, 1-14. Gazzaniga, M.S., Ivry, B.I. & Mangun, G.R. (Eds.). 2002. Cognitive Neuroscience: The Biology of Mind. New York, NY: Norton. Gogolin, I. 1994. Der monolinguale Habitus der multilingualen Schule. Münster: Waxmann. Green, D.W. 2011. Language control in different contexts: the behavioral ecology of bilingual speakers. Frontiers in Psychology, (2) 103, 1-4. Green, D.W. & Abutalebi, J. 2013. Language control in bilinguals: The adaptive control hypothesis. Journal of Cognitive Psychology, 25 (5), 515-530. Paap, K.R. & Greenberg, Z.I. 2013. There is no coherent evidence for a bilingual advantage in executive processing. Cognitive Psychology, 66 (2), 232-258. Papathanasiou, I. & De Bleser, R. (Eds.). 2003. The Science of Aphasia: From Therapy to Theory. Amsterdam: Elsevier. Paradis, M. 2004. A Neurolinguistic Theory of Bilingualism. Amsterdam: John Benjamins. Poarch, G.J. & van Hell, J.G. 2012. Executive functions and inhibitory control in multi‐ lingual children: evidence from second-language learners, bilinguals, and trilinguals. Journal of Experimental Child Psychology, 113 (4), 535-551. Posner, M.I. & Peterson, S.E. 1990. The attention systems of the human brain. Annual Review of Neuroscience, 13, 25-42. Silveri, M.C. & Misciagna, S. 2000. Language, memory, and the cerebellum. Journal of Neuroimaging, 13, 129-143. 104 Gerda Videsott / Rita Franceschini <?page no="105"?> Videsott, G., Della Rosa, P. A. & Franceschini, R. 2015. Il multilinguismo e i meccanismi attentivi dei bambini provenienti da un contesto migratorio. Form@are - Open Journal per la formazione in rete, 3 (15), 185-196. doi: 10.13128/ formare-17212. Videsott, G., Herrnberger, B., Hoenig, K., Schilly, E., Grothe, J., Wiater, W., Spitzer, M. & Kiefer, M. 2010. Speaking in multiple languages: neural correlates of language proficiency in multilingual word production. Brain and Language, 113 (3), 103-112. 105 The link between multilingualism multilingualism and attention in children <?page no="107"?> “Does speaking a foreign language at school make your kids smarter? ” The development of non-verbal intelligence in a primary school offering an immersion and a mainstream foreign language programme Anja Steinlen / Thorsten Piske Abstract This paper examines the role of non-verbal intelligence of children at‐ tending either a partial immersion (IM) programme or a regular foreign language (FL) programme in a school in Germany, with a focus on grade 1 and grade 4. Using Raven’s Progressive Matrices tests as an index of non-verbal intelligence, significant differences between the children in the two programmes were found for the non-verbal intelligence test in grade 4 but not in grade 1. Furthermore, regression analyses explored possible effects of teaching programme, parental background, gender, and language background, but these variables did not serve as significant predictors for non-verbal intelligence in grade 4. In addition, reading skills in German and English were examined as a proxy for academic competence. No differences between IM and FL children were noted for the German reading test but IM children outperformed FL children in the English reading test. Moreover, non-verbal intelligence positively correlated with German and English reading tests in grade 4. In sum, the results of this study suggest that not only cognitive variables such as executive functioning and metalinguistic awareness, but also non-verbal intelligence may be a good measure to examine cognitive benefits within the field of bilingual school programmes. <?page no="108"?> 1 This question was actually posed in a North American context in which using a foreign language, i. e. not English, as a medium of instruction in schools was an option. 2 Being multilingual makes you smarter. 1 Introduction “Does not speaking English at school make your kids smarter? ” (Nehamas, 2012). This question 1 and similar ones are often found in newspapers and magazines, and they imply that children in bilingual programmes might show cognitive and, therefore, academic advantages as compared to children in mainstream programmes. However, are such assumptions actually true or just commercial ‘gags’ in order to promote more intensive foreign language programmes? In Germany, similar discussions have been going on in recent years because currently about 2 % of all public or private primary schools in Germany are offering a bilingual programme in which content subjects (i.e. mathematics, science, music, physical education) are taught in a foreign language (FL), usually English or French (fmks, 2014). Parents want their children to attend such a programme, hoping that better FL skills not only lead to better job opportunities but that their children’s general cognitive abilities (such as concentration and intelligence) are also better fostered than in mainstream programmes (where the FL is taught as a subject, usually for two 45-minute lessons per week). This view is supported by a growing number of journal articles in Germany that report on the putatively positive impact of multilingualism on intelligence (e.g. “Mehrsprachigkeit macht schlau” 2 , Springer, 2006). This paper examines whether it is true that intensive foreign language programmes affect children’s cognitive development in a positive way. In this longitudinal study, two groups of children are compared who attended two different programmes with different foreign language intensity from grade 1 to grade 4, i. e. from age six to ten. The two programmes, offered by a primary school in Germany, comprise of a partial immersion (IM) programme and a mainstream foreign language (FL) programme. In the latter, the children receive two 45-minute English lessons per week. In the IM programme, 50 % of the teaching time is conducted in English, i. e. all subjects except for German language arts, mathematics and religious education are taught in English (see Steinlen, 2016, 2017, 2018a, b, 2021; Steinlen & Piske, 2013, 2014, 2015, 2016, 2018a, b; Tamm, 2010). The cognitive variable examined here is non-verbal intelligence, which generally describes thinking skills and problem-solving abilities that do not fundamentally require verbal language production and comprehension. This variable is chosen because approximately 50 % of the children attending the 108 Anja Steinlen / Thorsten Piske <?page no="109"?> primary school examined here have a minority language background. This means that these children speak German and another family language at home (see e. g. Steinlen, 2021). At the beginning of the research project it was unclear how well developed the minority language children’s German language skills actually were in grade 1, and this is why we decided to employ cognitive tests which are independent of language (among others a test on non-verbal intelligence). In the following, we will briefly define non-verbal intelligence and introduce a test that is often employed to assess non-verbal intelligence, namely Raven’s Progressive Matrices. Afterwards, we will review studies examining cognitive ef‐ fects of bilingualism, with particular reference to the development of non-verbal intelligence in different foreign language (bilingual) school programmes. As populations in schools are always heterogeneous, the relationship between gender, language and culture, on the one hand, and non-verbal intelligence, on the other hand, are explored. Finally, we will examine the relationship between cognitive abilities and academic success, focussing on non-verbal intelligence and reading comprehension because there are studies that suggest interactional effects. 2 Background of the study 2.1 Definition of non-verbal intelligence According to Kuschner (2013: 2037), non-verbal intelligence describes thinking skills and problem-solving abilities that do not fundamentally require verbal language production and comprehension. Non-verbal intelligence involves manipulating or problem solving of visual information, and it may vary in the amount of internalised, abstract, or conceptual reasoning and motor skills that are required to complete a task. Thus, non-verbal intelligence tests are often used for children and immigrant adults because IQ tests generally require the person to know the language well in order to be adequately able to answer the questions of the test (Kuschner, 2013). Measures that are most widely used to test non-verbal intelligence include the Matrix Analogies Test (Naglieri, 1985), the Test of Nonverbal Intelligence (e.g. TONI-3, Brown, Sherbenou & Johnson, 1997), the Universal Nonverbal Intelligence Test (UNIT, Bracken & McCallum, 1998), the Wechsler Intelligence Scale for Children (e.g. WISC-IV, Wechsler, 2003) and, last but not least, different versions of the Progressive Matrices (Raven 1976a, 1976b). The Progressive Matrices are also employed in the present study, where the subject’s task is to find a missing piece in a geometrical pattern, using one of six or eight possible alternatives. 109 “Does speaking a foreign language at school make your kids smarter? ” <?page no="110"?> 2.2 Non-verbal intelligence in studies on bilingualism It has been noted in many studies that children who are raised with two languages show advantages over monolingual peers in terms of their cognitive development (see e. g. Adesope, Lavin, Thompson & Ungerleider, 2010; Bialystok & Barac, 2012 for a review). For example, bilingual children have been found to have an increased attention control and metalinguistic awareness as compared to monolingual children (Bialystok, 2001; Bialystok & Martin, 2004). They may also outperform monolingual children in tasks on metalinguistic awareness (Bialystok, 2001) and executive functioning (Bialystok, 2001; Carlson & Meltzoff, 2008; Festman, Rodriguez-Fornells & Münte, 2010). In addition (and most important for the present study), studies have shown that bilingual children obtained better test results than monolingual children for non-verbal intelligence: In a study discussing the relationship between bilingualism and intelligence, Peal & Lambert (1962) administered the Coloured Progressive Matrices (CPM) on 10-year-old bilingual French-English and mon‐ olingual English students in Canada. They found that bilingual ‘balanced’ children scored higher than monolingual children did. Based on these results, they suggested that bilingualism may in some way influence non-verbal intel‐ ligence and, additionally, that bilinguals may have developed more flexibility in thinking as they regularly switch from one language to the other. These were radical results and interpretations because previous tests had generally found bilinguals at a disadvantage in non-verbal intelligence tests, as compared to monolingual subjects (see Hakuta & Diaz, 1985 for a review of these early studies). Today the study by Peal & Lambert (1962) is considered a landmark study because for the first time, not only gender, age, and socio-eco‐ nomic background were controlled for, but also linguistic proficiency through self-evaluations of the languages spoken by the children on the one hand and through tests of vocabulary and association on the other hand (e.g. Lee, 1996). A few subsequent studies have supported the findings obtained by Peal & Lambert (1962). For example, Hakuta & Diaz (1985) conducted a longitudinal study and administered the CPM (among other tests) at kindergarten level and in grade 1 to bilingual Spanish-English children in a bilingual English-Spanish programme and to monolingual English children in a mainstream English programme in a school in the U.S. Their aim was to determine whether bilingualism enhances cognitive development or whether more intellectually gifted children become more highly proficient bilinguals. Hakuta & Diaz (1985) found the degree of bilingualism indeed to be related to non-verbal cognitive ability, suggesting that “bilinguals’ greater sensitivity to linguistic structure and detail is then transferred and generalized to different verbal and nonverbal 110 Anja Steinlen / Thorsten Piske <?page no="111"?> tasks” (p. 340). Nevertheless, as of today, it is not yet clear as to how much L1 and L2 input is actually necessary in order to improve cognitive skills (e.g. Wode, 1994). As Cummins pointed out in 1981, there may be a certain threshold of second language proficiency that is necessary before cognitive benefits will develop. Nowadays, measures of non-verbal intelligence in studies on cognition and bilingualism in the school context are usually only taken as control variables. The majority of recent studies on this topic have rather focused on executive functions such as attentional control, inhibitory control, working memory, and cognitive flexibility, as well as on reasoning, problem solving, and planning (e.g. Barac, Bialystok, Castro & Sanchez, 2014; Bialystok, 2005; Bialystok & Barac, 2012; Goriot, this volume). 2.3 Non-verbal intelligence in studies on foreign language learning in schools In recent years, a small but growing number of studies have included groups of children, attending intensive foreign language (bilingual) programmes in schools. These studies have examined whether positive cognitive effects are also evident in the absence of full bilingualism. As noted above, in 2014 2 % of all primary schools in Germany offered such a programme where one or several content subjects are taught in a foreign language (fmks, 2014). Many studies have shown that bilingual programmes are particularly effective when they operate according to immersion (IM) principles, i. e. when content subjects are taught entirely in the target (L2) language (e.g. Genesee, 1987; Wesche, 2002; Wode, 2009). In such a context, the students acquire skills in the school language (often the students’ first language, in this study German) as well as in the target language (here English). In her literature review, Wesche (2002: 362) concluded that IM programmes represent “the most effective means of school second language instruction yet developed”. According to Genesee (2004), early total IM programmes are the most time-consuming variety, where the L2 is taught 100 % of the teaching time for several years until students’ L1 is formally introduced in the curriculum in grade 2 or later. In Germany, however, public schools are only allowed to offer partial IM programmes, because the subject German language arts has to be taught in German (Kultusministerkonferenz, 2013). This means that a maximum of 70-80 % of the teaching time can be conducted in the L2. More and more researchers are interested in the cognitive development of children in such IM programmes. It is often assumed that in such a context, school children, just like children growing up bilingually from birth, show 111 “Does speaking a foreign language at school make your kids smarter? ” <?page no="112"?> cognitive benefits because they are exposed to the new language early, continu‐ ously, and intensively. According to some studies, children in an early French IM programme indeed obtained better results in tests on metalinguistic awareness than their peers attending an English monolingual mainstream programme (e.g. Bialystok, Peets & Moreno, 2014 for 2 nd and 5 th graders, but see Carlson & Meltzoff, 2008, for different results for pre-school children). Research has also found evidence for greater cognitive flexibility (Bruck, Lambert & Tucker, 1973, 1974) and better non-verbal problem-solving abilities of IM students (Bamford & Mizokawa, 1991). In a study by Nicolay & Poncelet (2013), 3 rd grade French-speaking children in an English IM programme were tested and compared to a similar group following a monolingual curriculum. The authors assessed attentional and executive skills by means of six different tasks such as alerting, auditory selective attention, divided attention, mental flexibility, and found that the immersion children did better than their monolingual counterparts in four of the six tasks. Based on these results, Nicolay & Poncelet (2013) concluded that the IM experience had already produced some of the cognitive benefits associated with early bilingualism (see Festman & Kersten, 2010; Grosjean, 2014; for a review, see also Goriot, this volume). Less is known about the development of non-verbal intelligence in children enrolled in IM programmes. The results of several studies comparing children’s performance in Progressive Matrices tests (Raven, 1976a, b) in IM and monolin‐ gual programmes are inconclusive: No significant between-group differences were found by Lambert & Tucker (1972) for students in grade 1 (see also Lambert, Tucker & D’Anglejan, 1973, for similar results for grades 4 and 5). Other studies have found significant differences in grade 1 in favour of IM students (e.g. Zaunbauer & Möller, 2007, 2010, and see also Bamford & Mizokawa, 1991, for children in grade 2). Two alternative explanations may be offered to account for these differences: The first argument points to a possible positive preselection of students before they even enter an IM programme (e.g. Genesee, 1987; see also Jaekel, this volume; Swain & Lapkin, 1982; Zaunbauer & Möller, 2007), the second relates to positive long-term cognitive effects of intensive L2 exposure, similar to results obtained for bilingual children (e.g. Peal & Lambert, 1962). Both explanations will be discussed in the following two sections. 2.3.1 Preselection of students in foreign language programmes When more students have applied for admission to an immersion programme than a school can accept, interested students may randomly be assigned to an immersion or regular school programme: Many schools, particularly in Canada, employ a lottery system, as in the case of the St. Lambert school (e.g. Genesee, 112 Anja Steinlen / Thorsten Piske <?page no="113"?> 1987). Ideally, then, the IM students would be identical to regular students in all respects, except that they would be enrolled in different programmes. It is hypothesised that students in IM programmes, therefore, fare equally well as their peers in regular programmes in all tests (except for foreign language tests) and, consequently, no between-group differences would be found with respect to non-verbal intelligence tests. Indeed, the results of Raven tests carried out in the St. Lambert school did not show any differences between 1 st graders in the IM and in the regular programme (e.g. Genesee, 1987; Lambert & Macnamara, 1969; Lambert & Tucker, 1972). However, it is not always possible to randomly assign students to a particular programme in a school. In most schools, attending an IM programme is optional and often subject to selection biases (see e. g. Apsel, 2012; Genesee, 1987; see also Jaekel, this volume; Swain & Lapkin, 1982; Zaunbauer & Möller, 2007). Student-selection factors may include age-appropriate knowledge of the L1, the ability to concentrate, perseverance, commitment, communication abilities (e.g. Kersten, Fischer, Burmeister, Lommel, Schelletter, Steinlen & Thomas, 2010). Primary schools may then discourage parents of struggling learners (i.e. dyslexic children and children with auditory/ perceptual/ concentration problems) to attend an IM programme, often reasoning that such a programme would be too big a burden for such children (Fischer, 2007). This is one of the reasons why in Germany IM programmes are often considered to be ‘elitist’ because they are indeed often attended by students with particular personal, intellectual, or familial characteristics. In such cases, different results in cognitive tests (e.g. relating to non-verbal intelligence) between IM children and children in mainstream programmes can be expected and are indeed found. In a large longitudinal study comparing students in mainstream and IM primary schools in Germany, Zaunbauer and colleagues (Gebauer, Zaunbauer & Möller, 2012, 2013; Zaunbauer & Möller, 2006, 2007, 2010) reported that in grade 1, IM students outperformed their peers in mainstream programmes in the CPM. They concluded that these cognitive differences might point to prior selection effects, which could impede appropriate interpretation of the data. 2.3.2 Long-term cognitive effects of foreign language programmes Differences in the performance of children in IM and mainstream programmes may also be a result of positive long-term cognitive effects due to intensive L2 exposure in IM programmes. Already after one year of IM teaching, Samuels & Griffore (1979) found 2 nd graders in a New York State French immersion pro‐ gramme to obtain better results than their peers in a monolingual programme in subtests on verbal IQ and performance IQ of the Wechsler Intelligence 113 “Does speaking a foreign language at school make your kids smarter? ” <?page no="114"?> Scale. Similarly, Bialystok et al. (2014) reported positive effects for children in grade 2 with respect to their metalinguistic awareness (using a test on morphological awareness, a sentence-judgement task and a verbal fluency test). Nicolay & Poncelet (2013) showed that IM children outperformed their peers in a mainstream monolingual programme on tasks assessing alerting, auditory selective attention, divided attention and mental flexibility after the children had been in an IM programme for three years. Not surprisingly, the results are more pronounced for older learners: Bialystok et al. (2014) found differences between IM and non-IM students with respect to tasks on metalinguistic awareness to be greater in grade 5 than in grade 2. Finally, in grade 6, IM students performed similar to non-IM students in tasks on cognitive flexibility (consisting of four subtests on uses, similarities, lines, and patterns) but better in the Embedded Figure Test (Bruck, Lambert & Tucker, 1973). However, in the studies mentioned above, the children all attended early total IM programmes. It is, therefore, yet far from clear after how many years to L2 exposure IM-children may gain particular cognitive benefits and whether similar effects would also be obtained for children attending a partial IM programme. When focusing on studies using Progressive Matrices tests in early total IM settings, a different picture emerges, at least with respect to data obtained in the St. Lambert experiment in Canada: Here, no differences between IM and non-IM students have been found, and this applies to grade 2, grade 3 (Lambert & Tucker, 1972) and grade 5 (Lambert, Tucker & D’Anglejan, 1973). For IM students in grades 2 and 5 and in grade 3 similar results have been reported by Bialystok et al. (2014) and Nicolay & Poncelet (2013, 2015), respectively. To our knowledge, only one study has found differences between IM and non-IM students in Raven tests. Bamford & Mizokawa (1991) administered the CPM to 2 nd graders in a Spanish IM programme and to monolingual English peers attending a mainstream English programme, matching both groups for socio-economic status and parental educational background, and found IM children to demonstrate a superior growth in non-verbal intelligence over the course of the school year as compared to children in mainstream programmes. Turning to the long-term effects of FL teaching in mainstream programmes on non-verbal intelligence, we are only aware of one study: Kristiansen (1990) examined 600 Finnish and 166 Hindi children from India aged 12-13, who were in their 4 th year of studying English as their first FL. Unfortunately, the number of hours of FL teaching per week was not specified. Kristiansen used different English language tests and the CPM to test for non-verbal intelligence. Her analysis indicates that three subgroups with varying non-verbal ability (low, average, high) differed significantly in their FL skills (comprehension 114 Anja Steinlen / Thorsten Piske <?page no="115"?> and production), suggesting that irrespective of the children’s L1, there was a positive correlation between non-verbal intelligence and FL learning under normal school conditions. In conclusion, it is as yet not clear how much L2 input is needed in how many years in order to trigger positive cognitive long-term effects, in particular for non-verbal intelligence. 2.4 Effects of culture and language Around 50 % of the children who attended one of the foreign language pro‐ grammes offered by the primary school reported in this study are regarded as immigrants. This means that they or their parents migrated to Germany (Statistisches Bundesamt, 2020), and that they, more often than not, spoke German and another family language at home (e.g. Steinlen, 2016, 2017, 2018a, b, 2021; Steinlen & Piske, 2013, 2014, 2015, 2016, 2018a, b). As noted above, it was not clear at the beginning of the project reported here how well developed the German language skills of these immigrant children were. We therefore decided on cognitive tests that are generally considered “language-free” and/ or “culture-fair” (sometimes also called “culture-free”) be‐ cause we did not want children’s German language skills and cultural techniques such as reading and mathematics to affect the outcome of these cognitive tests (see above, Kuschner, 2013). However, the notion of language-free and/ or culture-free tests has been criticised because even in those tasks, subjects usually have to read and comprehend written test items and instructions (see e. g. Mushquash & Bova, 2007). Furthermore, psychological tests may capitalise on children’s familiarity with European games: For example, children in Western societies are generally exposed to games involving puzzles, which may not be the case in other cultures (Serpell, 1994). In Raven tests, children who know how to puzzle may be at an advantage because the test consists of visual geometric designs, each with a missing piece and six or eight alternative pieces to choose from - similar to a puzzle. Therefore, researchers nowadays agree that culture and language have a significant impact on the performance of a minority group on an instrument designed and standardised within the majority culture. This is probably the reason why inconclusive results are reported with respect to the effects of language and cultural background on Raven tests. For example, some studies have found that strong performance in the Standard Progressive Matrices (SPM) is correlated to good comprehension in the language in which the test was taken, making the test items unequally difficult across language groups (e.g. Israel, 2006; Knowles, 2008). Kluever & Green (1994) tested a sample of 500 Hispanic and Anglo firstthrough fifth-grade children from different schools in southern Colorado and reported statistically significant differences between 115 “Does speaking a foreign language at school make your kids smarter? ” <?page no="116"?> Anglo and Hispanic children in the Coloured Progressive Matrices (CPM) total score (see also Valencia, 1979 for similar results for 3 rd graders of the same area). Different results were reported for children with a migrant background in Germany (who often but not always speak an/ other language/ s apart from German at home, i. e. minority language children) and their performance in non-verbal intelligence tests. For example, Heppt, Stanat, Dragon, Berendes & Weinert (2014) and Melzer, Rißling & Petermann (2015) did not find any significant group differences in non-verbal intelligence tests for children in pre-schools and primary schools. The same results have been found in smaller-scale studies comparing children with and without migration back‐ grounds attending primary school programmes in Germany with different foreign language intensity. The children obtained age-appropriate results in Raven tests, without any significant group differences between majority and minority language children (Steinlen, 2016, 2017, 2018a, b, 2021; Steinlen & Gerdes, 2015; Steinlen & Piske, 2013, 2014, 2015, 2016, 2018a, b). Thus, the outcomes of non-verbal intelligence tests, at least for the German context, do not seem to depend on children’s language and/ or migration backgrounds. 2.5 Effects of gender Gender differences in cognitive abilities have been widely analysed in the psy‐ chological and neuropsychological literature. Usually, two major differences in cognitive abilities between boys and girls are reported: (1) higher verbal abilities, favouring girls; (2) higher spatial abilities, favouring boys (see e. g. Ardila, Rosselli, Matute & Inozemtseva, 2011, for a review). However, similar differences have not been reported yet for cognitive tests on non-verbal intelligence (as operationalised by the Progressive Matrices). In these tests boys and girls usually do not differ in their performance (see Lynn & Irwing, 2004 in a meta-analysis of studies for the SPM), although males seem to outperform females from the age of 15 onwards (e.g. Lynn, Allik & Irwing, 2004). For the age range that is of primary interest for this study (i.e. primary schoolers between six to ten years), Steinlen (2018a) examined gender and language background effects on 4 th graders’ English skills attending different school programmes and also tested the children on non-verbal intelligence. She did not find any gender-specific effects with respect to children in a regular programme in grade 4. However, according to SPM scores obtained in a partial IM programme boys outperformed girls, although both groups of children still obtained age-appropriate results. Possible reasons for the differences noted between the two programs were not discussed by Steinlen (2018a). 116 Anja Steinlen / Thorsten Piske <?page no="117"?> 3 It should be noted here that central determinants of reading include, among others, world knowledge, the ability to rapidly access lexical items, reading speed, short-term memory, broad and in-depth vocabulary knowledge, reading motivation, a positive attitude towards reading, knowledge of text features, and reading strategies (Bundesministerium für Bildung und Forschung, 2007). 2.6 Effects of cognitive abilities on academic success: L1 and L2 reading It is well known that intelligence (measured as the intelligence quotient or IQ) is one of the important prognostic variables in the academic outcome of children (e.g. Gamsjäger & Sauer, 1996; Lassiter & Bardos, 1995; von Stumm, Hell & Cha‐ morro-Premuzic, 2011). Students whose higher mental ability is demonstrated by the results of IQ tests tend to achieve highly in academic settings. However, non-verbal IQ has proven to be less effective in predicting academic outcomes than verbal IQ (e.g. DeThorne & Schaefer, 2004). This difference in predictive power is potentially due to two explanations, neither of which precludes the other: (1) Verbal subtests have a higher percentage of variance attributable to general intelligence than non-verbal subtests; and (2) language plays an important role in the standard academic learning environment. Note, however, that a least a few studies have found non-verbal intelligence to be a predictor for academic outcome, with respect to success in L1 reading 3 in school. For example, Stanovich, Cunningham & Feeman (1984) examined the relationship between Raven’s SPM, the Gates-MacGinitie Reading Tests, and the Reading Survey of the Metropolitan Achievement Test. Correlations ranged from 0.30 in primary grades to a high of 0.70 in grade 9 and above, indicating age-related changes in the relationship between reading and non-verbal intelligence. In a large-scale investigation, Naglieri (2001) demonstrated that both group and individually administered tests of non-verbal ability are significantly correlated with reading achievement. Using the Naglieri Non-Verbal Ability Test, Morvay (2015) examined 65 twelfth graders on their L1 Hungarian reading comprehension skills and found correlations between non-verbal intelligence and the Hungarian reading test to be moderate (r = 0.31, p<.05), disconfirming the prediction that there would be a non-correlation between reading, which is a verbal skill, and non-verbal intelligence. Non-verbal intelligence tests have also been used (although unsuccessfully) to distinguish between poor and average L1 readers (e.g. Fathi-Ashtiani & Ahmadi, 2006; Nation & Snowling, 1998). Based on the inconsistent results of these few studies, it seems that the relationship between L1 reading comprehension and non-verbal intelligence is still not well understood. With respect to the relationship between non-verbal intelligence and L2 reading comprehension, the authors are only aware of one study, which was conducted in the mainstream foreign language school context. In the same study 117 “Does speaking a foreign language at school make your kids smarter? ” <?page no="118"?> as mentioned above, Morvay (2015) also tested the 65 Hungarian twelfth-graders on L2 English reading comprehension. At the time of testing, the students had been studying English in school for eight years. Morvay (2015) reported significant correlations between non-verbal intelligence and L2 reading com‐ prehension, albeit not a very strong one (r = 0.282). She speculated that this result may be due to the fact that parallel to L1 reading, L2 reading also requires highly analytical skills, which may be the reason why in her study non-verbal intelligence significantly correlated with L2 reading. Other studies on younger children refer to the possible role of non-verbal intelligence as a predictor for L2 reading skills, but their results are not conclusive: On the one hand, Trites & Price (1978, 1980) administered a battery of tests to English four-year-old children in a French IM kindergarten. The three best predictors of their L2 French reading ability in grade 1 were non-verbal intelligence, an L1 English reading test, and teacher ratings of English auditory comprehension. On the other hand, MacCoubrey, Wade-Woolley, Klinger & Kirby (2004) did not find an effect of non-verbal intelligence. They administered an L1 English phonological awareness test to French IM children at the begin‐ ning of grade 1 and an L1 English and L2 French word identification test in grade 2. Non-verbal IQ scores were not successful in discriminating between successful and poor L2 word readers in grade 2, whereas rapid naming and phonological awareness proved to be better predictors. As these tests were conducted with very young and beginning learners of an L2, it is not clear how the relationship between non-verbal intelligence and L2 reading may change with more L2 experience (see e. g. Morvay, 2015 above). 2.7 Research questions The present study focuses on the development of cognitive skills (operational‐ ised as non-verbal intelligence) of children attending either a partial immersion (IM) programme or a mainstream foreign language (FL) programme in the same primary school. The longitudinal study follows one cohort of each programme (starting in the same year) from grade 1 to grade 4. Specifically, we address the following research questions: (RQ1) Do IM students outperform FL students in the non-verbal intelligence test conducted in grade 1, supposedly because students are positively pre-se‐ lected for IM-programmes (e.g. Genesee, 1987; Jaekel, this volume; Rumlich, this volume; Zaunbauer & Möller, 2007)? (RQ2) Do IM students outperform FL students in the non-verbal intelligence tests conducted in grade 4, because of the beneficial effects IM-programmes supposedly have on children’s cognitive abilities (e.g. Bialystok et al., 2014)? 118 Anja Steinlen / Thorsten Piske <?page no="119"?> (RQ3) Are the results of the non-verbal intelligence tests affected by the children’s gender and/ or their language/ cultural background (e.g. Kluever & Green, 1994; Lynn & Irwing, 2004)? (RQ4) How does non-verbal intelligence affect German and English reading comprehension skills at the end of grade 4 (e.g. Morvay, 2015)? 3 Method 3.1 Participants and procedure The data presented in this paper were collected in a (non-private) district primary school in the city of Tübingen in Germany. Since 2008/ 09, the school has offered both a mainstream foreign language programme and a partial immersion programme. In the partial immersion (IM) programme, all subjects except for German language arts, religious education and mathematics are taught in Eng‐ lish from the first day of grade 1 onwards, corresponding to 50 % of the teaching time. The students receive their instruction from native speakers of German who studied English in order to become English teachers (often focusing on bilingual teaching). These teachers exclusively speak English in class, although technical terms are always introduced in both English and German (see Dallinger, this volume; Steinlen, 2021; Tamm, 2010). In the mainstream foreign language (FL) programme, the subject English language arts is taught for two lessons per week from grade 1 to grade 4. The content of these lessons adheres to the standards for the subject English as decreed by the Ministry of Education, Youth and Sports Baden Württemberg (Ministerium für Kultus, Jugend und Sport Baden Württemberg 2004, 2016). When the IM-programme (which also includes the present sample) was first established, the number of applicants exceeded the number of available places for the bilingual cohort. At that time, the school board, together with parents and a neighbouring pre-school attended by many of the primary school children, decided on whether a particular child should attend the IM-programme by taking into consideration the child’s concentration, language, communication and hearing abilities (see Tamm, 2010). In recent years, however, the number of students enrolling in each of the two programmes offered by the primary school did not differ considerably so that such a procedure has not been used anymore. Altogether 151 children took part in this longitudinal study, comparing data of 74 children of the IM programme to data elicited from 77 children of the FL programme in grade 1 and in grade 4, disregarding those children who started the respective programme after grade 1. In the IM programme, there were 40 boys and 34 girls, 53 % of these children had a migration background. The FL 119 “Does speaking a foreign language at school make your kids smarter? ” <?page no="120"?> programme was attended by 38 girls and 39 boys, more than half (56 %) had a migration background. Such a background was attested when one or both parents were born abroad, and, more importantly, when a language other than German was spoken at home (see e. g. Mullis, Martin, Foy & Hooper, 2017 for PIRLS). The measures were taken at the end of the school year. In grade 1, the children in the IM programme were 7; 0 years on average (SD: 6 months, range: 70-93 months), in the FL programme, they were 7; 2 years on average (SD: 7 months, range: 71-102 months). In grade 4, the children were, accordingly, three years older. Because the tests were carried out on different days, the number of participants per test varied due to sickness or school-related activities. 3.2 Measures 3.2.1 Non-verbal intelligence tests For this study, Raven’s Progressive Matrices were chosen. They seemed to be particularly suitable for minority language children whose German skills might not have been developed in an age-appropriate way in grade 1 and the Matrices had already been employed in previous studies examining primary schoolers attending IM and FL programmes in Germany (e.g. Gebauer et al. , 2013; Kuska, Zaunbauer & Möller, 2010; Zaunbauer, Bonerad & Möller, 2005; Zaunbauer, Gebauer & Möller, 2012; Zaunbauer & Möller, 2006, 2007; ). Because test scores relating to linguistic or academic achievement may be affected by cognitive variables (e.g. Bleakley & Chin, 2004; Chudaske, 2012; Gamsjäger & Sauer, 1996), it is always important to include cognitive variables into studies comparing different groups of learners. Following Zaunbauer and colleagues, we administered the CPM (Coloured Progressive Matrices; Raven, 1976a) in grades 1, 2 and 3 and the SPM (Standard Progressive Matrices; Raven, 1976b) in grade 4. For the purpose of this study, only the results of grade 1 and grade 4 for children in the IM programme and in the FL programme are reported. The children’s task was to complete an incomplete geometrical pattern, using one of six or eight possible alternatives, respectively. The test consists of 36 or 48 items, which were presented in three or four sets of 12 in an increasing order of difficulty within each set. 30 minutes were allocated for both tests. The reliability values (Cronbach’s α) for German children were reported to be 0.90 for grade 1 (Bulheller & Häcker, 2010), and 0.93 for grade 4 (Heller, Kratzmeier & Lengfelder, 1998). According to the manuals, the CPM and SPM proved to be good indicators for Spearman’s g-factor which also yielded satisfying correlations with school performance tests. 120 Anja Steinlen / Thorsten Piske <?page no="121"?> 3.2.2 Reading tests German reading test: German reading comprehension was measured by means of the ELFE (Ein Lesetest für Erstbis Sechstklässler; Lenhard & Schneider, 2006). This test meas‐ ures reading comprehension at word, sentence, and text level. At word level, one of four words had to be assigned to a picture. At sentence level, the students selected one of four alternative words to complete a sentence. At text level, the children answered questions on short texts. Altogether, 16 minutes were allocated for this test. Internal consistency reliability estimates of 0.92 to 0.97 were reported, depending on the subtest (Lenhard & Schneider, 2006). Furthermore, the authors noted moderate to high correlations with respect to other reading test formats, teacher evaluations, and grades for German language arts (0.45-0.71). The ELFE has been used in many German schools to evaluate reading skills, and the teachers were all familiar with the test format. Only the results of the ELFE in grade 4 are reported in this study. English reading test: The PSAK (Primary School Assessment Kit, Little, Simpson & Čatibusič, 2003) was originally designed for assessing immigrant children’s English language skills in Ireland (i.e. English as an additional language). This non-standardised test was chosen because grading is carried out in relation to the three levels A1, A2 and B1 used within the Common European Framework of Reference (CEFR, Council of Europe, 2001). The PSAK was also chosen because of its colorful design, which, according to many students, motivated them to work through the booklet. The sub-test on reading (PSAK-R) measures reading comprehension on the word level (e.g. drawing a line from a word to a matching picture), on the sentence level (e.g. matching pictures with sentences) and on the text level (e.g. answering questions on texts). A total of 45 points can be obtained. More than 30 points correspond to level A2 and more than 43 points to level B1. As test quality criteria are not available for the PSAK-R, split half reliability values (Cronbach’s alpha) were computed with the present sample, which resulted in values of 0.79-0.81 (IM) and 0.72-0.77 (FL) for grade 4, yielding satisfying results. 3.3 Parents’ questionnaire In addition to the child’s age and country of birth, the parents provided informa‐ tion about the language/ s used at home and their educational background, with 0 corresponding to no school certificate and 6 to a university entrance certificate (e.g. Gebauer et al., 2013). In this study, only data on maternal education are provided. The parents also assessed their relative wealth compared to other families on a five-point scale, ranging from 1 (not wealthy at all) to 5 (very 121 “Does speaking a foreign language at school make your kids smarter? ” <?page no="122"?> wealthy). Moreover, the parents provided information on reading activities at home involving their child (e.g. Dickinson, Griffith, Golinkoff & Hirsh-Pasek, 2012) and parental supervision of the child’s homework (Fan & Chen, 2001), which they rated on a four-point scale (1 = never, 4 = very often). The questionnaire was answered by 122 parents altogether, i. e. there were 29 missing questionnaires regarding the tested children (16 minority language and 13 majority language children). 4 Results SPSS 26 version (2019) was used to compute statistical analyses, in particular ANOVAs and linear regression analyses. The data were not cleaned for outliers, and missing data were not imputed. In all of the analyses, it was seen to that the assumptions of homogeneity of variance and sphericity were met. The following results are reported in raw scores. 4.1 Test for family variables The family variables were taken from 122 parents’ questionnaires, corre‐ sponding to a response rate of 81 % (generally, 75 % - 80 % are considered acceptable, e. g. Draugalis, Coons & Plaza, 2008). According to the teachers, non-responses were usually due to scepticism about surveys in general and/ or parental time constraints, which may also account for the fact that parents did not always answer all of the questions. Table 1 compares the children in the two foreign language programmes (IM vs. FL) with respect to their parental background variables (i.e. self-estimated wealth, educational background, and parental activities with the child). As shown in Table 1, none of the comparisons reached a conventional significance level (p<0.05). In general, the parents considered their background as average in terms of wealth, and they had a high educational background, corresponding to a university entrance certificate. They greatly supported their child, as shown in the high values obtained for pre-school reading activities and homework supervision. The parents of the present study may, therefore, be described as middle-class, highly educated, and concerned about their child’s educational welfare, independent of whether their child attended the IM programme or the mainstream programme. As opposed to the notion that IM-programmes, in contrast to regular programmes, are often attended by children from a higher socio-economic status (see e. g. Apsel, 2012; Zydatiß, 2007), in this study the parents’ educational and socio-economic background is not a distinguishing factor. 122 Anja Steinlen / Thorsten Piske <?page no="123"?> Max. points IM pro‐ gramme M (SD) [N] FL programme M (SD) [N] Comparison (F-test) Self-estimated wealth 5 3.41 (0.63) [N = 54] 3.26 (0.77) [N = 51] F(1,104) = 2.575, p = 0.072 Maternal educa‐ tion 6 5.17 (1.30) [N = 60] 4.97 (1.30) [N = 51] F(1,109) = 3.266, p = 0.116 Parents read book to pre-schooler 4 3.30 (0.77) [N = 65] 3.11 (0.81) [N = 58] F(1,121) = 2.003, p = 0.160 Supervision of homework 4 3.48 (0.77) [N = 65] 3.34 (0.84) [N = 59] F(1,122) = 0.904, p = 0.344 Table 1: Mean values (M), standard deviations (SD), sample size (N), maximal points, norm values and the results of one-way ANOVAs, comparing IM (immersion) and FL (mainstream foreign language) children with respect to their family background and other child-related parental activities. 4.2 Non-verbal intelligence in grades 1 and 4 Table 2 presents the results of the two non-verbal intelligence tests in the IM and the FL group (conducted in grades 1 and 4), including the results of ANOVAs. The children in the IM group and the FL group in grade 1 obtain scores in the Raven tests that correspond to norm values, i. e. are age-appropriate. However, in grade 4, the children in the IM and the FL programme differ significantly regarding their test scores in the SPM, with the IM children outperforming their FL peers. Note, however, that the scores for both groups of students are age appropriate. Measure Max. points Norm values IM pro‐ gramme M (SD) [N] FL pro‐ gramme M (SD) [N] Comparison (F-test) CPM grade 1 36 24-26 26.6 (5.8) [N = 71] 24.9 (5.8) [N = 66] F(1,135) = 2.840, p = 0.094 SPM grade 4 48 38-41 41.7 (6.1) [N = 59] 37.6 (7.7) [N = 55] F(1,112) = 9.863, p = 0.002* Table 2: One-way ANOVAs of Raven scores by teaching programme (IM vs. FL) in grades 1 and 4. The asterisk indicates significant differences at the level p<0.05. 123 “Does speaking a foreign language at school make your kids smarter? ” <?page no="124"?> 4.3 Effects of gender and language background Next, we tested whether the results of the cognitive tests of the IM and FL children in grades 1 and 4 were affected by language status (i.e. minority vs. majority language background) or by gender. The results are illustrated in Table 3a (for gender) and in Table 3b (for language background). Group Girls M (SD) [N] Boys M (SD) [N] Comparison (F-test) Gender grade 1 (CPM for IM) 25.2 (6.1) [N = 33] 27.7 (5.3) [N = 38] F(1,69) = 3.468, p = 0.067 Gender grade 1 (CPM for FL) 24.4 (6.7) [N = 32] 25.5 (4.8) [N = 34] F(1,64) = 0.436, p = 0.511 Gender grade 4 (SPM for IM) 40.2 (7.3) [N = 29] 42.2 (4.3) [N = 30] F(1,57) = 3.608, p = 0.063 Gender in grade 4 (SPM for FL) 38.1 (7.3) [N = 28] 37.1 (8.2) [N = 27] F(1,53) = 0.258, p = 0.614 Table 3a: One-way ANOVAs of Raven scores by gender in grades 1 and 4 for IM and FL students. Significant differences between girls and boys were neither found for the FL programme nor for the IM programme. Apparently, gender does not affect the scores of the Raven tests. Group Majority language M (SD) [N] Minority language M (SD) [N] Comparison (F-test) Language grade 1 (CPM for IM) 28.4 (5.1) [N = 33] 24.6 (5.6) [N = 38] F(1,69) = 11.695, p = 0.001* Language grade 1 (CPM for FL) 25.4 (5.3) [N = 30] 24.6 (6.1) [N = 36] F(1,64) = 0.318, p = 0.575 Language grade 4 (SPM for IM) 42.6 (5.8) [N = 30] 40.8 (6.3) [N = 29] F(1,57) = 1.196, p = 0.279 Language grade 4 (SPM for FL) 38.9 (7.5) [N = 24] 36.6 (7.8) [N = 31] F(1,53) = 1.198, p = 0.279 Table 3b: Two-way ANOVAs of Raven scores by language background in grades 1 and 4 for IM and FL students. 124 Anja Steinlen / Thorsten Piske <?page no="125"?> As Table 3b shows, the variable language background only yielded significant group differences in grade 1, favouring majority over minority language chil‐ dren in the IM programme. However, this difference disappeared by grade 4, where majority and minority children performed alike. Significant language background-specific differences were not found for children in the FL pro‐ gramme, neither in grade 1 nor in grade 4. 4.4 Regression analyses on background variables By computing multiple linear regression analyses (Table 4), we tested whether non-verbal intelligence at the end of grade 4 can be predicted by the teaching programme (IM vs. FL), the children’s language background, their gender, their parental background, and/ or their scores obtained in the non-verbal intelligence test in grade 1. Dependent variable and predictors Beta R 2 corrected Non-verbal intelligence grade 4 (N = 72) Parents read to pre-schooler 0.152 = 0.164 Supervision of homework -0.149 = 0.167 Maternal education 0.199 = 0.098 Self-estimated wealth 0.035 = 0.743 Language background 0.109 = 0.287 Gender -0.033 = 0.734 CPM grade 1 0.426 = 0.000* Teaching programme (IM, FL) -0.141 = 0.172 0.343, p = 0.000 Table 4: Linear regression predicting non-verbal intelligence at the end of grade 4. Table 4 shows that neither the children’s background variables (such as parental factors, language background and gender) nor the teaching programme (all beta’s ns) exerted any influence on their non-verbal intelligence scores in grade 4. Non-verbal intelligence was only significantly predicted by the non-verbal intelligence test (CPM) conducted in grade 1, which accounted for 34.3 % of variance. 125 “Does speaking a foreign language at school make your kids smarter? ” <?page no="126"?> 4.5 German and English reading comprehension tests In this section, reading comprehension in German and English in grade 4 is used as an example for academic outcome in the two different foreign language programmes. Table 5 presents the results of the two reading tests conducted in grade 4 for the IM and the FL students. Measure Max. points Norm values IM children M (SD) [N] FL students M (SD) [N] Comparison (F-test) German reading com‐ prehension (ELFE) 120 88 95.4 (17.4) [N = 61] 97.0 (19.1) [N = 53] F(1,112) = 0.236, p = 0.628 English reading com‐ prehension (PSAK-R) 45 B1: 43+ A2: 30+ 39.2 (4.2) [N = 57] 28.9 (6.0) [N = 54] F(1,109) = 101.444, p = 0.000* Table 5: One-way ANOVAs of German and English reading test scores (grade 4) by teaching programme for IM and FL students. For the German reading comprehension test ELFE, the results did not show any significant differences between children in the two different programmes. Furthermore, both groups obtained age-appropriate results, even above average. Not surprisingly, the IM and FL students differed significantly regarding the outcomes of the English reading comprehension test PSAK-R in grade 4. Whereas IM students received scores at the upper level of A2 (in fact, half of the students reached level B1), FL students in general performed at the lower level at A2, half of them still ranging at level A1. 4.6 Non-verbal intelligence and reading Finally, in order to examine the relationship between non-verbal intelligence and reading outcomes in German and English at the end of grade 4, correlation analyses were conducted, taking into consideration the scores of the CPM in grade 1, the SPM in grade 4, and the results of the German and English reading comprehension tests in grade 4. Table 6a illustrates the results for IM students and Table 6b for FL students. 126 Anja Steinlen / Thorsten Piske <?page no="127"?> CPM grade 1 SPM grade 4 ELFE grade 4 PSAK-R grade 4 CPM grade 1 1 0.603 0.000* 0.118 0.368 0.142 0.297 SPM grade 4 0.603 0.000* 1 0.338 0.011* 0.341 0.011* ELFE grade 4 0.118 0.368 0.338 0.011* 1 0.487 0.000* PSAK-R grade 4 0.142 0.297 0.341 0.011* 0.487 0.000* 1 Table 6a: Correlation analyses of non-verbal intelligence by reading tests for IM students, indicating Pearson correlation and significance values. The results indicated, as expected, significant correlations between the CPM in grade 1 and the SPM in grade 4, i. e. the better the CPM test scores in grade 1, the better the SPM test scores in grade 4. Furthermore, we found significant correlations between the SPM of grade 4 and both reading tests, i. e. the higher the SPM scores, the higher the scores in the reading tests, independent of the target language. However, the CPM in grade 1 neither correlated with the German nor with the English reading test conducted in grade 4. For IM students, in sum, reading outcomes in grade 4 only correlated with non-verbal intelligence assessed in grade 4 but not in grade 1. CPM grade 1 SPM grade 4 ELFE grade 4 PSAK-R grade 4 CPM grade 1 1 0.468 0.000* 0.154 0.300 0.345 0.019* SPM grade 4 0.468 0.000* 1 0.355 0.011* 0.373 0.006* ELFE grade 4 0.154 0.300 0.355 0.011* 1 0.521 0.000* PSAK-R grade 4 0.345 0.019* 0.373 0.006* 0.521 0.000** 1 Table 6b: Correlation analyses of non-verbal intelligence by reading tests for FL students (Pearson correlation and significance). 127 “Does speaking a foreign language at school make your kids smarter? ” <?page no="128"?> Similar results were noted for the FL students. The SPM test scores correlated significantly with both reading tests (the better the SPM scores, the better the scores in the two reading tests), and the CPM in grade 1 did not correlate with the German reading test. However, in contrast to the IM group, the CPM scores in grade 1 also correlated significantly with the English reading test PSAK in grade 4, and a significant positive correlation was also noted between the English and the German reading test. 5 Discussion This chapter examined the role of non-verbal intelligence of children attending either a partial immersion programme or a foreign language programme in a primary school in Germany. The children were followed longitudinally from grade 1 to grade 4. Using Raven’s Progressive Matrices as an index of non-verbal intelligence, this study focused on possible pre-selectional effects in IM programmes as compared to mainstream FL programmes in non-verbal intelligence tests in grade 1 (RQ 1), on possible positive long-term effects of an IM programme on non-verbal intelligence in grade 4 (RQ 2), and generally, on possible effects of gender and language background on non-verbal intelligence (RQ 3). Finally, we examined the relationship between academic outcome (as operationalised by L1 and L2 reading comprehension) and non-verbal intelli‐ gence (RQ 4). The results of the analyses regarding these research questions are discussed in the subsequent sections. 5.1 Selection effects in grade 1 A comparison of the scores of Raven tests of 1 st graders in the IM programme with their peers in the FL programme did not yield any significant differences between the groups. This is an interesting result because, as noted in section 3.1, when the two cohorts started their respective programmes, more students than could be accepted into the partial IM programme had applied for admission and were preselected for age-appropriate knowledge of the L1, the ability to concentrate, perseverance, commitment and communication abilities (see Tamm, 2010). Therefore, we expected IM children and non-IM children in grade 1 to score differently in the test on non-verbal intelligence (see e. g. Gebauer et al., 2012, 2013; Zaunbauer & Möller 2006, 2007, 2010). However, the expected differences were not shown by the children in the two groups (see also Yadollahi Jouybari, Steinlen & Piske, 2020). Indeed, the findings of this study replicate the results from the St. Lambert experiment in Canada, where, in contrast to the present sample, students 128 Anja Steinlen / Thorsten Piske <?page no="129"?> were randomly assigned to an immersion or to a regular school programme. Lambert and colleagues did not find any significant differences in Raven tests for grade 1 students (e.g. Genesee, 1987; Lambert & Macnamara, 1969; Lambert & Tucker, 1972). We speculate that the preselection of students based on particular personal, intellectual, or familial characteristics do not always lead to cognitive abilities, favouring IM students, at least not with respect to non-verbal intelligence. 5.2 Benefits in grade 4 for non-verbal intelligence Positive long-term effects for different cognitive abilities due to intensive L2 exposure in IM programmes have been reported in many studies, where the per‐ formance of children in IM and mainstream programmes (usually monolingual ones) were compared in grade 2 (Samuels & Griffore, 1979), in grade 3 (Nicolay & Poncelet, 2013, 2015), and in grade 5 (Bialystok et al., 2014) with respect to tasks on metalinguistic awareness, alerting, auditory selective attention, divided attention, and mental flexibility. In these studies, IM students always outperformed non-IM students. Regarding non-verbal intelligence, our study also showed that the IM chil‐ dren in grade 4 outperformed their peers in the FL programme, although no significant group differences were found in grade 1. This result contradicts findings from studies for early total IM programmes in Canada where no group differences were noted between IM students and non-IM students in non-verbal intelligence tests (Genesee, 1987; Lambert & Tucker, 1972; Lambert et al., 1973). Independent of the programme, the students nevertheless obtained age-appropriate results in these tests, indicating that the cognitive benefits regarding non-verbal intelligence in FL programmes with different intensity are indeed rather small over the primary school years. It seems that, in general, positive long-term cognitive effects due to FL exposure in the school context are not yet well understood. We therefore suggest conducting longitudinal studies that examine children continuously over many years with a variety of cognitive tests and preferably even beyond their school years to better determine if and when particular types of cognitive benefits can be expected and why. 5.3 Effects of language background and gender As many children in our sample have a migration background, we also examined whether their language background may have had an effect on their non-verbal intelligence test scores. On the one hand, following Peal & Lambert (1962, see also Adesope et al., 2010; Bialystok & Barac, 2012), positive cognitive effects may be expected as these children grew up at home with more than one language 129 “Does speaking a foreign language at school make your kids smarter? ” <?page no="130"?> (usually a family language such as Arabic, Farsi, French, Spanish, Turkish etc. plus German), i. e. they classify as early bilinguals. On the other hand, it has often been pointed out (e.g. Kuschner, 2013) that non-verbal intelligence tests are neither entirely language-free nor culture-fair, because to a certain extent even these tasks do draw on language skills (e.g. to understand test instructions) and cultural techniques (e.g. prior knowledge of the concept of puzzles may facilitate test tasks), which favour children with a German background. The results of our study did, however, not indicate any impact of language background on the Raven tests conducted in grades 1 and 4, where majority and minority children performed equally well, independent of the foreign language programme they attended. A similar result has been reported for Raven tests taken by migrant vs. non-migrant (or minority vs. majority language) children in other IM cohorts and programmes in Germany (e.g. Steinlen, 2016, 2017, 2018a, b, 2021; Steinlen & Gerdes, 2015; Steinlen & Piske, 2013, 2014, 2016, 2018a, b, 2020). This finding also points to the fact that non-verbal intelligence does not seem to be affected by language background (see also Heppt et al., 2014; Melzer et al., 2015). The minority language children in this sample had two advantages when dealing with the Progressive Matrices. First, their competence in German even in grade 1 was apparently high enough to understand any test instructions (with reading and writing scores being age-appropriate, see also Yadollahi Jouybari et al., 2020). Second, and, more importantly, because they had attended a kindergarten before they entered school (as indicated in the family questionnaires), they were familiar with cultural techniques in Germany (i.e. puzzles), which may have facilitated the tasks of completing the Progressive Matrices (e.g. Kuschner, 2013; Melzer et al., 2015). Regarding gender, studies on verbal IQ in particular have repeatedly reported differences between boys and girls, with females usually outperforming males (see e. g. Ardila et al., 2011). However, such effects have not been noted for non-verbal intelligence tests, where boys and girls generally have not differed in their performance (see e. g. Lynn & Irwing, 2004). Our results confirm this finding: Although the children in our sample were much younger (i.e. age six to ten years), we did not find any significant gender-specific results (see also Bulheller & Häcker, 2010; Heller et al., 1998). Gender, therefore, does not seem to affect non-verbal intelligence in any way, at least not for children in primary school age, although more research with a larger sample is necessary to corroborate this finding. 130 Anja Steinlen / Thorsten Piske <?page no="131"?> 5.4 Non-verbal intelligence and reading In this study, reading skills were examined as a proxy for academic achievement, as reading is one of the basic skills acquired during the primary school years, and it plays a key role in the acquisition of academic knowledge and later participation in society (e.g. Bundesministerium für Bildung und Forschung, 2007). However, we not only examined reading comprehension in German (usually the children’s L1) but also in the foreign language English in grade 4. The analyses yielded significant differences between students in the FL and in the IM programme for the English reading test, with IM students obtaining scores between level A2 and B1, and FL students in between A1 and A2. Group differences for FL reading due to FL intensity of the programme have been reported in many studies (see e. g. Steinlen, 2021 for a review). No such group differences were noted for the German reading test, where IM and FL students obtained age-appropriate results. Thus, an intensive FL programme does not necessarily lead to negative results in the school language, at least as far as reading comprehension is concerned (see Steinlen, 2021 for a review). The relationship between academic outcome and cognitive abilities is a well-established one, in that students with higher mental ability (as demon‐ strated by IQ tests) tend to achieve highly in academic settings (e.g. Gamsjäger & Sauer, 1996; Lassiter & Bardos, 1995; von Stumm et al., 2011). For IM students in particular, it has been suggested that they perform better than their peers in mainstream programmes in tests on reading and mathematics because of greater cognitive abilities (e.g. Zaunbauer & Möller, 2007). We restricted our analysis of the relationship between cognitive abilities and academic outcome to non-verbal intelligence and reading comprehension, because some studies have suggested a positive relationship between non-verbal intelligence and L1 reading comprehension (Naglieri, 2001; Stanovich et al., 1984) and, moreover, between non-verbal intelligence and FL reading compre‐ hension (Morvay, 2015). Our results supported such a view: We found significant positive correlations between the scores of the Raven tests and the reading comprehension tests in English and German, all conducted in grade 4. It seems that in L1 and L2 reading comprehension, non-verbal intelligence does indeed play a role, although more studies with a larger number of students are needed to explore such a relationship in more detail, in particular with respect to foreign language learning in programmes with different intensity. 5.5 Limitations and conclusions of the present study Several limitations restrict the generalisability of the results obtained in the present study: More subjects in both foreign language programmes are war‐ 131 “Does speaking a foreign language at school make your kids smarter? ” <?page no="132"?> ranted to corroborate the findings reported here, including more in-depth statistical analyses appropriate for larger samples. Furthermore, this study was only concerned with non-verbal intelligence as an example of a cognitive ability. It is far from clear whether different cognitive tests (for example on attentional and executive abilities, e. g. Bialystok, 2001, or on attention and concentration) conducted with these children would yield different outcomes because many of these studies rather focused on bilingual children and less on children in different teaching programmes. Moreover, as the design of the study was a quasi-experimental one, causal effects and their interactions could not be examined (see also Zaunbauer & Möller, 2007), so that it is not clear whether better scores in non-verbal intelligence tests lead to better reading comprehension in German or English or vice versa. In addition, as Baumert, Köller & Lehmann (2012) point out, measures are still not available which separate effects of high-quality teaching from structural effects of bilingual programmes per se, which may have affected the results of students in the FL and IM programme in the present study. In sum, the results of our study examining the question “Does speaking an additional language at school make your kids smarter? ” are rather inconclusive. At least with respect to non-verbal intelligence, we did not find any differences between children who attended one of the two foreign language programmes after the first year of school. Thus, any pre-selectional effects may have taken place before the children even entered the programme. Long-term effects were noted in grade 4, with IM students outperforming FL students in the non-verbal intelligence test, although both groups obtained age-appropriate results. Only the CPM in grade 1 turned out to be a significant predictor of non-verbal intelligence in grade 4, with teaching programme and the children’s language background, family background and gender not being significant predictors. Finally, non-verbal intelligence affected German and English reading comprehension in grade 4. The results of this study suggest that not only cognitive variables such as executive functioning and metalinguistic awareness, but also non-verbal intelligence may be a good measure to describe cognitive benefits within the field of bilingual school programmes. 6 Acknowledgements The authors gratefully acknowledge the support received by the colleagues of the primary school in Tübingen, by the student assistants at the Frie‐ drich-Alexander-University of Erlangen-Nuremberg (Lea Pöschik) and by Nina Rogotzki for proofreading. All errors are, of course, solely attributable to 132 Anja Steinlen / Thorsten Piske <?page no="133"?> the authors. Most importantly, the children’s enthusiastic participation is greatly appreciated. This project was partially funded by the Faculty of Human‐ ities, Social Sciences and Theology of the Friedrich-Alexander-University of Erlangen-Nuremberg in 2013. References Adesope, O. O., Lavin, T., Thompson, T. & Ungerleider, C. 2010. A systematic review and meta-analysis of the cognitive correlates of bilingualism. Review of Educational Research, 80 (2), 207-245. doi: 10.3102/ 0034654310368803. Apsel, C. 2012. Coping With CLIL: Dropouts from CLIL streams in Germany. International CLIL Research Journal, 1 (4), 47-56. Ardila, A., Rosselli, M., Matute, E. & Inozemtseva, O. 2011. Gender differences in cognitive development. Developmental Psychology, 47 (4), 984-990. Bamford, K.W. & Mizokawa, D.T. 1991. Additive-bilingual (immersion) education: Cog‐ nitive and language development. Language Learning, 41 (3), 413-429. Barac, R., Bialystok, E., Castro, D.C. & Sanchez, M. 2014. The cognitive development of young dual language learners: A critical review. Early Childhood Research Quarterly, 29, 699-714. Baumert, J., Köller, O. & Lehmann, R. 2012. Leseverständnis im Englischen und Deut‐ schen und Mathematikleistungen bilingual unterrichteter Schülerinnen und Schüler am Ende der Grundschulzeit: Ergebnisse eines Zwei-Wege-Immersionsprogramms. Unterrichtswissenschaft, 40 (4), 290-314. Bialystok, E. 2001. Bilingualism in Development: Language, Literacy, and Cognition. New York, NY: Cambridge University Press. Bialystok, E. 2005. The impact of bilingualism on language and literacy development. In: Bhatia, T.K. & Ritchie, W.C. (Eds.), The Handbook of Bilingualism. Oxford, UK: Blackwell Publishing. 577-601. Bialystok, E. & Barac, R. 2012. Emerging bilingualism: Dissociating advantages for metalinguistic awareness and executive control. Cognition, 122, 67-73. Bialystok, E. & Martin, M.M. 2004. Attention and inhibition in bilingual children: Evidence from the dimensional change card sort task. Developmental Science, 7, 325-339. Bialystok, E., Peets, K.F. & Moreno, S. 2014. Producing bilinguals through immersion education: Development of metalinguistic awareness. Applied Psycholinguistics, 35, 177-191. Bleakley, H. & Chin, A. 2004. Language skills and earnings: Evidence from childhood immigrants. Review of Economics and Statistics, 86, 481-496. 133 “Does speaking a foreign language at school make your kids smarter? ” <?page no="134"?> Bracken, B.A. & McCallum, R.S. 1998. Universal Nonverbal Intelligence Test. Itasca, IL: Riverside Publishing. Brown, L., Sherbenou, R.J. & Johnsen, S.K. 1997. Test of Nonverbal Intelligence (TONI) - Third Edition. Austin, TX: Pro-Ed. Bruck, M., Lambert, W.E. & Tucker, G.R. 1973. Cognitive Consequences of Bilingual Schooling: The St. Lambert Project through Grade 6. Unpublished manuscript, Psy‐ chology Department, McGill University, Montreal. Bruck, M., Lambert, W.E. & Tucker, G.R. 1974. Bilingual schooling through the elementary grades: The St. Lambert Project at grade seven. Language Learning, 24, 183-204. Bulheller, S. & Häcker, H. 2010. Coloured Progressive Matrices mit der Parallelform des Tests und der Puzzle-Form. Manual zu Raven’s Progressive Matrices und Vocabulary Scales von J.C. Raven, J. Raven und J.H. Court. Frankfurt: Pearson. Bundesministerium für Bildung und Forschung. 2007. Förderung von Lesekompetenz - Expertise. Bildungsforschung Band 17. Bonn, Berlin: BMBF. Carlson, S. M. & Meltzoff, A.N. 2008. Bilingual experience and executive functioning in young children. Developmental Science, 11, 282-298. Chudaske, J. 2012. Sprache, Migration und schulfachliche Leistung: Einfluss sprachlicher Kompetenz auf Lese-, Rechtschreib- und Mathematikleistungen. Wiesbaden: VS-Verlag. Council of Europe. 2001. Common European Framework of Reference for Languages: Learning, Teaching and Assessment. Language Policy Unit, Strasbourg. http: / / www.co e.int/ t/ dg4/ linguistic/ source/ framework_en.pdf. May 31, 2021. Cummins, J. 1981. The role of primary language development in promoting educational success for language minority students. In: Schooling and Language Minority Students: A Theoretical Framework. Los Angeles, CA: California State University, Evaluation, Dissemination, and Assessment Center. 3-49. DeThorne, L.S. & Schaefer, B.A. 2004. Guide to child nonverbal IQ measures. American Journal of Speech-Language Pathology, 13, 275-290. Dickinson, D.K., Griffith, J.A., Golinkoff, R.M. & Hirsh-Pasek, K. 2012. How reading books fosters language development around the world. Child Development Research, doi: 10.1155/ 2012/ 602807. Draugalis, J.R., Coons, S.J. & Plaza, C.M. 2008. Best practices for survey research reports: a synopsis for authors and reviewers. American Journal of Pharmaceutical Education, 72 (1), 11. Fan, X. & Chen, M. 2001. Parental involvement and students’ academic achievement: A meta-analysis. Educational Psychology Review, 13 (1), 1-22. Fathi-Ashtiani, A. & Ahmadi, K. 2006. Verbal and nonverbal intelligence: dyslexic-dys‐ graphic students and normal students. Archives of Medical Science, 1, 42-46. Festman, J. & Kersten, K. 2010. Kognitive Auswirkungen von Zweisprachigkeit. In: Massler, U. & Burmeister, P. (Eds.), CLIL und Immersion: Fremdsprachlicher Sachfach- 134 Anja Steinlen / Thorsten Piske <?page no="135"?> unterricht in der Grundschule. Braunschweig: Westermann. 38-52. Festman, J., Rodriguez-Fornells, A. & Münte, T.F. 2010. Individual differences in control of language interference in late bilinguals are mainly related to general executive abilities. Behavioral and Brain Functions, 6 (5). www.behavioralandbrainfunctions.co m/ content/ 6/ 1/ 5. May 31, 2021. Fischer, U. 2007. Von Herausforderungen und Chancen: Was erwartet eine Schule bei der Einführung des Bilingualen Unterrichts? Grundschule, 39 (4), 32-35. FMKS, Verein für frühe Mehrsprachigkeit in Kindertageseinrichtungen und Schulen. 2014. Ranking: Bilinguale Kitas und Grundschulen im Bundesvergleich. www.fmks-on line.de/ download.html. May 31, 2021. Gamsjäger, E. & Sauer, J. 1996. Determinanten der Grundschulleistung und ihr prognos‐ tischer Wert für den Sekundarschulerfolg. Psychologie in Erziehung und Unterricht, 43 (2), 182-204. Gebauer, S.K., Zaunbauer, A. C.M. & Möller, J. 2012. Erstsprachliche Leistungsentwick‐ lung im Immersionsunterricht: Vorteile trotz Unterrichts in einer Fremdsprache? Zeitschrift für Pädagogische Psychologie, 3, 183-196. Gebauer, S.K., Zaunbauer, A. C.M. & Möller, J. 2013. Cross-language transfer in English immersion programmes in Germany: Reading comprehension and reading fluency. Contemporary Educational Psychology, 38, 64-74. Genesee, F.H. 1987. Learning Through Two Languages: Studies of Immersion and Bilingual Education. Cambridge, MA: Newbury House. Genesee, F.H. 2004. What do we know about bilingual education for majority language students? In: Bhatia, T.K. & Ritchie, W.C. (Eds.), Handbook of Bilingualism and Multiculturalism. Malden, MA: Blackwell. 547-561. Grosjean, F. 2014. Cognitive advantages of second language immersion education. http: / / www.psychologytoday.com/ blog/ life-bilingual/ 201401/ cognitive-advantages-secon d-language-immersion-education. May 31, 2021. Hakuta, K. & Diaz, R. 1985. The relationship between degree of bilingualism and cognitive ability: A critical discussion and some new longitudinal data. In: Nelson, K.E. (Ed.), Children’s Language, Volume 5. Hillsdale, NJ: Lawrence. 319-344. Heller, K.A., Kratzmeier, H. & Lengfelder, A. 1998. Matrizen-Test Manual Band 1. Ein Handbuch mit deutschen Normen zu den Standard Progressive Matrices von J. C. Raven. Göttingen: Beltz. Heppt, B., Stanat, P., Dragon, N., Berendes, K. & Weinert, S. 2014. Bildungssprachliche Anforderungen und Hörverstehen bei Kindern mit deutscher und nicht-deutscher Familiensprache. Zeitschrift für Pädagogische Psychologie, 28 (3), 139-149. Israel, N. 2006. Raven’s Advanced Progressive Matrices within a South African Con‐ text. Unpublished Master’s Thesis, University of the Witwatersrand, Johannes‐ burg. http: / / wiredspace.wits.ac.za/ bitstream/ handle/ 10539/ 1703/ RAPMResRepFinal2 135 “Does speaking a foreign language at school make your kids smarter? ” <?page no="136"?> 006NickyIsrael.pdf; jsessionid=D599D6F1D17A50D1226FB902697B9FD7? sequence=2. May 31, 2021. Kersten, K., Fischer, U., Burmeister, P., Lommel, A., Schelletter, C., Steinlen, A.K. & Thomas, S. 2010. Immersion in Primary School: A Guide. http: / / www.elias.bilikita.org / docs/ guidelines_for_bilingual_primary_schools_e.pdf. May 31, 2021. Kluever, R.C. & Green, K.E. 1994. An examination of ethnic and gender differences in the Raven Coloured Progressive Matrices Test. Paper presented at the Annual Meeting of the American Educational Research Association. New Orleans, LA. http: / / files.eric.e d.gov/ fulltext/ ED372085.pdf. May 31, 2021. Knowles, J. 2008. Assessment of Nonverbal Intelligence in South African Schools: Do Language and Gender Bias Performance on the Raven’s Standard Progressive Ma‐ trices? Unpublished Master’s Thesis, Johannesburg: University of the Witwaters‐ rand. wiredspace.wits.ac.za/ bitstream/ handle/ 10539/ 8238/ Microsoft_Word_-_jolene_ Corrected_Final_Draft_may09.pdf ? sequence=1. May 31, 2021. Kristiansen, I. 1990. Nonverbal Intelligence and Foreign Language Learning. Helsinki: Helsinki University Research Bulletin 73: Dissertation. Kultusministerkonferenz. 2013. Konzepte für den bilingualen Unterricht - Erfahrungsber‐ icht und Vorschläge zur Weiterentwicklung. http: / / www.kmk.org/ fileadmin/ veroeffent lichungen_beschluesse/ 2013/ 201_10_17-Konzepte-_bilingualer-_Unterricht.pdf. May 31, 2021. Kuschner, E.S. 2013. Nonverbal intelligence. In: Volkmar, F.R. (Ed.), Encyclopaedia of Autism Spectrum Disorder. New York, NY: Springer, 2037-2041. Kuska, S.K., Zaunbauer, A. C.M. & Möller, J. 2010. Sind Immersionsschüler wirklich leistungsstärker? Ein Lernexperiment. Zeitschrift für Entwicklungspsychologie und Pädagogische Psychologie, 42, 143-153. Lambert, W.E. & MacNamara, J. 1969. Some cognitive consequences of following a first-grade curriculum in a second language. Journal of Educational Psychology, 60, 86-96. Lambert, W.E. & Tucker, G.R. 1972. Bilingual Education of Children: The St. Lambert Experiment. Rowley, MT: Newbury House. Lambert, W.E., Tucker, G.R. & D’Anglejan, A. 1973. Cognitive and attitudinal conse‐ quences of bilingual schooling: The St. Lambert project through grade 5. Journal of Educational Psychology, 65 (2), 141-159. Lassiter, K.S. & Bardos, A.N. 1995. The relationship between young children’s academic achievement and measures of intelligence. Psychology in the Schools, 32, 170-177. Lee, P. 1996. Cognitive development in bilingual children: A case for bilingual instruction in early childhood education. Bilingual Research Journal, 20, 499-522. Lenhard, W. & Schneider, W. 2006. ELFE 1-6: Ein Leseverständnistest für Erstbis Sechst‐ klässler. Göttingen: Hogrefe. 136 Anja Steinlen / Thorsten Piske <?page no="137"?> Little, D., Simpson, B.L. & Čatibusič, B.F. 2003. PSAK: Primary School Assessment Kit. Dublin: Integrate Ireland Language and Training (IILT). Lynn, R. & Irwing, P. 2004. Sex differences on the progressive matrices: A meta-analysis. Intelligence, 32, 481-498. Lynn, R., Allik, J. & Irwing, P. 2004. Sex differences on the three factors identified in Raven’s Standard Progressive Matrices. Intelligence, 32, 411-424. MacCoubrey, S., Wade-Woolley, L., Klinger, D. & Kirby, J.R. 2004. Early identification of at-risk L2 readers. Canadian Modern Language Review, 61, 11-28. Melzer, J., Rißling, J-K. & Petermann, F. 2015. Einfluss der nonverbalen Intelligenz und des Migrationshintergrundes auf die sprachlichen Fähigkeiten. Das Gesundheitswesen, 77 (10), 793-798. Ministerium für Kultus, Jugend und Sport Baden-Württemberg. 2004. Bildungsstandards für Englisch Grundschule: Klassen 2, 4. www.bildungsplaene-bw.de/ site/ bildungsplan/ get/ documents/ lsbw/ Bildungsplaene/ Bildungsplaene-2004/ Bildungsstandards/ Grundschule_Bildungsplan_Gesamt.pdf. May 31, 2021. Ministerium für Kultus, Jugend und Sport Baden-Württemberg. 2016. Bildungsstandards für Englisch Grundschule: Klassen 2, 4. https: / / km-bw.de/ ,Lde/ startseite/ schule/ Bildun gsplan+2016+Grundschule. May 31. 2021. Morvay, G. 2015. The role of L1 reading ability, L2 proficiency and nonverbal intelligence in L2 reading comprehension. Argentinian Journal of Applied Linguistics, 3 (1), 26-47. Mullis, I.V.S, Martin, M.O., Foy, P. & Drucker, K.T. 2012. PIRLS 2011 International Results in Reading. Chestnut Hill, MA: TIMSS & PIRLS International Study Center. Mushquash, C.J. & Bova, D.L. 2007. Cross-cultural assessment and measurement issues. Journal on Developmental Disabilities, 13, 53-66. Naglieri, J.A. 1985. Matrix Analogies Test—Expanded Form. San Antonio, TX: The Psycho‐ logical Corporation. Naglieri, J.A. 2001. Do ability and reading achievement correlate? Journal of Learning Disabilities, 34 (4), 304-305. Nation, K. & Snowling, M.J. 1998. Semantic processing and the development of word recognition skills: Evidence from children with reading comprehension difficulties. Journal of Memory and Language, 39, 85-101. Nehamas, N. 2012. Does not speaking English at school make your kids smarter? Latitude News. http: / / www.latitudenews.com/ story/ does-not-speaking-english-at-school-mak e-your-kids-smarter/ . May 31, 2021. Nicolay, A.-C. & Poncelet, M. 2013. Cognitive advantage in children enrolled in a second-language immersion elementary school program for three years. Bilingualism: Language and Cognition, 16 (3), 597-607. 137 “Does speaking a foreign language at school make your kids smarter? ” <?page no="138"?> Nicolay, A.-C. & Poncelet, M. 2015. Cognitive benefits in children enrolled in an early bilingual immersion school: A follow up study. Bilingualism: Language and Cognition, 18 (4), 789-795. Peal, E. & Lambert, W.E. 1962. The relation of bilingualism to intelligence. Psychological Monographs, 76 (27), 1-23. Raven, J.C. 1976a. Coloured Progressive Matrices. Sets A, AB & B. Prepared by JC Raven. San Antonio, TX: Harcourt. Raven, J.C. 1976b. Standard Progressive Matrices. Sets A, B, C, D & E. Prepared by JC Raven. San Antonio, TX: Harcourt. Samuels, D.D. & Griffore, R.J. 1979. The Plattsburgh French language immersion program: Its influence on intelligence and self-esteem. Language Learning, 29, 45-52. Serpell, R. 1994. The cultural construction of intelligence. In: Lonner, W.J. & Malpass, R. (Eds.), Psychology and Culture. Boston, MA: Allyn and Bacon. 157-163. Springer, M. 2006. Mehrsprachigkeit macht schlau. Gehirn & Geist 2006, http: / / www.spe ktrum.de/ magazin/ mehrsprachigkeit-macht-schlau/ 848984. May 31, 2021. Stanovich, K., Cunningham, A.E. & Feeman, D.J. 1984. Intelligence, cognitive skills, and early reading progress. Reading Research Quarterly, 19 (3), 278-303. Statistisches Bundesamt. 2020. Bevölkerung mit Migrationshintergrund: Ergebnisse des Mikrozensus 2019, Fachserie 1, Reihe 2.2. Wiesbaden: Destatis. https: / / www.destatis.de/ DE/ Themen/ Gesellschaft-Umwelt/ Bevoelkerung/ Migration -Integration/ Publikationen/ _publikationen-innen-migrationshintergrund.html; jsessi onid=AAC2F6503C501D32338529F38C3E2217.internet8722. May 31, 2021. Steinlen, A.K. 2016. Primary school minority and majority language children in a partial immersion program: The development of German and English reading skills. Journal of Immersion and Content-Based Language Education, 4 (2), 198-224. Steinlen, A.K. 2017. The development of English grammar and reading comprehension by majority and minority language children in a bilingual primary school. Studies in Second Language Learning and Teaching, 7 (3), 419-442. Steinlen, A.K. 2018a. Grundschulen mit verschiedenen Englischprogrammen: Geschlecht und Migration auf dem Prüfstand. Zeitschrift für Fremdsprachenforschung, 29 (1), 3-26. Steinlen, A.K. 2018b. The development of English and German writing skills in a bilingual primary school in Germany. Journal of Second Language Writing, 39 (1), 42-52. Steinlen, A.K. 2021. English in Elementary Schools: Research and Implications on Minority and Majority Language Children’s Literacy Skills in Regular and Bilingual Programs. Tübingen: Narr. Steinlen, A.K. & Gerdes, U. 2015. Eine Pilotuntersuchung zur sprachlichen Entwicklung von Kindern mit und ohne Migrationshintergrund in einer Grundschule mit bilin‐ gualem Sachfachunterricht. In: Méron-Minuth, S. & Özkul, S. (Eds.), Fremde Sprachen 138 Anja Steinlen / Thorsten Piske <?page no="139"?> lehren und lernen - Aktuelle Fragen und Forschungsaufgaben. Frankfurt am Main: Peter Lang. 81-106. Steinlen, A.K. & Piske, T. 2013. Academic achievement of children with and without migration backgrounds in an immersion primary school: A pilot study. ZAA Zeitschrift für Anglistik und Amerikanistik, 61 (3), 215-244. Steinlen, A.K. & Piske, T. 2014. Deutsche und englische Lesefähigkeiten von Kindern mit und ohne Migrationshintergrund in einer bilingualen Grundschule: Eine Pilotstudie. In: Bürgel, C. & Siepmann, D. (Eds.), Sprachwissenschaft und Fremdsprachenunterricht: Spracherwerb und Sprachkompetenzen im Fokus. Baltmannsweiler: Schneider Verlag Hohengehren. 141-165. Steinlen, A.K. & Piske, T. 2015. Zur Entwicklung der Schulleistungen von Kindern mit und ohne Migrationshintergrund in einer bilingualen Grundschule: Eine Pilotstudie. In: Kötter, M. & Rymarczyk, J. (Eds.), Englischunterricht auf der Primarstufe. Neue Forschungen - weitere Entwicklungen. Frankfurt am Main: Peter Lang. 123-150. Steinlen, A.K. & Piske, T. 2016. Wortschatz und Leseverständnis des Englischen von Kindern mit und ohne Migrationshintergrund in einer bilingualen Grundschule. In: Steinlen, A.K. & Piske, T. (Eds.), Wortschatzlernen in bilingualen Schulen und Kindertagesstätten. Forum Angewandte Linguistik. Frankfurt am Main: Peter Lang. 123-166. Steinlen, A.K. & Piske, T. 2018a. Deutsch- und Englischleistungen von Kindern mit und ohne Migrationshintergrund im bilingualen Unterricht und im Fremdsprachenunter‐ richt: Ein Vergleich. In: Ballis, A. & Hodaie, N. (Eds.), Perspektiven auf Mehrsprachig‐ keit: Individuum - Bildung - Gesellschaft. Berlin: Mouton De Gruyter. 85-97. Steinlen, A.K. & Piske, T. 2018b. Die Entwicklung des englischen und deutschen Lesens bei ein- und mehrsprachigen Kindern in Grundschulen mit bilingualen Angeboten. In: Kutzelmann, S. & Massler, U. (Eds.), Mehrsprachige Leseförderung: Grundlagen und Konzepte. Tübingen: Narr. 51-62. Swain, M. & Lapkin, S. 1982. Evaluating Bilingual Education: A Canadian Case Study. Clevedon: Multilingual Matters. Tamm, C. 2010. Eine Schule macht sich auf den Weg: Einführung eines bilingualen Zuges an der Grundschule an der Hügelstraße. In: Massler, U. & Burmeister, P. (Eds.), CLIL und Immersion. Fremdsprachlicher Sachfachunterricht in der Grundschule. Braunschweig: Westermann. 30-37. Trites, R.L. & Price, M.A. 1978. Assessment of Readiness for Primary French Immersion. Toronto, Canada: Ontario Ministry of Education. Trites, R.L. & Price, M.A. 1980. Assessment of Readiness for Primary French Immersion: Grade One Follow-up Assessment. Toronto, Canada: Ontario Ministry of Education. 139 “Does speaking a foreign language at school make your kids smarter? ” <?page no="140"?> Valencia, R.R. 1979. Comparison of intellectual performance of Chicano and Anglo third-grade boys on the Raven’s Colored Progressive Matrices. Psychology in the Schools, 16, 448-453. von Stumm, S., Hell, B. & Chamorro-Premuzic, T. 2011. The hungry mind: Intellectual curiosity is the third pillar of academic performance. Perspective on Psychological Science, 6 (6), 574-588. Wechsler, D. 2003. Wechsler Intelligence Scale for Children - Fourth Edition. San Antonio, TX: Psychological Corporation. Wesche, M.B. 2002. Early French immersion: How has the original Canadian model stood the test of time? In: Burmeister, P., Piske, T. & Rohde, A. (Eds.), An Integrated View of Language Development. Papers in Honor of Henning Wode. Trier: WVT. 357-379. Wode, H. 1994. Speech perception and the learnability of languages. International Journal of Applied Linguistics, 4 (2), 143-196. Wode, H. 2009. Frühes Fremdsprachenlernen in bilingualen Kindergärten und Grund‐ schulen. Braunschweig: Westermann. Yadollahi Jouybari H., Piske, T. & Steinlen, A. 2020. Effects of bilingual programs and digital tools on minority language children’s and adult refugees’ emerging L2 and L3 literacy skills. In: Eisenmann, M. & Steinbock, J. (Eds.), Sprache, Kulturen, Identitäten: Umbrüche durch Digitalisierung? (Beiträge zur Fremdsprachenforschung, Band 15). Baltmannsweiler: Schneider Verlag Hohengehren. 133-145. Zaunbauer, A. C.M. & Möller, J. 2006. Schriftsprachliche und mathematische Leistungen in der Erstsprache: Ein Vergleich monolingual und teilimmersiv unterrichteter Kinder der zweiten und dritten Klassenstufe. Zeitschrift für Fremdsprachenforschung, 17, 181-200. Zaunbauer, A. C.M. & Möller, J. 2007. Schulleistungen monolingual und immersiv unter‐ richteter Kinder am Ende des ersten Schuljahres. Zeitschrift für Entwicklungspsycho‐ logie und Pädagogische Psychologie, 39, 141-153. Zaunbauer, A. C.M. & Möller, J. 2010. Schulleistungsentwicklung immersiv unterrichteter Grundschüler in den ersten zwei Schuljahren. Psychologie in Erziehung und Unterricht, 57, 30-45. Zaunbauer, A. C.M., Bonerad, E-M. & Möller, J. 2005. Muttersprachliches Leseverständnis immersiv unterrichteter Kinder. Zeitschrift für Pädagogische Psychologie, 19, 233-235. Zaunbauer, A. C.M., Gebauer, S.K. & Möller, J. 2012. Englischleistungen immersiv un‐ terrichteter Schülerinnen und Schüler. Unterrichtswissenschaft: Zeitschrift für Lernfor‐ schung, 40 (4), 315-333. Zydatiß, W. 2007. Deutsch-Englische Züge in Berlin (DEZIBEL). Eine Evaluation des bilingualen Sachfachunterrichts an Gymnasien: Kontext, Kompetenzen, Konsequenzen. Frankfurt am Main: Peter Lang. 140 Anja Steinlen / Thorsten Piske <?page no="141"?> Competence development in multilingual primary school classes in Germany: Linguistic competence and basic cognitive skills Jana Chudaske (translated by Nina Rogotzki) Abstract In Germany, various large-scale studies conducted to assess academic achievement (e.g. PIRLS, PISA, etc.) have shown significant differences between children and adolescents with and without a migration back‐ ground; and proficiency in German (as the language of instruction) has repeatedly been identified as one of the main predictors of academic achievement. In addition to the command of the language of instruction (e.g. German in Germany), basic cognitive skills also predict such out‐ comes, which are important prerequisites for successful first and second language acquisition. The focus of the present study is on 211 children in grade 3 and their performance in various German language tests and in cognitive tests. These data were collected as part of the KEIMS project (Kompetenzentwicklung in multilingualen Schulklassen). The results re‐ veal significant differences between children with and without a migration background in standardised German language tests and in terms of their teachers’ assessments of their German language proficiency. However, such language background effects disappeared once basic cognitive skills were controlled for. These results clearly indicate the need to disentangle effects of variables such as migrant, linguistic and cognitive background factors to better foster migrant children’s academic success in school and in particular their German language skills. <?page no="142"?> 1 Introduction When they were first published, the results of international school performance studies like TIMSS, PISA and PIRLS, which assessed academic skills in reading, mathematics and science as central factors for successful education, triggered an intense debate about the fairness of the German school system. This is the case because success at school is is to a considerable extent still influenced by regional, social and family factors as well as by ethnic origin. For example, pro‐ portionally fewer children with a migration background than children without a migration background attend grammar schools (Gymnasien) in Germany. Such educational outcomes are, to a large degree, determined by the command of the language of instruction (in this case German in Germany), for which, in turn, basic cognitive skills are essential. This chapter presents results of a project called KEIMS (Kompetenzentwicklung in multilingualen Schulklassen/ compe‐ tence development in multilingual classes) which was conducted in primary schools in Germany. The focus is on data of 211 children in grade 3 and their performance in various German language tests and cognitive tests. In the following, the notion of linguistic competence, some diagnostic tools to assess language competence (i.e. tests and teacher assessments), related cognitive skills and our research questions are introduced in sections 2, 3 and 4. In sections 5 and 6 the KEIMS study and its outcomes for third graders are presented. This chapter concludes with suggestions for further research (section 7). 2 Linguistic competence 2.1 Definitions In Hartig’s view (2008), the concept of competence is difficult to define due to its complexity in meaning. Generally, the concept of competence refers to the abilities of individuals in learning processes and the objectives they are to achieve. Having gained popularity in the field of linguistics in the 1960s (cf. Weinert, 2001), the concept of competence was first established in Germany in educational science at the beginning of the 1970s by Heinrich Roth, who included affective and motivational facets in addition to cognitive performance dispositions (cf. Klieme & Hartig, 2007). A theoretical framework for the classification of competences does not exist, instead multiple concepts of competence are nowadays applied, which cover a wide range of abilities, skills, and personal characteristics (i.e. professional competence, methodological com‐ petence, and individual competence). According to Hartig (2008), measurements of competence in educational science require a series of individual observations 142 Jana Chudaske (translated by Nina Rogotzki) <?page no="143"?> of different situations, or rather of different tasks. In this context, Ehlich (2005b: 11) defines linguistic competence as consisting of different linguistic abilities which entail phonic, pragmatic, semantic, morphologic-syntactic, discursive, and literal basic skills, all of which evolve parallel, influence each other reciprocally, and make up language proficiency as well as linguistic competence. 2.2 Diagnostics of language competence: Language tests and teacher assessments Language proficiency assessment is described by Ehlich (2005a: 36) as an evaluation of linguistic abilities at a special point in time, geared towards a socially as well as scientifically accepted “assumption of normality”, which entails a certain basic comprehension of age-appropriate phases of language acquisition and their specific characteristics. Professional language proficiency diagnostics employ adequate measuring procedures that take into account linguistic and psychometric aspects (Ehlich, Bredel & Reich, 2008). Language proficiency assessment procedures can be categorised according to various criteria: Ehlich (2005b) and Reich (2005), for example, adopt a classi‐ fication based on the type of procedure by differentiating between estimation procedures, observations, profile analyses, and tests, the latter of which are most commonly used when assessing language proficiency. By measuring linguistic skills in a standardised manner, the test performances of different children become comparable on the basis of pre-established standard values. Ehlich (2005b) also categorises the instruments of measurement according to the subjects’ age. Unfortunately, only very few procedures exist with regard to children of school age. Language acquisition at primary school-age seems to be less well researched than the early stages of language acquisition: The focus of most assessment batteries is on pre-school age, i. e. on children from three years old to school entry age. Although Neugebauer & Becker-Mrotzek (2013: 4) see progress in this context, they nevertheless see a need for further research: For example, more than half of the procedures analysed in their sample met merely 16 or even fewer of the 32 quality criteria developed for assessing test method performance; particularly linguistic basic skills, as well as validity, objectivity, and multilingualism, were rated unfavourably (for further details see Fried, 2004; Metz, Fröhlich & Petermann, 2009; Schnieders & Komor, 2005; Weinert, Doil & Frevert, 2008). With respect to language proficiency assessments of German as a first language (L1), most assessments are available for pre-schoolers, whereas the age from primary to secondary school has been given less consideration. The main problem with these assessments relates to students with a migration 143 Competence development in multilingual primary school classes in Germany <?page no="144"?> background (MB). They constitute a population with an increased need for language support, which makes the adequate language proficiency assessment of German as a second language (L2) all the more important. However, the fairness of the assessment of German language skills (or rather, the lack thereof) poses a problem when assessing German proficiency of MB students. This is due to the fact that language proficiency assessment procedures are generally not designed to include multilingualism (Ehlich, 2005b). With regard to the diagnostic procedures employed to test MB children, Schölmerich, Leyendecker, Citlak, Caspar & Jäkel (2008: 192) state that when used “cautiously and with sufficient attention to detail (like item analysis …), many instruments which are standardised for the general population can also be used with migrant and minority children” (see Schroeder & Stölting, 2005 for a similar view). List (2005: 53), however, rejects the use of such assessments, because in his view it is unfair to transfer norm values obtained for children for whom German is the L1 to children for whom German is the L2. While developmental delays constitute a central part of these assessments for L1 German, the assessment of German as an L2 is only concerned with the “regular” acquisition of the L2, disregarding, for example, any developmental language delays or disorders of bilingual children. Therefore, Reich (2008) sees a need to develop language proficiency assessments which are designed for both L1 and L2 acquisition. Linguistic abilities may also be evaluated by a third party: For ex‐ ample, various large-scale primary school studies in Germany, e. g. BeLesen (Schründer-Lenzen & Merkens, 2006) or SOKKE (Sozialisation und Akkultura‐ tion in Erfahrungsräumen von Kindern mit Migrationshintergrund - Schule und Familie, Herwartz-Emden & Küffner, 2006), utilised teachers’ assessments of students, i. e. informal assessments that are obtained incidentally and unsys‐ tematically in day-to-day school life. The accuracy of these evaluations relies heavily on the respective evaluator’s diagnostic competence, which, however, plays a major role in teachers’ day-to-day teaching instructions anyway with regard to pedagogical decisions, lesson design, and individual learning support (see also Arnold & Richert, 2009). Unfortunately, only a few empirical studies have yet examined the assessment of linguistic competence by teachers: For example, in the longitudinal study “BeLesen”, Schründer-Lenzen & Merkens (2006) examined 1200 primary school students’ academic achievements in Berlin, 70 % of them had a migration background. In addition to standardised achievement tests for various school subjects, teachers’ assessments regarding the students’ German language pro‐ ficiency were conducted in grade 1. In grade 3, C-Tests were employed to test language comprehension. In this context, Merkens (2005: 55) examined the 144 Jana Chudaske (translated by Nina Rogotzki) <?page no="145"?> relationship between the results of the C-Tests and the teachers’ assessments and found high correlations. In the longitudinal study SOKKE (see above), the focus was on the acculturation process of MB children throughout the four years of primary school. The degree of acculturation was defined by the child’s success in school, which in turn was measured by means of standardised academic achievement tests and teachers’ assessments of MB students’ German language skills. One of the findings showed that MB students, whose German language skills were rated best, had indicated German to be their dominant family language (Herwartz-Emden & Küffner, 2006). In yet another project called BiKS (“Bildungsprozesse, Kompetenzentwicklung und Selektionsentscheidungen im Vor- und Grundschulalter”, Lorenz & Artelt, 2009), inter-rater reliability of students’ achievements and teachers’ assessments regarding vocabulary and text comprehension skills was evaluated, and similar results were reported, i. e. high correlations between teachers’ assessment and language tests. However, teacher assessments are not without limitations: For example, Spinath (2005) speculates that the accuracy of teachers’ assessments may depend on students’ emotional and motivational characteristics (e.g. anxiety, motivation or self-concept) and the difficulty of the tasks. Furthermore, teachers are likely to base their assessment of each individual student’s achievements on the overall achievements of the class (cf. Ingenkamp & Lissmann, 2008). Such limitations in the assessment are, however, beyond the scope of this study. 3 Basic cognitive skills (intelligence and phonological working memory) and their relation to language acquisition A multitude of definitions of the concept of intelligence is mentioned in many studies - albeit none is generally agreed upon. Rost (2009), for example, describes intelligence as a cognitive capacity defined by genetic and environmental factors as well as their interdependencies. Wechsler, in turn, defines intelligence as the ability of an individual to act purposefully, to think rationally, and to deal effectively with his/ her environment (in Wild, Hofer & Pekrun, 2001). Klauer (2006) differentiates between factor-analytic models of intelligence (e.g. Spear‐ man’s construct of general intelligence (g factor); Thurstone’s model of primary mental abilities; Cattell’s concept of crystallised versus fluid intelligence) and cognitive models of intelligence (e.g. information processing models; working memory capacity, see Rost, 2009 for a review). Various empirical studies have indicated a relationship between intelligence and linguistic competence. Rost (2009) notes that intelligence is a precise predictor of academic achievement but points out that such a relationship 145 Competence development in multilingual primary school classes in Germany <?page no="146"?> is stronger for cognitive tests focusing on verbal intelligence than those on non-verbal intelligence. Furthermore, Brunner (in Rost, 2009) reports that general intelligence may explain up to 50 % of variance regarding verbal skills. In studies by Herrnstein & Murray and McGrew & Knopik (in Rost, 2009), correlations of r = .70 and r = .76 were noted between intelligence and oral and written language proficiency, respectively (see Rost, 2009 for a review). Similarly, in his study on the impact of the learning environments of parental home and kindergarten on the development of language (vocabulary, grammar, communication skills), Wolf (1987) observed correlations between intelligence and language proficiency, whose strength, however, depended on the skill in question (see also Degel, 1987). In yet another study, Marx (2006) examined the relation between language acquisition and inductive reasoning, i. e. the ability to detect regularities based on processes of comparison, which is a key component of human intelligence. Tasks devised to promote inductive strategising were able to improve the language skills of six-year-old children with regard to morphology and semantics in short term as well as in the longer run. Similarly, a study by Schiffer, Ennemoser & Schneider (2002) on the effects of television consumption on language development and reading skills of primary school students showed significant positive correlations with regard to intelligence test scores and language/ reading competence. That is, children with lower intelligence achieved lower scores in the language tests. Cognitivist approaches emphasise the importance of information processing and working memory (e.g. Grimm & Weinert, 2002; Klauer, 2006): For example, phonological working memory (which processes linguistic input and whose capacity is often gauged by means of memory span) is not only closely related to general cognitive achievements but also to language acquisition processes, especially regarding vocabulary and grammar. According to many authors (e.g. Hasselhorn & Werner, 2000), successful recalling of nonsense words (i.e. memory span) seems to be a good indicator of a functioning phonological working memory and surmise that these may have a great impact on speech development disorders, which are, however, not the focus of the present study. 4 Research questions Based on the literature review, the present study focuses on the following research questions: 1. Do 3 rd grade children with and without migration background differ with respect to their linguistic achievements in German in (a) standardised language tests and (b) teacher assessments? 146 Jana Chudaske (translated by Nina Rogotzki) <?page no="147"?> 2. Do possible effects of migration background on the linguistic abilities in German change when basic cognitive skills are controlled for? 5 Study design 5.1 The KEIMS project This study is part of a longitudinal research project called “Kompetenzentwick‐ lung in multilingualen Schulklassen” (KEIMS), which was conducted at the Institute of Educational Science, Department of Applied Educational Science, at the University of Hildesheim, Germany, from 2004 to 2010. The aim of the project was to investigate the development of social competence in relation to academic achievement throughout primary school, with a particular focus on the different degrees of multilingualism in the classes. Basic cognitive skills and joy for learning were assessed as prerequisites for learning and achievement, and, additionally, the primary school students’ academic achievements with regard to reading skills, orthography, and arithmetics as well as various aspects of social competence (prosocial behaviour, social information processing, self-concept of social competence, sociometry) and language proficiency were measured. Altogether more than 1300 primary school children in more than 60 classes took part in this project. 5.2 Study design and sample The following findings and analyses of the present study pertain to the data collected in 2006 (Arnold, Lindner-Müller, Hentschel & Chudaske, 2007), at the third point of measurement of the KEIMS project. With the support of the Ministry of Education and Cultural Affairs of Lower Saxony, six primary schools, with two classes respectively, were recruited in 2004 (first point of measurement, grade 1). The schools were chosen because many children in these classes had migration backgrounds and/ or were multilingual. The sample of this study comprises 211 third grade students from twelve classes of six primary schools in the Hildesheim/ Gifhorn/ Braunschweig area in Northern Germany. The average age of the children was 9.45 years. The gender ratio is relatively balanced with 115 girls (55 %) and 96 boys (45 %). 42 % of the children in this sample had a migration background, which was defined based on a procedure employed by the city of Bremen (Germany), in line with the PISA data (Husfeldt, Arnold, Möser & Brümmer, 2004). The following criteria were taken into consideration: 147 Competence development in multilingual primary school classes in Germany <?page no="148"?> 1. Family language: German is not spoken at home, or both German and (a) language(s) other than German are spoken at home. 2. Country of birth of the mother: The mother’s country of birth is not Germany. Thus, the children with a migration background were those who either did not speak German at home or who spoke German and another language at home and whose mothers were born outside Germany (Arnold et al., 2007). In the statistical analyses, migration status was, therefore, employed as a qualitative variable with two distinct, mutually exclusive classes (children with and without migration background, abbreviated as MB, see above). This information was provided in a students’ questionnaire, whereas information on the mother’s country of birth was collected in a teachers’ questionnaire (see below). 5.3 Methodology and instruments The variable basic cognitive skills was operationalised by the scores obtained in the Grundintelligenztest Skala 1 (CFT-1, Weiß & Osterland, 1997). This standar‐ dised test battery operates relatively language-free and is as such comparatively culture-fair, which is an important aspect to give equal opportunity to children with and without migration backgrounds. The test consists of five subtests (substitutions, mazes, classifications, similarities and matrices) to determine children’s basic intelligence, i. e. the children’s ability to recognise rules, identify characteristics and perceive them quickly. Linguistic competence was assessed using German tests and, in addition, teachers’ assessments. Three language proficiency tests were employed: (a) The “Heidelberger Sprachentwicklungstest” (HSET, Grimm & Schöler, 1991) is a test geared at various linguistic abilities. The focus of this study was on the three subtests “Formation of derivational morphemes” (DM), “Syntax” (S) and “Word selection” (WS). (b) The subtest “Language comprehension” (LC) is part of the Allgemeiner Schulleistungstest für dritte Klassen (AST 3, Fippinger, 1991). It is a language-related academic achievement test and assesses sentence-related and word-related learning content (e.g. recognition of word families and adjectives), which should have been mastered by the second semester of grade three. (c) The “Sprachstandsüberprüfung und Förderdiagnostik für Ausländer- und Aussiedlerkinder” (SFD, Hobusch, Lutz & Wiest, 2002) is a diagnostic in‐ strument usually employed in language training programmes. It was used because it provides separate norm values for children with and without 148 Jana Chudaske (translated by Nina Rogotzki) <?page no="149"?> MB. For this study, the subtests “Vocabulary” (V) and “Prepositions” (PP) were selected because vocabulary is important for the participation in academic as well as extra-curricular life whereas prepositions are pivotal in day-to-day school life, particularly with respect to the appropriate completion of tasks. The teachers’ assessments of the students’ language proficiency were coded by using a Likert rating scale, which was part of a teachers’ questionnaire. Following the methods adopted by the projects “BeLesen” (Merkens, 2005) and “em-soz” (Preuss-Lausitz, 2005), the teachers were asked to evaluate language comprehension, language production, command of grammar, and language behaviour towards classmates for each student on a four-point scale (ranging from “fully adequate” to “insufficient”). A raw score, summarising the scores for each item for each student, provided the basis for analysing teachers’ evaluation of children’s language proficiency in German. 5.4 Collection of data and statistical procedures Data collection took place in the spring of 2006. The language proficiency data were collected on the first day, starting with the SFD 3/ 4 and the AST 3, which were conducted as group tests during class, followed by the subtests of the HSET, which were carried out individually. The assignment of the students’ questionnaires (providing information about the student’s family language/ s), as well as the cognitive test (Grundintelligenztest Skala 1), took place on the following days of testing. The teachers received their questionnaires at the beginning of the survey and were asked to complete them by the fourth and final day of testing. The results of the language proficiency tests and the teachers’ assessments will be presented by using a “yes/ no” approach with respect to the variable “migration background”. Using the statistical software SPSS multivariate, single-factor variance analyses were computed. Subsequently, the control variable (covariate) “basic cognitive skills” as a learning and performance prerequisite was factored in. 6 Results Table 1 shows the results of multivariate variance analyses, presenting the results of the three language proficiency tests (including the subtests) for students with and without MB separately. In addition, the respective F-values, their significance, and the corresponding effect sizes are reported. 149 Competence development in multilingual primary school classes in Germany <?page no="150"?> HSET AST 3 SFD 3/ 4 DM S WS LC V PP MB N M SD M SD M SD M SD M SD M SD yes 87 45.2 (6.8) 44.2 (7.4) 46.2 (6.4) 50.1 (7.3) 12.8 (2.6) 6.9 (2.5) no 122 50.9 (9.1) 48.9 (8.8) 49.9 (6.8) 51.7 (8.6) 14.4 (2.1) 8.3 (1.9) total 209 48.6 (8.7) 46.9 (8.5) 48.3 (6.9) 51.0 (8.1) 13.8 (2.5) 7.8 (2.3) F 25.2 17.5 15.8 1.9 24.5 19.2 α =.05 P .000 .000 .000 .17 .000 .000 Ƞ 2 .11 .08 .07 .01 .11 .09 Table 1: Results for three German language tests. HSET results (T-value mean for the subtests DM = derivational morphemes, S = syntax, and WS = word selection), AST results (T-value mean for the subtests LC = language comprehension), and SFD results (raw score mean for the subtests V = vocabulary and PP = prepositions), listed separately by MB yes/ no (N = sample size, M = mean, and SD = standard deviation) as well as F-values (test statistics), p-values (statistical significance), and ƞ 2 (effect size). As Table 1 shows, students without MB consistently outperformed students with MB in all language tests. Almost all of the mean differences in the subtests were statistically significant, with the exception of the subtest “language comprehension” of the AST. Furthermore, effect sizes are largest for the subtest DM (derivational mor‐ phology) of HSET and V (vocabulary) of the SFD with an Eta value of ƞ 2 = .11. Thus, the MB variable accounts for 11 % of the variance in these tests. In the other tests, MB plays a smaller role: For example, MB affects performance in the subtest “prepositions” to 9 % (ƞ 2 = .09), in the subtests “syntax” to 8 % (ƞ 2 = .08), “word selection” to 7 % (ƞ 2 = .07), and “language comprehension” to only 1 % (ƞ 2 = .01). In Table 2 the results of the multivariate variance analysis regarding teachers’ assessments of students’ German language proficiency (subdivided into five linguistic categories) are shown for students with and without MB. 150 Jana Chudaske (translated by Nina Rogotzki) <?page no="151"?> Categories LC LP GRAM IPC OA MB N M SD M SD M SD M SD M SD yes 84 3.4 (0.6) 3.3 (0.6) 2.8 (0.8) 3.6 (0.5) 3.2 (0.7) no 117 3.7 (0.5) 3.7 (0.5) 3.4 (0.8) 3.8 (0.4) 3.6 (0.7) total 201 3.6 (0.6) 3.5 (0.6) 3.1 (0.8) 3.7 (0.4) 3.4 (0.7) F 11.0 24.1 25.4 15.2 18.3 α =.05 p .001 .000 .000 .000 .000 Ƞ 2 .05 .11 .11 .07 .08 Table 2: Results of teachers’ assessments of language abilities with regard to the catego‐ ries language comprehension (LC), language production (LP), command of grammar (GRAM), interpersonal communication (IPC), and overall assessment (OA), listed sepa‐ rately by MB yes/ no (N = sample size, M = mean, and SD = standard deviation) as well as F-values (test statistics), p-values (statistical significance), and ƞ 2 (effect size). The results in Table 2 illustrate that teachers assess the language proficiency of students with MB to be considerably lower than that of their peers without MB, irrespective of the linguistic category in question. All group differences are statistically significant. An inspection of the effect sizes indicates that the variable MB accounts for 11 % of the variance for LP (language production) and GRA (command of grammar) as assessed by the teachers. For the other three categories, i. e. language comprehension, interpersonal communication, and overall assessment of language proficiency, MB accounts for 5 % to 8 % of the variance. In the following, the impact of basic cognitive skills on the relationship between migration background and linguistic competence is examined, using covariance analysis (Table 3). For purposes of comparison, the results on “MB yes/ no” subsumes the values of the covariance analysis as well as the results of the simple variance analysis, which did not include the control variable (in parentheses). 151 Competence development in multilingual primary school classes in Germany <?page no="152"?> Variables AV F (1.207) p (α =.05) Ƞ 2 MB yes/ no Basic cognitive skills Vocabulary 12.71 27.27 .000 .000 .06 (.11) .12 MB yes/ no Basic cognitive skills Prepositions 12.17 45.39 .001 .000 .06 (.09) .18 MB yes/ no Basic cognitive skills Language comprehen‐ sion 0.33 28.20 .56 .000 .002 (.01) .12 MB yes/ no Basic cognitive skills Derivational morphemes 16.31 16.23 .000 .000 .07 (.11) .07 MB yes/ no Basic cognitive skills Syntax 10.10 27.87 .002 .000 .05 (.08) .12 MB yes/ no Basic cognitive skills Word selection 8.27 16.41 .004 .000 .04 (.07) .08 Table 3: F-values (test statistics), p-values (statistical significance), and ƞ 2 (effect size) of the variable “migration background” as well as of the covariate “basic cognitive skills” in relation to the dependent variables (subtests of language proficiency assessment). The control variable “basic cognitive skills” was not only statistically significant and showed small to medium effects in all tests; it also generated larger effects than those of migration background alone (except in the subtest “derivational morphemes”). The basic cognitive skills wield the strongest influence on the performance in the subtest “prepositions” (ƞ 2 = .18). Therefore, the greatest importance with regard to explaining the students’ performance in the German tests can be attributed to basic cognitive skills. The effect of the variable “MB yes/ no” was considerably smaller when basic cognitive skills are controlled for, yet remains as a small effect. Although the effect of MB on the performance in the subtest “language comprehension” is small, the group difference is not significant (see Table 1 for a similar result). Table 4 presents the results for the covariance analysis, which analyses the effects of basic cognitive skills and migration background on the teachers’ assessment of students’ language proficiency. 152 Jana Chudaske (translated by Nina Rogotzki) <?page no="153"?> Source of variance AV F (1.199) p (α =.05) Ƞ 2 MB yes/ no Basic cognitive skills Language comprehension 5.21 30.36 .024 .000 .03 (.05) .14 MB yes/ no Basic cognitive skills Language production 13.79 34.63 .000 .000 .07 (.11) .15 MB yes/ no Basic cognitive skills Command of grammar 15.33 32.16 .000 .000 .07 (.11) .14 MB yes/ no Basic cognitive skills Interpersonal Communication 8.79 19.37 .003 .000 .04 (.07) .09 MB yes/ no Basic cognitive skills Overall assessment 12.17 17.87 .001 .000 .06 (.08) .08 Table 4: F-values (test statistics), p-values (statistical significance), and ƞ 2 (effect size) of the variable “migration background” (MB) as well as of the covariate “basic cognitive skills” in relation to the dependent variables (teachers’ assessments of language ability). The results in Table 4 are similar to the ones presented for the standardised language tests (Table 3): Basic cognitive skills exert significant and medium effects on all teachers’ assessments of students’ language proficiency. Smaller effects are only noted in teachers’ assessment of students’ competence regarding German as the language of instruction (ƞ 2 = .08). The other effects range from ƞ 2 = .09 to ƞ 2 = .15. Thus, the basic cognitive skills of the students have a greater impact on the teachers’ assessments than the existence of a migration background. Again, even though MB loses influence when the control variable “basic cognitive skills” is taken into account, small effects ranging from ƞ 2 = .03 to ƞ 2 = .07 remain. 7 Discussion and conclusion Using standardised German language tests, which are supplemented by teach‐ ers’ assessments, this study provides important insights into the linguistic abilities of third graders with and without migration backgrounds. The find‐ ings show that students with MB generally performed lower in all language proficiency tests than students without MB (research question 1a). The only exception pertains to the subtest ‘language comprehension’ of the AST, where group differences were statistically not significant. Two explanations may account for this result: On the one hand, the question arises as to how well the variable “language comprehension” in the present data set represents the construct of linguistic competence as such. On the other hand, the performance 153 Competence development in multilingual primary school classes in Germany <?page no="154"?> of the students with MB in this test may characterise their language competence more precisely than other tests because the contents of this test include aspects taught at school which are likely to have been practised with ample frequency during the German lessons: For example, knowledge of punctuation, word families, and adjectives is usually not part of the linguistic knowledge acquired at home or outside school. In a similar vein, Komor & Reich (2008) argue that school attendance produces certain positive effects for MB students (even though monolingual peers’ language proficiency level is seldom reached). However, a comparable result would then be expected with regard to the subtest “prepositions”, because since school entry third graders have constantly been confronted with prepositions because they play a significant role in following instructions and completing work assignments. Yet, the primary school students with and without MB differ significantly with respect to this subtest. Thus, an effect of school attendance does not seem to play an important role. In general, the finding that the children with and without MB differ in five out of six subtests on German language proficiency is ultimately a result that also complies with prior findings obtained with younger children. For example, in the BiKS study, Dubowy et al. (2008) assessed linguistic and cognitive competence of children in day care centres and reported that children without MB consistently performed better in German tests than their peers with MB. In this study, the largest group differences were found with regard to vocabulary and derivational morphemes. It is possible that these constitute specific areas of difficulty, which may be focussed on more strongly in language training programmes. The teachers’ assessments of their students’ language proficiency show significant group differences in all categories (research question 1b). The teachers generally rate the linguistic competence of children without MB more favourably. The most significant differences are apparent for the categories “language production” and “command of grammar”, whereas the differences of children with and without MB in the categories “interpersonal communication” and “language comprehension” are less pronounced. It seems to be easier for students with MB to comprehend facts than to express them. That the MB students’ language comprehension was more favourably assessed than their language production possibly reflects a different level of development relation of comprehension and production at that particular point in the second language acquisition process of (at least) those children who encountered the German language for the first time at school entry. With respect to the teachers’ assessments of students’ general communicative skills regarding the language behaviour towards classmates, the differences between students with and without MB turned out to be rather small. However, written language skills 154 Jana Chudaske (translated by Nina Rogotzki) <?page no="155"?> or the comprehension and proper use of lesson-related technical terms have not been included in the teachers’ evaluations (Komor & Reich, 2008), which may be relevant for future studies. Basic cognitive skills constitute prerequisites for learning and academic achievement. These may also overrule other variables, such as the students’ language background. In this respect, the findings of the present study clearly indicate that the effect of migration background on all test results and teach‐ ers’ assessments was considerably reduced when basic cognitive skills were controlled for (research question 2). Hence, migration background does not play too great a role in the development of linguistic competence and is overruled by the effects of basic cognitive skills. For example, a comparison of the effect sizes also indicated that basic cognitive skills hold twice as much explanatory power regarding many linguistic competences as compared to migration background. Similar results were obtained in other large-scale studies, such as BeLesen, PISA or PIRLS. Some limitations affect the generalisability of the findings reported in this study: First, its descriptive design is, according to Rost (2007), rather weak because it neither meets the requirements of an experiment nor those of a quasi-experiment. Second, any interpretation of causal effects is difficult because the central variables were collected within a very narrow time frame. Third, as no random sampling or adequate cluster sampling took place (Bortz & Döring, 2003), the results of the study only allow limited conclusions. Fourth, the classification of the children according to the presence or absence of a migra‐ tion background produced satisfactory results (which confirmed expectations), however, possible effects of the mother’s country of birth should be pursued by further, more in-depth analyses. Fifth, the standard values of the HSET and the AST are outdated and therefore have to be treated with caution. These test methods were adopted, though, because they fulfil standard test criteria and, moreover, because of the limited availability of adequate alternatives. In retrospect, the SFD also needs to be addressed critically because this test was more difficult to master for the students than expected: Even a majority of the children with German as their first language had great difficulties completing the tasks correctly. Further examinations could help to ascertain whether these were effects of the instrument or of the composition of the sample. In sum, the key finding of this study is that basic cognitive skills, when integrated as a covariate in statistical analyses, generally override language background effects (such as a migration background). This result was found in standardised language proficiency tests as well as in teachers’ assessment of students’ linguistic abilities. Thus, any future study is well advised to include 155 Competence development in multilingual primary school classes in Germany <?page no="156"?> cognitive variables in their design in order to disentangle the effects of students’ linguistic and cognitive background. References Arnold, K-H. & Richert, P. 2009. Diagnostik. In: Heckt, D. H. & Sandfuchs, U. (Eds.), Grundschule von A bis Z. Braunschweig: Westermann. 33-34. Arnold, K-H., Lindner-Müller, C., Hentschel, M. & Chudaske, J. 2007. KEIMS 2006: Kom‐ petenzentwicklung in multilingualen Schulklassen - Ergebnisse der Sprachstandsunter‐ suchung in Klassenstufe 3 im Schuljahr 2005/ 06. Hildesheim: Universität, Centrum für Bildungs- und Unterrichtsforschung. Bortz, J. & Döring, N. 2003. Forschungsmethoden und Evaluation. 3 rd ed. Berlin: Springer. Degel, C. 1987. Zum Einfluss der familialen Lernumwelt auf die Sprachentwicklung: Eine Analyse intraindividueller Veränderungen. In: Wolf, B. (Ed.), Zuwendung und Anre‐ gung. Lernumweltforschung zur Sprachentwicklung im Elternhaus und Kindergarten. Weinheim: Deutscher Studien Verlag. 109-121. Dubowy, M., Ebert, S., von Maurice, J. & Weinert, S. 2008. Sprachlich-kognitive Kompe‐ tenzen beim Eintritt in den Kindergarten. Ein Vergleich von Kindern mit und ohne Migrationshintergrund. Zeitschrift für Entwicklungspsychologie und Pädagogische Psy‐ chologie, 40 (3), 124-134. Ehlich, K. 2005a. Eine Expertise zu “Anforderungen an Verfahren der regelmäßigen Sprachstandsfeststellung als Grundlage für die frühe und individuelle Sprachför‐ derung von Kindern und Jugendlichen mit Migrationshintergrund”. In: Gogolin, I., Neumann, U. & Roth, H-J. (Eds.), Sprachdiagnostik bei Kindern und Jugendlichen mit Migrationshintergrund. Münster: Waxmann. 33-50. Ehlich, K. 2005b. Sprachaneignung und deren Feststellung bei Kindern mit und ohne Migrationshintergrund: Was man weiß, was man braucht, was man erwarten kann. In: Ehlich, K. (Ed.), Anforderungen an Verfahren der regelmäßigen Sprachstandsfeststellung als Grundlage für die frühe und individuelle Förderung von Kindern mit und ohne Migrationshintergrund. Bildungsreform Bd. 11. Berlin: Bundesministerium für Bildung und Forschung (BMBF). 11-75. Ehlich, K., Bredel, U. & Reich, H.H. 2008. Sprachaneignung - Prozesse und Modelle. In: Ehlich, K., Bredel, U. & Reich, H.H. (Eds.), Referenzrahmen zur altersspezifischen Sprachaneignung. Bildungsforschung Bd. 29/ I. Berlin: Bundesministerium für Bildung und Forschung (BMBF). 9-34. Fippinger, F. 1991. Allgemeiner Schulleistungstest für 3. Klassen (AST 3). 2 nd ed. Weinheim: Beltz. 156 Jana Chudaske (translated by Nina Rogotzki) <?page no="157"?> Fried, L. 2004. Expertise zu Sprachstandserhebungen für Kindergartenkinder und Schu‐ lanfänger. Eine kritische Betrachtung. Deutsches Jugendinstitut. https: / / www.dji.de/ fileadmin/ user_upload/ bibs/ 271_2232_ExpertiseFried.pdf. May 31, 2021. Grimm, H. & Schöler, H. 1991. Heidelberger Sprachentwicklungstest. 2 nd ed. Göttingen: Hogrefe. Grimm, H. & Weinert, S. 2002. Sprachentwicklung. In: Oerter, R. & Montada, L. (Eds.), Entwicklungspsychologie. 5 th ed. Weinheim: PVU Psychologie Verlags Union. 517-550. Hartig, J. 2008. Kompetenzen als Ergebnisse von Bildungsprozessen. In: Jude, N., Hartig, J. & Klieme, E. (Eds.), Kompetenzerfassung in pädagogischen Handlungsfeldern. Theorien, Konzepte und Methoden. Bildungsforschung Bd. 26. Berlin: Bundesministerium für Bildung und Forschung (BMBF). 15-25. Hasselhorn, M. & Werner, I. 2000. Phonologisches Arbeitsgedächtnis und Sprachent‐ wicklung. In: Grimm, H. (Ed.), Sprachentwicklung. Enzyklopädie der Psychologie, Themenbereich C: Theorie und Forschung, Serie 3: Sprache, Bd. 3. Göttingen: Hogrefe. 363-378. Herwartz-Emden, L. & Küffner, D. 2006. Schulerfolg und Akkulturationsleistungen von Grundschulkindern mit Migrationshintergrund. Zeitschrift für Erziehungswissen‐ schaft, 9 (2), 240-254. Hobusch, A., Lutz, N. & Wiest, U. 2002. Sprachstandsüberprüfung und Förderdiagnostik für Ausländer- und Aussiedlerkinder (SFD 3/ 4). Horneburg: Persen. Husfeldt, V., Arnold, K-H., Möser, G. & Brümmer, F. 2004. PISA 2000: Ergänzende Analysen für das Land Bremen. Zusatzstudie im Auftrag des Senators für Bildung und Wissenschaft, Bremen. Göttingen: Universität Göttingen. Ingenkamp, K. & Lissmann, U. 2008. Lehrbuch der Pädagogischen Diagnostik. 6 th ed. Weinheim: Beltz. Klauer, K. J. 2006. Intelligenz und Begabung. In: Rost, D.H. (Ed.), Handwörterbuch Pädagogische Psychologie. 3 rd ed. Weinheim: Beltz. 275-280. Klieme, E. & Hartig, J. 2007. Kompetenzkonzepte in den Sozialwissenschaften und im erziehungswissenschaftlichen Diskurs. Zeitschrift für Erziehungswissenschaft, 10 (8), 11-29. Komor, A. & Reich, H.H. 2008. Semantische Basisqualifikation. In: Ehlich, K., Bredel, U. & Reich, H.H. (Eds.), Referenzrahmen zur altersspezifischen Sprachaneignung. Bildungsforschung Bd. 29/ I. Berlin: Bundesministerium für Bildung und Forschung (BMBF). 49-61. List, G. 2005. Was tun und was können Kinder sprachlich? Auf dem Weg vom linguis‐ tischen Testversuch zum entwicklungspsychologischen Sprachhandlungskonzept. In: Gogolin, I., Neumann, U. & Roth, H-J. (Eds.), Sprachdiagnostik bei Kindern und Jugendlichen mit Migrationshintergrund. Münster: Waxmann. 51-57. 157 Competence development in multilingual primary school classes in Germany <?page no="158"?> Lorenz, C. & Artelt, C. 2009. Fachspezifität und Stabilität diagnostischer Kompetenz von Grundschullehrkräften in den Fächern Deutsch und Mathematik. Zeitschrift für Pädagogische Psychologie, 23 (3/ 4), 211-222. Marx, E. 2006. Profitiert das kindliche Sprachsystem von anderen kognitiven Entwick‐ lungsbereichen? Pilotstudie zum Zusammenhang von Spracherwerb und induktivem Denken. Zeitschrift für Entwicklungspsychologie und Pädagogische Psychologie, 38 (3), 139-145. Merkens, H. 2005. Schulkarrieren von Kindern mit Migrationshintergrund in den ersten drei Jahren der Grundschule. Berichte aus der Arbeit des Arbeitsbereichs empirische Erziehungswissenschaft der Freien Universität Berlin. Heft 43. Berlin: Freie Universität. Metz, D., Fröhlich, L.P. & Petermann, F. 2009. Sprachstandserhebungsverfahren für Fünfbis Zehnjährige (SET 5-10). Konstruktion und Analyse. Kindheit und Entwicklung, 18 (4), 194-203. Neugebauer, U. & Becker-Mrotzek, M. 2013. Die Qualität von Sprachstands‐ verfahren im Elementarbereich. Eine Analyse und Bewertung. Mercator-In‐ stitut für Sprachförderung und Deutsch als Zweitsprache. https: / / www.mercator-i nstitut-sprachfoerderung.de/ fileadmin/ user_upload/ Mercator-Institut_Qualitaet_Spr achstandsverfahren_Web_03.pdf. May 31, 2021. Preuss-Lausitz, U. 2005. Verhaltensauffällige Kinder integrieren. Zur Förderung der emo‐ tionalen und sozialen Entwicklung. Weinheim: Beltz. Reich, H.H. 2005. Forschungsstand und Desideratenaufweis zu Migrationslinguistik und Migrationspädagogik für die Zwecke des “Anforderungsrahmens”. In: Ehlich, K. (Ed.), Anforderungen an Verfahren der regelmäßigen Sprachstandsfeststellung als Grundlage für die frühe und individuelle Förderung von Kindern mit und ohne Migrationshinter‐ grund. Bildungsreform Bd. 11. Berlin: Bundesministerium für Bildung und Forschung (BMBF). 121-169. Reich, H.H. 2008. Die Sprachaneignung von Kindern in Situationen der Zwei- und Mehrsprachigkeit. In: Ehlich, K., Bredel, U. & Reich, H.H. (Eds.), Referenzrahmen zur al‐ tersspezifischen Sprachaneignung. Forschungsgrundlagen. Bildungsforschung Bd. 29/ II. Berlin: Bundesministerium für Bildung und Forschung (BMBF). 163-169. Rost, D.H. 2007. Interpretation und Bewertung pädagogisch-psychologischer Studien. Eine Einführung. 2 nd ed. Weinheim: Beltz. Rost, D.H. 2009. Intelligenz. Fakten und Mythen. Weinheim: Beltz. Schiffer, K., Ennemoser, M. & Schneider, W. 2002. Die Beziehung zwischen dem Fernseh‐ konsum und der Entwicklung von Sprach- und Lesekompetenzen im Grundschulalter in Abhängigkeit von der Intelligenz. Zeitschrift für Medienpsychologie, 14 (1), 2-13. Schnieders, G. & Komor, A. 2005. Eine Synopse aktueller Verfahren der Sprachstands‐ feststellung. In: Ehlich, K. (Ed.), Anforderungen an Verfahren der regelmäßigen Sprach‐ standsfeststellung als Grundlage für die frühe und individuelle Förderung von Kindern 158 Jana Chudaske (translated by Nina Rogotzki) <?page no="159"?> mit und ohne Migrationshintergrund. Bildungsreform Bd. 11. Berlin: Bundesministe‐ rium für Bildung und Forschung (BMBF). 261-342. Schölmerich, A., Leyendecker, B., Citlak, B., Caspar, U. & Jäkel, J. 2008. Assessment of migrant and minority children. Journal of Psychology, 216 (3), 188-195. Schroeder, C. & Stölting, W. 2005. Mehrsprachig orientierte Sprachstandsfeststellungen für Kinder mit Migrationshintergrund. In: Gogolin, I., Neumann, U. & Roth, H-J. (Eds.), Sprachdiagnostik bei Kindern und Jugendlichen mit Migrationshintergrund. Münster: Waxmann. 59-74. Schründer-Lenzen, A. & Merkens, H. 2006. Differenzen schriftsprachlicher Kom‐ petenzentwicklung bei Kindern mit und ohne Migrationshintergrund. In: Schründer-Lenzen, A. (Ed.), Risikofaktoren kindlicher Entwicklung. Migration, Leis‐ tungsangst und Schulübergang. Wiesbaden: VS Verlag für Sozialwissenschaften. 15-44. Spinath, B. 2005. Akkuratheit der Einschätzung von Schülermerkmalen durch Lehrer und das Konstrukt der diagnostischen Kompetenz. Zeitschrift für Pädagogische Psychologie, 19 (1/ 2), 85-95. Weinert, F.E. 2001. Concept of competence: A conceptual clarification. In: Rychen, D.S. & Hersh Salganik, L. (Eds.), Defining and Selecting Key Competencies. Göttingen: Hogrefe & Huber. 45-65. Weinert, S., Doil, H. & Frevert, S. 2008. Kompetenzmessungen im Vorschulalter: Eine Analyse vorliegender Verfahren. In: Roßbach, H-G. & Weinert, S. (Eds.), Kindliche Kompetenzen im Elementarbereich: Förderbarkeit, Bedeutung und Messung. Bildungs‐ forschung Bd. 24. Berlin: Bundesministerium für Bildung und Forschung (BMBF). 89-209. Weiß, R.H. & Osterland, J. 1997. Grundintelligenztest Skala 1 (CFT 1). 5 th ed. Göttingen: Hogrefe. Wild, E., Hofer, M. & Pekrun, R. 2001. Psychologie des Lerners. In: Krapp, A. & Weidenmann, B. (Eds.), Pädagogische Psychologie. 4 th ed. Weinheim: PVU Psychologie Verlags Union. 207-270. Wolf, B. 1987. Theoretischer Gesamtrahmen. In: Wolf, B. (Ed.), Zuwendung und Anregung. Lernumweltforschung zur Sprachentwicklung im Elternhaus und Kindergarten. Wein‐ heim: Deutscher Studien Verlag. 6-28. 159 Competence development in multilingual primary school classes in Germany <?page no="161"?> Cognitive and linguistic profiles in early foreign language vocabulary and grammar Holger Hopp / Teresa Kieseier / Markus Vogelbacher / Dieter Thoma Abstract In this chapter, we assess the extent to which profile effects, that is domain-particular patterns of cognitive and linguistic predictors in early natural L2 acquisition, extend to classroom learners. Specifically, we test the relative impacts of cognitive and linguistic factors on achievement in the early foreign language acquisition of English vocabulary and grammar among German primary-school students. Using linear-mixed effects re‐ gression modelling, this chapter maps profile effects in the acquisition of vocabulary and grammar in native German-speaking students and bilingual students speaking a minority L1 other than German at the end of grades 3 and 4. The data reveal distinct profiles in (a) linguistic domain, (b) mode, i. e. productive versus receptive skills, (c) stages in development and (d) group, i. e. monolingual versus bilingual students, with both cognitive and linguistic factors contributing to English skills. 1 Introduction In this contribution, we assess the relative impacts of cognitive and linguistic factors on achievement in the early foreign language acquisition of English vocabulary and grammar in primary school. Previous research identified dif‐ ferent factors as being relevant for vocabulary development than for grammar acquisition (e.g. Paradis, 2011), so that different profiles emerge depending on the domain and type of acquisition (Gathercole, 2002; Marinis & Chondrogianni, 2011; Oller, Pearson & Cobo-Lewis, 2007; Unsworth, 2016). This chapter traces profile effects in the acquisition of vocabulary and grammar in students with native German and bilingual students with a minority L1 other than German at <?page no="162"?> 1 This study was funded by the German Ministry for Education and Science (BMBF) as part of the research initiative “Language Education and Multilingualism”, grant no. 01JM1401 (11/ 2014 - 10/ 2017). the end of grades 3 and 4 in German public primary schools. 1 In addition, our study comprises a longitudinal aspect in that we consider whether these profiles remain constant from grades 3 to 4. This way, we can map potential changes in the impact of factors on foreign language learning that may change over time (Paradis & Jia, 2017). In research on child L2/ 3 acquisition, cognitive and linguistic profiles have been studied under the rubric of individual differences or the influence of child-internal versus child-external factors (e.g. Paradis, Rusk, Duncan & Govindarajan, 2017; Unsworth, 2013; 2016). Internal factors can be grouped into personal factors (e.g. age), cognitive factors (e.g. working memory, basic cognitive reasoning) and linguistic factors (e.g. L1, vocabulary size), on the one hand, while external factors refer to the quantity and quality of input as well as the social contexts of language learning, on the other. Of note, previous studies on the early acquisition of English demonstrate that child-internal factors are more predictive than external factors (e.g. Chondrogianni & Marinis, 2011; Paradis, 2011; Paradis, Tulpar & Arppe, 2016; Paradis et al., 2017; but see Paradis & Jia, 2017). However, all of these studies have been conducted with immersed child L2/ 3 learners of English in contexts where English is the majority language (for review, see Murphy & Evangelou, 2016). In educational contexts of instructed foreign language acquisition, it is im‐ portant to assess whether these findings generalise to foreign language learning of English or whether the context of instructed learning in an environment with a different majority language engenders different profiles of early L2 acquisition. So far, large-scale studies on foreign language learning of English in Germany find that various factors play a role in English achievement. Bilingualism, proficiency in the majority language, use of and formal instruction in the minority L1 and age each affect foreign language development and outcomes to various extents (e.g. Göbel, Vieluf & Hesse, 2010; Hesse, Göbel & Hartig, 2008; Maluch & Kempert, 2017; Maluch, Kempert, Neumann & Stanat, 2015; Maluch, Neumann & Kempert, 2016). Due to the large samples, the studies typically assess only some contributing factors or tested them only indirectly via self or parental report, so that they cannot specify the precise impacts of the individual factors, let alone model their interactions. Moreover, large-scale studies typically employ only one general outcome variable, e. g. reading comprehension or performance in cloze tasks, so that they cannot test for profile effects across linguistic domains. 162 Holger Hopp / Teresa Kieseier / Markus Vogelbacher / Dieter Thoma <?page no="163"?> In this study, we investigate a wide variety of cognitive and linguistic factors in 200 students learning English as a foreign language in German public primary schools. The students were tested at the end of grade 3 and again at the end of grade 4. We assess English skills in receptive and productive vocabulary as well as in receptive grammar. In addition, this study measures cognitive predictors, i. e. working memory, basic cognitive processing, executive control and phonological awareness. As linguistic predictors, we include productive vocabulary measures in all previously learnt languages. Further, we consider social factors, e. g. socio-economic status (SES), and personal factors, e. g. gender and age, to delineate the relative scope of internal versus external factors. This chapter is structured as follows: In the following section, we review contributing factors to child L2 development, and we formulate the research question and the hypotheses. In section 3, we present the study and its findings. Section 4 discusses the results and concludes. 1.1 Profile effects and contributing factors in bilingual and child L2 development Profile effects emerge as the consequence of the distributed characteristics of bilingual knowledge (Oller et al., 2007), i. e. the fact that bilingual linguistic knowledge scopes across more than one language. In consequence, different profile effects may ensue: A) One linguistic domain may be more developed than some other domain in one language. B) There may be cross-language facilitation in one domain. Finally, C) there may be across-domain facilitation in bilinguals, with e. g. linguistic experience leading to cognitive benefits of early bilingualism that feed back into the acquisition of further languages. In research on child L2 acquisition, the investigation of profile effects has largely been focused on delineating systematic differences between linguistic domains in acquisition (e.g. Chondrogianni & Marinis, 2011; Oller et al., 2007). In this chapter, we take a wider view of the notion of profile effects and consider how diversified profiles in foreign language acquisition arise due to effects of different factors, e. g. cognitive, linguistic and group-specific factors. Cognitive factors At the cognitive level, a set of different factors has been found to give rise to profiles in the early acquisition of foreign languages. Working memory, i. e. the ability to store and manipulate verbal and non-verbal information (e.g. Baddeley, 2007), proves to be an important predictor of vocabulary acquisition in monolingual children (e.g. Gathercole, Willis, Emslie & Baddeley, 1992). Simple verbal memory span tasks, like the repetition of arbitrary digit or syllable sequences, also show a positive correlation with second language proficiency 163 Cognitive and linguistic profiles in early foreign language vocabulary and grammar <?page no="164"?> (see Linck, Osthus, Koeth & Bunting, 2014, for a meta-analysis). In particular for beginning foreign language learners, digit span tasks correlate with vocabulary and grammar development (e.g. Juffs & Harrington, 2011; Paradis, 2011; Speciale, Ellis & Bywater, 2004). In addition, aspects of executive control, i. e. the ability to focus, switch or inhibit attention (e.g. Miyake & Friedman, 2012), are implicated in bilingual acquisition. Especially early bilingualism appears to lead to cognitive advantages in executive control (e.g. Adesope, Lavin, Thompson & Ungerleider, 2010; Bialystok, 2009) that may circle back as increased language learning skills in the acquisition of a foreign language (e.g. Peets & Bialystok, 2010). Further, cognitive facility in analysing and segmenting linguistic elements contributes to language learning. Metalinguistic awareness, i. e. the ability to segment and manipulate language independent of meaning aspects (e.g. Jessner, 2006) contributes to the acquisition of vocabulary and grammar (e.g. Gathercole, Hitch, Service & Martin, 1997). In particular phonological awareness, usually tested as the skill to identify and manipulate phonemes or syllables in nonce words, acts as an important predictor of vocabulary and grammar acquisition in an L2 (e.g. Hu, 2003; Paradis, 2011; Verhoeven, 2007) and may lead to benefits in foreign language acquisition more generally (e.g. Cenoz, 2013b; Rauch, Naumann & Jude, 2012). Together, these factors may give rise to enhanced language learning ability, which may be subsumed under a general notion of language aptitude (e.g. Brooks & Kempe, 2013). Linguistic factors Among linguistic factors, cross-linguistic influence affects L2 acquisition. Spe‐ cifically, the degree of L1 proficiency in different domains, i. e. vocabulary, grammar and literacy, or the extent to which previously acquired languages differ in typological or structural similarity from the foreign language may mod‐ ulate foreign language acquisition. For bilingual children, Gathercole, Thomas, Roberts, Hughes & Hughes (2013) report that correlations in vocabulary size emerge as children grow older (for adult L2 learners, see Molnar, 2010; Szabo, 2016), which provides indirect evidence of the connectedness of lexicons in bilin‐ guals and foreign language learners. Direct evidence for cross-linguistic lexical interactions comes from psycholinguistic research on the multilingual lexicon. Bilinguals and foreign language learners demonstrate different responses in reaction time tasks, e. g. lexical decision or picture naming, for cognate words (e.g. FILM German/ English ) versus interlingual homographs (e.g. GIFT German/ English : pre‐ sent English - poison German ), suggesting that representations from both languages are accessed or that these representations are interconnected (for a review, see Dijkstra, 2005). These effects are in evidence also for children learning an L2 or 164 Holger Hopp / Teresa Kieseier / Markus Vogelbacher / Dieter Thoma <?page no="165"?> L3 in primary or early secondary school (Brenders, van Hell & Dijkstra, 2011; Poarch & van Hell, 2012), suggesting that the multilingual lexicon is connected at different levels from early on (e.g. Ecke, 2015; Kroll & Stewart, 1994). Similarly, early L2 learners initially transfer grammatical properties from the L1 (e.g. Haznedar, 1997; Paradis, 2011). Group factors Language learning differs depending on whether the foreign language is learnt as a second or a third language (Cenoz, 2013a). Studies comparing early bilinguals or heritage speakers of a minority language next to the majority language, on the one hand, to monolingual speakers of the majority language, on the other, find advantages for bilingual over monolingual students in foreign language performance when social and cognitive factors are controlled (e.g. Hesse et al., 2008; Maluch et al., 2015; 2016; for an overview, Cenoz, 2013b). Bilingual advantages in foreign language learning have variously been attributed to higher degrees of metalinguistic awareness (e.g. Jessner, 2008) or enhanced cognitive skills (e.g. Peets & Bialystok, 2010), greater experience with (learning) multiple languages and a larger linguistic reservoir for facilitative linguistic transfer. For instance, grammatical development may be affected by language transfer (e.g. Odlin, 1989), although the evidence of L1 transfer in early instructed foreign language acquisition is scarce (e.g. Lenzing, 2013; Wanders, 2006) and may be restricted to particular domains of finiteness marking (e.g. Paradis, 2011). Current theories of L3 acquisition predict different amounts of cross-linguistic transfer in the initial stages of foreign language acquisition for second language and third language learners of a foreign language (for an overview, Garcia Mayo, 2013). However, these models have rarely been applied to child trilinguals or early foreign language learners (though see Hopp, Kieseier, Vogelbacher & Thoma, 2018). Finally, many additional factors affect the time course and the trajectories of child second and foreign language acquisition, e. g. the quantity and quality of input of languages spoken at home and in school (e.g. Armon-Lotem, Wal‐ ters & Gagarina, 2011; Unsworth, Argyri, Cornips, Hulk, Sorace & Tsimpli, 2014), parental education and socio-economic status (e.g. Bohman, Bedore, Peña, Mendez-Perez & Gillam, 2010), teacher and school characteristics (Stanat, Rauch & Segeritz, 2010; Unsworth, Persson, Prins & de Bot, 2015), the gender and age of the children (e.g. Paradis, 2011; Unsworth, 2016) as well as the opportunities for child output in the second or foreign language (e.g. Bohman et al., 2010). Against the backdrop of this multitude of individual differences factors, the present study focuses on cognitive and linguistic predictors, while controlling for additional factors. 165 Cognitive and linguistic profiles in early foreign language vocabulary and grammar <?page no="166"?> 1.2 Research questions and hypotheses In the present study on students learning English as an early foreign language in German public primary schools, we ask the following research questions. 1) Which cognitive and linguistic factors contribute to English achievement in vocabulary and grammar? Based on previous research, we expect both cognitive and linguistic factors to affect early foreign language acquisition, with specific factors accounting for differences between groups and domains. These differ‐ ences are the focus of our investigation and give rise to the second research question. 2) Does the effect structure of cognitive and linguistic factors differ across linguistic domains, tasks, stages or groups in acquisition? Probing differential impacts of the predictors across various target language domains allows us to identify profile effects in early foreign language learning. Following from previous research, we predict cognitive factors to be more pronounced in lexical production and sentence comprehension because both require greater cognitive resources than isolated word comprehension (e.g. Chondrogianni & Marinis, 2011). In terms of linguistic predictors, we expect more cross-language effects for the lexicon than for grammar (e.g. Oller et al., 2007; Paradis, 2011). In terms of stages, we expect to see more within-language effects, i. e. English-to-English, over time than cross-linguistic effects as proficiency in English rises and reliance on previously acquired languages abates (e.g. Lenzing, 2013; Paradis & Jia, 2017). In terms of groups, we expect bilingual students to have an advantage in vocabulary and grammar over monolingual students when background factors are controlled for (e.g. Maluch et al., 2015, 2016). Next to these separate predictions for each aspect, we assess interactions between them. 2 The study 2.1 Participants Overall, we tested one sample of 200 students at six public primary schools in south-west Germany at the end of grade 3 and at the end of grade 4. In all schools, students learnt English as a foreign language from grade 1 onwards for two 45-minute lessons per week. At the end of grade 3, the sample comprised 200 students, of whom 88 were monolingual (44 %) and 112 were bilingual learners of English (56 %). At the end of grade 4, 81 monolingual and 103 multilingual students remained in the sample. In terms of age, the students ranged from seven years and ten months to eleven years and three months at the end of grade 3. The bilingual students had a mean length of exposure to German of eight years and four months (Table 1). 166 Holger Hopp / Teresa Kieseier / Markus Vogelbacher / Dieter Thoma <?page no="167"?> Full sample Bilingual Monolingual Male Female Male Female Male Female Sex 105/ 102 95/ 82 55/ 54 57/ 49 50/ 48 38/ 33 N 200/ 184 112/ 103 88/ 81 Age in months M SD N M SD N M SD N 111.24 / 123.12 5.93 / 5.76 198 182 112.03 / 123.72 6.09 / 5.92 111 / 102 110.24 / 122.35 5.59 / 5.51 87 / 80 Min Max Min Max Min Max 95/ 107 135/ 147 95/ 107 135/ 147 96/ 110 126/ 138 Contact with German in months M SD N 99.95 / 112.08 24.64 / 24.90 88 / 85 Table 1: Participants’ descriptives (grade 3/ grade 4). In this study, we defined all students who had acquired a language other than or in addition to German productively and/ or receptively before their first contact with English as bilingual students. The bilingual students had a wide range of first languages, i. e. Afghan, Albanian, Arabic, Bosnian, Bulgarian, Chinese, French, Greek, Italian, Croatian, Kurdish, Persian, Polish, Roma, Romanian, Russian, Serbian, Spanish, Tamil, Turkish, Hungarian and Vietnamese. The largest subgroups were speakers of Turkish (n = 40), Kurdish (n = 11), Albanian (n = 10) and Italian (n = 8). Across schools, the proportion of bilingual students varied and ranged from 22 % to 87 % (M = 56 %; SD = 27). 2.2 Language assessments We used standardised tests to assess receptive lexical performance at the end of grade 3 and to measure receptive grammatical performance at the end of grade 4. For receptive vocabulary in English, we administered the British Picture and Vocabulary Scale (BPVS III; Dunn, Dunn, Styles & Sewell, 2009). Based on pilot tests, we used sets 1-6 for the present study. For receptive grammar in English, we administered the Test for Reception of Grammar (TROG-2; Bishop, 2003). In addition, we collected measure of productive vocabulary in English, German and the non-German L1s of the bilingual students at the end of grades 3 and 4, using a category fluency task (adapted from Delis, Kaplan & Kramer, 2001). In this task, students named as many items as possible belonging to a semantic category 167 Cognitive and linguistic profiles in early foreign language vocabulary and grammar <?page no="168"?> (“animals”, “food”) within 60 seconds. For German and the L1s of the bilingual students, the semantic fields “plants” and “clothing” were used as categories. 2.3 Cognitive assessments We collected data on a number of cognitive functions. Phonological awareness was assessed in tasks testing phoneme deletion (i.e. the deletion of initial or final sounds) and phoneme manipulation (following Weber, Marx & Schneider, 2007). To measure non-verbal cognitive skills, we used part 1 of the non-verbal IQ test “Grundintelligenztest Skala 2 - Revision” (CFT 20-R, Weiß, 2006). Working memory was assessed in forward and backward digit span tasks (adapted from HAWIK-IV, Petermann & Petermann, 2008). Furthermore, executive function was measured using the Simon Task (Simon, 1969). 2.4 Social and background factors In addition to the language and cognitive factors described above, this study used social, family, linguistic and other background variables as control variables. Individual data were collected in a detailed parent questionnaire, which covered the language history of the child, the socio-economic and cultural conditions as well as the migration history of the family. Following Ganzeboom, de Graaf & Treiman (1992), we compiled a hierarchical score for the socio-economic status of the parents (ISEI) on the basis of parents’ profession and vocational training. In addition, we used the highest level of parental education, net household income and cultural capital (number of books in household) as controls. At the school level, we aggregated data on the proportion of bilingual students and the mean socio-economic status of the students. 2.5 Procedure Participation in the project was voluntary, and parents gave written consent for their children to take part in the study. The BPVS, TROG-2 and the CFT 20-R were administered in group settings in the classroom during regular lessons. For the BPVS and TROG, we created laminated books containing four pictures on each page. The students were orally presented with a vocabulary item (BPVS) or a sentence (TROG) and had to match it to one of the four pictures on an answer sheet. For the CFT 20-R, students were given task books as well as answer sheets and had to complete the tasks in four blocks with set time limits. For all other tasks, students were tested individually in a quiet room. Each session took about 30 to 45 minutes and was audio-recorded in full. All tasks were paper-and-pencil based, except for the computer-based Simon Task. For the category fluency tasks, the digit span and the phoneme tasks, all spoken responses were recorded, 168 Holger Hopp / Teresa Kieseier / Markus Vogelbacher / Dieter Thoma <?page no="169"?> transcribed and then coded. Parent questionnaires were handed out via the teachers to the students at the end of grade 3. The return rate was 82.5 %, i. e. 165 out of 200 questionnaires. 2.6 Analysis Raw test scores were computed according to the manuals of the respective tests. For the category fluency tasks, all responses were transcribed and coded for the respective semantic category. For the non-German L1s, native speakers of the respective L1s transcribed the data and provided literal German translations for each item, so that the responses could be coded. 3 Results Full sample Bilinguals Monolinguals T-Tests d Cohen English skills Grade 3: Receptive English Vocabulary (BPVS) 39.75(6.56) 38.28 (6.79) 41.61 (5.76) t (198) = -3.684, p < .001 .490 Grade 4: Receptive English Grammar (TROG-2) 46.92 (10.20) 43.92 (9.53) 49.12 (8.20) t (182) = -3.906, p < .001 .580 Grade 3: Productive English Vocabulary 8.11 (5.38) 7.58 (5.50) 8.78 (5.18) t (198) = -1.576, p = .117 .224 Grade 4: Productive English Vocabulary 11.25 (5.69) 9.91 (5.45) 12.36 (5.22) t (182) = -3.077, p = .002 .458 Linguistic factors: L1 and L2 skills Grade 3: Productive German Vocabulary 15.63 (6.35) 13.90 (6.07) 17.82 (6.05) t (198) = -4.534, p < .001 .646 Grade 4: Productive German Vocabulary 21.77 (7.31) 20.41 (7.27) 23.17 (7.10) t (182) = -2.588, p = .010 .384 Grade 3: Productive L1 Vocabulary (monolinguals = German) 12.05 (7.86) 7.51 (5.90) 17.82 (6.05) t (198) = - 12.136, p < .001 1.782 Grade 4: Productive L1 Vocabulary (monolinguals = German) 16.48 (9.34) 11.05 (7.17) 23.17 (7.10) t (179) = - 12.123, p < .001 1.699 Cognitive factors (all grade 3) Basic cognitive skills (CFT) 101.50 (14.98) 97.72 (14.38) 106.40 (14.38) t (193) = -4.180, p < .001 .604 Working memory (forward digit span) 7.79 (1.71) 7.54 (1.54) 8.11 (1.87) t (167.2) = - 2.347, p = .020 .337 Phonological awareness (phoneme manipulation) 0.56 (0.36) 0.47 (0.37) 0.66 (0.33) t (194.3) = - 3.856, p < .001 .538 Executive control (Simon Task) 66.81 (46.88) 69.91 (49.76) 62.85 (42.88) t (198) = 1.057, p = .292 - .151 Social factors SES of parents (highest ISEI) 49.43 (10.11) 45.73 (9.75) 53.59 (8.87) t (153) = -5.230, p < .001 .841 Education of parents (years in school) 11.52 (1.65) 11.08 (1.66) 12.00 (1.50) t (153) = -3.603, p < .001 .574 Net household income 3785.56 (2023.08) 3212.56 (1839.77) 4429.20 (2037.72) t (153) = -3.906, p < .001 .629 Cultural capital (number of books in German) 244.81 (390.32) 108.84 (159.09) 387.19 (506.58) t (83.9) = -4.851, p < .001 .770 Table 2: Descriptives; T-Tests bilinguals versus monolinguals; effect size. Mean (standard deviation). 169 Cognitive and linguistic profiles in early foreign language vocabulary and grammar <?page no="170"?> Table 2 presents the mean scores in English skills, linguistic, cognitive and social factors for the full sample as well as separately for the groups of monolingual and bilingual students. When comparing the means of the monolingual and the bilingual groups in pairwise T-Tests, we found significant between-group differences in most English skills, all linguistic factors, all cognitive factors - except for the Simon Task - as well as in all social factors. We assessed the relative impacts of linguistic, cognitive and social factors on English skills by regressing English receptive and productive vocabulary in grade 3 (Table 3) and productive vocabulary and grammar in grade 4 (Table 4) on the predictor variables in a hierarchical linear mixed regression. In order to take the group differences between monolingual and bilingual students into account, we added the factor bilingualism as a between-group factor to the linguistic factors. To control for effects of differences between schools, we used the factor school as a fixed main effect in all models. Further, institutional differences between schools in terms of the proportion of bilingual students and the mean SES of the students at the schools were added to the models at the institutional level, within which individual factors of the students were nested. Among individual factors, we also included gender and age to see whether these personal factors contribute to performance in English since previous research reports advantages for girls over boys (e.g. Chudaske, 2012) and for older children over younger child L2/ 3 learners (e.g. Blom & Bosma, 2016; Paradis, 2011; Paradis & Jia, 2017). Table 3 presents the regression models for the English receptive vocabulary (BPVS) and the English productive vocabulary in grade 3 English as the dependent variables, and Table 4 presents the regression models for the English productive vocabulary and the receptive grammar (TROG-2) in grade 4 as dependent variables. All models were constructed by starting with a model including all predictors and subsequent stepwise exclusion of predictors by means of model fit comparison (log-likelihood ratio tests). In all models, random effects were retained provided they significantly improved model fit. The models in (a) display the optimal models for data from all participants. The models in (b) present the optimal models for the sub-sample for which full parental data were available so that these models also considered social background factors. As can be seen in Tables 3 and 4, the predictor structure does not differ between models (a) and (b) and none of the individual-level social factors reaches significance so that we summarise the models in (a) for the full sample in the text. 170 Holger Hopp / Teresa Kieseier / Markus Vogelbacher / Dieter Thoma <?page no="171"?> BPVS - grade 3 Category fluency - grade 3 Model 1a Model 1b Model 2a Model 2b Constant term 37.97 *** (.79) 37.63 *** (.95) 5.67 *** (.88) 5.55 *** (.92) Institutional level Proportion of bilingual students a -1.87 + (.79) -1.91 + (.78) Mean SES school a 2.41 *** (.44) 2.83 *** (.63) Individual level Cognitive factors • Basic cognitive skills (CFT IQ) a 1.76 *** (.42) 1.36 ** (.51) • Phonological awareness a .89 * (.35) .84 * (.41) • Working memory (DSF) a • Executive control (Simon Task) a Personal factors • Gender female (reference= male) 1.32 + (.77) 1.91 * (.90) • Age (months) a Linguistic factors • Bilingualism (ref. = monolin‐ gual) 1.93 + (1.07) 2.39 + (1.30) 4.30 *** (.93) 4.92 *** (1.18) • Grade 3: Category fluency German a • Grade 3: Category fluency L1 (monolinguals= German) a 1.61 ** (.51) 1.76 ** (.61) 2.12 *** (.44) 2.56 ** (.52) 171 Cognitive and linguistic profiles in early foreign language vocabulary and grammar <?page no="172"?> BPVS - grade 3 Category fluency - grade 3 Parent unit nonresponse (ref. = parent data available) Social factors • SES (HISEI) a -.05 (.77) .67 (.64) • Education parents (years in school) a 1.00 (.69) .33 (.58) • Net household income a -.78 (.63) -.56 (.53) • Cultural capital parents (number of books in German) a -.73 (.51) -.47 (.43) N 194 149 b 194 149 b -2 Restricted Log Likelihood 1190.104 c 908.926 1128.924 c 867.130 Number of parameters 7 11 7 11 Dependent Variable: BPVS - grade 3; category fluency - grades 3; notation: unstan‐ dardised estimates; standard error in parantheses/ controlled for sign. random effects; type of covariance: variance components a standardised b only cases with full information from parents; missing values ML-estimated (EM-al‐ gorithm) c model optimised via chi²-comparison of -2 Restricted Log Likelihood; method: exclusion, stepwise + p < .10, * p < .05, ** p < .01, *** p < .001 Table 3: Predictors of English receptive vocabulary (BPVS) and English productive vo‐ cabulary (category fluency) in grade 3 on institutional and individual level (hierarchical linear mixed regression); controlled for school. In grade 3 (Table 3), the regression models for receptive vocabulary (BPVS) show that, among the individual-level factors, basic cognitive skills and productive vocabulary in the L1 are positively associated with English vocabulary. In addition, the mean socio-economic status of the school shows significant effects at the institutional level. Marginal associations were found for gender, with female students obtaining higher scores, and for bilingualism, with bilingualism contributing positively to English vocabulary. For productive vocabulary in grade 3, phonological awareness and L1 productive vocabulary significantly positively predict English skills, with bilingualism being a highly significant further predictor. At the institutional level, the proportion of bilingual students is marginally negatively associated with productive vocabulary in English. 172 Holger Hopp / Teresa Kieseier / Markus Vogelbacher / Dieter Thoma <?page no="173"?> Category fluency - grade 4 TROG-2 - grade 4 Model 3a Model 3b Model 4a Model 4b Constant term 12.04 *** (.71) 12.20 *** (.76) 44.73 *** (.85) 44.23 *** (.91) Institutional level Proportion of bilingual students a Mean SES school a 3.65 *** (.57) 3.69 *** (.71) Individual level Cognitive factors • Basic cognitive skills (CFT IQ) a 1.82 *** (.51) 1.79 ** (.56) • Phonological awareness a 1.67 ** (.52) 1.55 ** (.56) • Working memory (DSF) a 1.30 ** (.45) 1.04 * (.47) • Executive control (Simon Task) a Personal factors • Gender female (reference= male) 1.53 + (.88) 1.54 (.95) • Age (months) a Linguistic factors • Bilingualism (ref. = mono‐ lingual) -1,26 + (.68) -1.48 + (.85) 1.83 + (1.02) 2.81* (1.15) • Grade 4: Category fluency German a 1.28 *** (.32) 1.24 ** (.38) • Grade 4: Category flu‐ ency L1 (monolinguals= German) a • Grade 3: Category fluency English a 2.91 *** (.35) 2.95 *** (.39) • Grade 4: Category fluency English a 2.98 *** (.50) 3.20 *** (.53) Parent unit nonresponse (ref. = parent data available) -1.85 (1.15) 173 Cognitive and linguistic profiles in early foreign language vocabulary and grammar <?page no="174"?> Category fluency - grade 4 TROG-2 - grade 4 Social factors • SES (HISEI) a .62 (.58) .56 (.80) • Education parents (years in school) a -.37 (.53) .65 (.75) • Net household income a -.19 (.47) -.26 (.67) • Cultural capital parents (number of books in German) a -.23 (.38) -.11 (.54) N 176 140 b 176 140 b -2 Restricted Log Likelihood 975.700 c 777.653 1102.046 c 852.796 Number of parameters 6 10 10 13 Dependent Variable: category fluency - grade 4; TROG-2 - grade 4; notation: unstan‐ dardised estimates; standard error in brackets/ controlled for sign. random effects; type of covariance: variance components a standardised b only cases with full information from parents; missing values ML-estimated (EM-al‐ gorithm) c model optimised via chi²-comparison of -2 Restricted Log Likelihood; method: exclu‐ sion, stepwise + p < .10, * p < .05, ** p < .01, *** p < .001 Table 4: Predictors of English productive vocabulary (category fluency) and English grammar (TROG-2) in grade 4 on institutional and individual level (hierarchical linear mixed regression); controlled for school. In grade 4 (Table 4), productive English vocabulary is affected foremost by performance in the same task in grade 3. At the same time, productive German vocabulary is positively associated with productive English vocabulary. Further, bilingualism shows a marginally significant negative association with English vocabulary skills at the end of grade 4, when performance in English productive vocabulary in grade 3 is taken into account. Finally, for receptive grammar, many individual-level factors contribute to English skills. Among cognitive factors, basic cognitive skills, phonological awareness and working memory positively affect English grammar. Further, English vocabulary in grade 4 and bilingualism show (marginally) significant positive effects, and there is a tendency for girls to perform better than boys. Seeing that bilingualism is significantly associated with all of the English skills, 174 Holger Hopp / Teresa Kieseier / Markus Vogelbacher / Dieter Thoma <?page no="175"?> we next break down the students into groups of monolingual and bilingual students to assess whether the effect structure of cognitive and linguistic factors is comparable across the two groups. Tables 5 and 6 show the results of the model-fitting procedure for the monolingual group. Compared to the full sample, the effect structure differs in the following ways. For English receptive vocabulary, only basic cognitive skills affect English performance at the individual level (Table 5). Unlike in the full sample, L1 productive vocabulary, i. e. German vocabulary, is not significantly associated with English receptive vocabulary. For English produc‐ tive vocabulary in grade 3 (Table 5), however, German productive vocabulary is the only individual-level factor that has significant effects. Unlike in the full sample, phonological awareness does not act as a significant predictor of English vocabulary. For English productive vocabulary in grade 4, only English productive vocabulary in grade 3 contributed to performance (Table 6). In contrast, German productive vocabulary does not show any significant associations with English productive vocabulary in grade 4. Finally, for English receptive grammar in grade 4 (Table 6), the effect structure is comparable to the full sample in Table 4. BPVS - grade 3 Category fluency - grade 3 Model 1a Model 1b Model 2a Model 2b Constant term 40.02 *** (.65) 39.97 *** (.84) 7.10 *** (.64) 7.39 *** (.87) Institutional level Proportion of bilingual students a -1.78 ** (.59) -1.57 + (.83) Mean SES school a 1.95 ** (.63) 2.14 * (.95) Individual level Cognitive factors • Basic cognitive skills (CFT IQ) a 1.55 * (.59) 1.55 * (.71) • Phonological awareness a • Working memory (DSF) a • Executive control (Simon Task) a 175 Cognitive and linguistic profiles in early foreign language vocabulary and grammar <?page no="176"?> BPVS - grade 3 Category fluency - grade 3 Personal factors • Gender female (reference= male) • Age (months) a Linguistic factors • Grade 3: Category fluency German a 1.33 * (.53) 1.83 ** (.63) Parent unit nonresponse (ref. = parent data available) Social factors • SES (HISEI) a -.37 (1.00) .08 (.93) • Education parents (years in school) a .17 (.95) .17 (.90) • Net household income a .44 (.51) -.04 (.79) • Cultural capital parents (number of books in German) a -.34 (.51) -.37 (.48) N 85 70 b 85 70 b -2 Restricted Log Likelihood 510.859 c 412.054 498.400 c 405.526 Number of parameters 4 8 4 8 Dependent Variable: BPVS - grade 3; category fluency - grade 3; notation: unstandar‐ dised estimates; standard error in parentheses/ controlled for sign. random effects; type of covariance: variance components a standardised b only cases with full information from parents; missing values ML-estimated (EM-al‐ gorithm) c model optimised via chi²-comparison of -2 Restricted Log Likelihood; method: exclusion, stepwise + p < .10, * p < .05, ** p < .01, *** p < .001 Table 5: Predictors of English receptive vocabulary (BPVS) and English productive vo‐ cabulary (category fluency) in grade 3 on institutional and individual level (hierarchical linear mixed regression); monolinguals only; controlled for school. 176 Holger Hopp / Teresa Kieseier / Markus Vogelbacher / Dieter Thoma <?page no="177"?> Category fluency - grade 4 TROG-2 - grade 4 Model 3a Model 3b Model 4a Model 4b Constant term 12.12 *** (.46) 12.18 *** (.62) 44.91 *** (.72) 44.52 *** (.99) Institutional level Proportion of bilingual students a Mean SES school a 3.69 *** (.74) 4.24 *** (1.09) Individual level Cognitive factors • Basic cognitive skills (CFT IQ) a 2.85 *** (.67) 2.84 ** (.88) • Phonological awareness a 1.22 + (.72) .67 (.90) • Working memory (DSF) a 1.13 * (.55) 1.16 + (.65) • Executive control (Simon Task) a Personal factors • Gender female (reference= male) • Age (months) a Linguistic factors • Grade 4: Category fluency German a • Grade 3: Category fluency English a 3.58 *** (.50) 3.89 *** (.58) • Grade 4: Category fluency English a 2.45 *** (.66) 2.31 ** (.74) 177 Cognitive and linguistic profiles in early foreign language vocabulary and grammar <?page no="178"?> Category fluency - grade 4 TROG-2 - grade 4 Parent unit nonresponse (ref. = parent data available) Social factors • SES (HISEI) a .33 (.89) .61 (1.13) • Education parents (years in school) a .02 (.83) 1.13 (1.06) • Net household income a -.43 (.70) -.91 (.92) • Cultural capital parents (number of books in German) a -.19 (.45) -.13 (.58) N 79 66 b 79 66 b -2 Restricted Log Likelihood 444.638 c 369.753 471.688 c 386.716 Number of parameters 3 7 7 11 Dependent Variable: Category fluency - grade 4; TROG-2 - grade 4; notation: unstandardised estimates; standard error in parentheses/ controlled for sign. random effects; type of covariance: variance components a standardised b only cases with full information from parents; missing values ML-estimated (EM-al‐ gorithm) c model optimised via chi²-comparison of -2 Restricted Log Likelihood; method: exclusion, stepwise + p < .10, * p < .05, ** p < .01, *** p < .001 Table 6: Predictors of English productive vocabulary (category fluency) and English grammar (TROG-2) in grade 4 on institutional and individual level (hierarchical linear mixed regression); monolinguals only; controlled for school. For the bilingual students, Tables 7 and 8 display the final models, which differ less from the models for the full sample than the models for the monolingual students. Of note, the effects of L1 productive vocabulary are highly significant both for English receptive and productive vocabulary in grade 3 (Table 7). For English productive vocabulary in grade 4, however, German productive vocabulary has significant positive effects on English, with English productive vocabulary in grade 3 contributing further (Table 8). 178 Holger Hopp / Teresa Kieseier / Markus Vogelbacher / Dieter Thoma <?page no="179"?> BPVS - grade 3 Category fluency - grade 3 Model 1a Model 1b Model 2a Model 2b Constant term 40.30 *** (.86) 39.27 *** (1.27) 9.85 *** (.71) 9.78 *** (1.01) Institutional level Proportion of bilingual students a -2.18 *** (.54) -2.44 ** (.79) Mean SES school a 2.85 *** (.59) 3.67 *** (.87) Individual level Cognitive factors • Basic cognitive skills (CFT IQ )a 2.05 *** (.56) 1.36 + (.68) • Phonological awareness a 1.04 * (.45) .90 + (.52) • Working memory (DSF) a • Executive control (Simon Task) a Personal factors • Gender female (reference= male) 2.23 * (1.02) 3.44 * (1.31) 1.64 + (.84) 2.09 + (1.05) • Age (months) a Linguistic factors • Grade 3: Category fluency German a • Grade 3: Category fluency L1 a 2.67 *** (.69) 2.48 ** (.86) 2.98 *** (.56) 3.26 *** (.69) Parent unit nonresponse (ref. = parent data available) Social factors • SES (HISEI) a .20 (1.17) 1.30 (.95) • Education parents (years in school) a 1.12 (.99) .15 (.82) • Net household income a -1.36 (1.00 -.49 (.81) • Cultural capital parents (number of books in German) a -2.51 (1.84) -2.06 (1.53) N 109 79 b 109 79 b 179 Cognitive and linguistic profiles in early foreign language vocabulary and grammar <?page no="180"?> BPVS - grade 3 Category fluency - grade 3 -2 Restricted Log Likelihood 664.308 c 470.787 623.240 c 440.980 Number of parameters 6 10 6 10 Dependent Variable: BPVS - grade 3; category fluency - grade 3; notation: unstand. Estimates; standard error in parentheses/ controlled for sign. random effects; type of covariance: variance components a standardised b only cases with full information from parents; missing values ML-estimated (EM-al‐ gorithm) c model optimised via chi²-comparison of -2 Restricted Log Likelihood; method: exclusion, stepwise + p < .10, * p < .05, ** p < .01, *** p < .001 Table 7: Predictors of English receptive vocabulary (BPVS), English productive vocabu‐ lary (category fluency) in grade 3 on institutional and individual level (hierarchical linear mixed regression); bilinguals only; controlled for school. Category fluency - grade 4 TROG- 2 - grade 4 Model 3a Model 3b Model 4a Model 4b Constant term 10.87 *** (.41) 10.53 (.58) 47.32 *** (1.01) 47.46 *** (1.13) Institutional level Proportion of bilingual stu‐ dents a -4.43 *** (.90) -4.31 *** (1.10) Mean SES school a 1.08 * (.46) .58 (.66) Individual level Cognitive factors • Basic cognitive skills (CFT IQ) a • Phonological awareness a 2.16 ** (.70) 2.46 ** (.67) • Working memory (DSF) a 1.52 * (.72) .72 (.73) • Executive control (Simon Task) a .83 * (.36) 1.11 * (.49) Personal factors • Gender female (reference= male) 1.90 (1.30) 1.70 (1.41) • Age (months) a 180 Holger Hopp / Teresa Kieseier / Markus Vogelbacher / Dieter Thoma <?page no="181"?> Category fluency - grade 4 TROG- 2 - grade 4 Linguistic factors • Grade 4: Category fluency German a 1.77 *** (.43) 1.63 ** (.56) • Grade 4: Category fluency L1 a • Grade 3: Category fluency English a 2.25 *** (.45) 2.23 *** (.52) • Grade 4: Category fluency English a 3.29 *** (.75) 3.89 *** (.80) Parent unit nonresponse (ref. = parent data available) -3.58 * (1.54) Social factors • SES (HISEI) a .50 (.85) 1.19 (1.21) • Education parents (years in school) a -.02 (.74) -.25 (1.11) • Net household income a .18 (.76) .73 (1.06) • Cultural capital parents (number of books in German) a -1.26 (1.36) -2.15 (2.02) N 97 74 b 97 74 b -2 Restricted Log Likelihood 524.832 c 395.084 614.885 c 438.834 Number of parameters 6 10 8 11 Dependent Variable: Category fluency - grade 4; TROG- 2 - grade 4; notation: unstand. Estimates; standard error in parentheses/ controlled for sign. random effects; type of covariance: variance components a standardised b only cases with full information from parents; missing values ML-estimated (EM-al‐ gorithm) c model optimised via chi ² -comparison of -2 Restricted Log Likelihood; method: exclusion, stepwise + p < .10, * p < .05, ** p < .01, *** p < .001 Table 8: Predictors of English productive vocabulary (category fluency) and English grammar (TROG- 2) in grade 4 on institutional and individual level (hierarchical linear mixed regression); bilinguals only; controlled for school. 181 Cognitive and linguistic profiles in early foreign language vocabulary and grammar <?page no="182"?> In sum, the study shows a distinctive effect structure with institutional and per‐ sonal factors accounting for achievement in English vocabulary and grammar. In contrast, social factors at the individual level did not substantially contribute to foreign language performance. Among cognitive factors, all factors except for executive functions affected English achievement. However, the effect structure differed between vocabulary and grammar. Among the linguistic factors, L1 vocabulary was associated with English vocabulary in grade 3, while German vocabulary had significant relations with English vocabulary in grade 4. In addition to these cognitive and linguistic factors affecting the group of students as a whole, bilingualism was also associated with English skills, and, critically, the contributions of cognitive and linguistic factors to English achievement differed partially between monolingual and bilingual students. 4 Discussion In this study, we systematically assessed the relative impacts of cognitive and linguistic factors on achievement in English vocabulary and grammar in early foreign language learning in primary school. Both cognitive and linguistic factors were found to contribute to English skills. However, their relative impacts differed according to (a) linguistic domain, (b) mode, i. e. productive versus receptive skills, (c) stages in development and (d) group, i. e. monolingual versus bilingual students. In other words, these factors give rise to different profiles in early foreign language acquisition. In the following, we discuss the structure of effects according to these profiles in turn. 4.1 Linguistic domain In terms of cognitive factors, basic cognitive skills contributed to performance in receptive vocabulary, while phonological awareness contributed to perform‐ ance in productive vocabulary (see also Farnia & Geva, 2011). In this sense, both general and specific cognitive functions consistently facilitated foreign language vocabulary (Bohman et al., 2010). For grammar, too, both basic cognitive skills and phonological awareness were significantly associated with performance, yet working memory further contributed to grammar skills. This outcome may reflect demands imposed by tasks in specific linguistic domains because sentence comprehension taxes memory resources to a greater extent than the comprehension or production of single lexical items. Our finding from comprehension mirrors results from the production of complex sentences in child L2/ 3 learners of English (Paradis et al., 2017). 182 Holger Hopp / Teresa Kieseier / Markus Vogelbacher / Dieter Thoma <?page no="183"?> In terms of linguistic predictors, productive vocabulary of the first language contributed significantly to both receptive and productive vocabulary in Eng‐ lish. For the monolingual students, this effect was restricted to productive English vocabulary use. For the bilingual students, this L1 effect was pervasive as it indexes the effects of the heritage language productive vocabulary. These cross-linguistic effects of vocabulary highlight the interconnectedness of the multilingual lexicon. According to current models of multilingual lexical acquis‐ ition, such as the Parasitic Model (Ecke, 2015), novel lexical items in a foreign language link up to hosts from all previous languages which channel access to lemma and conceptual information (see also Kroll & Stewart, 1994, for a bilingual model). The present data suggest that the L1 exerts a privileged role in the early acquisition of foreign language vocabulary over and above the majority language German. This finding is all the more noteworthy in a context where German is the language often used for instruction or classroom management in primary schools. For grammar, however, linguistic effects were limited to within-language influences of English productive vocabulary on English receptive grammatical skills (see also Paradis et al., 2017). Such an asymmetry substantiates previous research on child and adult multilinguals showing that the degree of cross-lin‐ guistic connections differs between lexical and grammatical domains, with the bilingual lexicon being less language-selective than the Interlanguage grammar (e.g. De Houwer, 2009; Hopp, 2016; Kroll & Dussias, 2012). Finally, the effect sizes of linguistic factors outweigh those of cognitive factors in predicting both vocabulary and grammar. Therefore, linguistic factors have relatively stronger impacts than cognitive factors on early foreign language learning outcomes, irrespective of whether linguistic factors are across-language factors in grade 3 or within-language factors in grade 4. 4.2 Mode The effect structure of cognitive and linguistic factors also demonstrated dif‐ ferent profiles according to whether the task tapped into receptive or productive skills (see also Paradis & Jia, 2017). For receptive skills, basic cognitive skills were consistently found to have positive associations with performance, which likely reflects the role of problem-solving skills in untimed standardised receptive assessments. In contrast, in real-time production, such as the category fluency tasks, phonological awareness was the only cognitive skill contributing to English vocabulary. Even though the category fluency tasks do not require phonological segmentation and proceed by semantic association, a higher degree of phonological awareness implicates greater productive vocabulary size. 183 Cognitive and linguistic profiles in early foreign language vocabulary and grammar <?page no="184"?> More generally, this indicates that the ability to dissociate form and meaning acts as a resource in acquiring foreign language vocabulary in early foreign language acquisition (e.g. Jessner, 2008). At the same time, the directionality of the association is unclear, since a larger production vocabulary may lead to better performance on phonological awareness tasks, in particular as phonological awareness was measured using English lexical items in the present study. In any case, effects of phonological awareness were not limited to productive vocabulary but also surfaced for receptive grammar. This effect underscores that successful segmentation and decoding of words presented in sentential contexts is facilitated by higher degrees of phonological awareness. 4.3 Stages in development For productive vocabulary, we could assess developmental effects from grade 3 to grade 4. In terms of the linguistic factors, we found that while productive L1 vocabulary accounted positively for English vocabulary in grade 3, it was productive German vocabulary that had significant positive associations with English vocabulary in grade 4 when factoring in prior achievement in English vocabulary. This shift of linguistic predictors over time suggests that the status of German changes over time, either because students notice lexical correspondences between English and German more as proficiency in English rises, or because German is increasingly used in the foreign language classroom for translation or explanation in grade 4 when less frequent and more formal vocabulary items are introduced. In any case, these results resonate with previous findings that proficiency in the majority language affects foreign language achievement (e.g. Elsner, 2007) and that the influence of background languages changes in the course of schooling (Maluch et al., 2016). Moreover, both proportion of bilingual students and phonological awareness lose their status as significant predictors of productive vocabulary in grade 4 compared to grade 3. This may be interpreted as a levelling effect of school instruction. While foreign language teaching seems successful in balancing initial disadvantages in heterogeneous classrooms, the advantages of increased phonological awareness in grade 3 did apparently not receive enough support to maintain their facilitatory effects in grade 4. 4.4 Group Finally, the significant effects of bilingualism across all English measures indicate that monolingual and bilingual students differ in the way cognitive and linguistic factors impact foreign language achievement (see also Maluch & Kempert, 2017; Maluch et al., 2015). The separate models for the two groups 184 Holger Hopp / Teresa Kieseier / Markus Vogelbacher / Dieter Thoma <?page no="185"?> highlight important differences in the effect structure. The group of bilingual students showed significant effects of phonological awareness for English vocabulary, while the monolingual students did not. This asymmetry suggests that the effects of metalinguistic awareness in the present study were limited to bilingual students, with higher phonological awareness leading to gains in English vocabulary specifically in bilingual students. In this respect, it is important to note that the bilingual group of students did not overall have higher degrees of phonological awareness than the monolingual students (Table 2). Rather, it seems that the greater the phonological awareness is among the bilingual students, the more they develop a large L3 English production vocabu‐ lary. This contingency also holds for the size of the L1 production vocabulary, since the bilingual students showed a stronger effect of L1 vocabulary on English vocabulary than the German students did. Taken together, these effects suggest that the combination of greater L1 proficiency and higher metalinguistic awareness in bilingual students constitutes critical factors for achievement in a foreign language. Importantly, the role of group and the respective linguistic predictors for English vocabulary changes over time. When controlling for prior achievement in English vocabulary, the erstwhile positive effect of bilingualism on English vocabulary in grade 3 inverts to a negative effect in grade 4. In addition, rather than L1 vocabulary in grade 3, German vocabulary significantly predicts English vocabulary in grade 4. Of note, there is no comparable effect of German vocabulary on English vocabulary in grade 4 for the monolingual students (Table 6). Taken together, these effects suggest that proficiency in German becomes more important than L1 proficiency over time particularly for the bilingual students. This pattern of results points to the growing status of the majority language German for students with a minority L1 in foreign language learning (see also Maluch et al., 2016). Given the emphasis on the exclusive use of the foreign language in the primary school foreign language classroom (e.g. Böttger, 2012; Elsner, 2015; Legutke, Müller-Hartmann & Schocker-von Ditfurth, 2012), these results are unexpected. At the same time, they indicate that the reality in the classrooms is inflected by the use of the majority language. Such practices appear to put bilingual students with a language other than German as the L1 and comparatively lower German skills at a particular disadvantage in early foreign language learning. In sum, the present study finds that both cognitive and linguistic factors contribute to different profiles in vocabulary and grammar acquisition in early foreign language learning. The study thus highlights that both cognitive abilities and cross-linguistic influence modulate lexical and grammatical achievement 185 Cognitive and linguistic profiles in early foreign language vocabulary and grammar <?page no="186"?> in early foreign language acquisition. In turn, these factors also constitute resources in early foreign language teaching and can be used by teachers in foreign language teaching with a view to raising foreign language skills as well as overall academic achievement. References Adesope, O., Lavin, T., Thompson, T. & Ungerleider, C. 2010. A systematic review and meta-analysis of the cognitive correlates of bilingualism. Review of Educational Research, 80 (2), 207-245. Armon-Lotem, S., Walters, J. & Gagarina, N. 2011. Internal and external factors in child second language acquisition. Linguistic Approaches to Bilingualism, 1 (3), 291-317. Baddeley, A. D. 2007. Working Memory, Thought, and Action. Oxford Psychology Series: Vol. 45. Oxford: Oxford University Press. Bialystok, E. 2009. Bilingualism: The good, the bad, and the indifferent. Bilingualism: Language and Cognition, 12 (1), 3-11. Bishop, D.V.M. 2003. Test for Reception of Grammar: TROG-2: Manual (Version 2). London: Pearson. Blom, E. & Bosma, E. 2016. The sooner the better? An investigation into the role of age of onset and its relation with transfer and exposure in bilingual Frisian-Dutch children. Journal of Child Language, 43 (3), 581-607. Böttger, H. 2012. Englisch. Fachdidaktik für die Grundschule. Berlin: Cornelsen. Bohman, T.M., Bedore, L.M., Peña, E.D., Mendez-Perez, A. & Gillam, R.B. 2010. What you hear and what you say: Language performance in Spanish English bilinguals. International Journal of Bilingual Education and Bilingualism, 13 (3), 325-344. Brenders, P., van Hell, J.G & Dijkstra, T. 2011. Word recognition in child second language learners: Evidence from cognates and false friends. Journal of Experimental Child Psychology, 109 (4), 383-396. Brooks, P.J. & Kempe, V. 2013. Individual differences in adult foreign language learning: the mediating effect of metalinguistic awareness. Memory & Cognition, 41 (2), 281-296. Cenoz, J. 2013a. Defining multilingualism. Annual Review of Applied Linguistics, 33, 3-18. Cenoz, J. 2013b. The influence of bilingualism on third language acquisition: Focus on multilingualism. Language Teaching, 46 (1), 71-86. Chondrogianni, V. & Marinis, T. 2011. Internal and external factors in child second language acquisition. Linguistic Approaches to Bilingualism, 1 (3), 318-345. Chudaske, J. 2012. Entwicklungsbedingungen von Kindern mit Migrationshintergrund. In: Chudaske, J. (Ed.), Sprache, Migration und schulfachliche Leistung. Wiesbaden: VS Verlag für Sozialwissenschaften. 103-108. 186 Holger Hopp / Teresa Kieseier / Markus Vogelbacher / Dieter Thoma <?page no="187"?> De Houwer, A. 2009. Bilingual First Language Acquisition. MM Textbooks. Bristol: Multi‐ lingual Matters. Delis, D.C., Kaplan, E. & Kramer, J.H. 2001. Delis-Kaplan Executive Function System (D-KEFS). San Antonio, TX: The Psychological Corporation. Dijkstra, T. 2005. Bilingual word recognition and lexical access. In: Kroll, J.F. & de Groot, A.M.B. (Eds.), Handbook of Bilingualism. Psycholinguistic Approaches. New York, NY: Oxford University Press. 179-201. Dunn, L.M., Dunn, D.M., Styles, B. & Sewell, J. 2009. British Picture Vocabulary Scale: BPVS-III. 3 rd ed. London: GL Assessment. Ecke, P. 2015. Parasitic vocabulary acquisition, cross-linguistic influence, and lexical retrieval in multilinguals. Bilingualism: Language and Cognition, 18 (2), 145-162. Elsner, D. 2007. Hörverstehen im Englischunterricht der Grundschule: Ein Leistungsver‐ gleich zwischen Kindern mit Deutsch als Muttersprache und Deutsch als Zweitsprache. Frankfurt am Main: Lang. Elsner, D. 2015. Kompetenzorientiert unterrichten in der Grundschule: Englisch 1-4. Mün‐ chen: Oldenbourg Schulbuchverlag. Farnia, F. & Geva, E. 2011. Cognitive correlates of vocabulary growth in English language learners. Applied Psycholinguistics, 32 (4), 711-738. Ganzeboom, H.B., De Graaf, P.M. & Treiman, D.J. 1992. A standard international socio-economic index of occupational status. Social Science Research, 21 (1), 1-56. Garcia Mayo, M.d.P. 2012. Cognitive approaches to L3 acquisition. International Journal of English Studies, 12 (1). 129-146. Gathercole, V.C.M. 2002. Monolingual and bilingual acquisition: Learning different treatments of that-trace phenomenon in English and Spanish. In: Oller, D.K. & Eilers, R.E. (Eds.), Child Language and Child Development: Vol. 2. Language and Literacy in Bilingual Children. Clevedon: Multilingual Matters. 43-63. Gathercole, S.E., Hitch, G.J., Service, E. & Martin, A.J. 1997. Phonological short-term memory and new word learning in children. Developmental Psychology, 33 (6), 966-979. Gathercole, S.E., Willis, C.S., Emslie, H. & Baddeley, A. D. 1992. Phonological memory and vocabulary development during the early school years: A longitudinal study. Developmental Psychology, 28 (5), 887-898. Gathercole, V.C.M., Thomas, E.M., Roberts, E., Hughes, C. & Hughes, E. 2013. Why assessment needs to take exposure into account: Vocabulary and grammatical abilities in bilingual children. In: Mueller-Gathercole, V.C. (Ed.), Issues in the Assessment of Bilinguals. Bristol: Multilingual Matters. 20-55. Göbel, K., Vieluf, S. & Hesse, H.-G. 2010. Die Sprachtransferunterstützung im Deutsch- und Englischunterricht bei Schülerinnen und Schülern unterschiedlicher Sprachler‐ nerfahrung. Zeitschrift für Pädagogik, 55, 101-122. 187 Cognitive and linguistic profiles in early foreign language vocabulary and grammar <?page no="188"?> Haznedar, B. 1997. L2 acquisition by a Turkish-speaking child: Evidence for L1 influence. In: Hughes, E., Hughes, M. & Greenhill, A. (Eds.), Proceedings of the 21st Annual Boston University Conference on Language Development. Vol. 1. Somerville, MA: Cascadilla Press. 245-256. Hesse, H.-G., Göbel, K. & Hartig, J. 2008. Sprachliche Kompetenzen von mehrspra‐ chigen Jugendlichen und Jugendlichen nicht-deutscher Erstsprache. In: DESI-Kon‐ sortium (Ed.), Unterricht und Kompetenzerwerb in Deutsch und Englisch. Ergebnisse der DESI-Studie. Weinheim: Beltz (Verlags-)Lektorat, Empirische Bildungsforschung. 208-230. Hopp, H. 2016. Cross-linguistic lexical and syntactic co-activation in L2 sentence processing. Linguistic Approaches to Bilingualism, 7 (1), 96-130. Hopp, H., Kieseier, T., Vogelbacher, M. & Thoma, D. 2018. L1 effects in the early L3 acquisition of vocabulary and grammar. In: Bonnet, A. & Siemund, P. (Eds.), Foreign Languages in Multilingual Classrooms. Amsterdam: John Benjamins. 305-330. Hu, C.-F. 2003. Phonological memory, phonological awareness, and foreign language word learning. Language Learning, 53 (3), 429-462. Jessner, U. 2006. Linguistic Awareness in Multilinguals: English as a Third Language. Edinburgh: Edinburgh University Press. Jessner, U. 2008. Teaching third languages: Findings, trends and challenges. Language Teaching, 41 (1), 15-56. Juffs, A. & Harrington, M. 2011. Aspects of working memory in L2 learning. Language Teaching, 44 (2), 137-166. Kroll, J.F. & Dussias, P.E. 2012. The comprehension of words and sentences in two languages. In: Bhatia, T.K. & Ritchie, W.C. (Eds.), The Handbook of Bilingualism and Multilingualism. 2 nd ed. Chichester: Wiley-Blackwell. 216-243. Kroll, J.F. & Stewart, E. 1994. Category interference in translation and picture naming: Evidence for asymmetric connections between bilingual memory representations. Journal of Memory and Language, 33 (2), 149-174. Legutke, M.K., Müller-Hartmann, A. & Schocker-von Ditfurth, M. 2012. Teaching English in the primary school. 4 th ed. Uni-Wissen Anglistik, Amerikanistik. Stuttgart: Klett Lerntraining. Lenzing, A. 2013. The Development of the Grammatical System in Early Second Language Acquisition: The Multiple Constraints Hypothesis. Amsterdam: Benjamins. Linck, J.A., Osthus, P., Koeth, J.T. & Bunting, M.F. 2014. Working memory and second language comprehension and production: A meta-analysis. Psychonomic Bulletin & Review, 21 (4), 861-883. Maluch, J.T. & Kempert, S. 2017. Bilingual profiles and third language learning: The effects of the manner of learning, sequence of bilingual acquisition, and language use practices. International Journal of Bilingual Education and Bilingualism, 35 (2), 1-13. 188 Holger Hopp / Teresa Kieseier / Markus Vogelbacher / Dieter Thoma <?page no="189"?> Maluch, J.T., Kempert, S., Neumann, M. & Stanat, P. 2015. The effect of speaking a minority language at home on foreign language learning. Learning and Instruction, 36, 76-85. Maluch, J.T., Neumann, M. & Kempert, S. 2016. Bilingualism as a resource for foreign language learning of language minority students? Empirical evidence from a longitu‐ dinal study during primary and secondary school in Germany. Learning and Individual Differences, 51, 111-118. Marinis, T. & Chondrogianni, V. 2011. Comprehension of reflexives and pronouns in sequential bilingual children: Do they pattern similarly to L1 children, L2 adults, or children with specific language impairment? Journal of Neurolinguistics, 24 (2), 202-212. Miyake, A. & Friedman, N.P. 2012. The nature and organization of individual differences in executive functions: Four general conclusions. Current Directions in Psychological Science, 21 (1), 8-14. Molnar, T. 2010. Cognate recognition and L3 vocabulary acquisition. Acta Universitatis Sapientiae, Philologica, 2 (2), 337-349. Murphy, V.A. & Evangelou, M. 2016. Early Childhood Education in English for Speakers of Other Languages. London: British Council. Odlin, T. 1989. Language Transfer: Cross-linguistic Influence in Language Learning. Cambridge Applied Linguistics. Cambridge: Cambridge University Press. Oller, D.K., Pearson, B.Z. & Cobo-Lewis, A.B. 2007. Profile effects in early bilingual language and literacy. Applied Psycholinguistics, 28 (2), 191-230. Paradis, J. 2011. Internal and external factors in child second language acquisition. Linguistic Approaches to Bilingualism, 1 (3), 213-237. Paradis, J. & Jia, R. 2017. Bilingual children’s long-term outcomes in English as a second language: Language environment factors shape individual differences in catching up with monolinguals. Developmental Science, 20 (1). doi: 10.1111/ desc.12433. Paradis, J., Rusk, B., Duncan, T.S. & Govindarajan, K. 2017. Children’s second language acquisition of English complex syntax: The role of age, input, and cognitive factors. Annual Review of Applied Linguistics, 67, 1-20. Paradis, J., Tulpar, Y. & Arppe, A. 2016. Chinese L1 children’s English L2 verb morphology over time: Individual variation in long-term outcomes. Journal of Child Language, 43 (3), 553-580. Peets, K.F. & Bialystok, E. 2010. An integrated approach to the study of specific language impairment and bilingualism. Applied Psycholinguistics, 31 (2), 315-319. Petermann, F. & Petermann, U. 2008. HAWIK-IV. Kindheit und Entwicklung, 17 (2), 71-75. Poarch, G.J. & van Hell, J.G. 2012. Executive functions and inhibitory control in multi‐ lingual children: Evidence from second-language learners, bilinguals, and trilinguals. Journal of Experimental Child Psychology, 113 (4), 535-551. 189 Cognitive and linguistic profiles in early foreign language vocabulary and grammar <?page no="190"?> Rauch, D.P., Naumann, J. & Jude, N. 2012. Metalinguistic awareness mediates effects of full biliteracy on third-language reading proficiency in Turkish-German bilinguals. International Journal of Bilingualism, 16 (4), 402-418. Simon, J.R. 1969. Reactions toward the source of stimulation. Journal of Experimental Psychology, 81 (1), 174-176. Speciale, G., Ellis, N.C. & Bywater, T. 2004. Phonological sequence learning and short-term store capacity determine second language vocabulary acquisition. Applied Psycholinguistics, 25 (2), 293-321. Stanat, P., Rauch, D. & Segeritz, M. 2010. Schülerinnen und Schüler mit Migrationshin‐ tergrund. In: Klieme, E., Artelt, C., Hartig, J., Jude, N., Köller, O., Prenzel, M., Schneider, W. & Stanat, P. (Eds.), PISA 2009. Bilanz nach einem Jahrzehnt. Münster: Waxmann. 200-230. Szabo, C. 2016. Exploring the mental lexicon of the multilingual: Vocabulary size, cognate recognition and lexical access in the L1, L2 and L3. Eurasian Journal of Applied Linguistics, 2 (2), 1-25. Unsworth, S. 2013. Current issues in multilingual first language acquisition. Annual Review of Applied Linguistics, 33, 21-50. Unsworth, S. 2016. Early child L2 acquisition: Age or input effects? Neither, or both? Journal of Child Language, 43 (3), 608-634. Unsworth, S., Argyri, F., Cornips, L., Hulk, A., Sorace, A. & Tsimpli, I. 2014. The role of age of onset and input in early child bilingualism in Greek and Dutch. Applied Psycholinguistics, 35 (4), 765-805. Unsworth, S., Persson, L., Prins, T. & de Bot, K. 2015. An investigation of factors affecting early foreign language learning in the Netherlands. Applied Linguistics, 36 (5), 527-548. Verhoeven, L. 2007. Early bilingualism, language transfer, and phonological awareness. Applied Psycholinguistics, 28 (3), 425-439. Wanders, A. 2006. Die Rolle des Transfers. In: Pienemann, M., Keßler, J.-U. & Roos, E. (Eds.), Englischerwerb in der Grundschule. Ein Studien- und Arbeitsbuch. Paderborn: Schöningh. 97-109. Weber, J., Marx, P. & Schneider, W. 2007. Die Prävention von Lese-Rechtschreibschwier‐ igkeiten bei Kindern mit nichtdeutscher Herkunftssprache durch ein Training der phonologischen Bewusstheit. Zeitschrift für Pädagogische Psychologie, 21, 65-75. Weiß, R. 2006. CFT 20-R. Grundintelligenztest Skala 2. Manual. Göttingen: Hogrefe Verlag. 190 Holger Hopp / Teresa Kieseier / Markus Vogelbacher / Dieter Thoma <?page no="191"?> Part 3: Linguistic and cognitive abilities in secondary school children <?page no="193"?> Cognitive aspects of Processing Instruction Tanja Angelovska Abstract There is a consensus among SLA researchers that input plays a crucial role in the acquisition process. Likewise, there is no doubt that L2 learners do not attend to the whole input and that only the filtered input becomes available for their developing system. However, several questions arise as to what elements of the input learners attend to and what processing strategies underlie this comprehension process. The purpose of the present chapter is, on the one hand, to summarise the theoretical and empirical research on Input Processing, with a special emphasis on the cognitive aspects involved in Processing Instruction (an input-based grammar in‐ tervention that fosters form-meaning mapping). On the other hand, a study is presented which explores possible retention effects of Processing Instruction on the English -ed past tense marker across two different age groups (36 children and 13 adults with L1 German) and their performance rates at English sentence interpretation tasks with different cognitive task demands. The results indicate the same positive long-term effects of PI for both age groups, independent of the cognitive task demands. 1 Introduction SLA research from the 1970s was concerned with the negotiation and modifica‐ tion of input for learners (Long, 1983). In the 1980s, the notion of comprehensible input (Krashen, 1982, 2009) as automatically leading to learners comprehending the language input correctly was proposed. The emergence of the field of enquiry regarding “input processing” in SLA as such dates to the mid-1980s, dealing with more concrete questions such as: Which linguistic elements of the input are processed? What factors/ constraints guide that processing? Why do L2 learners seem unable to pick up elements of the input even when a sentence is relatively short and does not cause any working memory burdens? <?page no="194"?> Hence, a shift from a global focus on the role of input, in terms of negotiation of meaning, to a more psycholinguistics-oriented one, in terms of processing, was evident. Input Processing (IP) emerged as a field of enquiry to fill this gap from a micro-perspective. Since then, the research on L2 Input Processing and Processing Instruction (PI) has proliferated, resulting in numerous meth‐ odological and theoretical advancements (Benati & Lee, 2015; Lee & Benati, 2007; Lee & Benati, 2009). Several recent studies with reaction times (Angelovska & Roehm, 2020), visual moving window paradigm (Chiuchiù & Benati, 2020; Lee, Malovrh, Doherty & Nichols, 2020; Malovrh, Lee, Doherty & Nichols, 2020) or eye-tracking methodology (Benati, 2020a, 2020b; Ito & Wong, 2019; Lee & Doherty, 2019; Wong & Ito, 2018) confirm the effectiveness of PI, accounting for the moment-by-moment computations. 2 The Input Processing (IP) Model: Internal strategies and limited attentional resources Input Processing (IP) refers to “the initial process by which learners connect grammatical forms with their meanings as well as how they interpret the roles of nouns in relationship to verbs” (VanPatten, 2004: 5). He draws attention to the fact that input processing should not be viewed as a “model of acquisition per se” (VanPatten, 2012: 269) as IP deals with a certain and fixed set of principles used by learners during language comprehension and does not explain the production process. Nevertheless, IP is considered compatible with many other approaches and frameworks. The central foci of input processing are the psycholinguistic conditions un‐ derlying learners’ form-meaning mapping process. To understand why learners make certain connections in the input, and what internal learner-centred strategies are employed to make those form-meaning mappings, IP draws upon a set of principles. In other words, learners derive meaning from the input to which they are exposed. At the initial acquisition stages, it is rather likely that learners require a reasonable degree of mental effort, processing resources and demands when approaching the task of comprehension. Thus, learners rely on the easiest and efficient internal strategies to cope with the input that they are exposed to. Since learners know that the single words in a sentence carry meanings, they will approach the task of input processing by deriving meaning through interpreting utterances based on the lexical cues in the sentence. For example, in the sentence Yesterday John listened to music, L2 learners will rely on the lexical item (yesterday) to derive meaning, which will result in learners avoiding the redundant morphological marker for pastness (-ed). The two core 194 Tanja Angelovska <?page no="195"?> principles (Primacy of Meaning and The First Noun Principle, henceforth FNP) of input processing in L2 (cf. VanPatten, 2004) will be outlined below. Principle 1- Primacy of Meaning: Learners process input for meaning before they process it for form. a. The Primacy of Content Words Principle: Learners process content words in the input before anything else. This principle refers to the initial task that learners approach when attempting to comprehend the message. For example, L2 learners isolate words in the speech stream that they hear by connecting lexical items to their meanings and functions. What they do not and still cannot do at the initial stage of acquisition is to map a grammatical form to its meaning or function. This idea has already been substantiated in earlier psycholinguistic accounts (e.g. Faerch & Kasper, 1986). b. The Lexical Preference Principle: If grammatical forms express a meaning that can also be encoded lexically (e.g. when that grammatical marker is redundant), then learners will not initially process those grammatical forms until they have lexical forms to which they can match them (VanPatten, 2012: 271). Not all lexical items carry the same semantic information. ‘Meaningfulness’ refers to the relative communicative value that one form has (i.e. the overall meaning of the form that contributes to the compre‐ hension of the sentence. Two features operationalise the ‘communicative value’: [+/ -] semantic information and [+/ -] redundancy) (VanPatten, 2004). c. The Preference for Non-Redundancy Principle: Learners are more likely to process non-redundant meaningful grammatical forms before they process redundant meaningful forms. This principle states that learners have certain delays when processing redundant inflexions. Thus, learners must have processed and acquired the lexical items before they can acquire the corresponding grammatical markers. Previous SLA accounts have corroborated the idea of L2 learners being slower in the acquisition of redundant grammatical markers than of irredundant ones (e.g. Ellis, 1994). d. The Meaning-Before-Non-Meaning Principle: Learners are more likely to process meaningful grammatical forms before non-meaningful forms irre‐ spective of redundancy. e. The Availability of Resources Principle: For learners to process either re‐ dundant meaningful grammatical forms or non-meaningful forms, the pro‐ cessing of overall sentential meaning must not drain available processing resources. Proficiency level and learner’s ability to access lexical items in their mental lexicon influence the availability of the processing resources. Thus, if the sentence entails fewer unknown words to the learner, the 195 Cognitive aspects of Processing Instruction <?page no="196"?> learner will require fewer processing resources, which in turn would mean that the resources will be released for processing the grammatical form (VanPatten, 2004). f. The Sentence Location Principle: Learners tend to process items in sen‐ tence-initial position before those in final and medial position. This refers to the fact that elements appearing in the initial position of a sentence are more salient than those in the final position of a sentence. The final sentence items are more salient than the ones in the medial position (VanPatten, 2004). Applied to L2 English, the following grammatical features will be affected by the Primacy of Meaning Principle and its sub-principles: • Regular simple past form -ed. For example, in a sentence like John played football last week, the grammatical marker for pastness -ed will be redundant and as such not processed at first, because learners would rely on the lexical item (i.e. on the adverbial last week which also denotes meaning). Skipping the grammatical form -ed would mean that it has not been processed and thus, a form-meaning map cannot occur - thus, the learner relies on the Lexical Preference Principle. • Third-person singular -s. For example, in a sentence like Mary goes to school every day, -s is redundant because the adverbial (every day) is a lexical marker already denoting meaning. In such a sentence, the learner will rely on the lexical item to denote meaning (i.e. reliance on the Lexical Preference Principle). According to the Processability Theory in SLA (Pienemann & Keßler, 2011; Pienemann & Lenzing, 2015), this redundant form is rather processed at stage five (i.e. rather late, as stage five represents the last stage of morphological processing in acquisition). • English present subjunctive mood. The subjunctive mood denotes necessity, urgency and relevance. It is used after certain expressions, such as: It is essential that he study for his exams. This particular example refers to the present subjunctive form in English. The subjunctive mood is noticeable in the 3rd person singular, the verb ‘to be’, and the passive and formed in a subordinate clause that is tied to the main clause. The main clause includes lexical complements in various forms: adjectives (essential, important, im‐ perative, required, needed, etc.); nouns (requirement, advice, obligation, etc.); and verbs (recommend, demand, request, etc.). Learners of L2 English may face processing problems related to the first IP principle (Primacy of Meaning Principle) and especially with its two sub-principles (Lexical Preference Principle and the Sentence Location Principle). Because the subordinate clause 196 Tanja Angelovska <?page no="197"?> entails the subjunctive form and carries the same notion, the subjunctive marker is redundant and has a low communicative value. In contrast, the lexical items in the main clause, which carry the notion of relevance, urgency and need (i.e. the meaning of the mood), will be processed before the subjunctive marker can be processed. Likewise, the subjunctive is also prone to another sub-principle, namely the Sentence Location Principle, which refers to learners processing items in sentence-initial position before those in final position, and finally in medial position. The subjunctive marker is located within the subordinate clause in sentence medial (or final) position, which means that it is less likely that learners will process it before those meaning carrying lexical complements in the sentence-initial position (e.g. the adjectives, nouns or verbs denoting the mood). Principle 2- The First Noun Principle (FNP): Learners tend to process the first noun or pronoun they encounter in the sentence as the subject. a. The Lexical Semantics Principle: Lexical semantics of verbs may attenuate learners’ reliance on the First Noun Principle. This sub-principle refers to learners’ pre-knowledge of the meaning of the verbs prior to the task of processing. For example, if we take the verb bring, we can say that they know that the verb requires an agent that brings something (direct object). If two nouns were available to the verb bring, for example, teacher and books, learners will opt for teacher as the agent, because books are incapable of performing an action. b. The Event Probabilities Principle: Event probabilities may attenuate lear‐ ners’ reliance on the First Noun Principle. Like the first sub-principle, the Event Probabilities Principle refers to lexical semantics. Returning to the example with bring, we could safely assume that it is not possible that a book can bring a teacher. c. The Contextual Constraints Principle: Learners may rely less on the First Noun Principle if the preceding context constrains the possible interpreta‐ tion of the following clause or sentence. This means that the preceding context could weaken learners’ reliance on interpreting the first noun or pronoun in the sentence as an agent. The sub-principles of the FNP explain the fact that L2 learners are sometimes able to interpret sentences correctly by relying on the non-grammatical infor‐ mation (e.g. Yesterday), and not on the grammatical cues (e.g. -ed, cf. VanPatten, 2004). An example of a linguistic structure affected by the second IP principle (FNP) in English would be the passive construction. In a sentence like John was hit by 197 Cognitive aspects of Processing Instruction <?page no="198"?> Maria, learners might process John as the subject of the sentence and likely mis‐ interpret that John was the one who hit Maria. According to the FNP, learners process the first noun in a sentence as a subject. Such a misinterpretation would lead to a false mapping of syntactic structure to the meaning of the utterance and thus, cause a delay in acquisition. This phenomenon is also known as Subject Preference (see Lee, 2004, for the acquisition of English and O’Grady, 2011, for an overview). None of these principles function in isolation, but they may collude together to delay acquisition. 3 Processing Instruction (PI): The pedagogical application of IP (Input Processing) VanPatten’s model of Input Processing (1996, 2004, 2007, 2012, 2015 & 2017) directly informs the practices of Processing Instruction (PI), which is a grammar intervention in foreign language teaching. PI structured input activities help learners to map form (e.g. the grammatical marker for pastness -ed) and meaning (e.g. -ed denoting accomplished past events) - something learners would usually not do by default as they process input initially for meaning before they process it for form (Primacy of Meaning Principle). Processing Instruction is an effective input-oriented and meaning-based pedagogical intervention to grammar instruction attested across a variety of age groups and language combinations (see full reviews in Benati & Lee, 2015; Lee & Benati, 2009). The main purpose of PI is “to help learners circumvent ineffective processing strategies or to instil appropriate processing strategies, so that they derive better intake from the input” (Lee & Benati, 2007: 16). PI has often been called a ‘method’. However, as VanPatten pointed out (2015: 104): “PI is not a method. It is an intervention. As such, it is not meant to inform a curriculum but to be used by a communicative curriculum as necessary”. PI can be included in instructed settings when there is a processing problem in grammar. Processing Instruction (Lee & Benati, 2007; VanPatten, 2004) entails two main components: • Explicit information about a linguistic structure or form and the processing strategies underlying it. That information usually entails the information that the particular default processing strategy may work against the correct interpretation of the input. • Structured input tasks, which are input-oriented and which push the learners to process the form or structure. These tasks entail manipulated input so that learners must be dependent on the form or the structure to attend to the meaning of the sentence through processing, but not producing the 198 Tanja Angelovska <?page no="199"?> target structure or form. Structured input tasks are of two types, i. e. they are referential and affective. Referential activities are those in which the learner must rely on the grammatical form to get meaning, they entail isolated sentences or sentences embedded in a discourse that learners hear/ read and choose only one of two given answers (right or wrong). Affective structured input activities are those in which learners are engaged in processing information about the real world. The affective structured input activities usually entail a personal component whereby learners are asked to express an opinion, belief, or to give some other affective response. The goal of affective activities is to reinforce form-meaning connections established during referential structured input activities. Such structured input activities are developed by the following specific guidelines (Lee & VanPatten, 2003: 168): • Present one thing at a time; • keep meaning in focus; • move from sentences to connected discourse; • use both oral and written input; • have the learner do something with the input; and • keep the learner’s processing strategies in mind. Figure 1: A comparison of PI and traditional grammar instruction. 199 Cognitive aspects of Processing Instruction <?page no="200"?> In relation to other input-oriented techniques, PI is unique (apart from the accumulation of intake through form-meaning mapping) because of the nature of the structured input tasks, which require the learner to do something with the input while being exposed to it. This contrasts with traditional approaches (or traditional instruction, TI, see Figure 1 and Table 1), where the practice stage comes after the presentation of rules. Neither does PI entail any rule provision, nor is explicit information about the target feature necessary. Explicit information about the target feature is not a necessary component because “appropriately crafted structured input alone is sufficient to cause the changes in learner knowledge and performance” (VanPatten, 2012: 276). Table 1: Processing Instruction (PI) vs. Traditional Instruction (TI). In PI research, assessments usually entail auditory comprehension sen‐ tence-level interpretation tasks (and discourse-level tasks, such as video-narra‐ tions and text reconstructions, cf. Sanz & Morgan-Short, 2004; Sanz & VanPatten, 1998; VanPatten & Uludag, 2011) and controlled production tasks (mainly gap-filling tasks). VanPatten & Cadierno (1993) conducted the first classical quasi-experimental PI classroom study on the acquisition of object pronouns by second-year university-level English students attending Spanish classes. This study focused on the FNP and included three groups (two experimental groups and one control group). The two instructed groups differed in their type of instruction: The first group received traditional explicit instruction in object pronouns, the second group received PI on the same topic, and the control group did not receive any instruction on the target feature. VanPatten & Cadierno (1993) employed two assessment tasks: a sentence-level interpretation task and a sentence-level production task. Their findings showed that the TI group improved only on production measures and the PI group improved on the 200 Tanja Angelovska <?page no="201"?> interpretation and production tasks, thus outperforming the TI group. However, no significant group differences were found for the production task. The authors concluded that the type of instruction contributed to a correct form-meaning mapping. In a subsequent study, VanPatten & Wong (2004), who replicated Allen (2000), reported that the TI group improved because of conscious knowledge (i.e. explicit knowledge) of the target feature - which resulted in a successful outcome on the production task, but not on the interpretation one. As of today, the following conclusions may be drawn from existing research on PI: • PI is more effective than traditional instruction. Because of the positive effects of PI over traditional instruction for comprehension and production, recent PI studies (e.g. Angelovska & Benati, 2013; Benati & Angelovska, 2015) do not always include a comparison group instructed with another type of instruction (i.e. non-PI) if they are primarily interested in estab‐ lishing the positive effects of PI. • PI is effective irrespective of the learners’ age. Based on previous empirical findings (Benati, 2005; Benati, Lee & Houghton, 2008), Benati & Lee (2008) formulated the so-called Age Hypothesis for PI, according to which “Pro‐ cessing Instruction will be an effective intervention with younger learners as well as with adult learners” (Benati & Lee, 2008: 168). Angelovska & Benati (2013), Benati (2013), Benati & Angelovska (2015) and Mavrontoni & Benati (2013) reported positive effects of PI for younger learners. These studies used behavioural (off-line) measures and showed that PI is effective with both young (school children) and older learners (adolescent students). • PI is effective on both sentence and discourse-level tasks (interpretation and production in classroom and digital settings) and as evidenced by both off-line (e.g. paper and pencil tests) and on-line measures (e.g. eye-tracking data). • PI has primary (i.e. the ones resulting from the PI instruction itself on the target feature under question) and secondary effects (i.e. instructional effects on another non-target feature). A series of studies has shown that the instructional effects of PI on one structure can be transferred to another structure if it is affected by the same processing principle (cf. Benati & Lee, 2008). For example, both -ed and -s are affected by the same processing principle. PI has long-term retention effects sustained over a considerable amount of time (ranging from one week to eight months). 201 Cognitive aspects of Processing Instruction <?page no="202"?> 4 Cognitive aspects of Processing Instruction PI aims to circumvent inappropriate processing strategies and help learners map meaning and form correctly. Hence, the following cognitive aspects are relevant for PI, namely learner-related aspects, such as working memory capacity and prior language experience, as well as intervention-related cognitive aspects, such as cognitive task demands. 4.1 Working memory In general, all PI research studies have included besides the usual immediate post-test additional delayed post-test/ s. The period between the instruction and the delayed post-test/ s ranges from two weeks to eight months. Most PI studies included a two-week delayed post-test. To date, only one study has been conducted with long term effects of eight months after instruction (VanPatten & Fernández, 2004) - there is clearly a need to conduct additional PI studies that test durative effects. Research on working memory capacity (WMC) has shown that in L2 reading comprehension, WMC is more strongly associated with syntactic comprehen‐ sion in the L2 than in the L1 (Miyake & Friedman, 1998). Similarly, differences in WMC are known to be related to L2 processing (for a review see Michael & Gollan, 2005). The body of PI studies incorporating WMC as a measure is scarce (cf. Leeser & Sunderman, 2016; Santamaria & Sunderman, 2015; Sanz, Lin, Lado, Stafford & Bowden, 2016). Santamaria & Sunderman (2015) found that, compared to low-WMC participants, high-WMC participants had a better and longer-lasting learning effect in production tasks, but not in comprehension tasks. Interestingly, Sanz et al. (2016) conducted two experiments: [+/ grammar lesson + structured-input practice + explicit feedback]. The two experiments differed in the provision of rules and examples (e.g. in a traditional way) prior to PI practice. They found that the structured input activities were sufficient for form-meaning mapping in comprehension and production. In the experiment with the grammar lesson provided prior to practice, the chances for learners with higher WMC to outperform their counterparts decreased. They concluded that WMC could explain variation in comprehension and production only when grammar rules are not provided. 4.2 Prior language experience The effects of prior language experience (cf. Angelovska, 2017 for a review on the role of prior languages in L3 grammar acquisition) and learners’ continuous use of two or more languages has been shown to influence language and 202 Tanja Angelovska <?page no="203"?> cognitive abilities across the lifespan. Lee & McNulty (2013) examined the role of prior language experience and the effectiveness of PI. They found that PI appears to have “levelled-out the playing field” between learners with a more robust language repertoire and those with a native language other than English. What we do not know is whether “there is something special about PI” (Benati & Schwieter, 2017: 263). Fine-graded neurocognitive results based on so-called online measures, capturing moment-by-moment computations in processing, suggest that bilinguals with a lot of prior language experience show patterns for an additional language that are more similar to those of native speakers of languages (cf. Grey, Sanz, Morgan-Short & Ullman, 2017). However, no previous research has investigated the effects of PI with bilingual subjects (see also Benati & Schwieter, 2017) and it is still unclear whether PI can override any negative effects of processing strategies from typologically unrelated L2 and L3, which have been acquired subsequently in instructed settings. 4.3 Cognitive task demands In SLA research, the relationship between attentional capacity and L2 develop‐ ment has typically been studied by contrasting performance on tasks with different cognitive demands in different age groups. Such tasks may include written production tasks (Kuiken & Vedder, 2008; Ojima, 2006) and fluency and lexical complexity in writing (Ong & Zhang, 2010). The results of these studies seem to confirm Robinson’s view (2001, 2011) that higher task complexity might lead to higher L2 development. Task complexity refers to “the result of the atten‐ tional memory, reasoning, and other information processing demands imposed by the structure of the task on the language learner” (Robinson, 2001: 29). How‐ ever, other researchers (Foster & Skehan, 1996; Skehan, 1998) argue that higher cognitive task demands exhaust learner’s attentional capacity and decrease their ability to process language. Skehan’s Limited Attentional Capacity Model (1998) assumes that humans have limited information-processing capacity. According to Skehan (1998), the more cognitively complex a task is, the less likely a learner will have attentional resources to use for language processing. Révész, Sachs and Hama (2014) investigated adult English L2 learners who carried out simple tasks with lower reasoning demands or complex tasks with higher reasoning demands. The results indicated that the L2 learners were more successful under the simple task condition in oral production. Generally, inconsistent results are reported in previous studies measuring the extent to which task demands correlate with L2 development. 203 Cognitive aspects of Processing Instruction <?page no="204"?> 5 An exemplary study on cognitive task demands in PI To add to this line of research, the present study does not only compare age groups (Benati & Angelovska, 2015) but it also includes a second variable, namely cognitive task demands, with the aim to uncover possible interactions between PI effects and task demands. 5.1 The target feature The target feature (-ed) in this study was chosen because the use of this form is affected by a combination of different processing principles (The Lexical Preference Principle and the Preference for Non-Redundancy Principle). Both processing principles indicate that L2 learners will not process this redundant target form efficiently and appropriately when it co-occurs with a temporal adverb (e.g. last year) encoding the same semantic information. The English simple past tense marking poses an additional and potentially unique problem to L1 German learners of English, who are not always successful in distinguishing the present perfect (German, das Perfekt) and the simple past in English. This is due to the fact that the German Präteritum (simple past) is mostly reserved for written narrations (Rowley, 1983), but the German Perfekt is usually used in spoken language when it is used to express actions in the past. For example, the English sentence Mary cleaned the kitchen yesterday can be expressed in two ways in German: Mary hat die Küche gestern geputzt, or Mary putzte die Küche gestern (rarely used in oral communication). Therefore, a German learner of English will experience difficulties with mapping the simple past marker (-ed) in English to the past time framework, because “(L2 learners) may borrow the concept of past tense in their L1 as the starting point” (Benati, 2005: 76) or “L2 learners may be processing according to a strategy that has been developed during L1 acquisition” (Doughty, 2004: 262). Because in spoken German, present perfect tense expresses pastness, German learners of English could very likely end up with false mapping of the meaning ‘pastness’ to a verb in present perfect simple when presented with a target sentence in English containing a verb in present perfect. 5.2 Research questions and hypotheses In the attempt to examine the possible relationship between the age factor and cognitive task demands, the present study investigates whether two different age groups (children and adults) will benefit equally from PI. The following research questions are addressed: 204 Tanja Angelovska <?page no="205"?> 1 No control group was included because previous research (for an overview see Benati & Lee, 2008; Benati & Lee, 2015) from the large body on PI studies has proven the overriding effects of PI when compared to other types of input-based interventions (see Section 2). (Q1): Will Processing Instruction equally affect students in grade 5 and adult native speakers of German in their ability to process the English simple past regular tense -ed as measured by two interpretation sentence-level tasks with either high or low cognitive demands? (Q2): Will positive effects of Processing Instruction be retained over two weeks by both age groups on all tasks? Based on the results of the previous Processing Instruction research the following hypotheses are formulated: (H1): Both German children and adult learners of English will equally improve from pre-test to post-test, independent of whether the task is cognitively less demanding or cognitively more demanding. (H2): Both groups will equally retain the positive effects of Processing Instruction for both interpretation tasks over delayed post-tests administered two weeks after instruction. 5.3 Method 5.3.1 Participants The 49 participants belonged to two different age groups 1 , and all were native speakers of German: The 36 school-age learners were fifth-grade pupils with a mean age of ten and a half years, who attended the same secondary school in Southern Bavaria and who did not receive any extra-curricular courses in English. The second group consisted of 13 adults with a mean age of 26 years. Ethical approval to carry out this experiment was obtained, and all participants completed a consent form agreeing to take part in the experiment. Only L1 German native speakers without any knowledge of another foreign language than English and with no previous knowledge of the target form (-ed) were included in the final data pool consisting of the 49 subjects. Subjects were pre-tested on the ability to interpret and produce the target feature. To be included in the final data pool, participants had to score 60 % or lower on the pre-test’s battery (i.e. criterion used in all PI studies). All were tested with the Oxford Quick Placement Test, and they were all at an A1/ A2 language proficiency level. Learning was limited to classroom instruction and only learners exposed to all phases of the instructional treatment were included in the final data pool. 205 Cognitive aspects of Processing Instruction <?page no="206"?> 5.3.2 Procedure A pre-test/ post-test quasi-experimental design was used, whereby pre-tests were administered one week before the beginning of the instructional pe‐ riod (two consecutive days for a total of two hours of instruction for each group). During the instruction period, the same instructor (the researcher) in a classroom setting taught the groups. Paper and pencil post-tests were used immediately after instruction and two weeks after the instructional period. Short feedback from the instructor only indicated whether the participant’s response was correct or incorrect. No additional feedback was provided. 5.3.3 Instructional packet The same instructional treatment was used for both age groups. The material was developed following the original Processing Instruction guidelines (Lee & VanPatten, 1995; VanPatten & Sanz, 1995). The Processing Instruction treatment had the following characteristics: A) Presentation of the target feature simple past (-ed), where a sentence in isolation included the target feature, which was presented without any explicit information about what the target feature denotes. B) Use of referential and affective structured input activities in which learners have to respond to the content of sentences. The structured input activities in this instructional treatment were all communicative and meaningful, constructed in such a way to direct learners to attend to the form to com‐ plete the task (i.e. to understand the meaning of the target feature). As there were no lexical temporal indicators, learners had to use verbal morphology as the indicator of tense. The material included three referential and three affective structured input activities. High-frequent vocabulary (which occurred in learners’ EFL textbooks) was used in the activities. The teacher presented the instructions in each task in written English and translated them into German, on the one hand, to avoid possible comprehension difficulties (because participants were all beginning learners of English) and, on the other hand, to ensure that they had understood the task requirements correctly. Participants were allowed to ask questions in German related to the instructions to ensure that they did not have any comprehension difficulties with the task requirements. All participants were exposed to the same amount of explicit information (information about the processing strategy) and of structured input practice. 206 Tanja Angelovska <?page no="207"?> 5.3.4 Assessment and scoring procedures The cognitively less demanding interpretation task (Figure 2) was adopted from a previous study (Benati, 2005). The second (cognitively more demanding) interpretation task (see Figure 3) was developed especially for this experiment to account for the factor ‘cognitive task demands’. It is more complex than the first task because it included the distractors in the present perfect, which is used differently in German than in English. Examples of instructions and tasks are illustrated in Figures 2 and 3: First sentence-level interpretation task Listen to the sentences and decide. You will hear the sentences only once and you have 5 seconds to make the correct choice. Table 1. Descriptive statistics for first sentence-level interpretation task Table 3. Descriptive statistics for second sentence-level interpretation task LAST YEAR RIGHT NOW CAN’T TELL 1) ______ ________ ________ LAST YEAR RIGHT NOW CAN’T TELL 2) ______ ________ ________ LAST YEAR RIGHT NOW CAN’T TELL 3) ______ ________ ________ Sample sentences heard by the students: 1) I have traveled to Bristol. 2) I watch videos on my i-phone. 3) I ordered Chinese. Figure 2. Cognitively more demanding interpretation task LAST YEAR RIGHT NOW CAN’T TELL 1) ______ ________ ________ LAST YEAR RIGHT NOW CAN’T TELL 2) ______ ________ ________ LAST YEAR RIGHT NOW CAN’T TELL 3) ______ ________ ________ Sample sentences heard by the students: 1) I have traveled to Bristol. 2) I watch videos on my i-phone. 3) I ordered Chinese. Figure 3. Cognitively more demanding interpretation task PI (school age) ( n =36) 3.33 1.95 5.80 2.20 5.41 1.85 PI (adult) ( n =13) 4.15 2.03 5.92 2.53 5.53 2.66 Groups Pretest Posttest Delayed Posttest M SD M SD M SD PI school age ( n =36) 2.97 1.94 3.66 1.60 3.55 1.59 PI adult ( n =13) 3.76 2.08 6.07 2.43 6.06 1.03 Figure 2: Cognitively less demanding interpretation task adopted from Benati (2005). Second sentence-level interpretation task Read the questions before you listen to each of the sentences and then choose the correct answer. Try to choose the most suitable answer. You will hear the sentences only once. You have five seconds to read the question for each sentence and five seconds to make the correct choice. [Sample sentence heard by the students: John chopped his finger with a knife .] 1) Is John still in pain? YES ( ) NO ( ) CAN’T TELL ( ) [Sample sentence heard by the students: John lived in the States for 3 months .] 2) Is John still in the States? YES ( ) NO ( ) CAN’T TELL ( ) Figure 3: Cognitively more demanding interpretation task. The interpretation task included distractor sentences, which are known to pose difficulties in the interpretation of pastness. Temporal adverbs were removed in 207 Cognitive aspects of Processing Instruction <?page no="208"?> 2 Only such sentences in present perfect which carry the meaning that something started in the past and has continued up until now were included (e.g. I have had a cold for two weeks). Present perfect sentences expressing change over time, experience or actions, which happened at an unspecified time before now, were omitted. line with the FNP. All stimuli were pre-recorded by a native speaker of English and played for the subjects using a CD player. The questions aimed at testing learners’ cognitive abilities to map the form and meaning of the target verb. The on-task demands for the modified interpretation task are to match the grammatical form (-ed) with the meaning of the whole utterance (finished past action) without a given temporal adverbial as an option to choose from. Thus, learners have to attend to the meaning of the whole utterance by reasoning (i.e. to decide whether there is a possibility that the action, which started in the past may have continued in the present (present perfect 2 ) or the action was completed in the past). If the input sentence was in the simple past, the learner’s correct choice would be to exclude the possibility that the action has continued in the present. On the other hand, if the input sentence was given as a distractor, i. e. containing a verb in the present perfect, the learner should reason that the action, which began in the past, could/ may have continued in the present, as well. All tests were balanced in terms of difficulty and vocabulary, based on a pilot study. Only the target items were scored. The maximum possible score was ten for the first interpretation task and ten for the second interpretation task. Both interpretation tasks consisted of 20 sentences each, ten distractor items in the simple present or the present perfect tense and ten targeted forms (English simple past tense, only regular forms). The participants listened to the sentences and by ticking the appropriate item (e.g. last week vs. regularly) indicated (interpreted) whether the sentence they heard was related to a past or present action. No repetition was provided so that the test would measure real-time comprehension. The raw scores were calculated as follows: Zero points were assigned to incorrect responses and ‘can’t tell’ responses and one point to correct responses. No partial credits were given. 5.4 Results 5.4.1 Results from interpretation task 1 (cognitively less demanding) A one-way ANOVA conducted on the pre-test scores showed no significant differences between the two age groups before instruction (F (1, 48) = 1.645, p = .206). Both age groups began instruction with the same knowledge of 208 Tanja Angelovska <?page no="209"?> the target structure past tense -ed. Any differences which may be found after instruction may be attributed to the effects of instruction. Table 2 provides the mean scores and standard deviations for both age groups (adults and children) on the first sentence-level interpretation task for the first post-test and the delayed post-test. The results indicate an improvement from pre-test to post-test by the two age groups. However, the differences between the first post-test and the delayed post-test are extremely small for both groups. Groups Pre-test Post-test Delayed Post-test M SD M SD M SD PI (children) (n = 36) 3.33 1.95 5.80 2.20 5.41 1.85 PI (adult) (n = 13) 4.15 2.03 5.92 2.53 5.53 2.66 Table 2: Descriptive statistics for first sentence-level interpretation task. A two-way analysis of variances (ANOVA) with repeated measures was used to determine the effects of instruction, with the Age Group as the between-par‐ ticipants factor, and Time (pre-test, immediate post-test, and delayed post-test) as the within-participants factor. The two-way ANOVA revealed a significant main effect for Time (F (1, 48) = 21.438, p < .000). There were no significant effects for Age Group (F (1, 48) = .449, p = .506) and no significant interaction between Age Group and Time (F (1, 48) = .870, p = .356). Thus, these results demonstrate that both age groups improved in their ability to interpret English simple past tense forms as measured by a sentence-level interpretation task. In addition, these effects were retained over time. 5.4.2 Results from interpretation task 2 (cognitively more demanding) A one-way ANOVA was performed on the pre-test scores. The analysis did not yield any significant differences between the two age groups before instruction (F (1, 48) = 1.539, p = .221). Table 3 provides the mean scores and standard deviations for both age groups (adults and children) on the second sentence-level interpretation task. The scores show an improvement from pre-test to post-test scores by both age groups. 209 Cognitive aspects of Processing Instruction <?page no="210"?> Groups Pre-test Post-test Delayed Post-test M SD M SD M SD PI (children) (n = 36) 2.97 1.94 3.66 1.60 3.55 1.59 PI (adult) (n = 13) 3.76 2.08 6.07 2.43 6.06 1.03 Table 3: Descriptive statistics for second sentence-level interpretation task. Again, a two-way analysis of variance (ANOVA) with repeated measures was performed, with Age Group as the between-group factor and Time (pre-test, immediate post-test and delayed post-test) as the within-participants factor. The results of the ANOVA revealed a significant main effect for Time (F (1, 48) = 16.205, p < .000), for Age Group (F (1, 48) = 12.664, p < .001) and a significant interaction between Age Group and Time (F (1, 48) = 3.123, p < .000). The significant interaction indicates that there may be differential effects of the cognitive task demands for the different age groups. Tukey’s HSD post hoc tests confirmed that the scores differed significantly from pre-test to post-test 1 (immediate) and from pre-test to post-test 2 (delayed): PI (adult) > PI (school-age) (p < .000 and p < .000, respectively). The difference in scores from post-test 1 (immediate) and post-test 2 (delayed) was also significant: PI (adult) > (PI (school-age) (p = .004). Thus, both groups retained the effects of PI over time. 5.5 Discussion The results of this classroom experimental study provide affirmative answers to the two research questions of whether PI affects children and adult native speakers of L1 German in their ability to process the English simple past regular tense -ed in the same way and whether such effects can be retained. The analysis of the data collected through the interpretation tasks revealed three main results: First, both groups (L1 German 5 th graders and adults) improved equally from pre-test to post-tests in their ability to interpret the English simple past tense marker -ed for regular verbs at sentence-level. Second, the adults outperformed the children in the second interpretation task (which was cognitively more demanding than the first task). Third, the positive effects of PI were retained over time for both interpretation tasks by the two groups. The first hypothesis (H1) of this study dealt with the question of whether L1 German children and adult learners of English would equally improve from pre-test to post-test in the first interpretation task (cognitively less demanding) as in the second interpretation task (cognitively more demanding). The two 210 Tanja Angelovska <?page no="211"?> interpretation tasks (one reading and one listening) used in this study required both adults and children to choose the correct item while attending to the input. The learners’ improvement on the first interpretation task (25 %) was not as high as the 60 % in Benati’s study (2005). One reason for this difference may relate to the subjects’ age: Benati’s children were 12 to 13 years old and, therefore, may have had cognitive abilities at a higher maturational level than the learners in the present study with a mean age of only 10.5 years. In both studies, the results from the first sentence-level interpretation task show that PI has positive and equivalent effects on both age groups (children and adults). The findings from the second sentence-level interpretation task indicate that the adults showed more improvement than the children. This difference between the two age groups may be explained in terms of cognitive processing load (Skehan, 1998). In the sentence-level interpretation task, L2 learners were exposed to verb forms with ‘conflicting’ meanings, and this constituted a more demanding task for them. As VanPatten (2004: 22) states, “the issue of capacity (that is, limited resources) is not the same for everyone”. Thus, task demands and individual processing capacity can explain the results of the second task. In addition, maturity may have played a role, too, in that adults were more able to attend to both the -ed verb form and the auxiliary at the same time, whereas children may only have focused on the verb ending. This result seems to support the view that the more cognitively complex a task is, the less likely a learner will have attentional resources to use for language processing (Skehan, 1998; VanPatten, 2004). The second hypothesis (H2) in the present study dealt with the long-term effects of PI and examined whether children and adults would equally retain the positive effects of instruction for both interpretation tasks over a delayed post-test administered two weeks after instruction. Both age groups made similar improvements from the pre-test to the immediate post-test, and the positive effects of instruction were maintained at the delayed post-test. In sum, both groups showed the same positive effects of the instruction over time, regardless of the task complexity. 5.6 Summary The present study contributes to PI research because it provides further evidence that the acquisition of the English verbal morphology feature -ed for regular simple past tense is governed by a combination of processing principles. Moreover, the results of this study clearly showed that PI is an effective instructional intervention in helping L2 learners at different ages to establish accurate form-meaning connections in tasks which differ in cognitive demand. 211 Cognitive aspects of Processing Instruction <?page no="212"?> Despite these positive outcomes, there are several limitations. The first one is the limited number of participants and the imbalanced number of group members in the present study. A larger sample is necessary to corroborate the findings of this study. The second limitation of the present study is that long-term effects were measured after two weeks. Future studies should also examine whether long-term effects also hold after eight months (e.g. VanPatten & Fernández, 2004). The third limitation is that WMC scores for the two age groups were not assessed in this study. However, assuming that adults have indeed higher WMC scores than children may explain why they scored higher in the cognitively more demanding task than children did. The results yield important implications for teachers of English as a foreign language, who need to vary their task choices in relation to the different age groups and proficiency levels. 6 Conclusion and future directions Starting from the general position that L2 learners do not attend to the whole input they are exposed to, this chapter summarised research on Input Processing in SLA and focused on cognitive aspects involved in Processing Instruction - the pedagogical application of IP. Cognitive constructs, such as attentional resources, working memory, prior language experience, and cognitive task demands play an important role when learners process language input. The study in this chapter, therefore, dealt with retention effects of PI across two different age groups and their performance rates at interpretation tasks with cognitive task demands of two different degrees. Both learner-related and intervention-associated cognitive aspects offer ground for further research in the PI framework. Additional studies are war‐ ranted to examine cognitive aspects of Processing Instruction in more detail. For example, future PI studies should include measurements of WMC and struc‐ tured-input activities with differential cognitive task demands. Similarly, future projects need to address whether learners of L2 English (with no knowledge of other foreign languages) and trilingual learners (with different language backgrounds who use more than two languages daily) process grammatical input (English morphological forms and syntactic structures) differently. Finally, recent PI research employing online measures confirms the beneficial effects of structured input for correct form-meaning mapping across the board of target languages, age ranges and target features (Benati, 2020a, 2020b; Ito & Wong, 2019; Lee & Doherty, 2019; Wong & Ito, 2018). Despite all these reached milestones in PI research, it remains a puzzle why some “teachers 212 Tanja Angelovska <?page no="213"?> have not abandoned explicit grammar instruction as they have been advised to” (Larsen-Freeman, 2015: 265). It is about time that they did structured input practice. References Allen, L.Q. 2000. Form-meaning connections and the French causative. Studies in Second Language Acquisition, 22 (1), 69-84. Angelovska, T. 2017. Beyond instructed L2 grammar acquisition: Theoretical insights and pedagogical considerations about the role of prior language knowledge. Special Issue: Instructed L2 Grammar Acquisition and Beyond. Studies in Second Language Learning and Teaching, 7 (3), 397-419. Angelovska, T. & Benati, A. 2013. Processing instruction and the age factor: Can adults and school-age speakers of German process English simple past tense correctly? In: Lee, J. & Benati, A. (Eds.), Individual Differences and Processing Instruction. Sheffield: Equinox. 131-152. Angelovska, T. & Roehm, D. 2020. Processing instruction effects regardless of input modality and developmental processing constraints? A school lab classroom study on the morphosyntactic acquisition of L2 English. Instructed Second Language Acquisition, 4 (2), 180-202. doi.org/ 10.1558/ isla.40640. Benati, A. 2005. The effects of PI, TI, and MOI in the acquisition of English simple past tense. Language Teaching Research, 9 (1), 67-113. Benati, A. 2013. Age and the effects of processing instruction on the acquisition of English passive constructions among school children and adult native speakers of Turkish. In: Lee, J. & Benati, A. (Eds.), Individual Differences and Processing Instruction. Sheffield: Equinox. 83-104. Benati, A. 2020a. An eye-tracking study on the effects of structured input and traditional instruction on the acquisition of English passive forms. Instructed Second Language Acquisition, 4 (2), 158-179. Benati, A. 2020b. The effects of structured input and traditional instruction on the acquisition of the English causative passive forms: An eye-tracking study measuring accuracy in responses and processing patterns. Language Teaching Research, 1-21. doi: 10.1177/ 1362168820928577. Benati, A. & Angelovska, T. 2015. The effects of processing instruction on the acquisition of English simple past tense: Age and cognitive task demands. International Review of Applied Linguistics in Language Teaching, 53 (2), 249-269. Benati, A. & Lee, J. 2008. Grammar Acquisition and Processing Instruction: Secondary and Cumulative Effects. Bristol: Multilingual Matters. 213 Cognitive aspects of Processing Instruction <?page no="214"?> Benati, A. & Lee, J. (Eds.). 2015. Processing instruction: New insights after twenty years of theory, research and application. International Review of Applied Linguistics in Language Teaching (Special Issue), 53 (2), 87-275. Benati, A. & Schwieter, J. 2017. Input processing and processing instruction: Pedagogical and cognitive considerations for L3 acquisition. In: Angelovska, T. & Hahn, A. (Eds.), L3 Syntactic Transfer: Models, New Developments and Implications. Amsterdam: John Benjamins. 253-275. Benati, A., Lee, J. & Houghton, S. 2008. From processing instruction on the acquisition of English past tense to secondary transfer-of-training effects on English third person singular present tense. In: Benati, A. & Lee, J. (Eds.), Grammar Acquisition and Processing Instruction: Secondary and Cumulative Effects. Bristol: Multilingual Matters. 88-120. Chiuchiù, G. & Benati, A. 2020. A self-paced-reading study on the effects of structured input and textual enhancement on the acquisition of the Italian subjunctive of doubt. Instructed Second Language Acquisition, 4 (2), 235-257. Doughty, C. 2004. Effects of instruction on learning a second language: A critique of instructed SLA research. In: VanPatten, B., Williams, J. & Rott, S. (Eds.), Form-Meaning Connections in Second Language Acquisition. Mahwah, NJ: Lawrence Erlbaum. 189-212. Ellis, R. 1994. The Study of Second Language Acquisition. Oxford: Oxford University Press. Faerch, C. & Kasper, G. 1986. The role of comprehension in second language learning. Applied Linguistics, 7 (3), 257-274. Foster, P. & Skehan, P. 1996. The influence of planning time on performance in task-based learning. Studies in Second Language Acquisition, 18 (3), 299-323. Grey, S., Sanz, C., Morgan-Short, K. & Ullman, M. 2017. Bilingual and monolingual adults learning an additional language: ERPs reveal differences in syntactic processing. Bilingualism: Language and Cognition, 21 (5), 970-994. Ito, K. & Wong, W. 2019. Processing instruction and the effects of input modality and voice familiarity on the acquisition of the French causative construction. Studies in Second Language Acquisition, 41 (2), 443-468. Krashen, S. 1982. Principles and Practice in Second Language Acquisition. London: Per‐ gamon. Krashen, S. 2009. The comprehension hypothesis extended. In: Piske, T. & Young-Scholten, M. (Eds.), Input Matters in SLA. Bristol: Multilingual Matters. 81-94. Kuiken, F. & Vedder, I. 2008. Cognitive task complexity and written output in Italian and French as a foreign language. Journal of Second Language Writing, 17 (1), 48-60. Larsen-Freeman, D. 2015. Research into practice: Grammar learning and teaching. Language Teaching, 48 (2), 263-280. Lee, M-W. 2004. Another look at the role of empty categories in sentence processing (and grammar). Journal of Psycholinguistic Research, 33 (1), 51-73. 214 Tanja Angelovska <?page no="215"?> Lee, J. & Benati, A. 2007. Delivering Processing Instruction in Classrooms and Virtual Contexts: Research and Practice. London: Equinox. Lee, J. & Benati, A. 2009. Research and Perspectives on Processing Instruction. Berlin: Mouton de Gruyter. Lee, J. & Doherty, S. 2019. Native and nonnative processing of active and passive sentences: The effects of processing instruction on the allocation of visual attention. Studies in Second Language Acquisition, 41 (4), 853-879. Lee, J. & McNulty, E. 2013. The effects of language background on the results of processing instruction on the Spanish subjunctive/ indicative contrast after the adverb cuando. In: Lee, J. & Benati, A. (Eds.), Individual Differences and Processing Instruction. Sheffield: Equinox. 49-82. Lee, J. & VanPatten, B. 1995. Making Communicative Language Teaching Happen. New York, NY: McGraw-Hill. Lee, J. & VanPatten, B. 2003. Making Communicative Language Teaching Happen. 2 nd ed. New York, NY: McGraw-Hill. Lee, J., Malovrh, P. A., Doherty, S. & Nichols, A. 2020. A self-paced reading (SPR) study of the effects of processing instruction on the L2 processing of active and passive sentences. Language Teaching Research, 00 (0), 1-25. Leeser, M. & Sunderman, G. 2016. Methodological issues of working memory tasks for L2 processing research. In: Granena, G., Jackson, D.O. & Yilmaz, Y. (Eds.), Cognitive Individual Differences in Second Language Processing and Acquisition. Amsterdam: John Benjamins. 89-104. Long, M. 1983. Does second language instruction make a difference? TESOL Quarterly, 17 (3), 359-382. Malovrh, P., Lee, J., Doherty, S. & Nichols, A. 2020. A self-paced reading study of language processing and retention comparing guided induction and deductive instruction: Efficiency and depth of processing in computer-assisted learning. Instructed Second Language Acquisition, 4 (2), 203-234. Mavrontoni, M. & Benati, A. 2013. The effects of processing instruction and traditional instruction on two different school-age learners: The case of English present simple tense, third person singular. In: Lee, J. & Benati, A. (Eds.), Individual Differences and Processing Instruction. Sheffield: Equinox. 185-210. Michael, E.B. & Gollan, T.H. 2005. Being and becoming bilingual: Individual differences and consequences for language production. In: Kroll, J.F. & de Groot, A.M.B. (Eds.), Handbook of Bilingualism: Psycholinguistic Approaches. New York, NY: Oxford Univer‐ sity Press. 389-407. Miyake, A. & Friedman, N.P. 1998. Individual differences in second language proficiency: Working memory as language aptitude. In: Healy, A.F. & Bourne, L.E. (Eds.), Foreign 215 Cognitive aspects of Processing Instruction <?page no="216"?> Language Learning: Psycholinguistic Studies on Training and Retention. Mahwah, NJ: Lawrence Erlbaum. 339-364. O’Grady, W. 2011. Relative clauses: Processing and acquisition. In: Kidd, E. (Ed.), The Acquisition of Relative Clauses: Processing, Typology and Function. Amsterdam: John Benjamins. 13-38. Ojima, M. 2006. Concept mapping as pre-task planning: A case study of three Japanese ESL writers. System, 34 (4), 566-585. Ong, J. & Zhang, J. 2010. Effects of task complexity on the fluency and lexical complexity in EFL students’ argumentative writing. Journal of Second Language Writing, 19 (4), 218-233. Pienemann, M. & Keßler, J. (Eds.). 2011. Studying Processabilty Theory. An Introductory Textbook. Amsterdam: John Benjamins. Pienemann, M. & Lenzing, A. 2015. Processability theory. In: VanPatten, B. & Williams, J. (Eds.), Theories in Second Language Acquisition. 2 nd ed. New York, NY: Routledge. 159-179. Révész, A., Sachs, R. & Hama, M. 2014. The effects of task complexity and input frequency on the acquisition of the past counterfactual construction through recasts. Language Learning, 64 (3), 615-650. Robinson, P. 2001. Task complexity, task difficulty, and task production: Exploring interactions in a componential framework. Applied Linguistics, 22 (1), 27-57. Robinson, P. 2011. Second language task complexity, the Cognition Hypothesis, language learning, and performance. In: Robinson, P. (Ed.), Second Language Task Complexity: Researching the Cognition Hypothesis of Language Learning and Performance. Am‐ sterdam: John Benjamins. 3-38. Rowley, A. 1983. Das Präteritum in den heutigen deutschen Dialekten. Zeitschrift für Dialektologie und Linguistik, 50 (2), 161-182. Santamaria, K. & Sunderman, G. 2015. Working memory in processing instruction: The acquisition of L2 French clitics. In: Zhisheng, W., Borges Mot, M. & McNeil, A. (Eds.), Working Memory in Second Language Acquisition and Processing. Bristol: Multilingual Matters. 205-223. Sanz, C. & Morgan-Short, K. 2004. Positive evidence vs. explicit rule presentation and explicit negative feedback: A computer-assisted study. Language Learning, 54 (1), 35-78. Sanz, C. & VanPatten, B. 1998. On input and output practice in second language acquisition: A response to Salaberry. The Canadian Modern Language Review, 54 (2), 263-273. Sanz, C., Lin, H-J., Lado, B., Stafford, C.A. & Bowden, H.W. 2016. One size fits all? Pedagogical conditions and working memory capacity in early language development. Applied Linguistics, 37 (5), 669-692. 216 Tanja Angelovska <?page no="217"?> Skehan, P. 1998. A Cognitive Approach to Language Learning. Oxford: Oxford University Press. VanPatten, B. 1996. Input Processing and Grammar Instruction: Theory and Research. Norwood, NJ: Ablex. VanPatten, B. (Ed). 2004. Processing Instruction: Theory, Research, and Commentary. Mahwah, NJ: Lawrence Erlbaum. VanPatten, B. 2007. Input processing in adult second language aquisition. In: VanPatten, B. & Williams, J. (Eds.), Theories in Second Language Acquisition. Mahwah, NJ: Lawrence Erlbaum. 115-135. VanPatten, B. 2012. Input processing. In: Mackey, A. & Gass, S. (Eds.), Handbook of Second Language Acquisition. New York, NY: Routledge. 268-281. VanPatten, B. 2015. Input processing in adult SLA. In: VanPatten, B. & Williams, J. (Eds.), Theories in Second Language Acquisition. 2 nd ed. New York, NY: Routledge. 113-135. VanPatten, B. 2017. Situating instructed second language acquisition within second language acquisition: Facts about second language acquisition. Instructed Second Language Acquisition, 1 (1), 45-60. VanPatten, B. & Cadierno, T. 1993. Explicit instruction and input processing. Studies in Second Language Acquisition, 15 (2), 225-243. VanPatten, B. & Fernández, C. 2004. The long-term effects of processing instruction. In: VanPatten, B. (Ed.), Processing Instruction: Theory, Research, and Commentary. Mahwah, NJ: Lawrence Erlbaum. 273-289. VanPatten, B. & Sanz, C. 1995. From input to output: Processing instruction and communicative tasks. In: Eckman, F., Highland, D., Lee, P., Mileham, J. & Weber, R. (Eds.), Second Language Acquisition: Theory and Pedagogy. Mahwah, NJ: Lawrence Erlbaum. 168-185. VanPatten, B. & Uludag, O. 2011. Transfer of training and processing instruction: From input to output. System, 39 (1), 1-10. VanPatten, B. & Wong, W. 2004. Processing instruction vs. traditional instruction, once again: A study on the French causative. In: VanPatten, B. (Ed.), Processing Instruction: Theory, Research, and Commentary. Mahwah, NJ: Lawrence Erlbaum. 97-118. Wong, W. & Ito, K. 2018. The effects of processing instruction and traditional instruction on L2 online processing of the causative construction in French: An eye-tracking study. Studies in Second Language Acquisition, 40 (2), 241-268. 217 Cognitive aspects of Processing Instruction <?page no="219"?> Exploring the importance of prior knowledge and verbal cognitive abilities for foreign language learning Dominik Rumlich Abstract Domain-specific prior knowledge and verbal cognitive abilities represent two major determinants of school achievement across all subjects - yet they have often been neglected in empirical studies on foreign language learning. Against the background of a theoretical and empirical account of these two concepts, the data of a quantitative pre-post study with 155 participants at secondary schools in Germany will be evaluated with respect to the role they play in the development of English proficiency. Students’ general English as a foreign language (EFL) proficiency was measured with a series of C-Tests in a pre-test in year 6 (= prior knowledge; M age ≈ 12 years) and a post-test one year later (= final proficiency) together with their cognitive abilities (Kognitiver Fähigkeitstest; Heller & Perleth, 2000). The data shows that prior knowledge is the single most powerful predictor of students’ final general EFL proficiency; students’ verbal cognitive abilities are also highly relevant. Demographic factors only play a small role. The results, which make a strong claim for the incorporation of cognitive factors into studies on (general) EFL proficiency, will be discussed in the context of foreign language learning research. 1 Introduction and background Individual, i. e. person-related, factors have proven to be most decisive for secondary students’ learning success across all subjects in Germany and beyond (Fraser, Walberg, Welch & Hattie, 1987; Hattie, 2012: 25; Helmke & Weinert, 1997: 99; Schrader & Helmke, 2009: 494; Wang, Haertel & Wal‐ berg, 1993). On average, they explain about 50 % of achievement variance <?page no="220"?> 1 Studies on reading (comprehension) represent a notable exception; see, e. g. Barry & Lazarte (1998), Webb & Chang (2015) and the literature review in Carrell & Wise (1998). (Hattie, 2003: 1). Of these individual determinants of school achievement, cognitive factors can usually explain the largest amount of variance. Students’ domain-specific prior knowledge and verbal cognitive abilities (as an indicator of verbal intelligence) are the two most powerful (cognitive) predictors of school achievement in general (Baumert & Köller, 1998: 15; Helmke, 1992: 24; Helmke & Weinert, 1997: 106 f; Schiefele, Krapp & Schreyer, 1993: 121f) and of foreign language learning in particular (e.g. Dallinger, Jonkmann, Hollm & Fiege, 2016: 27 ff; Helmke, Schrader, Wagner, Nold & Schröder, 2008b: 244 f; Nold, 2003: 168). As students grow older, prior knowledge supersedes intelligence as the single most significant predictor (Köller & Baumert, 2008: 759). But whereas cognitive factors are frequently included in empirical research conducted in instructional psychology and the educational sciences, studies on foreign language learning incorporate them to a significantly lesser degree resulting in a substantial academic void in this area 1 . Since prior knowledge and verbal cognitive abilities are confounded with students’ affective-motivational dispositions such as academic self-concept and interest (Rumlich, 2016; Tobias, 1994), the failure to take the former into account is also prone to lead to an overestimation of the importance of affective-motivational characteristics and an underestimation with regards to cognitive variables for foreign language learning success (Rumlich, 2015). This might also be facilitated by the fact that motivation and effort (might also be thought to) become somewhat evident behaviour, i. e. are therefore easier to ‘see’ - which might seduce observers to draw ‘obvious’ causal inferences. The existing dearth of (recent) research on cognitive factors in foreign language learning despite their importance represents the major cause for this chapter. 2 Cognitive resources in foreign language learning The acquisition of one’s first language (L1) happens with “relative ease and rapidity and without the benefit of instruction” (White, 2003: 3). Basically, every child without any pathological conditions reaches a stage at which they possess functional control over their L1 (Spolsky, 1989: 104). Language learning is therefore often considered privileged learning, supported by nature and evolution so-to-speak (Bleyhl, 2007: 178). To what extent the principles and mechanisms of first language acquisition also apply to the learning of foreign languages at school is yet another matter due to fundamental differences between both processes (e.g. Cook, 2010; Ellis, 1994: 107). But the fact that (first) 220 Dominik Rumlich <?page no="221"?> language learning does have significant innate (brain) support - in contrast to learning in other school subjects such as mathematics, the sciences or the humanities - might be one major reason why cognition has traditionally played a less important role in research on language learning than in research on learning in other subjects. However, despite the paucity of recent research, there are numerous re‐ searchers and studies who have pointed to the paramount importance of cognitive factors in L2 learning: “When children begin the acquisition of a second language (L2), whether in the home or at school, their cognitive resources clearly play a central role in the rapidity and ultimate success with which that language is acquired” (Cummins, 1991: 70; cf. Genesee, 1976: 278; Robinson, 2001; Sang, Schmitz, Vollmer, Baumert & Roeder, 1986; Spolsky, 1989: 102 ff; Vollmer, 2001: 24). Apart from prior knowledge and verbal cognitive abilities/ intelligence, which represent two well-researched cognitive factors frequently considered in the area of educational sciences in general, the fields of applied lin‐ guistics and teaching EFL often also mention, among others, language (learning) aptitude. It denotes “strengths [that] individual learners have - relative to their population - in the cognitive abilities [that] information processing draws on during L2 learning and performance in various contexts and at different stages” (Robinson, 2005: 46). As such, aptitude refers to cognitive and conative dispositions as well as personality traits and denotes a largely predetermined ability or talent to learn languages (Robinson, 2005: 46). Hence, aptitude is a multi-faceted concept that draws on dispositions from different domains (cf. Schlak, 2008). Concerning its connection to intelligence, Skehan (1989: 110) notes that “intelligence and aptitude are related to one another, and that one can interpret language aptitude as consisting of specific components of intelligence which are especially relevant to learning situations.” Therefore, together with domain-specific prior knowledge, a more narrowly defined construct such as verbal cognitive abilities seem to be suitable for capturing a major share of the influence that intelligence-related individual differences exert on language learning. 2.1 Prior knowledge The concept of prior knowledge includes both declarative knowledge (“know what”) as well as procedural knowledge (“know-how”) and can be defined as the knowledge, skills and competences a person has in a specific area or domain (Renkl, 1996: 175); the word “prior” usually denotes the point of departure (t 0 ) before a learning episode. At the same time, there is no clear definition of domain; it is generally understood as the specific area in which (subsequent) 221 Exploring the importance of prior knowledge and verbal cognitive abilities <?page no="222"?> learning is supposed to take place. This is the major reason why the same or at least very similar preand post-tests are used to measure prior knowledge and final proficiency/ achievement in empirical studies. As students gradually develop their knowledge, skills, and competences during their school life, prior knowledge supersedes cognitive abilities/ intelli‐ gence as the most important single predictor of school achievement (Schrader & Helmke, 2008: 292) with correlation coefficients larger than .5, which equals more than 25 % shared/ explained variance. In other words: The specific knowl‐ edge, skills and competences students have with respect to what they are supposed to learn has a major effect on and hence predictive quality for their achievement. The great relevance of prior knowledge is also postulated by popular theories of learning such as constructivism (cf. e. g. Arnold, 1996: 31 ff; von Glasersfeld, 1995): “New understandings are constructed on a foundation of existing understandings and experiences” (Donovan & Bransford, 2005: 4). Learning is a continuous act of reorganising one’s thoughts and ideas based on new knowledge gained through sensual experiences and thought processes; it demands the active engagement of the learner with the learning matter (Mayer, 2004: 17). The successful establishment of connections between the old and the new is a key element for learning and remembering (Overmann, 2005: 26). At the same time, “prior knowledge, skills, beliefs, and concepts significantly influence what learners notice about their environment” (Cummins, Bismilla, Chow, Cohen, Giampapa, Leoni, Sandhu & Sastri, 2013: 4) with noticing as an important prerequisite for learning. The fact that prior knowledge, skills and competences, i. e. the status quo before a learning activity, is a key indicator for subsequent learning success also makes sense from a logical point of view: Prior knowledge as the starting point for a new learning episode simultaneously represents the end point of a previous (cumulative) learning episode. Hence, those who were successful and learnt a lot then obviously had what it took and acted upon it in a beneficial way so that one also states the obvious when saying that past achievement frequently possesses high predictive power for future achievement (Renkl, 1996: 177). In language subjects, prior knowledge also appears to be particularly relevant due to the cumulative nature of language learning (Köller, 1998: 56; Köller & Baumert, 2008: 740). In addition, it is important to take into account that language learning in EFL classes predominantly happens in and through the medium of the foreign language. This signifies that, to a considerable extent, students’ level of language proficiency also determines what they can potentially understand, which is one 222 Dominik Rumlich <?page no="223"?> 2 B = unstandardised regression coefficient; β = standardised regression coefficient; SE = standard error; p = p-value in significance testing. of the main determinants of learning in/ through a second or foreign language according to Cummins’ threshold hypothesis (1976; 1979). From an empirical point of view, the importance of domain-specific prior knowledge is substantial across all school subjects and has also generally been confirmed for EFL competences in different contexts (Dallinger et al., 2016: 27 ff; Helmke, Helmke, Schrader, Wagner, Nold & Schröder, 2008a: 384 f; Möller, Zaunbauer & Leucht, 2010: 222; Nold, 2000: 81; Nold, 2003: 178 f); it usually emerges as the most significant predictor (of all kinds of EFL achievement). Examining the predictive power of prior EFL listening skills at the beginning of year 9 for subsequent listening skills at the end of year 9, Helmke, Schrader, Wagner, Nold & Schröder (2008b: 254) report path coefficients of .81; in a study on “content and language integrated learning” (CLIL) over the course of year 8. Dallinger et al. (2016: 27 ff) also found that prior EFL listening skills and general EFL proficiency are the quintessential determinants of subsequent EFL listening skills (B = .31, SE = .03, p < .001) 2 and general EFL proficiency (B = .60, SE = .04, p < .001). In the same context, Rumlich (2016: 382) reports β-values of .40 (SE = .04, p < .001) for general EFL proficiency over a two-year span from the end of year 6 until the end of year 8. In the last three studies reported, prior knowledge exceeded the predictive power of any affective-motivational factor by far. 2.2 Cognitive abilities/ intelligence in general and verbal cognitive abilities/ verbal intelligence in particular So far, there is no uniform, unequivocal definition of intelligence. One of the most popular ones that was published in The Wall Street Journal in 1994 and agreed upon by 52 signatories (reprinted in the context of an editorial in Intelligence by Gottfredson, 1997: 13) reads: Intelligence is a very general mental capability that, among other things, involves the ability to reason, plan, solve problems, think abstractly, comprehend complex ideas, learn quickly, and learn from experience. It is not merely book-learning, a narrow academic skill, or test-taking smarts. Rather, it reflects a broader and deeper capability for comprehending our surroundings ‘catching on,’ ‘making sense’ of things, or ‘figuring out’ what to do. In order to establish a terminological foundation that a considerable number of experts in the area of intelligence would support, such a definition had to 223 Exploring the importance of prior knowledge and verbal cognitive abilities <?page no="224"?> remain broad and general at the expense of precision. In the following, further specifications, exemplifications and operationalisations of the construct are meant to elucidate the notion of intelligence from theoretical and empirical points of view. According to Helmke & Schrader (2001: 13553; cf. Gustafsson & Undheim, 1996), intelligence can also be “more or less equated with learning ability”: More intelligent students can adapt faster to new tasks, possess more efficient problem-solving strategies, identify rules relevant for the solution of problems more quickly and are equipped with higher processing capacities and more elaborate memory strategies (Hasselhorn & Gruber, 1997; Köller & Baumert, 2008: 759). Such a learning ability is also relevant in the context of incomplete, unclear and less than optimal lessons, from which more intelligent students can benefit to greater extents than less intelligent students (cf. Schrader & Helmke, 2010: 241): Low instructional quality forces students to fill in gaps for themselves, detect relations, infer key concepts, and develop their own strategies. On the whole, more intelligent students are able to recognise solution-relevant rules and to solve problems more quickly and more efficiently. Additionally, it is just that ability that helps them to acquire a rich knowledge base which is ‘more intelligently’ organised and more flexibly utilizable and so has an important impact on following learning processes (Helmke & Schrader, 2001: 13553). From these perspectives, it appears obvious that intelligence should be related to learning and achievement, which has been confirmed across numerous studies (Köller & Baumert, 2012: 651 f). Overall, the younger students are, e. g. at primary school, the better their intelligence predicts school achievement (Helmke & Weinert, 1997: 106; Köller & Baumert, 2008: 759) before it is overtaken by domain-specific prior knowledge. Independently of age, the influence of intelligence decreases with an increasing amount of prior knowledge that students acquire; in turn, the less prior knowledge there is, the more intelligence becomes predictive of learning and achievement again (Köller & Baumert, 2008: 759). From the introductory definitions, it has become clear that the construct of intelligence is multi-faceted, broad and abstract in nature with multiple definitions based on diverging epistemic convictions and academic traditions (Gruber & Stamouli, 2015: 26-30; Neisser, Boodoo, Bouchard, Boykin, Brody, Ceci, Halpern, Loehlin, Perloff, Sternberg & Urbina, 1996). The Berlin model of the structure of intelligence (Berliner Intelligenzstrukturmodell; Jäger, Süß & Beauducel, 1997) constitutes a recent and widely used hierarchical model, which 224 Dominik Rumlich <?page no="225"?> defines three sub-dimensions of intelligence represented by figural, numerical, and verbal cognitive abilities. In the context of research on children and adolescents in secondary education, these are the three dimensions considered in a frequently employed instrument in German contexts: The Kognitiver Fähigkeitstest (Heller & Perleth, 2000), short KFT, is a translated and adapted version of the Cognitive Abilities Test by Thorndike & Hagen (1971; 1993), which in turn is based on the Lorge-Thorndike-Intelligence-Test (Lorge, Thorndike & Hagen, 1964). It is geared towards measuring those abilities that are particularly relevant for learning at school (Heller & Perleth, 2000: 3). The verbal scales of the KFT are rooted in the construct of verbal intelligence as defined by Groffmann (1983 as indicated by Fehling, 2008: 110) and measure verbal comprehension and reasoning (sprachgebundenes Denken; Jäger, 1973) rather than mere word knowledge (Heller & Perleth, 2000: 3, 38, 45). In their overview on foreign language learning, Möller & Zaunbauer-Wo‐ melsdorf (2008: 594) underline that, in general, all types of students’ cognitive abilities are a strong predictor of their L2 achievement. A path model in the context of the DESI study with 15-year-olds (Klieme, 2008) showed that even students’ non-verbal, figural cognitive abilities are substantially related (path coefficient of .26) to their EFL achievement (Helmke et al., 2008b: 255; cf. Zaunbauer, Retelsdorf & Möller, 2009). The KFT manual reports correlation co‐ efficients of approximately .25 ((Gesamtschule comprehensive school) and .40 (Hauptschule - secondary modern school / Gymnasium grammar school) between students’ EFL grade and their verbal cognitive abilities. Correlations between numerical or figural abilities and EFL grades are between .09 and .31, coefficients between students’ overall ability score and their EFL grade are .24 (Gesamtschule), .36 (Hauptschule) and .38 (Gymnasium). Two large-scale studies with a focus on the evaluation of content and lan‐ guage integrated learning (CLIL) found that students’ verbal cognitive abilities constitute an important determinant of EFL proficiency in year 8 (Dallinger et al., 2016: 27 ff; Dallinger, Jonkmann & Hollm, 2018: 100; Rumlich, 2016: 382; see also Dallinger, this volume) and may potentially explain differential developments of achievement among students within a class (cf. Sang et al., 1986), but also between classes due to the resulting class environment (Dallinger et al., 2016; de Fraine, van Damme, van Landeghem, Opdenakker & Onghena, 2003). Dallinger et al.’s 2016 longitudinal CLIL study on general EFL proficiency (B = .12, SE = .03, p < .001) and listening comprehension (B = .14, SE = .03, p < .001), their 2018 cross-sectional CLIL study on general EFL proficiency (β = .19, SE = .03, p < .001) and listening comprehension (β = .18, SE = .03, p < .001) as well as Rumlich’s (2016) CLIL study on general EFL proficiency 225 Exploring the importance of prior knowledge and verbal cognitive abilities <?page no="226"?> (β = .19, SE = .02, p < .001) arrived at comparable results: They showed that verbal cognitive abilities make a substantial individual contribution to students’ EFL achievement even when the influence of prior knowledge is statistically taken into account, too. 3 The role of prior knowledge and verbal cognitive abilities for general language proficiency: A longitudinal study The above results indicate that prior knowledge, as well as verbal cognitive abilities, represent two individual cognitive dispositions that exert substantial influence on foreign language learning and might be responsible for catalytic effects: Students with higher verbal cognitive abilities (or prior knowledge) may be able to exploit English lessons to greater extents than students with lower verbal cognitive abilities (or prior knowledge) as suggested by the utilization of the learning opportunities model (Angebots-Nutzungsmodell; Helmke, 2012; English term by Helmke & Schrader, 2015). This might also explain differential effects among students within the same class, i. e. why students benefit to different extents from the same lessons and their proficiency develops signifi‐ cantly differently. Yet, outside of German CLIL contexts, domain-specific prior knowledge and verbal cognitive abilities have not been specifically examined in empirical foreign language learning research on general EFL proficiency - a significant overall indicator of language competence that all of the subskills tap into. Therefore, in the paper at hand, a sub-dataset from the DENOCS study (Development of North-Rhine Westphalian CLIL Students; Rumlich, 2016) will be analysed in order to address the following major research question: What influence do prior knowledge (general EFL proficiency in year 6) and verbal cognitive abilities have when trying to predict students’ general EFL proficiency in year 7? 4 Method and instruments DENOCS included a total of 1,403 students in 50 classes at German secondary schools (Gymnasium - grammar school and Realschule secondary modern school) in North-Rhine Westphalia, who were surveyed longitudinally with a questionnaire and language tests for the first time at the end of year 6 and then again one year later (from 2011 to 2012). As DENOCS was aimed at the evaluation of CLIL, the sample for the following sections will only include a subsample of seven classes (N total = 155) from two regular Gymnasiums without CLIL strands to avoid potential distortions due to CLIL-related influences. Since 226 Dominik Rumlich <?page no="227"?> 3 As the tests are protected by copyright, it is not possible to provide an item from the original test. However, the following examples illustrate the nature of the tests: V1 (lexicon): Aversion. Choose a word similar in meaning: peace - happiness - revulsion - foe - liking; V3 (analogies): Finger is to hand as toe is to ______. foot - face - thumb - body. Choose the answer that goes with the third underlined word the same way the first and second underlined words go together. the constructs in focus are general EFL proficiency and verbal cognitive abilities - as stated by the main research question - the two subsequent sections will be devoted to the statistical evaluation of the properties of the instruments to assure the quality of the data before it is analysed. All calculations of Rasch models were conducted with ACER ConQuest 3.0 (Wu, Adams & Wilson, 2007); SPSS Statistics (version 27) was used for all remaining analyses. 4.1 Statistical properties of the KFT (verbal cognitive abilities) As explained above, a validated and widely used German instrument to measure verbal cognitive abilities is the KFT (Heller & Perleth, 2000). Due to a large number of instruments in the test sessions and resulting time constraints, only the short version of the KFT with a total of 45 items from two scales was used (V1: lexicon, V3: analogies) 3 . The average correlation of the short version with the full three-scale version is ≥ .94 (Heller & Perleth, 2000: 12) and the short version is widely used for the given purpose. The scales have already undergone detailed statistical scrutiny in validation studies (cf. Heller & Perleth, 2000), but in order to check their properties in the context of the specific sample of the current study, the collected data were examined on the basis of descriptive statistics and a one-dimensional Rasch model. The mean difficulty of both scales is adequate (P V1 = 63.77, P V3 = 62.80; P Overall = 63.33) as is the difficulty of the individual items with the majority (75.56 %) between 20 < P < 80 (recommendation by Lienert & Raatz, 1998: 115). The remaining items display difficulties larger than 80, i. e. they constitute easy items, and usually function as ice-breakers at the beginning of the instrument. Of all participants, the lowest total number of correct answers is 13 and the highest 43, which also indicates preferable properties (Lienert & Raatz, 1998: 115). On the one hand, even the worst student answered a few items correctly; if this were not the case, their results would be likely to represent a distorted measure of their competence owing to an inappropriately high level of difficulty or a lack of motivation to sit the test, for instance. On the other hand, there was neither an item that remained unsolved nor did any student solve all of the items, which could allude to a lack of discriminatory power at the upper end of the competence continuum. 227 Exploring the importance of prior knowledge and verbal cognitive abilities <?page no="228"?> 4 EAP/ PV stands for “expected a posteriori/ plausible value” and is a measure of person separation reliability and discrimination; in simplified terms: how well/ accurately the test distinguishes people’s skills. The size of the values can be interpreted like those of Cronbach’s α, yet they tend to be slightly lower. A standard deviation of the scores of 6.01 indicates that there is a solid spread of the scores around the mean (M = 28.50, SE = 0.49). Still, more than 78 % of all scores fall into the interval between 22 and 35 points (approximately ± 1 SD around the mean), suggesting that the standard deviation is largely due to a certain number of outliers and the variability, in general, is lower than it seems at first sight. This lack of variability in students’ performance is also mirrored in the acceptable, but still unexpectedly low values of reliability of the person parameters reported below (Cronbach’s α and EAP/ PV reliability 4 ). The confidence interval of the mean is narrow, indicated by a small standard error of the mean, which implies that the estimate of the mean is trustworthy with rather high certainty. In a Rasch analysis, the 45 items of both tests exhibit satisfactory weighted mean square (MNSQ) values between .86 and 1.14. This is within the PISA criteria of 0.80 ≤ MNSQ ≤ 1.20 (Adams & Wu, 2002: 105). While the majority of the item discrimination coefficients is somewhat low for the scale V1 (.14 ≤ r pb ≤ .49), they are between .26 and .63 for V3 and thus acceptable, on average, for the entire short version of the test. The overall item separation reliability is high (.96), the reliability measures for the person estimates are acceptable (Cronbach’s α = .77; EAP/ PV reliability = .77). It appears that this is brought about by the small, rather homogenous sample, i. e. the number of participants is small and they display, on average, similar verbal cognitive abilities. This leads to comparatively low variance in the data (variance of the person estimates = .40), which reduces reliability and item discrimination since it becomes increasingly difficult to accurately distinguish between participants (due to their being similar in the first place). This is not a shortcoming of the test itself but more of a statistical artefact. Overall, the psychometric qualities of the verbal cognitive abilities test can be classified as good despite the reliability indices of the person estimates (which can nevertheless be assumed to be trustworthy); the data it elicited appears valid and reliable. As a result, there is ample reason to assume that participants’ performance on the tests used is a good estimate of their verbal cognitive abilities, so that further statistical analyses may be conducted on this data to obtain in-depth insights into the issues identified in the research question. Since the weighted likelihood person estimates (WLE) correlate very highly with the test takers’ sum scores of correct answers (one point for each correct answer; 228 Dominik Rumlich <?page no="229"?> 5 This is an uncontextualised exa_____ of wh_____ C-Tests lo_____ like. Usu_____, the fi_____ and la_____ two sent_____ do n_____ contain a_____ deletions. r = .94), which is a much more intuitive measure of students’ verbal cognitive abilities than the abstract Rasch estimate, all subsequent analyses will report students’ sum scores for the sake of simplicity and accessibility. 4.2 Statistical properties of the C-Tests (general EFL proficiency) A tool that has received a fair amount of scholarly attention and is frequently used to measure general EFL proficiency are C-Tests (Asano, 2014; Grotjahn, Klein-Braley & Raatz, 2002). Their results correlate substantially with overall TOEFL scores (Grotjahn, 2011: 135) as well as moderately to highly with different measures of language subskills, all of which tap into the same general dimension of general EFL proficiency (Eckes & Grotjahn, 2006). While lexical and grammatical competence are particularly conducive to C-Test performance, receptive skills seem to be less influential than productive skills (Eckes & Grot‐ jahn, 2006: 316). “A C-Test measures the ability to apply and integrate contextual, semantic, syntactic, morphological, lexical and orthographic information and knowledge pertaining to a particular written language” (Hastings, 2002: 66). C-Tests look like cloze tests and gap-fill exercises 5 , yet test-takers need to supply only the second half of every second word instead of entire words; they exploit the principles of reduced redundancy testing (Grotjahn, 2002: 211 f). Students’ general EFL proficiency in years 6 and 7 was measured with two validated C-Tests from the Hamburger Schulleistungstest (Hamburg scho‐ lastic achievement test; Behörde für Schule, Jugend und Berufsbildung. Amt für Schule, Hamburg, 1998; 2000), which had previously been used in other large-scale studies (e.g. KESS 7: Bos, Bonsen & Gröhlich, 2009; KESS 8: Bos & Gröhlich, 2010; LAU7: Lehmann, Gänsfuß & Peek, 2011). Due to low face validity and the unusual test format that students are often unfamiliar with, the task was called “Text research: A puzzle” (Text-Forschung: Ein Rätsel), and the instructions were rephrased to encourage student participation. They were given the role of an archaeologist who found incomplete or damaged ancient texts that they would need to complete in the best possible way on the basis of what was still legible. They solved a sample task with the entire group to illustrate what they were asked to do, but the main mechanism of the test that exactly half of the letters of each word (plus one in case of an even number of letters) had been deleted remained undisclosed. This was meant to enhance validity and reliability of the test data (Grotjahn, 2011: 134), as students would hopefully concentrate on the meaning of the words, contextual clues etc. while completing the task rather 229 Exploring the importance of prior knowledge and verbal cognitive abilities <?page no="230"?> than count letters or use other test strategies. In the scoring procedure, every fully correct and acceptable solution yielded one point; to enhance objectivity and prevent inconsistencies as to what counts as an error, even spelling mistakes should not be accepted (Grotjahn, 1995: 43). In both years, the two C-Tests that were used comprised about 25 items, amounting to a total of 50 items in each year. 100 items would have been desirable to secure high levels of reliability (Raatz & Klein-Braley, 1983), but this was not feasible due to time constraints and other important constructs that had to be included in the study. The use of well-established C-Tests of high psychometric quality, which performed well despite the incorporation of fewer than 100 items in previous studies (e.g. Bos et al., 2009; Bos & Gröhlich, 2010), appeared to be a feasible compromise without endangering the reliability of the results to a great extent. One of the tests constituted an anchor that was employed twice in order to be able to evaluate longitudinal changes in students’ proficiency scores on the basis of a one-dimensional Rasch model. In year 7, the other test was exchanged for a more challenging one to secure an appropriate level of difficulty and discrimination among students. As there was a one-year gap between both measurements and students did not obtain the solutions after the first test session, the influence of practice effects is bound to be minimal. Table 1 summarises the statistical (Rasch) properties of the C-Tests used in years 6 and 7. Along the veins of the above evaluations of the KFT scales, the figures below indicate good psychometric properties of the test items and the person estimates despite the small sample. Students’ lowest and highest scores (14 and 47 in year 6; 12 and 44 in year 7) indicate sufficient discrimination at the two ends of the proficiency spectrum. The average levels of difficulty of the tests are satisfactory (64.47 % of the items in year 6 and 59.52 % of the items in year 7 were solved correctly) as is their mean discrimination index (r pb = .32 in both years, 60 % in year 6 or 62 % in year 7, respectively, have an index equal to or larger than .30). The item fit is good (MNSQ within the PISA boundaries of 0.80 ≤ MNSQ ≤ 1.20; Adams & Wu, 2002: 105); the item separation reliability is high (.99). Internal consistencies of Cronbach’s α = .83 in year 6 or .82 in year 7, respectively, as well as EAP/ PV reliabilities of .83 (year 6) and .80 (year 7) can also be considered satisfactory and do not give rise to concern. 230 Dominik Rumlich <?page no="231"?> 6 The actual total number of participants is slightly lower in some of the analyses (as low as N = 141) due to individual missing values or longitudinal missings, i. e. students participated only in one of the two studies, which is a regular phenomenon in longitudinal research (especially over such long periods of time). As the number of total missings is small (max. 10 %) and missings appear to be random, the missings are no reason for concern. Correct answers Yr Min Max M P Item disc. 1 MNSQ sep. rel. 2 α 3 EAP/ PV 6 14 47 32.23 64.47 .07 - .49 0.88 - 1.16 .99 .83 .83 7 12 44 29.76 59.52 .13 - .49 0.90 - 1.12 4 .99 .82 .80 1 item discrimination (point-biserial correlation); 2 item separation reliability; 3 Cronba‐ ch’s α; 4 free estimation (if parameters are constrained, nine MNSQ values violate the PISA criteria of .80 ≤ MNSQ ≤ 1.20. However, as the respective parameters fall within the acceptable range when they are estimated freely, the violation reported is no reason for concern and the items can be kept (Pietsch, 2014, personal conversation). Table 1: Descriptive statistics and statistical properties (Rasch model) of the C-Tests. All in all, these evaluations support the conclusion that the instruments were suitable for the sample; the data they elicited appears valid and reliable. As a result, there is ample reason to assume that participants’ performance on the tests used is a good estimate of their general EFL proficiency, so that further statistical analyses may be conducted on this data to obtain in-depth insights into the issues identified in the research question. 4.3 The demographic setup of the sample (age, sex and L1) and their connection to C-Test and KFT scores The sample consists of seven classes with a total of N = 155 students 6 . The participants were asked to submit information on their age, sex (male/ female) and L1 background (German only/ no German/ German plus at least one other language) in a questionnaire. These demographic variables are commonly included in educational research in general and foreign language learning research in particular (cf. DESI: Klieme, 2008; KESS 7: Bos et al., 2009; KESS 8: Bos & Gröhlich, 2010; Zaunbauer et al., 2009) as they are potentially related to and confounded with the constructs under scrutiny. Hence, the composition of the sample could exert distorting influence on the results observed, which is why these three variables will now be examined more closely to obtain insights 231 Exploring the importance of prior knowledge and verbal cognitive abilities <?page no="232"?> into which variables need to be paid attention to in subsequent, more complex analyses. 51.5 % of the sample were male. In year 6, the mean age of the participants was 11.98 years (12.95 in year 7). 76.6 % solely speak German as L1, 11.7 % speak at least one second L1 in addition to German and 11.7 % do not speak German as L1. While students’ age turned out to be pretty much uncorrelated with students’ general EFL proficiency scores in year 6 (r = .06, p = .50) and year 7 (r = .03, p = .75), there is a tendency towards a significant correlation between students’ age and their KFT sum score (r = .13, p = .12). The association is generally weak, though. Table 2 displays the means (M), the standard errors of the mean (SE) and the standard deviations (SD) of students’ C-Test Rasch proficiency scores and their KFT sum scores according to sex (male/ female) and L1 setup. Rasch scores allow for direct longitudinal comparisons and were calculated so that the entire sample has a mean of zero in year 6; KFT sum scores are displayed as a raw sum score of correctly solved items. C-Test (yr 6) C-Test (yr 7) KFT (yr 7) Sub‐ group N M SE SD N M SE SD M SE SD Sex M 75 -0.07 0.09 0.81 80 0.44 0.10 0.86 29.28 0.69 6.13 F 79 0.08 0.10 0.86 75 0.68 0.08 0.71 27.68 0.69 5.95 L1 G 120 0.00 0.08 0.84 119 0.59 0.08 0.82 29.25 0.54 5.89 G- 16 0.14 0.23 0.90 18 0.35 0.18 0.77 26.78 1.55 6.59 G+ 18 -0.08 0.18 0.76 18 0.54 0.16 0.68 25.28 1.33 5.68 Total 154 0.00 0.07 0.84 155 0.56 0.06 0.80 28.50 0.49 6.08 Table 2: Descriptive statistics of C-Test Rasch proficiency scores and KFT scores ac‐ cording to sex and L1 background (G = German, G- = No German, G+ = German and at least one other language). Concerning students’ C-Test Rasch proficiency scores in year 6, it becomes clear that sexand L1-related difference are descriptively small (less than 20 % of a standard deviation) and not statistically significant (sex: t(152) = 1.10, p = .27, Cohen’s d = 0.18; L1: F(2, 152) = 0.31, p = .73, ω ² < .01). This situation prevails for L1 background in year 7 (F (2, 152) = 0.71, p = .49, ω ² < .01), but sex-related differences increase and display a tendency towards statistical significance (t(153) = 1.84, 232 Dominik Rumlich <?page no="233"?> p = .07, d = 0.30); the gap between girls (M = 0.68, SE = 0.08) and boys (M = 0.44, SE = 0.10) has widened over the course of year 7. While the variability of scores descriptively decreases among girls as they become more homogeneous (SD Yr6 = 0.86 vs. SD Yr7 = 0.71), it increases among boys (SD Yr6 = 0.81 vs. SD Yr7 = 0.86) as they become descriptively slightly more heterogeneous. Regarding KFT sum scores, one can detect statistically significant differences among students with different L1 setups despite the small group sizes (F (2, 152) = 4.35, p = .02, ω ² < .04). In comparison to monolingual German students (M = 29.25, SE = 0.54), students with German and an additional L1 (M = 25.28, SE = 1.55; t(135) = 2.68, p = .01, d = 0.68) scored significantly lower; a comparison with students without German as L1 (M = 26.78, SE = 1.33; t(135) = 1.84, p = .10, d = 0.41) merely approaches statistical significance. Sex-related differences display a tendency towards significant differences (t(153) = 1.64, p = .10, d = 0.27), but this time boys (M = 29.28, SE = 0.69) obtained higher scores than girls (M = 27.68, SE = 0.69). Overall, it becomes evident that age, sex, and L1 background only exert marginal influence on the two main constructs under scrutiny. One substantial connection exists between students’ L1 background and their KFT sum scores with another one emerging between sex and C-Test scores in year 7. 5 In-depth statistical analyses of C-Test and KFT scores A first statistical exploration of the three main variables of Rasch general EFL proficiency scores in year 6 (prior knowledge) and year 7 (final proficiency) and KFT sum scores shows that they are moderately to strongly related (Table 3). Pearson correlation coefficients among these variables are highly significant and range between .34 ≤ r ≤ .50, which corresponds to approximately 10-25 % shared variance (R ² ). As expected, the correlation is highest between prior knowledge and subsequent achievement (r = .50) and, presumably due to the time lag of one year, lowest between KFT sum scores and prior knowledge (r = .34). Variable KFT sum score Rasch C-Test year 7 Rasch C-Test proficiency scores year 6 (prior knowledge) .34 (p < .001) .50 (p < .001) KFT sum score 1 .43 (p < .001) Table 3: Correlations among main variables. 233 Exploring the importance of prior knowledge and verbal cognitive abilities <?page no="234"?> As suggested by the mean increase in overall Rasch C-Test scores from year 6 to 7 of .56 and the corresponding standard errors (Table 2, bottom line), a paired-sample t-test confirms this increase as highly statistically significant (t(141) = 8.32, p < .001) with a substantial effect size (d = 0.70). After these preliminary calculations, more complex analyses were computed in order to obtain a more detailed insight into the influence exerted by KFT sum scores (as a measure of cognitive abilities) and Rasch C-Test proficiency scores in year 6 (as a measure of prior knowledge/ general EFL proficiency) on Rasch C-Test proficiency scores in year 7 (as a measure of general EFL proficiency/ final achievement). For this reason, a stepwise regression analysis was performed ( *** p < .001). Table 4 includes Rasch C-Test proficiency scores in year 6, KFT sum scores and age (as the only one of the demographic covariates bordering on statistical significance). All of the common diagnostics for regression models indicate that the estimated model provides a satisfactory fit to the data (e.g. more than 10-15 cases per predictor; largest correlation among two variables equals .50; Durbin-Watson statistic = 2.12; 0.86 ≤ Tolerance ≤ 1.00; VIF max = 1.16, average VIF = 1.11; no standardised residuals > 2.50 and less than 5 % with absolute values > 2.00; Mahalanobis’ and Cook’s distances also within boundaries suggested by Field, 2013). Model 1 Model 2 Model 3 Variable B β SE p B β SE p B β SE p C-Test (yr 6) 0.51 .50 0.08 *** 0.39 .39 0.08 *** 0.38 .37 0.08 *** KFT 0.04 .32 0.01 *** 0.05 .34 0.01 *** Sex -0.26 -.16 0.11 .02 Adjusted R ² .24 .33 .35 *** p < .001 Table 4: Alternative regression models predicting students‘ Rasch C-Test scores in year 7 (indicating students’ general EFL proficiency). The stepwise regression procedure identifies Rasch C-Test proficiency scores in year 6 (prior knowledge) as a major predictor of Rasch C-Test proficiency scores in year 7 (β = .50 when no other predictor is included). Its regression coefficient drops to .37 when KFT scores (β = .34) are also included in the model, 234 Dominik Rumlich <?page no="235"?> making it a somewhat equally relevant predictor that also causes a substantial rise in explained variance (adjusted R ² ) from .24 to .33. Sex also turns out to be a relevant predictor (β = -.16) with female students (= 0) showing slightly better scores than male students. The change in R ² from model 2 (.33) to model 3 (.35) indicates a very small additional contribution of sex to the explanation of variance through the model. The above analysis of the two remaining demographic variables of age and L1 background suggested that they do not play a role in the prediction of students’ Rasch C-Test proficiency scores; hence they were automatically excluded by SPSS in the stepwise regression procedure. This is confirmed by the forced entry of these variables into the regression model: The β-values are close to 0 (or at least not statistically significantly different from 0, i. e. remain far from statistical significance) and only lead to a very small increase in the adjusted R ² value (.37). 6 Discussion and limitations The focus of this paper was the exploration of the importance of cognitive factors (prior knowledge and verbal cognitive abilities) for the prediction of students’ general EFL proficiency at the end of year 7. The above analyses and their implications will now be discussed in detail against the background of previous research and existing limitations of the study. The results corroborate the findings of other empirical studies that identified prior knowledge (e.g. Dallinger et al., 2016; Helmke et al., 2008a; Nold, 2000; Zaunbauer et al., 2009) and verbal cognitive abilities (e.g. Dallinger et al., 2016; Rumlich, 2016) as highly relevant for both more general and more specific EFL competences. While absolute effect sizes vary from study to study, also depending on the number and kind of additional factors included, prior knowledge usually plays the most important role as it did in the current study (cf. Schrader & Helmke, 2008; Helmke & Weinert, 1997; Wang et al., 1993) - often followed by affective-motivational factors with (verbal) cognitive abilities as a cognitive determinant somewhere in between (e.g. Dallinger et al., 2016; Helmke et al., 2008b; Zaunbauer et al., 2009). In the above models without any other (major) predictor, both cognitive variables almost seem on equal footing. But SPSS preferring prior knowledge as the first variable in a stepwise regression analysis, the high regression weight (β = .50) together with almost 25 % explained variance (R ² = .24) by just this one variable suggests a clear order of relevance. This is most likely also caused by the cumulative nature of language learning (Köller, 1998: 56; Köller & Baumert, 2008: 740). The high predictive power of prior knowledge in all of the above models is even more remarkable in view of the fact that it was measured one year 235 Exploring the importance of prior knowledge and verbal cognitive abilities <?page no="236"?> before students’ final EFL proficiency. It hence also represents a highly stable predictor over time as students’ longitudinal gains were substantial (d = 0.70) and in the middle range of what other studies reported for gains over one year in similar age groups (years 7/ 8: d ≈ 0.52, see Nikolova & Ivanov, 2010: 59; d ≈ 0.88, see Lehmann, Gänsfuß & Husfeldt, 2011: 398). The fact that a simple regression model including only the two predictors of prior knowledge and verbal cognitive abilities was able to explain 33 % of the variance in students’ general EFL proficiency (R ² = .33, model 2) makes a strong case for their inclusion into empirical research on language learning - especially when considering that, first, student-related factors account for 50 % of achievement variance on average (Hattie, 2003) and, second, even highly complex models of scholastic achievement with a broad variety of factors can rarely explain more than 50 % of the observed variance (Krapp, 1979; Schiefele et al., 1993; cf. Dallinger et al., 2016: 29, in a CLIL context: R ² = .55 for general EFL proficiency, but only R ² = .29 for listening comprehension despite the inclusion of 13 factors at student and class level). The shared variance of prior knowledge in the domain of general EFL proficiency and (German) verbal cognitive abilities can be explained by the fact that they partially tap into the same or similar competence dimensions, e. g. word knowledge ( Jude, 2008) and that cognitive abilities generally facilitate learning and (prior) achievement, i. e. function as its predictor. While a substantial amount of shared variance might not be desirable from a statistical and maybe also conceptual point of view (if the objective is to identify somewhat unique determinants of language competences), the interrelatedness of competences across (typologically closely related) languages is a default. But the importance of (German) verbal cognitive skills for general EFL proficiency in addition to prior knowledge strongly advocates their inclusion, which should not be ignored just for a methodological or conceptual sake striving for the independence or “purity” of the factors to be included. Since prior knowledge is an outcome of previous learning episodes whose success was determined by an entire array of influential factors - e. g. individual (cognitive, affective-motivational, volitional, constitutional), familial, social, institutional, teacherand teaching-related factors - it seems plausible that prior knowledge should by default be related to all potentially existing determinants of general EFL proficiency, no matter whether they were included in an analysis or not. This also underlines that its usual categorisation as a cognitive factor is slightly misleading and falls short of what it really is: a powerful overall factor. And while the exclusion of explicit affective-motivational factors represents a general limitation of the above analyses (despite the substantial amount of 236 Dominik Rumlich <?page no="237"?> variance already explained by the existing model), it can be argued that they were (indirectly or implicitly) included through prior knowledge. In order to examine potential distortions due to the composition of the sample, commonly influential demographic variables (sex, age, L1 background) and their relationships with the two predictors (prior knowledge, verbal cognitive abilities) and the central outcome variable of general EFL proficiency in year 7 were analysed. Approximately 20 % of the students are not monolingual German (G), i. e. do not speak German as L1 (G-; ~12 %) or speak German plus at least one other L1 (G+; ~12 %). While the even distribution of the L1 variable across the groups G+/ Gis positive, it has to be conceded that their absolute number (~18) is very small, which leads to a lack of statistical power to detect differences in proficiency scores that might actually exist (type II error). This can also not be ruled out by the results: Neither substantial descriptive differences in favour of Gin comparison to G (.14) and G (.22) at the end of year 6, nor substantially lower scores of Gin comparison to G (.24) and G+ (.19) result, as could be expected, in p-values anywhere close to statistical significance (also caused by the substantial within-group variation among Gand G+ as indicated by comparatively large standard errors and hence wide confidence intervals). In conclusion, these results should only be interpreted with great caution and, unfortunately, not much can be derived from them other than tendencies: Not speaking German as L1 coincides with lower EFL proficiency scores (over time; cf. Nikolova & Ivanov, 2010: 56) - as is the case with verbal cognitive abilities. However, differences are large enough to reach the level of statistical significance with regards to multilingual students (G+), who also display lower KFT scores. Due to the small sub-sample, even these results need to be interpreted with caution, but they are in line with findings from other empirical studies that report a smaller L1 lexicon in bilinguals (cf. Calvo & Bialystok, 2014: 285). The sample is almost balanced with regards to the male-female relation, which suggests that there is a substantial amount of statistical power to detect potentially existing differences. While they are negligible with respect to prior knowledge, they display a tendency towards statistical significance with respect to KFT scores (p = .10, d = 0.27; no differences reported in the norm studies; Heller & Perleth, 2000: 51 ff) and general EFL proficiency in year 7 (p = .07, d = 0.30), the latter in accordance with other studies that report similar effect sizes (e.g. d = 0.24 in Nikolova & Ivanov, 2010: 46; cf. Dallinger et al., 2016: 29; Helmke et al., 2008b: 255). Age does not seem to be related to general EFL proficiency scores, but a weak and small tendency towards higher KFT scores with age emerges from the data. But even though not all background factors 237 Exploring the importance of prior knowledge and verbal cognitive abilities <?page no="238"?> seem to exert influence, they might also be interrelated with each other, so that they should nevertheless be included in more complex analyses to account for potential influences. In addition, factors such as verbal cognitive abilities might mediate the influence of students’ L1 background on general EFL proficiency scores, for instance, which means that the differences in the latter are actually caused by differences in verbal cognitive abilities and are hence only indirectly related to students’ L1 background. The complex regression analysis shows that only sex emerges as a relevant demographic predictor of general EFL proficiency. So even when (differences in) prior knowledge and verbal cognitive abilities are considered, sex makes an individual contribution explaining some additional variance in EFL proficiency scores, which is in line with the large majority of studies on sex-related differences in EFL competence (e.g. Dallinger et al., 2016: 29; Helmke et al., 2008b: 255; Nikolova & Ivanov, 2010: 46). Above all, it has to be conceded that the small sample size also imposes constraints with respect to statistical procedures that can be used for the analysis of the data. The sample is not sufficient for large-sample techniques, e. g. structural equation modelling or multi-level regression analyses, that would allow for the consideration of the clustered nature of the sample (students, classes, schools) and lead to more accurate estimates. 7 Conclusion Despite certain shortcomings due to the relatively small sample, resulting issues with regards to demographic predictor variables and statistical procedures that could be used, the results of the present study still make a strong claim for the use of sufficiently complex statistical models and the incorporation of cognitive factors into studies on language learning and achievement. Prior knowledge and verbal cognitive abilities make substantial individual contributions to the prediction of students’ general EFL proficiency and are able to explain a major amount of variance despite their being related and sharing notable amounts of variance. If one variable has to be chosen over the other, e. g. due to constraints in an empirical study, the findings clearly lean to one side: As a result of the multi-dimensionality of domain-specific prior knowledge, it is highly advisable and economical to include it in research on language proficiency as it (indirectly) allows for the consideration of basically all factors that were influential in the past. 238 Dominik Rumlich <?page no="239"?> References Adams, R.J. & Wu, M. (Eds.). 2002. PISA 2000 Technical Report. Paris: Organisation for Economic Cooperation and Development (OECD). www.oecd.org/ pisa/ data/ 3368823 3.pdf. May 31, 2021. Arnold, R. 1996. Die Emergenz der Kognition. Skizze über Desiderata in der Erwachse‐ nendidaktik. Didaktisches Design, 1 (1), 31-42. Asano, Y. 2014. C-Tests und ‘allgemeine Sprachkompetenz’: Theoretische Überlegungen und empirische Analysen. In: Grotjahn, R. (Ed.), Der C-Test: Aktuelle Tendenzen. Frankfurt am Main: Peter Lang. 39-52. Barry, S. & Lazarte, A.A. 1998. Evidence for mental models: How do prior knowledge, syntactic complexity, and reading topic affect inference generation in a recall task for nonnative readers of Spanish? The Modern Language Journal, 82 (2), 176-193. Baumert, J. & Köller, O. 1998. Nationale und internationale Schulleistungsstudien: Was können sie leisten, wo sind ihre Grenzen? Pädagogik, 50 (6), 12-18. Behörde für Schule, Jugend und Berufsbildung. Amt für Schule, Hamburg (Ed.). 1998. Der Hamburger Schulleistungstest für sechste und siebte Klassen - SL-HAM 6/ 7. Hamburg. Behörde für Schule, Jugend und Berufsbildung. Amt für Schule, Hamburg (Ed.). 2000. Der Hamburger Schulleistungstest für achte und neunte Klassen - SL-HAM 8/ 9. Hamburg. Bleyhl, W. 2007. Bedingungen für einen optimalen Fremdsprachenunterricht. Ein Kri‐ terienkatalog für erfolgreiche Lernarrangements im Fremdsprachenunterricht. Erzie‐ hung und Unterricht, 157 (3-4), 174-183. Bos, W. & Gröhlich, C. (Eds.). 2010. KESS 8. Kompetenzen und Einstellungen von Schüler‐ innen und Schülern - Jahrgangsstufe 8. Münster: Waxmann. Bos, W., Bonsen, M. & Gröhlich, C. (Eds.). 2009. KESS 7. Kompetenzen und Einstellungen von Schülerinnen und Schülern an Hamburger Schulen zu Beginn der Jahrgangsstufe 7. Münster: Waxmann. Calvo, A. & Bialystok, E. 2014. Independent effects of bilingualism and socioeconomic status on language ability and executive functioning. Cognition, 130 (3), 278-288. Carrell, P.L. & Wise, T.E. 1998. The relationship between prior knowledge and topic interest in second language reading. Studies in Second Language Acquisition, 20 (3), 285-309. Cook, V.J. 2010. The relationship between first and second language learning revisited. In: Macaro, E. (Ed.), Continuum Companion to Second Language Acquisition. London: Continuum. 137-157. Cummins, J. 1976. The influence of bilingualism on cognitive growth: A synthesis of research findings and explanatory hypotheses. Working Papers in Bilingualism, 9, 1-43. Cummins, J. 1979. Linguistic interdependence and the educational development of bilingual children. Review of Educational Research, 49 (2), 222-251. 239 Exploring the importance of prior knowledge and verbal cognitive abilities <?page no="240"?> Cummins, J. 1991. Interdependence of first-and second language proficiency. In: Bia‐ lystok, E. (Ed.), Language Processing in Bilingual Children. Cambridge: Cambridge University Press. 70-89. Cummins, J., Bismilla, V., Chow, P., Cohen, S., Giampapa, F., Leoni, L., Sandhu, P. & Sastri, P. 2013. ELL students speak for themselves: Identity texts and literacy engagement in multilingual classrooms. http: / / learnteachlead.ca/ wp-content/ uploads/ 2013/ 08/ ELL identityTexts.pdf. May 31, 2021. Dallinger, S., Jonkmann, K., Hollm, J. & Fiege, C. 2016. The effect of content and language integrated learning on students’ English and history competences - Killing two birds with one stone? Learning and Instruction, 41, 23-31. Dallinger, S., Jonkmann, K., & Hollm, J. 2018. Selectivity of content and language integrated learning programmes in German secondary schools. International Journal of Bilingual Education and Bilingualism, 21 (1), 93-104. de Fraine, B., van Damme, J., van Landeghem, G., Opdenakker, M.-C. & Onghena, P. 2003. The effect of schools and classes on language achievement. British Educational Research Journal, 29 (6), 841-859. Donovan, M.S. & Bransford, J.D. 2005. How Students Learn: History, Mathematics, and Science in the Classroom. Washington, D.C.: The National Academy Press. Eckes, T. & Grotjahn, R. 2006. A closer look at the construct validity of C-tests. Language Testing, 23 (3), 290-325. Ellis, R. 1994. The Study of Second Language Acquisition. Oxford: Oxford University Press. Fehling, S. 2008. Language Awareness und bilingualer Unterricht: Eine komparative Studie. 2 nd ed. Frankfurt am Main: Peter Lang. Field, A.P. 2013. Discovering Statistics Using SPSS. 4 th ed. Los Angeles, CA: Sage. Fraser, B.J., Walberg, H.J., Welch, W.W. & Hattie, J.A. 1987. Syntheses of educational productivity research. International Journal of Educational Research, 11 (2), 145-252. Genesee, F. 1976. The role of intelligence in second language learning. Language Learning, 26 (2), 267-280. Gottfredson, L.S. 1997. Mainstream science on intelligence: An editorial with 52 signa‐ tories, history, and bibliography. Intelligence, 24 (1), 13-23. Groffmann, K.-J. 1983. Die Entwicklung der Intelligenzmessung. In: Groffmann, K.-J. & Michel, L. (Eds.), Enzyklopädie der Psychologie. Themenbereich B: Methodologie und Methoden. Serie II: Psychologische Diagnostik: Vol. 2. Intelligenz- und Leistungsdiag‐ nostik. Göttingen: Hogrefe. 1-103. Grotjahn, R. 1995. Der C-Test: State of the art. Zeitschrift für Fremdsprachenforschung, 6 (2), 37-60. Grotjahn, R. 2002. Konstruktion und Einsatz von C-Tests: Ein Leitfaden für die Praxis. In: Grotjahn, R. (Ed.), Der C-Test. Theoretische Grundlagen und praktische Anwendungen. Bochum: AKS. 211-225. 240 Dominik Rumlich <?page no="241"?> Grotjahn, R. 2011. C-Tests - Aspekte der Validität. Deutsch als Fremdsprache, 48 (3), 131-137. Grotjahn, R., Klein-Braley, C. & Raatz, U. 2002. C-Tests: An overview. In: Coleman, J.A., Grotjahn, R. & Raatz, U. (Eds.), University Language Testing and the C-test. Bochum: AKS. 93-114. Gruber, H. & Stamouli, E. 2015. Intelligenz und Vorwissen. In: Wild, E. & Möller, J. (Eds.), Pädagogische Psychologie. 2 nd ed. Berlin: Springer. 25-44. Gustafsson, J.-E. & Undheim, J. O. 1996. Individual differences in cognitive functions. In: Berliner, D.C. & Calfee, R.C. (Eds.), Handbook of Educational Psychology. New York, NY: Macmillan. 186-242. Hasselhorn, M. & Gruber, D. 1997. Entwicklung der Intelligenz und des Denkens. In: Weinert, F.E. & Helmke, A. (Eds.), Entwicklung im Grundschulalter. Weinheim: Beltz PVU. 13-41. Hastings, A.J. 2002. Error analysis of an English C-test: Evidence for integrated pro‐ cessing. In: Grotjahn, R. (Ed.), Der C-Test. Theoretische Grundlagen und praktische Anwendungen. Bochum: AKS. 53-66. Hattie, J.A. 2003. Teachers make a difference: What is the research evidence? Australian Council for Educational Research. https: / / research.acer.edu.au/ cgi/ viewcontent.cgi? a rticle=1003&context=research_conference_2003. May 31, 2021. Hattie, J.A. 2012. Visible Learning for Teachers: Maximizing Impact on Learning. London: Routledge. Heller, K.A. & Perleth, C. 2000. Kognitiver Fähigkeitstest (KFT) für 4.-12. Klassen, Revision. Göttingen: Beltz. Helmke, A. 1992. Determinanten der Schulleistung: Forschungsstand und Forschungs‐ defizit. In: Nold, G. (Ed.), Lernbedingungen und Lernstrategien. Tübingen: Narr. 23-34. Helmke, A. 2012. Unterrichtsqualität und Lehrerprofessionalität: Diagnose, Evaluation und Verbesserung des Unterrichts. 4 th ed. Seelze: Klett-Kallmeyer. Helmke, A. & Schrader, F.-W. 2001. School achievement: Cognitive and motivational determinants. In: Smelser, N.J. & Baltes, P.B. (Eds.), International Encyclopedia of the Social & Behavioral Sciences. Vol. 20. Amsterdam: Elsevier. 13552-13556. Helmke, A. & Schrader, F.-W. 2015. Angebots-Nutzungs-Modell der Wirkfaktoren aka‐ demischer Leistungen. In: Wirtz, M. A. (Ed.), Dorsch-Lexikon der Psychologie (online). Bern: Huber. https: / / portal.hogrefe.com/ dorsch/ angebots-nutzungs-modell-der-wirk faktoren-akademischer-leistungen. May 31, 2021. Helmke, A. & Weinert, F.E. 1997. Bedingungsfaktoren schulischer Leistung. In: Weinert, F.E. (Ed.), Enzyklopädie der Psychologie. Themenbereich D: Praxisgebiete. Serie 1: Päda‐ gogische Psychologie: Vol. 3. Psychologie des Unterrichtens und der Schule. Göttingen: Hogrefe. 71-176. 241 Exploring the importance of prior knowledge and verbal cognitive abilities <?page no="242"?> Helmke, A., Helmke, T., Schrader, F.-W., Wagner, W., Nold, G. & Schröder, K. 2008a. Wirksamkeit des Englischunterrichts. In: Klieme, E. (Ed.), Unterricht und Kompeten‐ zerwerb in Deutsch und Englisch. Ergebnisse der DESI-Studie. Weinheim: Beltz. 382-397. Helmke, A., Schrader, F.-W., Wagner, W., Nold, G. & Schröder, K. 2008b. Selbstkonzept, Motivation und Englischleistung. In: Klieme, E. (Ed.), Unterricht und Kompetenzerwerb in Deutsch und Englisch. Ergebnisse der DESI-Studie. Weinheim: Beltz. 244-257. Jäger, A.O. 1973. Dimensionen der Intelligenz. 3 rd ed. Göttingen: Hogrefe. Jäger, A.O., Süß, H.-M. & Beauducel, A. 1997. Der Berliner Intelligenzstruktur Test (BIS‐Test; Form 4). Göttingen: Hogrefe. Jude, N. 2008. Zur Struktur von Sprachkompetenz. Goethe-Universität-Frankfurt, Ger‐ many: Dissertation. http: / / publikationen.ub.uni-frankfurt.de/ files/ 6694/ Jude_Zur_Str uktur_von_Sprachkompetenz.pdf. May 31, 2021. Klieme, E. (Ed.). 2008. Unterricht und Kompetenzerwerb in Deutsch und Englisch: Ergebnisse der DESI-Studie. Weinheim: Beltz. Köller, O. 1998. Zielorientierungen und schulisches Lernen. Münster: Waxmann. Köller, O. & Baumert, J. 2008. Entwicklung schulischer Leistungen. In: Oerter, R. & Montada, L. (Eds.), Entwicklungspsychologie. 6 th ed. Weinheim: Beltz PVU. 735-768. Köller, O. & Baumert, J. 2012. Schulische Leistungen und ihre Messung. In: Schneider, W. & Lindenberger, U. (Eds.), Entwicklungspsychologie. 7 th ed. Weinheim: Beltz. 645-661. Krapp, A. 1979. Prognose und Entscheidung. Weinheim: Beltz. Lehmann, R.H., Gänsfuß, R. & Husfeldt, V. 2011. LAU 9. Aspekte der Lernerausgangs‐ lage und Lernerentwicklung - Klassenstufe 9. In: Behörde für Schule, Jugend und Berufsbildung (Ed.), LAU - Aspekte der Lernerausgangslage und der Lernerentwicklung. Klassenstufen 5, 7 und 9. Münster: Waxmann. 287-451. Lehmann, R.H., Gänsfuß, R. & Peek, R. 2011. LAU 7. Aspekte der Lernerausgangslage und Lernerentwicklung - Klassenstufe 7. In: Behörde für Schule, Jugend und Ber‐ ufsbildung (Ed.), LAU - Aspekte der Lernerausgangslage und der Lernerentwicklung. Klassenstufen 5, 7 und 9. Münster: Waxmann. 121-285. Lienert, G.A. & Raatz, U. 1998. Testaufbau und Testanalyse. 6 th ed. Weinheim: Beltz PVU. Lorge, I., Thorndike, R.L. & Hagen, E.P. 1964. The Lorge-Thorndike Intelligence Test. Boston: Houghton Mifflin. Mayer, R.E. 2004. Should there be a three-strikes rule against pure discovery learning? The case for guided methods of instruction. American Psychologist, 59 (1), 14-19. Möller, J., & Zaunbauer-Womelsdorf, A. C.M. 2008. Erwerb fremdsprachlicher Kompe‐ tenzen. In: Schneider, W. & Hasselhorn, M. (Eds.), Handbuch der pädagogischen Psychologie. Göttingen: Hogrefe. 587-596. Möller, J., Zaunbauer, A. C.M. & Leucht, M. 2010. Fremdsprachenerwerb. In: Rost, D.H. (Ed.), Handwörterbuch Pädagogische Psychologie. 4 th ed. Weinheim: Beltz. 219-225. 242 Dominik Rumlich <?page no="243"?> Neisser, U., Boodoo, G., Bouchard, T.J., Jr., Boykin, A.W., Brody, N., Ceci, S.J., Halpern, D.F., Loehlin, J.C., Perloff, R., Sternberg, R.J. & Urbina, S. 1996. Intelligence: Knowns and unknowns. American Psychologist, 51 (2), 77-101. Nikolova, R. & Ivanov, S. 2010. Englischleistungen. In: Bos, W. & Gröhlich, C. (Eds.), KESS 8. Kompetenzen und Einstellungen von Schülerinnen und Schülern - Jahrgangsstufe 8. Münster: Waxmann. 49-65. Nold, G. 2000. Ist schulisches Fremdsprachenlernen prognostizierbar? Überlegungen zum empirischen Forschungsprojekt “Lernstrategien zur Förderung sprachlicher Verstehensstrukturen in Englisch als Fremdsprache”. In: Aguado, K. (Ed.), Zur Method‐ ologie in der empirischen Fremdsprachenforschung. Baltmannsweiler: Schneider Verlag Hohengehren. 75-91. Nold, G. 2003. The impact of the learning culture on achievement in the English as a foreign language classroom - a view from Germany. In: Rymarczyk, J. & Haudeck, H. (Eds.), In Search of the Active Learner. Untersuchungen zu Fremdsprachenunterricht, bilingualen und interdisziplinären Kontexten. Frankfurt am Main: Peter Lang. 163-184. Overmann, M. 2005. Emotionales, transnationales, hyper-, tele- und multimediales Fremd‐ sprachenlernen. Frankfurt am Main: Peter Lang. Raatz, U. & Klein-Braley, C. 1983. Ein neuer Ansatz zur Messung der Sprachleistung. Der C-Test: Theorie und Praxis. In: Horn, R., Ingenkamp, K. & Jäger, R.S. (Eds.), Tests und Trends 3. Jahrbuch der pädagogischen Diagnostik. Weinheim: Beltz. 107-137. Renkl, A. 1996. Vorwissen und Schulleistung. In: Möller, J. & Köller, O. (Eds.), Emotionen, Kognitionen und Schulleistung. Weinheim: Beltz PVU. 175-190. Robinson, P. (Ed.). 2001. Cognition and Second Language Instruction. New York, NY: Cambridge University Press. Robinson, P. 2005. Aptitude and second language acquisition. Annual Review of Applied Linguistics, 25, 46-73. Rumlich, D. 2015. Zur affektiv-motivationalen Entwicklung von Lernenden im bilin‐ gualen Sachfachunterricht. In: Rüschoff, B., Sudhoff, J. & Wolff, D. (Eds.), CLIL Revisited: Eine kritische Analyse des gegenwärtigen Standes des bilingualen Sachfachun‐ terrichts. Frankfurt am Main: Peter Lang. 309-330. Rumlich, D. 2016. Evaluating Bilingual Education in Germany: CLIL Students’ General English Proficiency, EFL Self-concept and Interest. Frankfurt am Main: Peter Lang. Sang, F., Schmitz, B., Vollmer, H.J., Baumert, J. & Roeder, P.M. 1986. Models of second language competence: A structural equation approach. Language Testing, 3 (1), 54-79. Schiefele, U., Krapp, A. & Schreyer, I. 1993. Metaanalyse des Zusammenhangs von Interesse und schulischer Leistung. Zeitschrift für Entwicklungspsychologie und Päda‐ gogische Psychologie, 25 (2), 120-148. 243 Exploring the importance of prior knowledge and verbal cognitive abilities <?page no="244"?> Schlak, T. 2008. Fremdsprachenlerneignung: Tabuthema oder Forschungslücke? Zum Zusammenhang von Fremdsprachenlerneignung, Fremdsprachenlernen und Fremd‐ sprachenvermittlung. Zeitschrift für Fremdsprachenforschung, 19 (1), 3-30. Schrader, F.-W. & Helmke, A. 2008. Determinanten der Schulleistung. In: Schweer, M.K.W. (Ed.), Lehrer-Schüler-Interaktion. 2 nd ed. Wiesbaden: VS Verlag für Sozialwis‐ senschaften. 285-302. Schrader, F.-W. & Helmke, A. 2009. Erklärungsansätze für Schulleistung. In: Arnold, K.-H., Sandfuchs, U. & Wiechmann, J. (Eds.), Handbuch Unterricht. 2 nd ed. Bad Heilbronn: Klinkhardt. 493-496. Schrader, F.-W. & Helmke, A. 2010. Schulische Leistungen und individuelle Determi‐ nanten. In: Trolldenier, H.-P., Lenhard, W. & Marx, P. (Eds.), Brennpunkte der Gedächt‐ nisforschung. Göttingen: Hogrefe. 233-246. Skehan, P. 1989. Individual Differences in Second-Language Learning. London: Arnold. Spolsky, B. 1989. Conditions for Second Language Learning: Introduction to a General Theory. Oxford: Oxford University Press. Thorndike, R.L. & Hagen, E.P. 1971. Cognitive Abilities Test. Boston, MA: Houghton Mifflin. Thorndike, R.L. & Hagen, E.P. 1993. Form 5 Cog AT. Norms Booklet. Chicago, IL: Riverside Publishing Company. Tobias, S. 1994. Interest, prior knowledge, and learning. Review of Educational Research, 64 (1), 37-54. Vollmer, H.J. 2001. Zentrale Grundannahmen und Merkmale des Forschungsbereichs. Zeitschrift für Fremdsprachenforschung, 12 (2), 17-32. von Glasersfeld, E. 1995. Aspekte einer konstruktivistischen Didaktik. In: Landesinstitut für Weiterbildung (Ed.), Lehren und Lernen als konstruktive Tätigkeit. Soest: Bönen. 7-14. Wang, M.C., Haertel, G.D. & Walberg, H.J. 1993. Toward a knowledge base for school learning. Review of Educational Research, 63 (3), 249-294. Webb, S. & Chang, A. C.-S. 2015. How does prior word knowledge affect vocabulary learning progress in an extensive reading programme? Studies in Second Language Acquisition, 37 (4), 651-675. White, L. 2003. Second Language Acquisition and Universal Grammar. Cambridge: Cam‐ bridge University Press. Wu, M.L., Adams, R.J. & Wilson, M. 2007. ConQuest Manual. Camberwell: Acer Press. Zaunbauer, A. C.M., Retelsdorf, J. & Möller, J. 2009. Die Vorhersage von Englischleis‐ tungen am Anfang der Sekundarstufe. Zeitschrift Für Entwicklungspsychologie und Pädagogische Psychologie, 41 (3), 153-164. 244 Dominik Rumlich <?page no="245"?> The role of languages in bilingual History lessons and its effects on English and History achievement: Use of first language and correctness of foreign language Sara Dallinger (translated by Nina Rogotzki) Abstract Even though the field of Content and Language Integrated Learning (CLIL) has been the focus of research interest for a number of years, studies that examine CLIL-specific teaching methods are still rare. The present study analyses the effects of several language-related aspects on the competence development in English and History of 30 bilingually taught German eighth-grade classes (N = 703). Of particular importance were teachers’ comments on the frequency of first language use in several specific teaching situations, in particular when introducing new technical terms, and on teachers’ expectations of students’ correct foreign language use. Results showed that both an increased and systematic use of the first language for technical terms and high teacher expectations of correct foreign language use supported the development of English and History competences in CLIL classrooms. These results may lay the groundwork for the development of an as-yet largely non-existent, empirically based CLIL methodology. 1 Introduction Various models of bilingual teaching exist for the school context (Baker, 2006), including, for example Content and Language Integrated Learning (CLIL) where content subjects such as History or geography are taught by using a foreign language (see also Jaekel, this volume). CLIL can be adopted within the context of migration-related multilingualism, for instance by having students with a Turkish migration background attend Turkish-German classes as early <?page no="246"?> as in primary school (Gogolin, Neumann & Roth, 2009). In Germany, CLIL programmes very often combine the languages German and English as lingua franca (Nold, Hartig, Hinz & Rossa, 2008). Yet, does CLIL actually include German, i. e. the first language, or do the bilingual lessons rather turn out to be monolingual lessons taught in the foreign language? How much emphasis is truly placed on the students’ correct use of the foreign language? And how do these language-related aspects affect students’ development in English and in the content subject? Even though extensive research has been conducted on CLIL, these questions seem largely unanswered due to a lack of studies regarding the use of language and its effects on students’ achievements (e.g. Diehr, 2012). The present study sets out to contribute to closing this gap. By using multilevel analyses, the English and History achievements of 30 bilingual secondary school classes over the course of a school year were examined in relation to methodological and didactic aspects of the use of language. To this end, the teachers provided information on the frequency of first language use in different teaching situations, on the language chosen when introducing new technical terms, and on their expectations with regard to the correct use of the foreign language. 2 Theoretical and empirical background 2.1 CLIL and first language use Even though the term “bilingual” indicates actual bilingualism more often than not, CLIL seems to be construed as monolingual lessons taught in the foreign language, as, for example, reflected in the curricula of the federal states of Bremen, Hesse, Lower Saxony, and Rhineland-Palatinate (cf. Deutscher Bil‐ dungsserver, 1996-2014). A second position accepts the use of the first language as a stopgap measure used for clarification and for the explanation of technical terms (e.g. Bonnet, 2007; Theis, 2010). This position is represented in the curricula in Hamburg, Saxony, and Thuringia (e.g. Deutscher Bildungsserver, 1996-2014) as well as in information brochures of the Ministry of Education and Cultural Affairs of Baden-Wuerttemberg (KM, 2006). In practice, the actual bilingual lessons seem to adhere predominantly to this rule (Böing & Palmen, 2012; Diehr, 2012). A third position advocates a deliberate and systematic use of the first language, which is expected to foster the students’ learning process by system‐ atically supporting the areas of comprehension, expression, and technical vo‐ cabulary (Diehr, 2012; Gierlinger, Hametner & Spann, 2007; Otten & Wildhage, 2003). Therefore, the inclusion of the first language is considered indispensable 246 Sara Dallinger (translated by Nina Rogotzki) <?page no="247"?> (Heimes, 2013). This position is consistent with Butzkamm’s ‘Principle of Enlightened Monolingualism’ (1973, 2002), who considers CLIL an ideal instance of L2 learning in the classroom because of authentic communication contexts which may positively affect the acquisition of a foreign language as well. Hence, by including the first language in bilingual lessons, students would also be able to effectively improve their foreign language skills. It can thus be assumed that increased language comprehension in the classroom as well as the teaching of authentic subject matter leads to a greater appreciation and acceptance of the communication content and, moreover, to a rise of competence in the content subject. In conclusion, the existence of these three rather diverging positions on the use of languages in CLIL is, in all likelihood, due to the fact that no consistent methodological and didactic foundation regarding this particular type of teaching has been established so far (Breidbach, 2010; Thürmann, 2013). To date, few studies have been conducted on the actual use of languages in CLIL. In Müller-Schneck’s study (2006), 4 % of the 89 teachers participating in the survey stated that they did not use German at all. 89 % of the 25 teachers questioned by Bredenbröker (2000) used German infrequently and only as a rare exception, whereas the remaining 11 % admitted using German more frequently. In CLIL contexts the students’ first language was used to explain and clarify complex facts, to solidify the students’ knowledge through revisions and recaps, to introduce new technical terms (Bredenbröker, 2000; Gierlinger et al., 2007), and to mitigate content-related and language-related difficulties (Müller-Schneck, 2006). As far as the introduction of technical terms is concerned, it has yet to be ascertained, though, whether there are specific rules that should be applied regarding the use of one or both languages. This is even more surprising since terminology plays a central role in CLIL (Heimes, 2013) and should be introduced in both languages (KM, 2006; Otten &Wildhage, 2003; Theis, 2010). In sum, only a few, merely descriptive, empirical findings with regard to the use of the first language in CLIL are available, and none of them relate the use of language to students’ achievements in subject content and language. Thus, the effectiveness of teaching aspects, such as the use of the first language, remains largely unexplored. 2.2 Expectations of correct language use In contrast to the use of the first language, the use of the foreign language in CLIL is undisputed. However, it has been debated for a while now to what extent the accuracy of the students’ use of the foreign language should be taken into consideration and how this can be put into practice in the classroom, for example, regarding explicit, content-related language exercises. Focal point of 247 The role of languages in bilingual History lessons <?page no="248"?> the discussion is therefore the question whether the teacher prioritises a focus on form as opposed to a focus on meaning (Llinares, Morton & Whittaker, 2012). So far, CLIL has been almost exclusively based on the latter principle (Lorenzo, 2007) in order to ensure the successful communication of subject matter content. The paradigm message before accuracy (Bach, 2010), which only allows for a correction of foreign language mistakes when the comprehensibility is at stake, seems to predominate (KM, 2006). Nevertheless, in recent years an increased focus on form with regard to bilingual teaching has been called for. Muñoz (2007) argues that the lack of focus on form is one of the major shortcomings of CLIL. Her assessment is owed to research results obtained by studies on English as a foreign language (Doughty & Williams, 1998) and intensive bilingual immersion programmes in Canada (e.g. Lyster, 2007) that reported positive effects of an increased focus on form on students’ L2 language competence. Hardly any comparable studies on CLIL are known. Pérez-Vidal (2007), for example, conducted classroom observations in three bilingual lessons in one primary school and two secondary schools and concluded that the teachers mainly adhered to the principle focus on meaning and that a focus on form was barely observable. However, in Germany, similar studies on language correctness that consider a focus on form have been conducted with regard to English lessons. The DESI-study (2006) surveyed teachers’ expectations of correct language use from the teachers’ as well as the students’ point of view in grade 9. Results showed that high expectations had a positive effect on students’ English listening comprehension skills. Hochweber, Steinert & Klieme (2012) also analysed data obtained in the DESI-study and found that teachers’ expectations of correct language use predicted students’ performance in a C-test at the end of grade 9, thus being a significant predictor for students’ L2 development. Similar positive results were reported in the context of German lessons with respect to the students’ reading skills (Rjosk, Richter, Hochweber, Lüdtke, Klieme & Stanat, 2014). In sum, the studies just cited suggest that students’ foreign language skills benefit from high expectations of language correctness. However, this has yet to be verified for the CLIL context. The same applies to possible effects on achievement in the content subject. Concerning content learning, it could be ar‐ gued that excessive expectations of language correctness may have detrimental effects since the focus on correct language use distracts from the actual subject matter. However, Clemen & Sauer (2007) and Lamsfuß-Schenk (2008) showed how bilingually taught History lessons fostered the students’ grasp of History competences such as foreign language understanding, multi-perspectivity, and the ability to change perspectives. They attributed this to the fact that the 248 Sara Dallinger (translated by Nina Rogotzki) <?page no="249"?> students dealt with the foreign language study material longer and more thoroughly, which, according to the authors, led to a more in-depth processing and understanding of the subject matter. Hence, a stronger focus on form, i. e. higher expectations of language correctness, may have positive effects on content learning. 2.3 Research questions Taking into account the considerations above, there is a need for longitudinal studies that analyse the use of the first language and the expectations of foreign language correctness in CLIL programmes and investigate possible effects on students’ language and content subject competence. The present study takes on this task by examining the following research questions (RQ): (RQ1) How frequently and in which specific situations (e.g. explanation and clarification of complex subject matter, revisions and recaps, introduction of technical terms) is the first language used in bilingual History lessons and, furthermore, do teachers follow any rules with regard to the use of the first or foreign language when introducing technical terms? It is expected that the first language is rarely used and that technical terms are generally introduced in both languages. (RQ2) What are the teachers’ expectations with regard to students’ foreign language correctness? Considering the paradigms focus on meaning and message before accuracy that are predominant in the CLIL classroom, rather low expect‐ ations are predicted. (RQ3) Controlling for average prior achievement, socio-economic status, and student motivation, how do these language-related aspects affect their English competence? Following Butzkamm’s ‘Principle of Enlightened Monolingualism’ (1973, 2002), a frequent and systematic use of the first language in different classroom situations should have stronger effects on students’ L2 competence. Thus, lower L2 competence is expected when technical terms are merely introduced in English, whereas greater progress is expected when the language is chosen systematically or when technical terms are consistently introduced in both languages. Moreover, based on prior research (e.g. Hochweber et al., 2012), a positive effect of high expectations of language correctness on students’ English competence is predicted. Further possible effects on their History competence remain an open question, though. 249 The role of languages in bilingual History lessons <?page no="250"?> 3 Method 3.1 Sample and Procedure The sample was derived from the project COMBIH (Competences and Motiva‐ tion in Bilingual Instruction in History) and comprised 703 bilingually taught eighth graders from 30 classes in 28 secondary schools in the federal state of Baden-Wuerttemberg (50.1 % female; M age = 13.4; SD age = 0.48). 30.5 % of these students had a migration background (i.e. at least one parent was not born in Germany), and 74 % reported that one or both parents had qualified for college education. In grades 5 and 6 the students attended two weekly English lessons on top of their five English lessons. From grade 7 onwards they received CLIL-instruction in one content subject. In grade 7 geography was taught bilingually, in grade 8 History. Prior to this, in grades 5 and 6, the students had already attended regular History classes (i.e. taught in German). The teacher sample comprised 29 CLIL History teachers (41.4 % female; M age = 40.3; SD age = 9.12). One teacher did not participate in the survey. The data were collected in autumn 2012 and summer 2013, i. e. at the beginning and at the end of grade 8. Data collection was conducted in the form of group tests during regular school lessons by research assistants using a standardised manual. The teachers completed their questionnaires either at school or at home. 3.2 Instruments Teachers’ use of language. The teachers’ questionnaire was a non-standardised tool designed to assess the teachers’ frequency of use of German in different classroom situations. With regard to the question “How often do you use German in the following situations in the bilingual History lessons of this class? ” three situations were evaluated: “To explain and clarify complex subject matter” [Explanation/ clarification subject matter], “To revise and recap subject matter” [Revision & recap], and “To introduce technical terms in German as well” [Technical terms]. Responses were given on a 4-point scale ranging from 1 (very infrequently - once per month or less) to 4 (very frequently - more than twice per lesson). This item as well as the following one was developed by the research team. The teachers’ choice of language when introducing technical terms was assessed with the item “Which of the following statements best describes your approach to technical terms in the bilingual History lessons of this class? ” Five response options were given: “In general, I introduce technical terms solely in English” [English], “For the most part, I introduce technical terms solely 250 Sara Dallinger (translated by Nina Rogotzki) <?page no="251"?> in German” [German], “I introduce technical terms in English as well as in German” [English & German], “I do not follow a specific rule; it depends on the particular term” [No rule, term dependent], and “I do not follow a specific rule; I choose spontaneously during the lesson” [No rule, spontaneous]. The items were recoded as dummy variables (reference category: No rule, term dependent). The expectations of language correctness were measured with five items regarding the questions “How important do you rate the foreign language-re‐ lated aspects in your bilingual History lessons? ”: “It is important to me that the students’ pronunciation is correct”, “that the students produce grammatically correct written sentences”, “that the students produce grammatically correct spoken sentences”, “that the students master the English orthography”, and “that the students can express themselves more or less accurately in English”. Items 1-4 were adapted from the DESI-study (2006) and slightly modified for the teachers’ (rather than students’) questionnaires, item 5 was self-developed. Responses were given using a 4-point scale (1 (completely disagree) to 4 (com‐ pletely agree)). The five items were conflated in one scale “Language correctness” (α = .65). Students’ English achievement tests. A C-test (e.g. Grotjahn, 2011) was designed to assess students’ general English skills, which consisted of 159 items/ gaps distributed over seven text passages. In order to reduce test repetition effects, a booklet design was used: Three texts were administered at both times of measurements. Two texts were rotated between t1 and t2. The scaling of the tests was carried out for the entire sample of the COMBIH study, which comprised both CLIL-students and non-CLIL-students. The items were analysed and IRT-scaled using a 2-parameter logistic model (Birnbaum model) which allowed the item difficulty parameters and the coefficients of discrimination to vary (Birnbaum, 1968). The items were examined with respect to Differential Item Functioning (DIF; variation in item parameters between different subpopulations) regarding gender and group affiliation (i.e. CLIL students versus non-CLIL students) and with respect to Item Parameter Drift (IPD; variation in item parameters across time). The differences between groupand time-specific item parameters were tested for statistical signifi‐ cance and graphically represented using the delta theorem. Based on these procedures, 21 pre-test items and 18 post-test items were excluded. Further‐ more, two items showed considerable IPD. A variation in item parameters between t1 and t2 (Partial Measurement Invariance) was allowed for these items. Subsequently, person parameters were estimated using EAP estimators. In order to determine the reliability of the tests, the variance of the person 251 The role of languages in bilingual History lessons <?page no="252"?> parameters was related to the variance of the latent ability dimension theta. The reliability was high at both points of measurement (Rel(EAP) t1 = 0.94; Rel(EAP) t2 = 0.95). In addition, the students’ listening comprehension skills were assessed by means of nine tasks with 43 open or closed questions (following Köller, Knigge & Tesch, 2010). At t1, the students were assigned six tasks. At t2, two of these tasks with 13 items each were reused as anchor items and three new tasks were created to account for the students’ progress. The same IRT-based analyses were applied as for the C-test. Due to DIF four pre-test items were excluded. Since none of the items showed IPD, all parameters of the anchor items were set equal between pre-test and post-test. The reliabilities were satisfactory (Rel(EAP) t1 = 0.71) to good (Rel(EAP) t2 = 0.84). History achievement. The eighth-graders’ knowledge of History was evalu‐ ated using a cloze test and multiple-choice questions (36 items). The test was developed based on the curriculum for History of Baden-Wuerttemberg. The students were asked to provide information on important dates, names, and technical terms and had to identify correct historical contexts as well as reasons for and consequences of historic events. At t1, all students received the test in German. At t2, a third of the CLIL-students received the test once more in German. The remaining CLIL-students received either the cloze test or the multiple-choice questions in English. That way, future evaluations could analyse the achievement as a function of the tests’ language. Again, the same IRT-based analyses as above were applied. To avoid possible distortions in the parameter estimation bias, the English items were treated as separate items in the measurement model whose item parameters were allowed to differ from those of the German items. Three pre-test items as well as two German and two English items were excluded due to DIF. Since eight items exhibited IPD, Partial Measurement Invariance was allowed. The reliabilities were satisfactory (Rel(EAP) t1 = 0.70) to good (Rel(EAP) t2 = 0.79). Additional student related characteristics. At t1, the students completed figural and verbal cognitive skills tests, i. e. subtests of the KFT (Heller & Perleth, 2000). Parallel test versions were used to prevent copying and cheating. The figural test comprised 25 items with reliability values (Cronbach’s alpha) at α Version A = 0.91 and, respectively, α Version B = 0.90. The verbal test comprised 20 items (α = 0.69 and, respectively, α = 0.72). Based on the Expectancy-Value Model of Achievement Motivation (Wigfield & Eccles, 2000), the students’ intrinsic and extrinsic motivation was measured using two scales, i. e. intrinsic value (five items; History: α = 0.82; English: α = 0.72; e. g. “English/ History is one of my favourite subjects”) and extrinsic value (five items; History: 252 Sara Dallinger (translated by Nina Rogotzki) <?page no="253"?> α = 0.89; English: α = 0.88; e. g. “I think I will be able to put my English/ History knowledge to good use later in life”). Answers were given on a 4-point scale ranging from 1 (completely disagree) to 4 (completely agree). The items used to assess the students’ family background were adapted from large-scale studies such as PISA (cf. Klieme, Artelt, Hartig, Jude, Köller & Prenzel, 2010). Based on the provided information on the parents’ occupation and education, the international socio-economic index of occupational status (HISEI; Ganzeboom, de Graaf, Treiman & de Leeuw, 1992) was determined for each student with a range between 11 and 88 (M = 65.21; SD = 17.77). Moreover, information was obtained regarding the students’ gender and migration background. 3.3 Statistical analysis Since the language-related variables presented here always refer to the class as a whole, they are considered as class characteristics in two-level regression models (Raudenbush & Bryk, 2002). For each test (C-test, listening comprehen‐ sion, and History test) four models were calculated using Mplus 7.1 (Muthén & Muthén, 1998-2012). In each case, the test performance at t2 served as dependent variable. At class-level (level 2), the language-related variables were established as predictors. In model 1 the frequency of German in different teaching situations was analysed, in model 2 the choice of language for introducing technical terms, and in model 3 the expectations of language correctness. As the sample size at level 2 only amounted to N = 30 classes, the analyses were allocated to two separate models. In all models the average pre-test achievement of the class was controlled for in order to determine the predictability of the language-related variables for the competence development and to control for a possible cofounding of pre-test achievement and use of language. Again, due to sample size, a restriction to central control variables had to be applied at level 2. The average socio-economic status and the intrinsic motivation at class-level was factored in because these variables showed substantial variations between classes (intraclass correlation efficient ICC ≥ 0.05). In model 0 only the control variables were entered to enable the following models to determine how much additional variance at level 2 could be explained by language-related variables. In addition to analyses at class-level that were relevant with regard to the research questions, a more complex model at student-level (level 1) was specified that provided information on which students within each class showed greater achievement. Prior to analysis, all continuous variables were z-standardised, and all continuous independent variables were group mean centred. The percentage of missing values ranged between 0 % (gender) and 13 % (socio-economic status). At level 2, the percentage of missing values of all 253 The role of languages in bilingual History lessons <?page no="254"?> variables was at 3.3 % since one teacher of the 30 classes did not take part in the survey. To account for missing values, the full information maximum likelihood-procedure (FIML) implemented in Mplus 7 was applied. In opposition to multiple imputation, the FIML-procedure does not compile several complete data sets prior to analysis but estimates the value of the missing data based on all available information by means of a model-based approach. 4 Results In order to answer the first and second research question (i.e. use of the first language in bilingual History lessons and expectation of correct language use), Table 1 presents the mean values and the standard deviations and, respectively, the percentage frequencies. The lower mean values of the variables regarding the frequency of the use of German confirm that the first language was rather infrequently used. When used it was mostly for explaining and clarifying complex subject matter (M = 2.10; SD = 0.67). As expected, a large part of the teachers (41.8 %) stated that they introduced technical terms in both languages. 34.5 % of the teachers introduced technical terms only in English, whereas slightly less than 25 % declared to choose the language term dependent. None of the teachers stated that they introduced technical terms only in German or that they chose the language spontaneously during the lesson. In consequence, these response options were not entered into the regression analyses. Again, as expected, the mean value of M = 2.38 (SD = 0.42) indicated low to medium expectations of language correctness. Furthermore, Table 1 displays the correlations between language-related and dependent variables at level 2. Significant correlations were found between the items for the use of German in different classroom situations and for introducing technical terms. With regard to the dependent variables, positive effects of the use of German and the expectations of language correctness on the competence development could be shown, even though several correlations proved to be statistically insignificant due to the small sample size. The introduction of technical terms solely in English correlated negatively with all test values. The item for the use of German for the explanation and clarification of complex subject matter exhibited insignificant correlations. 254 Sara Dallinger (translated by Nina Rogotzki) <?page no="255"?> Note: M = mean value (Range: 1 = infrequent use of German or low demands on linguistic correctness - 4 = frequent use of German or high demands), SD = standard deviation; t2 = second measuring point; N = 30 classes. * p < .05, ** p < .01, *** p < .001. Table 1: Descriptive results and correlations between language-related and dependent variables. Prior to answering the third research question (effects of the language-related aspects on the students’ competence development), the analyses at level 1 regarding possible correlations between individual student characteristics and their achievement will be presented. Subsequently, the results of the analyses at level 2 will be displayed in Table 3 (C-test, listening comprehension) and Table 4 (History test). The model was specified in a way that allowed a separate examination of the levels, thus, the findings at level 1 are not influenced by model variations at level 2 and vice versa. For this reason, Table 3 and 4 forego a repeated presentation of the level 1-values as they are identical for the models 1-3, respectively. Table 2 documents a stability of the C-test achievement amounting to B = 0.58. The C-test competence development was predicted by gender, figural and verbal skills and extrinsic value. That is, after controlling for all other predictors, better progress was shown by girls (in comparison with boys), by students with higher cognitive skills, and by those students who considered English as 255 The role of languages in bilingual History lessons <?page no="256"?> particularly useful. 43 % of the C-test variance at level 1 was explained by student characteristics. For the listening comprehension test, the stability was B = 0.28. In this respect, the variables for socio-economic status, verbal cognitive skills, and extrinsic value correlated with the students’ achievement and accounted for 20 % of the variance. With regard to the History test, a stability of B = 0.39 was displayed. The verbal skills and the intrinsic value for History predicted the achievement and explained 28 % of the variance. Regarding the language-related variables at level 2, Table 3 shows that the proportion of variance in the C-test that can be traced back to the difference between classes was at 11 % (ICC = 0.11). C-test achievement was greater in those classes whose teachers had higher expectations of language correctness (B = 0.12; model 3). All other language-related variables did not exhibit any statistically significant correlations. Due to the small sample size (N = 30 classes) the variance additionally explained by the language variables needs to be considered in terms of the interpretation of effects, though. Consideration of these variables explained an extra 5 % or 11 % and 12 %, respectively, of the variance at level 2 (model 1-3 and model 0). In addition, classes with high socio-economic status showed greater progress (B = 0.34; model 3). The proportion of variance in the listening comprehension test was also at 11 % (see Table 3). Competence development was particularly positive in those classes in which German was used intensively for the introduction of technical terms (B = 0.15). The use of German explained an additional 28 % of the variance. Similarly, for classes whose teachers introduced technical terms exclusively in English (B = -0.54) lesser progress was recorded than for those classes whose teachers stated that they followed no specific rule but chose the language depending on the term. The systematic use of both languages for the introduction of technical terms proved to be negative for listening comprehension achievement (B = -0.37). The use of language with regard to technical terms explained an extra 39 % of the variance. No significant effects could be noted for the scale depicting the expectation of language correctness, which only accounted for 5 % of the variance. In model 1, a significant effect was detected with respect to the prior achievement of the class. For the History test (see Table 4) a higher proportion of level 2 variance (21 %) could be attested than for English achievement. With regard to the frequency of German, none of the variables were statistically significant and only an extra 4 % of the variance could be explained. Once again, it became evident that the introduction of technical terms in German and English correlated with a poorer performance development of the classes (B = -0.43) in comparison with the classes whose teachers stated that they did not follow a specific rule. 256 Sara Dallinger (translated by Nina Rogotzki) <?page no="257"?> Note: B = regression coefficient, SE = standard error; t1 = first measuring point, HISEI = international socio-economic measure of occupational status, KFT = Cognitive Ability Test; N = 703 learners. * p < .05, ** p < .01, *** p < .001. Table 2: Effects of individual pupil characteristics on achievement development (pupil level). 257 The role of languages in bilingual History lessons <?page no="258"?> A comparatively large, although statistically insignificant, negative effect was found for the exclusive introduction of technical terms in English. Together, these variables explained an additional 14 % of the variance. Moreover, achieve‐ ment was again greater in those classes with teachers with high expectations of language correctness (B = 0.22), which explained an additional 24 % of the variance. Finally, in model 3 the average socio-economic status predicted the competence development of the class. Further analyses considered all language-related variables simultaneously in one model. However, in light of the sample size N = 30 classes, these findings have to be approached with caution. They showed, though, that with respect to the C-test a total of 74 % of the level 2 variance could be explained, i. e. an extra 35 % in comparison to model 0. Overall, 97 % (an additional 66 %) of the total variance at level 2 could be explained regarding the listening comprehension test and 73 % (an additional 42 %) regarding the History test. Note: B = regression coefficient, SE = standard error; t1 = first measuring point, HISEI = international socio-economic measure of occupational status; N = 30 classes. a Reference category: No rule, term dependent. * p < .05, ** p < .01, *** p < .001. Table 3: Effects of language-related aspects on performance development in English (class level). 258 Sara Dallinger (translated by Nina Rogotzki) <?page no="259"?> Note: B = regression coefficient, SE = standard error; t1 = first measuring point, HISEI = international socio-economic measure of occupational status; N = 30 classes. * p < .05, ** p < .01, *** p < .001. Table 4: Effects of language-related aspects on performance development in the History test (class level). 5 Discussion The aim of this paper was to analyse the role of teachers’ use of languages in CLIL History lessons with regard to students’ competence in English and in their content subject History. To this end, a longitudinal study was conducted 259 The role of languages in bilingual History lessons <?page no="260"?> with two points of measurement, i. e. at the beginning and at the end of grade 8. The first research question (RQ1) assessed the use of the first language, i. e. German. The results confirmed expectations and were able to replicate previous findings, which ascertained that even though the foreign language is primarily used in CLIL, German is also used in specific classroom situations (Bredenbröker, 2000; Gierlinger et al., 2007; Müller-Schneck, 2006). These results validate the standards specified in the curricula of several states, which stipulate the predominant but not exclusive use of the foreign language in CLIL (cf. Deutscher Bildungsserver, 1996-2014) and support the standard school practice which, to a large extent, has implemented these plans (Böing & Palmen, 2012; Diehr, 2012). Therefore, the findings reported here corroborate the second position presented in section 1 that postulates the use of German for technical terms and otherwise at best as a stopgap measure. Moreover, the COMBIH study was able to add to previous findings by pro‐ viding more detailed information on the choice of language for the introduction of technical terms, according to which technical terms are mainly introduced either in English or in both languages, but never exclusively in German. In the context of the RQ2, it could be confirmed that CLIL-teachers have comparatively lower expectations of correct language use and, thus, mostly adhere to the principles commonly applied in CLIL, i. e. focus on meaning (Llinares et al., 2012) and message before accuracy (Bach, 2010). For the first time, the COMBIH study was able to relate these descriptive findings for teachers’ language use to students’ English and History achieve‐ ment and, hence, to analyse the effectiveness of these language-related aspects (cf. RQ3). The importance of this correlation is stressed by the great proportions of variance, which were additionally controlled for by taking the language-re‐ lated variables into consideration. The positive effect of first language use for technical terms, which was determined with regard to students’ listening comprehension, supports a continuation of the predominant school practices which stipulate that German is merely used for technical terms and when needed (e.g. to mitigate comprehension difficulties). The fact that no effects could be attested to the frequent and systematic use of German for explanations and clarifications and for revisions and recaps limits but does not preclude the transferability of Butzkamm’s Principle of Enlightened Monolingualism (1973, 2002). This is indicated by the present results, which showed that with regard to the introduction of technical terms not only the exclusive use of English but also the consistent use of both languages yielded less progress. Positive effects could only be shown for classes whose teachers stated to choose the language deliberately and in dependence on the respective terms. It becomes apparent, 260 Sara Dallinger (translated by Nina Rogotzki) <?page no="261"?> then, that it is not the inclusion of the first language per se that matters, but rather the deliberate and systematic use of the first language. Since higher teacher expectations of language correctness correlated with higher scores in the C-test as well as in the History test, the present findings encourage a critical examination of the paradigms focus on meaning and message before accuracy that are predominant in CLIL. Correspondingly, future CLIL has to place more emphasis on the principle focus on form (e.g. Doughty & Williams, 1998; Lyster, 2007). Moreover, previous findings regarding the effects of language correctness standards on students’ English competence could be reinforced by the present results (e.g. Hochweber et al., 2012) and, for the first time, transferred to the CLIL context. On the one hand, the positive effect on the students’ History competence could be accounted for by the fact that the students dealt with the foreign language study material longer and more exhaustively and, thus, were able to process it more thoroughly (e.g. Clemen & Sauer, 2007; Lamsfuß-Schenk, 2008). On the other hand, such effects could imply generally higher standards of teacher expectations for students’ competence, which may affect the quality of the instruction they provide (e.g. Hochweber et al., 2012). It cannot be ruled out that such confounding variables may have had an impact on the findings. Future studies would therefore have to conduct experimental manipulations regarding methodological and didactic aspects of CLIL. In sum, the present findings are significant because, for the first time, they allowed an assessment of the effects of language-related teaching aspects of CLIL History lessons on students’ achievement in English (as the foreign language) and History (as the content subject in question). The hierarchical data structure was adequately represented by using multilevel analyses. However, only a few aspects could be analysed. Hence, there is a need for further studies that investigate the role of language use in CLIL contexts in even more detail. For example, the ratio of teacher and student speaking time during bilingual lessons in relation to the respective ratio of first language vs. foreign language use would be of interest. Moreover, language-related aspects should be examined not only from the teachers’, but also from the students’ point of view and, additionally, by means of classroom observations. Finally, future studies need to take into account the effects of language-related teaching aspects with regard to competences in further foreign languages and content subjects. Such findings could serve as a foundation for the still pending conception of a specific CLIL methodology and didactics aiming at an optimal promotion of the students’ competence. The present results point to the fact that the systematic use of the first language German for technical terms and higher teacher expectations of foreign language correctness might play an important role in this context. 261 The role of languages in bilingual History lessons <?page no="262"?> References Bach, G. 2010. Bilingualer Unterricht. Lehren, Lernen, Forschen. In: Bach, G. & Niemeier, S. (Eds.), Bilingualer Unterricht. Frankfurt/ M.: Peter Lang. 9-22. Baker, C. 2006. Foundations of Bilingual Education and Bilingualism. Clevedon: Multilin‐ gual Matters. Birnbaum, A. 1968. Some latent trait models and their use in inferring an examinee’s ability. In Lord, F.M. & Novick, M.R. (Eds.), Statistical Theories of Mental Test Scores. Reading, MA: Addison-Wesley. 397-479. Böing, M. & Palmen, P. 2012. Bilingual heißt zweisprachig! Überlegungen zur Verwen‐ dung beider Sprachen im bilingual deutsch-französischen Geographieunterricht. In: Diehr, B. & Schmelter, L. (Eds.), Bilingualen Unterricht weiterdenken. Frankfurt/ M.: Peter Lang. 73-90. Bonnet, A. 2007. Fach, Sprache, Interaktion - Eine Drei-Säulen-Methodik für CLIL. Fremdsprachen Lehren und Lernen, 36, 126-141. Bredenbröker, W. 2000. Förderung der fremdsprachlichen Kompetenz durch bilingualen Unterricht. Frankfurt/ M.: Peter Lang. Breidbach, S. 2010. Bilinguale Didaktik - bald wieder zwischen allen Stühlen? Zu den Aussichten einer integrativen Didaktik des bilingualen Sachfachunterrichts. In: Bach, G. & Niemeier, S. (Eds.), Bilingualer Unterricht. Frankfurt/ M.: Peter Lang. 165-176. Butzkamm, W. 1973. Aufgeklärte Einsprachigkeit. Zur Entdogmatisierung der Methode im Fremdsprachenunterricht. Heidelberg: Quelle & Meyer. Butzkamm, W. 2002. Psycholinguistik des Fremdsprachenunterrichts. Tübingen: Francke. Clemen, F. & Sauer, M. 2007. Förderung von Perspektivendifferenzierung und Perspekti‐ venübernahme? Bilingualer Geschichtsunterricht und historisches Lernen. Geschichte in Wissenschaft und Unterricht, 58 (12), 708-723. DESI-Konsortium (Ed.). 2006. Unterricht und Kompetenzerwerb in Deutsch und Englisch. Zentrale Befunde der Studie Deutsch Englisch Schülerleistungen International (DESI). Frankfurt/ M.: DIPF. Deutscher Bildungsserver. 1996-2014. Bildungspläne der Bundesländer für allgemeinbil‐ dende Schulen. http: / / www.bildungsserver.de/ Bildungsplaene-Lehrplaene-der-Bunde slaender-fuer-allgemeinbildende-Schulen-400.html. May 31, 2021. Diehr, B. 2012. What’s in a name? Terminologische, typologische und programmatische Überlegungen zum Verhältnis der Sprachen im Bilingualen Unterricht. In: Diehr, B. & Schmelter, L. (Eds.), Bilingualen Unterricht weiterdenken. Frankfurt/ M.: Peter Lang. 17-36. Doughty, C. & Williams, J. (Eds.). 1998. Focus on Form in Classroom Second Language Acquisition. Cambridge: CUP. 262 Sara Dallinger (translated by Nina Rogotzki) <?page no="263"?> Ganzeboom, H.B.G., de Graaf, P.M., Treiman, D. & de Leeuw, J. 1992. A standard international socio-economic index of occupational status. Social Science Research, 21 (1), 1-56. Gierlinger, E., Hametner, C. & Spann, H. 2007. Englisch als Arbeitssprache - viele Inseln und (noch) kein Festland? In: Gierlinger, E., Hainschink, V. & Spann, H. (Eds.), Content and Language Integrated Learning (CLIL) auf der Sekundarstufe I. Linz: Trauner. 63-118. Gogolin, I., Neumann, U. & Roth, H.-J. 2009. Schulversuch bilinguale Grund‐ schulklassen in Hamburg. Abschlussbericht über die türkisch-deutschen Modell‐ klassen. https: / / www.ew.uni-hamburg.de/ ueber-die-fakultaet/ personen/ gogolin/ pdfdokumente/ bericht-2009-bilinguale-grundschulklassen.pdf. May 31, 2021. Grotjahn, R. 2011. C-Tests: Aspekte der Validität. Deutsch als Fremdsprache, 48 (3), 131-137. Heimes, A. 2013. Didaktiken und Methodiken bilingualer Fächer: Geschichte. In: Hallet, W. & Königs, F.G. (Eds.), Handbuch Bilingualer Unterricht. Seelze: Kallmeyer. 345-352. Heller, K.A. & Perleth, C. 2000. Kognitiver Fähigkeitstest für 4. - 12. Klassen. Revision. Göttingen: Hogrefe. Hochweber, J., Steinert, B. & Klieme, E. 2012. Lehrerkooperation, Unterrichtsqualität und Lernergebnisse im Fach Englisch. Unterrichtswissenschaft, 40 (4), 351-370. Klieme, E., Artelt, C., Hartig, J., Jude, N., Köller, O. & Prenzel, M. (Eds.) 2010. PISA 2009. Münster: Waxmann. KM = Ministerium für Kultus, Jugend und Sport Baden-Württemberg. 2006. Bilingualer Unterricht. Realschule. Braunschweig: Westermann. Köller, O., Knigge, M., & Tesch, B. (Eds.). 2010. Sprachliche Kompetenzen im Länderver‐ gleich. Münster: Waxmann. Lamsfuß-Schenk, S. 2008. Fremdverstehen im bilingualen Geschichtsunterricht. Eine Fall‐ studie. Frankfurt/ M: Peter Lang. Llinares, A., Morton, T. & Whittaker, R. 2012. The Roles of Language in CLIL. Cambridge: CUP. Lorenzo, F. 2007. The sociolinguistics of CLIL. In: Lorenzo, F., Madinabeitia, S.C., de Alba Quiñones, V. & Moore, P. (Eds.), Models and Practice in CLIL. Logroño: LESRA. 27-38. Lyster, R. 2007. Learning and Teaching Languages through Content. Philadelphia, PA: John Benjamins. Müller-Schneck, E. 2006. Bilingualer Geschichtsunterricht. Frankfurt/ M.: Peter Lang. Muñoz, C. 2007. CLIL. Some thoughts on its psycholinguistic principles. In: Lorenzo, F., Madinabeitia, S.C., de Alba Quiñones, V. & Moore, P. (Eds.), Models and Practice in CLIL. Logroño: LESRA. 17-26. Muthén, L.K. & Muthén, B.O. 1998-2012. Mplus User’s Guide. Los Angeles, CA: Muthén & Muthén. 263 The role of languages in bilingual History lessons <?page no="264"?> Nold, G., Hartig, J., Hinz, S. & Rossa, H. 2008. Klassen mit bilingualem Sachfachunterricht. Englisch als Arbeitssprache. In DESI-Konsortium (Ed.), Unterricht und Kompetenzer‐ werb in Deutsch und Englisch. Weinheim: Beltz. 451-457. Otten, E. & Wildhage, M. 2003. Content and Language Integrated Learning. Eckpunkte einer “kleinen” Didaktik des bilingualen Sachfachunterrichts. In Wildhage, M. & Otten, E. (Eds.), Praxis des bilingualen Unterrichts. Berlin: Cornelsen. 12-45. Pérez-Vidal, C. 2007. The need for focus on form in content and language integrated approaches. In: Lorenzo, F., Madinabeitia, S.C., de Alba Quiñones, V. & Moore, P. (Eds.), Models and Practice in CLIL. Logroño: LESRA. 39-54. Raudenbush, S.W. & Bryk, A.S. 2002. Hierarchical Linear Models. Thousand Oaks, CA: Sage. Rjosk, C., Richter, D., Hochweber, J., Lüdtke, O., Klieme, E. & Stanat, P. 2014. Socio‐ economic and language minority classroom composition and individual reading achievement. The mediating role of instructional quality. Learning and Instruction, 32, 63-72. Theis, R. 2010. Bilingualer Geschichtsunterricht. In: Doff, S. (Ed.), Bilingualer Sachfachun‐ terricht in der Sekundarstufe. Tübingen: Narr. 44-57. Thürmann, E. 2013. Spezifische Methoden für den bilingualen Unterricht/ CLIL. In: Hallet, W. & Königs, F.G. (Eds.), Handbuch Bilingualer Unterricht. Seelze: Kallmeyer. 229-235. Wigfield, A. & Eccles, J.S. 2000. Expectancy-value theory of achievement motivation. Contemporary Educational Psychology, 25 (1), 68-81. 264 Sara Dallinger (translated by Nina Rogotzki) <?page no="265"?> Does a positive selection bias into CLIL streams explain higher language proficiency? The impact of cognitive abilities and SES on the selection process Nils Jaekel Abstract CLIL streams in Germany have long been understood as an elitist approach. Although many schools have started to open their CLIL streams to a larger number of students, in general allowing all students together with their parents to choose CLIL schooling, other schools still select their students based on a variety of high achievement criteria. Considering this selective nature, CLIL students’ success in acquiring high levels of language profi‐ ciency while maintaining or even surpassing their peers in mainstream EFL classes in content knowledge is not surprising. Researchers thus need to consider to what extent these often found above average achievement outcomes may be explained by positive selection biases of students into CLIL streams. This study investigated the composition of CLIL versus EFL classes with regard to a number of students’ individual difference and background variables (i.e. gender, age, SES, home language (L1) and cognitive abilities). The overall sample consisted of 380 year 9 students from 16 classes and nine different grammar schools from the Ruhr area in the federal state of North-Rhine Westphalia, Germany. Language profi‐ ciency was measured by the last grade in English and a C-Test battery. The typical CLIL students in the present study were younger, of higher SES, and scored higher on the cognitive functions test (CFT) than their peers in EFL classes. As expected, CLIL students also received better grades in English and scored significantly higher on the general language proficiency test (C-Test) than EFL peers. Structural equation modelling however showed that controlling for the positive selection bias into CLIL classes did not change the strong positive effect of CLIL on language proficiency. These <?page no="266"?> results suggest that the previously reported positive effects of CLIL are not explained by the existing selection bias. 1 Introduction The concept of Content and Language Integrated Learning (CLIL) has been a tremendous success story. Coined as an umbrella term to cover the various approaches of teaching content through the medium of a foreign language, today the approach has reached mainstream education in many parts of Europe and has gained increasing attention outside of Europe, for example in South America or Australia. CLIL has been moving into the spotlight of researchers and practitioners alike as it promises a) synergistic effects of teaching one content subject while improving students’ foreign language proficiency and b) deeper processing of both domains. In addition, the development of intercultural communicative competence, highlighted in the CEFR and a central part of Coyle’s prominent 4Cs model of CLIL (Council of Europe, 2001: 23; Coyle, 2007: 551), is a central aim. Its reputation is based on a large body of research from around Europe emphasising linguistic benefits and comparable, if not higher, achievement in the content subject (Pérez-Cañado, 2012). Cross-na‐ tional studies which thoroughly scrutinise CLIL are however missing. CLIL approaches differ greatly, as per its intended definition, and in 2006 they could be found throughout Europe in 19 countries in many shapes and different sizes (Eurydice, 2006: 14). Multilingual societies such as Luxembourg or Malta have long-standing traditions of multilingual education. Similarly, countries such as Belgium and Spain have been offering content teaching for regional or minority languages (Eurydice, 2006: 14). While various approaches had been established throughout Europe, content teaching through an L2 only really gained momentum after the European Council in 1995 announced their plan of ‘the multilingual European citizen’, i. e. to be proficient in one’s mother tongue and at least two additional languages (European Commission, 1995). These efforts went hand in hand with the aims proposed for CLIL which were to foster internationalisation, job perspectives, intercultural understanding as well as both language skills and subject-related knowledge (Eurydice, 2006: 22). The ever-faster spread of English through media, the internet and the importance of high English language proficiency have surely contributed significantly to this international success. In Germany, for example, in 2013 more than 1,500 schools taught through CLIL or “bilingualer Sachfachunterricht” as it is called here (Sekretariat der Ständigen Konferenz der Kultusminister der Länder und der Bundesrepublik 266 Nils Jaekel <?page no="267"?> Deutschland, 2013: 4). Within 14 years the number of schools offering CLIL thus quadrupled from only 366 in 1999 (Sekretariat der Ständigen Konferenz der Kultusminister der Länder und der Bundesrepublik Deutschland, 2013: 4). CLIL is understood as a story of success which educational policies and practitioners alike aim to spread and expand from selective grammar school contexts to comprehensive schools and the lower tier “Hauptschule” (Ministerium für Schule und Weiterbildung des Landes Nordrhein-Westfalen, 2015; Schwab, 2013). Similar egalitarian approaches are being pursued in Italy (Cinganotto, 2016), and collaboratively investigated across Europe in the ADiBE Project, CLIL for all: Attention to Diversity in Bilingual Education (Pérez-Cañado, Coyle, Ting, Nikula, Dalton-Puffer, Matz & Jerez Montoya, 2019). Despite the wide spread of the approach, CLIL for various reasons still maintains an elitist image in Germany. Although many schools, particularly comprehensive schools, offer CLIL on a voluntary basis open for any student, the bulk of grammar schools maintains a rigorous selection process. An alternative to these sought after CLIL streams is modular CLIL, which has also spread rapidly throughout Germany (Sekretariat der Ständigen Konferenz der Kultusminister der Länder und der Bundesrepublik Deutschland, 2013: 10). CLIL modules usually limit teaching through the L2 to a short timeframe focused on a narrow content topic. Evidence for their popularity can be seen in the implementation of smaller CLIL modules in current EFL schoolbooks (for example, Englisch G 21, Biederstädt, Harger & Schwarz, 2008; Green Line 3, Horner, Ashford, Baer-Engel, Daymond, Dennis & Plitsch, 2007). The popularity of and demand for traditional CLIL streams, however, remain high even after the advent of modular CLIL. This article aims to shed some light on what characteristics students in CLIL streams at grammar schools have and how they differ from regular EFL students with regard to gender, age, cogni‐ tive abilities, their home language (L1), and their parents’ socio-economic background. 2 Background The CLIL approach deliberately aims to cover the manifold approaches of bilingual teaching currently available throughout Europe. CLIL was proposed “as a generic umbrella term which would encompass any activity in which a foreign language is used as a tool in the learning of a non-language subject in which both language and the subject have a joint curricular role” (Marsh, 2002: 58). 267 Does a positive selection bias into CLIL streams explain higher language proficiency? <?page no="268"?> CLIL approaches may be as little as language showers and as all-encompassing as full immersion (Marsh, 2002: 13). This alone complicates research and in particular the ability to compare and evaluate CLIL programmes across borders. However, even within countries, CLIL differs considerably. In Germany, CLIL has a longstanding tradition foremost at grammar schools, which is one of the key reasons it still maintains an elitist reputation. In secondary education, CLIL is currently implemented in three main forms: CLIL streams, ‘reduced’ CLIL streams and modular CLIL. The full CLIL streams usually start with two years of intensive preparation in English in years 5 and 6 before the first content subject is introduced in year 7. Successively in years 8 and 9, respectively, one additional content subject is introduced to the bilingual curriculum. All three content subjects are then maintained until the end of year 9 (see Figures 1 and 2). Should students opt for CLIL during their upper-secondary years, one or two content subjects will be taught using the L2, and one of these subjects will be assessed in the L2 during their final school-leaving exams (Abitur). Figure 1: Overview of the number of English lessons and teaching through English in full CLIL streams in North-Rhine Westphalia, Germany (horizontal line - average number of lessons in regular EFL classes). 268 Nils Jaekel <?page no="269"?> 1 Verordnung über den Bildungsgang in der Grundschule, 2005; Verordnung über die Ausbildung und die Abschlussprüfungen in der Sekundarstufe I (Ausbildungs- und Prüfungsordnung Sekundarstufe I - APO-S I), 2005; Merkblatt zum bilingualen Unter‐ richt in der gymnasialen Oberstufe, 2009. 2 Taking a 35 week, out of 40, school year into account as the basis as done so by the Department for Education in North-Rhine Westphalia to account for project weeks, school trips, and other missed or cancelled lessons www.schulentwicklung.nrw.de/ cm s/ umsetzungsbeispiele-zu-den-klp-g8/ englisch/ index.html [accessed May 31, 2021]. Figure 2: Differences of English lessons received by EFL and full CLIL stream students following curricular guidelines - Comparing full CLIL streams with average lessons in EFL 1 . In addition to full CLIL streams, variations other than modular CLIL exist. Some schools, for example, have decided to offer preparatory English classes in years 5 and 6 and offer alternating content subjects taught through the L2 during years 7 through 9 without increasing the overall number of CLIL subjects. The difference in the amount of exposure to English in full CLIL streams over the course of the first five years of secondary education is significantly larger. Over the course of seven years from year 3 in primary school until the end of year 9, CLIL students, on average, receive a total of 1505 lessons in English including content lessons taught through the medium of English compared to only 770 lessons of English for the mainstream EFL classes (see Figure 2). 2 3 Selection The process of becoming a CLIL student involves several key stakeholders: parents, students, primary and secondary school teacher(s), but also educational 269 Does a positive selection bias into CLIL streams explain higher language proficiency? <?page no="270"?> policymakers, i. e. local or regional governments as well as the European Council’s support for CLIL (also see Rumlich, 2013: 183-184). The European Union policies have been extremely supportive of multilingualism and CLIL in particular. Evidence for this support can be found in its language policies (Eu‐ ropean Commission, 1995), the backing of the Common European Framework of References for Languages (Council of Europe, 2001) and funding for research projects pertaining to CLIL and its implementation in school. Parents as stakeholders first of all need to be aware of CLIL streams, see or understand the value of the approach, and identify potential schools after they assess their child’s suitability for the more demanding CLIL stream in conjunction with teachers both at the primary and secondary level. Aware‐ ness for CLIL has increased significantly also due to programmes such as Bilingual für alle! (CLIL for everyone! ) in the state of North-Rhine-Westphalia, which encourages secondary schools to introduce CLIL as modules or in elective courses (Ministerium für Schule und Weiterbildung des Landes Nordrhein-Westfalen, 2015). The value of education and in particular of an approach that has the potential benefit of significantly increasing communicative compe‐ tence in a foreign language and concomitantly increases future prospects of working internationally or benefiting from English proficiency at the university plays an important role in opting for CLIL. Parents’ socio-economic status (SES) affects their expectations regarding the education their children will receive and thus impact the choices they make (for example Arnold & Doctoroff, 2003: 522). A higher SES may thus not only be linked to sending one’s children to a more prestigious grammar school, despite longer commutes to school but also to signing their children up for CLIL (also see Zydatiß, 2007: 113-114). On the basis of the results of the DEZIBEL study in Berlin, Zydatiß concludes that human resources and the socio-economic background of the students’ family have the highest influence on the selection of a learner for a CLIL stream (Zydatiß, 2007: 149). Taking these findings into consideration it seems that many CLIL programs select students thus promoting educational and socio-economic elites. This is very unfortunate since especially students with a migrant or lower socio-economic background could gain from this approach as was shown in the DESI study for regular EFL teaching (Hesse, Göbel & Hartig, 2008: 214). Students are, of course, also involved as stakeholders in this decision process. How much primary school children base their decision on their parents’ educa‐ tional aspiration cannot easily be answered. Commencing English learning and teaching in year 1 has at least increased the ability of students to assess their interest in English and possibly the extent they want to be involved in learning content through the foreign language. Ultimately, the question however remains 270 Nils Jaekel <?page no="271"?> unanswered whether students decide themselves to join a CLIL stream or if their parents simply enrol them and what role teacher recommendations from primary school have on this decision process. Throughout Europe, schools or rather teachers select or recommend students for CLIL on the basis of different methods. In France for example L1 and/ or L2 competence is used as a criterium (Eurydice, 2006: 21). In Hungary, content-spe‐ cific knowledge is included in the selection decision, whereas in Poland both content and language competences factor into the decision (Eurydice, 2006: 21). Even within countries such as Germany individual states do not have a standardised system in place which would allow for a transparent selection process. Methods of pre-screening suitable CLIL candidates might take German, Maths and English grades from primary school or primary school teachers’ recommendations into account. Interviews or written tests may provide further evidence for CLIL suitability (Wolff, 1997: 256). As the number of placements in CLIL streams is often limited to one class per year and due to the added workload for students in CLIL, schools resort to a selection process. Some schools, however, may offer more classes depending on their teaching staff ’s qualifications or the school’s profile (European School / in‐ ternational schools). Recently, schools in Germany have also started to offer the additional two hours of preparatory extra lessons allotted for CLIL to all students and postpone their selection process to year 6 instead of mid-year 4, prior to secondary school. In the majority of cases, however, particularly at grammar schools, selection currently seems to be an inevitable procedure of CLIL until the approach will be taken fully to the mainstream. In 2002, Marsh already argued that [g]uidelines should be drawn up to facilitate the inclusion of a broad range of learners in a framework that encompasses diverse models. This would help to unlock the potential of CLIL/ EMILE and facilitate mainstreaming (Marsh, 2002: 201). While selection is still a current issue in CLIL in many European countries, studies have shown that even low achievers or students who are considered to be at risk for academic underachievement can benefit greatly from this approach (Marsh, 2002: 175; Schwab, 2013). In the context of immersion programmes in Canada, Genesee has provided evidence that at-risk students generally perform equally well on the content side despite the foreign language environment while benefiting from the increased exposure to L2 and thus the potential to develop their own “functional L2 proficiency” (Genesee, 2006: 561, 566). More evidence supporting the findings from Canada is provided by the results of the DESI study. Multilingual students performed slightly better in English tests than their 271 Does a positive selection bias into CLIL streams explain higher language proficiency? <?page no="272"?> German L1 counterparts (Hesse et al., 2008: 214). Therefore, every student may benefit from approaches such as CLIL. 4 Research into CLIL Over the past decade, research into CLIL has flourished, focusing mostly on the impact this approach has on language proficiency, while content subject outcomes have received less attention. Studies have considered different aspects of CLIL such as individual language skills (for example Bos, Bonsen & Gröh‐ lich, 2009; Dalton-Puffer, 2007; Klieme, 2008; Rumlich, 2016; Zydatiß, 2007), content learning (Bonnet, 2002; Dallinger, Jonkmann, Hollm & Fiege, 2016; Osterhage, 2009; Pérez-Cañado, 2018), attitudes and motivation (Lasagabaster, 2011, 2019), language learning strategy use ( Jaekel, 2018; Zydatiß, 2007) or academic self-concept and self-efficacy ( Jaekel, 2018; Rumlich, 2016). Large scale and longitudinal studies scrutinising and validating the approach are however rare. Furthermore, cross-national studies validating the approach thoroughly, across borders and within a longitudinal study design have not been conducted. Investigating CLIL across borders, researchers face significant obstacles due to different approaches and understanding of the approach. In order to validate the approach internationally, more rigorous testing beyond (limited) regional studies is required. Currently, results need to be considered within the respective national or regional context and can only to a limiting degree be transferred. In Sweden, for example, results in CLIL have not confirmed the success story written in other European countries (Sylvén, 2013). There is thus still a dire need for research, even more so as CLIL is expanding further it will require more accompanying scientific research. 4.1 Research on language achievement in CLIL Research into CLIL has rapidly increased with several large scale and even more smaller-scale studies investigating the difference in English language pro‐ ficiency between CLIL and regular EFL students. Generally, CLIL students have been found to make faster progress in acquiring a foreign language. Considering the increased exposure and opportunity to use the foreign language, the rapid progression is a calculated outcome which leads to attaining levels of language proficiency that students in regular EFL classes only realise years later, if at all (Theis & Werkmann, 2004: 147-149). A popular and highly validated instrument to assess students’ general language proficiency in CLIL and EFL classes used in many studies is the C-Test. Based on reduced redundancy, the mutilated text has to be “repaired” 272 Nils Jaekel <?page no="273"?> by the students. Studies at different stages from years 6 to 9 have employed different C-Tests with CLIL students overwhelmingly outperforming EFL students in every study (Bos et al., 2009: 30; Jaekel, 2018: 252; Rumlich, 2013: 191-193; Zydatiß, 2007: 170, see also Dallinger, this volume and Rumlich, this volume). These results are also reflected in studies investigating other language skills. Studies comparing writing skills of CLIL and EFL students have generally concluded that the former group’s skills are superior, although with varying degrees. Texts from CLIL students are usually longer and stand out due to their richness both lexically and syntactically including the use of a wider array of tenses (Ackerl, 2007: 10; Mewald, 2007: 161; Zydatiß, 2007: 195). Overall higher complexity of text production correlates with a high degree of accuracy (Zydatiß, 2007: 191). Similar results have been found with regard to speaking skills. Studies have found that CLIL students speak faster, a key indicator for fluency (Mewald, 2007: 160; Zydatiß, 2007: 251), use more content words (Zydatiß, 2007: 257) and produce longer sentences than their EFL peers (Burmeister & Daniel, 2002; Mewald, 2007: 160). Concerning receptive language skills, the DESI study reported that CLIL students performed one competence level higher than the EFL group in both reading and listening (Nold, Hartig, Hinz & Rossa, 2008: 454). These results have been confirmed through the DEZIBEL study (Zydatiß, 2007: 251), and for reading Bredenbröker concludes that CLIL students outperform their peers and make significantly faster progress within the first year (Bredenbröker, 2002: 144). The latter results suggest that students were either selected prior to CLIL or at least ‘well-guided’. In her meta-analysis, Ruiz de Zarobe comes to the conclusion that for many skill areas, such as reading, listening, speaking, writing, receptive vocabulary as well as affective outcomes, CLIL students are at a clear advantage (Ruiz de Zarobe, 2011: 145). Syntax, productive vocabulary, informal/ non-technical language, accuracy in writing and pronunciation were however not as positively affected as the aforementioned skill areas (Ruiz de Zarobe, 2011: 146). 4.2 Research on student background While research has focused mainly on language and content achievement, less is known about the profile of CLIL students. The question is whether or not there is an inherent bias of those parents or students who opt for CLIL. Potential benefits but also the additional workload students will have to endure may be decisive factors for or against CLIL. Few studies have reported extensively on student characteristics such as gender, SES, cognitive abilities or affective 273 Does a positive selection bias into CLIL streams explain higher language proficiency? <?page no="274"?> factors comparing CLIL and EFL students (for example Möller, Fleckenstein, Hohenstein, Preusler, Paulick & Baumert, 2018; Zydatiß, 2007). These studies are no evidence of active student selection, but there may be a much more complex interplay of individual difference variables in choosing or being chosen for CLIL. The extensive DEZIBEL study reported on additional factors involving CLIL choice and individual differences. While there was no difference according to SES, fewer non-L1 German students were signed-up for CLIL (Zydatiß, 2007: 107, 116). Furthermore, the decision to join a CLIL stream was more often influenced by parents and primary school teachers than within regular EFL courses (Zydatiß, 2007: 113). Bredenbröker’s study has provided evidence for pre-selection or ‘guidance’ for CLIL streams, as even at the beginning of secondary school bilingual students already performed better than their non-CLIL peers. After only one year of bi‐ lingual education, students’ reading comprehension had improved significantly, while the traditionally taught students only slightly improved; after a second year, reading comprehension in both groups stabilised (Bredenbröker, 2002: 144). These results were confirmed by another large-scale (N = 2,149), longitudinal study, KESS 7 in Hamburg, Germany, which investigated student development from year 4 to the end of year 6. While the focus was not set on CLIL, this popular approach was included in analyses. At the end of year 4, future CLIL students (n = 218) already significantly outperformed their peers in English listening (Cohen’s d = .50; Bos et al., 2009: 6). Over the course of the first two years of secondary school, this difference continued to increase significantly (Cohen’s d = .98; Bos et al., 2009: 41). This significant difference in English proficiency in year 4 is evidence that CLIL streams draw or select excellent students who go on to thrive in the CLIL environment. Similarly to KESS 7, the GanzIN study investigated the development of students from year 5 to year 9 at grammar schools in North-Rhine-Westphalia, Germany (N = 3,218). Students enrolled in CLIL in year 5 (n = 223) performed significantly better in listening tests at the beginning of year 5 compared to EFL students (Ritter, Jaekel, Meister & Lewandowska, 2015: 433). An interesting development could be recognised for a third group (n = 63) which received two hours of preparation regardless of future CLIL participation. This group started with the lowest scores in listening in year 5, by year 7, however, this group reported the best results, scoring slightly higher than the predetermined CLIL group (Ritter et al., 2015: 432-433). 274 Nils Jaekel <?page no="275"?> 5 Empirical study The current sample was collected at nine grammar schools in Duisburg, Essen, Herne and Bochum, all situated in the characteristically working-class Ruhr district in Germany. Overall, nine CLIL and seven regular EFL classes participated in the study. The EFL sample was solely collected at schools that did not offer any form of CLIL at the time in order to avoid complications with regard to student selection. In total, 378 students participated. Parents’ SES was assessed through the employment information / highest education provided by the students. Using ISCO conversions the parents’ employment was coded into ISEI scores which are regularly used by large scale studies such as PISA to ascertain family SES (see Ganzeboom, de Graaf & Treiman, 1992; Maaz, Trautwein, Gresch, Lüdtke & Watermann, 2009). To assess students’ cognitive abilities the CFT-20 was administered. Due to time constraints, part one of the instrument was used as it is more economical and provides a reliable account of non-verbal cognitive abilities, using figural analogy (Weiß, 2006). Language proficiency was assessed through the last grade in English and a four text C-Test battery (see Zydatiß, 2007: 165). Both descriptive statistics and regression and correlation-based structural equation models (SEM) were used to understand whether English language proficiency in CLIL streams and regular EFL can mostly be explained through CLIL as a didactic approach or students’ individual differences. The great advantage of SEM models is that predictor variables can be controlled for and constructs such as language proficiency consisting of both grades and language tests can be validated through confirmatory factor analysis (In’namia & Koizumi, 2011). The present study is limited to five predictor variables, CLIL as a mediating variable and the outcome variable language proficiency consisting of the English grade and a C-Test battery. Comparing both groups through T-Tests, the descriptive results already highlight significant differences between the EFL and CLIL group (see Table 1). In the present study, learners in CLIL were significantly younger (Cohen’s d = .32), came from a higher SES family (Cohen’s d = -.51) and scored higher on the CFT-20 (Cohen’s d = -.28). Furthermore, CLIL students on average received better English grades (d = .51) and outperformed their EFL peers in the C-Test (Cohen’s d = -.94), which is understood as a general language proficiency test (for example Eckes & Grotjahn, 2006: 290-291). 275 Does a positive selection bias into CLIL streams explain higher language proficiency? <?page no="276"?> EFL CLIL t p d N = 176 N = 202 Age 14.49 (.594) 14.29 (.569) 3.352 .001 .34 Gender 1.51 (.501) 1.56 (.497) -1.139 .255 -.10 SES 47.19 (12.39) 53.90 (13.88) -4.969 <.001 -.51 L1 German / non-German .31 (.565) .32 (.606) -.072 .943 -.02 Cognitive abilities 105.39 (12.80) 109.24 (14.55) -2.737 .006 -.28 Last grade in English (1-6) 2.82 (.906) 2.37 (.873) 4.929 <.001 .51 C-Test score (max. 100) 57.98 (11.47) 68.52 (11.02) -9.079 <.001 -.94 Table 1: Mean values (standard deviations) of outcome and main predictor variables including independent samples T-Tests. Controlling for all predictor variables and using CLIL as a mediating variable, the differences between the EFL and CLIL groups uncovered through T-Testing are confirmed in the SEM model below. Participation in CLIL was significantly predicted by SES (β = .24), cognitive ability (β = .12) and age (β = -.14). CLIL students came from a higher SES family background, scored higher on the CFT-20 and were younger than their peers in regular EFL classes. The SEM model was able to explain 10 % of the CLIL/ EFL affiliation’s variance. The language proficiency construct of C-Test and English grade was con‐ firmed through factor analysis. The SEM model was able to account for a total of 34 % of variance in language proficiency. The language proficiency construct was significantly predicted by family SES β = .21 (direct β = .12; indirect β = .09), cognitive abilities β = .28 (β = .23; β = .05), gender β = .25 (β = .22; β = .03) and CLIL/ EFL β = .38. The overall effect consists of a direct and indirect component, the latter is mediated through the CLIL/ EFL variable. Direct effects can be seen in Figure 4, for details on direct indirect effects see Figure 3. Students of a higher SES, scoring higher on the CFT, and girls attained a better overall language proficiency compared to their peers. CLIL as a mediating variable has the strongest overall positive impact on language proficiency. 276 Nils Jaekel <?page no="277"?> 3 These fit indices with X ² non-significant, a CFI > .95 and RMSEA <.05 support that the model is statistically sound (Iacobucci (2010: 91); In’namia & Koizumi (2011: 252)). Figure 3: SEM model predicting language proficiency of CLIL and EFL students SEM modelfit: X ² = 8.585; df = 5 p = .127; CFI = .987; RMSEA = .044, PCLOSE = .518  3 . Figure 4: Direct and indirect effect of the SEM model (Figure 3) predicting language proficiency. 277 Does a positive selection bias into CLIL streams explain higher language proficiency? <?page no="278"?> 6 Discussion This study set out to investigate whether student characteristics in CLIL streams differed significantly from those in regular EFL classes and, most importantly, if those differences explained the expected better language proficiency scores among CLIL students. Student characteristics were shown to differ significantly between CLIL streams and regular EFL classes with regard to age, SES and cognitive abilities. As expected, CLIL students performed significantly better on the C-Tests and reported receiving better English grades than their peers in EFL classes. Structural equation modelling, however, showed that controlling for the positive selection bias into CLIL classes did not change the strong positive effect of CLIL on language proficiency. A fairly consistent finding in large-scale assessments has been that CLIL stu‐ dents already perform better in different language skills prior to entering CLIL, underlining the selective nature of CLIL streams. Opting for CLIL was so far not shown to be directly linked to SES whereas parent and teacher involvement was found to be a decisive factor (Ioannou Georgiou, 2012: 496; Zydatiß, 2007: 147-148). Additionally, Zydatiß reported that there were fewer non-L1 German CLIL students compared to regular EFL classes (2007). These latter findings are unfortunate in the light of evidence that well-implemented content-based bilingual education can be as beneficial for high and low achievers (Genesee, 2006: 561, 566; Schwab, 2013). Moreover, the DESI study also highlighted slight advantages of students with a different home language than German in regular EFL classes (Hesse et al., 2008: 214). The present study did not determine an impact of L1 on language proficiency or a selection based on L1. Individual differences such as age, gender, SES and cognitive abilities, amongst other factors, are well known to impact academic achievement in general, but also language proficiency. Large-scale assessments such as DESI, KESS or GanzIn have provided ample evidence for a prevailing link of these factors in language learning within the German school context (Bos et al., 2009; Nold et al., 2008; Ritter et al., 2015). As expected, the descriptive statistics results already emphasised significant differences between the two student populations (see Table 1). CLIL students were significantly younger, came from a higher SES family background and have significantly higher cognitive abilities. The reported English grades were also significantly better in contrast to what Zydatiß called ‘Gerechtigkeitsfalle’, i. e. despite higher language proficiency CLIL students may receive worse grades than lower-achieving EFL students due to grade differentiation within classes (Zydatiß, 2006: 381). CLIL students’ general language proficiency C-Test score also significantly exceeded EFL 278 Nils Jaekel <?page no="279"?> students’ scores. Cohen’s d as a measure of the size of the effect with .51 and -.94, respectively, both reflect the meaningful difference in language proficiency between the two groups. Although significant differences were found, it is difficult to ascertain whether school selection, ambitious parents or interested students most ac‐ counted for CLIL participation. The results suggest that families, even within the more prestigious grammar school context, differ significantly with regard to the educational aspirations they have for their offspring and this is why SES seems to be a decisive factor. Moreover, CLIL is a more demanding stream, not only with respect to time but also cognition. This may be one reason why students with higher scores on the cognitive abilities tests were more often enrolled in CLIL. C-Tests have, however, also been shown to be predicted by cognitive abilities (Raatz, 2002: 180-182). Additionally, teacher recommen‐ dations from primary schools encouraging students who excel academically as well as grade-dependent preselection at grammar schools may be another explanation for the over-representation of certain student groups in CLIL. Another interesting result is that CLIL students were slightly younger and had attained better grades and higher C-Test scores. Similar results were reported for the GanzIn study for students enrolled in regular EFL classes. One explanation for this finding could be that due to the German school entrance regulations, younger versus older students in one cohort, i. e. those entering first grade early versus delayed ( Jaekel, Strauss, Johnson, Gilmore & Wolke, 2015) may be of higher SES family background and, in addition, have higher academic abilities. In the SEM analyses, the language proficiency latent variable consisted of both C-Test scores and the students’ last English grades. Taking both direct and indirect effects into account, cognitive abilities, gender and SES all showed a similarly strong positive impact on language proficiency. Cognitive abilities, however, had the strongest positive impact. Cognitive abilities have been shown to correlate with academic achievement and C-Testing (Raatz, 2002: 180-182). Gender showed the second-largest impact on language proficiency. Language learning has been considered a rather female domain which these results confirm (Pajares, 2002: 122-123). The strongest overall impact on language proficiency, however, was exerted by CLIL/ EFL - enrolment in CLIL streams was thus highly beneficial for students’ language competence, even after statistically controlling for confounding factors. The CLIL approach with its extensive L2 exposure and authentic content-related communication is thus a key factor in predicting high language proficiency scores. Certain limitations to this study must be considered. In future studies, an even wider field of measures should be accounted for, including primary school 279 Does a positive selection bias into CLIL streams explain higher language proficiency? <?page no="280"?> teacher evaluations, parent and/ or student questionnaires, statements of the involved schools regarding their approach to student selection, as well as school grades from primary school to provide an even more detailed understanding of student, parent or school-based choices. CLIL itself is a success and participating in a stream has a considerable impact on students’ achievement, which is, con‐ sidering the double amount of language exposure, an expected and calculated outcome. In addition, this study showed that the previously reported positive effects of CLIL are not explained by the existing selection bias. Future research should thus expand the research focus much more on lower-tier schools’ implementation of CLIL than has been done in the past. References Ackerl, C. 2007. Lexico-grammar in the essays of CLIL and non-CLIL students: Error analysis of written production. Vienna English Working Papers, 16 (3), 6-11. Arnold, D.H. & Doctoroff, G.L. 2003. The early education of socioeconomically disadvan‐ taged children. Annual Review of Psychology, 54 (1), 517-545. Biederstädt, W., Harger, L. & Schwarz, H. 2008. English G 21 A für Gymnasien. Berlin: Cornelsen. Bonnet, A. 2002. 47 % - Das Spracherwerbspotenzial englischsprachigen Chemieunter‐ richts. In: Breidbach, S., Bach, G. & Wolff, D. (Eds.), Bilingualer Sachfachunterricht. Didaktik, Lehrer-/ Lernerforschung und Bildungspolitik zwischen Theorie und Empirie. Frankfurt am Main: Peter Lang. 125-139. Bos, W., Bonsen, M. & Gröhlich, C. (Eds.). 2009. KESS 7 - Kompetenzen und Einstellungen von Schülerinnen und Schülern an Hamburger Schulen zu Beginn der Jahrgangsstufe 7. Münster: Waxmann. Bredenbröker, W. 2002. Förderung der fremdsprachlichen Kompetenz durch bilingualen Sachfachunterricht: Empirische Untersuchungen. In: Breidbach, S., Bach, G. & Wolff, D. (Eds.), Bilingualer Sachfachunterricht. Didaktik, Lehrer-/ Lernerforschung und Bil‐ dungspolitik zwischen Theorie und Empirie. Frankfurt am Main: Peter Lang. 141-150. Burmeister, P. & Daniel, A. 2002. How effective is late partial immersion? Some findings from a secondary school program in Germany. In: Burmeister, P., Piske, T. & Rohde, A. (Eds.), An Integrated View of Language Development. Papers in Honor of Henning Wode. Trier: Wissenschaftlicher Verlag Trier. 499-515. Cinganotto, L. 2016. CLIL in Italy: A general overview. Latin American Journal of Content and Language Integrated Learning, 9 (2), 374-400. Council of Europe. 2001. Common European Framework of Reference for Languages: Learning, Teaching, Assessment. Cambridge: Press Syndicate of the University of Cambridge. 280 Nils Jaekel <?page no="281"?> Coyle, D. 2007. Content and language integrated learning: Towards a connected research agenda for CLIL pedagogies. International Journal of Bilingual Education and Bilin‐ gualism, 10 (5), 543-562. Dalton-Puffer, C. 2007. Discourse in Content and Language Integrated Learning (CLIL) Classrooms. Amsterdam: John Benjamins Pub. Dallinger, S., Jonkmann, K., Hollm, J. & Fiege C. 2016. The effect of content and language integrated learning on students’ English and history competences - Killing two birds with one stone? Learning and Instruction, 41, 23-31. Eckes, T. & Grotjahn, R. 2006. A closer look at the construct validity of C-tests. Language Testing, 23 (3), 290-325. European Commission. 1995. White Paper on Education and Training: Teaching and Learning. Towards the Learning Society. https: / / op.europa.eu/ en/ publication-detail/ -/ publication/ d0a8aa7a-5311-4eee-904c-98fa541108d8/ language-en. May 31, 2021. Eurydice. 2006. Content and Language Integrated Learning (CLIL) at School in Europe. Bruxelles: European Unit. Ganzeboom, H.B.G., de Graaf, P.M. & Treiman, D.J. 1992. A standard international socioeconomic index of occupational status. Social Science Research, 21 (1), 1-56. Genesee, F. 2006. What do we know about bilingual education for majority language students? In: Bhatia, T.K. & Ritchie, W.C. (Eds.), Blackwell Handbooks in Linguistics. The Handbook of Bilingualism. Malden, MA: Blackwell Pub. 547-576. Hesse, H.-G., Göbel, K. & Hartig, J. 2008. Sprachliche Kompetenzen von mehrsprachigen Jugendlichen und Jugendlichen nicht-deutscher Erstsprache. In: Klieme, E. (Ed.), Unterricht und Kompetenzerwerb in Deutsch und Englisch. Ergebnisse der DESI-Studie. Weinheim: Beltz. 208-230. Horner, M., Ashford, S., Baer-Engel, J., Daymond, E., Dennis, P. & Plitsch, A. 2007. Green Line 3. Stuttgart: Ernst Klett Verlag. Iacobucci, D. 2010. Structural equations modeling: Fit indices, sample size, and advanced topics. Journal of Consumer Psychology, 20 (1), 90-98. In’namia, Y. & Koizumi, R. 2011. Structural equation modeling in language testing and learning research: A review. Language Assessment Quarterly: An International Journal, 8 (3), 250-276. Ioannou Georgiou, S. 2012. Reviewing the puzzle of CLIL. ELT Journal, 66 (4), 495-504. Jaekel, J., Strauss, V.Y.-C., Johnson, S., Gilmore, C. & Wolke, D. 2015. Delayed school entry and academic performance: A natural experiment. Developmental Medicine and Child Neurology, doi: 10.1111/ dmcn.12713. Jaekel, N. 2018. Language learning strategy use in context: The effects of self-efficacy and CLIL on language proficiency. International Review of Applied Linguistics in Language Teaching, 58 (2), 195-220. 281 Does a positive selection bias into CLIL streams explain higher language proficiency? <?page no="282"?> Klieme, E. (Ed.). 2008. Unterricht und Kompetenzerwerb in Deutsch und Englisch: Ergebnisse der DESI-Studie. Weinheim: Beltz. Lasagabaster, D. 2011. English achievement and student motivation in CLIL and EFL settings. Innovation in Language Learning and Teaching, 5 (1), 3-18. Lasagabaster D. 2019. Motivation in Content and Language Integrated Learning (CLIL) research. In: Lamb M., Csizér K., Henry A. & Ryan S. (Eds.), The Palgrave Handbook of Motivation for Language Learning. London: Palgrave Macmillan. 347-366, doi: 10.1007/ 978-3-030-28380-3_17. Maaz, K., Trautwein, U., Gresch, C., Lüdtke, O. & Watermann, R. 2009. Intercoder-Relia‐ bilität bei der Berufscodierung nach der ISCO-88 und Validität des sozioökonomischen Status. Zeitschrift für Erziehungswissenschaft, 12 (2), 281-301. Marsh, D. (Ed.). 2002. CLIL/ EMILE: The European Dimension: Actions, Trends and Foresight Potential. https: / / jyx.jyu.fi/ bitstream/ handle/ 123456789/ 47616/ 1/ david_marsh-report. pdf. May 31, 2021. Mewald, C. 2007. A comparison of oral foreign language performance of learners in CLIL and in mainstream classes at lower secondary level in Lower Austria. In: Dalton-Puffer, C. & Smit, U. (Eds.), Empirical Perspectives on CLIL Classroom Discourse. Frankfurt am Main: Peter Lang. 137-177. Ministerium für Schule und Weiterbildung des Landes Nordrhein-Westfalen. 2015. Bilingualer Unterricht in Nordrhein-Westfalen. https: / / www.schulministerium.nrw/ themen/ schulsystem/ unterricht/ lernbereiche-un d-unterrichtsfaecher/ bilingualer-unterricht-nordrhein. May 31, 2021. Möller, J., Fleckenstein, J., Hohenstein, F., Preusler, S., Paulick, I. & Baumert, J. 2018. Varianten und Effekte bilingualen Lernens in der Schule. Zeitschrift für Erziehungs‐ wissenschaft, 21 (1), 4-28. Nold, G., Hartig, J., Hinz, S. & Rossa, H. 2008. Klassen mit bilingualem Sachfachunterricht: Englisch als Arbeitssprache. In: Klieme, E. (Ed.), Unterricht und Kompetenzerwerb in Deutsch und Englisch. Ergebnisse der DESI-Studie. Weinheim: Beltz. 451-457. Osterhage, S. 2009. Sachfachkönnen (scientific literacy) bilingual und monolingual unterrichteter Biologieschüler: Ein Kompetenzvergleich. In: Caspari, D., Hallet, W., Wegner, A. & Zydatiß, W. (Eds.), Bilingualer Unterricht macht Schule. Beiträge aus der Praxisforschung. 2 nd ed. Frankfurt am Main: Peter Lang. 41-50. Pajares, F. 2002. Gender and perceived self-efficacy in self-regulated learning. Theory Into Practice, 41 (2), 116-125. Pérez-Cañado, M.L. 2012. CLIL research in Europe: Past, present, and future. International Journal of Bilingual Education and Bilingualism. 15 (3), 315-341. Pérez-Cañado, M.L. 2018. The effects of CLIL on L1 and content learning: Updated empirical evidence from monolingual contexts. Learning and Instruction, 57 (1), 18-33. 282 Nils Jaekel <?page no="283"?> Pérez-Cañado, M.L., Coyle, D., Ting, T., Nikula, T., Dalton-Puffer, C., Matz, F. & Jerez Montoya, T. 2019. ADiBE Project, CLIL for All: Attention to Diversity in Bilingual Education. https: / / adibeproject.com. May 31, 2021. Raatz, U. 2002. C-tests and intelligence. In: Coleman, J.A., Grotjahn, R. & Raatz, U. (Eds.), University Language Testing and the C-test. Bochum: AKS-Verlag. 169-185. Ritter, M., Jaekel, N., Meister, C. & Lewandowska, Z. 2015. Zwischenbilanz der fachdi‐ daktischen Arbeit im Fach Englisch. In: Wendt, H. & Bos, W. (Eds.), Wissenschaftsband Projekt Ganz In. Münster: Waxmann. 413-442. Ruiz de Zarobe, Y. 2011. Which language competencies benefit from CLIL? An insight into applied linguistics research. In: Ruiz de Zarobe, Y., Sierra, J.M. & Puerto, F.G. (Eds.), Content and Foreign Language Integrated Learning. Contributions to Multilingualism in European Contexts. Bern: Peter Lang. 129-155. Rumlich, D. 2013. Students’ general English proficiency prior to CLIL: Empirical evidence for substantial differences between prospective CLIL and non-CLIL students in Germany. In: Breidbach, S. & Viebrock, B. (Eds.), Content and Language Integrated Learning (CLIL) in Europe. Research Perspectives on Policy and Practice. Frankfurt am Main: Peter Lang. 181-201. Rumlich, D. 2016. Evaluating Bilingual Education in Germany. CLIL Students’ General English Proficiency, EFL Self-Concept and Interest. Frankfurt am Main: Peter Lang. Schwab, G. 2013. Bili für alle? Ergebnisse und Perspektiven aus einem Forschungsprojekt zur Einführung bilingualer Module in einer Hauptschule. In: Breidbach, S. & Viebrock, B. (Eds.), Content and Language Integrated Learning (CLIL) in Europe. Research Perspec‐ tives on Policy and Practice. Frankfurt am Main: Peter Lang. 297-314. Sekretariat der Ständigen Konferenz der Kultusminister der Länder und der Bundesre‐ publik Deutschland. 2013. Konzepte für den bilingualen Unterricht - Erfahrungsbericht und Vorschläge zur Weiterentwicklung. https: / / www.kmk.org/ fileadmin/ Dateien/ ver oeffentlichungen_beschluesse/ 2013/ 201_10_17-Konzepte-bilingualer-Unterricht.pdf. May 31, 2021. Sylvén, L.K. 2013. CLIL in Sweden - why does it not work? A metaperspective on CLIL across contexts in Europe. International Journal of Bilingual Education and Bilingualism, 16 (3), 301-320. Theis, R. & Werkmann, R. 2004. Kann der Englischunterricht von bilingualen Modellen lernen? Eine empirische Untersuchung. Neusprachliche Mitteilungen aus Wissenschaft und Praxis, 57 (3), 145-155. Weiß, R. 2006. CFT 20-R. Grundintelligenztest Skala 2. Manual. Göttingen: Hogrefe Verlag. Wolff, D. 1997. Strategien des Textverstehens: Was wissen Fremdsprachenlerner über den eigenen Verstehensprozess? In: Rampillon, U. (Ed.), Strategien und Techniken beim Erwerb fremder Sprachen. Ismaning: Hueber. 270-289. 283 Does a positive selection bias into CLIL streams explain higher language proficiency? <?page no="284"?> Zydatiß, W. 2006. Bilingualer Unterricht = höhere Sprachfähigkeit: Konsequenzen für die Bewertungspraxis an deutschen Schulen. In: Timm, J.-P. & Vollmer, H.J. (Eds.), Gießener Beiträge zur Fremdsprachendidaktik. Fremdsprachenlernen und Fremdspra‐ chenforschung. Kompetenzen, Standards, Lernformen, Evaluation: Festschrift für Helmut Johannes Vollmer. Tübingen: Narr. 372-388. Zydatiß, W. 2007. Deutsch-Englische Züge in Berlin (DEZIBEL): Eine Evaluation des bilingualen Sachfachunterrichts an Gymnasien: Kontext, Kompetenzen, Konsequenzen. Frankfurt am Main: Peter Lang. 284 Nils Jaekel <?page no="285"?> Speed of automatisation predicts performance on “decorative” grammar in second language learning Ewa Dąbrowska / Ashley Blake Abstract A large body of research shows that adult second language learners have particular difficulty with what we call ‘decorative’ grammar (aspects of grammar which have abstract and largely redundant meanings), while they often do well on tests of ‘functional’ grammar (those aspects of grammar which make clear contributions to meaning). We suggest that ‘decorative’ grammar is challenging for L2 learners because when grammatical routines are not automatised, speakers will tend to focus their processing resources on those aspects of language that are most relevant to meaning. We examine the relationship between ‘functional’ and ‘decorative’ grammar, assessed using a grammaticality judgement task (GJT) and a picture selection task (PST), respectively, and two cognitive abilities which are hypothesised to support language learning, namely the ability to autom‐ atise a complex cognitive procedure (assessed using the Multiple-trial Tower of Hanoi task, MToH) and explicit language aptitude (assessed using the Language Analysis test from the PLAB). As predicted, we found a positive correlation between performance on the MToH and the GJT, while Language Analysis predicted performance on both the PST and the GJT. We did not find a relationship between the MToH and the PST, suggesting that ‘functional’ grammar is less reliant on automatised routines. 1 Introduction 1.1 Language as a skill Since the cognitive revolution, linguists have tended to think of language, and in particular, grammar, as a body of knowledge - either a set of rules or principles (e.g. Chomsky, 1965, 1981) or an inventory of conventional linguistic units <?page no="286"?> (e.g. Langacker, 1987). It is undeniably true that knowing a language involves possessing a large body of knowledge, including a considerable amount of knowledge about the idiosyncratic properties of individual lexemes or classes of lexemes. However, knowledge alone is not sufficient. In order to be able to make a contingent contribution to a conversation, a speaker must decode the previous speaker’s utterance, decide on their own response, retrieve the relevant lexical items, combine them in a grammatically appropriate way, and launch the appropriate articulatory sequence. This procedure is executed very quickly: The gaps between individual speakers’ turns in conversational interchange are on average only 0.5 seconds long, and often they are even shorter. It is not unusual for conversational turns to overlap by about a syllable, which means that the second speaker has to do all of the above while the previous speaker is still executing their turn (Dąbrowska, 2004). This kind of performance requires an enormous amount of skill, and a number of researchers have drawn parallels between language use and skilled behaviours such as driving a car, touch-typing or playing a musical instrument. Johnson (1996), for example, points out both are goal-directed behaviours that are hierarchically organised and non-stereotyped, and both involve a selection from a large repertoire of options and the ability to combine units, often under time pressure. He explains how insights from research on skill acquisition can be applied to second language learning and teaching. A broadly similar view of L2 learning is advocated by DeKeyser (2015), while other researchers (e.g. Chater & Christiansen, 2018; Chater, McCauley & Christiansen, 2016; Kamhi, 2019) have argued that language is a complex cognitive skill that develops through repeated use and practice before becoming effortless and automatic. Thus, research on skill learning is also relevant to first language development and processing. Psychologists working on skill learning have identified three stages in the acquisition of skills, namely the cognitive, associative and autonomous phase (Anderson, 1982, 1993; Anderson, Bothell, Byrne, Douglass, Lebiere & Qin, 2004; Anderson, Greeno, Reder & Simon, 2000; Fitts & Posner, 1967; Taatgen, Huss, Dickison & Anderson, 2008). In the cognitive stage, a new procedure is generated using declarative knowledge and working memory. During this stage, performance is slow and effortful, and error rates are relatively high. During the associative stage, declarative knowledge is incorporated into the procedures necessary to perform the skill. This stage sees a reduction in cognitive involvement and a gradual increase in the speed of processing. Finally, in the autonomous stage, procedural knowledge becomes increasingly automatic. This is evidenced by a further decrease in reaction times and error rates. In this stage, performance depends entirely on procedural memory and 286 Ewa Dąbrowska / Ashley Blake <?page no="287"?> makes minimal demands on attentional resources, enabling skilled performers to direct their attention to other matters. The behavioural changes observable during skill acquisition (i.e. reduction in error rates and speed) are very rapid in the early stages of learning (i.e. in the cognitive phase), then gradually level off. This decrease can be modelled by a power function (the ‘power law of practice’: Newell & Rosenbloom, 1981). Beyond the cognitive stage, the improvement in performance is very slow; however, psychological and neuroscientific research suggests that important qualitative changes are still happening “underground”, so to speak. This is evidenced by the fact that different cognitive abilities are associated with performance at different phases of skill learning (Ackerman, 1992; Beaunieux, Hubert, Witkowski, Pitel, Rossi, Danion, Desgranges & Eustache, 2006; Hubert, Beaunieux, Chetelat, Platel, Landeau, Danion, Viader, Desgranges & Eustache, 2007). In the cognitive phase, individual differences in performance are predicted by differences in (non-verbal) intelligence, working memory and episodic memory. During the associative phase, episodic memory is no longer relevant, but intelligence and working memory continue to play a role (albeit not as pronounced as in the cognitive phase). In addition, psychomotor abilities become significant predictors (although their role is less prominent than in the autonomous phase). The associative phase can thus be considered a ‘mixed phase’, as only parts of the skill are proceduralised and there is continued involvement of working memory and intelligence. Finally, in the autonomous phase, differences in performance depend primarily on psychomotor abilities. There is also evidence that different areas of the brain are involved during different phases of skill learning. For example, Hubert et al. (2007) found that the regions activated during the cognitive phase included the prefrontal cortex, the cerebellum, and the parietal regions. During the associative phase, higher levels of activation were observed in the occipital regions, right thalamus, and caudate nucleus, while the final, autonomous phase activated the left thalamus and caudate nucleus. These findings dovetail with the behavioural changes described earlier: The involvement of the frontoparietal network in the cognitive stage indexes the use of problem-solving strategies; the involvement of the occipital regions during the associative and the autonomous phase suggests the use of mental imagery, while the involvement of the cerebellum in the autonomous phase fits in well with the fact that performance during this phase depends primarily on psychomotor abilities. 287 Speed of automatisation predicts performance on “decorative” grammar <?page no="288"?> 1.2 Language acquisition from a usage-based perspective According to usage-based models of language development (Behrens, 2009; Bybee, 2010; Ellis, Römer & O’Donnell, 2016; Goldberg, 2006; Tomasello, 2003), grammar learning involves learning constructions or form-meaning pairings of varying size and degree of specificity. Learning a construction involves at least three stages. In the first stage, the learner acquires a database of analysed exemplars. This involves segmenting complex forms into smaller chunks and matching these chunks of form with salient semantic substructures. For example, a learner who hears the utterance where’s the cat? in a situational context where it is clear that the speaker is asking about the location of a particular member of the species felis catus might be able to link part of the phonological form, namely / ˈkæt/ to the semantic substructure corresponding to the animal, and the form / ˈwɛǝr/ to the semantic substructure corresponding to a request for information about where something is located, as in the left-hand part of Figure 1. (Note that the segmentation and matching process could be partial, in the sense that some parts of the phonological form may initially not be attached to any semantic substructure. In the hypothetical example represented in the figure, the phonological forms corresponding to the definite article and the copula are not yet associated with any specific meaning). The second stage involves generalisation. This is a crucial stage since this is what allows learners to produce novel utterances and can be thought of as involving a proportional analogy, where a previously learnt exemplar (or several exemplars) are used as a model in order to infer a missing term in a target. For example, imagine a learner who has learnt the phrase where’s the cat? but does not initially know how to ask about the location of a pteranodon, although s/ he knows the word pteranodon. Our learner can construct the novel form where’s the pteranodon? on analogy to where’s the cat? by retaining the matching elements (where’s the…) and replacing the non-matching element (cat) with new content (in this case, pteranodon), as illustrated in Figure 1. Analogical generalisation is a process that involves retrieving an appropriate analogical base, establishing correspondences between the component units of the base and target at both the semantic and the phonological level and projecting inferences about the missing term. This process is quite slow and effortful the first time it is performed. However, each subsequent analogical mapping is facilitated by previous experience, so that over time, the procedure for performing the mapping gradually becomes routinised and optimised. Thus, the final stage of the acquisition of a construction involves the entrenchment, or automatisation, of the operations needed to generate novel forms. 288 Ewa Dąbrowska / Ashley Blake <?page no="289"?> Figure 1: Generalisation as a proportional analogy problem. The first two stages of the acquisition of constructions (accumulation of ana‐ lysed exemplars and generalisation) correspond to the cognitive phase of skill acquisition, the stage that relies on declarative knowledge and slow, effortful processing. The final stage of entrenchment corresponds to the associative and autonomous phases. This is when performance gradually becomes more fluent and more reliable as a result of practice. Importantly, a construction that is highly entrenched can be processed quickly and efficiently, with minimal effort; as a result, the speakers can concentrate their attention on the content of the message and the meaning they wish to convey. Automatisation is a slow process and requires vast amounts of practice. As an example, consider the acquisition of regular past tense marking in English. The relevant rule is relatively simple and can be acquired after just a few exposures to relevant exemplars. However, it takes much longer for learners to reach a stage where they supply the correct markers consistently whenever required. Many adult L2 learners never reach this stage: Instead, they show residual optionality (i.e. supply the correct form most of the time but not all of the time). This is sometimes attributed to less effective procedural memory, although there is little evidence that older children or adults are less effective procedural 289 Speed of automatisation predicts performance on “decorative” grammar <?page no="290"?> 1 Most studies which report an age-related decline in procedural/ implicit learning com‐ pared young and elderly adults (see Howard & Howard, 2013), and hence their results are not relevant to determining whether or not children are better at procedural learning than adults. Most studies that directly compared young children and young adults either found no difference or better learning in older children and adults compared to younger children (Lukács & Kemény, 2015; Thomas, Hunt, Vizueta, Sommer, Durston, Yang & Worden, 2004). To our knowledge, the only study that found an advantage in younger learners was Janacsek, Fiser & Nemeth (2012). However, the advantage only showed up when comparing raw reaction times, but not on accuracy measures or on Z-transformed RT measures. learners than young children 1 . It is important to note that children acquiring their first language also go through a (sometimes lengthy) period of inconsistent performance. For example, English-speaking children start using the regular past tense inflexion at about age 2; 4, and they begin to use it productively (as evidenced e. g. by overgeneralisation errors) a few months later. They reach Brown's (1973) criterion for mastery (i.e. supply the correct form 90 % of the time in obligatory contexts) by about age 4 and supply the target form at adult-like levels (i.e. close to 100 % correct) after 5; 6 (Rice, Wexler & Hershberger, 1998). This means that there is a period of inconsistent use spanning at least three years. Considering the frequency of the regular past tense marker, this is a very long time. Note, too, that if we assume that language acquisition begins in earnest at age 1 and that children are exposed to the language for eight hours every day (a rather conservative estimate! ), by the time they are 5; 6 they will have had 13140 hours of exposure (4.5 years * 365 days * 8 hours). Many late L2 learners are not exposed to this quantity of input; so even if they learn at the same rate as children, they may not have had enough input to fully automatise even very frequent inflexions. Whatever the reason, it is well-known that many L2 learners do not fully automatise the grammatical routines of the second language: L2 processing relies to a greater extent on controlled processes, hence processing is slower, more effortful and more prone to error (Ellis, Loewen, Elder, Erlam, Philp & Reinders, 2009). Another important consideration is that controlled processes place heavy demands on working memory. Working memory capacity is strictly limited; that is to say, we can only consciously attend to a small number of elements at a time (Baars, 1997; Just & Carpenter, 1992). The implications for linguistic processing are clear: When processing is not fully automatised, speakers will attend to those aspects of an utterance which are most relevant to getting the message across, such as content words and aspects of grammar which make a clear contribution to meaning (‘functional’ grammar) at the expense of what Dąbrowska, Becker & Miorelli (2020) call ‘decorative’ grammar, 290 Ewa Dąbrowska / Ashley Blake <?page no="291"?> 2 It should be emphasised that the ‘functional’/ ‘decorative’ distinction is a continuum rather than dichotomy: Grammatical morphemes such as past tense, agreement markers and articles do add to the meaning, at least in some contexts. See Dąbrowska, Becker & Miorelli (2020) for further discussion. or those aspects of grammar whose contribution to meaning is relatively abstract and often redundant, such as tense and agreement marking, articles and subcategorization restrictions (Goldschneider & DeKeyser, 2001; VanPatten, 1996, 2004). 2 1.3 Tower of Hanoi as a model of skill acquisition In this study, we used a non-linguistic task (which we call ‘multiple trial Tower of Hanoi’, or MToH) to assess participants’ skill learning abilities, and in particular, the speed of automatisation. The Tower of Hanoi (ToH) (see Figure 2) is a puzzle consisting of three vertical rods mounted on a base and a number of discs that differ in size (in this study, we used a four-disc version). In the initial configuration, all four discs are stacked on the leftmost rod in ascending order of size, with the smallest one at the top. The participant’s task is to move the discs to the rightmost rod following certain rules, namely: (i) only one disc can be moved at a time, and (ii) one cannot place a larger disc on top of a smaller one. The solution to the ToH puzzle has a recursive structure in the sense that goals have subgoals embedded in them. For example, in order to reach the target configuration, one must first move the largest disc to the rightmost rod; in order to be able to do that, one must first remove the three discs that are on top of it, and so on. Thus, participants must hold goals and subgoals in memory while planning their moves. Figure 2: The Tower of Hanoi puzzle. 291 Speed of automatisation predicts performance on “decorative” grammar <?page no="292"?> The ToH puzzle is often used in psychological assessment to measure executive functioning (Lezak, Howieson, Loring & Fischer, 2004) and problem-solving abilities (Donnarumma, Maisto & Pezzulo, 2016; Marengo, 2015). In such studies, participants solve the puzzle just once, or at most two or three times. The ToH puzzle can also be used to measure cognitive procedural learning (Beaunieux et al., 2006; Hubert et al., 2007). In this case, participants solve the puzzle many times and the measure of interest is how quickly performance improves, as assessed by the number of moves per trial and/ or time per trial. We adopt a similar experimental paradigm in our study. Beaunieux et al. (2006) and Hubert et al. (2007) examined correlations between time per trial and performance on various cognitive measures in order to estab‐ lish when a particular phase begins and ends. Given that in the cognitive phase the time taken to complete the puzzle depends primarily on IQ, while in the autonomous phase psychomotor abilities are the best predictor, they stipulated that the cognitive stage comprises those trials in which IQ is a significantly better predictor of completion time than psychomotor abilities, while in the autonomous phase the reverse would be true. The intervening trials, where there is no significant difference between the effect of IQ and psychomotor abilities, correspond to the associative phase. Using these anchoring points, they established behavioural criteria for delimiting the phases as follows: • The cognitive phase ends when a participant finds the optimal solution to the puzzle (i.e. one which involves 15 moves). • The associative phase begins immediately after the cognitive phase and ends when a participant applies the optimum solution in at least four out of five consecutive trials. • The autonomous phase begins immediately after the end of the associative phase. 1.4 The current study The current study has two main goals. Our first goal is to explore the predictive validity of the MToH task. To this end, we examine the extent to which measures of proceduralisation and automatisation derived from the task predict performance on measures of foreign language achievement. As a measure of proceduralisation, we use the length of the associative phase, as defined by Hubert et al. (2007). Our measure of the degree of automatisation is based on the fact that automatised skills require less conscious attention and hence are less susceptible to interference from other tasks (cf. Anderson, 1982; DeKeyser, 2015). To assess the degree to which a skill is automatised, we compare performance on the ToH when the participant’s attention is partly 292 Ewa Dąbrowska / Ashley Blake <?page no="293"?> occupied by a concurrent task with baseline performance (i.e. without the concurrent task). We measure grammatical achievement in L2 English using two different tasks. The first task assessed participants’ grammatical comprehension using picture selection. In this task, participants were presented with a sentence and two pictures and were asked to choose the picture that went with the sentence. To be able to do this, participants must process the sentences for meaning. Therefore, the picture selection task is a measure of ‘functional’ grammar. The second task was a grammaticality judgment task in which participants were presented with sentences such as those in (1) and (2) and had to decide whether or not they were well-formed. (1) (a) *Last night the old lady die in her sleep. (b) Last night the old lady died in her sleep. (2) (a) *The man lets his son to watch TV. (b) The man lets his son watch TV. In order for participants to be able to reject an ungrammatical sentence without any additional information about the intended meaning, the sentence itself must contain some clue about what the correct form should have been. For example, sentence (1a) is ungrammatical because the adverbial last night indicates that the action occurred in the past, but the verb is in the present tense form; and (2a) is ungrammatical because the verb let takes bare infinitives rather than to infinitives. In other words, for the task to work, the information provided by the element which makes the sentence ungrammatical must be redundant and/ or incompatible with some other element of the sentence. Therefore, grammaticality judgment tasks assess participants’ knowledge of ‘decorative’ grammar. The second goal of our study is to test a specific hypothesis about the role of two distinct aspects of language aptitude, namely inductive learning ability, or the ability to infer the rules governing language structure, and the speed of proceduralisation/ automatisation in L2 learners’ acquisition of ‘decorative’ and ‘functional’ grammar. We hypothesise that both aspects of aptitude will be positively correlated with performance on both tasks; however, for the reasons explained in section 1.2, we expect inductive language learning ability to correlate more strongly with functional grammar and speed of proceduralisation and automatisation with decorative grammar. 293 Speed of automatisation predicts performance on “decorative” grammar <?page no="294"?> 2 Method 2.1 Participants We recruited 36 students (24 males and 12 females) from a Grammar School in the Erlangen area of Germany who were learning English as a foreign language. The students were in the tenth grade and aged from 14 to 18 years (mean 15.5). All were fluent in German. 32 of the participants reported speaking German as their first language; of these, six indicated an additional first language to German (Vietnamese, Thai, Russian, Polish, Romanian and Turkish). The remaining four participants reported a first language other than German (two Greek, one Turkish, one Bosnian/ Croatian). All participants had had at least five years of regular English instruction (estimated 722 classroom hours); six had technically had an additional year since they had repeated a school year. 2.2 Materials 2.2.1 Background questionnaire The participants completed a background questionnaire that elicited informa‐ tion about their age, gender, native language(s) and any foreign languages they were learning in addition to English. They also chose an alias that was used to match their records for sessions 1 and 2. 2.2.2 Grammaticality Judgement Task (GJT) The grammaticality judgement task (GJT) used the same stimuli as Dąbrowska et al. (2020), who in turn used a subset of the stimuli created by DeKeyser (2000) and Johnson & Newport (1989). The stimuli assessed basic aspects of grammar, including past tense, plural and third-person singular inflexions, determiners, particle movement, subcategorisation, word order, the progressive, and pronominalisation. 80 of the 200 items used by DeKeyser (2000) were selected, making sure that each of the eleven categories from the original study was represented in proportions similar to those in the original test. The sentences were recorded by a female native speaker of English. Half of the stimuli were grammatical, and the other half were ungrammatical (see Table 1 for examples). 294 Ewa Dąbrowska / Ashley Blake <?page no="295"?> Category Grammatical variant Ungrammatical variant Past tense Last night the old lady died in her sleep. Last night the old lady die in her sleep. Plural Three boys played on the swings in the park. Three boy played on the swings in the park. Third person singular John’s dog always waits for him at the corner. John’s dog always wait for him at the corner. Present pro‐ gressive Tom is working in his office right now. Tom working in his office right now. Determiners Tom is reading a book in the bathtub. Tom is reading book in the bathtub. Pronominalisa‐ tion The girl cut herself on a piece of glass. The girl cut himself on a piece of glass. Particle move‐ ment The man climbed up the ladder carefully. The man climbed the ladder up carefully. Subcategorisa‐ tion The man lets his son watch TV. The man lets his son to watch TV. Yes-No ques‐ tions Will Harry be blamed for the accident? Will be Harry blamed for the accident? WH questions What is Martha bringing to the party? What Martha is bringing to the party? Word order The woman asked the policeman a question. The woman the policeman asked a question. Table 1: Examples of items used in the grammaticality judgment task. Following earlier studies, participants heard both the grammatical and ungram‐ matical version of each sentence. The items were divided into two blocks so that if the grammatical version of an item occurred in block 1, the ungrammatical version occurred in block 2 and vice versa. Within each block, the sentences were presented in the same random order to all participants. Participants were instructed to type “g” if they thought the sentence was grammatical or “u” if they thought it was ungrammatical. They were asked to keep their index fingers on the relevant keys throughout the task. Each sentence was played twice, with a two second pause between repetitions. The next trial began as soon as the participant responded. If the participant did not respond within ten seconds from the offset of the first presentation, the trial timed out and the next trial began. There was a short break after the first 40 items. 295 Speed of automatisation predicts performance on “decorative” grammar <?page no="296"?> 2.2.3 The Picture Selection Task (PST) The picture selection task is an online version of Dąbrowska’s (2018) Pictures and Sentences Test. The PST tests the comprehension of ten different structures (actives, passives, subject clefts, object clefts, subject relatives, object relatives, simple locatives, locatives with quantifiers, possessive sentences with quanti‐ fiers, post-modifying prepositional phrases). Each item in the test consists of a sentence (e.g. The woman was the one that the man caught) and two pictures, e. g. a woman catching a man and a man catching a woman (see Figure 3). There are eight items per structure (hence 80 items in total), presented in a semi-random order, so that no two sentences of the same structure were presented immediately next to each other (see Table 2 for examples of all sentence types). Construction Example Pictures Active The man fed the girl. Target: man feeding girl Distractor: girl feeding man Passive The girl was fed by the man. Subject cleft It was the man who fed the girl. Object cleft It was the girl that the man fed. Subject relative The man was the one who fed the girl. Object relative The girl was the one that the man fed. Locative w/ quantifier Every pencil is in a box. Target: three boxes, each containing a pencil plus an extra box Distractor: three boxes, each con‐ taining a pencil plus an extra pencil Possessive w/ quantifier Every box has a pencil in it. Target: three boxes, each containing a pencil plus an extra pencil Distractor: three boxes, each con‐ taining a pencil plus an extra box Simple locative The spoon is in the cup. Target: spoon in cup Distractor: spoon next to cup Postmodifying PP The spoon in the cup is red. Target: red spoon in green cup Distractor: yellow spoon in red cup Table 2: Examples of items used in the picture selection task. 296 Ewa Dąbrowska / Ashley Blake <?page no="297"?> The presentation order was the same for all participants. The sentences were spoken by a female native speaker of English. Participants responded using arrow keys (the left arrow to choose the picture on the left and the right arrow to choose the picture on the right). They were asked to keep their index fingers on the keys throughout the task. The response deadline was ten seconds per item. If the participant did not respond within this time, the trial timed out and the next trial began. Figure 3: An example of an item from the picture selection task. 2.2.4 The Language Analysis Task The Language Analysis task was adapted from the Language Analysis subtest of the Pimsleur Language Aptitude Battery (Pimsleur, Reed & Stansfield, 2004). Participants were presented with words and simple sentences in an unfamiliar language and their English translations. Their task was to translate 15 novel sentences into the language. Participants selected the correct answer from an array of four by clicking an on-screen button. All stimuli were presented in writing and the vocabulary and model sentences were visible on the screen at all times. There was no time limit. The maximum possible score in this task was 15. 2.2.5 Multiple-trial Tower of Hanoi (MToH) The four-disc version of the puzzle was used. Participants were asked to solve the puzzle 40 times, with a secondary task introduced in the last five trials. The secondary task involved tapping in response to a recording consisting of a random sequence of the words one and two, with four second pauses between the words. Participants were instructed to tap with their non-dominant hand in response to the numbers (one tap in response to word one and two taps in response to the word two) while solving the puzzle with their dominant hand. 297 Speed of automatisation predicts performance on “decorative” grammar <?page no="298"?> 2.3 Procedure Participants’ parents were contacted by letter and asked to sign a consent form. The experiment consisted of two sessions. Session 1 was administered via an online platform (Pavlovia). Participants accessed the experiment by following a link given in the letter to their parents. The link took them to a web page that explained what the experiment would involve. They were then asked to click a button to indicate their consent to take part. They were advised that they could withdraw from the study at any time simply by closing the browser. The tasks were administered in the following order: • Background questionnaire • Spoken Grammaticality Judgement Task (Block 1) • Language Analysis Task • Picture Selection Task • Spoken Grammaticality Judgement Task (Block 2) Session 2 comprised the MToH task. Participants were tested individually in a quiet room at their school during the school day. The experimenter explained the rules, showed examples of legal and illegal moves and offered further clarifications if necessary. Following Beaunieux et al. (2006), the experimenter also demonstrated the first move (placing the smallest disc on the middle rod); this was done in order to prevent a possibly random search through the possibility space. Participants completed the puzzle 40 times, with the secondary task introduced after the 35 th trial. The experimenter counted the number of moves per trial using a hand-held counter and recorded the length of each trial using a stopwatch on a smartphone. In addition, the participants’ hand movements were video recorded for later checking. 3 Results In the GJT, data for two participants were excluded because they had either more than 20 consecutive timeouts or more than 20 consecutive RTs shorter than 1000 ms, suggesting they had not engaged with the task. In the Language Analysis Task, two participants were excluded as they completed the task in 22.5 seconds and 90 seconds respectively, which also suggests a lack of engagement. (The task normally takes 10-12 minutes to complete.) Two participants were excluded from the MToH as they indicated that they were already familiar with the puzzle. 298 Ewa Dąbrowska / Ashley Blake <?page no="299"?> 3.1 Performance on the MToH task Figures 4 and 5 show how performance on the MToH improved as a function of practice. Figure 4 shows the decrease in the average number of moves per trial, and Figure 5 the decrease in the average time per trial. Both curves show a typical power function shape. At the beginning of the MToH task, participants needed on average 21 moves per trial and took about 70 seconds to complete the puzzle. Both the mean number of moves per trial and the mean time per trial decreased sharply over the first six trials and then started to level off. After 15 trials, there is very little improvement: that is to say, the participants’ performance begins to stabilise at about this time. Most participants stabilised at the optimal (i.e. 15-move) solution; three participants stabilised at a 17-move solution and at a 19-move solution. Figure 4: MToH: Average number of moves per trial. Figure 5: MToH: Average time per trial. In the last five ‘normal’ trials (trials 31-35) participants averaged about 15.5 moves and 17.5 seconds per trial - i. e. they needed just over one second per move. There was a slight increase in both the number of moves and completion time on trial 36 (when the secondary task was introduced), but most participants 299 Speed of automatisation predicts performance on “decorative” grammar <?page no="300"?> returned to previous levels of performance very quickly, indicating a relatively high degree of automatisation. 3.2 Descriptive statistics and correlational analysis For the Picture Selection Task and the Language Analysis Task, participants were given one point for each correct answer (therefore maximum possible scores are 80 and 15 respectively). For the Grammaticality Judgement Task, performance was assessed using the sensitivity index (d’). The sensitivity index is a measure derived from signal detection theory. Unlike a simple proportion of correct responses, d’ takes a participant’s response bias into account, thus providing a more sensitive measure of how well a participant is able to discriminate between grammatical and ungrammatical items (Huang & Ferreira, 2020). The remaining four measures are derived from performance on the MToH task. LAssoc1 and LAssoc2 are measures of the length of the associative phase. According to the criteria introduced by Hubert et al. (2007), the cognitive stage ends when a participant achieves the optimal solution for the first time, and the associative phase ends when he/ she achieves the optimal solution in four out of five consecutive trials. LAssoc1 is based on this criterion, with one difference. As indicated earlier, some of our participants never discovered the optimal solution and stabilised on a solution with 17, 18 or 19 moves (see below). For this reason, we define the associative stage with respect to the stable solution achieved by a particular participant (i.e., the solution which they replicate on most subsequent trials). The associative stage, therefore, begins on the trial immediately following the trial in which a participant achieved the stable solution for the first time, and ends on the trial at which the participants achieved a stable solution on at least four out of five consecutive trials. Thus, LAssoc1 corresponds to the number of trials between these two points in time, including the last trial of the associative stage. LAssoc2 is based on a stricter criterion for the end of the associative phase, whereby a participant is required to achieve the stable solution in five out of five consecutive trials and corresponds to the number of trials between the trial on which the participant achieved the stable solution for the first time and the trial on which the participant achieved a stable solution on five consecutive trials (including the fifth consecutive successful trial). Two additional measures (difference in the number of moves: DiffMoves and difference in time: Diff Time) were motivated by the fact that when a procedure is automatised, its execution does not require much conscious attention and therefore performance should not be disrupted if a secondary (tapping) task 300 Ewa Dąbrowska / Ashley Blake <?page no="301"?> 3 The statistics for reaction time reported in Tables 1 and 2 are based on raw reaction times. We also computed RT values using the standard filtering procedure (i.e. consid‐ ering only RTs for correct items and excluding values that were more than two standard deviations above or below the mean for each participant). However, since the two measures were virtually perfectly correlated (r = .99), we chose to report the unfiltered values. is introduced (cf. Anderson, 1982; DeKeyser, 2015). These two measures were computed by subtracting the average number of moves (DiffMoves) or the average time per move (Diff Time) in the trials with the secondary task (i.e. trials 36-40) from the average number of moves or the average time per move in the last five ‘normal’ trials (i.e. trials 31-35). A negative figure for these two measures indicates that performance was affected by the secondary task, and the lower the figure, the more it was affected. The data were analysed in the R software (R Development Core Team, 2020); d’ values were computed using the psycho package (Makowski, 2018). The descriptive statistics for performance on the four experimental tasks are given in Table 1. 3 As can be seen from the table, there are considerable individual differences in performance on all tasks. Measure Mean SD Range PST Accuracy 69.44 6.18 58-79 GJT Accuracy 53.97 9.99 26-71 GJT d’ 0.90 0.72 -0.92-2.33 GJT RT 4812.28 1033.11 3046-6707 PST RT 2995.32 578.58 2323-4815 Language Analysis 10.91 3.49 4-15 LAssoc1 6.59 3.73 4-16 LAssoc2 12.21 10.37 4-45 DiffMoves -0.27 1.20 -5.60-2.20 Diff Time -1.15 2.74 -13.27-3.51 Table 3: Descriptive statistics. Table 4 presents the pairwise Pearson product-moment correlations between all variables. As expected, there was a positive relationship between performance on the two language tasks for both accuracy (r = .29) and reaction time (r = .60), 301 Speed of automatisation predicts performance on “decorative” grammar <?page no="302"?> although only the latter was statistically significant. This is not surprising as both tasks measure different aspects of grammatical proficiency. Correlations between accuracy and reaction times on the two tasks were non-significant (and negative), which indicates that there was no speed-accuracy trade-off. As predicted, we found a significant correlation between the Language Analysis task and accuracy on the picture selection task (r = .45). The correlation between Language Analysis and performance on the GJT was weaker but also significant (r = .35). GJT d’ GJT RT PST Acc PST RT LgAnal LAssoc1 LAssoc2 Diff‐ Moves Diff‐ Time GJT d’ 1 -0.15 0.29† 0.07 0.36* 0.01 0.01 0.47** 0.49** GJT RT -0.15 -0.15 -0.43** 0.60*** -0.11 -0.09 -0.14 0.23 0.19 PST Acc 0.29† -0.43** 1 -0.27 0.45** -0.14 -0.19 0.09 -0.08 PST RT 0.07 0.6*** 0.27 1 0.09 0.1 -0.1 0.18 0.17 LgAnal 0.36* -0.11 0.45** 0.09 1 -0.49** -0.24 -0.08 -0.07 LAssoc1 0.01 -0.09 -0.14 0.1 -0.49** 1 0.27 0.07 0.01 LAssoc2 0.01 -0.14 -0.19 -0.1 -0.24 0.27 1 0.3† 0.18 Diff‐ Moves 0.47** 0.23 0.09 0.18 -0.08 0.07 0.3† 1 0.86*** Diff‐ Time 0.49** 0.19 -0.08 0.17 -0.07 0.01 0.18 0.86*** 1 *** Correlation is significant at .001 level ** Correlation is significant at .01 level ** Correlation is significant at .05 level † Correlation approaches significance (p < 0.10). Table 4: Pairwise correlations between all variables. The two measures of the length of the associative phase based on Hubert et al. (2007), LAssoc1 and LAssoc2, do not correlate with performance on any of the language tasks. They are also not correlated with the interference-based measures and only weakly correlated with each other. DiffMoves and Diff Time, in contrast, are strongly correlated with each other (r = .86). There is a significant relationship between DiffMoves and performance on the GJT (r = .47) and Diff-Time and performance on the GJT (r = .49). There were no significant 302 Ewa Dąbrowska / Ashley Blake <?page no="303"?> correlations between any of the interference measures and PST or between the interference measures and reaction times on either task. 4 Discussion The main objectives of our study were to investigate the predictive validity of different measures of proceduralisation and automatisation and to examine the effects of various aspects of language aptitude on the acquisition of ‘functional’ grammar (aspects of grammar that provide a clear contribution to meaning) and ‘decorative’ grammar (aspects of grammar where the contribution is abstract and often redundant). In the following section, we discuss the validity of our measures. The next two sections are devoted to a discussion of our results in the context of the predictions outlined in the introductory section. 4.1 Measures of proceduralisation and automatisation The interference-based measures introduced here (DiffMoves and Diff Time) show moderately strong correlations with performance on the GJT, and they are also strongly correlated with each other, suggesting that they are both reliable and valid measures of automatisation. In contrast, the two measures based on Hubert et al.’s (2007) criteria (LAssoc1 and LAssoc2) did not correlate with performance measures on any of the language tasks. They are also not correlated with the interference-based measures (DiffMoves and Diff Time). It could be argued that this is because they measure something different, namely the speed of proceduralisation (as opposed to automatisation); however, the fact that they are only weakly correlated with each other raises doubts about their reliability. It is worth noting that the effects of automatisation may be even stronger than our results suggest. A closer inspection of the participants’ performance on the MToH by trial shows that performance stabilises around trial 15. This means that participants had about 20 trials left to automatise the procedure before the secondary task was introduced. This might have been too much so that virtually all participants reached high levels of automaticity. The average DiffMoves score was -.27, meaning that participants made, on average, only .27 more moves per trial when the secondary task was introduced. The average Diff Time was -1.15 seconds, which means that they needed just over one extra second per trial. Over the five trials with the secondary task, this equates to 1.35 moves and approximately six extra seconds. Furthermore, only 15 participants out of 34 had negative difference scores, and for twelve of them, the effect was minimal (less than one move per trial). In other words, the disruption in performance produced by the secondary task was very small and affected less than half of the 303 Speed of automatisation predicts performance on “decorative” grammar <?page no="304"?> participants. This suggests that we may have introduced the secondary task too late in the experiment: in other words, if the secondary task were introduced earlier (e.g. after 20 or 25 trials), we are likely to have observed more interference and larger individual differences - and possibly even stronger effects of speed of automatisation. 4.2 Inductive language learning ability and speed of automatisation as predictors of performance on ‘functional’ and ‘decorative’ grammar The Language Analysis subtest of the PLAB is an established test of explicit language aptitude which is particularly relevant to L2 grammar, particularly in earlier stages of acquisition. Since fluent language processing relies on automatised knowledge, we hypothesised that the speed of automatisation (as measured by the MToH) should also make an important contribution to explaining grammatical attainment. Given that explicit attention is a limited commodity, language learners tend to attend primarily to those aspects of form which are most relevant to meaning. Therefore, we predicted that Language Analysis scores would correlate more highly with performance on ‘functional’ grammar (assessed using a picture selection task), while MToH should be more strongly associated with performance on ‘decorative’ grammar (assessed with a traditional grammaticality judgement task). These predictions were mostly met. Language Analysis predicted performance on both the PST and the GJT, with the former relationship being stronger than the latter (.45 vs. .35). There was also a robust relationship between both interference measures (DiffMoves and Diff Time) and performance on the GJT (.47 and .49 respectively). In contrast, our data show no significant relationship between the interference measures and performance on the PST. These results suggest that explicit language aptitude (and in particular, inductive learning ability as measured by the Language Analysis test) play a role in the acquisition of both ‘decorative’ and ‘functional’ grammar, while speed of automatisation appears to be relevant for ‘decorative’ grammar only. 4.3 Reaction time measures Our primary measure of proficiency was accuracy. However, one would also expect to find a relationship between the reaction time measures and aptitude, and particularly speed of automatisation. This was not the case, which is somewhat surprising, since RT measures are typically more sensitive than accuracy measures. Furthermore, as noted earlier, our results show no evidence of a speed-accuracy trade-off that might have masked the effects of one or both of these measures. It is possible that all our participants concentrated 304 Ewa Dąbrowska / Ashley Blake <?page no="305"?> more on accuracy than on speed. Whatever the reason, our findings suggest that reaction-time based measures are not always more sensitive than accuracy measures. 5 Conclusion Our findings suggest that interference from the secondary task on the MToH is a promising measure of the speed of automatisation. Measures based on the length of the associative phase, on the other hand, had no predictive validity. This could indicate that the benefits of proceduralisation only become evident once the performance is highly automatised. On the other hand, the fact that the two measures of the speed of proceduralisation do not correlate with each other suggests that they are simply not very reliable. As discussed earlier, our interference measure may have been introduced too late, this should be taken into account in future research. Our results also support the hypothesis that speed of automatisation is relevant for mastery of ‘decorative’ grammar in L2 learning in instructional settings. Further research will be needed to determine if it is also relevant for more functional aspects of grammar and whether it also holds for other age groups (e.g. younger learners) and in other settings (e.g. naturalistic exposure). 6 Acknowledgements We would like to thank Anna Schreiber for collecting and coding the data and Laura Becker for programming the experimental tasks. This project was funded by an Alexander von Humboldt Professorship (ID-1195918) awarded to the first author. References Ackerman, P.L. 1992. Predicting individual differences in complex skill acquisition: Dynamics of ability determinants. Journal of Applied Psychology, 77 (5), 598-614. Anderson, J.R. 1982. Acquisition of cognitive skill. Psychological Review, 89 (4), 369-406. Anderson, J.R. 1993. Problem solving and learning. American Psychologist, 48 (1), 35-44. Anderson, J.R., Bothell, D., Byrne, M.D., Douglass, S., Lebiere, C., & Qin, Y. 2004. An integrated theory of the mind. Psychological Review, 111 (4), 1036-1060. Anderson, J.R., Greeno, J.G., Reder, L.M., & Simon, H.A. 2000. Perspectives on learning, thinking, and activity. Educational Researcher, 29 (4), 11-13. 305 Speed of automatisation predicts performance on “decorative” grammar <?page no="306"?> Baars, B.J. 1997. In the Theater of Consciousness: The Workspace of the Mind. Oxford: Oxford University Press. Beaunieux, H., Hubert, V., Witkowski, T., Pitel, A-L., Rossi, S., Danion, J-M., Desgranges, B., & Eustache, E. 2006. Which processes are involved in cognitive procedural learning? Memory, 14 (5), 521-539. Behrens, H. 2009. Usage-based and emergentist approaches to language acquisition. Linguistics, 47 (2), 383-411. Brown, R. 1973. A First Language. The Early Stages. Cambridge, MA: Harvard University Press. Bybee, J. 2010. Language, Usage and Cognition. Cambridge: Cambridge University Press. Chater, N. & Christiansen, M.H. 2018. Language acquisition as skill learning. Current Opinion in Behavioral Sciences, 21, 205-208. Chater, N., McCauley, S. M. & Christiansen, M.H. 2016. Language as skill: Intertwining comprehension and production. Journal of Memory and Language, 89, 244-254. Chomsky, N. 1965. Aspects of the Theory of Syntax. Cambridge, MA: MIT Press. Chomsky, N. 1981. Lectures on Government and Binding: The Pisa Lectures. Berlin: Walter de Gruyter. Dąbrowska, E. 2004. Language, Mind and Brain. Some Psychological and Neurological Constraints on Theories of Grammar. Edinburgh: Edinburgh University Press. Dąbrowska, E. 2018. Experience, aptitude and individual differences in native language ultimate attainment. Cognition, 178, 222-235. Dąbrowska, E., Becker, L. & Miorelli, L. 2020. Is adult second language acquisition defective? Frontiers in Psychology, 11, 1839. DeKeyser, R. 2000. The robustness of critical period effects in second language acquisi‐ tion. Studies in Second Language Acquisition, 22 (4), 499-533. DeKeyser, R. 2015. Skill Acquisition Theory. In: VanPatten, B. & Williams, J. (Eds.), Theories in Second Language Acquisition: An Introduction. 2 nd ed. New York: Routledge. 94-112. Donnarumma, F., Maisto, D. & Pezzulo, G. 2016. Problem solving as probabilistic inference with subgoaling: Explaining human successes and pitfalls in the Tower of Hanoi. PLoS Computational Biology, 12 (4), doi: 10.1371/ journal.pcbi.1004864. Ellis, N., Römer, U. & O’Donnell, M.B. 2016. Usage-Based Approaches to Language Acquis‐ ition and Processing: Cognitive and Corpus Investigations of Construction Grammar. Malden, MA: Wiley-Blackwell. Ellis, R., Loewen, S., Elder, C., Erlam, R., Philp, J. & Reinders, H. 2009. Implicit and Explicit Knowledge in Second Language Learning, Testing and Teaching. Bristol: Multilingual Matters. Fitts, P.M. & Posner, M.I. 1967. Human Performance. London: Prentice Hall. 306 Ewa Dąbrowska / Ashley Blake <?page no="307"?> Goldberg, A.E. 2006. Constructions at Work. The Nature of Generalization in Language. Oxford: Oxford University Press. Goldschneider, J.M. & DeKeyser, R. 2001. Explaining the “natural order of L2 morpheme acquisition” in English: A meta‐analysis of multiple determinants. Language Learning, 51 (1), 1-50. Howard Jr, J.H. & Howard, D.V. 2013. Aging mind and brain: is implicit learning spared in healthy aging? Frontiers in Psychology, 4 (817), doi: 10.3389/ fpsyg.2013.00817. Huang, Y. & Ferreira, F. 2020. The application of signal detection theory to acceptability judgments. Frontiers in Psychology, 11 (73), doi: 10.3389/ fpsyg.2020.00073. Hubert, V., Beaunieux, H., Chetelat, G., Platel, H., Landeau, B., Danion, J.M., Viader, F., Desgranges, B. & Eustache, F. 2007. The dynamic network subserving the three phases of cognitive procedural learning. Human Brain Mapping, 28 (12), 1415-1429. Janacsek, K., Fiser, J. & Nemeth, D. 2012. The best time to acquire new skills: Age‐related differences in implicit sequence learning across the human lifespan. Developmental Science, 15 (4), 496-505. Johnson, J.S. & Newport, E.L. 1989. Critical period effects in second language learning: The influence of maturational state on the acquisition of English as a second language. Cognitive Psychology, 21 (1), 60-99. Johnson, K. 1996. Language Teaching and Skill Learning. Oxford: Blackwell. Just, M. A. & Carpenter, P. A. 1992. A capacity theory of comprehension: Individual differences in working memory. Psychogical Review, 99 (1), 122-149. Kamhi, A.G. 2019. Speech-language development as proceduralization and skill learning: Implications for assessment and intervention. Journal of Communication Disorders, 82 (8), doi: 10.1016/ j.jcomdis.2019.105918. Langacker, R.W. 1987. Foundations of Cognitive Grammar. Volume 1: Theoretical Prerequi‐ sites. Stanford, CA: Stanford University Press. Lezak, M.D., Howieson, D.B., Loring, D.W. & Fischer, J.S. 2004. Neuropsychological Assessment. Oxford: Oxford University Press. Lukács, Á. & Kemény, F. 2015. Development of different forms of skill learning throughout the lifespan. Cognitive Science, 39 (2), 383-404. Makowski, D. 2018. The psycho package: An efficient and publishing-oriented workflow for psychological science. Journal of Open Source Software, 3 (22), 470. Marengo, L. 2015. Representation, search, and the evolution of routines in problem solving. Industrial and Corporate Change, 24 (5), 951-980. Newell, A. & Rosenbloom, P.S. 1981. Mechanisms of skill acquisition and the law of practice. In: Anderson, J.R. (Ed.), Cognitive Skills and their Acquisition. Hillsdale, NJ: Lawrence Erlbaum. 1-55. Pimsleur, P., Reed, D.J. & Stansfield, C.W. 2004. Pimsleur Language Aptitude Battery: Manual 2004 Edition. Bethesda, MD: Second Language Testing, Inc. 307 Speed of automatisation predicts performance on “decorative” grammar <?page no="308"?> Rice, M., Wexler, K. & Hershberger, S. 1998. Tense over time: The longitudinal course of tense acquisition with specific language impairments. Journal of Speech, Language, and Hearing Research, 41 (6), 1412-1431. Taatgen, N.A., Huss, D., Dickison, D. & Anderson, J.R. 2008. The acquisition of robust and flexible cognitive skills. Journal of Experimental Psychology: General, 137 (3), 548-565. Thomas, K.M., Hunt, R.H., Vizueta, N., Sommer, T., Durston, S., Yang, Y. & Worden, M.S. 2004. Evidence of developmental differences in implicit sequence learning: An fMRI study of children and adults. Journal of Cognitive Neuroscience, 16 (8), 1339-1351. Tomasello, M. 2003. Constructing a Language: A Usage-Based Theory of Child Language Acquisition. Cambridge, MA: Harvard University Press. VanPatten, B. 1996. Input Processing and Grammar Instruction in Second Language Acquisition. Norwood, NJ: Ablex Publishing Corporation. VanPatten, B. 2004. Input processing in second language acquisition. In: VanPatten, B. (Ed.), Processing Instruction: Theory, Research, and Commentary. Mahwah, NJ: Lawrence Erlbaum. 5-31. 308 Ewa Dąbrowska / Ashley Blake <?page no="309"?> Collocational proficiency: The effects of ‘target language input’ and ‘age’ Cordula Glass Abstract Monolingual adult native speakers of a language are often considered a unanimous group of language experts who share the same knowledge about the structures and idiosyncrasies of a language. Non-native speakers, on the other hand, seem to struggle to gain native-like competence. One particular problem area are phraseological sequences, such as collocations. Yet studies (Chipere, 2003; Dąbrowska, 2012; Pakulak & Neville, 2010) suggest that, depending on their age or educational background, native speakers also show linguistically diverse behaviour. This article examines how native (L1) and non-native (L2) speakers from different age groups (15-year-old students in Year 11 and adults) perform in a test on colloca‐ tional competence (CollMatch) and whether the environment in which an L2 is learnt influences these results. The findings of this study may have implications for foreign language learning and teaching, because they help to identify the impact of different influencing factors, such as quality of input and stages of language attainment. 1 Introduction A good command of phraseological phenomena is often considered the cherry on top of the proverbial cake of language proficiency. Since even very advanced learners of a language seem to struggle with correct phrasing or the most likely choice of a collocate, combinatorial restrictions and idiosyncrasies of a language seem to be one of the more challenging parts of foreign language learning and teaching. Some of the most frequently discussed phraseological phenomena in foreign language learning are collocations. These reoccurring combinations of lexical units can very generally be defined as “a construction, which consists of a lexical form and a function or semantic meaning. The lexical form of a <?page no="310"?> collocation can be rather fixed - and therefore closely related to a compound - semi-fixed or delexicalised” (Glass, 2019: 52; based on Goldberg, 2006; Siepmann, 2005). Quite often, these collocational constructions are considered quite diffi‐ cult to master for L2 learners. Pawley & Syder, for example, observed that even if a learner of a language produces seemingly grammatically correct sentences, “(t)he trouble is that native speakers just do not say things that way” (Pawley & Syder, 1983: 195). Also, Bahns & Eldaw (1993) found that non-native speakers of English, despite good knowledge of English vocabulary, have difficulties with collocational combinations. Howarth (1998: 186) points out that learners particularly struggle with semantically restricted collocations, such as blow a fuse or make a remark, while de Cock (2004: 243) emphasises that non-native phraseology, in general, is a complex network of “overuse, underuse, misuse of target language NS sequences”. To learners and teachers of English, these observations raise the question of whether a good, native-like command of English can be achieved at all by a non-native speaker, since most non-native speakers of a language seem to have trouble mastering the idiomatic subtleties of a language even at a very advanced level. Thus Pawley & Syder conclude: “The nature of the problem will be less obvious to those who have learnt their language(s) by immersion in the speech community more or less from the start” (Pawley & Syder, 1983: 195). The idea of immersing learners in a language has indeed become a common concept in modern language education. More than 50 years ago, immersion programmes were put to a test in North America, for example by Lambert & Tucker (1972) and their well-known project in St. Lambert in Canada. But also reseachers such as Krashen & Terrell (1983) or Wode (1995) promoted more authentic native-like input as well as a largely target language-based approach within the foreign language classroom. Especially against the background of an increasingly more global and international world, the demand for bilingual education is still growing. Here immersion programmes as an intensive form of bilingual education seem to satisfy this demand, since they combine authentic input with a foreign language’s genuine function as a medium for communica‐ tion (e.g. Wode, 1995). Furthermore, findings in cognitive linguistics support the concept of a language training which focuses on input representing a realistic amount of authentic language, with all its repetitions, restrictions and idiosyncrasies (e.g. Bybee, 2013; Ellis & Larsen-Freeman, 2009). Within a cognitive emergentist framework, for example, the mind operates as a kind of data structuring unit, which processes any kind of linguistic input and abstracts categories and systematicity from it. Thus, the brain relies on nothing but its own, 310 Cordula Glass <?page no="311"?> general mechanisms, which are not thought to be supported by any kind of lan‐ guage-specific, inborn mental faculty or device. Based on the input, the human brain then (re-)categorises information until a complex system of linguistic constructions emerges (for example Bybee, 2010; Ellis, 2006; MacWhinney, 2001; Tomasello, 2005), with collocations at the intersection between lexically fixed constructions and potential creative language use (Glass, 2019). However, these findings predominantly refer to the process of first language acquisition and the observable difficulties of L2 learners might well be connected to a difference in cognitive processing and learning, like potential L1 interfer‐ ence or overshadowing and blocking from already acquired L1 constructions (Ellis, 2008). Thus, the question remains how non-native speakers can be supported on their way to attaining native-like competence, and whether native-like competence is a desirable goal in the first place because the com‐ munity of native speakers itself is more diverse than some researchers like to admit. For simple syntactic structures, Pakulak & Neville (2010) were able to show that native speakers with a high-level proficiency differ considerably from native speakers with a lower level of proficiency. Dąbrowska (2012) also suggests that the assumption that all adult monolingual native speakers of a language linguistically share the same mental structures could be a serious misconception of language acquisition research. Howarth (1996: 124-132) admits that in his study, native speakers at times deviated from the norm and produced non-standard combinations. Nevertheless, mother-tongue competence is often considered the gold standard for L2 learner assessment. But if even native speakers have different levels of language competence, this would mean that L2 learners would be measured against a standard that even some native speakers will reach late or even never. Therefore, this article addresses the question of how different age groups of native speakers, namely academically trained adults and 15-year-old students, master one of the most challenging phraseological phenomenon, i. e. colloca‐ tions, and how non-native speakers of English with a high target language input fare in comparison. Furthermore, an additional sample of non-native adults who learn English in a highly immersed context in Great Britain was compared to a dataset of native English speakers and also compared to non-native speakers in Germany who had received traditional English-as-a-subject lessons in Germany. However, before the study and its results will be presented, the next sections focus on collocations and why these combinations should be considered essential for foreign language learning and teaching. 311 Collocational proficiency: The effects of ‘target language input’ and ‘age’ <?page no="312"?> 1 In the British National Corpus (BNC) agree ranges among the top five verb collocates for entirely in terms of log-likelihood value (431.41), t-score (10.59) and z-score (37.81). Generally, the log-likelihood measures lexical association. It is therefore higher for lexical items which co-occur together quite often. In comparison to entirely + agree, a more arbitrary combination such as often + agree scores a loglikelihood value of 0.0069, t-score of 0.0824 and z-score of 0.0416. 2 Collocations 2.1 Collocations in linguistics One of the early definitions proposed by the English lexicographer Samuel Johnson defines collocations simply as “a placing or setting in order” ( Johnson, 1755, in Besalke, 2017). While over the years, the depth and scope of collocational studies have changed, the gist of these special combinations within a language is still about putting words together in an orderly way. The scope of an “orderly way” however, has been the major point of discussion throughout different approaches and disciplines. Today most definitions circle around three main factors: fixedness, opacity and the likelihood of co-occurrence. Within the last decades, it has become a common typology for the concept of collocations to distinguish two definitions (Barnbrook, Mason & Krishnamurthy, 2013; Granger & Paquot, 2008; Herbst, 1996; Herbst, 2011): Context-oriented approaches (e.g. Firth, 1968; Halliday, 1966; Sinclair, 1966) regard collocations as a significant combination of words, which emerges through context and use. Strictly speaking, every co-occurrence of a word with another word would be seen as a collocation from such a context-oriented point of view, while some words indeed co-occur more often with certain words than would be expected from a statistical point of view ( Jones & Sinclair, 1974: 19). Here the word order within a collocation is derived from its likelihood to co-occur with other items and the degree of its fixedness. Consequently, the lexical surrounding of a word defines its meaning, and the more often two or more words occur together, the closer they become associated with another. This is, for example, the case with combinations such as entirely, which shows a high likelihood 1 to be followed by a verb like agree. From a cognitive point of view, this combination also ranges, as emergentist approaches would predict, among the most frequent answers from native speakers of English in a gap-filling task (Herbst, 1996: 390). Significance-oriented approaches (e.g. Cowie, 1983; Hausmann, 1984), on the other hand, focus on the semantic opacity of a combination. They only consider two or more words a collocation if their combination carries a meaning which exceeds the meaning of its individual constituents or if the combination of 312 Cordula Glass <?page no="313"?> the collocates is combinatorically restricted, for example white wine, which is usually of a rather yellowish colour or make tea, which cannot be rephrased as *cook tea. Thus Hausmann, for example, regards collocations as “typical, specific and characteristic relations between two words” (Hausmann, 1985: 118). Furthermore, according to Hausmann, these relations can hardly be predicted by anyone but a native speaker, which brings this definition to the centre of attention for foreign language learning and teaching. Significance-oriented approaches are, therefore, often regarded to be more suitable for the language classroom, since the aspect of unpredictability makes them supposedly difficult to learn. Context-oriented approaches, on the other hand, are often used to produce word lists, which are then revised by a teacher or author in order to produce a list with statistically and semantically significant candidates. However, Ellis, Simpson-Vlach & Maynard (2008) were able to show that different statistic parameters of recurrent formulaic language also influence their processability in native and second language speakers of English. 2.2 Collocations in L2 research It has only been within the last decades that phraseological phenomena in general and collocations in particular came to the attention of L2 research. Until the 1990s, “slot-filler model[s]” (Sinclair, 1991: 109), which postulated a fixed grammar with potentially unlimited lexical filling, dominated the English classroom. Up to this point, teaching English often consisted of a structural component (grammar) and a more content related side, the lexical juice to the grammatical bone (vocabulary). Only gradually studies emerged, showing the importance of phraseological competence for learners of English as well as their lack of a native-like use for idiomatic phenomena (e.g. de Cock, 2004; Granger, 1998; Howarth, 1996; Nesselhauff, 2004). From an early stage on, collocations gained a great amount of attention. One reason for this might be that their status inbetween rather fixed idiomatic expressions, like idioms or catchphrases (such as the early bird catches the worm) and rather free, grammatically correct combinations of words like She bought a house are fascinating, yet difficult to discern at the same time. Thus, collocational combinations may cause many difficulties for learners of English at all levels. They are not as easily memorised as more idiomatic phrases, since they do not operate as one multi-lexeme form with one clearly assigned meaning but rather show a certain flexibility, which reaches its limits eventually. Thus, it comes as no surprise that one of Howarth’s (1996) major findings was that L2 speakers of English are less aware of collocational restrictions and more prone to creating more or less acceptable blended combinations like *make 313 Collocational proficiency: The effects of ‘target language input’ and ‘age’ <?page no="314"?> a reaction instead of give a reaction or make a response. Furthermore, Granger (1998) points out that the respective L1 of a learner plays an important role in the mastery of collocations. In a study on collocations consisting of an adjective and an amplifier, like closely linked or bitterly cold, she was able to show that not only do learners not exploit the full range of a collocation but also that her native French participants were not aware of any kind of salience and rather preferred combinations which they knew from their native tongue like punished as a collocate for severely, which corresponds to the French sévèrement puni. However, in an early study of non-native speakers’ knowledge of collocations, Bahns & Eldaw (1993) found that among their 15 test items, some collocations were more easily translated or paraphrased than others. Also de Cock (2004) observed that L2 speakers do not deal with all multi-word sequences in the same way. While some, as Granger (1998) showed, are drastically underused, other combinations tend to be used more often by non-native speakers of English compared to their native speaker counterparts, while a third group, analogously to Howarth (1996), use a combination which could be traced back to a collocation, but would not be regarded as an acceptable combination. Interestingly enough, all of these studies attribute their results to the fact that their test-takers belong to a group of L1 or L2 speakers of English. However, the differences between the non-native speakers and native language data might be more complex than a simple native/ non-native distinction. For example, even more advanced learners of English are generally exposed to the target language for a much shorter amount of time compared to adult native speakers. Furthermore, in cases where corpora or linguistically trained experts have been used as a benchmark, the level of the L2 participants generally does not match the competence of professional L1 users like authors, journalists or academic researchers. These L1 professionals, in turn, may be less diverse or more accurate in their use of collocations as compared to native users of a language without such a background. But as neuro-cognitive studies (Bowden, Steinhauer, Sanz & Ullman, 2013; Clahsen & Felser, 2006) based on event-related potentials (ERPs) showed, semantic as well as some areas of structural processing are identical in L1 and L2 speakers or at least converge with growing proficiency. Furthermore, Chipere (2003), as well as Dąbrowska & Street (2006) even found examples of non-native speakers outperforming native speakers in analysing passive sentences and more complex sentence structures. However, these observations only held for a comparison of non-native speakers and native speakers with a lower educational level. It is not yet clear as to whether such outcomes also apply to collocations as a broader area between lexical and syntactic phenomena. While older studies 314 Cordula Glass <?page no="315"?> suggest that there are observable differences between L1 and L2 speakers’ collocational proficiency, the question remains whether these differences hold for all groups of native and non-native speakers or if there might be differences depending on age and/ or classroom setting. Therefore, this paper will put L1 and L2 speakers’ collocational competence to a test in order to explore the level of collocational proficiency of native as well as non-native speakers of different age groups (focusing on 15-year-old students and young adults) and to examine potential effects of a larger amount of target language input on L2 attainment. 3 The study Assuming that cognitive learning mechanisms are common to all human beings, proficiency in a language, as in other cognitive abilities, can be achieved through language exposure and use. Therefore, it would be quite likely that, like walking or riding a bike, linguistic skills develop over time. This study focuses on collocational proficiency and the influence of age as well as of linguistic input. Consequently, the research questions to be addressed are as follows: RQ 1: Does age (adults vs. students) affect L1 collocational proficiency in English? RQ 2: How does the amount of English language input affect L2 speakers’ collocational proficiency in English? a) Are there any differences between non-native adults vs. non-native stu‐ dents in Great Britain? b) Are there any differences between native and non-native students in Great Britain in year 11? c) Are there any differences between highly immersed L2 adults and L1 adults/ students? d) Are there any differences between adult speakers of English with regular English-as-a-subject language training and L1 adults/ students and non-na‐ tive highly immersed adults? A total of 109 participants from Great Britain and Germany took part in the study. In order to test their collocational competence, all were presented with a test called CollMatch (Gyllstad, 2007). 315 Collocational proficiency: The effects of ‘target language input’ and ‘age’ <?page no="316"?> 2 For this study native speaker participants (L1, abbreviated GB) are defined as partici‐ pants with at least one English native speaker parent, who indicated in an accompa‐ nying questionnaire that they actively use English in private communication. 3 This school, like the university, is located in Hertfordshire. It is a mixed, non-denom‐ inational secondary school, with academy status. The focus of this school lies on business and enterprises. With an Ofsted label of good with a number of outstanding features, it can be considered a good example of an average school within the outskirts of London. 3.1 Participants The database for this study consists of five sets: adult native speakers 2 of English, adult non-native speakers of English who live in Great Britain (GB), adult German native speakers who study English at a German university, and a group of native as well as non-native speakers from year 11 at a British secondary school 3 . In total, 109 participants contributed to the database of this study. They are distributed across the five sets as follows: • GB adult (native): 24 adult native speakers of English, undergraduate students, average age 21.17 years. • GB adult (non-native): 24 adult non-native speakers of English, undergrad‐ uate students, average age 21.77 years. • GER adult (non-native): 24 adult German undergraduate students of English, average age 21.87 years. • GB year 11 (native): 24 young native speakers of English, average age 15.35 years. • GB year 11 (non-native): 13 young non-native speakers of English, average age 15.41 years. In order to control as much as possible for regional varieties, the four groups from Great Britain (GB) all come from the same community, situated 20 miles north of London. Unfortunately, the sample of young non-native speakers is, with only 13 participants, considerably smaller than the other four because there were only 13 students in year 11 at the time of test-taking whose mother tongue was not English. However, the samples of adult native and non-native speakers in Great Britain, as well as the group of non-native speakers from Germany (GER), are all randomly selected extracts from much larger samples (Glass, 2019), in order to ensure that the group sizes would be comparable for this analysis. None of these groups differs in their results significantly from the results achieved by the larger database. In the following, the acceptance ratings of CollMatch are compared in percent. 316 Cordula Glass <?page no="317"?> 4 The JACETlist (Ishikawa, Uemura, Kaneda, Shimizu, Sugimori & Tono, 2003) is a word list of 8000 items, initially designed for learners of English in Japan (for a detailed description see Uemura & Ishikawa, 2004). 3.2 Test materials To collect the data for this study, the CollMatch 3 (Gyllstad, 2007) was distributed to five groups of native and non-native speakers of English. Initially, this test was devised at the University of Lund in order to provide a suitable tool to allocate a learner’s receptive collocational competence as well as her/ his progress in language learning. With a Cronbach’s alpha of .82, it is highly reliable, and even though it only tests the participant’s ability to recognise collocations, CollMatch 3 correlates well with tests for vocabulary size (r = .90) and vocabulary depth (r = .85 - .90) (see Gyllstad, 2007). The task as such is designed as a judgement test. Participants are presented with 100 verb-noun combinations: 70 of them are collocational items, and 30 are pseudo-items, which usually do not occur significantly often within the English language. Each combination has to be categorised as “English” or “not English”. The items have been predominantly selected on a statistical basis. Taken from the British National Corpus (BNC), they all yield a z-score of 2.58 or over within a span of three words to the right from the node (here: the verb), which translates into 1 % of error tolerance (Berry-Rogghe, 1973; Gyllstad, 2007: 107; Oakes, 1998: 8-9). Furthermore, all lemmata are taken from the JACETlist 4 , to assure that not only native speakers but also (advanced) learners of English are familiar with their form and at least one of their meanings. Collocation items include, for example lose sleep, draw a breath, run a bath; whereas examples such as impose success, rule an award, stand an occasion were classified as pseudo-items. CollMatch was selected for two main reasons. First, it had proven to be a highly reliable test in previous studies (Barfield & Gyllstad, 2009; Gyllstad, 2007), but second, it also provides a good starting point for investigating the extent of a speaker’s knowledge of collocations. Of course, the design of the test refers to passive knowledge of language, which might be regarded as less authentic, since the fact that a subject knows about a combination does not necessarily imply that s/ he actively uses this combination. Yet, at the same time, CollMatch forces participants to judge a relatively wide spectrum of combinations, while production tasks are prone to yield only a limited range of collocational data, which often, depending on the task and/ or topic, represent variations of closely related concepts. Finally, data from judgment tasks are easier to compare, since all participants work on the same test items. 317 Collocational proficiency: The effects of ‘target language input’ and ‘age’ <?page no="318"?> 3.3 Procedure The testing took place in a normal classroom situation. All participants received a copy of CollMatch, which, apart from the 100 verb plus noun combinations (V+N), contained detailed instructions as well as two example items. The instructions encouraged every subject to work through the 100 combinations as swiftly and yet as concentrated as possible. Most participants finished within ten minutes, but there was no actual time limit. Furthermore, none of the subjects knew about the precise linguistic focus of the study. They had been told that this was a test for an academic project, which served the purpose of comparing different students at different schools and universities, but not that this test focused on collocations. Yet the academically trained subjects might have guessed the underlying concept due to their prior training in English linguistics. But this should not affect the test results because even with the concept of collocations in mind, each participant had to rely on either her/ his knowledge or intuition to separate the more acceptable from the less established combinations. There are two types of measures taken from CollMatch. The participants’ scores were measured by the sum of correct evaluations. Once a participant classified a collocation correctly as “English” or “not English”, she/ he was assigned a point. For misclassification, no points were awarded but also none taken off. In total, each subject can obtain a maximum of 100 points (see Table 1). The collocational combinations as such were also measured individually, based on the percentage of positive evaluations (rated as “English”) they received from the participants (acceptance score). Thus, items that are generally known to a group of speakers and classified as “English” by all members of a group, receive 100 %, while pseudo-items would be expected to score close to 0 % (see Figures 1-3). 4 Results Table 1 summarises the average score of the participants for each of the five groups and lists the results for the correct assessment of collocations (items) and pseudo-items. If both categories are taken together (i.e. collocations and pseudo-items, n = 100), L1 adults scored best (70.62 %), followed by L2 highly immersed adults living in Great Britain (66.19 %), L2 highly immersed young students in year 11 living in in Great Britain (53.21 %) and L2 adults living in Germany (50.63 %). L1 young students in year 11 living in Great Britain obtained the lowest scores (49.28 %). With respect to collocations (items, n = 70), adult native speakers of English (adult L1) reached close to ceiling results, with a mean acceptance score of 93.26 %. The young native L1 speakers (young (L1) only obtained 61.68 % 318 Cordula Glass <?page no="319"?> (see also Glass 2019; Gyllstad, 2007). For the non-native speakers, adult L2 professionals (adult (L2, highly immersed)) obtained the highest acceptance rate for collocations (83.97 %), followed by young L2 students (young L2, highly immersed) with a rate of 64.64 %, with the older L2 learners (adult L2, Germany) scoring lowest (63.83 %). Table 1: Overview of CollMatch scores (in percent) for all five groups, subdivided into pseudo-items (right), collocation items (middle) and a sum of both (left). The pseudo-items (n = 30) were correctly identified by most adult native speakers (with the lowest scores), followed by L2 adults from Germany and L2 highly immersed students. L2 highly immersed adults and young L1 students were least able to correctly identify pseudo-items (and received the highest scores). One-way ANOVAs and post-hoc Tukey HSD tests (Table 2) yielded non-sig‐ nificant group differences for the pseudo-items (except for L1 adults vs. L1 young students). In addition, no significant differences were noted for L2 adults vs. L1 students, L2 adults vs. L2 highly immersed students, and L1 students vs. L2 highly immersed students for collocation items and pseudo-items. All other group differences were significant. Table 2: Results from post-hoc Tukey HSD tests. 319 Collocational proficiency: The effects of ‘target language input’ and ‘age’ <?page no="320"?> In general, group differences were more pronounced for collocations than for pseudo-items. Only one group difference was significant for pseudo-items, in contrast to seven significant group differences for the collocations. To be more precise, the results between L2 adults and L1 students, between L2 adults and L2 highly immersed students, and between L1 students and L2 highly immersed students were insignificant for collocations and pseudo-items. Furthermore, the two native speaker groups showed significant differences regarding collocations and pseudo-items, with L1 adults outperforming L1 young students. Finally, L1 highly immersed adults performed significantly better than any other group (except for L1 adults). These findings correspond to Dąbrowska (2012) and Dąbrowska & Street (2006), whose academically educated non-native speakers performed similar to native speakers with a similar educational background. 4.1 Levels of collocational proficiency at different stages of L1 attainment As noted above, there is a significant difference between younger and older native English speakers regarding their collocational competence. This outcome points to observable differences in collocational proficiency at different stages of L1 attainment. Figure 1 illustrates these results (in percent) for the two L1 (native) groups, with all 100 items in CollMatch being consecutively listed. The results for the L1 adults are shown in the top figure and those for the L1 students in year 11 in the bottom figure. 320 Cordula Glass <?page no="321"?> Figure 1: Results from CollMatch for adult native speakers of English (top) and young native speakers of English (aged 15) (bottom). The y-axis shows the rate of acceptance for an item in %. The x-axis represents all 200 items in the test. Collocations are represented in grey, pseudo-items in black. As shown in Figure 1, most of the items in CollMatch are correctly identified as collocations (grey) or as pseudo-items (black) by the native speakers of English, and the items are distributed at the top and on the bottom of Figure 1, respectively. The older native speakers show a clear distribution of collocations vs. pseudo-items (Figure 1, top), with most collocations being correctly rated as English (90 % - 100 %), and most pseudo-items being correctly rated as “not English (0 % - 20 %). However, the younger native speakers (Figure 1, bottom) seem to be less certain in their evaluation of certain test items, with more items receiving an acceptance rate between 20 % - 80 %, and, therefore, a more scattered distribution of the test items. This indicates that even at the age of 15 years, native speakers’ collocational competence is still developing and that they are not as advanced as adult native speakers. In addition, there are some collocational combinations of which both speaker groups seem to be very sure about, while they seem to be less decided on the status of other collocational combinations (which range at about 50 % or even lower). These include items such as supply one’s assistance, afford an opportunity and lay pressure, which have also been found to be problematic for native adults 321 Collocational proficiency: The effects of ‘target language input’ and ‘age’ <?page no="322"?> in Gyllstad’s study (2007: 170), possibly because they may even occur in natural speech, as queries in the BNC revealed (albeit with very low hit rates, see Glass, 2019: 150). 4.2 Collocational proficiency in L2 speakers with a very high target language input Figure 2 illustrates the test scores for the two non-native speaker groups living in Great Britain, namely the highly immersed L2 adults (top, GB adult, non-native) and the highly immersed L2 students (bottom, GB year 11, non-native). Age seems to influence the results as well: Highly immersed L2 adults differentiate collocations from pseudo-items better, and their acceptance rates are more correct than those of younger L2 speakers who also live in Great Britain. Thus, in Figure 2, the items of the L2 adult group are less scattered than those of the L2 younger speaker group. This pattern is similar to that of the two L1 groups, although their level of proficiency differs from native speakers’ performance (see section 4.1). 322 Cordula Glass <?page no="323"?> Figure 2: Results from CollMatch for adult non-native speakers of English living in Great Britain (top) and young non-native speakers of English (aged 15) living in Great Britain (bottom). Collocations are represented in grey, pseudo-items in black. 4.3 Non-native speakers of English with a regular English-as-a-subject language training In Figure 3, the results of the German L2 adult learners, who study at a German university, are illustrated. They identified 63.83 % of the collocations correctly; thus outperforming younger native speakers, who only scored 61.68 % (Table 1). The distribution pattern of the scores for collocations and pseudo-items in Figure 3 looks similar to those from L1 and highly immersed L2 students in year 11 (see Figure 1, bottom and Figure 2, bottom). Thus, pseudo-items (black) range towards the bottom of the diagram, while collocations (grey) are scattered throughout the graph. Recall that this impression was confirmed in post-hoc Tukey HSD tests, which revealed statistical differences between German adults and L1/ L2 adults living in Great Britain, but non-significant differences between German adults and L1/ L2 students from year 11. 323 Collocational proficiency: The effects of ‘target language input’ and ‘age’ <?page no="324"?> Figure 3: Results from CollMatch for advanced learners of English with German as their L1. Collocations are represented in grey, pseudo-items in black. 5 Discussion The aim of this study was to determine whether L1 and L2 speakers’ gradual development in competence and proficiency may also be reflected in the area of collocations. All subjects took the CollMatch, a test which assesses receptive collocational competence. The 109 subjects differed regarding their age (i.e. students in year 11 at age 15 vs. adults at age 21), whether they used English as the L1 or L2 while living in Great Britain, and whether they received a larger or smaller amount of target language input in formal classroom settings (L2 adults having learnt English in school) or in natural settings (i.e. highly immersed L2 adults in Great Britain). As the outcomes of this study indicate, L1 and L2 speakers’ collocational competence develops gradually as a function of age and of target language exposure, confirming previous assumptions with empirical data. 5.1 Effects of exposure and age in L1 collocational proficiency For the L1 group, the results showed that in a simple judgement task like CollMatch, highly educated adult native speakers performed with high accu‐ racy and close-to-ceiling results. While similar results have been observed in previous studies (Glass, 2019; Gyllstad, 2007), this study also points to the fact 324 Cordula Glass <?page no="325"?> that not all L1 participants reached this level: Younger, and potentially less experienced native speakers scored significantly lower than the more proficient native adults. Similar results were also found for syntactic and morphological constructions (Chipere, 2003; Dąbrowska, 2012; Dąbrowska & Street, 2006; Pakulak & Neville, 2010). Thus, it may be assumed that even at the age of about 15, L1 acquisition regarding collocations is not yet completed (see also Wray & Perkins, 2000 on formulaic language). Thus, the findings from this study support the assumption that age remains an important factor at later stages of first language attainment, at least with respect to phraseological phenomena such as collocations. This result also casts doubts on the existence of the native speaker as a universal concept (see also Dąbrowska, 2012): In many studies, monolingual adult native speakers of a language are often considered a unanimous group of language experts who share the same knowledge about the structures and idiosyncrasies of a language, and whose data, therefore, are often used as benchmarks in L2 research. As the present study highlighted, though, even L1 speakers at the age of 15 years still seem to be less proficient in collocations in contrast to academically trained L1 speakers at the age of 21, which is why a careful selection of native speakers (also regarding their age) is essential when using them as a control group in L2 research studies. 5.2 Effects of exposure (and age) in L2 colloquial proficiency Several effects of target language exposure were examined in this study: First, two groups of 15-year-old students in Great Britain were compared, who either used English as a native (L1) or as a non-native (L2) language. The group differences were non-significant for both collocation items and for pseudo-items. This result indicates that at age 15, L1 and L2 speakers of English have a very similar receptive collocational proficiency level in English. This is very likely due to the fact that the L2 speakers in this study have been highly immersed in the English language because they live in Great Britain and use English in all circumstances except at home, where another family language is spoken in addition to English. Second, highly immersed L2 adults living in Great Britain were compared to L1 native English adults (all at age 21). The differences between the two groups were generally non-significant (with the exception of the collocation items where the highly immersed L2 adults did not perform as well as the L1 adults). Thus, a speaker’s L1 may be a less influencing factor, because, for the data at hand, intensive and continuous exposure to the target language resulted in a very good performance in the collocation test, independent of the participants’ 325 Collocational proficiency: The effects of ‘target language input’ and ‘age’ <?page no="326"?> L1. These findings confirm earlier studies by Lambert & Tucker (1972: 204) or Swain & Lapkin (1982: 82), who argue that under certain circumstances L2 speakers can at times even outperform native speakers or at least achieve a native-like level of proficiency. Third, a comparison of the CollMatch’s results for the four groups from Great Britain (L1 and L2, 15 and 21 years) suggests that highly immersed non-native speakers of English indeed performed rather similar compared to their native speaking peers. A difference of six years between the participants on the other hand yielded statistical differences between these groups, highlighting the effect of ‘age’, independent of whether English was the subjects’ L1 or L2. Fourth, this study compared L2 adults (all at age 21 years) who either lived in Great Britain and used English as an additional language in a natural English context or who lived in Germany and had learnt English in a formal classroom context. Not surprisingly, the L2 adults from Germany obtained much lower scores in the collocation test than the L2 adults from Great Britain. Indeed, the overall scores from the L2 adults from Germany resemble the scores obtained by the young L1 speakers at age 15, without any significant group differences between young L1 speakers, highly proficient L2 adults from Great Britain and L2 adults from Germany regarding the test scores of the CollMatch. Finally, the accurate identification of pseudo-items did not seem to be problematic for any of the five groups in this study. That is, independent of their age and their exposure to English, they were able to identify these incorrect collocations as such. Although there were some group differences (with adult native speakers performing best, followed by L2 adults from Germany and L2 highly immersed students, and L2 highly immersed adults and young L1 students showing the lowest performance), statistical analyses did not reveal any significant group differences. Several points are worth mentioning with respect to a qualitative ap‐ proach regarding the correct identification of collocations as “English” and pseudo-items as “not English”: For example, there is a tendency for restricted collocations (e.g. blow a fuse or make a remark) to be less accurately identified by the three non-native groups than by the two native speaker groups. In addition, four collocations (i.e. assume responsibility, fly a flag, employ a technique, afford an opportunity) are also rather abstract, which may explain why three of them were equally problematic for highly immersed L2 speakers of English. In contrast, non-native speakers identified restricted phrases like keep pets, catch fire and shift gear more correctly than restricted phrases. That would indicate that the non-native speakers were more confident about phrases that they presumably encountered on a daily basis. A similar result was observed by 326 Cordula Glass <?page no="327"?> Howarth (1998: 186), who pointed out that restricted collocations are responsible for most L2 deviations (see also Howarth, 1996: 159-161), which also applies to all L2 speaker groups in this study, ranging from traditionally schooled learners to highly immersed non-native speakers. Moreover, a qualitative comparison between the young L1 speakers and the traditionally schooled L2 adults from Germany revealed more differences than similarities, although their scores in the CollMatch were almost the same: Indeed, almost half of the test items were evaluated rather differently in both groups. For example, the traditionally schooled L2 learners from Germany were better at correctly identifying less restricted but abstract and rather academic vocabulary (e.g. raise objections, launch a campaign or adopt an approach) as well as more restricted collocates for several parts of the body (e.g. clear one’s throat, blow one’s nose, snap one’s fingers, clean windows, kick one’s heels or shrug one’s shoulders) than the young L1 speakers were. Unfortunately, there is no clear tendency for a consistent overor undervaluation from any group. Nevertheless, it would be interesting to investigate whether German learners identified these combinations more readily because these items occur more frequently in their L2 input. One could argue, for example, that collocations connected to body parts are indeed learnt at a very early stage of classroom-based English lessons (that is, in primary school in year 3 and 4). Verb plus noun combinations, which refer to a more academic domain, are presented later in the L2 classroom input (in secondary school, from year 5 onwards), while abstract concepts - especially in a university context - are part of the regular input that students of English receive, either in the shape of lectures, seminar discussions or term papers. In general, in this study, all groups differed qualitatively regarding their accuracy rates of different collocational types, suggesting that the type of exposure to the English language that subjects receive should always be investigated in detail. Although statistically not significant, pseudo-items (e.g. *pick a glance, *express a worry, *lay pressure) received higher scores from the adults living in Great Britain whose L1 was not English. However, this picture changes with age: The two groups of 15-year-old students in year 11 (with English as the L1 or the L2) scored relatively low regarding pseudo-items, thus identifying them correctly as “not English”. Nevertheless, they tended to reject some established collocations; for example, one-third of the pseudo-items were overrated by young native speakers (again, not statistically significant). Because the number of pseudo-items in CollMatch (with only 30 items) is rather low, more research is needed in order to examine possible age-related effects as well as effects due to the subjects’ L1 in more detail because at this stage, any language-specific effects (for example of German or any other L1) are difficult to discern. 327 Collocational proficiency: The effects of ‘target language input’ and ‘age’ <?page no="328"?> 5.3 Limitations and scope of further research The results of this explorative study are based on pseudo-longitudinal data with a small sample size, which is why a longitudinal study with more subjects in each group is warranted. A future study could, for example, follow these groups over several years in order to examine the development of their receptive collocation competence from the very beginning to a very advanced level. In her pseudo-longitudinal study, Glass (2019) also included younger English learners from Germany in year 7 and 9 and found that, although English collocational proficiency develops well and although it does show similarities to L1 acquisition, it does not progress linearly. For example, she found the greatest difference not between learners in year 7 and 9 but between year 11 students and learners at university level. This may of course be attributed to the length of exposure to the target language and the subjects’ age, but it may also be due to the exposure to different tasks, i. e. different types of L2 input at school vs. university level (Glass, 2019: 186). Ideally, further studies on the acquisition of collocations would also include a productive collocation test, to obtain a holistic view as to how collocations in general and different types of collocations in particular are acquired produc‐ tively and receptively in different language acquisition settings and at different ages. In this respect, it may also be useful for future studies to investigate productive knowledge of collocations and its relationship with productive knowledge of single-word items. Finally, it would also be interesting for language teaching research to de‐ termine to what extent L2 input in the classroom influences the learning of collocations. The findings of the present study suggest that, despite similar scores in the collocation test, native speakers and L2 learners differ qualitatively in their mastery of different types of collocation, probably attributable to the nature and frequency of the linguistic input during the acquisition process. Such a study could compare different classroom settings (i.e. regular vs. bilingual contexts) in order to discern whether more input alone can contribute to better collocational proficiency among EFL learners. 6 Conclusions The present study shows that native and non-native speakers generally achieve different proficiency levels regarding collocations. Yet a more detailed look at these differences suggests that depending on the factors of ‘age’ and the degree of target language ‘immersion’, some speakers performed better than others, independent of whether English was their L1 or L2. For the conception of the 328 Cordula Glass <?page no="329"?> native speaker this might have two interesting implications: First of all, that the attainment of collocational proficiency could last well into native speakers’ late teens, but also - as for example, Dąbrowska (2012) suggests - that the native speaker, as a unanimous concept, may not exist after all. Independent of the factor ‘age’, non-native speakers of English who attain the target language in a highly immersed setting perform equally as well as their native-speaking peers. On the other hand, students who learn English almost exclusively as a subject at school perform quantitatively at the same level as younger L1 speakers, but even then, the quality of their evaluations differs in many cases. This observation could have two major implications for the teaching of English as a foreign language: First, students who do not live in a country where the target language is spoken may also profit from immersion programmes, since they use the language as a genuine medium for communication rather than a mere subject. For the same reasons, an immersion classroom also provides a more authentic L2 input, which may lead to a better understanding of collocational combinations, similar to the rather native-like collocational proficiency of both highly immersed L2 groups in this study. Thus, independent from the classroom setting, this study also advocates a more authentic L2 input in general. In addition, Källkvist, Gyllstad, Sandlund & Sundqvist (2017) point out that the use of learners’ L1 may also have a positive effect on the learners’ target language in a classroom since it could facilitate the learning of vocabulary and certain grammatical structures, especially in awareness-raising activities. Therefore, Corcoll López & González-Davies (2015) encourage a “target language mainly” as opposed to a “target language only” approach. As this paper has shown, the acquisition of collocations is rather complex, since their attainment is not only based on isolated vocabulary or on isolated syntactic structures in a traditional sense but relies on an intricate and lan‐ guage-specific combination between vocabulary and grammar. Especially with regard to more restricted word combinations, where even native speakers seem to struggle to explain their cause or function, attainment via authentic language input could prove to be a viable way for all L2 learners, independent of their proficiency level. 329 Collocational proficiency: The effects of ‘target language input’ and ‘age’ <?page no="330"?> References Bahns, J. & Eldaw, M. 1993. Should we teach EFL students collocations? System, 21 (1), 101-114. Barfield, A. & Gyllstad, H. 2009. Researching Collocations in Another Language - Multiple Interpretations. Basingstoke: Palgrave Macmillan. Barnbrook, G., Mason, O. & Krishnamurthy, R. 2013. Collocation - Applications and Implications. London: Palgrave Macmillan. Berry-Rogghe, G.L.M. 1973. The computation of collocations and their relevance to lexical studies. In: Aitken, A.J., Bailey, R.W. & Hamliton-Smith, N. (Eds.), The Computer and Literary Studies. Edinburgh: Edinburgh University Press. 103-112. Besalke, B. (Ed.). 2017. A Dictionary of the English Language: A Digital Edition of the 1755 Classic by Samuel Johnson. johnsonsdictionaryonline.com. May 30, 2021. Bowden, H., Steinhauer, K., Sanz, C. & Ullman, M.T. 2013. Native-like brain processing of syntax can be attained by university foreign language learners. Neuropsychologia, 51 (13), 2492-2511. Bybee, J. 2010. Language, Usage and Cognition. Cambridge: Cambridge University Press. Bybee, J. 2013. Usage-based theory and exemplar representation of constructions. In: Hoffmann, T. & Trousdale, G. (Eds.), The Oxford Handbook of Construction Grammar. Oxford: Oxford University Press. 49-69. Chipere, N. 2003. Understanding Complex Sentences - Native Speaker Variation in Syntactic Competence. Basingstoke: Palgrave Macmillan. Clahsen, H. & Felser, C. 2006. How native-like is non-native language processing? Trends in Cognitive Sciences, 10 (12), 564-571. Corcoll López, C. & González-Davies, M. 2016. Switching codes in the plurilingual classroom. ELT Journal, 70 (1), 67-77. Cowie, A.P. 1983. General introduction. In: Cowie, A.P., Mackin, R. & McCaig, I.R. (Eds.), Oxford Dictionary of English Idioms. Oxford: Oxford University Press. x-xvii. Dąbrowska, E. 2012. Different speakers, different grammars. Linguistic Approaches to Bilingualism, 2 (3), 219-253. Dąbrowska, E. & Street, J.A. 2006. Individual differences in language attainment: Com‐ prehension of passive sentences by native and non-native English speakers. Language Science, 28 (6), 604-615. De Cock, S. 2004. Preferred sequences of words in NS and NNS speech. Belgian Journal of English Language and Literature, 225-246. Ellis, N. 2006. Language acquisition as rational contingency learning. Applied Linguistics, 27 (1), 1-24. Ellis, N. 2008. Usage-based and form-function focused language acquisition: The asso‐ ciative learning of constructions, learned attention, and the limited L2 endstate. In: 330 Cordula Glass <?page no="331"?> Robinson, P. & Ellis, N. (Eds.), Handbook of Cognitive Linguistics in Second Language Acquisition. New York, NY: Routledge. 372-405. Ellis, N. & Larsen-Freeman, D. 2009. Language as a Complex Adaptive System. Malden, MA: Wiley-Blackwell. Ellis, N., Simpson-Vlach, R. & Maynard, C. 2008. Formulaic language in native and second language speakers: Psycholinguistics, corpus linguistics and TESOL. TESOL Quarterly, 42 (3), 375-396. Firth, J.R. 1968. A synopsis of linguistic theory, 1930 - 55. In: Palmer, F.R. (Ed.), Selected Papers of J.R. Firth 1952 - 59. London: Longman. 168-205. Glass, C. 2019. Collocations, Creativity and Constructions: A Usage-Based Study of Collo‐ cations in Language Attainment. Tübingen: Narr Francke Attempto. Goldberg, A. 2006. Constructions at Work. Oxford/ New York, NY: Oxford University Press. Granger, S. 1998. Prefabricated patterns in advanced EFL writing: Collocations and formulae. In: Cowie, A.P. (Ed.), Phraseology - Theory, Analysis, and Applications. Oxford: Clarendon Press. 145-160. Granger, S. & Paquot, M. 2008. Disentangling the phraseological web. In: Granger, S. & Meunier, F. (Eds.), Phraseology - An Interdisciplinary Perspective. Amsterdam: John Benjamins. 27-49. Gyllstad, H. 2007. Testing English Collocations - Developing Receptive Tests for Use with Advanced Swedish Learners. Lund: Språkoch Litteraturcentrum, Lund University. Halliday, M.A.K. 1966. Lexis as a linguistic level. In: Bazell, C.E., Catford, J.C., Halliday, M.A.K. & Robins, R.H. (Eds.), In Memory of J.R. Firth. London: Longman. 148-162. Hausmann, F.J. 1984. Wortschatzlernen ist Kollokationslernen. Praxis des Neusprachlichen Unterrichts, 31, 395-406. Hausmann, F.J. 1985. Kollokationen im deutschen Wörterbuch. Ein Beitrag zur Theorie des lexikographischen Beispiels. In: Bergenholtz, H. & Mugdan, J. (Eds.), Lexikographie und Grammatik - Akten des Essener Kolloquiums zur Grammatik im Wörterbuch, 28. - 30.6.1984. Tübingen: Max Niemeyer Verlag. 118-129. Herbst, T. 1996. What are collocations: Sandy beaches or false teeth? English Studies, 77 (4), 379-393. Herbst, T. 2011. Choosing sandy beaches - collocations, probabemes and the idiom principle. In: Herbst, T., Faulhaber, S. & Uhrig, P. (Eds.), The Phraseological View of Language - A Tribute to John Sinclair. Berlin: De Gruyter Mouton. 27-57. Howarth, P.A. 1996. Phraseology in English Academic Writing - Some Implications for Language Learning and Dictionary Making. Tübingen: Max Niemeyer Verlag. Howarth, P.A. 1998. The phraseology of learners’ academic writing. In: Cowie, A.P. (Ed.), Phraseology - Theory, Analysis, and Applications. Oxford: Clarendon Press. 161-186. Ishikawa, S., Uemura, T., Kaneda, M., Shimizu, S., Sugimori, N. & Tono, Y. 2003. JACET 8000 - JACET List of 8000 Basic Words. Tokyo: JACET. 331 Collocational proficiency: The effects of ‘target language input’ and ‘age’ <?page no="332"?> Jones, S. & Sinclair, J.M. 1974. English lexical collocations. A study in computational linguistics. Cahier de Lexicologie, 24, 15-61. Källkvist, M., Gyllstad, H., Sandlund, E. & Sundqvist, P. 2017. English only in multilingual classrooms? LMS - Lingua, 2017 (4), 27-31. Krashen, S.D. & Terrell, T.D. 1983. The Natural Approach - Language Acquisition in the Classroom. New York, NY: Phoenix ELT. Lambert, W.E. & Tucker, G.R. 1972. Bilingual Education of Children - The St. Lambert Experiment. Rowley, MA: Newbury House. MacWhinney, B. 2001. Emergentist approaches to language. In: Bybee, J. & Hopper, P. (Eds.), Frequency and the Emergence of Linguistic Structure. Amsterdam: John Benjamins. 449-470. Nesselhauff, N. 2004. Collocations in a Learner Corpus. Amsterdam: John Benjamins. Oakes, M.P. 1998. Statistics for Corpus Linguistics. Edinburgh: Edinburgh University Press. Pakulak, E. & Neville, H.J. 2010. Proficiency differences in syntactic processing of monolingual native speakers indexed by event-related potentials. Journal of Cognitive Neuroscience, 22 (12), 2728-2744. Pawley, A. & Syder, F.H. 1983. Two puzzles for linguistic theory - nativelike selection and nativelike fluency. In: Richards, J.C. & Schmidt, R.W. (Eds.), Language and Communication. London: Longman. 191-226. Siepmann, D. 2005. Collocation, colligation and encoding dictionaries, Part 1: Lexicological aspects. International Journal of Lexicography, 18, 409-443. Sinclair, J.M. 1966. Beginning the study of lexis. In: Bazell, C.E., Catford, J.C., Halliday, M.A.K. & Robins, R.H. (Eds.), In Memory of J.R. Firth. London: Longman. 410-430. Sinclair, J.M. 1991. Corpus, Concordance, Collocation. Oxford: Oxford University Press. Swain, M. & Lapkin, S. 1982. Evaluating Bilingual Education - A Canadian Case Study. Clevedon: Multilingual Matters. Tomasello, M. 2005. Constructing a Language - A Usage-Based Theory of Language Acquisition. Cambridge, MA: Harvard University Press. Uemura, T. & Ishikawa, S. 2004. JACET 8000 and Asia TEFL vocabulary initiative. The Journal of ASIA TEFL, 1 (1), 333-347. Wode, H. 1995. Lernen in der Fremdsprache - Grundzüge von Immersion und bilingualem Unterricht. Ismaning: Max Hueber Verlag. Wray, A. & Perkins, M. 2000. The functions of formulaic language: an integrated model. Language and Communication, 20 (1), 1-28. 332 Cordula Glass <?page no="333"?> Part 4: Metaphors used in the context of a putative ‘bilingual advantage’ <?page no="335"?> The path to bilingualism, a road to better cognitive performance? Metaphors of language learning in Cognitive Science Silke Jansen Abstract This paper proposes to anlyse research on the relationship between lan‐ guage and cognition as a social practice of knowledge construction, which largely relies on metaphor as a cognitive tool. It centers on metaphorical representations of language acquisition and learning (LAL) within research into the so-called “bilingual advantage”. Traditionally concerned only with “native” bilinguals, this line of research has recently expanded its focus to include L2 learners, arguing that they may be considered as a kind of bilinguals “in training”. Eighteen scientific papers that represent this recent development were analysed and contrasted with the DemBil corpus, an electronic corpus containing more than 500 scientific papers on possible cognitive differences between monolinguals and bilinguals published between 1923 and 2018. It appears that views of LAL in these papers largely rely on two conceptual metaphors, namely LANGUAGE LEARNING IS MOVING FORWARD ON A PREDEFINED PATH and LANGUAGE LEARNING IS BEING EXPOSED TO AN ACTIVE SUB‐ STANCE . The recourse to these metaphors remains remarkably stable, regardless of different research designs and contrasting findings that prevailed at different moments in time. We explore the roots of these metaphors in academic and folk conceptions of language learning, point out the tensions between these models and current theories of LAL, and critically discuss some undue generalisations which were drawn because researchers have taken such metaphors literally. <?page no="336"?> 1 Introduction Since the development of the first intelligence tests in the 1910s, continuous attempts have been made to detect, measure and explain possible cognitive differences between monolingual and bilingual speakers. Within this period, views on bilingualism have radically changed: While early studies from the 1920s and 1930s claimed that growing up with more than one language caused serious harm to childrens’ mental and emotional development, research from the 1960s onwards has seen bilingualism as a source of “mental training” that turns bilingual speakers into cognitive high performers. In previous publications ( Jansen, 2021; Jansen, Higuera del Moral, Barzen, Reimann & Opolka, 2021), we have argued that this complete turnaround within only one century may be, at least in part, due to a cultural bias, because of the striking parallels between these findings and overarching explanatory paradigms that circulated in Western societies and supported different stances towards bilingualism at different moments in time: While in the 19 th and early 20 th century nationalism and impe‐ rialism advocated monolingualism as an expression of distinct national, cultural and linguistic identities, the more recent philosophies of multiculturalism and postmodernity celebrate fluid identities and linguistic pluralism (Kubota, 2016), and neoliberalism promotes competition and performance orientation (Charteris-Black, 2018: 210). In addition (and more importantly perhaps), the models that researchers have built to explain how bilingualism affects cognition also seem to emerge from social experience. The debate on the bilingual (dis)advantage is essentially based on the idea that the bilingual’s mind contains two languages that stand in conflict with each other. In the first half of the 20 th century, when bilingualism was believed to be harmful to cognitive development, this conflict is metaphorically construed as a struggle or war between the two languages in the mind: And for a child learning to attach labels to objects, actions, learning to think by means of words, is it really an advantage to have two words, three words, for the same thing? … Of myself I can say that I have had great difficulties to overcome in matters of thought, speech, expression, and that I attribute these difficulties to early bilingualism and the constant struggle of two languages for precedence (O’Grady, 1917, in: Smith, 1923: 273). In contrast to this, later publications that celebrate cognitive benefits of bilin‐ gualism metaphorically understand language conflict in the speakers’ brain as a competition between language components. For example, Engel de Abreu et al. attribute alleged higher performance of bilinguals to 336 Silke Jansen <?page no="337"?> […] the experience of managing several languages on a regular basis [which] trains executive functions that are needed to resolve conflict between competing language systems and improves their functioning across other tasks and domains […] (Engel de Abreu, Cruz-Santos, Tourinho, Martin & Bialystok, 2012: 1364). This competition is generally framed in terms of sports or business competition. However, brains have nothing literally in common with a battlefield, an arena for sports contests, or a market. When researchers talk about languages or linguistic components as if they were struggling or competing with each other, they attribute human characteristics to them, such as self-interest and purposeful behaviour. What is more, they interpret the consequences of “struggle & war” and “competition” in the bilingual brains in direct analogy to fundamental social experiences: While war - a central experience for researchers living in Europe and the USA during the first decades of the 20 th century - brings destruction and suffering, competition, which pervades contemporary societies, is considered to be a means of enhancing the fitness and performance of social, economic, or other kinds of systems. In summary, the scientific debate on bilingualism and cognition is essentially based on an anthropomorphic metaphor, which re‐ ceives different interpretations according to the changing social circumstances researchers face. These observations illustrate three fundamental aspects of scientific practice that we want to address in the present contribution, too: 1. Especially when focusing on phenomena which are not amenable to direct observation (such as speech production processes in the brain), researchers have to rely on conceptual models in order to design their experiments, and to provide scientific explanations for their data. Metaphor is the most common tool in scientific knowledge construction (Brown, 2003; Goschler, 2007; Jäkel, 2003): To facilitate understanding, the research object (e.g. processes in the brain) is conceived of in terms of another domain of human experience (e.g. conflict and competition between humans). 2. Just as any other form of human knowledge production, psychometric research into bilingualism and cognition is socially situated, because the selection and interpretation of scientific metaphors emerges from cultural and disciplinary traditions, as well as researchers’ experience in ever-changing historical situations. 3. Especially when metaphor awareness is hardly developed and metaphors are taken literally, the metaphorical nature of scientific models can easily lead researchers astray because of unjustified entailments derived from them. We will discuss these issues on the basis of some recent developments within the “bilingual advantage” debate. Following the ideal of the “native” speaker, 337 The path to bilingualism, a road to better cognitive performance? <?page no="338"?> this line of research has mostly focused on cognitive differences between monolinguals and bilinguals, the last being defined as speakers who acquired two languages from early childhood and are equally skilled in both of them. In recent years, however, some researchers have begun to highlight the gradual nature of bilingual competences, and to acknowledge that there are many ways of becoming bilingual even later in life. As a consequence, more and more studies have included L2 learners into psychometric experiments, trying to determine whether they show differences in cognitive performance in comparison with monolinguals and/ or “native” bilinguals. This new focus on how different kinds of bilingualism affect cognitive functioning has forced researchers to place more emphasis on LAL. The aim of the present contribution is to analyse LAL metaphors within this line of research, to explore their historical roots and their relationship to the development of the “classical” bilingual (dis)advantage debate, and to discuss their merits and fallacies for scientific reasoning. As may be apparent from these observations, this paper is not a direct contribution to the debate on cognition and bilingualism. Instead, it adopts a meta-perspective, taking scientific discourse as a research object in itself. Applying basic notions of Cognitive Linguistics within a social-constructivist approach, we propose to analyse science as a social practice in which researchers jointly construct assumptions about the world. The theories and models they produce can have predictive or explanatory power to a greater or lesser extent, but they always remain human constructions that arise from bodily perceptions as well as socio-cultural experience. By this means, we would like to encourage discussion on the heuristic status of scientific models and findings, as well as on their social embeddedness. 2 Theoretical framework and methods 2.1 Metaphors in Cognitive Science According to Lakoff & Johnson (1980: 5), metaphors can be understood as “understanding and experiencing one kind of thing or experience in terms of another”. For example, when we say that somebody speaks a language fluently, or when we characterise languages or linguistic practices as pure or mixed, we metaphorically treat language as a substance. Likewise, when cognitive scien‐ tists write in their papers that information is stored or processed in the brain for speech production, in analogy to raw material in industrial production processes, they rely on the conceptul metaphor THE HUMAN BRAIN IS A WORKSHOP (cf. Goschler, 2007; Jäkel, 2003). In these kinds of metaphorical models of reality, a piece of well-known everyday experience (typically a concrete thing such as 338 Silke Jansen <?page no="339"?> 1 Following the general practice in cognitive linguistics, we designate source and target domains with capital letters, and write surface metaphorical expressions (as well as any other linguistic items in metalinguistic use) in italics. a substance or a workshop) is used as a conceptual model for understanding an object of observation (typically an abstract and/ or inscrutable phenomenon such as language or cognitive processes in the brain). The field of knowledge from which the metaphor is taken is referred to as “source domain”, and the phenomenon under study as “target domain”. Metaphorical mappings between different domains (e.g. language and substances) can be traced in discourse through lexical items (e.g. fluency, mixed, pure, etc.). 1 Many of these expressions are so highly conventionalised that we hardly perceive them as metaphorical. Based on whether the source domain belongs to the area of bodily and per‐ ceptual experience with the physical world (e.g. the SUBSTANCE metaphor), or constitutes a complex and culturally bound concept (e.g. the WORKSHOP met‐ aphor), Lakoff & Johnson (1980) distinguish between ontological and structural metaphors. Ontological metaphors play a major role in what Johnson (1987) calls image schemas, i. e. “directly meaningful […] preconceptual structures, which arise from, or are grounded in, human recurrent bodily movements through space, perceptual interactions, and ways of manipulating objects” (Hampe, 2005: 1). Central image schemas are, for example, the CONTAINER schema, the PATH schema or the FORCE schema. In line with these abstract schemas, languages are often construed as CONTAINERS (for example when we say that a text is written in or translated into English), and LAL as a PATH (for example when we distinguish between more or less advanced learners), or the effect of a FORCE acting upon learners (e.g. when researchers interpret LAL as a consequence of language exposure). By highlighting similarities and concealing aspects of the target domain that do not coincide with the source domain, a metaphor suggests a particular understanding of the object in discussion, rather than another (cf. Brown, 2003: 17-18; Lakoff & Johnson, 1980: 10). For example, the metaphors of conflict and rivalry that have dominated in research into bilingualism and cognition hide other possible approaches to language coexistence in the brain that could be based on cooperation and mutualism. To which extent these metaphors have shaped researchers’ understanding of bilingualism becomes clear if we call in mind their theory-constitutive function: Because alleged negative or positive effects of bilingualism on cognition are essentially explained as a consequence of an underlying conflict or competition between the bilingual’s languages, the metaphor has “become an inherent part of [the] theory and cannot be replaced 339 The path to bilingualism, a road to better cognitive performance? <?page no="340"?> by non-metaphorical descriptions without affecting this theory” (Drogosz, 2018: 37; cf. also Brown, 2003: 15). The fundamental heuristic function of metaphor can energise scientific practice, for example when new metaphors stimulate new ways of conceiving the target domain and inspire researchers to formulate new hypotheses through entailments, i. e. logical inferences derived from the underlying cognitive model (e.g. if languages are in competition, this enhances cognitive fitness). However, its suggestive power may also lead researchers to draw a distorted or fictitious picture of the research object, especially when a model is erroneously taken for reality (cf. Brown, 2003; Lakoff & Johnson, 1980, and our observations in section 5). This happens all the easier when a metaphor is highly conventionalised. 2.2 Methods Our aim is to analyse metaphors of LAL in two directions of the debate on the “bilingual advantage”. The first direction is what we call the “classical” debate which has been going on since the 1920s. These studies are based on a maximum definition of bilingualism and thus consider only so-called “native” bilinguals, i. e. speakers who grew up with and have equal skills in two languages. The second direction, which we call the “L2 advantage debate” for the purpose of this contribution, operates with a gradual definition of bilingualism, according to which L2 learners can be considered to be bilingual to a certain degree. As a consequence, these studies acknowledge different kinds of bilingualism and compare L2 learners to monolinguals and/ or “native” bilinguals. Studies of this type can only be found since 2008 and are much rarer than studies of the “classical” type. For a systematic analysis of the “classical” debate, we rely on the DemBil corpus, an electronic corpus that was created for the research project Demysti‐ fying Bilingualism (cf. Jansen et al., 2021). This corpus contains 534 scientific papers (comprising 3,433,225 tokens and 38,329 types) which were published between 1923 and 2018 and deal with cognitive differences between monolin‐ guals and bilinguals (generally understood as “native bilinguals”). The corpus is available in CQPweb (cwb.sourceforge.net/ cqpweb.php), a web-based corpus analysis system that allows for different kinds of word queries and various other analysis options. The “L2 advantage debate” is represented by eighteen papers, published between 2008 and 2020, that we identified through a systematic search in scientific databases, based on keywords, title and abstract. Although we cannot guarantee that our selection is exhaustive, we are confident that it covers the majority of publications of this type. As the text amount for this line of research 340 Silke Jansen <?page no="341"?> is relatively small and does not seem to be worth the effort to build a corpus in CQPweb, the texts were analysed manually. Metaphors in both corpora were detected by a combination of inductive and deductive as well as qualitative and quantitative approaches. After familiarising ourselves with the texts, we compiled a list of metaphorical expressions related to the conceptualisation of LAL. Together with related lexical items (e.g. synonyms, antonyms, words from the same lexical field, etc.), these expressions were used as starting points for computer-assisted analyses of the DemBil corpus. Given that concordance programmes such as CQPweb only show a decontextualised chunk for every occurrence of a search word, and that metaphoric uses become only visible within a larger context, promising chunks were checked against the original texts. This often led to the discovery of new metaphorical expressions, which were themselves submitted to quantitative and qualitative analysis. Frequency patterns were examined for central candidates using different functions of the CQPweb interface, among them key word and collocation analysis, frequency breakdowns (a function which allows to rank different types of the same query according to their frequency in the corpus), or distribution according to time of publication. It is important to note that authors often rely on the same source domain to describe a set of different target domains. For example, in our corpora, the PATH schema is used for three different target domains: 1. LAL, 2. cognitive development, and 3. scientific practice. As a consequence, PATH surface metaphorical expressions can have a variety of functions in the discourse. For example, the query for the verb follow delivers more than 2,000 matches in the corpus, most of them in expressions such as as follows, it follows that etc., which belong to the metaphor SCIENTIFIC ARGUMENTATION IS A JOURNEY . Disambiguation was achieved through a systematic search for combinations with other words from the same source domain, sometimes by using wildcards (e.g. the query “{follow} * * * * * * * path” delivers matches such as “follow the same path”, “follow a linear path”, “follows a privileged developmental path”). When this was not possible, the matches were checked manually, and only those occurrences were counted that belonged to the target domain under scrutiny. In the following sections, relevant surface metaphors will be presented in italics, with absolute frequencies in brackets behind each word. Note that the significance of a given conceptual metaphor for knowledge construction does not primarily emerge from the absolute frequency of lexical items, but from systematic metaphorical mappings, as reflected in the sustained use of a broad range of expressions from the source domain in descriptions of the target domain. 341 The path to bilingualism, a road to better cognitive performance? <?page no="342"?> 2 Relevant surface metaphors are highlighted in bold for better recognition. To anticipate, we will already reveal at this point that the texts draw on a plethora of metaphors for LAL, many of which have been mentioned in previous publications on learning metaphors (e.g. LEARNING IS INCORPORATING, LEARNING IS CONSTRUCTING , cf. Marsch, 2009). However, only two of these metaphors are theory-constitutive in the texts analysed here, namely LAL IS MOVING FORWARD ON A PREDEFINED PATH , and LAL IS BEING EXPOSED TO AN ACTIVE SUBSTANCE . As we will argue in section 5, these metaphors can be considered as instantiations of the image schemas PATH and FORCE . Given that these metaphors are pivotal for how researchers model the effects of bilingualism on cognition and thus have shaped scientifique practice to a greater extent than others, they will be the main focus of our contribution. 3 The PATH schema: Learning a language is moving forward on a path 3.1 The PATH schema in the “classical” debate on the bilingual (dis)advantage In the DemBil corpus, processes of LAL are construed in terms of a predefined route from a source or starting point (being a non-speaker) to an endpoint (full mastery of the language), passing through a sequence of contiguous locations or developmental stages. The following passage illustrates how different surface metaphors related to movement in space and time combine to evoke an understanding of learning according to the PATH image scheme (cf. Johnson 1987: 113 ff): 2 […] bilingual children exposed to two languages from birth achieve each and every major linguistic milestone in their one language, on the same time table as their other language, and both languages proceed on the identical time table as observed in the monolingual child […] (Kovelman, Baker & Petitto, 2008: 204, our emphasis). According to the descriptions found in our corpus texts, learners follow the path (9) as they progress (88), proceed (9) or advance (11) in the acquisition process. They reach (11), attain (118), achieve (14) or pass (3) predetermined points on the path, which are referred to as milestones (60) or landmarks (3), according to a fixed timetable (8) before achieving the final goal (8) or aim (2) of (ideally full) proficiency in the language. In the case of the speaker’s first language(s), this process begins with birth and ends at some point in late childhood. These examples illustrate how the discourse constructs LAL as a straightforward, 342 Silke Jansen <?page no="343"?> unidirectional and predictable process on the bases of lexical choices. This is not an exception within systems of knowledge production: According to other studies in the field of metaphor, the PATH schema is one of the most commonly used source domains for the target domain of learning, both in folk and academic reasoning, as well as across national and cultural contexts (cf. Alghbban, Salamh & Maalej, 2015: 11; Berendt, 2008: 76; Cortazzi & Jin, 1999: 159-160; Smith, 2013-2014: 27, 29). Interestingly, the PATH image scheme and related surface metaphors cannot only be found in descriptions of LAL but are also used to conceptualise other aspects of cognitive development. Compare the following, exemplary passage, in which children are assumed to move forward on a predefined trajectory until reaching an adult level of both mental and linguistic competence: Following from additional increases in frontal lobe function around ages 5-6 years […] begins the important near final period in linguistic development by the end of which time children complete the acquisition of among the most complex grammatical principles of their native language […]. Ages 5-6 years not only constitute a time whereupon specific brain changes are linked with linguistic and cognitive milestones […], but such advances in higher cognition, in turn, provide the foundation for affording the child greater social and personal independence (Kovelman et al., 2008: 205-206, our emphasis). As illustrated by this quotation, different dimensions of cognitive and linguistic development are conceived of as a bundle of distinctive paths that all individuals follow in parallel. Speakers who grow up with two languages are considered to follow several acquisition paths simultaneously and independently from each other: Thus, the number of languages in the environment modified children’s expectations about words and their meanings, possibly setting the stage for different paths of language learning (Bialystok, Craik, Green & Gollan, 2009: 120, our emphasis). This is also illustrated by the recurrent use of the adjectives parallel in collo‐ cations with nouns that refer to different aspects of cognitive and linguistic functioning and development. A frequency breakdown applied to the nouns that combine with parallel delivers the following results: activation (position 1), language (pos. 3), acquisition (pos. 5), activity (pos. 6), development (pos. 7). The idea that bilingual language acquisition is moving forward on different parallel acquisition paths, rather than developing one single complex linguistic repertoire, emerges from a combination of the PATH schema with a conceptu‐ alisation of languages according to the CONTAINER schema. Under this view, 343 The path to bilingualism, a road to better cognitive performance? <?page no="344"?> a fuzzy and messy phenomenon such as human language is reified as a discrete entity with fixed boundaries that separate the inside from the outside of the container, in line with widespread and culturally well-entrenched language myths (cf. Watts, 2003: 6-7). A frequency breakdown of the adjectives that combine with the form languages in the DemBil corpus illustrates this view, because these lexemes highlight the conceptualisation of languages as separated and thus countable entities: Different (pos. 1), multiple (pos. 2), other (pos. 3), second (pos. 5). The following two examples from our corpus illustrate how these (probably unquestioned) assumptions contribute to the discursive construction of mono‐ lingual and bilingual language acquisition as distinctive processes, although under an alternative view they could also be described as belonging to one global acquisition process: […] both monolingual and bilingual children go through the same major milestones in language development at approximately the same time. Commonly recognized stages are: babbling (playing with sounds apparently without intending to convey meaning) during the period of roughly 6-12 months […] (Sorace, 2007: 195, our emphasis). Children learning two languages simultaneously are delayed in terms of vocabulary in each of their languages, although their total vocabulary is comparable to their monolingual peers (Weigmann, 2014: 1018). If researchers did not assume languages to exist as discrete and distinguishable entities that are learned on parallel predefined paths, the bilingual children mentioned in the quotations above would just appear as “ordinary” language acquirers both in terms of speech and amount of acquisition. This example is a telling illustration of the socially constructed nature of scientific categories and the role that metaphor plays in this construction. It also shows that bilingual language acquisition is ultimately understood as dual monolingual acquisition, a view that implicitly sets the monolingual speaker as a norm, in line with the monolingual mindset that has characterised Western societies for at least two centuries (cf. Radwańska-Williams, 2008; Watts, 2003: 278). In addition to language acquisition from early childhood, the PATH scheme also underlies representations of language learning processes that begin later in life. In this case, speakers are supposed to enter a new path at a moment when they have already covered a certain portion or even the whole of their first path. To the extent that they move forward towards “full” mastery of the second language, they are considered to simultaneously move towards bilingualism. This idea is illustrated by the use of collocations that combine the noun bilingual with different kinds of classifying adjectives that situate learners on different 344 Silke Jansen <?page no="345"?> 3 Note that the unequal mastery of two languages is often referred to as pseudo-bilin‐ gualism in papers published before the middle of the 1960s. positions on the path (cf. intermediate bilinguals (38), near monolinguals (13), advanced bilinguals (3), etc.; compare also the adjective combination near native (9), that is used with speaker (5) or nouns such as competence (1), fluency (1), proficiency (1), etc.). Given that movement along the L1 path coincides with movement in time across the lifespan, temporal adjectives (e.g. early (558), late (499) bilingual(ism), later (14) and new (3) bilingual(s); parallel (1), bilingual and simultaneous (171) bilingual(ism), as opposed to successive (195) bilingual(ism) or expressions that refer to particular periods in life (cf. crib bilingual(ism) (28), child bilingual(ism) (13), adult bilingual(ism) (214); cf. also lifelong bilingual(ism) (62)) are used to locate the L1 and L2 acquisition paths relative to each other (cf. also highly frequent collocations such as first language (493), second language (2,404), third language (99), as well as prior language(s) (18), subsequent language(s) (19), further language(s) (6), etc.). Together with language names such as English (pos. 1 in frequency breakdown for adjectives that precede the lemma bilingual) and Spanish (position 10), late (pos. 2), early (pos. 3), successive (pos. 8) and simultaneous (pos. 9) are among the most frequently used adjectives with bilingual. In line with the general understanding in the discourse, Sorace (2007: 200) states that “simultaneous early exposure to more than one language seems to provide an effortlessly natural path to becoming bilingual.” Accordingly, the adjectives natural (20) or naturalistic (6) are used to describe this acquisition scenario, suggesting by implication that other settings are “unnatural”. The term native-like, which comes up in the 1970s and used as a technical term until today (104 matches in the DemBil corpus), implies that L2 learners may come close to bilingualism as they move forward on the path, but will never actually reach this goal. This is another expression of the implicit monolingual norm that underlies the discourse: “Natural” bilinguals are only those whose language experience with their two languages is identical to monolinguals of either language. Although the PATH metaphor inherently implies that linguistic competence is a gradual phenomenon, it is explicitly or implicitly claimed in the “classical” debate that the best representatives of the monolingual-bilingual continuum can be found at the poles: When the authors use expressions such as genuine (11), real (3), true (10), ideal (6) or extreme (1) bilingual(s)  3 , they generally refer to speakers with maximum skills in two languages, while extreme monolinguals (1) are speakers with no knowledge whatever in a second language. As a consequence, studies systematically recruit participants from these extreme 345 The path to bilingualism, a road to better cognitive performance? <?page no="346"?> groups, may it be for practical or ideological reasons, or both. Nevertheless, these kinds of speakers are probably of negligible significance in today’s real-world contexts, where the vast majority of people have contact to more than one language in their daily life or in educational contexts and develop different kinds of skills in their languages. This may be the reason why, from around 2008 onwards, researchers started to become interested in possible cognitive (dis)advantages of speakers with different kinds of language biographies and levels of L2 proficiency. In the following section, we examine how the PATH schema is used in the “L2 advantage debate”. 3.2 The PATH schema in studies on the “bilingual advantage” in L2 learners In recent times, the traditional dichotomous understanding of monolingualism and bilingualism has increasingly been challenged, and L2 learners have shifted into the focus of research into the bilingual (dis)advantage. However, the un‐ derlying conceptualisation of LAL as a progressive movement through space has remained fundamentally unchanged: In line with the PATH image scheme, the process of L2 learning is characterised as the path (Gilmore Robinson & Sorace, 2019: 268; Hansen, Macizo, Duñabeitia, Saldaña, Carreiras, Fuentes & Bajo, 2016: 4), the road (Bialystok, Peets, & Moreno, 2014: 189) or the transition (Khare, Verma, Kar, Srinivasan & Brysbaert, 2013: 729) to bilingualism. L2 learners are characterised as speakers located at some point between the beginning and the end of the path, for example when they are referred to as “bilinguals in training” (Sullivan, Janus, Moreno, Astheimer & Bialystok, 2014: 84), who are in the process of “acquiring their second language” (Kaushanskaya, Gross & Buac, 2014: 458), “becoming bilingual” (Hansen et al., 2016: 4), or “becoming proficient in an additional language” (Poarch & van Hell, 2012: 535). Within this trajectory, incipicient or early stages (Gilmore Robinson & Sorace, 2019: 259; Kalashnikova & Mattock, 2014: 114) are distinguished from later phases (Nicolay & Poncelet, 2013: 606). Foreign language teaching and especially immersion programmes are expected to help these persons to “attain the final aim of full bilingualism in due course” (Kalashnikova & Mattock, 2014: 115), and to “produce bilinguals” (Bialystok et al., 2014). L2 learners can be linguistically and cognitively more or less advanced (Carlson & Meltzoff, 2008; Gilmore Robinson & Sorace, 2019: 268; Nicolay & Poncelet, 2013: 597; Woumans, Surmont, Struys & Duyck, 2016: 68) on this path in comparison to each other and to “true” bilinguals. They may also trail behind (Kaushanskaya et al., 2014: 577) or surpass their monolingual peers (Purić, Vuksanović & Chondrogianni, 2017: 8). Poarch & van Hell (2012, cf. also Gilmore Robinson & Sorace, 2019) condensate these ideas in their model 346 Silke Jansen <?page no="347"?> of the Multilingual Language Continuum, which assigns different categories of speakers a relative position on the path to bilingualism: The unbalanced multilingual children in our study would be placed somehow in between the ‘bilingual-from-birth’ and ‘immersion-L2-learners’ levels along the con‐ tinuum: even though they share similar English language proficiency levels, the intensity of the language learning experience puts them in a more advantageous position at an earlier stage (Gilmore Robinson & Sorace, 2019: 268, our emphasis). Thus, there is a direct continuity between the DemBil corpus and the “L2 advantage debate” with regard to the PATH schema, while the focus of attention has shifted from the endpoints to the intermediate stages of the path. In all papers from the corpus on the “L2 advantage debate”, it is at least implicitly assumed (and sometimes also explicitly stated, cf. Nicolay & Poncelet, 2013: 605) that the processes of L1 acquisition and L2 learning are fundamentally similar. This is a necessary condition to claim that L2 learning can shape a speaker’s cognition in the same way as bilingual L1 acquisition: […] children who have started the acquisition of their second language at an early preschool age in the context of formal education are on their way to becoming bilingual both in terms of linguistic and cognitive abilities […] and enjoy some of the cognitive advantages that have been previously attributed only to children who have acquired two languages from birth (Kalashnikova & Mattock, 2014: 122). The idea that L2 learning modifies speakers’ cognitive structures, so that they will become more similar to a bilingual not only linguistically but also cognitively, is even more strongly stated in the FORCE schema, which we will discuss in the next section. 4 The FORCE schema: Learning a language is being exposed to an active substance 4.1 The FORCE schema in the “classical” debate on the bilingual advantage According to another structural metaphor frequently used in the DemBil corpus, speakers acquire languages if and because they are “exposed” to them. Within this metaphorical frame, language appears to be a substance that acts upon people’s minds, bringing it to its designated condition. The verb expose (726) and the noun exposure (1849) are the most emblematic and frequent surface metaphors that instantiate the exposure metaphor, together with immerse (129), immersion (991) and emergent bilinguals (5). The EXPOSURE metaphor 347 The path to bilingualism, a road to better cognitive performance? <?page no="348"?> construes speakers as “blanks” that are plunged into a liquid (e.g. a language) and emerge from it with a new property (being speakers of this language), just as it happens in industrial production processes. An analysis of the verbs and adjectives that most frequently combine with immersion in the DemBil corpus illustrate these conceptualisations: The collocations L2 immersion (102) and language immersion (68) are on position 1 and 2 in the frequency breakdown of the query “noun + immersion”, and together make up 92.63 % of this collocational type. The same collocates are most frequently used in combination with exposure (language exposure (211), pos. 1; L2 exposure (53), pos. 2). In collocations of the type “adjective + immersion”, language names (cf. French immersion (116), pos. 1; English immersion (19), pos. 4; Spanish immersion (17), pos. 6; cf. also L2 immersion (18), pos. 5) and adjectives of number and degree (cf. partial immersion (25), pos. 2, dual immersion (23), pos. 3, total immersion (15), pos. 7, bilingual immersion (9), pos. 8, full immersion (6), pos. 9) are most frequently used. This also applies, although to a lesser degree, to adjective collocations with exposure (cf. bilingual exposure (115), pos. 1; English exposure (84), pos. 2). These results indicate that different languages are conceived of as distinct substances (according to the understanding of languages as bound entities, cf. section 3.1.), in which speakers are immersed to varying degrees. In line with this, learners and their minds are described as being sensitive (66) to the effects of language exposure. Building on the ontological metaphor LANGUAGE IS A (LIQUID) SUB‐ STANCE , which is all-pervading both in academic discourse and in everyday speech (cf. section 2.1. and Alghbban et al., 2015: 562; Coffey, 2015: 506), the EXPOSURE metaphor relies on the capacity of particular substances to act upon sensitive surfaces. On an abstract level, it is an instantiation of the FORCE image schema, which describes that a force operates on some target, making it move or change (cf. Johnson, 1987: 43-44; Peña Cervel, 1999: 189-190). The EXPOSURE metaphor seems to be culturally linked to the folk conceptual schema LEARNING IS A PROCESS , which is widespread in the English-speaking world and “intuitively seem[s] to be related to chemical & industrial “processes” which change materials into something else, such as into a finished product” (Berendt, 2008: 78). The title of Ellen Bialystok’s article “Producing bilinguals through immersion education” (2014) is a nice illustration of this idea. In analogy to chemical processes in production that transform raw materials into something more useful and of greater value, it is assumed that exposure to language causes durable changes in brain structures and processes, which are interpreted in terms of learning (cf. for example “exposure to, and hence learning 348 Silke Jansen <?page no="349"?> about, the specific properties of their two native languages”, Sebastián-Gallés, Albareda-Castellot, Weikum & Werker, 2012: 995). More specifically, language learning is conceived of as a “curing” or “hardening” process by which language structures consolidate (2) or become established (36) in the brain, so that they cannot be changed anymore - just like a photo which, once processed, remains fixed and insensitive to light. This metaphor seems to be related not only to the popular belief according to which a second language should not be learned before the first one is firmly established, but also to academic theories about a sensitive (3) or critical period (48) in language acquisition, as well as the idea that the L1 remains stable and unaffected by L2 learning, which is a general but implicit assumption in research (cf. Kroll, Bogulski & McClain, 2012; Kroll, Dussias, Bogulski & Kroff, 2012 for a review and discussion). It is also highly compatible with the reification view of languages as fix entities. Just as, on an abstract level, “the FORCE image schema calls for the PATH schema for its development and understanding” (Peña Cervel, 1999: 189), the EXPOSURE and the PATH metaphor of LAL draw on a shared underlying time structure, because the change of state triggered by the force has both a directional and a temporal dimension. In our data, temporal expressions belonging to the PATH image scheme are systematically used in combination or in close proximity with exposure and expose, suggesting that exposure to language takes place in a linear timeline that coincides with the acquisition path. Compare the following example for illustration: To establish bilingual norms, it must be considered that a child may begin to learn two languages simultaneously from birth, or may be exposed to them consecutively […]. Whereas the simultaneous learner is initially exposed to two languages, the consecutive or sequential bilingual starts out with one language (L1), but adds another language (L2) after acquiring the fundamental structures of the first language within roughly the first 3 years of life […] ( Junker & Stockman, 2002: 382, our emphasis). As a consequence, exposure to language is considered to be the motor (or FORCE ) that drives learners forward on the acquisition path: Importantly, an EARLY age of first language exposure is considered to be essential in order for children to achieve each of these language milestones on the typical (healthy) developmental time course described above […] (Kovelman et al., 2008: 204). The interaction between the EXPOSURE metaphor and the PATH schema is further illustrated by cross-metaphorical collocations or other lexical combi‐ nations that contain surface expressions from both source domains (cf. early immersion (5), late immersion (5), duration of exposure (4), length of exposure 349 The path to bilingualism, a road to better cognitive performance? <?page no="350"?> (28), prolonged exposure (13), simultaneous exposure (8), continuous exposure (8), current exposure (6), initial exposure (6), prior exposure (5), short exposure (1) etc.). These kinds of metaphorically “hybrid” expressions occupy relatively high ranks in the frequency breakdown for collocations with exposure (cf. early exposure (45), pos. 3; previous exposure (20), pos. 6; first exposure (19), pos. 7 among adjective combinations with exposure; onset of exposure (15), pos. 4, age of exposure (9), pos. 6 among prepositional noun-noun combinations with exposure). Just as the acquisition path, the exposure timeline is mapped onto the course of speakers’ lives (cf. lifelong exposure (3), exposure from birth (5), exposure from age 3 (1), history of language exposure since birth (1)). Given that language is metaphorically construed as a continuous and steady force that acts upon humans, learners accumulate linguistic influence as they travel forward on the path, so that “timing of development corresponds to level of exposure” (Mueller Gathercole, Thomas, Jones, Guasch, Young & Hughes, 2010: 621). In line with this, language exposure is measured quantitatively, either in terms of absolute time of exposure, or intensity (i.e. language use per time unit). This is illustrated by the prepositional noun-noun combinations with exposure used in the DemBil corpus, which can virtually all be related to quantity, time length or intensity (cf. amount of exposure (63), pos. 1; length of exposure (28), pos. 2; intensity of exposure (27), pos. 3; years of exposure (10), pos. 5, degree of exposure (9), pos. 7; rate of exposure (7), pos. 8; amounts of exposure (5), pos. 9, etc.). Frequent adjective-noun combinations point in the same direction (cf. daily exposure (31), pos. 4; intensive exposure (22), pos. 5; minimal exposure (18), pos. 8; regular exposure (17), pos. 9, equal exposure (15), pos. 10, limited exposure (15), pos. 11; extensive exposure (12), pos. 12; prolonged exposure (12), pos. 13, cumulative exposure (8), pos. 16, etc.), especially because comparative forms are quite frequent (cf. less exposure (19), more exposure (12), longer exposure (3), etc.). The categorisation of speakers as monolingual or bilingual according to the amount or intensity of exposure, a relatively common procedure in the studies, reflects the assumption that a certain amount of exposure will automatically lead to language acquisition. Compare, for example, the following statements in which accumulation of linguistic knowledge appears to be a necessary consequence of language exposure: As a general principle, the more frequently and consistently one is exposed to a language, the better is one’s knowledge of that language (Kay-Raining Bird, Cleave, Trudeau, Thordardottir, Sutton & Thorpe, 2005: 188). All of this suggests that the authors assume a direct, straight-line relationship to exist between advance of time, the amount of language exposure and the 350 Silke Jansen <?page no="351"?> progress on the LAL path. The inherent strength of the force that acts upon the learner depends on the cumulative effect of overall amount, intensity and quality. Just as the effects of a chemical substance can be expected to be stronger with longer exposure and higher dosing, LAL will be more successful with more prolonged and intensive contact to language. In perfect alignment with this chemical metaphor, Bialystok (e.g. 2017: 234, 250) metaphorically describes the consequences of bilingualism on individuals as dose-related effects. In addition to quantity and intensity, quality is another characteristic that relates to language exposure. Although quality is not explicitly defined, it seems to be implicitly understood as closeness to the norm of the monolingual (prob‐ ably educated) native speaker - another example of an implicit monolingual norm (and probably an orientation towards standard language) that pervades this line of research: One factor affecting possible input quality is whether individuals are exposed from native or non-native speakers. It is often the case that the bilingual children are exposed to input from both, and predominant exposure to non-native speakers might not be as effective a tool to support language acquisition (Gilmore Robinson & Sorace, 2019: 256). More significantly perhaps, the way how the FORCE schema is applied in the texts analysed here suggests that monolingual cognitive and linguistic development is considered to be the normal case, which again reproduces monolingual language myths. According to Johnson (1987: 45 ff; cf. also Peña Cervel, 1999: 190 ff), the FORCE image schema has different forms of appearance (e.g. COMPULSION, BLOCKAGE, ATTRACTION , etc.), depending on the nature of the force and its effects. Within the conflation of the PATH and FORCE image schemas, the effect of exposure to more than one language is conceived of as a deviation from the expectable norm - that means, the monolingual path: Broadly, these findings add to the growing body of empirical evidence showing that early experiences, including multiple language exposure, dramatically influence cognitive trajectories […] (Brito & Barr, 2014: 1161, our emphasis). As reflected in this exemplary quotation, the impact of bilingualism on cognition is understood in terms of the DIVERSION type of the FORCE image schema, according to which “a force vector is diverted as the result of the causal interaction of two or more vectors” ( Johnson, 1987: 46; cf. also Peña Cervel, 1999: 200). In the DemBil corpus, this schema is instantiated by frequent transitive constructions with verbs that refer to changes of state induced by an external 351 The path to bilingualism, a road to better cognitive performance? <?page no="352"?> 4 In the queries that led to these results, wildcards were used in order to find relevant expressions where the noun and the verb form were separated by other elements, e. g. bilingualism clearly affects, bilingualism inevitably impacts, exposure to another language enhances, etc. It is important to note that the list of expressions analysed here is far from being exhaustive, due to the large range of noun phrases that can function as synonyms of bilingualism. As a systematic analysis of these expressions goes beyond the scope of this paper, not all potentially relevant expressions could be taken into account (cf. knowing more than one language, bilingual experience, the daily use of two languages from childhood, to mention just a few examples from the corpus) were not taken into account. Nevertheless, we think that these examples suffice to give an impression of the importance of the DIVERSION schema in the texts. force. Such verbs may take nouns such as bilingualism or exposure (to two/ dif‐ ferent/ several languages  4 ) as a subject or different kinds of noun phrases, which refer to cognitive structures or functions (e.g. brain functions/ areas, cognitive functions/ abilities/ performance/ control, etc.), as an object. Verbal lexemes that appear in these kinds of constructions are affect (129), impact (25), modify (11), alter (8), shape (5), and change (3) (cf. also the expressions have an effect (9), have an influence (2), have consequences (3)). According to a frequency breakdown of all finite verb forms that occur with bilingualism and exposure, affects is even the third most frequent item, directly after is and has. Other verbal lexemes that occur in these kinds of constructions combine change of state with (generally positive) evaluation, among them enhance (64), boost (8) or improve (18). The same verbs also occur in passive constructions with bilingualism, language exposure or related noun phrases as by-agents (e.g. affected (71), influenced (18), caused (11), modified (8), impacted (4), altered (2) etc. by bilingualism or (bilingual / previous) exposure; enhanced (4), improved (2), boosted (1) by bilingualism / bilingual exposure. In line with the FORCE schema, these constructions present language and language exposure as an agent that changes the expected trajectory of learners, while the leaners themselves have (24), receive (2) or experience (1) exposure, or are subjected to (2) the influence of languages. Indeed, these metaphorical expressions imply that learners are passive experiencers of changes inflicted on them by an external force and have no agency in the acquisition process. It seems that brain changes happen just magically: […] the current research demonstrates that regular exposure to an additional language is alone sufficient to alter the language learning mechanism (Kuo & Anderson, 2012: 465). In summary, the path and the force schemas are all-pervasive in the DemBil corpus. Yet, they are not theory-constitutive: No direct relationship is estab‐ 352 Silke Jansen <?page no="353"?> lished between metaphorical models of the processes of learning and acquisition on the one hand, and possible cognitive differences between monolinguals and bilinguals on the other. On the contrary, these differences are explained by underlying competition between languages, according to the competition metaphorical model presented in section 1. In the following section, we will argue that to the extent that L2 leaners shift into the focus of the “bilingual advantage” debate, the PATH and the FORCE metaphors gain a more prominent role, and even become constitutive for the interpretation of data and theory building. 4.2 The FORCE schema in studies on the “bilingual advantage” in L2 learners The more prominent role that recent papers confer to the FORCE schema is most directly reflected in the definition and naming of the speaker groups under study. While the studies compiled in the DemBil corpus generally select participants based on their alleged location at opposing ends of the acquisition path, referring to them as the bilingual and the monolingual group, most of the 18 studies that represent the “L2 advantage debate” derive their criteria for group definition from the exposure metaphor. This becomes visible in labels such as (native) bilinguals vs. immersion group (Carlson & Meltzoff, 2008), monolinguals vs. immersed children (Hansen et al. 2016), monolingual group vs. immersion group (Nicolay & Poncelet, 2013), monolinguals vs. high exposure group vs. low exposure & non-intensive exposure group (Puric et al., 2017), immersed vs. non-immersed (Simonis, van der Linden, Galand, Hiligsmann & Szmalec, 2020), and monolingual kindergarten children vs. immersion kindergarten children (Woumans et al., 2016). However, some of these studies still use terms that are related to the PATH metaphor (cf. monolinguals vs. sequential bilinguals (Kalashnikova & Mattock, 2014), monolinguals vs. early bilinguals vs. late bilinguals (Luk, de Sa & Bialystok, 2011; Pelham & Abrams, 2013)). The direct confrontation of monolinguals and speakers immersed in or exposed to an L2 implicitly establishes a metonymic relationship between being immersed or exposed and being bilingual, which epitomises the core idea of the FORCE-PATH model of language acquisition: exposure is directly related to acquisition and, hence, to bilingualism (of varying degrees). This new way of categorising speakers reflects the shift in focus that has taken place in recent years within research into bilingualism and cognition, from a dichotomous to a gradual understanding of bilingualism, and from an emphasis on language production under the COMPETITION metaphor towards a focus on language learning and the use of the PATH-FORCE schema as a central 353 The path to bilingualism, a road to better cognitive performance? <?page no="354"?> metaphorical model. We may assume that the hypothesis according to which cognitive performance gradually increases with the degree of bilingualism could emerge precisely because PATH-FORCE models of LAL are so widespread and deeply entrenched in both academic and folk representations of (language) learning. All of the 18 papers that focus on possible cognitive advantages in L2 learners, with no exception, explicitly refer to the PATH-FORCE schema in central parts of the papers, particularly in the discussion. Compare the following example: The results show a continuum in which more experience in using two languages is associated with greater benefit and greater approximation to the pattern reported for bilingualism. It is an evolving system in which experience gradually and continu‐ ally modifies ability. Language-education programs are not only teaching children language, but they are also making them bilingual. The road to bilingualism is incremental, and so are the accrued advantages (Bialystok, 2014: 189, our emphasis; compare also e. g. Nicolay & Poncelet, 2013: 592; Poarch & van Hell, 2012: 539; Sullivan et al., 2014: 95, to mention just a few). This does not mean that the COMPETITION structural metaphor has disap‐ peared - on the contrary, most authors refer to it at some point of their argumentation, often combining it with the path-force schema: We predicted that all bilinguals would show comparable within-language competition as a function of phonological overlap but that native French bilinguals […] would show more cross-language competition as a function of their daily exposure to the task language (English) (Mercier, Pivneva & Titone, 2014: 91, our emphasis; cf. also Kamat, Shinde, Gaikwad & Guhilot, 2012: 308). The general reasoning behind the studies is that if bilingualism leads to enhanced cognitive functions as a result of competition between languages, and exposure to a second language leads to bilingualism, then there must be a causal link between exposure and cognitive performance. Within this reasoning, the PATH-FORCE schema has a much more prominent role than in previous studies, because of its explanatory function for the specific problem addressed in the papers. Interestingly, the EXPOSURE metaphor has generated new perspectives on bilingualism and cognition, which have begun to alter the format and presentation of the studies. In line with the all-pervading COMPETITION metaphor, most of the studies in the DemBil corpus are structured like sports competitions, where monolinguals and bilinguals engage in a “race” for the best scores in cognitive tasks (generally in terms of reaction times in Simon 354 Silke Jansen <?page no="355"?> tasks, Flanker tasks, etc., cf. Jansen et al., 2021). In contrast, recent studies sometimes resemble clinical trials of pharmaceuticals, following the metaphor L2 LEARNING IS A MEDICAL TREATMENT , which is itself an instantiation of the metaphor LAL IS EXPOSURE TO AN ACTIVE SUBSTANCE . Bubbico, Chiacchiaretta, Parenti, di Marco, Panara, Sepede, Ferretti & Perrucci’s (2019) study on the effects of second language learning on brain plasticity in elderly participants is a telling example of this: Not only do the authors label their study as an intervention study and their comparison groups as intervention group and control group, following the example of clinical studies, but they also discuss “exposure” to an L2 in terms of intervention and treatment. Their considerations culminate in the following statement, which explicitly equates a second language experience with a medical treatment: These results should consider that a second language learning program, even late in life, can be considered a non-pharmacological treatment able to counteract cognitive aging along with the onset of dementia. Learning a second language is a powerful tool that can be part of a healthy lifestyle program that can preserve brain plasticity in aging individuals (Bubbico et al., 2019: 10). It seems, thus, that the authors take the SUBSTANCE metaphor literally - a common fallacy in metaphor use in science, which we will address in the following section. 5 Merits and fallacies of learning metaphors in research into bilingualism, language learning and cognition With its emphasis on the PATH and the EXPOSURE metaphor, the discourse analysed here is not an exception within systems of knowledge production in Western and, to some extent, other societies. Linguistic studies have shown that the PATH metaphor is, cross-culturally, among the most widespread metaphors of learning. Likewise, the EXPOSURE metaphor (as an instantiation of the FORCE schema) has circulated within the English-speaking world since the 19 th century Industrial Revolution (cf. Berendt, 2008: 78). Originally representing folk conceptualisations of learning processes, both the PATH and the EXPO‐ SURE metaphor became part of the common stock of assumptions that has shaped scientific practice in Cognitive Sciences (especially Developmental Psy‐ chology, Cognitive Psychology and Linguistics). Outstanding figures in these areas, among them Piaget with his PATH -based stage theory and Chomsky with his EXPOSURE -based conception of a “language acquisition device”, further contributed to the implementation of these metaphorical models in academic 355 The path to bilingualism, a road to better cognitive performance? <?page no="356"?> 5 Cf. 101.58 instances per million words in the DemBil corpus between 1923 and 1961, against 388.67 between 1962 and 1999, and 792.27 between 1962 and 2018. Note also that Chomsky himself uses the term when he advances the hypothesis of an innate Language Acquisition Device (1964: 42). discourse. For example, it is certainly not a coincidence that the use of the terms exposure and expose increases sharply in the DemBil corpus from the early 1960s onwards - just at the time when Chomsky formulated his influential ideas on innate linguistic structures that emerge automatically when speakers receive language exposure. 5 From a current standpoint, Chomsky’s theories of language acquisition are generally considered to be disproved (cf. also Kany & Schöler, 2014: 495-496). Modern approaches, for example in the context of Constructivism and Connectivism, have highlighted that if it is true that language acquisition is fundamentally processual in character, it cannot be described as a linear, steady, goal-oriented and inter-individually uniform progression. Learners often make detours and setbacks and go through phases of stagnation as well as spurts (cf. Kany & Schöler, 2014: 495 and Klann-Delius (2016) on the representation of different aspects of language acquisition progress as a u-shaped rather than a linear curve). Connectivist theories understand language acquisition as a plu‐ rality of both parallel and overlapping processes, which constantly contribute to a dynamical reorganisation of the whole system (cf. Klann-Delius, 2016: 113). Consequently, learning a new language has a transformative potential for the totality of the speaker’s repertoire - even for the L1, which is far from being “stabilised” and does not remain unaffected from further language learning (cf. Kroll, Bogulski & McClain, 2012; Kroll, Dussias, Bogulski & Kroff, 2012). This is incompatible with some important entailments of the PATH schema, among them the idea that there is a particular and independent track for every language in bilingual acquisition, and that the acquisition process will reach an endpoint where no further growth or development is possible. As originally acquired languages can be completely or partially lost, language acquisition is not necessarily a unidirectional process either. In addition, constructivist approaches emphasise that, contrary to what the EXPOSURE metaphor suggests, learners are not passive experiencers whose brains blindly react to linguistic stimuli but take an active role in language learning: they do not “acquire” or “receive” knowledge, but actively construct hypotheses about target language structures, employing different kinds of problem-solving strategies. Furthermore, sociocultural theories of language highlight that languages are learned in social situations and that speakers play an active part in shaping the interactions in which learning takes place. 356 Silke Jansen <?page no="357"?> Thus, it seems that the metaphors that prevail in current debates on cognitive effects of language learning reflect a prescientific understanding of LAL (or at best an older stage of research) and transport a somewhat reductionist view of what language acquisition is. If reduction of complexity is not necessarily a problem in scientific modelling, the question remains if implicit assumptions related to the PATH and to the EXPOSURE metaphors may have misguided researchers in their process of knowledge construction. Obviously, the idea that mental performance continuously increases together with bilingualism is an entailment of the PATH-FORCE schema of cognitive and linguistic development. As long as research focused only onto the endpoints of the language acquisition path, this possible entailment was not relevant. However, when researchers became interested in L2 learners, it was easily available as an explanation for data that hinted at a correlation between language skills and cognitive performance. This is what Bialystok explains in one of her review articles: Something about bilingual environments or bilingual experience accelerates the development and maintenance of attention. Moreover, because it is conceptualized as a continuum rather than as a discrete process, it is easy to imagine a quantitative relation between the intensity of experience and nature of the outcomes; in a discrete componential model, different degrees or types of experience may lead to qualitatively different outcomes and would have difficulty accounting for the dose-related effects found in the literature […] (Bialystok, 2017: 249-250). Precisely because so far unexploited entailments of established metaphors can provide hypotheses that immediately seem to make sense, researchers may forget that their models are indeed metaphorical and, consequently, overstep the boundaries of empirically supported interpretations. The PATH-FORCE schema views L2 acquisition as a process that, fueled by L2 exposure, advances in a linear fashion towards the goal of bilingualism, because the strength and intensity of a force acting upon an entity determines the strength, and intensity of the impact is part of our pervasive everyday experience. If (as all the authors of the 18 texts on possible advantages in L2 learners do) we accept the idea that bilinguals are cognitive high performers in comparison to monolinguals, nothing seems to be more natural than assuming that more powerful exposure brings speakers closer to the desired goal of being bilingual and, hence, to better cognitive performance. Apart from the fact that this explanation is descriptive rather than explana‐ tory, the PATH-FORCE model seems to be taken for granted (or maybe even to be literally true) in the studies analysed here. Even authors who were not able 357 The path to bilingualism, a road to better cognitive performance? <?page no="358"?> to find cognitive differences between monolinguals and L2 learners rely on it when they argue that cognitive advantages will only become visible at particular points or stages within the process of language acquisition - either because they would emerge abruptly from a certain degree of bilingualism onwards (Carlson & Meltzoff, 2008: 16; Kaushanskaya et al., 2014: 576), or because they would manifest themselves only “during the first phases of foreign-language learning, [when] specific control processes may be more strongly solicited […] due to lack of automaticity in language use than in later stages” (Simonis et al., 2020: 367). In other words, when no cognitive differences are observed between monolingual and “exposed” speakers, it is assumed that the “exposed” participants were assessed at a point on the path where those differences were simply not visible, for whatever reason. Thus, any kind of experimental outcomes are always interpreted as evidence for the PATH-FORCE model, either in its continuous or in its discrete-componential guise (cf. Bialystok, 2017). This is a result of circular reasoning, probably promoted by lack of metaphor awareness. 6 Conclusion For more than a hundred years, psychologists have investigated the impact of bilingualism on cognition, with varying results. Since its beginnings, the debate around the so-called bilingual (dis)advantage has largely drawn on dichotomous constructions, both of monolingual and bilingual speakers, and of the two allegedly competing languages in the bilingual brain. In recent times, many researchers have challenged the categorical understanding of bilingualism, which is ideologically derived from the glorification of the “native speaker” as a model speaker. Arguing that learning a second language is becoming bilingual (at least to a certain extent), they postulate that L2 learners’ brains will become more and more similar to “native” bilinguals’ brains as their experience with an additional language increases - with all the cognitive benefits that bilingualism purportedly entails. Obviously, this assumption is based on a general acceptance of the “bilingual advantage” hypothesis as formulated in the “classical” debate, and reflects an entailment of the PATH-FORCE schema of linguistic and cognitive development, which has been available in the discourse at least since the 1960s. Engaging in a classic case of circular reasoning, the authors use the PATH-FORCE schema to explain both positive and negative results of psychometric experiments that aim at finding cognitive differences between L2 learners and monolinguals and / or “native” bilinguals. Because of its explanatory function regarding the relationship between the process of language learning and cognitive transformations, the PATH-FORCE schema 358 Silke Jansen <?page no="359"?> is much more prominent in the papers that are contained in our corpus than the CONTENTION metaphor. Still, all authors refer and adhere to the CONTENTION model of bilingual language production. Both the PATH and the FORCE schema are deeply entrenched in Western academic and non-academic thought. However, they are at odds with modern theories of LAL in many respects: The PATH schema constructs learning “as purely sequential and accumulative” (Coffey, 2015: 502), disregarding non-linear processes of learning as well as the social and emotional context of language learning as a personal experience. The EXPOSURE metaphor, a specific instan‐ tiation of the FORCE schema, suggests that language learning is something that simply happens, without any active involvement of the learner. None of the metaphorical models provides any details relating to concrete processes of LAL, while both of them highlight the gradual alignment with a stable, predefined norm. In contrast to what the rejection of the dichotomy between monolinguals and bilinguals suggests, the papers that focus on L2 learners in their capacity as “partial” bilinguals do not overcome traditional language myths, but rather contribute to reproduce them: they reify languages as objects with fixed, clear-cut boundaries that coexist and develop independently from each other in the bilingual mind and still define “full” bilingualism as double monolingualism, implicitly setting the (educated) native speaker as a norm. Finally, and more noteworthy perhaps, they carry on and perpetuate the spirit of neoliberal multiculturalism that has characterised research on the bilingual advantage for the last 30 years, which finds its most visible expression in the CONTENTION metaphor (cf. Jansen, 2021; Jansen et al., 2021). Within such a framework, multilingualism is propagandised mainly for competitive and economic reasons in a globalised world, and bilinguals are construed as high performing, healthy individuals that act as ideal economic subjects. In line with these ideas, research on cognitive benefits of L2 learning presents language learning as a PATH and a FORCE that does not only lead to bilingualism, but also to high performance and mental health. Further echoes of this ideological stance can be detected in recent papers on cognitive advantages in L2 learners, when authors highlight the relevance of their findings for educational and cultural politics, for example regarding immersion education (cf. Nicolay & Poncelet, 2013: 605), life long L2 learning (Sullivan et al., 2014: 96), language conservation and revitalisation policies (Garraffa, Obregón, O’Rouke & Sorace, 2020, especially page 10), as well as for public health (Bubbico et al., 2019). In Cognitive Science, as in other academic and non-academic contexts, the production of knowledge is deeply entrenched with overarching explanatory paradigms, social circumstances and 359 The path to bilingualism, a road to better cognitive performance? <?page no="360"?> zeitgeist, which all act as the cement that allows researchers to narrate a coherent and appealing story as to how their data on L2 acquisition and cognitive performance should be interpreted. References Primary sources: Corpus texts DemBilCorpus DeMultCorpus [https: / / corpora.linguistik.uni-erlangen.de/ cqpweb/ demult_v5/ ] Bialystok, E. 2017. The bilingual adaptation: How minds accommodate experience. Psychological Bulletin, 143 (3), 233-262. Bialystok, E., Craik, F.I.M., Green, D.W & Gollan, T.H. 2009. Bilingual minds. Psychological Science in the Public Interest, 10, 89-129. Bialystok, E., Peets, K.F. & Moreno, S. 2014. Producing bilinguals through immersion education: Development of metalinguistic awareness. Applied Psycholinguistics, 35, 177-191. Brito, N. & Barr, R. 2014. Flexible memory retrieval in bilingual 6-month-old infants. Developmental Psychobiology, 56, 1156-1163. Engel de Abreu, P.M.J., Cruz-Santos, A., Tourinho, C.J., Martin, R. & Bialystok, E. 2012. Bilingualism enriches the poor: enhanced cognitive control in low-income minority children, Psychological Science, 23, 1364-1371. Junker, D.A. & Stockman, I.J. 2002. Expressive vocabulary of German-English bilingual toddlers. American Journal of Speech and Language Pathology, 11, 381-394. Kay-Raining Bird, E., Cleave, P., Trudeau, N., Thordardottir, E., Sutton, A. & Thorpe, A. 2005. The language abilities of bilingual children with Down syndrome. In: American Journal of Speech and Language Pathology, 14, 187-199. Kovelman, I., Baker, S. A., Petitto, L-A. 2008. Age of first bilingual language exposure as a new window into bilingual reading development. Bilingualism, 11, 203-223. Kuo, L-J. & Anderson, R.C. 2012. Effects of early bilingualism on learning phonological regularities in a new language. Journal of Experimental Child Psychology, 111, 455-467. Marsch, S. 2009. Metaphern des Lehrens und Lernens. Vom Denken, Reden und Handeln bei Biologielehrern. Berlin: Freie Universität, Diss. Mercier, J., Pivneva, I. & Titone, D. 2014. Individual differences in inhibitory control relate to bilingual spoken word processing. Bilingualism, 17, 89-117. Mueller Gathercole, V.C., Thomas, E.M., Jones, L., Guasch, N.V., Young, N. & Hughes, E.K. 2010. Cognitive effects of bilingualism: digging deeper for the contributions of language dominance, linguistic knowledge, socio-economic status and cognitive abilities. International Journal of Bilingual Education and Bilingualism, 13, 617-664. 360 Silke Jansen <?page no="361"?> Sebastián-Gallés, N., Albareda-Castellot, B., Weikum, W.M. & Werker, J.F. 2012. A bilingual advantage in visual language discrimination in infancy. Psychological Science, 23, 994-999. Smith, F. 1923. Bilingualism and mental development. British Journal of Psychology. General Section, 13, 271-282. Sorace, A. 2007. The more, the merrier: facts and beliefs about the bilingual mind. In: Della Sala, S. (Ed.), Tall Tales about the Mind and Brain: Separating Fact from Fiction. Oxford: Oxford University Press. 193-203. Weigmann, K. 2014. The Babylonian benefit. Neurological research shows that being bilingual enhances mental performance and may protect from Alzheimer’s disease. EMBO reports, 15, 1015-1018. “L2 advantage corpus” Bialystok, E. 2014. Producing bilinguals through immersion education: Development of metalinguistic awareness. Applied Psycholinguistics, 35, 177-191. Bubbico, G., Chiacchiaretta, P., Parenti, M., di Marco, M., Panara, V., Sepede, G., Ferretti, A. & Perrucci, M.G. 2019. Effects of second language learning on the plastic aging brain: Functional connectivity, cognitive decline, and reorganization, Frontiers in Neuroscience, 13, 1-13. Carlson, S.M. & Meltzoff, A.N. 2008. Bilingual experience and executive functioning in young children, Developmental Science, 11, 282-298. Garraffa, M., Obregón, M., O’Rouke, B. & Sorace, A. 2020. Language and cognition in Gaelic-English young adult bilingual speakers: A positive effect of school immersion program on attentional and grammatical skills. Frontiers in Psychology, 11, 1-12. Gilmore Robinson, M. & Sorace, A. 2019. The influence of collaborative language learning on cognitive control in unbalanced multilingual migrant children. European Journal of Psychological Education, 34, 255-272. Hansen, L.B., Macizo, P., Duñabeitia, J.A., Saldaña, D., Carreiras, M., Fuentes, L.J. & Bajo, M.T. 2016. Emergent bilingualism and working memory in emergent bilinguals, Language Learning, 66, 51-75. Kalashnikova, M. & Mattock, K. 2014. Maturation of executive functioning skills in early sequential bilingualism. International Journal of Bilingual Education and Bilingualism, 17, 111-123. Kaushanskaya, M., Gross, M. & Buac, M. 2014. Effects of classroom bilinguals on task-shifting, verbal memory, and word learning in children. Developmental Science, 17, 564-583. Khare, V., Verma, A., Kar, B., Srinivasan, N. & Brysbaert, M. 2013. Bilingualism and the increased attentional blink effect: Evidence that the difference between bilinguals 361 The path to bilingualism, a road to better cognitive performance? <?page no="362"?> and monolinguals generalizes to different levels of second language proficiency. Psychological Research, 77, 728-737. Luk, G., de Sa, E. & Bialystok, E. 2011. Is there a relation between onset age of bilingualism and enhancement of cognitive control? Bilingualism: Language and Cognition, 14, 588-595. Nicolay, A-C. & Poncelet, M. 2013. Cognitive advantage in children enrolled in a second-language immersion elementary school program for three years. Bilingualism: Language and Cognition, 16, 597-607. Pelham, S.D. & Abrams, L. 2013. Cognitive advantages and disadvantages in early and late bilinguals. Journal of Experimental Psychology: Learning, Memory, and Cognition, 40, 313-325. Poarch, G.J. & van Hell, J.G. 2012. Executive functions and inhibitory control in multi‐ lingual children: Evidence from second-language learners, bilinguals, and trilinguals. Journal of Experimental Child Psychology, 113, 535-551. Puric, D., Vuksanović, J. & Chondrogianni, V. 2017. Cognitive advantages of immersion education after 1 year: Effects of amount of exposure. Journal of Experimental Child Psychology, 159, 296-309. Simonis, M., Van der Linden, L., Galand, B., Hiligsmann, P. & Szmalec, A. 2020. Executive control performance and foreign-language proficiency associated with immersion education in French-speaking Belgium. Bilingualism: Language and Cognition, 23, 355-370. Sullivan, M.D., Janus, M., Moreno, S., Astheimer, L. & Bialystok, E. 2014. Early stage second-language learning improves executive control: Evidence from ERP. Brain & Language, 139, 84-98. Woumans, E., Surmont, J., Struys, E. & Duyck, W. 2016. The longitudinal effect of bilingual immersion schooling on cognitive control and intelligence. Language Learning, 66, 76-91. Secondary sources Alghbban, M.I, Salamh, S.B. & Maalej, Z. 2015. Metaphoric modeling of foreign language teaching and learning, with special reference to teaching philosophy statements. Applied Linguistics, 38, 559-580. Berendt, E.A. (Ed.) 2008. Metaphors for Learning. Cross-cultural Perspectives. Amsterdam & Philadelphia, PA: John Benjamins. Brown, T. L. 2003. Making Truth. Metaphor in Science. Urbana & Chicago, IL: University of Illinois Press. Charteris-Black, J. 2018. Competition metaphors and ideology: life as a race. In: Wodak, R. & Forchtner, B. (Eds.), The Routledge Handbook of Language and Politics. New York, NY & London: Routledge. 202-217. 362 Silke Jansen <?page no="363"?> Chomsky, N. 1964. Current Issues in Linguistic Theory. The Hague: Mouton. Coffey, S. 2015. Reframing teachers’ language knowledge through metaphor analysis of language portraits. The Modern Language Journal, 99, 500-514. Cortazzi, M. & Jin, L. 1999. Bridges to learning. Metaphors of teaching, learning and language. In: Cameron, L. & Low, G. (Eds.), Researching and Applying Metaphor. Cambridge: Cambridge University Press. 149-176. Drogosz, A. 2018. The metaphor of struggle in communication of the evolutionary theory. Styles of Communication, 10, 36-49. Goschler, J. 2007. Metaphors in cognitive and neurosciences. Which impact have meta‐ phors on scientific theories and models? Metaphorik.de 12. www.metaphorik.de & sites & www.metaphorik.de & files & journal-pdf & 12_2007_goschler.pdf. May 30, 2021. Hampe, B. 2005. Image schemas in Cognitive Linguistics: Introduction. In: Hampe, B. (Ed.), From Perception to Meaning. Image Schemas in Cognitive Linguistics. Mouton & De Gruyter. 1-12. Jäkel, O. 2003. Wie Metaphern Wissen schaffen. Hamburg: Kovac. Jansen, S. 2021. Are bilinguals brighter than monolinguals? The bilingual (dis)advantage debate in light of changing attitudes and policies towards bilingualism. In: Becker, L., del Valle, J. & Knauer, G. (Eds.), La glotopolítica: Un desafío para la renovación de la lingüística hispánica. Frankfurt: Peter Lang. Jansen, S., Higuera del Moral, S., Barzen, J., Reimann, P. & Opolka, M. 2021. Demys‐ tifying Biligualism: How Metaphor guides Science towards Mythification. London: Palgrave/ Macmillan. Johnson, M. 1987. The Body in the Mind. The Bodily Basis of Meaning, Imagination, and Reason. Chicago, IL & London: University of Chicago Press. Kamat, R.K., Shinde, S.A., Gaikwad, P.K. & Guhilot, H. 2012. Harnessing VLSI System Design with EDA Tools. Heidelberg: Springer. Kany, W. & Schöler, H. 2014. Skinner und Chomsky: Zwei Protagonisten der Spracher‐ werbsforschung. In: Ahnert, L. (Ed.), Theorien in der Entwicklungspsychologie. Berlin: Springer. 486-501. Klann-Delius, G. 2016. Spracherwerb. Eine Einführung. Stuttgart: Metzler. Kroll, J.F., Bogulski, C.A. & McClain, R. 2012. Psycholinguistic perspectives on second language learning and bilingualism. In: Linguistic Approaches to Bilingualism, 2 (1), 1-24. Kroll, J.F., Dussias, P.E., Bogulski, C.A. & Kroff, J.R.V. 2012. Juggling two languages in one mind. Psychology of Learning and Motivation, 56, 229-262. Kubota, R. 2016. The multi & plural turn, postcolonial theory, and neoliberal multicul‐ turalism: Complicities and implications for applied linguistics. Applied Linguistics, 37, 474-494. 363 The path to bilingualism, a road to better cognitive performance? <?page no="364"?> Lakoff, G. & Johnson, M. 1980. Metaphors We Live By. Chicago, IL & London: University of Chicago Press. Peña Cervel, S. 1999. Subsidiarity relationships between image-schemas: An approach to the force schema. Journal of English Studies, 1 (1), 187-207. Radwańska-Williams, J. 2008. The “native speaker” as a metaphorical construct. In: Berendt, E.A. (Ed.), Metaphors for Learning. Cross-cultural Perspectives. Amsterdam & Philadelphia, PA: John Benjamins. 139-158. Smith, B. 2013-2014. The journey metaphor’s entailments for framing learning. The Journal of the Assembly for Expanded Perspectives on Learning, 19, 23-37. Watts, R.J. 2003. Language Myths and the History of English. Oxford: Oxford University Press. 364 Silke Jansen <?page no="365"?> Contributors Angelovska, Tanja is Associate Professor of English Linguistics and Foreign Language Teaching at the University of Salzburg (Austria), an associate member of the Centre for Research and Enterprise in Language (University of Green‐ wich, UK), the Language Acquisition, Multilingualism, & Cognition Laboratory (Wilfrid Laurier University, Canada), and Brain & Cognition Lab (Texas A&M University, USA). Her research interests include L2/ L3 input processing by children and adults, second and third (or additional) language acquisition in children and adults using behavioural and (neuro-) cognitive measures, and the effects of bilingualism on cognitive functions in relation to factors, such as frequency, language dominance and proficiency. Biedinger, Nicole is Professor of Sociology and Empirical Social Research at the Catholic University of Applied Sciences Mainz (Germany) at the Department of Social Work and Social Sciences. She obtained her PhD in 2010 on the subject of “Ethnic and Social Inequality in Pre-school Education”. Her research focuses on the sociology of education and migration with a particular emphasis on the pre-school sector. Blake, Ashley is a PhD Researcher at the University of Birmingham (UK). Her research focuses on individual differences in children’s language acquis‐ ition, as part of a large-scale research project led by Ewa Dąbrowska, Frie‐ drich-Alexander-Universität Erlangen-Nürnberg (Germany). Her research com‐ bines insights from skill acquisition theory and the usage-based model of language acquisition to investigate how the speed of automatization predicts differences in children’s linguistic ability. Blumenthal, Yvonne is Lecturer at the Universities of Rostock and Greifswald (Germany). She completed her teacher training after studying special education with a focus on children with learning difficulties and social emotional disorders. Her research interests include diversity in schools, especially aspects of partic‐ ipation (e.g. social inclusion and relationships) for students with special educa‐ tional needs. Furthermore, she is interested in inclusive and evidence-based school development and databased decision-making processes to support the academic, language and social-emotional development of students. <?page no="366"?> Chudaske, Jana is a psychologist. She has been working at the Institute of Educational Science at the University of Hildesheim (Germany) since 2005 and completed her PhD thesis entitled “Sprache, Migration und schulfachliche Leistung. Einfluss sprachlicher Kompetenz auf Lese-, Rechtschreib- und Mathematikleistungen” in 2011. Her research interests in the field of teaching meth‐ odology include self-directed and cooperative learning. She also provides sub‐ ject-specific study counselling and subject coordination for student mobility in the programme Pedagogy of Teacher Education at the University of Hildesheim. Dąbrowska, Ewa is an Alexander von Humboldt Professor of Language and Cognition at the Friedrich-Alexander-Universität Erlangen-Nürnberg and Professor of Linguistics at the University of Birmingham. Previously she worked at the universities of Northumbria, Sheffield, Sussex, Glasgow and Gdańsk, and served as President of the UK Cognitive Linguistics Association and editor-in-chief of Cognitive Linguistics. Her research focuses on individual differences in linguistic knowledge in both native and non-native speakers, the mental status of linguistic units and generalisations, and the role of lexically specific units in language acquisition and processing. Dallinger, Sara is currently working as a high school teacher in the federal state of Baden-Württemberg (Germany). She obtained her PhD at the University of Education Ludwigsburg (Germany) in 2015. Her project dealt with selec‐ tional effects, achievement development, and the role of languages in bilingual (German-English) history teaching. Her research interests include bilingual teaching, quantitative research paradigms and teaching literature. Franceschini, Rita is Professor of Linguistics at the Free University of Bozen-Bolzano (Italy). After obtaining her PhD at the University of Zurich (Switzerland), she worked at the Universities of Bergamo (Italy), Basel (Switzer‐ land) and at the University of Saarland (Germany), where she was appointed Full Professor of Applied Romance Linguistics in 2004. She has served on the university councils of the Universities of Basel and Salzburg and, since 2021, on the council of the University of Hildesheim (Germany). Her interdisciplinary research interests include social, linguistic and neurobiological aspects of multilingualism. Glass, Cordula is Research Coordinator at the Friedrich-Alexander-Univer‐ sität Erlangen-Nürnberg (Germany). Prior to that, she studied Cultural Engi‐ neering at Otto-von-Guericke University Magdeburg (Germany), where she initially worked on linguistic and cultural aspects of knowledge management. She completed her PhD on the topic of Cognitive Linguistics at the Fried- 366 Contributors <?page no="367"?> rich-Alexander-Universität Erlangen-Nürnberg (Germany) in 2016. Other aca‐ demic stations include research at the University of Birmingham (UK) and the University of Bayreuth (Germany). Goriot, Claire is interested in (second) language development, bilingualism, and (language) education. She obtained her PhD from Radboud University Nijmegen (the Netherlands), where she investigated the cognitive and linguistic development of (functionally) monolingual pupils, pupils in early-English schools, and children growing up with Dutch and English. Her research focuses on various aspects of (language) education and bilingualism, and includes both pupils and teachers. Hopp, Holger is Professor of English Linguistics at the Technical University Braunschweig (Germany) and co-principal investigator of the project “Spra‐ chliche und kognitive Ressourcen der Mehrsprachigkeit im Englischerwerb in der Grundschule”. In his research, he investigates child and adult L2/ 3 acquisition and processing as well as heritage language acquisition and attrition. He uses several psycholinguistic methods to determine the directionality, scope and degree of cross-linguistic influence in biand multilingual speakers of different ages. Jaekel, Nils is a University Lecturer at the University of Oulu in Finland. He received his PhD in Applied Linguistics & Second Language Acquisition from the Ruhr-University Bochum, Germany. For his dissertation, he investigated lan‐ guage learning strategy use among students in Content and Language Integrated Learning (CLIL) and regular English as a Foreign Language classrooms. His research interests include content-based foreign and second language teaching and learning, early second language acquisition, and individual differences in language acquisition, particularly language learning strategies and self-efficacy in language learning. Jansen, Silke is a Romance linguist at Friedrich-Alexander-Universität Er‐ langen-Nürnberg. She specialises in sociolinguistics, language contact, and linguistic ideologies, with a focus on Latin America and the Caribbean. Among her research interests is the role of language variation and linguistic difference in processes of identity construction, social differentiation, and “Othering”, particularly regarding language ideologies and myths. Especially keen on interdisciplinary research, she has led several projects at the interface between linguistics, social sciences, and psychology, among them the project “Demys‐ tifying bilingualism”, which critically examines language myths in cognitive science. She is currently leading the project VIOLIN (“Linguistic violence against 367 Contributors <?page no="368"?> migrants in institutions”), which focuses on problematic interactions between migrants and institutional representatives. Kieseier, Teresa is a PhD researcher in the project “Sprachliche und kognitive Ressourcen der Mehrsprachigkeit im Englischerwerb in der Grundschule” at the Department of English at the University of Mannheim (Germany). Her doctoral thesis deals with the acquisition of English in a multilingual primary school context and includes linguistic and metalinguistic perspectives. In her research, she is interested in L1 and child L2/ L3 acquisition, multilingualism and phonology. Piske, Thorsten is Professor and Chair of Foreign Language Education at the Friedrich-Alexander-Universität Erlangen-Nürnberg (Germany). His research focuses on first and second language acquisition as well as on bilingual educa‐ tion. In various projects, he has examined the effectiveness of German-English bilingual programmes at kindergartens, primary schools and secondary schools. Since his post-doctoral period in the USA in the late 1990s, funded by a research grant from the National Institutes of Health, he has also participated in several studies on the linguistic development of migrants. Rogotzki, Nina is currently working as the spokesperson of the customer dia‐ logue and the works council at Schleswig-Holstein Local Transport Association (NAH-SH). She co-edited two volumes exploring the relationship between Star Trek and the sciences (2003). Her research interests include gender roles, literary criticism and translation. Rumlich, Dominik has held the Chair of English Didactics at the University of Paderborn (Germany) since 2018. Prior to that, he held a junior professorship for Teaching English at the Westfälische Wilhelms-University Münster (Germany) and represented Prof Dr Bärbel Diehr’s Chair of Psycholinguistics and Second Language Acquisition at the University of Wuppertal (Germany). He obtained his PhD on “Evaluating bilingual education in Germany: CLIL students’ general English proficiency, EFL-self-concept and interest” at the University of Duis‐ burg-Essen (Germany) in 2015. His research focuses on bilingual teaching/ CLIL, assessment and affective-motivational factors of foreign language learning. Schmidt, Katja is Lecturer at the University of Rostock (Germany). After she studied English and Russian, she completed teacher training and worked as an English teacher for several years. In 2012, she earned her PhD with a thesis entitled “Diskurskompetenz im bilingualen Biologieunterricht. Eine empirische Untersuchung zum Definieren”. Her current work focuses on the 368 Contributors <?page no="369"?> second language development of children with special needs in early immersion programmes. Steinlen, Anja is Senior Lecturer at the Department of Foreign Language Edu‐ cation at the Friedrich-Alexander-Universität Erlangen-Nürnberg (Germany). In 2002, she completed her PhD on L2 speech learning at the University of Aarhus (Denmark). Her work currently focuses on first, second and third language acquisition in bilingual and regular educational institutions (kindergartens, primary schools and secondary schools), especially with regard to children with a migration background and learning disabilities. The results of her PhD and post-doc projects have been published in two books (2005, 2021). Thoma, Dieter is Senior Lecturer at the University of Mannheim (Germany) and co-principal investigator of the project “Sprachliche und kognitive Re‐ ssourcen der Mehrsprachigkeit im Englischerwerb in der Grundschule”. In his research, he is interested in language acquisition and (bilingual) language processing. In addition to language learning in educational settings, he has been studying how (second) language processing interacts with cognitive, emotional and social processes in decision making and consumer behaviour. He uses different experimental methods including reaction-time tasks, eye tracking and pupillometry as well as large-scale survey studies. Videsott, Gerda graduated in 2003 in “Lettere e Filosofia” from the University of Trento (Italy). In 2009 she was awarded her PhD at the University of Augsburg (Germany) with a thesis entitled “Mehrsprachigkeit aus neurolinguistischer Sicht. Eine empirische Untersuchung zur Sprachverarbeitung viersprachiger Probanden”. She is currently a researcher at the Faculty of Design and Arts of the Free University of Bozen-Bolzano (Italy) in the field of education, inclusive education and education research. Her research interests include neurocognition of multilingualism in children. Vogelbacher, Markus is the coordinator of the project “From Kindergarten to Elementary School and Lower Secondary Education” as part of the National Ed‐ ucational Panel at the Leibniz Institute for Educational Trajectories in Bamberg (Germany). Prior to that, he worked as a post-doc on the project “Sprachliche und kognitive Ressourcen der Mehrsprachigkeit im Englischerwerb in der Grundschule” at the Department of English at the University of Mannheim (Germany). His research focuses on family stress processes, development and socialisation of values in children, early social disparities and language acquisition. 369 Contributors <?page no="370"?> Subject index age 9 f., 13-16, 21 ff., 25, 27 f., 30-33, 37 f., 41 f., 44 f., 48 f., 51, 54 f., 59, 62-65, 67, 69-74, 76-80, 82, 93 f., 103, 108, 110, 113, 116 f., 120 f., 123, 126, 128-132, 143, 147, 162 f., 165 f., 170, 186, 189 f., 193, 198, 201, 203-206, 208-213, 215, 219, 224, 231-237, 265, 267, 276, 278, 290, 294, 305, 309, 311, 315 f., 321, 324-329, 347, 349 f., 362 alerting 11, 89, 94, 100 f., 112, 114, 129 aptitude 14, 41, 56, 164, 215, 221, 285, 293, 303 f., 306 attentional mechanisms 89 f., 93, 96, 100, 102 f. Attentional Network Test 11, 89 f., 94 f., 98, 100 Automated Working Memory Assessment → AWMA automatisation 14, 285, 288, 291 ff., 300, 303 ff. AWMA 66, 68, 82 bilingualism 8, 10, 16 f., 60, 63 f., 74, 82 f., 85, 109-112, 133, 135, 138, 163 f., 170, 172, 174, 182, 184 ff., 189 f., 239, 246, 335-340, 342, 344 ff., 351-355, 357-363, 365 bilingual programme 108 BPVS 44 ff., 48-51, 57, 167 f., 170, 172, 176, 180, 187 British Picture Vocabulary Scale → BPVS CLIL 13 f., 16, 55, 133 f., 139, 223, 225 f., 236, 243, 245-252, 259-283, 367 f. cognitive abilities 7, 9 f., 12 ff., 19, 24, 26, 25, 30 f., 37, 41, 52, 62, 87, 108 f., 116 ff., 129, 131, 185, 191, 203, 208, 211, 219- 223, 225-228, 234-238, 265, 267, 273, 275 f., 278 f., 285, 287, 315, 347, 360 collocational proficiency 315, 320, 324 f., 328 f. Coloured Progressive Matrices → CPM conflict 11, 61, 94, 98-101, 336 f., 339 content and language integrated learning → CLIL CPM 110, 113 f., 116, 120, 125-128, 132 crystallised intelligence 45 cultural capital 23, 168 declarative knowledge 221, 286, 289 Dutch 10, 59, 62 ff., 67-73, 75-84, 186, 190 early bilingual 60, 62, 64, 78, 138, 189 EFL 14, 206, 216, 219, 221 ff., 225 ff., 229, 231-238, 243, 265, 267-270, 272-279, 282 f., 328, 330 f., 368 Ein Lesetest für Erstbis Sechstklässler → ELFE ELFE 121, 126, 136 English 8, 10-15, 29 f., 37, 39, 42-46, 48 f., 51 f., 54, 56, 59, 61-65, 67 ff., 71-84, 92, 94, 107 f., 110 ff., 114, 116, 118 f., 121, 126 ff., 131 f., 135, 137 f., 161-164, 166 f., 170, 172, 174 ff., 178, 180-189, 193, 196 f., 200, 203-215, 219, 226, 240 f., 243, 245 f., 248-252, 254 ff., 258-261, 265 f., 268-272, 274 ff., 278-283, 289, 293 f., 297, 307, 310-318, 320 f., 323- 332, 339, 345, 347 f., 354 f., 361, 364-369 English as a foreign language → EFL ESKOM-V 21, 23, 28, 30 f. ethnic differences 9, 21 ff., 25, 28-31 <?page no="371"?> executive functioning 10, 59-65, 69 f., 74 ff., 78, 80-84, 107, 110, 132, 134, 239, 292, 361 expressive vocabulary 24, 29, 52 First Noun Principle 195, 197 fluid intelligence 12, 42, 45, 68, 145 force schema 339, 347, 351-355, 357 ff., 364 foreign language 11-14, 38, 42, 53, 55 f., 60 f., 63, 84 f., 107 ff., 111 ff., 115-119, 122 f., 126, 128, 130 ff., 161-166, 182- 186, 188 ff., 198, 205, 212, 214, 219 f., 222, 225 f., 231, 243, 245-249, 251, 260 f., 266 f., 270 ff., 282, 292, 294, 309 ff., 313, 329 f., 362, 368 French 8, 42 f., 108, 110, 112 f., 118, 130, 138 ff., 167, 213 f., 216 f., 314, 348, 354, 362 gender 11, 14, 40 f., 49, 107, 109 f., 116, 119, 124 f., 128 ff., 132, 136, 147, 163, 165, 170, 172, 251, 253, 255, 265, 267, 273, 276, 278 f., 294, 368 German 10-13, 21, 23 ff., 27 ff., 31 f., 37, 39, 43 f., 46, 49, 51, 61, 92 ff., 97-102, 107 ff., 111, 115 f., 119 ff., 126 ff., 130 ff., 138, 141-144, 146-150, 152-155, 161 ff., 166 f., 169, 174 f., 178, 182-185, 190, 193, 204-207, 210, 213, 225 ff., 231 ff., 236 f., 240, 245, 247 f., 250-256, 260 f., 271 f., 274, 278 f., 294, 316, 323 f., 327, 366, 368 GJT 285, 294, 298, 302 ff. grammar 12 ff., 29, 38 f., 43, 56, 138, 142, 146, 149, 151, 154, 161, 163 f., 166 f., 170, 174 f., 178, 181-185, 188, 193, 198 f., 202, 213 f., 265, 267 f., 270 f., 274 f., 279 f., 285, 288, 290, 293 f., 303 ff., 313, 329 grammaticality judgement task → GJT Hungarian 117 f., 167 immersion education 37 f., 41-44, 52, 61 f., 85, 135, 348, 359-362 immigrant 21, 23, 28-34, 109, 115, 121 immigration background → migration background inhibition 8, 16, 60-63, 80, 93, 133 Input Processing 13, 193 f., 198, 212, 217, 308 intelligence 7 f., 10 ff., 17, 37, 41 ff., 45 f., 49, 52, 54 ff., 69 ff., 75 ff., 107-111, 114, 116 ff., 120, 123, 125 f., 128-132, 134, 136 ff., 145, 148, 220-224, 240, 283, 287, 336, 362 Italian 11, 89, 92 ff., 97-102, 167, 214 K-ABC 23 ff., 34 Kaufman Assessment Battery for Children → K-ABC KFT 225, 227, 230-234, 237, 241, 252, 257 kindergarten 10, 34, 46, 54, 59, 64, 91, 110, 118, 130, 146, 353 Kognitiver Fähigkeitstest → KFT L2 reading 11, 117 f., 128, 131, 137, 202 L2 vocabulary 10, 37-43, 49, 51 f., 55 language balance 59, 61-65, 69, 73, 78 f., 81 f., 85 language proficiency assessment 144, 152 lexical balance 59 f., 67, 69, 76, 78, 80 f. linguistic competence 142-145, 151, 153 ff., 343, 345 listening 39, 211, 223, 225, 236, 248, 252 f., 255 f., 258, 260, 273 f. mainstream programme 112, 122 majority language 8 f., 11 f., 62, 122, 124, 130, 135, 138, 162, 165, 183 ff., 281 metalinguistic awareness 7, 10, 51, 53, 107, 110, 112, 114, 129, 132 f., 165, 185 f., 360 f. migrant background → migration back‐ ground migration background 21 f., 38, 44, 91, 93 f., 101 f., 119, 129, 141 f., 144, 146- 371 Subject index <?page no="372"?> 149, 151 ff., 155, 245, 250, 253, 369 minority language 11 f., 62 f., 83, 109, 116, 120, 122, 130, 138, 140, 165, 189 monolingual 7 f., 10, 56, 59-65, 67-72, 74 f., 78-83, 90, 92 f., 110, 112 ff., 129, 140, 154, 161, 163, 165 f., 170, 175, 178, 182- 185, 214, 233, 237, 246, 282, 311, 325, 332, 336, 342, 344 ff., 350 f., 353, 358 motivation 40, 117, 145, 220, 227, 249, 252 f., 264, 272, 282 MToH 285, 291 f., 297-300, 303 ff. multilingualism 7 ff., 89-93, 100, 102 ff., 108, 143 f., 147, 186, 245, 270, 359, 366, 368 f. Multiple-trial Tower of Hanoi task → MToH non-verbal intelligence 10 ff., 37 f., 42-52, 107-114, 116-119, 123, 125-130, 132, 146 one-person-one-language 38, 43 orienting 11, 89, 94, 100 f. path schema 339, 341 ff., 346 f., 349, 356, 359 Peabody Picture Vocabulary Task → PPVT picture selection task → PST picture vocabulary 24 positive selection 14, 265, 278 PPVT 67 f., 84 pre-school 9 f., 19, 22 f., 28, 30 f., 35, 38 f., 42 ff., 46, 50 ff., 63, 112, 119, 122, 143, 365 Preschool Education and Educational Careers among Migrant Children → ESKOM-V Primacy of Meaning 195 primary school 7, 10 ff., 21 f., 28 f., 34, 63, 65, 87, 89, 94, 100, 102 f., 107 ff., 115 f., 118 f., 128-132, 138 f., 141, 143 f., 146 f., 154, 161, 182, 185, 188, 224, 246, 248, 269 ff., 274, 279, 327, 368 Primary School Assessment Kit → PSAK prior knowledge 13, 130, 219-224, 226, 233-239, 244 proceduralisation 292 f., 303, 305 procedural knowledge 221, 286 Processing Instruction 193 f., 198, 200 ff., 205 f., 212-215, 217, 308 productive 12, 39 f., 57, 161, 163, 167, 170, 172, 174 ff., 178, 180-184, 229, 273, 328 productive vocabulary 57, 163, 167, 170, 172, 174 ff., 178, 180-184, 273 PSAK 121, 126, 128, 137 PST 285, 296, 303 f. questionnaire 11, 65, 94, 97, 121 f., 148 f., 168, 226, 231, 250, 294, 298, 316 Raven’s Progressive Matrices 52, 107, 109, 120, 128 reading 11, 16, 22, 24, 26 f., 29 f., 34, 39, 42 f., 55, 107, 109, 115, 117 ff., 121 f., 126 ff., 130 ff., 134 f., 137 f., 142, 146 f., 162, 190, 211, 214 f., 220, 239, 244, 248, 264, 273 f., 360 receptive 10, 12, 37-40, 43, 45, 48-52, 54, 59, 64, 67, 70, 78, 161, 163, 167, 170, 172, 174 ff., 178, 180, 182 f., 229, 273, 317, 324 f., 328 receptive vocabulary 45, 48-52, 67, 78, 167, 170, 172, 175 f., 180, 182, 273 Russian 31, 62, 80, 167, 294 secondary school 7, 9, 12, 21, 24, 143, 165, 189, 191, 205, 246, 269, 271, 274, 280, 316, 327 second language 37 f., 40, 54, 56, 59, 61 ff., 79, 84, 111, 135 f., 141, 144, 154, 163, 165, 186, 188 ff., 214-217, 221, 239 f., 243, 285 f., 290, 306 ff., 313, 331, 344-347, 349, 354 f., 358, 361 ff., 368 f. 372 Subject index <?page no="373"?> sequential bilingual 189, 349 Snijders-Oomen Non-Verbal Intelligence Test → SON-R socio-economic 9, 14, 28, 50, 80, 91, 110, 114, 122, 163, 165, 168, 172, 187, 249, 253, 256, 258 f., 263, 267, 270, 360 SON-R 42, 44-47, 49, 57 Spanish 61 f., 110, 114, 130, 167, 186 f., 200, 215, 239, 345, 348 speaking 38 f., 43 f., 63, 65, 93 f., 98, 100, 107 f., 112, 132, 137, 161, 188 f., 237, 261, 273, 290, 294, 312, 326, 329, 348, 355, 362 SPM 115 ff., 120, 123, 126 ff. Standard Progressive Matrices → SPM switching 59, 61 ff., 66, 69, 71, 74 ff., 78, 80 ff., 84 f., 91, 93, 103 Test for Reception of Grammar → TROG-2 TROG-2 167 f., 170, 174, 178, 186 Turkish 9, 21, 23 ff., 27 ff., 31 f., 62, 80 f., 83, 130, 167, 188, 190, 213, 245, 294 verbal cognitive abilities 13, 219 ff., 225 f., 228, 235-238, 275 verbal intelligence 37, 43, 45, 52, 107, 109 f., 115-118, 125, 128 f., 131 f., 146, 223, 225 verbal memory 29, 81, 163, 361 vocabulary 10, 12, 25 f., 29, 37 ff., 41 ff., 50 ff., 54 f., 59, 62, 64, 67, 69, 71 ff., 75- 80, 82 f., 110, 117, 145 f., 149 f., 154, 161- 164, 166, 168, 170, 172, 174 f., 178, 182- 185, 187-190, 206, 208, 244, 246, 297, 310, 313, 317, 327, 329, 332, 344, 360 vocabulary size 55, 162, 164, 183, 317 word knowledge 38 ff., 42, 225, 236, 244 working memory 8, 10, 12, 42, 59, 61 ff., 66, 69, 71, 74, 76 ff., 80-83, 111, 145 f., 162 f., 174, 182, 188, 193, 202, 212, 215 f., 286 f., 290, 307, 361 writing 39, 42, 130, 138, 203, 216, 273, 297, 331339 written → writing 373 Subject index <?page no="374"?> Multilingualism and Language Teaching herausgegeben von Thorsten Piske (Erlangen), Silke Jansen (Erlangen) und Martha Young-Scholten (Newcastle) Bisher sind erschienen: 1 Jessica Barzen, Hanna Lene Geiger, Silke Jansen (Hrsg.) La Española - Isla de Encuentros / Hispaniola - Island of Encounters 2015, 227 Seiten €[D] 64,- ISBN 978-3-8233-6901-1 2 Anja Steinlen, Thorsten Piske (Hrsg.) Bilinguale Programme in Kindertageseinrichtungen Umsetzungsbeispiele und Forschungsergebnisse 2016, 306 Seiten €[D] 68,- ISBN 978-3-8233-6902-8 3 Christine Möller Young L2 learners‘ narrative discourse Coherence and cohesion 2015, 294 Seiten €[D] 68,- ISBN 978-3-8233-6903-5 4 Thorsten Piske, Anja Steinlen (Hrsg.) Cognition and Second Language Acquisition Research on Bilingual and Regular Language Programs 2022, 386 Seiten €[D] 78,- ISBN 978-3-8233-8194-5 5 Viviane Lohe Die Entwicklung von Language Awareness bei Grundschulkindern durch mehrsprachige digitale Bilderbücher Eine quasi-experimentelle Untersuchung zum Einsatz von MuViT in mehrsprachigen Lernumgebungen 2018, 320 Seiten €[D] 78,- ISBN 978-3-8233-8208-9 6 Cordula Glass Collocations, Creativity and Constructions A Usage-based Study of Collocations in Language Attainment 2019, 306 Seiten €[D] 78,- ISBN 978-3-8233-8171-6 7 Anja Steinlen English in Elementary Schools Research and Implications on Minority and Majority Language Children’s Reading and Writing Skills in Regular and Bilingual Programs 2021, 208 Seiten €[D] 49,- ISBN 978-3-8233-8451-9 8 Patrick Wolf-Farré, Katja F. Cantone, Anastasia Moraitis, Daniel Reimann (Hrsg.) Sprachkontrast und Mehrsprachigkeit Linguistische Grundlagen, didaktische Implikationen und Desiderata 2021, 332 Seiten €[D] 78,- ISBN 978-3-8233-8349-9 9 Ruth Trüb An Empirical Study of EFL Writing at Primary School 2022, 293 Seiten €[D] 58,- ISBN 978-3-8233-8543-1 <?page no="375"?> This volume examines interactions between second/ foreign language acquisition and the development of cognitive abilities in learners who acquire an additional language in pre-schools, primary or secondary schools. The chapters explore possible links between cognitive and linguistic skills displayed by multilingual learners. This book will appeal to different kinds of readers such as language acquisition researchers, linguists, psychologists and language teachers. ISBN 978-3-8233-8194-5 Multilingualism and Language Teaching 4