The Acquisition and Use of Yes-no Questions in English
A Corpus-Study from a usage-based perspective
0815
2016
978-3-8233-9068-8
978-3-8233-8068-9
Gunter Narr Verlag
Ursula Kania
This monograph offers a comprehensive account of the L1-acquisition and use of yes-no questions in English from a usage-based, construction grammar perspective. On the basis of the BNC and a high-density, longitudinal CHILDES corpus, the book explores two issues which have largely been neglected in previous research: 1. the prevalence of non-canonical questions (such as elliptical and declarative questions) in adult-to-adult as well as child(-directed) speech and the L1-acquisition of these structures. 2. The discourse-functional properties of both canonical and non-canonical yes-no questions, especially with regard to their influence on the acquisition process.
<?page no="0"?> 36 Ursula Kania The Acquisition and Use of Yes-no Questions in English A Corpus-Study from a usage-based perspective Language Development <?page no="1"?> The Acquisition and Use of Yes-no Questions in English <?page no="2"?> Language Development Herausgegeben von Cristina Flores (Braga), Tanja Kupisch (Konstanz), Jürgen M. Meisel (Hamburg/ Calgary), Esther Rinke (Frankfurt am Main) Band 36 <?page no="3"?> Ursula Kania The Acquisition and Use of Yes-no Questions in English A Corpus-Study from a usage-based perspective <?page no="4"?> Bibliografische Information der Deutschen Nationalbibliothek Die Deutsche Nationalbibliothek verzeichnet diese Publikation in der Deutschen Nationalbiliografie; detaillierte bibliografische Daten sind im Internet über http: / / dnb.dnb.de abrufbar. 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, Übersetzungen, Mikroverfilmungen und die Einspeicherung und Verarbeitung in elektronischen Systemen. Gedruckt auf säurefreiem und alterungsbeständigem Werkdruckpapier. © 2016 · Narr Francke Attempto Verlag GmbH + Co. KG Dischingerweg 5 · D-72070 Tübingen Internet: www.narr.de E-Mail: info@narr.de Printed in Germany ISSN 2197-6384 ISBN 978-3-8233-8068-9 b <?page no="5"?> v Table of Contents List of tables ...................................................................................................... viii List of figures .......................................................................................................ix List of abbreviations ...........................................................................................xi Acknowledgements ......................................................................................... xiii 1 Introduction and Outline.............................................................................. 1 1.1 Outline..................................................................................................... 2 2 Theories of Language Acquisition .............................................................. 6 2.1 The Universal Grammar (UG) Approach ........................................... 6 2.1.1 Generative-Transformational Grammar................................ 10 2.1.2 Summary ................................................................................... 19 2.2 The Usage-Based Approach ............................................................... 19 2.2.1 Intention Reading ..................................................................... 22 2.2.2 Pattern finding .......................................................................... 23 2.2.3 Analogy-Formation and Schematization .............................. 24 2.2.4 Learning from the input .......................................................... 24 2.2.5 Construction Grammar............................................................ 28 2.2.6 Summary ................................................................................... 45 3 The Acquisition and Use of Yes-No questions ....................................... 47 3.1 The Prevalence of Questions in Child-Directed Speech ................. 47 3.2 Developmental Paths .......................................................................... 52 3.2.1 Error-patterns in child questions............................................ 58 3.3 Non-Canonical Yes-No Questions..................................................... 62 3.3.1 Non-canonical yes-no questions in developmental research ...................................................................................... 65 3.3.2 Summary ................................................................................... 73 3.4 The Discourse Functions of Canonical Yes-No Questions ............. 73 3.4.1 Canonical yes-no questions as requests for information..... 75 3.4.2 Yes-no questions as ‘indirect speech acts’ ............................. 76 3.4.3 Developmental research on the functions of questions....... 81 3.5 Terminological note............................................................................. 83 3.6 Summary............................................................................................... 84 3.7 Research Questions.............................................................................. 85 4 Yes-No Questions in Adult-to-Adult Speech.......................................... 86 4.1 The Corpus ........................................................................................... 87 <?page no="6"?> vi 4.2 Data Extraction..................................................................................... 88 4.3 The Overall Frequency of Questions................................................. 90 4.4 Coding and Analysis of Different Question Types ......................... 91 4.4.1 Wh-questions ............................................................................ 93 4.4.2 Tag-questions ............................................................................ 95 4.4.3 Alternative questions ............................................................... 97 4.4.4 Yes-no questions ....................................................................... 98 4.5 Summary............................................................................................. 102 5 Yes-No Questions in Child(-Directed) Speech ..................................... 103 5.1 The Importance of Case Studies ...................................................... 103 5.2 The Corpus ......................................................................................... 104 5.3 Data Extraction and Coding ............................................................. 107 5.4 Overall Frequency and Distribution of Question Types .............. 113 5.5 Yes-No Questions .............................................................................. 117 5.5.1 Thomas’ Yes-No Question Errors......................................... 120 5.5.2 Longitudinal changes in the relative frequency of noncanonical yes-no questions .................................................... 122 5.6 Summary............................................................................................. 126 6 The Acquisition and Use of Weak Assertives....................................... 127 6.1 Facilitation Paths in the Learning of Canonical Questions .......... 128 6.2 Variation Sets with Weak Assertives .............................................. 133 6.2.1 You want (to) X? as an information-seeking question ........ 136 6.2.2 You want (to) X? as a request for clarification ..................... 137 6.2.3 Variation sets with You want (to) X? as a request for information.............................................................................. 138 6.2.4 Variation sets with You want (to) X? as a request for clarification .............................................................................. 140 6.2.5 Thomas’ use of You want (to) X? ........................................... 141 6.3 Summary............................................................................................. 144 7 The Acquisition and Use of Strong Assertives ..................................... 145 7.1 Inversion Rates ................................................................................... 146 7.2 The Discourse Function of Strong Assertives ................................ 149 7.2.1 That’s X? and It’s X? as requests for clarification (echouses).......................................................................................... 150 7.2.2 That’s X? and It’s X? as initiating questions (biased)......... 151 7.2.3 That’s X? and It’s X? as requests for clarification (echouses) in Thomas’ speech ........................................................ 152 7.2.4 That’s X? and It’s X? as initiating questions (biased) in Thomas’ speech....................................................................... 153 <?page no="7"?> vii 7.3 Summary............................................................................................. 154 8 The Role of Functional Factors in the Acquisition of Canonical Yes- No Questions .............................................................................................. 156 8.1 Correlation Analysis.......................................................................... 158 8.2 Form-function Mappings in Canonical Yes-No Questions .......... 162 8.3 Formal Coding and Analyses........................................................... 163 8.4 Functional Coding and Analyses..................................................... 166 8.4.1 Functional Frequency of Can-I as a Permission-Seeking Formula.................................................................................... 172 8.5 Summary............................................................................................. 174 9 Summary and General Discussion.......................................................... 175 9.1 Summary............................................................................................. 175 9.2 General Discussion ............................................................................ 177 9.2.1 The notion of grammaticality................................................ 177 9.2.2 Frequency and entrenchment ............................................... 178 9.2.3 The importance of discourse (function)............................... 179 9.3 Suggestions for further research ...................................................... 180 9.3.1 Conventionalized indirect speech acts ................................ 180 9.3.2 The role of prosody ................................................................ 181 9.3.3 Sampling Children’s Language ............................................ 181 9.3.4 Representativeness of child-language corpora ................... 182 9.4 Conclusion .......................................................................................... 182 References ......................................................................................................... 183 <?page no="8"?> viii List of tables Table 2.1: Different types of constructions......................................................... 33 Table 2.2: Different levels of the ditransitive (double object) construction ... 34 Table 3.1: Mean proportions of different construction types found in child-directed speech ................................................................................... 50 Table 4.1: Overall frequency of questions in the BNC-sample........................ 90 Table 4.2: Coding scheme and results for yes-no questions in the BNC sample ............................................................................................................ 99 Table 5.1: Overall frequency of questions in the Thomas corpus ................. 108 Table 5.2: Distribution of different types of yes-no questions in the Thomas corpus ........................................................................................... 118 Table 6.1: Distribution of PREDs, weak assertives and canonical constructions .............................................................................................................. 129 Table 6.2: Distribution of PREDs, weak assertives and canonical constructions with want ........................................................................................... 134 Table 6.3: Distribution of discourse functions expressed by You want (to) X? (INP) ....................................................................................................... 138 Table 6.4: Distribution of discourse functions expressed by You want (to) X? (INP; CHI) .............................................................................................. 141 Table 7.1: Inverted yes-no questions according to auxiliary + subject......... 147 Table 7.2: Distribution of discourse functions expressed by It’s X? and That’s X? ...................................................................................................... 150 Table 8.1: Subject+auxiliary-combinations ‘overused’ or ‘underused’ by Thomas ........................................................................................................ 161 Table 8.2: Formal frequencies of Can-PERSON-PROCESS? .......................... 163 Table 8.3: Formal frequencies of Can-PERSON-PROCESS? in the input .... 164 Table 8.4: Form-function mappings of Can-PERSON-PROCESS? constructions ............................................................................................... 170 Table 8.5: Functional frequencies of all permission-seeking constructions 173 <?page no="9"?> ix List of figures Figure 2.1: I-to-C movement for auxiliaries ....................................................... 17 Figure 2.2: do-insertion followed by I-to-C movement .................................... 18 Figure 2.3: V-to-I-to-C movement for copula be ................................................ 18 Figure 2.4: The structure of linguistic constructions ........................................ 32 Figure 4.1: The distribution of different question types in the BNCsample ............................................................................................................ 93 Figure 4.2: The distribution of different types of wh-questions ..................... 95 Figure 4.3: The distribution of different types of tag-questions...................... 96 Figure 4.4: The distribution of canonical and non-canonical yes-no questions in the BNC sample.................................................................... 101 Figure 5.1: Mean Length of Utterance (MLU) in morphemes, calculated per session ................................................................................................... 107 Figure 5.2: Mean Length of Utterance (MLU) in morphemes, calculated per month .................................................................................................... 107 Figure 5.3: The distribution of different question types in Thomas’s input 114 Figure 5.4: The distribution of different question types in Thomas’s speech........................................................................................................... 114 Figure 5.5: The distribution of different types of wh-questions in Thomas’s input ........................................................................................... 115 Figure 5.6: The distribution of different types of wh-questions in Thomas’s speech......................................................................................... 115 Figure 5.7: The distribution of different kinds of tag-questions in Thomas’s input ........................................................................................... 116 Figure 5.8: The distribution of different kinds of tag-questions in Thomas’s speech......................................................................................... 116 Figure 5.9: The distribution of canonical and non-canonical yes-no questions in Thomas’ input....................................................................... 119 Figure 5.10: The distribution of canonical and non-canonical yes-no questions in Thomas’ speech .................................................................... 119 Figure 5.11: Mean percentage of yes-no questions to utterances in Thomas’ input, calculated per month...................................................... 123 Figure 5.12: Mean percentage of yes-no questions to utterances in Thomas’s speech, calculated per month.................................................. 123 Figure 5.13: Mean percentage of non-canonical yes-no questions in the Thomas corpus (including fragments), calculated per month ............. 124 <?page no="10"?> x Figure 5.14: Mean percentage of non-canonical yes-no questions in the Thomas corpus (excluding fragments), calculated per month............. 125 Figure 8.1: Percentages of Can you X? accounted for by the four most frequent verbs ............................................................................................. 165 Figure 8.2: Percentages of Can I X? accounted for by the four most frequent verbs ............................................................................................. 165 <?page no="11"?> xi List of abbreviations APRED auxiliary-predicate yes-no question (Don’t like this? ) Aux auxiliary BNC British National Corpus CHAT Codes for the Human Analysis of Transcripts CHI child (speaker ID in the CHILDES-Corpus) CHILDES Child Language Data Exchange System C complement phrase CL Cognitive Linguistics CLAN Computerized Language ANalysis CxG Construction Grammar Det determiner GB Government and Binding HPSG Head-Driven Phrase-Structure Grammar I inflection INP input, i.e., utterances addressed to Thomas INV investigator (speaker ID in the CHILDES-Corpus) IP inflectional phrase LAD Language Acquisition Device M modal auxiliary MLU mean length of utterance MOSAIC Model Of Syntax Acquisition In Children MOT mother (speaker ID in the CHILDES-Corpus) NP noun-phrase PRED predicate yes-no question (Want some? ) P&P Principles and Parameters PP prepositional phrase RTL right-to-left elaboration S sentence SAI subject-auxiliary inversion SPRED subject-predicate yes-no question (You cold? ) ST Standard Theory UG Universal Grammar VP verb-phrase WSERT weak assertive yes-no question (You want some? ) <?page no="13"?> xiii Acknowledgements This monograph is a slightly revised and updated version of my doctoral dissertation, which I successfully defended at the University of Leipzig in early 2013. I am extremely grateful to the many people who supported me during the years I spent working on this project and I would like to thank some of them at this time. The greatest debt of gratitude is reserved for my supervisor, Doris Schoenefeld, who gave me the freedom to pursue my own line of research while at the same time providing me with guidance and mentorship whenever I needed it. I would also like to thank Elena Lieven, in particular for agreeing, without hesitation, to be the second advisor, and for inviting me to present and discuss preliminary results at one of the meetings of the child language group at the Max-Planck-Institute for Evolutionary Anthropology in Leipzig. I am also indebted to a number of friends and fellow researchers who have given me valuable feedback and provided me with technical and/ or emotional support at various stages and in a number of different contexts, e.g., at various conferences and workshops or during coffee/ lunch breaks at the English Department in Leipzig (in alphabetical order): Anne-Kristin Cordes, Antje Endesfelder-Quick, Katharina Genske, Klaus Heimeroth, Nina Julich, Heike Steinhoff, and Anja Thelen. My thanks also go to my parents, who never questioned my ability to do this or anything else in life, and whose continued support means the world to me. Finally I would like to thank my two ‚Lieblingsmenschen‘: my wonderful wife Sarah, who somehow manages to keep me happy and healthy in even the toughest of times, and my daughter Nora, who I love to bits and whose own discovery of the wonderful world of language made me fall in love with this subject all over again. Liverpool, June 2016 Ursula Kania <?page no="15"?> 1 1 Introduction and Outline The study of yes-no and wh-questions has a long tradition in the field of language acquisition research. Questions are very prevalent in child-directed speech and play an important role in the development of conversational skills (Olsen-Fulero & Conforti, 1983). In addition to that, the ways in which questions are used and responded to by children have been suggested to provide insights into cognitive development on a more general level (Cross, 1977; Ervin-Tripp, 1970; Furrow, Nelson, & Benedict, 1979; Newport, Gleitman, & Gleitman, 1977). Within generativist approaches, which have dominated the field for much of the second half of the 20 th century, the focus has mostly been on the formal characteristics of questions (Chomsky, 1957, 1965, 1995). Within this framework, the formation of syntactic questions requires the application of abstract (transformational) rules, particularly subject-auxiliary-inversion (SAI) and (for non-subject wh-questions) wh-fronting. Consequently, the investigation of children’s application of these rules has been the focus of many nativist studies of question development (e.g., Bellugi, 1965; Brown, 1968; Erreich, 1984). Since generativism posits that children possess the relevant knowledge about syntactic categories and rules from birth (as part of Universal Grammar (UG)), the fact that children have been found to produce non-target-like questions (e.g., ‘uninversion errors’) has been accounted for by appealing to performance limitations (e.g., Nakayama, 1987) or children’s operating with non-adult-like rules (e.g., Santelmann, Berk, Austin, Somashekar, & Lust, 2002). The generative approach stands in stark contrast to the usage-based, constructivist perspective adopted in this book (Tomasello, 2003). Being closely associated with the framework of Construction Grammar (e.g., Goldberg, 1995, 2006), this approach posits that children make use of general cognitive abilities i.e., pattern-finding and intention-reading skills in order to learn linguistic constructions (i.e., form-function pairings) from the input. More specifically, children are suggested to abstract away from concrete, rotelearned chunks towards more abstract constructional schemas only gradually (Lieven, Pine, & Baldwin, 1997; Tomasello, 1992). This is attested particularly well for (canonical) questions (Dąbrowska, 2000; Dąbrowska & Lieven, 2005; Steinkrauss, 2009). Within this non-derivational view of question formation, errors are thought to result from children combining and/ or elaborating partially schematic constructions in a non-adult-like way (e.g., Ambridge & Rowland, 2009; Ambridge, Rowland, Theakston, & Tomasello, 2006; Rowland & Pine, 2000; Rowland, 2007; Rowland, Pine, Lieven, & Theakston, 2005). <?page no="16"?> 2 The current study aims at extending the usage-based, constructivist account of question acquisition by exploring two issues which have largely been neglected in previous research. First of all, the formal variation encountered in yes-no questions has hardly been addressed explicitly (but see Estigarribia, 2007, 2010). In many cases, non-canonical yes-no questions (particularly rising declaratives such as You’re leaving? and ‘reduced questions’ like Want some juice? ) were analyzed as errors (Dąbrowska, 2000) or excluded from analysis (Rowland, 2007). 1 From a truly usage-based perspective, this is surprising, given the fact that the high frequency of non-canonical yes-no questions is wellattested for both adult-to-adult and child-directed speech (for adult-to-adult speech, see, e.g., Weber, 1993; for child-directed speech, see, e.g., Broen, 1972; Cross, 1977; Estigarribia, 2007, 2010; Furrow et al., 1979). Secondly, previous research has mainly focused on the formal properties of constructions, while discourse-functional characteristics have rarely been considered (for notable exceptions, see Abbot-Smith & Behrens, 2006; Cameron-Faulkner, Lieven, & Theakston, 2007; Steinkrauss, 2009, 2011). This is striking, since meaning is “probably the single most important characteristic of language from the child’s point of view” (Lieven, 2010, p. 100; also see, e.g., Behrens, 2009c). Using a combination of quantitative and qualitative methods, the current study explicitly addresses these two issues. It is the first high-density corpus study on the L1-acquisition of yes-no questions in English from a usagebased perspective which investigates the formal and functional variation within this family of constructions in a systematic way. Not only is it shown that we have to re-evaluate what children actually have to acquire (i.e., the adult ‘endpoint’ of acquisition) but also that the functional dimension of language needs to be taken into account in order to fully explain the observed developmental patterns. 1.1 Outline This book is structured as follows: Chapter 2 provides an introduction to the two major competing paradigms in language acquisition research and the associated grammatical frameworks. First, the Universal Grammar approach is addressed, followed by a brief discussion of generative-transformational grammar, with a special 1 Following Estigarribia (2007, 2010), ‘non-canonical’ as used in this book refers to yes-no questions which do not exhibit the characteristics commonly ascribed to these constructions (i.e., cases which ‘lack’ inversion and structures without subject and/ or auxiliary). It is not used in the sense of non-canonical clauses that is put forward by Huddleston and Pullum (2006, p. 46) according to which all questions are considered to be non-canonical clauses, since they deviate from the ‘canonical’, declarative pattern. <?page no="17"?> 3 focus on the treatment of yes-no questions. In Section 2.2, the usage-based approach adopted in this book is introduced, followed by a brief summary of Construction Grammar. The Goldbergian (2006) approach to SAI is outlined in some detail and unresolved issues are discussed. In Chapter 3, previous research on the development and use of (yes-no) questions is reviewed. Starting out from a brief discussion of the prevalence of questions in child-directed speech, research on developmental paths and error patterns is considered (Sections 3.1, and 3.2, respectively). Sections 3.3 and 3.4 zoom in on the two issues which are at the core of this book, i.e., non-canonical yes-no questions and the discourse-functional properties of (both canonical and non-canonical) yes-no questions. This critical discussion is concluded by the research questions presented in Section 3.7, which serve as the basis for the empirical part of this book. Chapter 4 contains the first corpus-based study, focusing on the overall frequency of questions and the distribution of canonical vs. non-canonical yes-no questions in adult-to-adult speech. After defining the target structures and establishing the coding criteria, the results obtained from the analysis of a subsample of the British National Corpus (BNC) are discussed. While the overall frequency of questions is not as high as that reported for child-directed speech, non-canonical questions are found to be very prevalent within the family of yes-no questions. Chapter 5 zooms in on the developmental data; using all question constructions extracted from the high-density Thomas corpus (Lieven et al., 2009), the distribution of canonical vs. non-canonical questions is assessed. Consequently, developmental changes in the relative frequency of noncanonical questions are analyzed. It is shown that the relative frequency of non-canonical questions in the input stays relatively constant. In contrast to this, non-canonical questions are very prevalent in Thomas’ early speech and their relative frequency gradually decreases over time. However, at the end of the recording period (4; 11) 2 , Thomas still uses particular noncanonical questions with a higher frequency than his caretakers, indicating that the acquisition process is far from complete at this point. Overall, the findings from Chapters 4 and 5 suggest that non-canonical questions are part of the adult target repertoire and thus should be integrated into usagebased accounts of language acquisition. Chapters 6 and 7 provide in-depth analyses of particular types of noncanonical yes-no questions. Focusing on PREDs (Want some? ) and weak assertives (WSERTs; You want some? ), Chapter 6 offers evidence suggesting that these non-canonical structures may be used by Thomas as a basis for learning their more complex, canonical counterparts (Do you want some? ) by means of an adjunction strategy. First of all, it can be shown that Thomas 2 Children’s age is always provided in the format Years; Months or Years; Months.Days, i.e., 4; 11 means 4 years and 11 months. <?page no="18"?> 4 uses item-specific PREDs first and that the corresponding weak assertives and canonical questions emerge only later on. Furthermore, weak assertives enter into paraphrase relations, i.e., they are often used alongside their more complex counterparts in child-directed speech, providing Thomas with models for an adjunction strategy. Extending previous work by Estigarribia (2007, 2010), it is proposed that this strategy can be viewed as a modified version of the juxtaposition-operation suggested by Dąbrowska & Lieven (2005). Besides being used as functional equivalents to canonical questions, it is found that weak assertive can also serve as echo-questions (A: I want some. B: You want some? ), lending support to the claim that they can be conceptualized as both ‘reduced’ questions and rising declaratives. Chapter 7 concentrates on another class of non-canonical question, socalled strong assertives (That’s a ball? ), which have often been treated as uninversion errors. The ‘inversion rates’ of particular auxiliary+subject combinations are analyzed, expanding findings from an earlier corpus-based study by Dąbrowska (2004, Chapter 9). The results show that the caretakers hardly ever reach an inversion rate of 100%, indicating that assumptions about the adult target inventory should always be evaluated on the basis of real usage-data. Further analyses reveal that strong assertives are used by both the caretakers and Thomas to either echo a previous utterance (A: That’s rice. B: That’s rice? ) or to bias the interlocutor towards a positive reply to a tentative assumption (I can have that? ), corroborating claims made concerning the discourse function fulfilled by strong assertives. The fact that Thomas mostly uses strong assertives in a pragmatically appropriate way underlines the need to consider the discourse context before classifying children’s productions as ‘errors’. The role of discourse-functional factors is further elaborated on in Chapter 8. The focus is on different Can-PERSON-PROCESS? -constructions, whose emergence and frequency of use in Thomas’s speech cannot be accounted for by the relative frequency of these structures in his input. Providing an in-depth analysis of developmental patterns and form-function mappings, it is shown that Thomas is influenced by the functional rather than formal frequencies found in the ambient language, indicating that functional factors need to be addressed in order to fully account for the observed patterns. Since particular item-based constructions are consistently used to express requests (for permission or action), it is suggested that these ‘indirect’ uses (Can I please have a go? ) are learned as form-function pairings from the input, making it possible to conceptualize them as constructions whose (presumably ‘indirect’) function is directly associated with their form (as suggested by Stefanowitsch, 2003). In Chapter 9, the major empirical findings of the studies presented in Chapters 4-8 are summarized. Furthermore, the results are discussed with regard to general issues within usage-based, constructivist approaches to <?page no="19"?> 5 language development and use. After addressing open questions and possible directions for further research, a brief, general conclusion is offered. <?page no="20"?> 6 2 Theories of Language Acquisition There are two major competing paradigms within the field of language acquisition: Universal Grammar (UG) approaches (which is also referred to as generativism or nativism) and usage-based (also: constructivist, emergentist) approaches, the latter of which are compatible and partially overlapping with functional approaches. Like the different terms and the plural form ‘approaches’ already suggest, diverse manifestations are subsumed under each heading, each focusing on different aspects of language acquisition and proposing slightly different views concerning, for example, the formalization of grammatical structures and the nature of linguistic knowledge. Still, there are certain basic assumptions that are shared by all UG approaches on the one hand and all usage-based approaches on the other. Furthermore, those aspects of the associated grammatical frameworks which are relevant for the current debate on the formation and acquisition of questions are relatively homogeneous within each of the two paradigms. The following sections provide a brief summary of the main characteristics of UG and usage-based approaches, respectively, in both cases followed by a short outline of the corresponding grammatical framework (generativetransformational grammar and construction grammar), with a focus on the formalization of (canonical) yes-no questions. This overview will provide the basis for the discussion of previous research on the acquisition and use of yes-no questions in Chapter 3. 2.1 The Universal Grammar (UG) Approach Inextricably linked to the work of Noam Chomsky, the foundations for the generative enterprise were laid by his publications in the late 1950s and early 1960s (Chomsky, 1957, 1964, 1965), with his fundamental views about language (acquisition) already firmly in place in his devastating review (1959) of Burrhus F. Skinner’s Verbal Behavior (1957). Chomsky’s comprehensive criticism was not only a reaction (and objection) to Skinner, but to behaviorist notions in general, making him a key figure in the so-called cognitive revolution within the cognitive sciences. In behaviorist psychology, only observables (like external stimuli and behavior) are studied while mental processes and representations are not considered, since they take place within the ‘black box’ of the mind and cannot be investigated directly. Based on Pavlov (1927) and Watson and Rayner (1920), Skinner introduced the notion of operant (or: instrumental) conditioning, which - unlike classical conditioning - focuses on the modification of non-reflexive (i.e., voluntary) behavior. Depending on whether a behavior is seen as desired or undesired, <?page no="21"?> 7 it can be either strengthened (through positive or negative reinforcement) or weakened (through punishment or extinction), leading to learning by association. Complex behavior can be taught via successive approximations, i.e., by rewarding each small advancement towards the desired behavior. Generalization to new instances (i.e., appropriate reactions to new stimuli) is seen as the result of stimulus generalization (learning by induction). Skinner proposed that the same mechanisms underlying (the manipulation of) animal behavior (like getting rats to press levers in order to receive food pellets) can be used to explain human behavior as well, including the acquisition and use of language (Skinner, 1957, p. 3). Adult-like verbal behavior is thus thought to be achieved by the selective reinforcement of certain vocalizations of the child, which “gradually assume forms which produce appropriate consequences in a given verbal community” (Skinner, 1957, p. 31). In his review, Chomsky pointed out that successful language acquisition cannot be the result of behaviorist learning, since not every correct verbalization is reinforced and errors are rarely corrected explicitly (Chomsky, 1959, p. 56). At the core of his criticism lies the notion of language as a uniquely human ability that is much more complex than the mere repertoire of verbal responses described by Skinner. Most importantly, language exhibits productivity, i.e., speakers are able to understand and produce a potentially infinite number of sentences, most of which they have never encountered before (Chomsky, 1959, p. 56). Furthermore, native speakers of a language can reliably judge whether a structure is well-formed (i.e., grammatical) or not, regardless of its semantics (Chomsky, 1959, p. 54). Chomsky proposed that this goes far beyond the superficial analogy of stimulus generalization and that “we recognize a new item as a sentence not because it matches some familiar item in any simple way, but because it is generated by the grammar that each individual has somehow and in some form internalized” (Chomsky, 1959, p. 56). This grammar relies on a set of potentially recursive rules that operate on linguistic categories and phrases, generating an infinite number of sentences from the limited lexical items in a given language, thus being reminiscent of the Humboldtian notion of making ‘infinite use of finite means’. This assumption about the representations and processes underlying linguistic productivity is shared by all generativist proposals and also entails a view of language acquisition that is radically different from the account provided by behaviorism. From a generative point of view, mental events and representations, i.e., the contents of the ‘black box’, lie at the heart of linguistic abilities and thus must be taken into account, even if they can only be investigated by indirect means. What is more, the abstract knowledge present in the human mind is thought to go beyond what can be learned from experience; while lexical items (i.e., words, fixed phrases, and idiomatic expressions) can (and in fact have to) be learned from the input using general learning mechanisms, knowledge about grammatical categories and rules cannot be learned this way: <?page no="22"?> 8 A consideration of the character of the grammar that is acquired, the degenerate quality and narrowly limited extend of the available data, the striking uniformity of the resulting grammars, and their independence of intelligence, motivation, and emotional state, over wide ranges of variation, leave little hope that much of the structure of the language can be learned by an organism initially uninformed as to its general character. (Chomsky, 1965, p. 56) Within this approach, the data provided by speakers’ performance is thus thought to be neither sufficient to learn a language nor is it considered to be of direct relevance to linguistic theory, which “is concerned primarily with an ideal speaker-listener ... unaffected by such grammatically irrelevant conditions as memory limitations ... and errors (random or characteristic) in applying his knowledge of the language in actual performance” (Chomsky, 1965, p. 3). The resulting assumption that language cannot be learned from the input is known as the argument from the poverty of the stimulus (Chomsky, 1980, 1986; Crain & Pietroski, 2001; Lightfoot, 1989), which entails the claim put forward by all nativists that some linguistic knowledge is unlearnable and thus must be innate. 3 Abstract grammatical knowledge is seen as the core of this so-called linguistic competence, while languagespecific knowledge (most notably the lexicon) is viewed as belonging to the periphery. Chomsky (1965) posits an innate language faculty (termed language-acquisition device (LAD)) that is hard-wired in the brain and can account for the fact that all normally-developing children achieve full mastery of their native language(s) within a short time span and with relative ease despite limited exposure to linguistic input. Sometimes, the LAD is compared to an organ, whose general structure is “genetically determined, on a par with the elements of our common nature that causes us to grow arms and legs rather than wings” (Chomsky, 1988, p. 4). Consequently, the process of language acquisition is not primarily a matter of learning, but of maturation and growth, in effect marginalizing the importance of social interaction. The LAD is sometimes also referred to as a cognitive module which is largely separate from other cognitive modules and exhibits domain-specific characteristics (Fodor, 1983). The notion of maturation has also been taken to entail that a language can only be fully acquired within a biologically predetermined timeframe. Popularized by Lenneberg (1967), this so-called critical period hypothesis has since been subject to much criticism, mainly because it is only vaguely defined and empirical evidence is sparse (for a comprehensive overview, see Singleton, 2005). The linguistic knowledge initially contained in the LAD is termed Universal Grammar (UG), which “is sufficiently rich to account for the acquisi- 3 This ‘logical problem of language acquisition’ is also known as Plato’s problem: In Plato’s Meno (470-399 BC), Socrates demonstrates that an uneducated slave boy knows geometric principles, suggesting that this knowledge is not learned from experience but present from birth. <?page no="23"?> 9 tion of language, yet not so rich as to be inconsistent with the known diversity of language” (Chomsky, 1965, p. 58). There is little - if any - agreement concerning the exact contents of UG and suggestions range from very comprehensive and inclusive views (see for example, Cowper, 1992) to the identification of recursion as the only innate property of grammar (Hauser, Chomsky, & Fitch, 2002). The vast majority of nativist accounts posits that at least some functional or lexical categories (or the features that the categories consist of) are innate (such as noun or verb). Instead of having to build categories on the basis of the input, children just have to map the specific items they encounter to the innate abstract categories, supposedly making the task easier. This innate knowledge about categories leads to the so-called linking problem: How are children able to connect the highly abstract categories provided by UG to the respective instances in their native language? 4 A logical consequence of the innateness hypothesis is the claim that theories of adult grammars should be applicable to children’s emerging linguistic systems as well, since they possess the same basic linguistic categories as adults do this is known as the (the strong version of the) continuity hypothesis (Pinker, 1984; Hyams, 1986). 5 From this assumption it follows that the obvious differences between children’s and adults’ linguistic competence and developmental changes in children’s speech have to be accounted for without reference to differences in grammatical knowledge. One possible ‘way out’ is to blame performance factors (such as limited working memory, processing difficulties, or lack of pragmatic or lexical knowledge) that hinder children from showing their full competence. Along similar lines, the maturation hypothesis (Borer & Wexler, 1987, 1992; Babyonyshev, Ganger, Pesetsky, & Wexler, 2001) attributes the differences at least partly to the fact that parts of the brain have to mature in order for certain aspects of UG to become available. However, performance limitations and maturation have been criticized as ‘fudge factors’, since there is usually no independent measure of the postulated processing load nor reference to genetic research and alternative explanations are not considered (Tomasello, 2003, p. 187; also see Braine, 1994, and Sampson, 1997). The problems faced by UG-accounts derive mainly from the formal description of the final state they are based on, since the characterization of how language is acquired of course depends on what it is that must be acquired. In other words, any theory of language acquisition is based on a particular theory of language. The richer and more abstract the final state of language acquisition is posited to be, the more likely it is that a quite rich 4 The only theory directly addressing this problem is Pinker’s semantic bootstrapping account (Pinker, 1984, 1989) see Tomasello (2005) for a critical discussion. 5 A weaker version - aptly named ‘weak continuity hypothesis’ assumes that children do not necessarily operate with full adult syntax. However, they only make use of grammatical structures that are licensed by the general principles of UG (Clahsen, 1990; Haan, 1987; Jordens, 1990). <?page no="24"?> 10 initial state is assumed. Therefore, the primary aim of generativetransformational accounts of grammar is to develop a theory of language that is both observationally and descriptively adequate, i.e., to specify rules that produce only those structures that are considered to be well-formed and to capture a native speaker’s intuition (Chomsky, 1964, p. 29). Based on such an adequate model of the linguistic competence of adult native speakers, a linguistic theory should aim at explanatory adequacy, i.e., it should have predictive power and “provide a principled basis, independent of any particular language, for the selection of the descriptively adequate grammar of each language” (Chomsky, 1964, p. 29). Although this third level of adequacy was seen as ‘utopian’ at the time Chomsky formulated the first generative-transformational theory of language (Chomsky, 1965, p. 26), it was set as the ultimate aim and was one of the driving forces behind the constant revisions of the theory, whose most important developments are outlined below. 2.1.1 Generative-Transformational Grammar The most influential generative-transformational theories of syntax are known as Standard Theory (ST; Chomsky, 1965, 1957), Principles and Parameters (P&P, based on Government and Binding [G&B]; Chomsky, 1981) and Minimalism (Chomsky, 1995). 6 While some aspects of ST were abandoned decades ago, it will be outlined below in some detail mainly for two reasons. First of all, it was the first generative-transformational model of syntax and served as the basis for all subsequent developments. It was also the framework adopted by the majority of research on language acquisition in the 1960’s and 1970’s, a time during which much influential research on question-formation was conducted. 2.1.1.1 Standard Theory In his seminal work Syntactic Structures (1957), Chomsky proposed a model of syntax that combines phrase structure rules (as applied in immediate constituent analysis) with grammatical transformations (and morphophonemic rules) in order to provide derivations for all well-formed sentences of a language (Chomsky, 1957, p. 46). The application of phrase structure rules can be formalized as (1): (1) X Y (Z) 6 Generative grammars that are non-transformational (such as Head-driven Phrase Structure Grammar (HPSG); Pollard and Sag, 1994) will not be considered here. For a very comprehensive treatment of interrogatives from a HPSG-perspective, see Ginzburg (2001). <?page no="25"?> 11 X stands for the name of the phrase and the arrow stands for the application of the phrase structure rule, also called rewriting rule because X is rewritten as Y (Z). Y and Z can either be lexical categories (like noun or determiner) or phrasal categories (also called constituents, which are then further expanded by the application of another phrase structure rule in the next step). Round brackets indicate that an element is optional. The basic phrase structure rules (according to Chomsky 1957 and 1965) underlying so-called kernel (i.e., simple, active, declarative, affirmative) sentences are given in (2) to (6): 7 (2) S NP VP (3) NP (Det) N (4) VP Verb (NP) (PP) (5) Verb Aux V (6) Aux C (M) (have + en) (be + ing) (be + en) In this formalization, Aux is an obligatory part of Verb (and thus of VP) and contains an obligatory tense-element (symbolized by C in Chomsky, 1957, p. 46; this was later replaced by tense, Chomsky, 1965, p. 43) which can be rewritten as a zero morpheme, the third person singular present tense inflection or the past tense inflection. The other elements in Aux are optional but if they are chosen, they must occur in the given order. The structure resulting from the application of phrase structure rules is termed the ‘base’ or ‘deep structure’ (later referred to as D-structure), which is mapped onto a ‘surface structure’ by transformational rules (Chomsky, 1965, p. 16). 8 The different levels of representation linked by transformations allows for an economical description of syntax, since transformations operate on strings of constituents and can capture more general processes and thus have a much wider scope than idiosyncratic phrase structure rules. For example, sentences (7) and (8) can be traced back to one common underlying structure which in one case has been changed by a transformation (known as particle-movement), a formalization that is more parsimonious than positing two separate phrase structure rules and is also thought to mirror a native speaker’s intuition according to which both structures are closely related. (7) He gave up the plan. (8) He gave the plan up. Unlike phrase structure rules, transformations can also handle the interdependency of discontinuous elements, which is necessary, for example, to 7 Please refer to the list of abbreviations for an explanation of all symbols. 8 Note that in this account, the surface structure serves as input to the phonological component and thus is not equivalent to the actual realization of a sentence. In more recent work within P&P, this is referred to as S-structure, while the term surface structure is applied to the output of the phonological component (Radford, 1988, p. 419). <?page no="26"?> 12 account for the fact that auxiliaries determine the form of the following main verb. Consider example (9): (9) The man is dancing. The inflectional affix -ing on the verb depends on the preceding selection of be. According to the formalization in (6), the selection of be automatically adds -ing, i.e., the elements are treated as continuous on the phrase structure level. By an affix transformation (also referred to as affix hopping; Sobin, 2010, p. 92), the inserted inflections are moved onto the next verb, separating be from -ing and thus recovering the discontinuity between the auxiliary and the inflection on the main verb. 9 Within this framework, transformations which can provide derivations for structures that deviate from kernel sentences, like negated sentences, passives, and interrogatives, are of particular interest. For example, yes-no questions are formed by applying the transformation T q to declarative strings with the structures (10) to (13) (Chomsky, 1957, pp. 61-63): (10) NP - C + M - … (11) NP - C + have - … (12) NP - C + be - … (13) NP - C - V -… T q changes the order of the first and second segment (14) 10 - a declarative sentence (15) can thus be transformed into a yes-no question (16). (14) X 1 - X 2 - X 3 X 2 - X 1 - X 3 (15) She is a good linguist. 11 (16) Is she a good linguist? 9 For a detailed formalization and explanation of this transformation, see Chomsky (1957, pp. 39-42). 10 Estigarribia (2007) criticizes that this transformation operates on the segment C + M/ have/ be “even though C + M/ have/ be is not a constituent and transformations are supposed to be structure dependent” (Estigarribia, 2007, p. 11). However, structure dependence merely indicates that constituent boundaries are recognized, making it possible to “formulate a transformation that can insert all or part of [emphasis added] the Auxiliary Verb to the left of a Noun Phrase that precedes it” (Chomsky 1965, p. 55). 11 In this, and many of the following examples, only the resulting structures will be given. Still, it is important to keep in mind that transformational rules operate on the structures underlying sentences and not on actual sentences. Since the term SAI may be misleading in this respect, some researchers used alternative terms (e.g., ‘whquestion modal auxiliary placement rule’) instead of SAI in order to distance themselves from the notion that “each correct whquestion containing a modal auxiliary has that modal auxiliary in a declarative position at some point in its formation” (Kuczaj and Brannick, 1979, p. 45). <?page no="27"?> 13 In declarative strings containing a lexical verb (13), another operation is applied, because inverting the first and second segment (which consists only of C and not C+V) results in an unaffixed element in initial position. Consider sentence (17), whose derivation after application of all phrase structure rules looks like (18): (17) She loves chocolate. (18) She + S + love + chocolate. Applying T q results in the following sequence: 12 (19) * -s she love chocolate? In these cases “do is introduced as the ‘bearer’ of an unaffixed affix” (Chomsky, 1957, p. 62), resulting in a well-formed yes-no question: (20) Does she love chocolate? This is also commonly referred to as do-support, while T q has become known as subject-auxiliary inversion (SAI) or NP-AUX inversion. 13 Transformations were seen as an elegant generalization for the description of well-formed structures, especially since they operate on (parts of) constituents, not words (Chomsky, 1965, pp. 55-56, 1971, pp. 27-28), thus exhibiting a characteristic that Chomsky believed to be inherent to all natural languages: structure dependence. One prime example of the need for structure-dependent rules are interrogative structures with a subject NP containing a relative clause with an auxiliary, like in Chomsky’s (1980) example: (21) The man who is smoking is tall. Operating with a structure-independent rule (i.e., a rule that just takes into account the linear ordering of elements and not their inherent hierarchical structure) like ‘move the first auxiliary to the front’ would result in an ungrammatical sequence such as: 12 Affix hopping is applied after T q , otherwise the inflected verb (in this case loves) would be inverted, which is the standard way of forming yes-no questions in other Germanic languages such as German. Up to (and including) the Early Modern English Period, SAI was not obligatory for question-formation (Radford, 1997) and some modern varieties of English (such as Hawai’i and Jamaican Creole) also do not have SAI in questions (Carr, 1972; Radford, 1997). 13 Note that in most accounts referring to this (or a very similar) process as SAI or NP- AUX-Inversion, the term ‘auxiliary’ is used to refer not only to genuine auxiliary verbs (HAVE, BE), but also to modal auxiliaries (can, could, may, might, shall, should, etc.), copula BE and the dummy auxiliary DO. <?page no="28"?> 14 (22) * Is the man who ___ smoking is tall? Only a rule that recognizes that not the first auxiliary (which in this case belongs to the subject) but the auxiliary of the main predicate needs to be moved leads to the correctly inverted sentence: (23) Is the man who is smoking ___ tall? Supposedly children almost never hear sentences like (23) - and even if they did, they could not infer the correct, structure-dependent rule from these structures without innate constraints limiting the hypothesis space. 14 This feeds directly into the poverty of the stimulus argument, since children arrive at the correct rule without the help of relevant structures in the input. Apart from studies showing that structures like (23) do occur and arguing that children could infer the correct pattern from this (Pullum & Scholz, 2002), this Chomskyan argument has been criticized for its assumption that structure-dependence is a purely formal phenomenon: If we allow for children’s recognition of structure based on function, the problem does not arise, since children recognize that the whole NP makes one act of reference (Stemmer, 1981; Tomasello, 2005, p. 190; also see Lewis, 2002, on how children may use stochastic information in the input to learn structures which are not encountered in the input). 15 In comparison with yes-no questions, the formation of wh-questions is essentially viewed as an additional transformation (often referred to as wh-fronting) which is applied to the structure generated by T q / SAI (Chomsky, 1957, p. 69), a notion that is also found in subsequent accounts, albeit with small modifications. This is also mirrored in a line of experimental research addressing the psychological reality of transformations. According to the so-called Derivational Theory of Complexity the processing difficulty associated with a sentence should depend on the number of transformations that have to be applied, tying in with research in psychology showing that only a limited amount of information can be kept track of in a single mental activity (Miller, 1956). While early experiments provided promising results (Miller & McKean, 1964), subse- 14 This is also pointed out by Ambridge and Lieven (2011) who state that - according to the generativist approach - “since there are an infinite number of [wrongheaded] hypotheses, there can never be enough input data to rule them out, no matter how much language the child hears” (Ambridge and Lieven, 2011, p. 120). 15 Moreover, the universal status of structure-dependence is questionable, since there are languages which do not exhibit this characteristic (for example, see Everett, 2005, 2007, on Pirah-, an Amazonian language that consequently does not make use of recursion and embedding). Interestingly, it has also been shown that children can be made to produce structure-dependence errors within an experimental setting (Ambridge, Rowland, and Pine, 2008). <?page no="29"?> 15 quent research led to the abandonment of the Derivational Theory of Complexity (for an overview, see Fodor, Bever, & Garrett, 1974). Nevertheless, the successful application of transformational rules is still viewed as the main task in language acquisition in most generative research. Therefore, intermediate stages are of particular interest. This is mirrored in the high number of studies focusing on different stages in the acquisition of questions and a comparison of inversion rates between yes-no questions (which require one main transformation) and wh-questions (which require an additional transformation) - for a short discussion of this research, see Section 3.2. Within a later manifestation of ST (resulting in what is known as Revised Standard Theory or Extended Standard Theory, Chomsky, 1973), the formalization of (the structures underlying) SAI was slightly modified. For example, in an account based on Emonds (1976) and Culicover (1976), SAI was formulated as the transformation that turns (24) into (25) (Akmajian, Steele, & Wasow, 1979, pp. 17; 54): (24) X - NP - AUX - Y (25) X - AUX + NP - Y In this account, the syntactic category AUX is maintained but only includes the tense element and either a modal or do, while be and have are viewed as verbs. An additional transformation - do-replacement - causes auxiliary have and be to take the place of do in AUX. Adopting X-bar-conventions, the original phrase structure rule determining the ordering of constituents (see (6)) is abandoned; instead, the ordering is accounted for by subcategorization rules affecting three different levels of verbal phrases (containing perfective have, progressive be, and passive be, respectively). This account minimizes the number of elements assigned to AUX and allows SAI to be formalized as operating on the whole constituent AUX instead of only its first member(s). Despite much research that aimed at refining ST, the main problem could not be solved within this framework: Although transformations are more powerful than phrase-structure rules, many specific transformations were posited, leading to a very rich, non-restrictive grammar. While this grammar can model native speaker intuitions about the well-formed structures of the language (thus achieving observational as well as descriptive adequacy, at least for English), it does not achieve explanatory adequacy. In other words, a child acquiring the language is simply presented with too many options and no convincing mechanisms are suggested that constrain the hypothesis space in order for the appropriate grammar to be chosen. Subsequent accounts thus aimed at limiting the number of categories and tried to subsume very specific transformations under more general operations. This was <?page no="30"?> 16 achieved by focusing on linguistic universals instead of language-specific characteristics. 2.1.1.2 Principles and Parameters By the time Chomsky published his influential Lectures on Government and Binding (1981), the basic tenets of generative-transformational grammar had undergone comprehensive revisions. X-bar theory (for example, Chomsky, 1970; Jackendoff, 1977) had been established as one of the leading approaches to syntax, and transformations were consequently formalized as holding over phrase structure trees with a hierarchical organization. Universal Grammar was assumed to contain only a limited amount of principles that are common to all languages (i.e., linguistic universals). In order to account for the variation between languages, UG also contains parameters, which are usually described as binary switches, since they can take one of two possible values, thus limiting the hypothesis space for children acquiring a language. There is little agreement concerning the number and exact nature of parameters (Fodor, 2001, p. 734). It is commonly assumed that each parameter is set to a default position at birth and that very limited exposure to the target language is sufficient in order to confirm the setting or to trigger a resetting of the parameter. Examples of parameters accepted by most generative linguists include the so-called ‘pro-drop’ (also: null subject) parameter and the ‘head-direction’ parameter. The pro-drop parameter determines whether the subject of a verb must be overtly realized or not. While English (presumably) is non-pro drop, Spanish and Italian are described as pro-drop languages. Not only does the setting of the parameter to ‘pro-drop’ license null thematic subjects in tensed clauses, it is also thought to specify other features (for example, that-trace violations). 16 With one parameter fixing a bundle of related features, the number of parameters is minimized, thus leaving less room for errors during the language acquisition process. Different surface phenomena can serve as evidence for the correct setting of one underlying parameter, making it possible to formulate hypotheses that can be tested empirically by analyzing children’s production data. Although parameters presumably make the task of language acquisition easier, they cannot solve the linking problem, since there are still “ambiguities of string-to-structure alignment” (Fodor, 2001, p. 765). The head direction parameter, for example, specifies the order of head and complement in various phrases and clauses of a language however, in order to set the parameter, the child must determine which elements are heads and which are complements. If the child can do this, “she will also know which order they came in. If she already knows which order the head and the complements come in a sentence, there is no need to set the parameter” (Mazuka, 1996, p. 25). 16 For a critical discussion of cross-linguistic research showing that this correlation does not hold, see Newmeyer (1998, pp. 357-359). <?page no="31"?> 17 Within this framework, the formation of yes-no questions is no longer seen as the result of a very specific SAI-transformation; instead, it is conceived as I-to-C-movement, which is held by some to be a parameter of UG (for example, Fodor, 2003). Following conventions established within G&B by the end of the 1980’s, the term movement rule is used instead of transformation, since the operation is viewed not as structure changing, but as the copying of a constituent followed by deletion (leaving a trace that is co-indexed with the moved element). The very specific transformation rules are replaced by the basic operation ‘Move α’, which is constrained by general principles. Following Chomsky (1981), Santelmann et al. “assume that auxiliaries and modals are directly generated in the head of the inflection phrase (IP), and then move to the C position in questions” (Santelmann et al., 2002, p. 816) (Figure 2.1). Lexical main verbs do not move; instead, do is inserted into I, followed by I-to-C-movement (Figure 2.2). Finally, copula be is the only element that is allowed to move from VP to I, again followed by I-to-C movement (Figure 2.3). Figure 2.1: I-to-C movement for auxiliaries <?page no="32"?> 18 Although this formalization looks quite different from the original ST account of SAI, one common feature is the view that interrogative structures are essentially derived from basic underlying structures which are turned into questions by purely formal operations. Within all generativetransformational approaches, structures like reduced yes-no questions are treated as deviations from the full, canonical forms, i.e., as (ungrammatical) performance phenomena - an important point which will be discussed in more detail in Chapter 3. The P&P approach was popularized in the 1980s and ’90s mainly by Stephen Pinker (for example, 1984, 1994); for language acquisition researchers working within the generative paradigm, P&P remains the most influential generative account and the reference work by Guasti (2002) is still widely used (for a recent review of work in language acquisition within the P&P framework, see Hyams et al., 2015). The most recent development within the Figure 2.3: V-to-I-to-C movement for copula be Figure 2.2: do-insertion followed by I-to-C movement <?page no="33"?> 19 generative enterprise - Minimalism (Chomsky, 1995, 2000) - has not been widely used in language acquisition research, mainly because it “is still more of a research program than a theory” (Snyder, 2007), aiming at descriptive elegance (i.e., maximal economy) rather than explanatory adequacy. Therefore, it will not be considered here. 17 2.1.2 Summary Nativist accounts based on generative-transformational theories of grammar posit innate linguistic knowledge about representations or principles in the form of a Universal Grammar, which resides in a (relatively) autonomous language faculty and enables children to acquire their native language(s) rapidly and with ease. This is based on the assumption that the linguistic environment provides neither unambiguous nor sufficient information on the underlying linguistic structures (poverty of the stimulus argument). According to the so-called continuity hypothesis, children possess the same basic grammatical knowledge as adults do, with differences in linguistic performance presumably resulting from processing limitations. Since actual language use is seen as an imperfect manifestation of a speaker’s tacit grammatical knowledge, UG approaches focus on linguistic competence (as opposed to performance) and on the formal rather than the functional dimension of language. Consequently, linguistic productivity is based on purely formal rules (transformation or movement) that operate on structures on a subphonological level. Nativism thus contrasts sharply with the usagebased approach adopted here, which will be outlined in the next section. 2.2 The Usage-Based Approach Many linguists and psychologists believe that there is a biological basis for language, just not in the form of an autonomous Generative Grammar. Just as plausible for these linguists is the hypothesis that language rests on more general biological predispositions, such as the abilities to create and learn symbols, to form concepts and categories, to process information rapidly, and to interact and communicate with other persons intersubjectively. (Tomasello, 1998a, p. xi) Linguistic nativism has dominated research on language acquisition since it was first established in the late 1950’s and it still is the approach against which alternative theories (have to) position themselves. It is only during the 17 It is interesting to note that recent work within the Minimalist Program focusing on language acquisition proposes as the guiding principle that “UG should be resorted to as an explanation for the observed facts only after one has considered the possibility that they are the result of input or general cognitive principles” (Westergaard, 2009, p. 16). Therefore, it remains to be seen, what - if anything - will be left for UG. <?page no="34"?> 20 last decade that a gradual shift towards the usage-based, constructivist approach can be observed, mainly due to Tomasello’s early publications on this subject (for example: 1995, 1998a, 1998b, 2000) and the first book-length treatment offering a comprehensive account of this theory (Tomasello, 2003). Recent methodological and empirical advances (like connectionist modeling, imaging techniques, and analysis of large corpora) have also contributed to the shift towards emergentist approaches (MacWhinney, 2010, p. 26), whose basic tenets can be traced back to previous research within developmental psychology (e.g., Piaget, 1935/ 1952, 1937/ 1954). 18 The term usage-based was coined by Langacker (1987, 1991) in his outline of Cognitive Grammar, one of a group of related theories within cognitive-functional linguistics adopting the premise that “language structure emerges from language use” (Tomasello, 2003, p. 5, on the usage-based model; also see Bybee, 1985, 2006). 19 This non-nativist theory holds that language is integrated with general cognition and thus assumes no (strong) modularity, viewing a possible localization of language functions in specific parts of the brain as a result of experience with language rather than stemming from an innate language module. Along similar lines, no strict critical or sensitive period is posited; instead, a gradual decline in plasticity is to be expected, in the learning of language as well as other cognitive skills. Concerning language acquisition, the usage-based approach to language acquisition makes two fundamental assumptions (Tomasello, 2003, p. 3): 1. Children possess more than just the very simple learning mechanisms proposed by behaviorists like Skinner (i.e., isolated association and induction). 20 2. The adult endpoint of the acquisition process can be formalized in far less abstract terms than is the case in generative grammar accounts, a point that will be dealt with shortly here and which will be discussed in more detail in the section on Construction Grammar below. Since language structure is based on actual language use, the utterance (rather than sentence) as the most fundamental psycholinguistic unit of 18 Despite the strong Chomskyan influence, much work on language acquisition during the 1970s also had a strong cognitive-functional orientation; this is mirrored in research by, e.g., Bates (1976, 1979), Clark (1978), and Slobin (1973). In contrast to the usagebased, constructivist approach, explanations were driven by semantic and pragmatic considerations only, mostly leaving out the formal side of linguistic structures. 19 It has been noted that “although construction grammar and the usage-based model are in principle independent of each other, they are often combined in linguistic analyses” (Diessel, 2004, p. 13) - this also applies to this work, in which the two frameworks are viewed as complementary. 20 Although Skinner’s account of verbal behavior is now generally considered obsolete, it has seen a revival in research of treatment options for children with autism (e.g., Sundberg, 2001). <?page no="35"?> 21 language development and use is central in this approach, defined as “a linguistic act in which one person expresses towards another, within a single intonation contour, a relatively coherent communicative intention in a communicative context” (Tomasello, 2000, p. 63). 21 The idea of utterances as basic units can also be found in very early accounts of language acquisition, which view the production of single words by very young children as oneword utterances, also referred to as holophrases, since their function is equivalent to that of ‘whole’ sentences (Stevenson, 1893; Laguna, 1927). Furthermore, the pivot-grammar proposed by Braine (1963) is closely related to the usage-based notion that early language is very item-based and that language development proceeds in a piecemeal fashion. In this account, children’s early productivity is based on items consisting of a constant (the pivot) and a slot that can be filled with different words and thus allows for a limited degree of productivity (in other words: low-level abstractions). For example, the formula More X consists of the pivot More, which is always used in first position, and the variable slot X, which can be filled with a label for the desired object/ substance (for example cereal or cookie). Since the later version of pivot grammar grouped productive word-combinations according to their semantic bases (Braine, 1976), thereby stressing the importance of communicative function, it can be seen as a precursor to usage-based emergentist approaches. But even though Braine concluded that “children's syntactic competence is more limited and much more concretely based than current theories suppose" (Braine, 1976, p. 4), this approach had to “suppose some discontinuity in development, where a transition to a more adult type of linguistic system takes place” (Braine, 1976, p. 88). However, the conclusion that “you can’t get there from here” (Tomasello, 2003, p. 3) was drawn by Braine merely because empirical work on subsequent stages of development was missing at the time (Braine, 1976, p. 89). Along similar lines, other early studies suggest that children’s first multi-word utterances are mostly rote-learned strings or item-based phrase which merely act as a makeshift linguistic repertoire and are abandoned once abstract rules are mastered (Ewing, 1982; MacWhinney, 1978). 22 Based on research in cognitive psychology which focuses on human cognitive development and learning, a core claim of the usage-based approach to language development is that it is possible to get there from here since not only the initial stages but all of language development can be accounted for in terms of general learning mechanisms. So while “it made sense for researchers to explore the possibil- 21 The use of utterances as basic units for analysis goes back to research in conversation analysis (see, e.g., Schegloff, 1968), but the usage-based constructionist perspective is new in the sense that the utterance-level constructions in actual language use also serve as the basis for the associated theory of (mental) grammar. 22 One notable exception to this is a paper by Clark (1974), in which she states that child’s speech “becomes creative through the gradual analysis of the internal structure of sequences which begin as prepackaged routines” (p. 9). <?page no="36"?> 22 ity of a universal grammar at the time it was proposed (Chomsky 1965), when an understanding of the power of statistical learning and induction were a long way off” (Goldberg, 2009a, p. 203), recent findings suggest that early rote-learned formulas and item-based constructions serve as the starting point for a gradual development of linguistic competence on the basis of general learning mechanisms. The most important learning mechanisms are intention-reading and pattern-finding, both consisting of various related, domain-general skills and interacting in various ways during language development. While patternfinding skills are also found in other species, intention-reading skills are thought to be unique to humans, having their basis in a social-cognitive adaptation that evolved phylogenetically (Tomasello, 1999, 2005). This approach thus does not deny a biological foundation for language; it just assumes that this foundation does not take the form of an autonomous language faculty. 2.2.1 Intention Reading Between the age of 9-12 months children across cultures develop a set of social-cognitive abilities which allow them to communicate symbolically (Tomasello, 1995), with intention-reading as the most important and prerequisite skill. At this point, children begin to understand that humans are intentional agents and that they use language (and also other communicative means like gestures) in order to manipulate the mental or intentional states of others (Tomasello, 2003, p. 3). For example, a few months before they start using their first words, children will follow an adult’s gaze or pointing gesture towards an event or object (Corkum & Moore, 1995; Meltzoff & Brooks 2006; Scaife & Bruner, 1975) and will also actively direct the visual attention of others to distal objects, for example by pointing (Bates, 1976, 1979; Butterworth & Cochran, 1980), showing the emerging ability to share attention to events or objects with others (Bakeman & Adamson, 1984). This makes it possible to establish a so-called joint attentional frame in which adult, child, and object/ event enter into a triadic interaction. 23 This collaborative engagement is essential for the child’s breaking into language and children whose skills of nonlinguistic joint attention emerge earlier also start acquiring language earlier (Carpenter, Nagell, & Tomasello, 1998, p. 109). Furthermore, studies have found that there is a high correlation between time infants spend in joint engagement with their caretakers and later vocabulary size (for example, Tomasello & Todd, 1983; see Tomasello, 1988, for a review). Crucially, the child can use the common ground not only to imitate an action but to recognize the intention behind the action and thus to 23 The idea that language learning takes place in interaction and is socially grounded is not new - e.g., the notion of ‘joint attentional frame’ goes back to Bruner’s ‘joint attentional format’ (Bruner, 1983, Chapter 4). <?page no="37"?> 23 participate in ‘role reversal imitation’, which is part of cultural learning (Tomasello, Kruger, & Ratner, 1993). Experiments have shown that children do not reproduce adult’s accidental actions, but only intentional ones (Carpenter, Akhtar, & Tomasello, 1998) and that they carry out an action an adult intends to perform, even if the adult is prevented from actually carrying it out (Meltzoff, 1995). Besides, there is experimental evidence that 14-montholds already have an understanding of and take into account shared experience in inferring an adult’s social intention: after engaging in the activity of cleaning up toys with one experimenter, either the known or an unknown experimenter - who had not participated in the activity - produced a pointing gesture towards an object. Children interpreted the point as belonging to the activity, i.e., a request to clean up the object, significantly more often in the shared experience than in the unshared condition (Liebal, Behne, Carpenter, & Tomasello, 2009). Word learning studies have shown that children are also good at identifying communicative intentions and are able to identify intended referents even if word and referent are not present simultaneously (Akhtar & Tomasello, 1996; Tomasello & Barton, 1994). Most of the time, however, children are faced with larger units in communication. While children can use their intention-reading skills in order to recognize the intention behind whole utterances, they also need to find out which roles are played by the individual parts of an utterance. This requires an additional set of skills, referred to as pattern-finding. 2.2.2 Pattern finding Pattern-finding skills are required in order to understand the grammatical dimension of language. This set of skills is domain-general and not unique to humans. 24 For example, some animal species have been found to be capable of categorical perception (Kuhl, 1987), indicating that they perceive incoming auditory stimuli in terms of categories of similar items rather than unique and unrelated instances, just like humans do (Eimas, Siqueland, Juszcyk, & Vigoroto, 1971). In a study with 8-month-olds using the preferential-looking technique (Saffran, Aslin, & Newport, 1996), it was shown that prelinguistic infants preferred to listen to an audio stream containing trisyllabic nonsense-words they had heard in a training session, indicating that they recognized the sequences from the training stream. The speech streams were presented in a continuous manner, meaning the infants must have been relying on transition probabilities between the syllables of the ‘words’, since no cues indicating boundaries (like prosody or pauses) were given. What is more, children are also able to abstract away from concrete patterns: Using similar proce- 24 The fact that language relies on domain-general abilities such as categorization has already been noted by Bruner (1983), who states that distinctions encoded linguistically “have analogues in the child’s way of ordering his world of experience” (p. 30). <?page no="38"?> 24 dures, Marcus (1999) and Gomez and Gerken (1999) showed that 7-montholds preferred to listen to a stream exhibiting the same patterning of syllables (e.g., ABB) also in cases in which the syllables used differed from the ones in the training session. These and other studies suggest that children employ statistical learning (such as distributional analysis) and that they are able to form perceptual and conceptual categories grouping together similar auditory and visual stimuli. 2.2.3 Analogy-Formation and Schematization On the basis of the systematic variation in the linguistic input, the interplay of intention reading and pattern finding gives rise to schematization: For example, a child hearing many instances of the type Where is X? will notice that the overall function of the utterance (i.e., looking for something) stays constant, while the thing being searched for varies across situations. These slot-filler categories (Nelson, 1985) show how early abstraction processes are not only made possible but even made easier by a high degree of repetition in the input. Based on these early abstractions, children can form analogies by recognizing functional similarities between items which do not share (as much) concrete linguistic material. For example, a child may recognize that both The girl is kicking the ball and The boy is eating an apple have the structure A is B-ing the C, with A being the actor, B denoting the activity and C playing the role of the undergoer (Tomasello, 2006, p. 287). Eventually, paradigmatic categories like nouns and verbs may be learned through ‘functionally based distributional analyses’, i.e., by using functional and distributional cues to group together items that fulfill similar communicative functions in a variety of constructions. Crucially, the claim is that only the combination of functional AND distributional cues leads to the emergence of adult-like categories, since there is only an overlap but not a perfect correlation between semantic (e.g., ACTION) and syntactic categories (e.g., VERB); for example, Tomasello (2005, p. 194) explicitly states that connectionist models only taking into account purely formal distributional properties do not fully capture the abstraction process. The processes enabled by (the interplay of) intention reading and pattern-finding are thought to be powerful enough to develop adult-like linguistic skills, making it unnecessary to posit innate linguistic knowledge. 2.2.4 Learning from the input Within the usage-based approach, the socio-cultural embeddedness of language use is seen as the driving force in language development, with linguistic representations emerging in a bottom-up manner. In other words, abstract grammatical knowledge is thought to be a result of rather than a necessary prerequisite for the language learning process. Since children only slowly abstract away from concrete exemplars and require a large amount of <?page no="39"?> 25 input, Tomasello questions the nativist assumption that language development is rapid, stating that it “takes many years of daily interaction with mature language users for children to attain adult-like skills, which is a longer period of learning with more things to be learned - by many orders of magnitude - than is required by any other species on the planet” (Tomasello, 2003, p. 2). Furthermore, in contrast to the notion of impoverished input found in nativist accounts, the linguistic environment is characterized as very rich, both quantitatively and qualitatively. Estimating that a child is exposed to language about ten hours per day, she will have heard millions of words by the time she turns five. 25 In nativist approaches, input frequencies are not thought to be of major importance - particular structures in the input merely have to occur with a certain frequency in order to trigger the (re)setting of a parameter (e.g., Gibson & Wexler, 1994). 26 In contrast to this, usage-based approaches assume that the linguistic environment provides all the positive and indirect negative evidence necessary to learn the target language(s) and that the input shapes the language used by the child and also the emerging linguistic representations (e.g., Theakston, Lieven, Pine, & Rowland, 2004, 2005; Rowland, Pine, Lieven, & Theakston, 2003). Even though it has been suggested that features which are characteristic of childdirected speech (at least in most Western cultures) may facilitate language development, it should be noted that it is not deemed necessary for language learning and that this register is not found in all language communities (for a review on child-directed speech, see Lieven, 1994). Within this framework, cross-linguistic differences can easily be explained by the different linguistic environments and the linking problem simply does not arise, since categories are not innate but emerge during language development (Tomasello, 2005). 27 In order to address developmental changes in a psycho- 25 This is, of course, just an estimate, and studies show that there is a considerable amount of individual variability (Hart and Risley, 1995, 1999; Huttenlocher, Haight, Bryk, Seltzer, and Lyons, 1991). Based on their study on the child-directed speech by 12 mothers, Cameron-Faulkner, Lieven, and Tomasello (2003) conclude that Englishspeaking children are exposed to roughly 5000 to 7000 utterances a day. The densest corpus project to date, called the Speechnome project, provides a nearly complete coverage of a child’s linguistic environment from 9-24 months; early estimates suggest that the total number of words (spoken by and to the child) amount to about 10 million for this 15 month-period alone (Roy, 2009). 26 The generative position on this is made very explicit by Crain, Thornton, and Khlentzos (2009) who state that “experience plays a critical role in determining a learner’s core grammar, because parameters are fixed on the basis of experience. Experience also matters in learning the idiosyncratic linguistic patterns of the local language, i.e., the periphery. It is unlikely, however, that the ‘universal, or innate principles or constraints’ of core grammar are acquired through experience” (Crain, Thornton, and Khlentzos, 2009, p. 146). 27 This should not be taken to indicate that there are no linguistic universals, but they are attributed to characteristics of human cognition, culture, and communication rather than to an innate language faculty (Tomasello, 2003, pp. 18-19). <?page no="40"?> 26 logically plausible way, the nativist continuity assumption is abandoned in favor of the developmental assumption, proposing “that whereas the processes working at different developmental stages are constant, the actual structures and representations are different at these different stages” (Tomasello, 2003, p. 324). From this follows that although there is a high degree of homogeneity of mental grammars within a given linguistic community, they may not be as uniform as posited by nativist accounts. This has already been indicated by Langacker more than two decades ago (1987, p. 376) and is borne out by experimental data showing that there is a great amount of variability within native speakers, depending on a number of factors like previous experience with particular structures (e.g., Chipere, 2003; also see Dąbrowska 2008, 2010, on questions with long-distance dependencies and Street & Dąbrowska, 2010, on passives and quantifiers). 28 Language development thus proceeds in a rather piece-meal fashion, with grammatical categories and more abstract representations only gradually emerging over time. Within the nativist framework, a child’s production of creative utterances she could not have heard in the adult input was taken to mean that she must be operating with abstract categories and rules. However, overgeneralizations which are of particular interest in the context of syntactic analyses, i.e., those involving non-canonical uses of verbs in argument structure constructions (such as She giggled me) are very rare below the age of 3; 0 (Pinker, 1989, pp. 17-26). Furthermore, observational as well as empirical studies show that early productivity is best accounted for in terms of lexically specific syntactic categories rather than abstract categories and across-the-board generalizations. Based on a diary study of his daughter’s early language development, Tomasello (1992) found that most of her speech was very item-based and her productivity rather limited. She seemed to be operating with sentence frames build around verbs and she did not readily generalize a structure she used with one verb to other verbs, leading Tomasello to formulate his Verb Island hypothesis. The strict version of the hypothesis has since been abandoned because it has been found that not only verbs, but also other lexical items (pronouns and pronoun-clitics) can serve as the basic building blocks for early sentence frames (Lieven et al., 1997; Pine & Lieven, 1993; Pine, Lieven, & Rowland, 1998) - still, the main hypothesis that children’s early linguistic productivity can be accounted for in item-based terms has been corroborated in subsequent research. A study using a combination of maternal diaries and naturalistic sampling (Lieven et al., 1997) found that 12 English-speaking children’s use of language from the age of 2-3 years was very item-based, with 92% of their multi-word utterances stemming from one of their first 25 lexically-based patterns. Experimental as well as corpus-based studies on languages other than English 28 For a comprehensive discussion of research on individual differences in native language attainment, see Dąbrowska (2012). <?page no="41"?> 27 have replicated the results concerning the item-based learning and use of language (for example, see Berman, 1993, for an experimental study on transitive verbs in Hebrew and Pizutto & Caselli, 1992, for a corpus-based analysis of verb uses in Italian). Still, it has been suggested that there may be differences concerning the degree of formulaicity in early language development, depending on the structure of the input (see, for example, Steinkrauss, 2009, for a case study of wh-question acquisition in German, which suggests that a higher degree of variation in the input leads to earlier abstraction in German compared to English). Of course, children occasionally do make overgeneralization errors as soon as they start using language productively but the point is that children are rather conservative language users; experimental evidence for this comes, for example, from a study showing that five-year-olds are less likely than both seven-year-olds and adults to generalize a novel argument structure construction after having been exposed to the same input (Boyd & Goldberg, 2012). In contrast to nativist accounts, within the usage-based framework their productivity is not constrained by formal rules but, for example, by the cognitive mechanism of statistical pre-emption, which is closely linked to the notion of entrenchment. 29 Entrenchment refers to the formation of cognitive (in this case linguistic) routines resulting from repeated exposure to similar exemplars the more often a structure is activated, the stronger its neural representation gets (Bybee, 1985, p. 117). Based on Braine (1988), Braine and Brooks (1995) suggested that children strongly prefer to use verbs in the argument structure constructions they have heard them used in, i.e., they stick to the uses that are entrenched. Children are, for example, not very likely to use a verb they have only heard used in intransitive sentences in transitive ones (preventing them from producing utterances like He disappeared the rabbit). Theakston (2004) proposed that children’s tendency to overgeneralize non-frequent rather than frequent verbs is due to the higher degree of entrenchment of the patterns associated with the frequent verbs. In reaction to this, Goldberg has pointed out that the high frequency with which a verb occurs in a particular construction cannot be the decisive factor, since there are many cases in which such a verb can also be used in other patterns without resulting in an ill-formed structure (Goldberg, 2009b, pp. 102-103); she suggests that the entrenched forms only constrain the use of functionally related forms. In other words, “once one linguistic form for expressing a meaning has been learned, it pre-empts other forms that express the same meaning, unless the language input offers positive evidence for a second form” (Braine & Brooks, 1995, p. 361). A possible competition between forms is thus resolved by relying on those forms which are attested in the input (also see MacWhinney, 1987, on the Competition Model). For 29 Other factors influencing linguistic productivity, such as type/ token-frequency and the degree of openness of a pattern, will be discussed in the subsequent section. <?page no="42"?> 28 example, the high number of instances of make someone laugh preempts the potential transitive to laugh someone (which would be functionally equivalent). In contrast to this, the high number of instances of the intransitive X sneezes does not preempt the generalization to the caused-motion construction She sneezed the napkin off the table, since it is formed in analogy to other instantiations of this construction and it is used to express a meaning which is distinct from the intransitive X sneezes. To sum up, within the usage-based approach the task faced by the child in language development is not the linking of language-specific surface strings to categories offered by UG but the gradual learning of meaningful structures of varying degrees of complexity and abstractness, also referred to as constructions, from the linguistic environment. Children are able to learn the constructions of their native language(s) using their social-cognitive and general cognitive skills, most notably intention-reading and pattern finding, which encompass the ability to categorize items based on formal or functional similarities, to form analogies, and to schematize. Cognitive processes such as entrenchment and preemption are thought to enable children to avoid or retreat from overgeneralization errors. Since all levels of language are thought to be learnable this way, ranging from simple lexical items to very complex and abstract structures, the usage-based theory of language acquisition is consequently based on a theory of (mental) grammar contrasting sharply with generative-transformational accounts. This framework, termed construction grammar, will be outlined below. 2.2.5 Construction Grammar In comparison with generative-transformational approaches, Construction Grammar is a relatively new development. It belongs to the field of Cognitive Linguistics (CL), which evolved in the late 1970s as an alternative to Chomskyan linguistics and was primarily established and expanded through work by Fillmore, Langacker, Talmy, and Lakoff (Fillmore, Kay, & O'Connor, 1988; Langacker, 1987, 1991; Talmy, 1988; Lakoff & Johnson, 1981, 1987). Lakoff was also one of the key figures in Generative Semantics (Lakoff, 1971, 1976), an attempt to integrate the study of semantics into a generative framework. Cognitive linguistics rejects the nativist notion that language is an autonomous faculty as well as the strong focus on (allegedly) purely formal properties of language, instead adopting the view that language is integrated with and based on the same mechanisms and representations as other cognitive abilities. Apart from that, it aims at a holistic account of language, with a strong interest in the establishment of meaning through linguistic expressions in interaction. Interdisciplinary in nature, CL was influenced by work in cognitive psychology in its early days (e.g., Rosch, 1975) and continues to make use of findings and methods from other disciplines within the cognitive sciences, aiming at integrating what is known <?page no="43"?> 29 about language with what is known about human cognition and social interaction. The 1980s and ‘90s saw a growing proliferation of research within CL, which is now established as a family of theories united by common principles and aims rather than a coherent theory, subsuming different but related approaches to semantics and grammar. 30 The dominant set of approaches to grammar is known as (a family of) Construction Grammar(s), including Croft’s Radical Construction Grammar (with a stronger focus on typology, Croft, 2001), Embodied Construction Grammar (Bergen & Chang, 2005), Construction Grammar as developed by Fillmore and Kay (Fillmore, 1988; Fillmore et al., 1988), Goldberg’s (Cognitive) Construction Grammar (Goldberg, 1995, 2006), and Langacker’s Cognitive Grammar (Langacker, 1987, 1991), which constitutes the most comprehensive account to date. 31 The current book is largely based on Goldberg while also referring to related accounts occasionally. The different approaches are seen here as mutually informative perspectives rather than mutually exclusive alternatives, since they focus on different aspects of language structure and use while sharing many central assumptions (for an overview of common properties, see Goldberg, 2006, p. 215). She also notes that “the (mostly minor) differences among various cognitive, functionalist or usage-based approaches pale in comparison to the stark contrasts between these approaches and traditional generative grammar” (Goldberg, 2009a, p. 219). Indeed one can view Goldberg’s Cognitive Construction Grammar as being based on Langacker’s work, which is especially interested in the semantic dimension of language use. In contrast to generative and other traditional approaches, constructions are not seen as purely formal entities without meaning (‘taxonomic ephiphenomena’, Chomsky, 1989, p. 43). Instead, it is established that grammar has a symbolic dimension, which in generativism is only attributed to the lexicon. Consequently, no sharp division is assumed between grammar and lexicon and the notion of construction is extended to include all levels of linguistic description. The existence of a grammar/ lexicon continuum was first suggested by Fillmore, Kay, and O’Connor (1988) in their study on the formal (i.e., schematic) idiom let alone. In contrast to so-called substantive idioms, which are lexically filled and compositional in the sense that their meaning can be derived from their component parts (e.g., all of a sudden, kith and kin), sche- 30 See Evans (2006) and Barcelona and Valenzuela (2011) for more comprehensive overviews and further references. 31 The latter is sometimes seen as a separate theory of grammar within in a cognitive framework (like Richard Hudson’s Word Grammar; Hudson, 1990), but can also be seen as a type of Construction Grammar. Goldberg even goes as far as using the terms Construction Grammar and Cognitive Grammar interchangeably (Goldberg, 1998, p. 205) and Langacker notes that “there is less disagreement between CG [Cognitive Grammar] and CCxG [Goldbergian Cognitive Construction Grammar] than meets the eye” (Langacker, 2009, p. 174). <?page no="44"?> 30 matic idioms (like let alone, but also the X-er, the Y-er) possess semantic and pragmatic characteristics that cannot be derived this way (Fillmore et al., 1988, p. 505). They point out that schematic idioms cannot be assigned to the lexicon because they are not fully lexically fixed and exhibit (limited) productivity. Neither is it possible to assign them to the grammar, since they contain fixed lexical material and their syntactic behavior cannot be captured by (and in some cases runs counter to what is predicted by) general rules. They conclude that “in the construction of a grammar, more is needed than a system of general grammatical rules and a lexicon of fixed words and phrases” (Fillmore et al., 1988, p. 534). The assumption that lexicon and grammar form a continuum also has consequences for the learnability of linguistic knowledge generative approaches assign idiomatic expressions to the lexicon and assume that everything in the lexicon can be learned using general learning mechanisms. In the case of schematic idioms, this includes syntactic, semantic, and pragmatic idiosyncrasies. If this complex task of learning about items belonging to the linguistic periphery (i.e., the lexicon) can be mastered without the help of innate linguistic knowledge, it is only logical to conclude that this should also be the case for learning the core of linguistic competence. In other words, “if the child has to learn the irregular and peculiar aspects of a language by general learning mechanisms, these mechanisms should suffice to learn the more general and predictable patterns of that language as well” (Behrens, 2009c, p. 384). Central in a Construction Grammar framework is the description of all linguistic structures as constructions, which are defined as “learned pairings of form with semantic or discourse function, including morphemes or words, idioms, partially lexically filled and fully general phrasal patterns” (Goldberg, 2006, p. 5). 32 In contrast to Goldberg’s earlier account (Goldberg, 1995), this more recent definition also views fully compositional patterns as constructions “as long as they occur with sufficient frequency” (Goldberg, 2006, p. 5). This is based on psycholinguistic evidence suggesting that frequently occurring patterns are stored as fixed units (e.g., Bybee, 1995, and Barlow & Kemmer, 2000). Importantly, this happens regardless of whether the structure is fully compositional: Research on prefabricated word combinations such as beyond repair has shown that speakers recognize them as familiar, meaning that they are “stored in memory despite being largely predictable in form and meaning” (Bybee, 2006, p. 713). The strong influence 32 In Goldberg’s view, “it’s constructions all the way down” (Goldberg, 2006, p. 18), i.e., all linguistic knowledge can be accounted for in terms of constructions, while Langacker reserves the term ‘construction’ for composite (i.e., complex) structures. Langacker sometimes uses the term ‘symbolic assemblies’ rather than constructions to stress that composite structures should not be thought of as being constructed out of component structures in Langacker’s words: “stepping-stones are not the same as building blocks” (Langacker, 2008, p. 166). <?page no="45"?> 31 of frequency of occurrence on the formation of linguistic representations is a central assumption within Construction Grammar (CxG). On a general level, input frequency has been shown to influence children’s language development (Theakston et al., 2004; Theakston et al., 2005; Rowland et al., 2003; for a review on the prevalence of frequency effects in first language development, see Ambridge et al., 2015) as well as adult’s linguistic performance (Bybee & Hopper, 2001), leading to the conclusion that “frequency affects linguistic representations at all points in development” (Abbot-Smith & Tomasello, 2006, p. 279). However, many issues pertaining to frequency are still unresolved. Crucially, there is no agreement on what constitutes a ‘sufficient’ number of occurrences for a structure to achieve unit status, i.e., individual cognitive representation. This is partly due to the fact that research explicitly investigating the influence of potentially relevant factors such as “construction type, lexical and construction frequencies, sentence length and complexity” (Gurevich, Johnson, & Goldberg, 2010, pp. 72-73) is still largely missing. Furthermore, it is very likely that additional factors such as attention, previous linguistic knowledge and perceptual as well as functional salience affect the processing of linguistic structures. Gries (2008, p. 424) suggests that research discussing independent evidence for the unit status of certain constructions may offer interesting insights. He points out the study by Bybee and Scheibman (1999), in which it is shown that the string I don’t know, for example, undergoes phonological reduction, thus providing independent evidence for its status as a fixed unit. The basic architecture of constructions is illustrated in Figure 2.4. 33 The inclusion of different types of formal as well as functional characteristics allows for a comprehensive description of all linguistic structures encountered within one coherent framework. Usually, only a few of the properties are usually focused on, depending on which characteristics are deemed necessary for a comprehensive analysis of a particular (type of) construction. 33 The figure is an adapted version of the one found in Croft (2001, p. 18). <?page no="46"?> 32 It is particularly striking that so far only a few studies have taken into account the contribution of (suprasegmental) phonological properties such as prosody. For example, Fried & Östman (2004, p. 20) state that structures like Is the sauna hot are ambiguous, since they can be instances of genuine questions or exclamations and they suggest that we are in fact dealing with two different constructions differentiated (mainly) by prosodic features. 34 This example shows that the sequential ordering of elements is in some cases not sufficient for the description (and identification) of a yes-no question. However, the contribution of prosody and consequently the correspondence between utterance type and intonation is not that straightforward - in his account of the so-called ‘Mad magazine construction’ (also known as Incredulity Response Construction, e.g., Him be a doctor! ? ), Lambrecht (1990, p. 225) points out that while sentence fragments such as A job? can be uttered with a rising intonation in order to express incredulity, the exact same structure can also be used for other pragmatic purposes, e.g., to express a request for clarification. Constructions range from very simple and (lexically and semantically) concrete items like monomorphemic words to very complex and abstract structures like passives or ditransitives. The notion of abstractness may refer to both form and function, i.e., to lexical as well as semantic specificity - usually, the two characteristics are correlated, since constructions which are (partially) lexically filled tend to exhibit more specific semantic properties than constructions which are not lexically specified at all. Examples of various constructions are provided in Table 2.1 (based on Goldberg, 2006, 34 A similar observation has been made by Huddleston, who states that “in abstraction from their prosody, ... Don't you tell her [is ambiguous] between interrogative and imperative, How much remains to be done between interrogative and exclamative, and so on” (Huddleston, 1994, p. 412). Figure 2.4: The structure of linguistic constructions <?page no="47"?> 33 p. 5), with both complexity and abstractness/ schematicity increasing from bottom to top: 35 Table 2.1: Different types of constructions type of construction example passive Subj aux VPpp (PP by ) 36 ditransitive (double object) Subj V Obj1 Obj2 covariational conditional The Xer the Yer idiom (partially filled) jog <someone’s> memory idiom (filled) it takes one to know one complex word eyeball word ball morpheme -ing Crucially, since complex and schematic constructions like the ditransitive or the passive are based on (and have the same basic structure as) their actual instantiations, they are represented (in analysis, but potentially also in the speaker’s mental grammar) at multiple levels simultaneously. There is a considerable degree of heterogeneity concerning the different terms for various points along the simple/ complex and concrete/ abstract continuum, respectively - Table 2.2 shows three different levels of the ditransitive (double object) construction and provides the terms most frequently found in the literature to denote the respective level of abstractness. As already outlined in the previous section, children are thought to master these different levels by slowly abstracting away from rote-learned, invariant formulas and item-based constructions in a bottom-up manner, enabling them to gradually achieve adult-like productivity. 35 Note that relatively simple constructions can also be abstract (e.g., N-s for regular plurals), while complex constructions can also be concrete (e.g., lexically filled idioms such as the already mentioned It takes one to know one). 36 Even though conventional labels such as ‘subject’ are used by Goldberg and also here for ease of presentation, I follow the stance most explicitly formulated in Croft’s Radical Construction Grammar that “there are no formal syntactic categories such as ‘subject’ or ‘object’ per se” but that they are “derived from constructions” (Croft, 2005, p. 276); for a more comprehensive discussion, see Croft (2001, Chapter 1). <?page no="48"?> 34 Table 2.2: Different levels of the ditransitive (double object) construction terminology based on Goldberg (2006) alternative terms Subj V Obj1 Obj2 grammatical construction/ higher-level constructional schema Subj give Obj1 Obj2 formulaic (lexically-based) frame/ item-based construction/ low-level constructional schema Mummy gave me a kiss. invariant formula/ fixed phrase (potentially rote-learned) 2.2.5.1 Type and token frequencies Since statistical learning is held to play a crucial role in this process, different types of frequency of occurrence are of particular interest. Usage-based, CxG accounts differentiate between type and token frequency. 37 Token frequency is closely linked to the mechanism of entrenchment introduced in the previous section. Type frequency, on the other hand is linked to the abstraction processes of analogy and schematization, which enable linguistic productivity. Within CL, type frequency was initially applied to the analysis of (inflectional) morphology in English (e.g., Bybee, 1985, 1995); e.g., the total number of instances of the irregular past tense broke (= its token frequency) is quite high, the number of different verbs exhibiting the same pattern (= its type frequency) is relatively low (e.g., spoke, rode,…), while the regular past tense exhibits the reverse pattern (i.e., a low token but high type frequency) indicating that the irregular form is highly entrenched as a fixed unit whereas the regular pattern is very productive (Bybee, 2004, pp. 604-605). The notion of type-token ratio has also been adapted as a measure of productivity in the analysis of more complex constructions - type-frequency then refers to the number of different items that appear in a given slot. It has been suggested that a high type-token ratio promotes the extraction of more abstract constructional schemas and thus productivity in language learning. However, the relationship between type-token-frequency and the ability to generalize from the input is not that straightforward: In a series of experiments, Casenhiser and Goldberg (2005) and Goldberg, Casenhiser, and Sethuraman (2004) found that - holding overall type and token frequencies constant - skewed 37 It is important to keep in mind, however, that there are other types of frequencies whose relevance must be evaluated (such as relative vs. absolute frequency of occurrence). <?page no="49"?> 35 input (with one type occurring in a disproportionally large number of instances) facilitated the learning of a novel construction in both child and adult learners compared to balanced input (with each type contributing to about the same proportion of tokens). Furthermore, it has been found that presenting the same amount of exemplars of a construction in a distributed rather than massed manner during training sessions leads to better learning outcomes (Ambridge, Theakston, Lieven, & Tomasello, 2006). Apart from that, it is important to note that the predicted productivity of a pattern also depends on additional factors like potential pre-emption by competing structures and a pattern’s degree of openness (Bybee, 1995, p. 430; Goldberg, 2006, pp. 93-94); the latter refers to the fact that the degree of (semantic or phonological) similarity between attested and potentially new instances will influence a language user’s likeliness to generalize (‘conservative induction’ Brooks & Tomasello, 1999; see Goldberg, 2009b, p. 104 for a short discussion). The claim that both the relationships between constructions (at different levels of abstraction) and their instantiations as well as between different constructions are of importance derives from the assumption that constructions are organized in a network-like structure. In the subsequent section, I will elaborate on this by discussing the different suggestions that have been made concerning 1. the storage of information on the different levels of abstraction and 2. the nature of links between formally and/ or functionally related constructions. 2.2.5.2 The usage-based model Within constructionist approaches, suggestions concerning the storage of linguistic knowledge at different levels of the taxonomic hierarchy range from a full-entry model, in which information about each individual instance is retained in memory and abstractions are mainly created ad-hoc and on-line (Goldberg, 1995, p. 74) to a Complete Inheritance Model, in which information is stored non-redundantly only at the highest-possible level of the constructional network, with subordinate structures inheriting features while (detailed) knowledge of individual exemplars is not retained (Fillmore, 1999). 38 An intermediate position is occupied by the default inheritance model, in which all instances inherit their features from their dominating construction, unless it is overridden by more idiosyncratic information at lower levels. Goldberg (2009b, p. 99) notes that the discussion about the representation of linguistic knowledge bears a striking similarity to the exemplar-based vs. prototype-models developed in research on nonlinguistic categorization (Anderson, 1991; Murphy, 2002; Ross & Makin, 38 Note that in any case the knowledge retained about occurring exemplars is selective, so that there is always “a degree of abstraction over the actual input” (Goldberg, 2006, p. 46); otherwise, the formation of categories would not be possible, since all usage events are unique at a very fine-grained level. <?page no="50"?> 36 1999). Recently in both fields a preference for a usage-based approach has developed, which is a kind of “hybrid model comprising both abstractions and the retention of the exemplars of which those abstractions are composed” (Abbot-Smith & Tomasello, 2006, p. 276), allowing for even more redundancy than the default inheritance model, since common features are shared with rather than merely inherited from dominating constructions. It is adopted by Goldberg (Goldberg, 2006) but also Langacker (1987; 1991) and Croft (2001), based on the assumption that “redundancy is not to be disparaged, for in one way or another every language makes extensive use of it” (Langacker, 2008, p. 188). The usage-based model is also mirrored in the inclusion of frequently occurring, fully compositional structures as constructions. The assumption that it is preferable to choose between itemspecific and very general knowledge is referred to by Langacker (Langacker, 1987, p. 29) as the ‘rule-list fallacy’ of the generative words-and-rules model (Pinker, 1999). It has been noted that this nonreductive approach to linguistic structure lacks the parsimony and mathematical elegance aimed at (and indeed often provided) by generative approaches - however, the goal of usage-based constructionist approaches is to provide a psychologically plausible description of linguistic knowledge, tying in with what is known about memory in general (Langacker, 1987, p. 56). Research on cognitive domains other than language has shown that information is likely to be stored redundantly and it has been suggested that the brain has a nearly infinite storage-capacity so that there is no need to minimize the amount of stored information (Schneider & Bjorklund, 1998, p. 469). This is corroborated by studies indicating that humans possess an extremely powerful implicit and explicit verbatim memory - for example, in a series of experiments conducted by Gurevich, Johnson, and Goldberg (2010), participants recognized and recalled full sentences that they were exposed to only once and they reliably reproduced sentences they had heard six days before in a scene-description task that did not involve explicit recall. Apart from that, the intuitively plausible assumption that “it is easier for us to look something up than it is to compute it” (Bresnan, 1978, p. 14) is also backed up by evidence suggesting that the storing of complex expressions facilitates processing. Bod (2001) found that when participants were confronted with either high-frequency sentences (e.g., I like it) or low frequency sentences (e.g., I keep it), they were quicker in deciding on the well-formedness of highfrequency sentences, analogous to findings on frequency effects in lexical decision tasks (Gerhand and Barry, 1999). This suggests that high-frequency multi-word utterances, regardless of their compositionality, are also stored in memory and it has been estimated that speakers have ‘hundreds of thousands’ ready-made prefabs available (Pawley & Syder, 1983, p. 192; for a very comprehensive treatment of formulaic language, see Wray, 2002, and Schmitt, 2004; for a summary of recent research on formulaic language, see Wray, 2012). Apart from allowing for faster processing, entrenched chunks <?page no="51"?> 37 may also play a role in the categorization of new usage-events: A corpusbased study by Zeschel (2009) suggests that speakers store frequent chunks such as good for a laugh as units and that less typical variations (e.g., good for a cheer) can best be accounted for as analogical extensions of the entrenched (and thus prototypical) exemplar rather than instances generated by a more abstract schema. The usage-based model also entails a specific view of language development: Since linguistic information is assumed to be stored redundantly in adults, children are not expected to abandon concrete, item-based constructions as soon as they have abstracted away from them. Instead, item-based constructions are retained while more abstract constructional schemas are extracted and primarily used in the mental organization of linguistic knowledge (for experimental evidence suggesting that adult speakers store and use multi-word units, see Arnon & Snyder, 2010). What is more, specific knowledge is even thought to take precedence over more general knowledge, i.e., “units at lower hierarchical levels will generally prevail” when it comes to “categorizing a usage-event or computing a novel expression” (Langacker, 1987, p. 414). 39 In sum, “humans organize their knowledge into patterns while retaining an impressive amount of specific information” (Goldberg, 2009b, p. 99). This is also in line with research on structural priming in children and adults: While the reliance on item-specific knowledge (in the form of lexical overlap) decreases with age (Savage, Lieven, Theakston, & Tomasello, 2003), it still influences priming behavior in adults (Pickering & Branigan, 1998), suggesting that both knowledge of item-based constructions and of more abstract representations is present. 40 Since both very broad generalizations and idiosyncrasies can be captured, constructionist approaches achieve the descriptive adequacy demanded of a valid theory of language (by Chomsky, 1965, pp. 26-27). Explanatory adequacy is also achieved to a considerable degree, albeit in a different way than Chomsky presumably envisioned (i.e., focusing less on purely formal and more on functional explanations; for a short but illumi- 39 On a general note, Langacker emphasizes that actual instantiations rather than broad generalizations should be the focus of analysis: “Lower-level schemas, expressing regularities of only limited scope, may on balance be more essential to language structure than high-level schemas expressing the broadest generalizations. A higher-level schema implicitly defines a large ‘space’ of potential instantiations. Often, however, its actual instantiations cluster in certain regions of that space, leaving other regions sparsely inhabited or uninhabited altogether. An adequate description of linguistic convention must therefore provide the details of how the space has actually been colonized” (Langacker, 2000, p. 29). 40 An alternative explanation would be that there are connections between lexical items and more complex abstract representations which can themselves be primed (Pickering and Branigan, 1998, p. 646). <?page no="52"?> 38 nating discussion of what constitutes an ‘explanation’, see Goldberg, 2006, pp. 11-12). It is important to keep in mind that even though most studies aim at psychological plausibility by basing their claims on findings from experimental (production or comprehension) studies or corpus data, the level of granularity chosen for analysis may not correspond to the level of representation used by a speaker in actual language processing (Imo, 2009). Furthermore, the question is not so much at which level of abstractness a construction will be represented but rather which level of granularity is relevant in which situation of language use. Since information is stored redundantly on different levels of linguistic representation, speakers probably shift between these levels continuously, being influenced by a variety of factors such as their prior linguistic experience and the recent activation of particular constructions within the current discourse situation (for a comprehensive discussion of the emerging and emergent nature of constructions, see the contributions in Auer & Pfänder, 2011). In this process, the second aspect mentioned above (i.e., the nature of links between constructions) comes into play: Since constructions are organized in a symbolic network, formally and/ or functionally similar constructions are perceived as related and may therefore prime one another. The exact nature of the different relationships between (parts of) constructions is still a matter of debate. However, it is to be expected that the same associative links that exist between lexical items can also be found at the level of more complex constructions, which is mirrored in the nature of the links posited between constructions. 2.2.5.3 Relationships between constructions One of the most comprehensive treatments of the relationships that may hold between (utterance-level) constructions is offered by Goldberg (1995), where she proposes the existence of four different types of inheritance links. It is important to note that the original account is based on the so-called Principle of Maximized Economy, according to which “the number of distinct constructions is to be minimized as much as possible” (Goldberg, 1995, pp. 67-68) and the mechanism of default inheritance is entertained in order to achieve efficiency of storage (Goldberg, 1995, p. 72). Even if the aim of (local) parsimony is abandoned in favor of a usage-based model, inheritance links are still a useful means of capturing similarities between (formally) related constructions. Since the proposed structure of the constructional network is based on general cognitive principles of organization (e.g., the use of metaphor and metonymy, Lakoff, 1987), a motivation for the existence of particular constructions is offered and the posited relationships are made psychologically plausible (in how far they are psychologically real is a different matter - see, e.g., Sandra & Rice, 1995, for a discussion of polysemy at the lexical level). <?page no="53"?> 39 Goldberg (1995, pp. 75-81) describes four major types of inheritance links that make explicit the relationships between different (parts of) constructions which are related on the formal level: “polysemy links, metaphorical extension links, subpart links, and instance links” (Goldberg, 1995, p. 75). Polysemy links provide the connection between the (assumed) central sense of an argument structure construction and related senses. For example, while the central sense of the ditransitive construction (‘X CAUSES Y TO RECEIVE Z’, Goldberg, 1995, p. 75) is instantiated by a prototypical sentence like Mommy gave me the ball, a sentence like Mommy baked me a cake is an instantiation of the related sense ‘X INTENDS TO CAUSE Y TO RECEIVE X’ (Goldberg, 1995, p. 75). On a schematic level, the formal side of the construction remains the same (Subj V Obj1 Obj2) while on the functional side, the semantic features of the central sense are extended so that “each of the extensions constitutes a minimally different construction, motivated by the central sense” (Goldberg, 1995, p. 76). An example of a subpart link (also: part-whole link) between constructions is the relationship between the caused-motion construction (V SUBJ OBJ OBL, e.g., Bob kicked the ball across the lawn) and the intransitive-motion construction (V SUBJ OBL, e.g., The ball rolled across the lawn). Both the syntactic and semantic specifications of the intransitive-motion construction are a subpart of the caused-motion construction (with the external cause of movement present in the latter, but not the former), leading to the treatment of the intransitive-motion construction as a related but independent construction. Instance links were originally posited “between constructions iff [= if, and only if] one construction is a more fully specified version of the other” (Goldberg, 1995, p. 79). For example, there is a specific sense of the verb drive to which is only found in the resultative construction - while it is possible to drive someone crazy/ bonkers/ mad, it is not possible to *drive someone silly/ angry - for this specific sense, the existence of a partially lexically filled ‘drive-‘crazy’-construction’ is posited. Within a usage-based model, which Goldberg adopts in her more recent work (e.g., Goldberg, 2006), instance links can also be seen as connections between different levels of granularity or lexical specificity in cases where there is no specific sense; e.g., “Where's X? is an instance of the where-question construction, and perhaps a whquestion construction more generally” (Tomasello, 2003, p. 106). Lastly, metaphorical extension links capture the connection between constructions related by metaphorical mappings. For example, the resultative construction (26a) is argued to be a metaphorical extension of the causedmotion construction (26b): (26) a. She kicked him unconscious. b. She kicked him out of the house. <?page no="54"?> 40 The actual goal (i.e., a location) in (26b), expressed by the prepositional phrase (PP) out of the house, is extended metaphorically to the goal (i.e., a state) in the resultative construction, expressed by the adjectival phrase unconscious. Since the (metaphorical) goal is already encoded in the adjectival phrase, the addition of a (directional) PP is not permitted (for an extensive discussion of this analysis, see Goldberg, 1991): c. * She kicked him unconscious out of the house. It is important to note that these links do not constitute the only connections between constructions: There are semantically related utterance-level constructions which do not share syntactic structure and are thus not connected via an inheritance link but rather via semantic associations. For example, Hare & Goldberg (1999) have shown that provide with-structures (e.g., His editor credited Bob with the hot story) can be used to prime ditransitives (e.g., His editor offered Bob the hot story), which is most likely due to similar assignments of semantic roles in the two constructions. Furthermore, it is likely that there are not only connections between abstract schemas and their instantiations and between different utterance-level constructions, but also between lexical items and the more complex constructions they appear in (see, e.g., Gries and Stefanowitsch, 2004; Stefanowitsch and Gries, 2003, on collostructional analysis, a procedure used to measure the strength of association between constructions and lexemes). 41 To sum up: Based on the assumption that “our knowledge of linguistic constructions, like our knowledge generally, forms an integrated and motivated network” (Goldberg, 2009b, p. 98), there are some proposals concerning the nature of links between formally and/ or functionally related constructions. However, the network metaphor is still an underdeveloped concept and more research is needed in order to assess the different associative links between constructions. One way of exploring the relationships between constructions is to analyze families of constructions sharing certain formal properties. A prime example of this is Goldberg’s (2006) comprehensive account of constructions exhibiting SAI, which I will discuss in some detail since SAI is still often assumed to be the core feature of yes-no questions. Apart from that, the criticism Goldberg’s proposal received helps point out general issues which have been viewed as problematic within usage-based, CxG accounts, including areas which are still underresearched, thus paving the way for my own empirical investigations presented in this book. 41 For a recent elaboration of the network metaphor of usage-based construction grammar, see Diessel (2015). <?page no="55"?> 41 2.2.5.4 Subject-Auxiliary Inversion Within the generative paradigm, questions are of primary interest because they are assumed to be derived from more basic structures via transformations or - within more recent developments - other general syntactic operations involving the copying/ moving of constituents. In contrast to this, questions have not played such a prominent role in constructionist accounts, since all linguistic structures - regardless of their complexity and abstractness - can be captured within the same mono-stratal, non-transformational framework. There are mainly two reasons why questions have nevertheless received a considerable degree of attention in constructionist work: First of all, the learning of (canonical) yes-no questions has been used as a ‘test case’ for usage-based constructionist approaches, exploring the possibility that the learning of structures presumably relying on very abstract formal rules can also be accounted for within a non-transformational framework this point will be discussed more comprehensively in the next chapter. Secondly, syntactic questions have been discussed in order to suggest that even allegedly purely formal syntactic generalizations (such as SAI, which is not only used in the formation of questions) can be shown to be motivated functionally, i.e., by their deviance from prototypical declarative sentences. Previous work - also within CxG (Fillmore, 1999) - had assumed no functional generalization to be associated with the syntactic SAI-pattern, i.e., it was seen as a case of constructional ambiguity (e.g., Green, 1985; Newmeyer, 2000). In contrast to this, Goldberg presents a functionally-motivated analysis of constructions exhibiting SAI, viewing them as cases of constructional polysemy (Goldberg, 2009b, p. 111; Goldberg & Del Guidice, 2005, 2006). It should be noted that while the conventional term SAI is used in CxG accounts, implying some sort of movement operation, SAI constructions are not seen as being derived from structures with declarative word order. Rather, the word order is specified directly in the constructions exhibiting SAI. <?page no="56"?> 42 In total, there are nine different types of constructions containing SAI (based on Goldberg, 2006, p. 166): (1) yes-no question Was she home? 42 (2) (non-subject) wh-question What did she say? (3) counterfactual conditional Had she been home, they would not have punished her. (4) initial negative adverbs Seldom had she been home… (5) wishes/ curses May she never find true love! (6) exclamatives Boy, was she good! (7) comparatives He was home sooner than was she. (8) negative conjunct Neither does she go home regularly. (9) positive rejoinder So does he. Goldberg claims that there is a synchronic functional motivation behind all these constructions which “naturally form a coherent functional category of the type familiar from general categorization and lexical concepts” (Goldberg, 2006, p. 167). This implies that the generalization will not be captured in terms of (necessary and sufficient) features shared by all members of the functional category but rather in terms of conventional extensions from the prototype. Based on previous work by Lakoff and Brugman (1987), Lambrecht (1994), Michaelis and Lambrecht (1996), and Diessel (1997), Goldberg suggests that all SAI constructions deviate from the prototypical sentence, which is seen as an independent, “declarative, positive assertion with predicate-focus information structure” (Goldberg, 2006, p. 168). The most prominent feature is non-positive polarity, which is present in the majority of cases. 43 A non-prototypical sentence thus functions as the prototype for SAI constructions, with each of the nine types of SAI constructions sharing only a subset of the prototypical attributes, giving rise to a radial network. In addition to the four types of inheritance links mentioned above, markedness links are suggested, in these cases linking the prototype to the extensions (Goldberg, 2006, p. 178). This is due to the fact that we do not encounter instances of the prototype ‘non-prototypical sentence’, i.e., the core of the 42 Goldberg subsumes both yes-no questions and non-subject wh-questions under one type, which mirrors the fact that they are viewed as very closely related. I would like to argue that the two types of constructions are formally and functionally quite different and therefore treat them as two different types. Furthermore, tag questions are not mentioned by her at all, even though they are used quite frequently and differ from yes-no questions in terms of complexity and the sequential ordering of the elements. 43 While she attributes non-positiveness to seven out of the nine different SAI constructions (excluding positive rejoinders and comparatives) (Goldberg, 2006, p. 179), she later revises the status of comparatives by stating that “[i]t is only possible to get SAI in comparatives when the main clause is positive—the positive main clause yields a negative entailment in the lower, SAI clause”, thus rendering comparatives with SAI also non-positive (Goldberg, 2009b, p. 112). <?page no="57"?> 43 category is a generalization without direct instantiations; consequently, the category of SAI is seen as contrasting with the prototypical sentence in terms of form (SAI) and function (non-prototypical), i.e., SAI constructions are marked. For example, yes-no questions are non-declarative and non-positive, since they do not presuppose or assert that the proposition is true. Exclamatives are not directly linked to the SAI prototype but only indirectly so through a link to ‘yes-no questions’. This analysis is based on the observation that exclamatives (27) are closely related to rhetorical questions (28), which - in Goldberg’s view form a subtype of yes-no questions (29) (Goldberg, 2006, p. 173). (27) Boy, are you hungry (or what? )! (28) Are you hungry or what? (29) Are you hungry? The fact that “many exclamative utterances are at the same time rhetorical questions” (Goldberg, 2006, p. 173) and that 13% of exclamatives extracted from Google included the phrase or what is taken as evidence that “speakers synchronically perceive a relationship between exclamatives and questions” (Goldberg, 2006, p. 174). For a more comprehensive discussion of this issue, see Goldberg & Del Guidice (2005). She concludes that her functionally motivated analysis is superior to purely formal generative accounts of SAI, since the latter is only descriptively adequate but does not possess any explanatory power due to the lack of a motivation for the distribution of and restrictions on SAI (Goldberg, 2006, pp. 181-182). Goldberg’s functional account of SAI constructions has been criticized on many grounds, most prominently in a series of articles published in an issue of Cognitive Linguistics. For example, Borsley & Newmeyer state that Goldberg’s analysis is flawed since there are also other constructions such as ifclauses which share a subset of the features attributed to the SAI prototype but do not exhibit SAI (Borsley & Newmeyer, 2009, pp. 136-137). On the other hand, there are prototypical (i.e., positive) sentences which do exhibit SAI, for example sentences with positive adverbs like Particularly did he urge his co-workers to sign-up (Borsley & Newmeyer, 2009, p. 138). In her reaction to this, Goldberg points out that these sentences share other properties of the SAI-family, but not the ‘non-positive’ criterion, which implicitly points to the fact that SAI constructions may not be a radial category (with all members sharing one or more features of the prototype; Goldberg, 2009a, p. 209). This ties in with Croft’s observation that the relationship between SAIconstructions might best be captured in terms of family resemblance since “each [type of SAI-construction] shares some attributes with other members, but no member has all of the prototypical properties” and that “SAI is only one structural property of otherwise diverse constructions” (Croft, 2009, <?page no="58"?> 44 p. 161). This also entails that the positing of markedness-links between the SAI-prototype and the individual types of SAI-constructions would be unnecessary, since no central member of the category would have to be assumed. Note that this does not question the basic assumption that SAI is not a purely syntactic generalization, since “the facts surrounding SAI do not involve a system of syntactic generalizations. Moreover, the surface form of SAI is learnable, and its complicated distribution requires recourse to a usage-based constructionist account” (Goldberg, 2009a, p. 208). 44 However, positing a network of constructions related by family resemblance instead of a radial category does have consequences for how the relationship is assumed to be represented mentally. Entertaining the notion of a radial category, Goldberg’s analysis implies that all SAI-constructions are linked via the same (uninstantiated) prototype and are thus closely related. Croft, on the other hand, does not expect a more schematic SAI construction to be psychologically real, nor does he posit that all SAIconstructions are perceived as related (to the same degree), since “there is little functional unity to the different constructions that exhibit SAI” (Croft, 2009, p. 162). He also notes that there is little formal unity, because “SAI is only one structural property of otherwise diverse constructions” (Croft, 2009, p. 161). In any case, the question whether yes-no questions, for example, “ever, for any speakers other than linguists, become related in some way to constructions such as: <Expletive aux pronoun> (e.g., Boy are you nice! , Wow is that scrummy! ) is an empirical issue” (Lieven, 2009, p. 196) which has not been explored so far, and Goldberg herself recognizes the need for psycholinguistic experiments in order to clarify this issue (Goldberg, 2009a, p. 211). Apart from the fact that not much is known about the psychological reality of the proposed links between constructions in the mental grammar of adults, we are also far away from having a comprehensive answer to the question of how linguistic representations emerge in child speech. This is partly due to the fact that although developmental psychologists working on language development have embraced CxG as a psychologically plausible model of grammar, studies on language acquisition have not played a prominent role in Cognitive Linguistics for a long time, a fact which has been aptly commented on by Tomasello: The irony ... derives from the fact that Cognitive Linguistics stresses language use, entrenchment, constructional schema formation, and other usagebased processes that obviously have their origins in childhood, but empirical studies of how these things actually take place have traditionally not been considered of direct relevance to the theory. (Tomasello, 2007, p. 1108) 44 “The existence of purely syntactic generalizations would only be a powerful blow to the constructionist (non-UG) approach if it turned out that such syntactic generalizations formed an unlearned system of domain-specific representations or principles” (Goldberg, 2009a, p. 208). <?page no="59"?> 45 While Croft’s criticism already implies that Goldberg’s search for a common functional motivation for SAI may have been misguided, Bybee criticizes Goldberg more explicitly for being influenced by generative thinking, “fall[ing] back on generativist means of argumentation and the structuralist assumption that the more general a rule, constraint or principle is, the more valuable it is” (Bybee, 2007, p. 694). To her, it is hardly surprising that SAI does not have a single function, so that “assigning it a function, or even a cluster of related functions, as Goldberg does, requires the postulation of features so abstract as to be virtually meaningless” (Bybee, 2007, p. 696). In general, she accuses Goldberg of shifting her focus from “generalizations that speakers make into the dubious territory of generalizations that linguists come up with” (Bybee, 2007, p. 694). In this book I follow Croft and Bybee in assuming that the different constructions exhibiting SAI are related by family resemblance instead of deriving from one common prototype. Along similar lines, I will describe the structures dealt with here, i.e., yesno questions, as a family of constructions related by formal and/ or functional characteristics. 45 Overall, one can still detect a strong influence of generativist thinking on usage-based, constructivist accounts of language development and use, as evidenced by the appeal to principles of economy (e.g., in models of defaultinheritance or generalizations proposed to underlie SAI; Goldberg 1995, 2006, 2009). Furthermore, the strong need to position oneself against generativist approaches has led to a focus on the very same structures that have dominated generative research, thus excluding both non-canonical structures and neglecting functional factors. This is also mirrored in developmental research on questions, which will be reviewed in the next chapter. 2.2.6 Summary Within Construction Grammar, linguistic knowledge is conceptualized as “a structured inventory of conventional linguistic units” (Langacker, 1987, p. 57) of varying degrees of complexity and abstractness, abandoning the generative-transformational ‘rules and words’ account. It is situated within the larger field of Cognitive Linguistics and adopts a non-modular view of language. One of the primary goals is to integrate what is known about language with what is known about human cognition and interaction more generally. Drawing on findings concerning information processing and representation in non-linguistic domains, CxG stresses the role of frequency of occurrence and associative links in the formation of linguistic representa- 45 The fact that questions are a family of constructions rather than extensions of one underlying prototype has also been noted by Dąbrowska (2004, p. 159). For an alternative treatment of SAI from a cognitive, functional perspective, see Chen (2013), who proposes the existence of two separate SAI constructions. <?page no="60"?> 46 tions, aiming at offering a psychologically plausible model of mental grammar. <?page no="61"?> 47 3 The Acquisition and Use of Yes-No questions This chapter provides a critical discussion of previous research on the L1acquisition of yes-no questions in English. Starting out from a general overview of studies on the prevalence of questions in child-directed speech (Section 3.1), the acquisition of questions by children (Section 3.2) and competing approaches concerning the explanation of error patterns (Section 3.2.1) are discussed. After that, I will define non-canonical yes-no questions (i.e., fragments, reduced questions and rising declaratives) and discuss their treatment in previous research (Section 3.3), concluding that the acquisition and use of these structures is still underresearched. Subsequently, a summary of previous research on the discourse function(s) associated with yesno questions is presented (Section 3.4), pointing out issues which have not been addressed in detail so far. The chapter concludes with a terminological note (Section 3.5), a brief summary (Section 3.6) and the research questions addressed in the empirical part of this book (Section 3.7). Please note that research on wh-questions is only included in some detail in cases where their formation is considered in comparison to yes-no questions. Furthermore, studies on the comprehension of questions and children’s answering behavior are not discussed. While the focus is on work analyzing the acquisition and use of questions in English, studies on other languages are included in cases where evidence on English is sparse, which pertains particularly to the functional dimension of questions in child(directed) speech. Since assumptions about what it is that has to be acquired rely heavily on the description of the adult target system, summaries of the respective generative and constructionist formalizations (which have already been discussed in more detail in the previous chapter) are given, and research on yes-no questions in adults is considered in cases pertaining to one of the two foci named above (i.e., non-canonical yes-no questions and discourse-functional characteristics). 3.1 The Prevalence of Questions in Child-Directed Speech Yes-no and wh-questions have received and continue to receive much attention in developmental research for a number of reasons. First of all, the production (and especially comprehension and answering) of questions has been suggested to provide valuable insights into children’s mental development (Ervin-Tripp, 1970). Most importantly, within generativetransformational approaches, questions are seen as a prime example for the application of formal rules, most prominently SAI and (in non-subject whquestions) wh-fronting. For usage-based, constructionist approaches, they <?page no="62"?> 48 are considered a ‘test case’ since the learning of these supposedly purely formal generalizations needs to be accounted for in terms of the gradual abstraction of item-based constructions which are picked up from the ambient language. In terms of sheer quantity, the input is very rich, and there are numerous studies mentioning the high number of questions in childdirected speech (e.g., Sachs, Brown, & Salerno, 1976). However, only rather comprehensive studies which also provide quantitative information will be included here. 46 While all studies find that questions are rather frequent, there is a considerable range in the percentages provided (from approx. 21- 44%). 47 This is potentially due to actual differences in the data, with different coding procedures also influencing the frequency counts. 48 For example, Broen (1972) analyzed child-directed speech by ten mothers, obtained in two five-minute-recording-sessions (free play and storytelling) and found that on average, 36.9% of all utterances addressed to the children (mean age = 1; 9) were questions. It has to be noted that the five categories that were used (i.e., single word, imperative, question, declarative, and grammatically incomplete sentence) were not mutually exclusive (Broen, 1972, p. 29) and that the exact nature and amount of overlap was not reported on. Furthermore, the definition of what counts as a ‘question’ is very inclusive: ‘reduced‘ structures such as Want me to shut it? were coded as both ‘grammatically incomplete sentence’ and ‘yes-no questions’, and intonation-only questions (i.e., rising declaratives) as well as tag-questions were also subsumed under the heading ‘question’. Newport (1977) investigated the relationship between mother's speech and the course and rate of children’s language acquisition. Two-hour sessions of fifteen mother-daughter dyads were recorded with an experimenter present, with a six-month-gap between sessions (there were three age groups, with children at the first recording session being 1; 0-1; 3, 1; 6-1; 9, and 2; 0-2; 3 old). About 100 maternal utterances each were considered in the analyses of child-directed speech. Questions comprised a total of 44% of all utterances deemed complete and grammatical, with 21% yes-no questions (Newport, 1977), the use of which also significantly correlated with children’s language at the time the second sample was obtained, measured in terms of auxiliaries (per verb phrase). Apparently, reduced questions were 46 Davis (1932) provides a summary of very early studies on questions and also reports data on the percentages of questions in child speech (ranging from 11-22%). 47 Research on other languages indicates that this prevalence of questions in childdirected speech is not exclusive to English (e.g., Savic, 1975, for Serbo-Croatian, and Buium, 1976, for Hebrew). 48 The subsequent section focuses on the percentages of questions found in child-directed speech. See Pine (1994) and Richards (1994) for more comprehensive reviews on the makeup of child-directed speech and its relationship to children’s language development. <?page no="63"?> 49 subsumed under the general heading ‘yes-no questions’, since the authors give examples from this category (Newport et al., 1977, p. 127). Cross (1977) based his analysis on samples of 300 words each of sixteen mother-child-dyads (age range of the children: 19-32 months) and found 18.2% of all utterances to be yes-no questions, subsuming intonation-only as well as tag-questions in this category. He found no correlation of the maternal use of questions with any measures of the children’s language development. In a further study on the relationship between child-directed speech and children’s syntactic development, Furrow et al. (1979) analyzed speech samples of 100 utterances each collected from 7 mothers interacting with their children at 1; 6 and 2; 3, respectively. All utterances were assigned to (mutually exclusive) syntactic categories; questions were subdivided into ‘wh-question’, ‘yes-no question’ (which also included tags) and ‘yes-no question other’, the latter category including reduced structures as well as rising declaratives. Interestingly, the percentage of ‘other yes-no questions’ is similar to (and at 1; 6 even higher than) that of ‘yes-no question’ (with 7.6% vs. 9.3% and 8.1% vs. 7.6%, respectively). Furthermore, they found a significant correlation between mother’s use of ‘other yes-no questions’ at 1; 06 and the children’s language at 2; 3, measured in terms of MLU and the children’s use of verbs, nouns (per utterance) and auxiliaries (per verb phrase). The authors state that “especially those [yes-no questions] which were not simple inversions ... were associated with greater language achievement by the child” (Furrow et al., 1979, p. 433), even though it has to be kept in mind that such a correlation does not necessarily entail a cause-effect relationship. This analysis constitutes a rare case in which non-canonical input is seen as beneficial to the child’s language development, an aspect which is later on taken up by Estigarribia (2007, Chapter 6, 2010); for a comprehensive study of factors influencing auxiliary verb learning, see Richards (1990). A more recent study by Cameron-Faulkner et al. (2003) is based on a larger amount of data: two-hour-recordings from twelve mothers (age range of the children: 1; 9.28-2; 6.23), taken from the Manchester corpus (Theakston, Lieven, Pine, & Rowland, 2001) were analyzed in terms of six different utterance-level-constructions. Questions accounted for 31% of all maternal utterances (16% wh-questions and 15% yes-no questions). However, the authors discuss the possibility that this high proportion of questions may have resulted from the discourse situation, since an experimenter was present, which could have prompted the mothers to ask more questions than they usually would (Cameron-Faulkner et al., 2003, pp. 850-851). In order to confirm the representativeness of their sample, they compared it to a subsample in the Wells corpus (Wells, 1981). Apart from a slightly lower percentage of wh-questions and a slightly higher percentage of fragments and imperatives, the overall proportions were strikingly similar (Cameron- Faulkner et al., 2003, p. 851). <?page no="64"?> 50 An overview of the utterance types found in child-directed speech in all studies is provided in Table 3.1. For the categories ‘fragment’, ‘imperative’, and ‘declarative’, only the overall proportions are provided, regardless of whether the original study offered more detailed information (i.e., different subcategories such as ‘one-word’ and ‘multiword’ fragment). For the category ‘question’, all subcategories provided in the respective studies were kept. Table 3.1: Mean proportions of different construction types found in child-directed speech Utterance type Broen (1972) Newport (1977); Newport et al. (1977) Cross (1977) Furrow et al. (1979) Cameron- Faulkner et al. (2003) Wells (1981); (analysis by Cameron- Faulkner et al., 2003) (mean) age 1; 9 1; 7 2; 1 1; 6 2; 3 range: 1; 10- 2; 7 range: 1; 6-5; 0 49 fragments 60% excluded (approx. 40%) 39% (incl. unintelligible utterances) 20.6% 24.8% 20% 27% imperative 24% 18% 7.4% 21.1% 19% 9% 14% declarative 30% 38% 33.6% 22.3% 25% 39% 38% question 37% 44% 33.4% 35.4% 30.3% 32% 21% y-n question 13% 21% 7.7% 7.6% 8.1% 15% 13% whquestion 9% 15% 15.4% 18.5% 14.6% 16% 8% tagquestion 2% not coded 5.6% coded as y-n coded as declaratives coded as declaratives intonation questions; reduced 13% not coded 5.6% (only intonation questions) 9.3% 7.6% coded as declaratives coded as declaratives further categories 8% (deictic questions) 49 Since Cameron-Faulkner et al. (2003) only state that “the CDS [=child-directed speech] from four mothers in the Wells (1981) corpus were coded” (p. 851), no precise information on the age of the children can be provided here. <?page no="65"?> 51 Two things are particularly noteworthy here: First of all, prevalence of noncanonical yes-no questions and secondly, the overall high proportion of questions found in child-directed speech. 50 It is striking that while there are only minor differences concerning the coding of the other categories - mainly with regard to the number of subcategories coded for - the definition of what counts as a question is a slightly different one in each of these studies, with the only common denominator being ‘full’ yes-no and wh-questions. Broen (1972), Furrow et al. (1979) and Cross (1977) included reduced and declarative yes-no questions in their counts and even though the different coding procedures do not allow for a direct comparison, all studies found non-canonical structures to be almost as frequent as or even more frequent than full yes-no questions (see the respective columns of the shaded rows in Table 3.1 for comparison). Unfortunately, these studies only investigate sentence types on a very abstract level and are based on rather small samples, while the only large-scale study applies criteria which actually mask the percentage with which non-canonical questions occur: Cameron-Faulkner et al. (2003) based their coding system on the sequential ordering of elements, with a focus on utterance-initial material and the exclusion of prosodic factors. Consequently, reduced questions were coded as fragments while intonation-only questions as well as tag-questions were coded as declarative. The heterogeneity of the category ‘question’ found in all studies already implies that there are different underlying definitions of what counts as a ‘question’. In most cases, different types of structures are subsumed under the heading, i.e., questions do not form a homogeneous category with a common formal or functional feature but rather a family of formally and/ or functionally related constructions. Furthermore, particular members of this family are usually focused on: While canonical yes-no questions have been thoroughly investigated, non-canonical questions have largely been neglected, even though they have been found to be quite frequent in the input. Apart from this relative prevalence of non-canonical structures, the data suggest that - while there are differences due to individual variability, discourse contexts, and the different coding procedures - there is a rather high proportion of questions within child-directed speech. What is more, even if one only counts full yes-no and wh-questions as belonging to the category ‘question’, the proportion of questions is found to be higher than that of full subject-predicate structures in all analyses, with a “surprisingly small proportion of fully transitive utterances supposedly the canonical English sentence type” (Cameron-Faulkner et al., 2003, p. 851). Therefore, the appropriateness of the assumption that (canonical) questions are derived from the supposedly ‘prototypical’, i.e., declarative SVO-sentence or should 50 There are further studies on the makeup of child-directed speech which corroborate this: For example, Remick (1976) found that between 26 and 57% of all complete clauses in the child-directed speech of 20 minute-tapes of 8 mothers were questions. <?page no="66"?> 52 at least be described as ‘deviating’ from this more basic pattern, is highly questionable, at least from the perspective of a child learning English (in a Western setting). 51 However, this view has dominated generative-transformational formalizations, which focused on subject-auxiliary inversion, turning more basic structures into well-formed (i.e., ‘full’) yes-no questions. Reduced structures and rising declaratives were not regarded as legitimate targets for acquisition, despite the fact that they are found to be present (and considered to be acceptable) in the speech addressed to children. Consequently, (early) research on the acquisition of yes-no questions focused on children’s inversion rates and used supposedly well-formed structures (‘full sentences’ exhibiting SAI) as the reference point and the adult target structure. 3.2 Developmental Paths Early generativist-transformational accounts of the acquisition of questions focused mainly on children’s ability to apply SAI (in yes-no and non-subject wh-questions), and wh-fronting (in non-subject wh-questions), leading to a focus on structures found in child speech and different developmental stages while largely neglecting the influence of child-directed speech and the semantic and pragmatic dimension of questions. The first comprehensive studies on the acquisition of questions were provided by Bellugi (1965), and Bellugi and Klima (1966). In the theoretical part of Bellugi (1965), she summarizes question formation within the generative-transformational framework. Sentences are seen as “strings of symbols ... upon which certain operations can be performed” (Bellugi, 1965, p. 104). Interestingly, Bellugi considered a change in intonation as one such operation, with a rising pitch converting any declarative sentence into a yes-no question (Bellugi, 1965, p. 107): (1) The boy played with his dog. (2) The boy played with his dog? Further operations for yes-no and non-subject wh-questions are the movement of the auxiliary in front of the subject-NP (i.e, SAI) and in cases where no auxiliary is present the insertion of do, which takes the tense marking with it. Even without a change in intonation, these operations result in a ‘question-like utterance’ (Bellugi, 1965, p. 112) - however, it is generally assumed that the resulting yes-no questions will exhibit both SAI and rising 51 On a more general level, it has been noted that “virtually all formal approaches would take declarative positive assertions as [in] some sense ‘basic’ or ‘canonical’ ... but that decision is based more on descriptive simplicity than on any other criterion” (Borsley and Newmeyer, 2009, p. 136). <?page no="67"?> 53 intonation. For wh-questions, the questioned element is replaced by a whword, which is fronted, followed by SAI (in non-embedded non-subject whquestions). Tag-questions are (usually) attached to the end of a sentence and basically follow the same operations as yes-no questions, but the subject noun phrase of the main clause is pronominalized and the polarity of the main clause is reversed. Finally, negated questions are achieved by attaching not or -n’t to the auxiliary. It is assumed that “these few operations will enable one to transform an infinite set of sentences into an infinite set of questions” (Bellugi, 1965, p. 110) and it is suggested that once children have mastered an operation, they will be able to apply it across-the-board. In the empirical part of her paper, Bellugi is concerned with the different stages through which children pass until they achieve an adult-like ability to form questions. Based on data from the three Harvard-children Adam, Eve, and Sarah (available on CHILDES; Brown, 1973), the developmental paths are described in three stages, which correspond to a mean length of utterance (MLU) around 2.0, and above 2.5, and 3.5, respectively (Brown, 1973). In stage I (MLU 1.5-2.5), the rising intonation contour is the only operation involved in the formation of questions that seems to be present in the children’s speech - there appears to be little structure to sentences, and whquestions are restricted to a few rote-learned routines like What’s that? (Bellugi, 1965, p. 115). In stage II (MLU 2.5-3.0), two of the three children still use the intonation-only strategy for yes-no questions, while Adam makes use of D’you want X? as “pivot plus open class element” (Bellugi, 1965, p. 118). In a similar manner, the children use wh-words in utterance-initial position as a question marker, but do not seem to make use of wh-fronting and inversion. The use of tag-questions is restricted to ‘fixed tags’ such as You did it, huh? . In stage III (MLU 3.0-4.0), there is an “impressive and sweeping set of developments in the child’s grammar” (Bellugi, 1965, p. 120). Auxiliary verbs and inflections are now used in all sentence types and inversion is largely present in yes-no but not yet in wh-questions questions. Along similar lines, do-support is present in yes-no but not wh-questions while ‘full’ tag-questions are still absent. This patterning of inversion and dosupport in yes-no and wh-questions in stage III suggests that while “the evidence is good for considering that the child performs certain operations on strings that are definable as transformational operations ... there seems to be a limit on the number of operations on one string” (Bellugi, 1965, p. 126). In a similar vein, Brown suggested that children go through a stage in which they can only perform wh-fronting (‘preposing’) but not SAI (‘transposing’) in wh-questions (Brown, 1968, p. 284; also see Brown & Hayes, 1970). Along the lines of the Derivational Theory of Complexity (Fodor et al., 1974; Miller & McKean, 1964), it is suggested that each transformation is associated with a certain processing cost, assuming that SAI and wh-fronting “are ‘psychologically real’ operations for the child” (Slobin, 1979, p. 96). However, psy- <?page no="68"?> 54 cholinguistic evidence for the psychological reality of transformations could not be found in experimental studies (Fodor et al., 1974). Furthermore, the distinct developmental stages proposed on the basis of the Harvard children and the corresponding error patterns (i.e., higher non-inversion rates for whthan for yes-no questions) were not found in the majority of subsequent studies (e.g., Derwing & Smyth, 1988; Erreich, 1984; Ingram & Tyack, 1979; Klee, 1985). It should also be noted that even if children were indeed found to pass through a stage in which they invert in yes-no but not non-subject wh-questions - there is no reason why wh-fronting (‘preposing’) and not SAI (‘transposing’) should occur first in wh-questions, instead of the other way around (i.e., inversion but no wh-fronting); amongst others, this has been pointed out by Bellugi herself (1971) as well as Dąbrowska (2004), and Fay (1978, p. 149). Overall, it was assumed that once the necessary prerequisites were in place (e.g., knowledge of the auxiliary-system), children would be able to apply the rules for question-formation across-the-board (e.g., not restricted to particular subject-auxiliary combinations) gradual increases in performance were consequently attributed to performance-factors or the inflation of inversion-rates through rote-learned strings such as What’s that? . These early formulas were thought to be abandoned once more abstract rules were mastered, meaning that item-based effects were only expected at the very early developmental stages. 52 Since SAI in yes-no questions seemed to be mastered from quite early on (in contrast to wh-questions), most subsequent studies in the 1970s and 1980s focused on wh-questions. The main aim was to shed light on factors explaining the order of emergence of particular whquestions. Some studies only focused on the order of acquisition found in children’s productions (e.g., Labov & Labov, 1978) while others investigated children’s answering of wh-questions, mainly in order to assess the influence of semantic and pragmatic complexity associated with particular whwords and consequently the cognitive development of the children (Ervin- Tripp, 1970; Horgan, 1977; Cairns & Ryan Hsu, 1978; Parnell, Patterson, & Harding, 1984); also see Tyack & Ingram (1977) on both production and comprehension of wh-questions. The relatively robust order of acquisition in both comprehension and production was attributed to differences in the semantic and/ or syntactic complexity rather than input-characteristics, since no correlation between input frequency and order of acquisition was found (Savic, 1975, for Serbo-Croatian). More recent studies from a generative perspective also stress that while task demands and characteristics of the input might influence children’s performance to some degree, they generally 52 It should be noted that there are early studies suggesting that, for example, the progressive ending -ing is first acquired on a verb-by-verb basis Brown (1973) and that similar findings were obtained concerning the regular plural (Cazden, 1968). Still, the assumption was that item-based effects would disappear once the rule had been generalized (or, within P&P, once the setting of the parameter had been triggered). <?page no="69"?> 55 already possess the relevant syntactic knowledge and can quickly make across-the-board generalizations (Valian & Casey, 2003). In sum, the generativist position holds that children possess the competence to comprehend and produce questions from rather early on and that the formation of questions relies on knowledge of abstract rules (formalized as transformations in earlier accounts and a ‘copying followed by deletion’operation in later ones). Error patterns are explained as stemming from performance factors hindering children from applying their knowledge in language production and/ or the overor underapplication of certain rules, affecting performance on whole categories (such as auxiliary vs. copula be; see Section 3.2.1). Even though the fact that early item-specificity can be found in child language is not denied (and is indeed confirmed by many of the studies mentioned, e.g., Bellugi, 1965; Kuczaj & Brannick, 1979), it is stressed that the production of rote-learned formulas and item-based structures is only used as an initial strategy which is abandoned once children start using the rules productively. 53 The formation of questions is the one domain where the existence of such abstract rules was hardly ever cast in doubt, even by those who suggested that they might not play a role in the formation of other structures (such as passives; Maratsos, 1978, p. 263). Therefore, input frequencies are not thought to influence developmental paths to a considerable degree, a claim that is mirrored in such statements as “each child hear[s] a different set of sentences as “input”; and yet the language milestones ... are constant” (Bellugi, 1965, p. 113). In contrast to generativism, usage-based, constructivist approaches not only stress the high degree of formulaicity of early child language but assume that children only slowly abstract away from rote-learned formulas to gradually achieve adult-like productivity. Furthermore, item-specific knowledge is not thought to be abandoned in favor of more abstract knowledge in the course of development and all linguistic knowledge is assumed to be learned on the basis of the ambient language with the help of mostly domain-general cognitive skills. Consequently, input-output correlations and the interaction between children and their caretakers are of primary importance. While some early accounts of question-acquisition also suggest that adult-like productivity is achieved gradually and that questions may not be formed by applying rules of inversion (Clark, 1974, p. 7), only more recent usage-based constructivist accounts offer a comprehensive account of the representations and mechanisms which might be involved. Questions are potentially problematic for constructivist accounts emphasizing low-level, item-based patterns since their word-order is non-canonical and their formation presumably rests on the application of abstract rules (e.g., Dąbrowska & Lieven, 2005, p. 440). Therefore, they were seen as ‘test 53 But see, e.g, Todd (1982) for a generativist study suggesting that the application of rules is context-specific rather than dependent on the complexity of the structures involved. <?page no="70"?> 56 cases’ for these approaches whose findings cast doubt on the generativist assumption that the formation of questions relies on a rule-based manipulation of underlying declarative forms - some English-speaking children learn particular (partially schematic) question constructions as their first multiword combinations, making it unlikely that their declarative counterparts play a role in their formation (e.g., Dąbrowska, 2000; Tomasello, 1992). Furthermore, research suggests that the emergence of questions in child speech can indeed be accounted for in lexically specific terms. One of the most comprehensive studies on question acquisition within the usage-based framework is provided by Dąbrowska (2000), who presents an analysis of syntactic questions (i.e., “utterances which are at least two words long and which begin with either a WH word or an auxiliary and end with a question mark”, Dąbrowska, 2000, p. 87) in the speech of a single child, Naomi (age span 1; 6.16 - 3; 08.19; Sachs corpus; Sachs, 1983). It was found that overall 88% of all syntactic questions were formulaic, a formula being defined as “any sequence of simple units, with or without a slot” (Dąbrowska, 2000, p. 88) occurring at least five (‘minor formula’) or ten times (‘major formula’) in the data. The degree of formulaicity gradually decreased over time (from 94.8% at 1; 6-1; 11 to 63.5% at 3; 0-3; 8), and the analysis of the two formulas Can-I- PROCESS and What’s-THING-doing? suggests that these structures progress from invariant formulas to increasingly abstract formulaic frames by schema extraction, with slots being opened up one by one to allow for more productivity. The use of ‘inversion’ seems to emerge in a bottom-up, item-based manner as well, with inverted vs. uninverted word order being confined to particular auxiliary-subject combinations initially (Dąbrowska, 2000, p. 97; also see Dąbrowska, 2004, Chapter 9, and Lieven, 2008). In order to shed further light on the children’s linguistic knowledge and the processes involved in the construction of novel utterances, Dąbrowska & Lieven (2005) analyzed spontaneous production data obtained at relatively dense intervals from two children (Annie and Brian, each recorded five times a week for a period of six weeks at the ages of 2; 0 and 3; 0), focusing on the production of question constructions. 54 For each child at each age, the respective corpus was split into two parts: The main corpus and the test corpus, the latter of which consisted of the last two (Annie) and the last 5 recordings (Brian), respectively. The syntactic questions in the test corpus (i.e., the ‘targets’) were then analyzed in terms of how far the structures could be derived from utterances already found in the main corpus. This was done on the basis of simple operations performed on component units identified in the main corpus. A component unit was defined as “an expression which shares lexical material with the target and is attested at least 54 Brian is the pseudonym that was used for Thomas before permission to use his real name was obtained (Lieven and Behrens, 2012, p. 233). His whole corpus is used as the basis for my own empirical investigations in Chapters 5-8. <?page no="71"?> 57 twice in the main corpus” (Dąbrowska & Lieven, 2005, p. 447). Such a unit could either be a fixed phrase or a frame with a slot. The production of novel expressions was captured in just two operations, juxtaposition and superimposition. In juxtaposition, two component units are combined linearly in either order (example taken from Dąbrowska & Lieven, 2005, p. 442): (3) now + are you downstairs? now are you downstairs? or are you downstairs now? In superimposition, an underspecified part of one unit is elaborated by another unit (a ‘filler’) which has the properties the slot requires (example taken from Dąbrowska & Lieven, 2005, p. 443): (4) shall I PROCESS? + open that shall I open that? Overall, applying these two operations accounted for about 90 percent of all question types in the test corpus, thus indicating that children’s early question constructions are very item-based and lending further support to the claim that “rather than assembling their questions from atomic elements according to abstract syntactic rules and then applying further syntactic transformations such as WH fronting and inversion, these children combined partially specified units using the two operations described above” (Dąbrowska & Lieven, 2005, p. 456). 55 There was also an increase in the number of operations needed for the derivation of utterances from 2; 0 to 3; 0 suggesting that children gradually develop more flexibility. 56 Furthermore, constructivist research suggests that relative input frequency, and not syntactic and/ or semantic complexity of wh-words is the best predictor for the order of acquisition of particular wh-auxiliarycombinations (e.g., Rowland & Pine, 2000; Rowland et al., 2003; also see Clancy, 1989, for initial wh-words in Korean and Theakston et al., 2005, for initial subject-auxiliary combinations in English declaratives). These findings stand in contrast to earlier generative research, in which no such correlation was found (Savic, 1975). However, “this conclusion seems to have been based on the fact that there is no very close match between child and adult data, rather than on statistical analysis” (Rowland et al., 2003, p. 612) and a reanalysis shows that there is indeed a strong correlation which is statistically highly significant (Rowland et al., 2003, p. 612). Moreover, most generative research did not even consider input frequency, i.e., the possible influ- 55 Question types instead of tokens were counted and self-repetitions as well as imitations were excluded (Dąbrowska and Lieven, 2005, p. 451). 56 This gradual abstraction process has been observed in other languages as well - see, e.g., Steinkrauss (2009) on wh-questions in German and Clancy (1989) on wh-questions in Korean. <?page no="72"?> 58 ence of this factor on order of acquisition was not analyzed (Bloom, Merkin, & Wootten, 1982). A similar relationship between the relative input frequencies of utterance-initial word-combinations and the acquisition of canonical as well as non-canonical yes-no questions is likely, but to my knowledge no study to date has investigated this issue. To sum up, in contrast to generative accounts, which stress the reliance on abstract transformational rules and the quick development of across-theboard generalizations in children’s speech, usage-based accounts assume the gradual learning of constructions in a bottom-up manner, with relative input frequencies exerting a great influence on the child’s language development. These differences are also mirrored in the different approaches to the explanation of children’s error patterns, which will be briefly summarized below. 3.2.1 Error-patterns in child questions Strikingly, almost none of the studies focusing on the explanation of error patterns in children’s questions offer a definition of what constitutes an error. Implicitly, ‘error’ is used to refer to utterances in child speech which are (supposedly) not attested in the input and which are deemed ‘ungrammatical’ by English native speakers. In the case of questions, so-called auxiliary-doubling errors (examples (5) and (6)) and non-inversion errors in whquestions (7) have been focused on, the latter of which were often compared with non-inversion in yes-no questions. 57 (5) *What does she doesn’t like? (6) *Does she doesn’t like cake? (7) *What she doesn’t like? The findings are often contradictory and studies are often not directly comparable - mainly due to different methodologies employed (e.g., naturalistic vs. experimental studies), and different error coding as well as different criteria for the inclusion or exclusion of data. For example, wh-questions without medial auxiliary and declarative word order such as What she likes? are counted as non-inversion errors in some studies (e.g., Thornton, 2008), while they are excluded from these counts in others (e.g., Valian, Lasser, & Mandelbaum, 1992). Moreover, the density of data obtained in naturalistic sampling as well as the analysis method used may influence the calculation of error rates (Rowland, Fletcher, & Freudenthal, 2008, p. 3). 57 Only generative-transformational and usage-based, constructivist explanations for error patterns are considered here (e.g., for explanations based on Role and Reference Grammar, see Valin, 1998, 2002). Other error types, such as auxiliary omission (? What he doing? ) and double tensing (*What does he likes? ) as well as non-raising errors in whquestions (? What he likes? ) are not considered in detail here, while ‘omission errors’ (? <Has> Anyone seen Waldo? ), ‘raising-errors’ (? She likes cake? ) and ‘non-inversion’ errors in yes-no questions (? She can dance? ) will be dealt with in Section 3.3. <?page no="73"?> 59 Still, a few trends have emerged for auxiliary-doubling and non-inversion errors: Auxiliary DO displays a rather high error rate (Rowland et al., 2005; Santelmann et al., 2002; Stromswold, 1990), while both auxiliary and copula BE are associated with rather low error rates (Maratsos & Kuczaj, 1978; Rowland et al., 2005). In cases where auxiliary and copula BE are distinguished, the latter tends to displays a higher error rate (Klee, 1985). Negated questions are associated with high error rates, especially attracting auxiliarydoubling errors (Bellugi, 1971; Erreich, 1984). For a more comprehensive overview and more references, see Ambridge et al. (2006, with a focus on wh-questions) and Stromswold (1990, p. 288, on both whand yes-no questions). Stemming from the generative assumption that, for example, the subjectauxiliary inversion rule is a general principle of Universal Grammar which is available to children from very early on, the fact that these errors are wellattested in child speech poses a problem for these accounts. In order to account for the observed patterns, early studies suggested that children may initially only apply the copying but not the deletion operation associated with SAI, leading to aux-doubling errors. According to the so-called Basic Operations Hypothesis, children possess a non-adult-like rule, causing across-the-board errors (Mayer, Erreich, & Valian, 1978). Under the competing Transformational Hypothesis, the errors were interpreted as a performance problem rather than a result of a non-adult grammar, with difficult sentences masking children’s grammatical competence and leading to higher error rates (Fay, 1978; Hurford, 1975; Nakayama, 1987). The fact that negative questions are particularly difficult for children was accounted for by positing an additional transformation called ‘Auxiliary Preposing’, which moves “the tense marker (and auxiliary, if present) to the left of the negative marker not” (Fay, 1978, p. 147). In this view, children again perform the copying but not the deletion operation associated with the respective transformation, in this case resulting in double-tense errors like *I didn’t got X. 58 In a more recent account, Guasti et al. (1995) propose that because of a mis-set parameter, children may be reluctant to move the negation marker or both the negation marker and the auxiliary out of IP (i.e., I to C-movement), leading to aux-doubling or non-inversion errors, respectively. Once the parameter is reset, errors are expected to disappear across-the-board. Overall, most generative accounts of question errors assume that children possess knowledge about SAI but fail to apply it consistently, due to processing 58 A problem of all copying-without deletion accounts is that they fail to account for the fact that there are no such errors in questions involving WH-movement, i.e. utterances like *What this is what? are unattested. This is similar to the failure of Brown’s (1968, p. 284) and Bellugi’s (1965, p. 126) proposal concerning limits on the number of transformations which can be performed on one utterance to account for the fact that SAI and not WH-movement is affected. <?page no="74"?> 60 limitations (Bellugi, 1965) or the failure to view the rule as obligatory (Valian et al., 1992). However, all these proposals expect errors to occur (and then disappear) across-the-board. More recent generative research contains auxiliary-specific approaches in order to accommodate the fact that errors are not evenly distributed, claiming that idiosyncratic properties of the dummy-auxiliary DO and copula BE are responsible for the observed patterns. While DO may cause problems because it is usually not overtly present in the underlying declarative sentence, copula BE is the only main verb which inverts in English, thus potentially conflicting with the general rule that main verbs do not invert in English (Santelmann et al., 2002). Studies trying to account for non-inversion in wh-questions suggested that the uninverted word-order encountered in some adjunct wh-questions (8) is initially extended to other wh-questions by children, who only gradually retreat from this overgeneralization (Villiers, 1991). (8) How come he is leaving? While making slightly more specific predictions than earlier accounts, generative research still revolves around abstract categories (such as SUBJECT and AUX) or lexemes rather than particular word forms. For example, Stromswold (1990) suggests that when children “hear a particular auxiliary, they generalize across tense, number, and person within the Be, Do, and Have subtypes” (Stromswold, 1990, p. 20), implying that categorygeneral rules are involved. Therefore, they are not able to incorporate the finding that errors pattern by different lexical forms of the same auxiliary subtype (e.g., is vs. are; Rowland et al., 2005). Furthermore, generative research does not take into account the role potentially played by the frequency of particular structures found in the ambient language. This stands in contrast to usage-based, constructivist approaches, which assume that children take over patterns found in the input. Since children start out with item-based, low-level schemas from which they abstract away only gradually, error patterns are expected to be influenced by the concrete lexical material involved. For example, non-inversion errors in wh-questions are expected to pattern by wh-word + auxiliary subtype, being influenced by the absolute and relative frequencies of these word-combinations in the input. It is important to note that even though terms like ‘non-inversion error’ are often used (e.g., Ambridge et al., 2006, p. 535), canonical questions are viewed as independent constructions which do not result from transforming an underlying declarative sentence; rather, the word-order is specified directly within the respective construction. It has been suggested that both auxiliary-doubling errors and noninversion errors may result from similar processes, e.g., from children’s combining schemata in a non-adult like way. For example, in cases where <?page no="75"?> 61 the respective wh-aux-combination has not been learned yet, children’s use of so-called groping patterns (Braine, 1976) may lead them to combine a whword or a wh-aux-subject combination that is already well-entrenched with a part of a declarative from the input, resulting in uninverted utterances like *What she likes? (Rowland & Pine, 2000, p. 164) or auxiliary-doubling errors like *Why don’t you don’t like cakes? (Rowland & Pine, 2000, p. 179) - these are essentially results of the juxtaposition and superimposition operations proposed by Dąbrowska & Lieven (2005, pp. 442-443). A more comprehensive version of this schema-combination account has been tested experimentally by Ambridge & Rowland (2009), who elicited both positive and negative what, why, and yes-no questions from 50 children (age 3; 3 - 4; 3) and found that children who possessed a particular positive frame (e.g., What does THING PROCESS? ) made significantly more auxiliary-doubling errors with the corresponding negative question (e.g., *What does she doesn’t like? ) than children who did not produce the positive frame, indicating that schema combination (i.e., the “superimposition of a positive question frame ... and a negative declarative frame” Ambridge & Rowland, 2009, p. 254) may indeed be involved in the production of these errors. In a similar vein, the prediction that non-inversion errors pattern by specific wh-word + auxiliary combinations is borne out by the data: See, e.g., Rowland & Pine Rowland & Pine (2000) for an analysis based on the Adam corpus (Brown, 1973; age 2; 3.4 - 4; 10.23); also see Ambridge et al. (2006), for an elicited production study with 28 children (mean age = 4; 1) which additionally found a significant influence of lexical auxiliary subtype alone. On the whole, Rowland & Pine (2000) found that the child produced inverted questions with the wh-aux combinations most frequently encountered in the input. Ambridge et al. (2006) found a relatively high error rate for what+do, despite the fact that this combination is very frequent in the input. This can be accounted for by looking not at simple input frequencies, but at the type and token-frequencies of the respective constructions - what+do occurs almost exclusively in the form What do you X? , i.e., with a high token, but a low type-frequency. This might prevent the child from developing a more abstract, and thus more productive What do PERSON X? formula (Ambridge et al., 2006, pp. 544-545), causing errors with subjects other than you. This is one way in which naturalistic and experimental methods can be combined in order to shed further light on the observed patterns. Along similar lines, Ambridge & Rowland (2009) correlated the input frequency of 12 lexical frames (based on the maternal data of the Manchester Corpus) with the mean percentages of correct questions and auxiliaryplacement errors obtained from the children in the experimental setting. They found that the input frequency of 11 out of 12 lexical frames of the type What does THING PROCESS? correlated with correct production of the respective frames in the children’s speech (the only outlier being Why can THING PROCESS? ). Similarly, in her study on yes-no and wh-questions in <?page no="76"?> 62 longitudinal data from 10 children, Rowland (2007) found that two-word frames with a high input frequency were more likely to be picked up by the children than frames with a low input frequency; furthermore, questions which could be produced using these entrenched frames showed significantly lower error rates than questions requiring the use of generalization strategies. These findings indicate that input frequency is an important determinant of error rates. It is, however, certainly not the only factor: For example, some negative questions attract higher error rates than would be expected based on input frequency alone (Ambridge & Rowland, 2009, pp. 247-248). The challenge for usage-based, constructivist approaches, then, is to “move beyond input frequency, and seriously consider other factors” while generativist accounts “will have to incorporate some role for the learning of particular lexical strings” (Ambridge & Rowland, 2009, p. 257). 59 The next section considers structures which have also commonly been classified as errors, i.e., ‘non-inversion’ errors (? She can dance? ), ‘raisingerrors’ (? She likes cake? ), and errors of omission (? <Has> Anyone seen Waldo? ) in yes-no questions. It will be suggested that these structures are potentially well-formed and should therefore be considered legitimate targets for acquisition. 3.3 Non-Canonical Yes-No Questions Generative approaches to question acquisition have focused very strongly on the mastering of SAI; even though usage-based approaches stress that questions are not formed by applying SAI, they have concentrated on the very same structures, i.e., yes-no questions with auxiliary-subject order, as legitimate targets for acquisition. This has led to a neglect of yes-no questions deviating from this pattern, with the exception of studies treating them as ‘errors’. This is particularly ironic from a usage-based perspective stressing that “language structure emerges through language use” (Tomasello, 2003, p. 5), since these non-canonical structures have been well-documented in research on child-directed speech (see Section 3.1) and in descriptions of adult language use (albeit often without providing information on their frequency of occurrence). These structures, which I am calling non-canonical (following Estigarribia, 2007, 2010), fall into three broad structural categories. 59 The strong focus on input frequencies has also led to usage-based, constructivist approaches being (mis)understood as relying solely on this factor. For example, Westergaard (2009) claims that “input is given primary importance and frequency is often used as the major explanation for acquisition orders as well as children’s occasional non-target-consistent forms” and she actually contrasts usage-based approaches with other frameworks in which input frequency is “seen as an important factor, but one that has an effect only in connection with other factors” (p. 65). <?page no="77"?> 63 First of all, there are yes-no questions in which the subject precedes the auxiliary: (9) You can’t see it? They have been referred to as ‘non-inverted yes-no questions’ and ‘rising declaratives’. Also often subsumed under this heading are utterances that contain an overtly tensed verb but no medial auxiliary (10). In generative accounts, they are sometimes treated separately under the heading ‘raising errors’, which presumably result from a failure to raise the tense (and or agreement) marking from the main verb inside VP. 60 Following Bolinger (1957) and Estigarribia (2007, p. 19, 2010), I call these structures ‘(strong) assertive questions’ (with and without medial auxiliary, respectively). (10) You saw it? Secondly, there are questions containing a (main clause) predicate, with either auxiliary, subject or both components missing: (11) <Do> You want an apple? (12) Don’t <you> want an apple? (13) Gone already? Since these structures have been regarded as the result of phonological, morphological, or syntactic deletion procedures, they are commonly referred to as ‘reduced questions’, which is the term which I will also adopt here. I do this without wishing to imply that these structures are a result of an online deletion process during language production. From a Construction Grammar perspective, there are no phonetically null elements which are present at some underlying level of representation (the ‘what you see is what you get’ approach to syntactic form, Goldberg, 2003). 61 It is, however, possible to capture the relationship between the ‘reduced’ and the corresponding fuller forms in terms of inheritance links; to my knowledge, there are only two studies to date which address the issue of reduced questions from a construction grammar perspective. The first is provided by Heine (2011), who assumes default inheritance. As a consequence, she regards reduced forms 60 For want of a better term, the label ‘raising error’ is also used in usage-based studies on question acquisition (Ambridge and Rowland, 2009, p. 259). 61 Reviewing different accounts on reduced questions as the result of a phonological process, Estigarribia (2007) comes to the conclusion that “English presumably evolved from having an online deletion process contingent on prior contraction, to directly generating reduced questions without relying on phonological deletion” (p. 43). He also observes that reduced questions could consequently be viewed as constructions in the sense of Goldberg (2006), meaning that they could be acquired independently of and prior to more complex, canonical questions (Estigarribia, 2007, pp. 50-52). <?page no="78"?> 64 (presumably) exhibiting the same or at least very similar functional and distributional features as their fuller counterparts merely as subtypes of the full(er) mother construction. In contrast to this, Caines (2012) provides corpus-based and experimental evidence on zero-auxiliary questions of the type You talking to me? , suggesting that ‘reduced’ constructions occurring with a high frequency are cognitively entrenched. In line with these findings, I would like to argue that while it is likely that reduced forms and their full(er) counterparts are (perceived by speakers as) related, reduced forms will be learned as (partially) independent constructions as long as they occur with sufficient frequency. It has been noted that structures like (13) “would be regarded by purists as incorrect English despite their frequency in real conversation of educated speakers” (Biber, Johansson, Leech, Conrad, & Finegan, 1999, p. 158). Strikingly, they later on repeat their claim that “yes/ no-questions are frequently elliptic” (Biber et al., 1999, p. 206) but do not offer detailed frequency information. Their overview of question types used in conversation only differentiates declarative questions (comprising about 10%) and fragments (comprising about 15%), without offering a more fine-grained categorization. Stating that with examples like (13), “we are reaching a point ... where ellipsis merges with non-clausal material” (Biber et al., 1999, p. 158), it is not clear how these cases were categorized (i.e., as declarative questions or fragments). Furthermore, it is not made explicit how they coded utterances like (11). These constitute a special case in that the form can either be categorized as an ‘assertive question’ (without a medial auxiliary) or a ‘reduced question’, which is due to what Fodor and Crain have called “various morphological accidents of English” (Fodor & Crain, 1987, p. 40). Within generative research, this has caused problems, since utterances like (11) “can be construed as non-inverted questions even if the speaker’s own mental representation of his utterance involved inversion” (Fodor & Crain, 1987, p. 40), resulting in a heterogeneous treatment of these structures in studies on inversion rates (this will be discussed in more detail below). In order to acknowledge the ambiguous status of these utterances, I refer to them as ‘weak assertives’ (WSERTs), again following Bolinger (1957) and Estigarribia (2007, p. 19, 2010). Lastly, there are fragments, which can either be single words (14), verbless phrases (15) or consist solely of a subordinate clause (16). (14) Now? (15) The guy from the shop? (16) Because he hit on you? In contrast to ‘reduced yes-no questions’, which are (usually) also interpretable without the corresponding discourse context and whose ‘full’ form can be reconstructed (so-called ‘context-independent ellipsis’; Gunter, 1963), <?page no="79"?> 65 fragments are often not amenable to reconstruction without the discourse context (Bolinger, 1957, p. 67). 3.3.1 Non-canonical yes-no questions in developmental research Within the context of acquisition studies, fragmentary questions are mostly ignored since they (presumably) do not offer any insights into the development of syntax. The existence of reduced and assertive questions in child (-directed) speech poses a problem particularly for generative approaches, in which all (well-formed) yes-no questions presumably exhibit SAI. Within this framework, the interest in the comparison of inversion rates of yes-no questions (presumably involving one major transformation) vs. whquestions (presumably involving two major transformations) has led many studies to classify instances of non-canonical yes-no questions in child speech as errors, especially strong assertives containing a medial auxiliary (e.g, Bellugi, 1971; Ingram & Tyack, 1979; Klee, 1985; Stromswold, 1990). As has already been noted by Estigarribia, even researchers acknowledging that non-canonical yes-no questions are part of the adult target repertoire classify them as ungrammatical (Estigarribia, 2007, pp. 20-21). For example, Brown refers to weak assertives and reduced yes-no questions as “mild departures from grammaticality which are acceptable in colloquial speech [emphasis added]” (Brown, 1973, p. 240) while at the same time calling these structures in the children’s speech ‘not well-formed’ (Brown, 1973, p. 180). In her study comparing inversion rates of yes-no and wh-questions using elicitation and spontaneous production data in 18 children (aged 1; 5 - 3; 0), Erreich (1984) admits that strong assertives with medial auxiliaries are not ungrammatical; still, they are not treated as correct but are categorized as non-inversion errors: The simplifying assumption was made in the present research that a comparison of the development of yes-no and wh-questions could be assessed independently of the speech acts which the utterances performed. ... It is possible that, especially when children fail to invert in yes-no and wh-questions, they are performing a different speech act. (Erreich, 1984, p. 585) In their overview of studies analyzing error patterns in questions, Ambridge and Lieven state that “the picture [regarding research on error patterns in questions] is complicated by the fact that non-inverted yes-no questions are not always counted as errors as they are permissible in certain contexts (e.g., She can’t do it? )” (Ambridge & Lieven, 2011, p. 283). One way in which researchers tried to sidestep the problem was to conduct experimental studies which could be designed to make the need for inversion obligatory (e.g., Derwing & Smyth, 1988; Erreich, 1984; Kuczaj & Maratsos, 1975; Nakayama, 1987; Santelmann et al., 2002). However, experimental tasks often have a rather artificial setting, which may lead children to employ strategies <?page no="80"?> 66 different from the ones they use in spontaneous conversation; as Stromswold has pointed out, using embedded questions like Can you ask him what that is? or Ask mommy if she likes coffee as prompts (e.g., Erreich, 1984, p. 583; Nakayama, 1987, p. 114) “could likely lead to more failures to invert, since the eliciting sentence has a noninverted auxiliary” (Stromswold, 1990, p. 138). In general, it can be observed that non-canonical yes-no questions have mainly been regarded as a complication, both with respect to the assessment of inversion rates in child speech but also with respect to the acquisition of SAI from the perspective of children acquiring English. For example, Valian and Mandelbaum propose that the fact that yes-no questions exhibit optional inversion might be a cause for uninversion errors in wh-questions (Valian et al., 1992). Within usage-based research on error patterns in questions, non-inverted yes-no questions have also mostly been excluded due to difficulties in evaluating them unambiguously. In her naturalistic study of errors in yes-no questions, Rowland states that “inversion is usual (although optional) in English yes-no questions” (Rowland, 2007, p. 108) and furthermore that “non-inverted yes-no questions were also excluded because it is often difficult to determine from the transcript whether a true question was asked or whether the utterance had been given a question mark in error” (Rowland, 2007, p. 116). What is more, even in usage-based frameworks, we find statements claiming that “in some constructions [subject-auxiliary] inversion is obligatory (y/ n questions)” (Goldberg, 2009a, p. 210). Consequently, there are studies referring to strong assertive (with or without medial auxiliaries) used by children as inversion errors. For example, in her study on the Naomi corpus (Sachs, 1983), Dąbrowska (2000) states that “‘correct as question’ means that the utterance contains the correct form of the auxiliary in the correct (i.e., ‘‘inverted’’) position [and] an explicit subject” (Dąbrowska, 2000, p. 91). She then makes the following statement about error patterns: “The proportion of ‘‘uninverted’’ questions also fails to show signs of dramatic improvement. ... The proportion of uninverted yes/ no-questions at the end of the study (28. l percent) is not much lower than during the first six months (33.8 percent ... )” (Dąbrowska, 2000, p. 90). Albeit implicitly, this evaluation is based on the traditional, i.e., generative-transformational definition of well-formedness, focusing on the (assumed) underlying linguistic system of an ‘ideal speaker-listener’ (Chomsky, 1965, p. 3) rather than on actual language usage. Recall that within a Chomskyan framework, actual performance data is not regarded to be of direct relevance, a point which has repeatedly been stressed by Chomsky himself: The fact that a certain noise was produced, even intentionally, by an English speaker does not guarantee that it is a well-formed specimen of his language. <?page no="81"?> 67 Under many circumstances it is quite appropriate to use deviant utterances. Furthermore, under normal conditions speech is subject to various, often violent distortions that may in themselves indicate nothing about the underlying linguistic patterns. The problem of determining what data is valuable and to the point is not an easy one. What is observed is often neither relevant nor significant, and what is relevant and significant is often very difficult to observe, in linguistics no less than in the freshman physics laboratory, or, for that matter, anywhere in science. (Chomsky, 1964, p. 28) According to this view, then, the only options are to either regard noncanonical yes-no questions as utterances deviating from well-formed patterns, i.e., as errors, or to treat them as ‘irrelevant’, i.e., as mere performance phenomena which are simply not of interest to the serious linguist. 62 As the previous section has shown, these are exactly the options that developmental studies on question acquisition from a generative perspective have chosen from. Unfortunately, the same goes for the vast majority of usagebased approaches to question development as well. However, if we take the basic tenet of the usage-based approach seriously, i.e., that language structure emerges from language use, it follows that non-canonical yes-no questions should not be regarded as errors, since they are part of the target inventory. Importantly, language use in this context refers primarily to informal, spontaneous conversation, a perspective which has been aptly summarized by George Sampson: For a small child learning to speak English, one of the things to be learned is that it is allowable and normal to begin an informal spoken sentence with its main verb; in the language he is learning, ‘Want your lunch now’ is grammatical. The child is learning to speak English, not (at this stage) learning to write for The Times. (Sampson, 2005, p. 44) But even though it has been noted that “it is the essential nature of language to be spoken, and ... written language is secondary” (Hopper, 1998, p. 171), views concerning grammaticality are often strongly influenced by norms stemming from (highly formal) written language. 63 62 The fact that utterances which do occur might still be considered ‘irrelevant’ is mirrored in statements contrasting grammaticality with acceptability: “[T]here are mild departures from grammaticality which are acceptable in colloquial speech. ... Both subject and auxiliary or auxiliary alone may be omitted: - Do you want your lunch? - You want your lunch? - Want your lunch? ” (Brown, 1973, p. 240) A rare exception is the study by Hendrick (1982), which focuses on the theoretical implications of reduced yes-no and wh-questions within a Chomskyan framework. 63 The implications of regarding spontaneous spoken speech as primary are aptly summarized by Chafe who states that “ordinary talk ... occupies a special place as the kind of language that is most natural in both form and function, the kind of language <?page no="82"?> 68 From a usage-based perspective, grammaticality is to be equated with familiarity (Bybee, 2007, p. 695), i.e., all structures uttered intentionally by (native) speakers of English should be regarded as potentially well-formed. Familiarity is closely linked to entrenchment, which in turn largely depends on frequency of occurrence (also see Langacker, 1987, pp. 59-60, on the relationship between entrenchment and grammaticality). Considering the fact that non-canonical questions have been observed to occur regularly in spontaneous speech, it is highly likely that they are perceived as wellformed by the majority of speakers. Therefore, they should not be regarded as (grammatical) errors in child speech. Note that this claim does not entail that these structures are also to be considered pragmatically appropriate whenever they occur in child speech. 64 With regard to the function of non-canonical questions, it has been suggested that non-inverted (or: declarative) questions (in this case including strong assertives with or without medial auxiliary) are used (mainly) as echoquestions, i.e., they ask for verification, repetition or clarification of what has been said previously (e.g., Quirk, Greenbaum, Leech, & Svartvik, 1985, pp. 814; 835-837; also see Gunlogson, 2003, for an in-depth analysis of rising versus falling declaratives, and Weber 1993, for a more differentiated treatment from a discourse analytic perspective). Reduced questions, on the other hand, are commonly assumed to be functionally interchangeable with canonical questions, since they are viewed as surface reductions of the underlying ‘full’ forms (e.g., Stirling & Huddleston, 2006, p. 1541). 65 Reduced structures (with subject, or auxiliary, or both components elided) and the retrievability of elided elements have been described as follows: The auxiliary is retrievable from the form of the verb when there is one: do with a plain form (want), have with a past participle (seen, driven), be with a gerund-participle (feeling); when the predicate is verbless (at home, you) the understood auxiliary is be. The pronoun ellipted in subject position will normally be understood as you, and if understood as 3 rd person virtually requires an indexical act. (Stirling & Huddleston, 2006, p. 1541) humans must be designed by evolution to produce and comprehend. ... Because conversation is the form of language least influenced by acquired skill, it provides us with the most direct and uncontaminated access to natural mental processes” (Chafe, 1998, pp. 96-97). 64 In fact, in studies on error patterns in child speech, one commonly finds comments on utterances like She doesn’t like cake? and She likes cake? , stating that they are “acceptable in certain contexts [emphasis added] (e.g. as echo question s)” (Ambridge, Rowland, Theakston, and Tomasello, 2006, p. 533). 65 Interestingly, Quirk et al. (1985) note that some questions can be classified as either reduced canonical or reduced non-canonical questions; “For example, Want some? could be a reduced version of You want some? rather than of Do you want some? ” (p. 898). <?page no="83"?> 69 Since weak assertives are ambiguous between non-inverted and reduced questions, it is to be expected that they can fulfil the functions associated with either of these question types. 66 However, there is hardly any research investigating the function of reduced questions on the basis of empirical data (among the notable exceptions are discourse analytic studies subsuming them under declarative questions, e.g., Weber, 1993). Since most studies on children’s use of questions have focused on decontextualized instances, we also do not know much about the ways in which all types of noncanonical questions are used in child(-directed speech). From the view that non-canonical yes-no questions are well-formed (and potentially pragmatically appropriate) target structures, it follows that the richness of formal (and functional) variation within English yes-no questions should be integrated into accounts of question acquisition. To my knowledge, there are only two studies which not only treat ‘non-inverted’ yes-no questions as well-formed but also include their percentage in the input in the analyses (Estigarribia, 2007, 2010; Dąbrowska, 2004): Dąbrowska (2004, Chapter 9), in a re-examination of the Naomi corpus (Dąbrowska, 2000), compares the inversion rates in Naomi’s yes-no questions with those found in her mother’s language, grouped by auxiliary, and observes that “towards the end of the period covered by this study, [the inversion rate] reached levels similar to those found in the input” (Dąbrowska, 2004, p. 188). Such an approach takes into account the structural variation found in the ambient language and does not evaluate the child’s performance with respect to an artificial baseline of an inversion rate of 100%, which is presumably found in the target language. Furthermore, her analyses offer interesting information on the distribution of different noncanonical structures: Whereas the mother’s ‘inversion rate’ for yes-no questions with the modal auxiliaries can and will is very high (97% and 100%, respectively), the rate for do/ does is rather low (62%). Since both strong (17) and weak assertives (18) are treated by her as ‘uninverted’ and the number of the former is reported to be very low (Dąbrowska, 2004, p. 189), this essentially suggests that weak assertives like (18) occur quite frequently in the input. Since she does not provide any precise breakdown of the category ‘uninverted’, it is not possible to ascertain how many of the ‘uninverted’ questions are strong assertives (with/ without medial auxiliary) and weak assertives, respectively. (17) You do want some? (18) You want some? 66 Whether speakers interpret the intended meaning of utterances like You want some? solely on the basis of context or whether, in addition, different formal (most likely prosodic) characteristics are associated with the different functions is an open question. <?page no="84"?> 70 This study gives us some idea of how non-canonical yes-no questions can be taken into account in developmental research within a usage-based framework. However, we are still far away from having a complete picture. First of all, the samples used by Dąbrowska are rather thin and she herself states that “one would ideally want to compare the proportions of inverted and uninverted questions for each auxiliary + subject combination at different stages of development” (Dąbrowska, 2004, p. 188). Along similar lines, a corpus with a higher density is needed in order to evaluate in how far the learning of non-canonical constructions can also be accounted for in terms of the combination and gradual abstraction of item-based constructions. The only comprehensive study not only taking into account but actually focusing on the role of non-canonical yes-no questions is provided by Estigarribia (Estigarribia, 2007, 2010). His analyses are based on data taken from CHILDES-corpora of six children acquiring American English (Sarah, Adam, Eve: Brown, 1973; Shem: Clark, 1978; Abe: Kuczaj, 1976; Naomi: Sachs, 1983). 67 At the heart of his study is the claim that non-canonical yes-no questions in the input may actually make the learning of canonical yes-no questions easier, since children can start out with structurally less complex forms which can then be expanded by what Estigarribia terms right-to-left elaboration. For example, children may start out from (19) and then add a subject (20) and later also an auxiliary (21) (example taken from Estigarribia, 2010, p. 85). (19) Cold? (20) You cold? (21) Are you cold? 68 His prediction is that children will begin with the production of less complex non-canonical yes-no questions, which are gradually expanded by right-to-left elaboration, eventually leading to the mastery and productive use of canonical questions. Indeed, the children in his study are found to 67 Estigarribia (2010) only takes into account the data obtained from the first five corpora. 68 Estigarribia takes tag-questions (You’re cold, aren’t you? ) as belonging to the canonical (auxiliary-initial) class, stating that “although technically not independent clauses, [they] also have an auxiliary-subject structure” (Estigarribia, 2010, p. 71). In other words, parts of his evidence for the emergence of structures resulting from right-to-left elaboration comes from forms which can not result from this exact process. Even though he states that tag-questions were not produced very frequently in his sample (2007, p. iv), this decision may have inflated the counts for canonical questions, overestimating the amount of structures which can be accounted for by right-to-left elaboration. Therefore, I do not consider tag-questions to belong to the same category as auxiliary-initial questions (such as (21)). <?page no="85"?> 71 produce non-canonical questions before canonical ones and thus to (presumably) make use of the right-to-left elaboration strategy. 69 While Estigarribia is generally sympathetic towards the usage-based, constructivist approach, he initially claims that the observed patterns could not be accounted for within this framework, since accounts involving itembased learning “propose that the acquisition of yes/ no questions proceeds just like that of wh-questions. In both cases, children learn particular question tokens that have an initial aux-subj cluster” (Estigarribia, 2007, p. 68). However, one of the studies which he refers to (Dąbrowska, 2000) is based on the Naomi corpus, i.e., the same child whose data is included in his study and for whose development he concludes that “unlike the other children, Naomi produces early canonical questions, slightly before the emergence of non-canonical classes” (Estigarribia, 2007, p. 251). This suggests that the developmental path of questions in this particular child may indeed best be accounted for in terms of the item-based learning of utterance-initial auxsubj clusters rather than right-to-left elaboration based on non-canonical yesno questions. More importantly, however, Estigarribia himself mitigates his initial claim that non-canonical structures will be used productively prior to the productive use of canonical questions (2007, p. 68), concluding that “when [canonical questions with initial auxiliaries] emerge, they are not necessarily [emphasis added] restricted to a few lexically based frames” (Estigarribia, 2010, p. 79). So while “there is no a priori reason to suppose that processes of structural abstraction must wait until initial auxiliaries are produced” (Estigarribia, 2010, p. 79), there is no a priori reason to assume that auxiliaryinitial questions (and also non-canonical questions) are based on abstract representations, either. Of course, “naturalistic observations of children’s speech can never prove that the limited patterns they use really reflect lowscope knowledge rather than underlying abstract structure” (Cameron- Faulkner et al., 2003, p. 844) and thin samples in particular can lead to an underestimation of children’s productivity (Rowland & Fletcher, 2006; Tomasello & Stahl, 2004). However, children are found to be less productive than adults even when large samples are used and children’s productions are compared with matched samples from adult speech (Rowland & Fletcher, 2006). There is also accumulating evidence that the ability to generalize indeed develops only gradually (see, e.g., Bannard & Matthews, 69 Right-to-left elaboration bears a resemblance to Freudenthal at al.’s (2002) application of the connectionist MOSAIC model to account for optional infinitive phenomena in English and Dutch, in which utterances were assumed to be processed (and acquired) from right-to-left. However, as pointed out by Estigarribia (2010, p. 83), the MOSAIC model aims at accounting for errors, not for the successful learning of well-formed structures and also assumes that less complex strings are ‘unlearned’ once the more complex variants are mastered. <?page no="86"?> 72 2008, for evidence that children store multi-word units and lexically specific frames). Furthermore, previous research from a decidedly usage-based, constructivist perspective simply did not take non-canonical questions into account, so that the potential item-based nature of non-canonical questions is still largely unexplored and Estigarribia himself concludes that “RTL [=right-toleft-] elaboration is not inconsistent with item-based learning processes” but that “further studies with denser sampling regimes are needed to possibly further reconcile both positions” (Estigarribia, 2010, p. 86). In fact, he states that the operations of juxtaposition and superimposition proposed by Dąbrowska & Lieven (2005) are in effect not much different from right-to-left elaboration. In his view, the main differences are that rightto-left elaboration “restricts adjunction to the interrogative domain and ... restricts adjunction to operate at the left-edge for [yes-no questions]” (Estigarribia, 2010, p. 86). He particularly regards the second constraint to be of importance, since errors which would result from unconstrained juxtaposition (e.g., *She drinking is? ) are unattested. However, right-to-left elaboration only contains the required restriction concerning the linear ordering of elements on a theoretical level, without providing an explanation for how the child might come to the conclusion that *She drinking is? is not possible. Along similar lines, “RTL elaboration assumes but does not explain [emphasis added] why children only elaborate interrogative strings, and not other kinds of structures (this is an important difference with respect to the itembased juxtaposition mechanism)” (Estigarribia, 2010, p. 88). Therefore, it seems preferable to aim at refining the more general mechanism of juxtaposition, not least since the usage-based framework also offers the opportunity to integrate item-based effects and the semantics of the respective constructions, while string containment in right-to-left elaboration does not integrate semantic factors (Estigarribia, 2010, p. 76). Apart from the fact that the string containment relations that the different classes of yes-no questions (see examples (19) through (21)) enter into are not constrained by semantics, discourse-functional factors are also not taken into account in the analysis of individual (classes of) questions (Estigarribia, 2007, pp. 322-323). Therefore, the discourse-functional properties, particularly of non-canonical questions, are still largely unexplored (the assumption being that ‘uninverted’ questions mostly function as echo-questions, while weak assertives may also be used interchangeably with their ‘full’, canonical counterparts; Estigarribia, 2007, p. 111). In his suggestions for further research, Estigarribia briefly points out a few examples suggesting that different question constructions “alternate in similar discourse contexts, and that they are sometimes used to paraphrase one another” (Estigarribia, 2007, p. 324). In Chapters 6 and 7 of this book, these paraphrase relations are analyzed on the basis of work conducted on so-called variation sets (Küntay & Slobin, 2002), shedding further light on the discourse relations that weak <?page no="87"?> 73 assertives and strong assertives with medial auxiliaries enter into and the discourse function fulfilled by this type of non-canonical question. 3.3.2 Summary Canonical as well as non-canonical questions are extremely prevalent in caretaker-child interaction, comprising between 21 and 44% of the language addressed to children. While generative approaches to the acquisition of questions mainly focus on the mastering of abstract rules (most notably SAI), usage-based, constructivist research stresses the importance of itembased constructions, arguing that children start out and only gradually abstract away from particular auxiliary-subject combinations, a prediction that is indeed corroborated by empirical research. These competing assumptions also entail different explanations for children’s error patterns: From a generative perspective, children are generally thought to possess adult-like rules from rather early on, with errors stemming from performance factors. In cases where errors are thought to result from non-adult-like rules, this is predicted to influence children’s performance on questions containing particular categories of items (e.g., all questions involving the copula BE). In contrast to this, usage-based, constructivist accounts predict error rates to be influenced by the specific lexical material involved, an expectation that is indeed borne out by experimental as well as naturalistic data. In addition to this, the relative input frequency of particular question-frames is found to influence error rates, with highly frequent frames exhibiting low error rates. However, some errors (especially those involving negation) are still more frequent than would be expected on the grounds of input frequency alone. Strikingly, generative as well as usage-based research on questions have treated non-canonical yes-no questions (i.e., ‘non-inverted’, reduced, and fragmentary questions) either as errors or have excluded them from analyses, despite the fact that these structures have been reported to be quite frequent in child-directed speech. Apart from that, non-canonical yes-no questions are found to be acceptable in descriptions of adult-to-adult speech. There are hardly any studies taking these structures into account as legitimate targets for acquisition (but see Dąbrowska, 2004; Estigarribia 2007, 2010). Therefore, the acquisition and use of these structures is still underresearched, both in adult-to-adult and child(-directed) speech. 3.4 The Discourse Functions of Canonical Yes-No Questions Meaning is what language is all about; the analyst who ignores it to concentrate solely on matters of form severely impoverishes the natural and necessary subject matter of the discipline and ultimately distorts the character of the phenomena described. (Langacker, 1987, p. 12) <?page no="88"?> 74 As has been shown in the preceding sections, most (generative as well as usage-based) research on yes-no questions is based on the assumption that the vast majority of cases exhibit certain formal characteristics, i.e., SAI, and that this is largely due to established descriptions of the adult target repertoire. Along similar lines, the basic assumption found in most developmental research on the function of yes-no questions is that they are prototypically used to request missing information (with the function of non-canonical questions being rarely considered at all). Much research focusing on the formal characteristic just assumes questions to be requests for missing information, without paying much attention to discourse-functional properties of these constructions at all. Functionally oriented research, in contrast, mostly falls into one of the following two categories: On the one hand, there are studies which are interested in children’s emerging ability to answer questions appropriately. Since the focus is on the semantic and pragmatic appropriateness of verbal responses, questions which do not fulfill an information-seeking function are often excluded. On the other hand, there are studies focusing exclusively on the (presumably) non-prototypical functions of questions, treating them as ‘indirect speech acts’ whose correct interpretation (presumably) requires particular inferencing skills. Strikingly, there are hardly any studies exploring form-function mappings found in canonical yes-no questions. This is mainly due to the fact that generative approaches, which dominated the field for much of the latter half of the 20 th century, were hardly interested in the functional aspects of linguistic structures at all. Research from a functional perspective, which proliferated in the 1970s and ‘80s, in turn mostly neglected formal characteristics. In many cases, this resulted from a rejection of generative grammar as a theoretical framework of linguistic description, leading to a situation in which many developmental psycholinguists have abandoned linguistic theory altogether, basing their explanations on more general principles of representation and learning taken from developmental psychology and cognitive science. Although these fields also have a lot to offer, they do not provide the detail or the rigour that we once derived from a fruitful relationship with linguistics. (Bates, 1998, p. 462) With the advent of Construction Grammar, an alternative framework of linguistic description has become available: Besides offering the tools for the description of formal as well as functional characteristics of linguistic structures, the symbolic view of constructions (as form-function pairings) ultimately requires both sides to be considered for a full description of the structures encountered (Behrens, 2009a, p. 432). Not only does this provide the opportunity to describe linguistic structures in a very comprehensive way within one coherent framework, it also offers the possibility to analyze <?page no="89"?> 75 prototypical form-function mappings on a rather fine-grained level. However, to date only a few usage-based studies have explored functional factors in detail (Abbot-Smith & Behrens, 2006; Steinkrauss, 2009, 2011), with the current work aiming at making a contribution concerning this matter. The next subsections are structured as follows: Section 3.4.1 provides a summary of canonical yes-no questions as requests for information, both from a descriptive and developmental perspective. Section 3.4.2 deals with the treatment of yes-no questions as ‘indirect speech acts’, with a focus on studies from a developmental perspective. Lastly, Section 3.4.3 discusses previous research on the functions (and form-function mappings) of yes-no questions in child(-directed) speech. 3.4.1 Canonical yes-no questions as requests for information In most descriptions of adult language (as provided by reference grammars), we can find the assumption that formal interrogatives prototypically function as requests for information and it has been referred to as “one aspect of questions (where it is discussed) about which theoreticians and researchers of all types are agreed” (Sinclair & van Gessel, 1990, pp. 938-939). For example, Quirk et al. (1985) state that “simple sentences may be divided into four major syntactic types differentiated by their form” and that “their use correlates largely [emphasis added] with different discourse functions” (p. 803). In their view, a “direct association between syntactic class and semantic class is the norm”, with questions being “primarily used to seek information on a specific point” (Quirk et al., 1985, p. 804). Even though the use of the term ‘norm’ leaves room for cases in which syntactic class and semantic class do not match, those cases can only be viewed as exceptions deviating from the prototypical mapping. Consequently, syntactic questions exhibiting other functions (such as directives) are interpreted as ‘indirect speech acts’ (Quirk et al., 1985, pp. 805-806). Similar views concerning the prototypical formfunction mappings in questions are also expressed by Huddleston and Pullum (2006, p. 853) and Biber et al. (1999, p. 202). In both cases, we also find the acknowledgement “that interrogatives aren’t always used as questions” (Huddleston & Pullum, 2006, p. 854) and that questions “are often [emphasis added] used for purposes other than asking for information” (Biber et al., 1999, p. 207) but the respective cases are still considered to constitute ‘indirect’ uses. In developmental research as well it is the assumed prototypical use of questions that has been focused on, especially since the information-eliciting function of questions was believed to encourage children’s vocalizations and children’s answering behavior was thought to be a good indicator of their mental development. For example, it has been suggested that “answering questions is among the first clearly discourse-bound obligation to which children are sensitive” (Ervin-Tripp & Miller, 1977, p. 14) and by about three <?page no="90"?> 76 years of age, children are much more likely to produce an adjacent, contingent utterance following a question than following a non-question (Bloom, Rocissano, & Hood, 1976). Often, questions are asked just to keep the conversation going, suggesting that caretakers practice conversational conventions with their children by making heavy use of questions as turn-allocating devices (Olsen-Fulero & Conforti, 1983, p. 496). While it is necessary to restrict the focus of investigation to particular structures (in these cases interrogatives expression requests for information), closer inspection of individual studies reveals that slightly different criteria for the consideration of a structure as a ‘real’ question are applied. For example, questions functioning as directives (e.g., Will you bring it over? ) were regarded as nonquestions by Bloom et al. (1976, p. 539), who studied children’s responses to questions vs. non-questions. Along similar lines, in a study by Olson-Fulero and Conforti (1983) on children’s responsiveness to questions, “only those mother utterances which actually elicited a verbal response from the child were considered to be questions” (Olsen-Fulero & Conforti, 1983, p. 500); this subsumed, for example, test questions and requests for clarification, while excluding directives and utterances directing the child’s attention (such as See? ). Ervin-Tripp and Miller (1977, p. 15) additionally discarded yes-no questions that functioned as requests for confirmation, thus applying even stricter criteria concerning what counts as a genuine question. Interestingly, they note that a “diversity of function was especially apparent in the polar (“yes”/ “no”) questions” (Ervin-Tripp & Miller, 1977, p. 15) but since detailed quantitative information is not provided, the percentage of questions fulfilling ‘non-prototypical’ functions cannot be assessed. A mirror-image of the same problem is encountered in studies on so-called ‘indirect speech acts’, in which information-seeking questions are excluded. Indirect speech acts are supposed to be rather difficult to comprehend for young children, since their correct interpretation presumably requires the application of inferencing skills. The next section provides a brief summary of the conceptualization of ‘indirect’ speech acts before discussing developmental studies on children’s responses to these structures. 3.4.2 Yes-no questions as ‘indirect speech acts’ The term ‘indirect speech act’ was coined by John Searle (1969, 1975). Within the larger framework of ‘Speech Act Theory’ (which Searle developed on the basis of work by Austin, 1962), all utterances can be described to perform four different acts: First of all, the utterance act, i.e., the actual production of meaningful speech sounds; secondly, the propositional act, which encompasses reference and predication; thirdly, the illocutionary act, i.e., the discourse function to which end the speaker uses the utterance (such as questioning, commanding, or requesting). The fourth act is the perlocutionary act which is also called perlocutionary effect, since it describes the effect <?page no="91"?> 77 the utterance has on the interlocutor (such as convincing or scaring; Searle, 1969, pp. 23-25). ‘Indirect speech acts’ are defined as “cases in which one illocutionary act is performed indirectly by way of performing another” (Searle, 1975, p. 60). For example, with uttering Can you reach the salt? , the speaker performs the illocutionary act of requesting indirectly by using a structure which is commonly associated with the illocutionary act of questioning. In trying to determine how it is possible for the hearer to arrive at the correct interpretation of these utterances, Searle proposes a serial processing model of speech act comprehension, in which the literal meaning serves as the basis for an inferencing process, which proceeds from the rejection of the literal interpretation (on the grounds of information such as background knowledge and the assumption that the speaker is communicatively cooperative) to the establishment of the primary illocution, i.e., the intended meaning (Searle, 1975, pp. 62-63; also see Clark & Lucy, 1975, for a similar proposal and experimental evaluation). 70 In this Standard Pragmatic Model, the term ‘indirect speech act’ thus not only serves as a mere descriptive term but is claimed to possess psychological reality in the sense that the indirectness leads to a higher processing cost in the interpretation of the respective utterances. While later research suggested that the literal and the non-literal meaning of indirect speech acts are computed in parallel, this view still entails that indirect speech acts are associated with a higher processing load than direct speech acts (Clark, 1979; Clark & Schunk, 1980). The same was suggested for other instances of non-literal language, such as idioms and figurative uses (see Gibbs, 1994, for a comprehensive discussion). That some indirect speech acts can even be conceptualized in the same way as idioms was proposed by Sadock (1970, 1974b), who referred to expressions like Would you mind closing the window? as ‘whimperatives’ and suggested that they are used like idiomatic expressions, which are also closely associated with a non-literal interpretation. The treatment of conventionalized indirect speech acts as idioms was later largely rejected on the grounds that the former, but not the latter, can evoke both literal and non-literal interpretations, depending, for example, on contextual factors (Munro, 1979). 71 Nevertheless, psychological evidence concerning the interpretation of idioms was suggested to be indicative of the way people handle indirect speech acts. Since it has been found that people are faster in processing the ‘non-literal’ meaning of conventional idiomatic expression than the ‘literal’ meaning (Gibbs, 1980), it has been argued “that the traditional distinction between literal and metaphoric 70 Interestingly, Searle has made a very similar proposal concerning the interpretation of another case of ‘non-literal’ language use, i.e., metaphors (Searle, 1979). For a discussion (and rejection) of this proposal, see Lakoff (1993, p. 205). 71 Nevertheless, we still find statements saying that expressions like Can you turn the light on? “are idiomatically or conventionally [emphasis added] used in the performance of indirect speech acts” (Huddleston and Pullum, 2006, p. 865). <?page no="92"?> 78 language is better characterized as a continuum between conventional and unconventional utterances” (Gibbs, 1980, p. 149). This difference between conventionalized and non-conventionalized indirect speech acts has already been noticed by Searle. Somewhat surprisingly, he states that “the problem is made more complicated [emphasis added] by the fact that some sentences seem almost to be conventionally used as indirect requests” (Searle, 1975, p. 60), even though he does not elaborate on why this should be the case. The distinction between conventionalized and non-conventionalized indirect speech acts has been addressed by other researchers working within the classic paradigm as well (Clark, 1979; Morgan, 1978) but “no explicit theory of conventionalized indirect speech acts has been proposed” (Stefanowitsch, 2003, p. 105). More recent cognitive approaches conceptualize indirectness differently: Within the theory of speech act metonymies (Panther & Thornburg, 1998, 2003a, 2004; Thornburg & Panther, 1997), it is assumed that speakers possess cognitive models of specific scenarios, e.g., for the act of ‘giving’. At the core of the model is the act of giving itself; however, there are certain, more peripheral notions associated with it, i.e., that the speaker wants something from the hearer or that the hearer is both able and willing to perform the respective act. This latter part of the model is evoked by conventionalized indirect speech acts such as Can you hand me the keys? . By evoking an activation of any part of the model, an activation of the whole model is achieved through association (metonymic links). In this view, the ‘non-literal’ interpretation of an indirect speech act can be derived ad-hoc, but it “can also, through frequency of use, become a conventionalized meaning, stored separately in the lexicon” (Panther & Thornburg, 2004, p. 97). This possibility is explored further by an explicit theory of conventionalized indirect speech acts, outlined in some detail by Stefanowitsch (2003). Within this approach, “the theory of speech act metonymies ... provides the motivation for the similarity in form between the conventionalized indirect speech acts and the direct speech act on which they are based” (Stefanowitsch, 2003, p. 105). However, conventionalized indirect speech acts are viewed as constructions whose “illocutionary force is directly linked to their form (and [which] are thus not really ‘indirect’ speech acts at all)” (Stefanowitsch, 2003, p. 106). Formally, these conventionalized indirect speech acts are found to differ from both the corresponding direct constructions and nonconventionalized indirect speech acts. For example, only conventionalized indirect speech acts, but not the corresponding direct constructions, can contain the request marker please (i.e., Can you open the door? can function as both a polite request for action and a request for information on the hearer’s ability to perform the action, while only the first interpretation is possible for Can you please open the door? ). To my knowledge, the implications of this approach have not been explored from a developmental perspective. After all, it is during early child- <?page no="93"?> 79 hood that the assumed linguistic conventions are established. And while we indeed find much evidence suggesting that even rather young children are able to respond appropriately to conventionalized indirect speech acts, this has been regarded as a somewhat surprising finding, stemming from the classic assumption that complex inferencing processes must be involved. In contrast to this, a usage-based, constructivist perspective on conventionalized speech acts in the sense of Stefanowitsch (2003) would expect children to learn the meaning of these constructions in the same way as other conventionalized constructions, thus being able to interpret them correctly from early on. There is much anecdotal evidence suggesting that conventionalized indirect speech acts are understood from early on. For example, Wells (1976) suggests that the possibility of a literal interpretation of utterances such as Why don’t we have a cup of coffee? is realized only later on, with the emergence of metalinguistic skills, leading to replies such as Because we can’t afford it (Wells & Ferrier, 1976, p. 330). Concerning syntactic questions of the type Why don’t you X? , addressed to the child Adam, Roger Brown made the following remark: What has always amazed me about these pseudo questions is that Adam, long after he was answering genuine Why questions with ‘Because’, and some explanation or other, did not once apply this syntactic and semantic analysis to fake Why questions, but invariably treated them as orders - which is what they, pragmatically speaking, were. (Brown, 1977, p. 21) 72 While children’s emerging linguistic abilities are generally ‘amazing’, Adam’s correct interpretation of rhetorical Why-questions is not particularly surprising from a usage-based perspective assuming that they are learned as form-function pairings. Thus, if the item-based construction Why don’t you X? is always used to suggest actions, this form-function mapping can be learned directly from the input. However, even though the formulaic nature of many conventionalized indirect speech acts has been noticed, indirect speech acts were investigated almost exclusively on a rather abstract level. For example, Garvey (1975) studied the production of children’s directives (in question or imperative form) in naturalistic samples obtained from 36 dyads of nursery school children (age range: 3; 6-5; 7). She found that while the older children produced more indirect requests than the younger ones, all children seemed to possess the necessary prerequisites to comprehend and produce indirect requests. However, she does not analyze the different question forms used to express directives. 72 Further evidence for the fact that Why don’t you X? may be learned as a separate formula comes from Thornton (2008, p. 125), who reports that her subject AL did not invert in other negative questions, but consistently used Why don’t you X? in the correct, i.e., ‘inverted’, form. <?page no="94"?> 80 In a study based on naturalistic sampling of five mother-child dyads with younger children (age range 1; 7-2; 4), Shatz (1978) investigated children’s comprehension of mother’s imperatives and questions used as directives, also treating all questions as belonging to one formal category. She found “the likelihood of obtaining an appropriate response to directives was roughly 50% and was not influenced by the choice of imperative versus question form” (Shatz, 1978, p. 42), indicating that both forms were understood equally well by the children. This impression is corroborated by other naturalistic observations: In a study on the early answering behavior of her daughter Kelly, Horgan (1977) observes that, even at only 1; 3-1; 8 years of age, Kelly gave appropriate non-verbal responses to yes-no questions such as Do you want to go outside? (by going to the door; Horgan, 1977, pp. 160- 161), suggesting that she understood the intended meaning through the contextual embeddedness of the utterance (for similar observations, also see Dore, 1974; Ervin-Tripp, 1977; Garvey, 1975). Another strand of research on the comprehension and production of indirect speech acts in children is constituted by experimental studies (e.g., Bara & Bucciarelli, 1998; Bernicot & Legros, 1987; Bucciarelli, Colle, & Bara, 2003; Carrell, 1981; Reeder, 1980; for a review, see Gibbs, 1994). Some of these studies were designed to force children to rely solely on the linguistic form for the interpretation of the respective utterances (e.g., Carrell, 1981). They tap into children’s metalinguistic skills (since little to none contextual information is provided) and often explicitly included only nonconventionalized indirect speech acts (e.g., Bernicot & Legros, 1987), so it is not surprising that even older children (age 4-7) have been found to be worse than adults in interpreting different kinds of indirect requests. More recent empirical evidence suggests that children from about 2; 6 onwards understand (conventional) indirect speech acts correctly (Bara & Bucciarelli, 1998). A follow-up study, in which 160 children (age range: 2; 6-7; 0) were asked to attribute communicative intentions to characters in videotaped stories, revealed that children perform significantly better on conventional indirect speech acts (e.g., Sorry, could you close the window? ) than on non-conventional indirect speech acts (e.g., Excuse me, I am studying in reaction to noisy behavior by the interlocutor; Bucciarelli et al., 2003). What is more, conventional indirect speech acts were also understood significantly better than direct utterances (such as Mummy, pick me up! ). Even though they do not appeal to Cognitive Linguistics or Construction Grammar, they interpret the results within a Cognitive Pragmatics framework, suggesting that both direct and conventional indirect speech acts “are simple acts, immediately linking the move to the game” (Bucciarelli et al., 2003, p. 213). To sum up: Questions expressing functions other than requesting information (such as offers and directives) are assumed to be ‘indirect’ speech acts in the sense that their correct interpretation requires complex inferential <?page no="95"?> 81 processing (Searle, 1969, 1975). More recent, cognitive approaches suggest that indirect speech acts may be motivated by metonymy, linking peripheral parts of a scenario (e.g., the hearer’s ability to perform an action) to the core (e.g., the act of ‘giving’). In this view, linguistic convention is a decisive factor and it has been suggested that indirect speech commonly associated with their ‘indirect’ interpretation may be learned and used as form-function pairings (Panther & Thornburg, 2004; Stefanowitsch, 2003). Naturalistic as well as experimental evidence on the comprehension of conventional indirect speech acts by children indeed suggests that this might be the case, since these structures are understood (and used appropriately) from about 2; 6 onwards. However, to date there are only a few studies which have addressed the issue of convention explicitly. In addition to that, form-function mappings have only been analyzed on an abstract level, meaning that not much is known about which item-based constructions commonly function as conventional indirect speech acts. Along similar lines, many studies assume that questions are primarily used to request missing information, with (even conventional) indirect uses constituting a rather small percentage. Concerning these aspects, some information can be gleaned from studies focusing on the functions that questions fulfil. The next section provides a brief discussion of this research. 3.4.3 Developmental research on the functions of questions As we have seen in the preceding section, it is commonly assumed that there are prototypical mappings of sentence type to discourse function, the most common function of formal yes-no questions being that they request missing information. Consequently, yes-no questions fulfilling other functions were conceptualized as indirect speech acts. Since the aim was to evaluate children’s comprehension and production of this utterance type, the percentage of yes-no questions actually fulfilling direct vs. indirect functions was not assessed. In developmental research focusing on the functional characteristics of syntactic questions, on the other hand, we find closer analyses of the functional variability of questions and form-function mappings on different levels of granularity. Importantly, these studies proceed from a formal perspective, i.e., they look at the range of functions expressed by particular forms, not at competing forms that might be used to express similar functions. Proceeding from a rather strict definition of questions as sincere requests for missing information, we are almost inevitably in the realm of what has been termed ‘indirect’ as soon as we postulate uses which deviate from this (presumably prototypical) mapping. Besides, even the class of ‘real’ questions exhibits some heterogeneity at a fine-grained level, since they can be used to ask for different kinds of information (e.g., clarification, causal explanations, justification). This has also <?page no="96"?> 82 been noted in developmental studies, in which different subtypes of information-seeking questions were identified (e.g., Bishop, Chan, Hartley, & Weir, 2004; Davis, 1932; Meyer & Shane, 1973; Piaget, 1928). One of the first systematic studies on the different functions expressed by syntactic questions was provided by Holzman (1972), who classified interrogatives in four samples of 100 child-directed utterances each according to their discourse function (the samples are taken from the Adam, Eve, and Sarah corpus, respectively, two from a period in which the child’s MLU was 2 morphemes, the other two from a period in which the child’s MLU was 4 morphemes). Even though requests for behaviour were subsumed under the category ‘real question’, she found the majority (i.e, 50-84%) of mother’s interrogatives not to serve as ‘real’ questions (but, for example, as test or rhetorical questions; Holzman, 1972, p. 314). An additional analysis of two 350-utterance samples of the three children (MLU=3) revealed that children primarily used syntactic questions to request missing information but that they still made use of the other functions as well. She concluded that this flexibility is made possible by the “speech act's embeddedness in a communication context which is defined by the action under way and the personal associations of the actors” (Holzman, 1972, p. 312). While providing a rather detailed account of the functional side of interrogatives, she does not differentiate between subclasses of syntactic questions such as yes-no questions, wh-questions and tag questions (Holzman, 1972, p. 313). To my knowledge, one of the only studies providing an analysis of formfunction mappings in syntactic questions on a more fine-grained level is Shatz (1979). Serving as the data were all questions (n = 1142) produced by 17 mothers in 15-minute play sessions with their children (age range: 1; 6- 2; 10). The children were assigned to either the ‘low’ or the ‘high’ group (i.e., an MLU of 2 and 3-4, respectively). All questions were first assigned to one of 11 functional groups (e.g., testing, directive, calling attention). Formally, questions were grouped as belonging to different ‘question-frames’. The aim “was to capture similarities among forms while staying as close to the data as possible, since it seemed more likely that the child listeners might construct early classificatory systems for form types at this level rather than at a more abstract one” (Shatz, 1979, p. 1094). This type of coding, which bears a striking similarity to usage-based descriptions of item-based constructions and low-level schemata, was used as the basis for the identification of prototypical form-function mappings (‘characteristic pairs’) by applying the following criteria: “a given frame had to be paired with a given function more often than with other functions, and the pair had to occur at least once in the speech of more than half of the mothers in a group” (Shatz, 1979, p. 1094). One of the predominant frames was Can you +V (+X)? , which was used as a directive in 33% (low group) and 35% (high group) of all cases, respectively. Strikingly, 38% of all questions addressed to children of the low <?page no="97"?> 83 group were characteristic pairs, while the corresponding figure for children in the high group was only 25% (Shatz, 1979, p. 1096) and low-group children were better at responding appropriately to characteristic than noncharacteristic pairs, suggesting that the “production of characteristic pairs can be useful at the early stages of acquisition for broadening children's understanding of conversational demands and leading them away from language-action mappings” (Shatz, 1979, pp. 1098-1099). First of all, these studies provide support for the claim that utterances commonly functioning as ‘indirect’ speech acts are used (and thus may be learned) as conventionalized form-function pairings, i.e., constructions. Secondly, they suggest that the use of syntactic questions as ‘real’ questions may not be that prototypical after all. Research on other languages suggests that the functional variability of ‘interrogatives’ is not exclusive to English; see, e.g., Vaidyanathan (1988), for a study on form-function mappings in the questions of Tamil-speaking children, and Clancy (1989), for a functional analysis of wh-questions in Korean children. In their study on peer-to-peer communication of French children (aged 3-4; 10), Sinclair & van Gessel (1990) concluded that “the traditionally defined central function of questions to obtain information not previously known (or not certainly known) by the speaker - ... is not obviously or clearly present in our data” (Sinclair & van Gessel, 1990, p. 939). 3.5 Terminological note Before coming to the summary and the research questions concluding this chapter, a brief note on terminology is in order: Yes-no questions are also commonly referred to as ‘closed’ or ‘polar interrogatives’ (e.g., Ginzburg & Sag, 2001) and sometimes the impression is given that the two terms ‘question’ and ‘interrogative’ may be used interchangeably. It has been pointed out that the term ‘question’ refers to a semantic type “whereas interrogative ... is to be defined in terms of syntactic form” (Huddleston, 1994, p. 413). As we have seen in the preceding sections, not all questions are interrogative in form and not all interrogatives serve as questions, i.e., there is not always a match between form and function. Therefore, it is indeed advisable to keep the syntactic and semantic/ pragmatic dimensions separate by differentiating between ‘questions’ and ‘interrogatives’ (for an attempt at a purely functional definition of questions, see Tsui, 1992, for a comprehensive discussion of interrogatives in terms of syntactic form, see Huddleston, 1994). I have chosen to use the label ‘(non)canonical yes-no question’ throughout this book to refer to the structures which are the focus of my investigation for the following reasons: As has already been pointed out, (non)canonical yes-no questions are related by family resemblance, i.e., there is neither a common formal property (like SAI) nor a common functional property (like <?page no="98"?> 84 requesting missing information) that is shared by all members, so that no clear case can be made for either the term ‘question’ or the term ‘interrogative’. Therefore, I adopt the term ‘yes-no question’ rather than ‘interrogative’, not only because it is the label commonly used in language acquisition studies, but also to stress the importance of the functional dimension of these constructions, which has largely been neglected so far. 3.6 Summary Canonical as well as non-canonical questions are used with a high frequency in child-directed speech. Generative approaches to the acquisition of questions mainly focus on abstract rules (most prominently SAI), while usagebased, constructivist research is based on the assumption that children gradually abstract away from particular auxiliary-subject combinations learned from the ambient language. Within both paradigms, non-canonical yes-no questions (i.e., ‘non-inverted’, reduced, and fragmentary questions) were treated either as errors or have excluded them from analyses, despite the fact that these structures occur frequently in child-directed speech and are found to be acceptable adult-to-adult speech as well. Consequently, there are hardly any studies taking these structures into account as legitimate targets for acquisition, and the acquisition and use of these structures is still underresearched, both in adult-to-adult and child(-directed) speech. The same can be said for the functional characteristics of canonical yes-no questions, even though there is much work discussing their use on a theoretical level. Since they are often assumed to prototypically function as sincere requests for missing information, yes-no questions used for different purposes (e.g., as directives) are commonly conceptualized as ‘indirect speech acts’, whose interpretation requires the application of inferencing strategies on the part of the hearer. Empirical studies show that at least ‘conventionalized indirect speech acts’ (i.e., questions commonly fulfilling a function other than requesting information) are not more difficult to process than direct speech acts and even rather small children have been reported to react appropriately to such indirect speech acts. It has been suggested that these conventionalized cases can be formalized as constructions, i.e., formfunction pairings whose illocutionary force is directly associated with their form (Stefanowitsch, 2003). However, this claim has not been evaluated on the basis of developmental data, even though it is during early childhood that linguistic conventions are established. Furthermore, there are hardly any developmental studies analyzing form-function mappings in questions at all. In particular, form-function pairings have not been analyzed on a finegrained level, leaving the characteristics of these mappings and their role in language development largely unexplored. <?page no="99"?> 85 3.7 Research Questions The overall aim of this book is two-fold: The first goal is to integrate noncanonical yes-no questions into a usage-based, constructivist account of language development and use. The second aim is to augment our understanding of functional factors by analyzing how discourse function influences the acquisition of particular (canonical) question-constructions. This evaluation and expansion of previous research in the field is based on 1. a corpus of adult-to-adult speech (the BNC) and 2. a high-density corpus of a boy learning English as his first language. The aim is pursued by addressing the following research questions: (1) How frequently are different types of (canonical as well as non-canonical) yes-no questions encountered in adult-to-adult speech? (Chapter 4) (2) How frequently are different types of (canonical as well as non-canonical) yes-no questions encountered in child(-directed) speech? (Chapter 5) a. (How) do the relative frequencies of non-canonical yes-no questions change over time, both in the caretakers’ and the child’s language? (3) (How) are PREDs (such as Want some? ) and weak assertives (such as You want some? ) related to their canonical counterparts (Do you want some? )? (Chapter 6) a. Are canonical questions of the type Do you want some? learned by Thomas through the gradual expansion of PREDs and weak assertives? b. Which paraphrase relations do weak assertives enter into? c. Which discourse functions do weak assertives fulfill? (4) What is the relationship between strong assertives (such as That’s a ball? ) and their ‘inverted’ counterparts? (Chapter 7) a. How frequent are strong assertives in child(-directed) speech? b. (How) does the ‘inversion rate’ change over time? c. Which discourse functions do strong assertives fulfill? (5) (How) do functional factors influence the acquisition of canonical yes-no questions? (Chapter 8) a. (How) is the relative input frequency of initial auxiliarysubject combinations related to the child’s learning and frequency of use of the respective canonical constructions? b. Can functional factors explain cases in which the child’s developmental path differs from the one predicted by the relative input frequency of the respective constructions? c. Does the predominantly ‘indirect’ use of particular questions in child(-directed) speech lend support to the claim that conventionalized indirect speech acts (Can you pass the salt? ) can be conceptualized as constructions (as proposed by Stefanowitsch, 2003)? <?page no="100"?> 86 4 Yes-No Questions in Adult-to-Adult Speech Canonical as well as non-canonical yes-no questions have been found to be very prevalent in child-directed speech. In contrast to this, it has been suggested that questions are nearly absent in adult-to-adult language (Newport et al., 1977, p. 122). Apart from the fact that this claim is based on sparse data, no difference is made between the various types of syntactic questions leading Estigarribia to conclude that “we have no estimates to date of the prevalence of yes/ no questions in [adult-to-adult speech]” (Estigarribia, 2007, p. 139). This is not entirely true, since, for example, corpus-based reference grammars provide frequency information on utterance types found in different types of language (formal/ informal; written/ spoken): Since questions are commonly found to be more frequent in spontaneous, informal conversations than in (formal) written language, the prevalence of questions in the former is often stressed (Biber et al., 1999, p. 211). In comparison with the percentages obtained in studies on child-directed speech (ranging from 21-44%), they indeed appear to be used with a rather low frequency. All previous estimates on adult-to-adult speech, however, deal with different types of syntactic questions on a rather abstract level. To my knowledge, there are no comprehensive quantitative corpusbased studies concerning the frequency with which non-canonical yes-no questions occur. It is important to assess whether and to which degree they are part of the adult target repertoire. Otherwise, the fact that non-canonical structures are found in acquisition data could point to the fact that they are a transitional phenomenon, for example due to prosodic and/ or discourse factors particular to child-directed speech. If, on the other hand, adults’ use of yes-no questions is found to exhibit more formal variability than is commonly assumed, these constructions need to be learned by children acquiring English and thus need to be integrated into accounts of language development. On a more fine-grained level, it is of interest to assess the internal makeup of the category yes-no questions, i.e., if a yes-no question is asked, which form is likely to be chosen? There are two competing views concerning the distribution of non-canonical vs. canonical yes-no questions in childdirected vs. adult-to-adult speech. On the one hand, it has been suggested that child-directed speech exhibits a higher percentage of ‘grammatically correct’ utterances than adult-to-adult discourse (Newport et al., 1977). If non-canonical yes-no questions are counted as less well-formed and less acceptable than canonical ones, we might therefore expect their percentage in child-directed speech to be lower than that found in adult-to-adult conversation. <?page no="101"?> 87 On the other hand, child-directed speech has also been described as more redundant, simpler, and formally less complex than adult-to-adult speech (Snow, 1972). For example, it has been noted that “a sizeable proportion of the child’s input contains utterances which are not complete examples of the language the child is acquiring” (Cross, 1977, p. 176). Thus, if non-canonical yes-no questions (particularly ‘reduced’ forms) are conceptualized as formally simpler than their ‘full’ counterparts, we would expect their percentage in child-directed speech to be on par with or even higher than in adultto-adult speech (Estigarribia, 2007, p. 114). In fact, some studies on childdirected speech have reported that, for example, omission of the auxiliary do is frequently found in motherese, whereas it is “non-existent in adultdirected questions” (Newport et al., 1977, p. 127). This chapter presents a study based on a corpus of adult-to-adult speech (BNC). In Sections 4.1 and 4.2, the most important information on the (sub)corpus used is summarized and the criteria for data extraction are discussed. Section 4.3 presents the results concerning the overall frequency of questions, while Section 4.4 deals with the coding and analysis of different types of questions found within the broader categories, with a special focus on non-canonical yes-no questions. Overall, the chapter provides an indepth analysis of (different types of) yes-no questions in adult-to-adult conversation and thus establishes the ‘baseline’ for comparison with the developmental data discussed from Chapter 5 onwards. 4.1 The Corpus The analysis of adult-to-adult speech is based on the British National Corpus (BNC). In total, the BNC contains nearly a hundred million words of spoken and written British English collected in the early 1990s. It was chosen because it is one of the biggest accessible corpora of present-day English and thus (presumably) provides a representative sample of the linguistic repertoire of native speakers of British English. Furthermore, the detailed classification of the different texts that make up the corpus allows for a careful selection of suitable transcripts for analysis. In addition to this, the webbased software BNCweb makes it possible to create and analyze subcorpora based on the BNC and also offers tools for the export of query results to spreadsheet applications such as MS Excel and SPSS (for a detailed guide to BNCweb, see Hoffmann et al., 2008). 73 73 Data cited herein have been extracted from the British National Corpus Online service, managed by Oxford University Computing Services on behalf of the BNC Consortium. All rights in the texts cited are reserved. <?page no="102"?> 88 4.2 Data Extraction For the current analysis, a subcorpus consisting of the face-to-face spontaneous conversations collected in the demographically sampled component of the BNC was created. The demographically sampled component contains unscripted spontaneous conversations recorded by informants who carried a portable tape-recorder and were asked to tape conversations with friends over a certain period of time (Hoffmann et al., 2008, p. 32). Since spontaneous conversation is “the most common type of language produced by most speakers of a language” (Hoffmann et al., 2008, p. 16), this subcorpus can be regarded as the adult-to-adult counterpart to the adult-child interactions focused on here. This subpart of the BNC was also used for the survey of questions in the corpus-based Longman Grammar of Spoken and Written English, in which the frequency of question marks per million words was used to assess the prevalence of questions, finding that “there is on average one question per every 40 words in conversation” (Biber et al., 1999, p. 211). However, since we do not have any information on the length of utterances in conversation, we do not know anything about the frequency of questions in terms of utterances. Therefore, I assessed the frequency of all utterances transcribed with a question mark as the utterance terminator. 74 According to the transcription conventions applied to the spoken component of the BNC (Crowdy, 1995), this yields all structures which are commonly grouped under the heading of interrogative, as defined, for example, by Langacker: An interrogative sentence in English is characterized by the occurrence of a question word (what, who, why, when, how, etc.) and/ or by the positioning of the subject nominal after the first auxiliary in the verb group. ... Intonation ... is sometimes the sole marker of interrogation (You’re finished? ). (Langacker, 1991, pp. 504-506) This subsumes wh-questions, alternative questions, tag-questions as well as the canonical and non-canonical yes-no questions described in Section 3.3. Importantly, rising intonation is used as a defining (i.e., necessary) feature 74 The definition of ‘utterance’ used here is different from the one applied to the coding of the spoken component of the BNC. The BNC defines an utterance (u-unit) as “a continuous stretch of speech produced by one participant in a conversation”, i.e. one speaker turn, regardless of length (Burnard, 2007, 4 Spoken texts, para. 2), while I follow Tomasello (and the CHAT-conventions) in defining an utterance as “a linguistic act in which one person expresses towards another, within a single intonation contour [emphasis added], a relatively coherent communicative intention in a communicative context” (Tomasello, 2003, p. 63). This latter definition corresponds more closely to the definition of sentence-units (s-units) in the BNC, which are seen as “stretches of discourse which the transcribers have identified as sentence-like in form (e.g., because they are bounded by pause) or function (e.g., a one-word utterance such as Yeah)” (Berglund Prytz, Hoffmann, Lee, and Smith, 2002). Therefore, the frequencies are based on s-units rather than u-units. <?page no="103"?> 89 only of non-canonical yes-no questions. Besides the fact that wh-questions are commonly associated with a falling intonation (e.g., Wells, 2006, p. 42), tag-questions, alternative questions and particularly canonical yes-no questions have been found not to exhibit rising intonation in all cases (contrary to what is often assumed in the literature; see, e.g., O'Connor & Arnold, 1973). For example, in an analysis of unscripted spoken material from the Survey of English Usage, Quirk et al. found that 290 of 720 yes-no questions terminated with a fall and “further analysis showed that yes-no questions with the modal operators can, could, may, might, and would tended to have almost as many falling tones as rising tones” (Quirk et al., 1985, p. 807). After more fine-grained analyses based on parts of the same corpus, Geluykens comes to the conclusion that “the claim that Rising intonation (and, more particularly, final Rises) is the ‘normal’ pattern for polar questions lacks empirical evidence” (Geluykens, 1988, p. 467). 75 Meanwhile, this claim has been corroborated by further research on intonation patterns found in spontaneous speech (e.g., Bartels, 1999; Hedberg & Sosa, 2002). 76 It is important to keep in mind that canonical and non-canonical yes-no questions do not belong to a uniform category, with all members sharing the formal (prosodic) feature ‘terminal rise’ or ‘auxiliary-subject word order’, for example. Rather, they are members of a family of related constructions; for example, the non-canonical question You want that? might share the feature ‘terminal rise’ with the canonical question Do you want that? , which in turn shares the feature ‘subject-auxiliary word order’ with the canonical question Can I have that? , which may be articulated with a terminal fall. This is also mirrored in descriptions of questions stating that “there are two external signals of a Q-marker: rising intonation and/ or [emphasis added] inversion” (Erreich, 1984, p. 584). It is important to note that the coding/ extraction procedure applied creates a(n artificial) cutoff-point, i.e., there are constructions which are formally as well as functionally related to yes-no questions which are not taken into account. These cases include 75 Furthermore, “intonation is not used to distinguish genuine Inversion-questions from interrogatives without Question-status, such as Rhetorical questions and Requests” (Geluykens, 1988, p. 467), i.e., there do not seem to be prosodic characteristics which differentiate distinct uses of the same question construction. 76 It has to be kept in mind that these findings are based on adult-to-adult speech, i.e., we do not know whether this is true of child-directed speech and children’s speech as well. There is some evidence suggesting that adults make use of exaggerated intonation when talking to small children (Garnica, 1977; Remick, 1976) and it has been noted that “perceptual properties such as prosody must also contribute to the acquisition of question versus declarative constructions” (Abbot-Smith and Tomasello, 2006, p. 282). Apart from this, children have been reported to strongly rely on rising intonation when producing questions (Bellugi, 1965, p. 111). If/ in how far question intonation in childdirected and children’s speech differs from adult-to-adult speech is an issue awaiting further empirical investigation. <?page no="104"?> 90 non-rising declaratives that function as questions, e.g., You’re feeling ill. (‘queclaratives’; Geluykens, 1987; Weber, 1993) embedded interrogatives, e.g., He doesn’t know what to say (also called ‘indirect questions’; Huddleston, 1994, pp. 414-415). non-questions exhibiting SAI, e.g., exclamatives like Is it hot in here! 77 The fact that there are no clear demarcation points (with excluded structures sharing formal and/ or functional characteristics with the yes-no questions considered here) is to be expected if we assume that our ‘constructicon’ is a complex network of constructions which are highly interlinked. 4.3 The Overall Frequency of Questions Table 4.1: Overall frequency of questions in the BNC-sample All utterances 610557 (100 %) Transcribed with question mark 105430 (17 %) With 17% the overall frequency in adult-to-adult spontaneous conversation is indeed somewhat lower than that found in most studies on child-directed speech (ranging from 21-44%). Nevertheless, it is far from the ‘near-absence’ postulated in earlier research, which did not test this claim empirically (Newport et al., 1977, p. 122; Olsen-Fulero & Conforti, 1983, p. 496). More interesting than the mere frequency with which questions occur in adult-toadult speech is the formal distribution within this category. It has already been found that “nearly half the questions in conversation consist of fragments or tags” (Biber et al., 1999, p. 211), but again Biber et al. do not offer detailed information. For example, even though they state that there are many cases where “ellipsis merges with non-clausal material” (Biber et al., 1999, p. 158) so that instances like <We’re> Too old to change aren’t we? cannot be categorized unambiguously, they do not give any information regarding their classification criteria. They also do not give any information on the internal structure of the category ‘declarative question’ which subsumes both structures with and without medial auxiliary (e.g., You want that? and You don’t want that? ). The section below thus analyses the formal distribution within the category ‘question’ in more detail. 77 Quirk et al. (1985) note that “in AmE an exclamatory question can be pronounced with a rising tone: Wasn’t the concert teRífic? But in this case, the expectation of a response is stronger” (p. 825). <?page no="105"?> 91 4.4 Coding and Analysis of Different Question Types For this more fine-grained coding of questions, a random selection of 5000 hits was exported to MS Excel and then coded by hand. While very timeconsuming, the manual coding of a rather inclusive dataset has the advantage of avoiding the systematic biases that often result from relying exclusively on automatic extraction and coding procedures. Automatic retrieval strategies are often influenced by a priori assumptions about linguistic structures, thus missing interesting cases. In order to avoid this, it is preferable to pursue a strategy that leads to the inclusion of all potentially relevant items (thus maximizing recall) even though this may make it necessary to discard irrelevant items manually (because precision has been reduced). For example, the assumption that question-tags occur exclusively in utterance final position may lead to a search for auxiliary-pronoun combinations immediately followed by a question mark. However, since tags may also occur in utterance-internal position (1), this retrieval strategy may result in a “skewed picture of actual tag question usage” (Smith, Hoffmann, & Rayson, 2007, p. 165). (1) KDY 173 You haven't pushed that in have you dear? 78 Biases resulting from automated retrieval strategies that are too restrictive can also be found in studies on question development: For example, in her study on inversion errors in yes-no questions, Stromswold excluded “all lines that contained a wh-word” (Stromswold, 1990, p. 146), thus eliminating all yes-no questions with embedded wh-questions and relative wh-words, e.g., Do you know what that is? . On a more general note, MacWhinney (2008) remarked that complex searches tend to be prone to error (MacWhinney, 2008, p. 169). The very inclusive retrieval strategy used in the current study maximizes recall and also ensures that the variation encountered in actual language use is taken into account. In order to keep the dataset as inclusive as possible, only very few items were discarded after careful consideration: First of all, a list containing known duplicate stretches in the BNC was consulted in order to ensure that there were no duplicates inflating the count for particular structures (Hoffmann & Evert, 2007). During coding one duplicate was spotted that was not contained in the list - the larger KBD transcript obviously contains (parts of) the shorter KSP transcript and one item was consequently discarded (examples (2) and (3)): (2) KSP 273 Mm. What, then you'd give him a Millwall shirt? I'd give it back. 78 In line with the requirements for citations from the BNC, the text identifier (a 3 letter code) as well as the sentence number is provided. <?page no="106"?> 92 (3) KBD 8988 Mm. What, then you'd give him a Millwall shirt? I'd give it back. Also excluded were utterances consisting solely of non-lexical conversational sounds such as Mmh? and Huh? containing quoted material, e.g., And I said to him what 's it going up ? (KE6 7882) consisting solely of social routines such as Sorry? , Excuse me? , or (Beg your) pardon? consisting entirely of direct self-repetitions or imitations 79 Lastly, utterances containing unintelligible material (marked by [unclear]) were excluded in case a relevant part of the target utterance was affected. For example, an utterance like (4) was discarded because the structure of the question could not be determined unambiguously. In contrast, example (5) was kept, because the turn presumably consisted of two clauses joined by the coordinating conjunction and, and it was still possible to identify the target utterance as a wh-question. (4) KP2 422 [unclear] be your partner on that day? (5) KP6 1607 [unclear] and where's she at school now? [unclear] In total, applying these criteria lead to the exclusion of 335 utterances. The remaining 4665 utterances were assigned to the four major categories: 1. whquestion, 2. tag-question, 3. alternative question and 4. yes-no question. 80 Figure 4.1 shows the overall distribution of question-types within the dataset. Alternative questions are very rare and account only for about 1% of all questions, while the other three question types occur with roughly the same frequency, each making up about a third of the dataset. 79 Since BNCweb restricts the context to one preceding and one following s-unit, only directly adjacent self-repetitions could be identified. There were no direct imitations in this sample. 80 This fourway categorization mirrors the one found in the major reference grammars (e.g., Biber, Johansson, Leech, Conrad, and Finegan, 1999, p. 204); much work on the makeup of child-directed speech treats alternative questions as a subtype of yes-no questions (e.g., Furrow, Nelson, and Benedict, 1979). <?page no="107"?> 93 Figure 4.1: The distribution of different question types in the BNC-sample Wh-questions, tag-questions and yes-no questions were further coded into different sub-categories. This fine-grained categorization makes it possible to compare the obtained percentages with results from other studies. 81 Additionally, it allows for the application of filtering options in order to conduct analyses on different subcategories, which is especially relevant for the class of yes-no questions focused on here. 4.4.1 Wh-questions 4.4.1.1 Full Wh-questions This category includes the structures considered as ‘full wh-questions’ in most studies on the composition of child-directed speech and children’s acquisition of wh-questions. However, even within this category we find different structural types: non-subject wh-questions (6) are often regarded as the most relevant subtype. Subject wh-questions (7) and structures of the type What/ How about X? (8), on the other hand, have sometimes been excluded because they are seen as ‘irrelevant’ for the development of whquestion syntax, since they both do not exhibit inversion and the latter contains neither a subject nor a verb (Valian & Casey, 2003, p. 121). (6) KCU 8246 What are you doing with a diploma? (7) KP9 107 Who was up here? 81 It is important to note, however, that in some cases it is not possible to achieve a oneto-one-correspondence of categories due to differences in the extraction procedure and the coding criteria applied. Furthermore, there are studies which do not make their criteria explicit so that the treatment of certain structures remains unclear. 32% 30% 1% 37% Question types (BNC) wh-question tag-question alt. question yes-no question <?page no="108"?> 94 (8) KCN 2161 What about her doll and things? 4.4.1.2 Auxiliaryless wh-questions Even though wh-questions are not a focus of this work, the number of whquestions with an omitted auxiliary (9) has been assessed in adult-to-adult speech, since the treatment of these structures in children’s productions bears a striking resemblance to the discussion of well-formedness in children’s production of yes-no questions. (9) KSS 1528 What you doing? Similar to the case of reduced yes-no questions, it has been pointed out repeatedly that “auxiliaries (but not pronouns) delete after a sentenceinitial interrogative wh element” (Zwicky & Pullum, 1983, p. 158) and that these structures are commonly found and thus are acceptable in spontaneous speech. Nevertheless, auxiliaryless wh-questions have mainly been treated as errors in child speech, without taking into account the frequency of such structures in the input (e.g., Bellugi, 1965; Rowland et al., 2005). This is not meant to imply that auxiliaryless wh-questions should be regarded as well-formed whenever they occur in children’s speech, it merely suggests that formal variation in the input should be considered in order to get an accurate picture of the adult target repertoire (for a proposal concerning restrictions on the deletion of auxiliaries in wh-questions, see Hendrick, 1982). 4.4.1.3 Echo-, clarification, and quiz-questions This category comprises structures commonly referred to as echo-questions (10) and quiz questions, in which the wh-word is not fronted (11). Clarification questions like (12) have also been subsumed under this heading. (10) KCX 8391 It's been what? (11) KSS 4337 But most of the time it's what part of Yorkshire? (12) KBK 1487 What bits? 4.4.1.4 One-word and ‘oblique’ wh-questions This category subsumes utterances consisting solely of a wh-word, a combination of how + adjective or adverb (e.g., How many? ; How old? ) or an ‘oblique’ wh-question in which the wh-word is governed by a preposition (13): <?page no="109"?> 95 (13) KSV 496 What for? Figure 4.2 shows the distribution within the class of wh-questions. Figure 4.2: The distribution of different types of wh-questions As can be seen, with 71% ‘full’ wh-questions constitute the vast majority of all wh-questions (n = 1481). Nevertheless, there are quite a number of echo- / quiz and clarification questions (5%) and in about 4% of all wh-questions, the auxiliary has been omitted. A fifth of all wh-questions consist only of the question word (and in some cases a preposition). 4.4.2 Tag-questions 4.4.2.1 Full tag-question This category subsumes the (presumably) prototypical type of tag-question, consisting of a full declarative clause followed by the auxiliary (or a form of do) and the pronominalized subject of the declarative (14). Also included are fragments followed by a tag (15). I did not differentiate types of polarity (matching vs. non-matching); therefore, (14) to (16) were subsumed under this heading: Analogous to the category full wh-questions, the label full tagquestion refers to the structures focused on in most studies on the acquisition and use of tag-questions (Dennis, Sugar, & Whitaker, 1982). (14) KBE 4469 They're waterproof aren't they? (15) KCN 3748 Coloured glass isn’t it? (16) KBK 4308 They're still in the same place are they? 71% 4% 5% 20% Wh-questions (BNC) full without aux. echo/ quiz one-word <?page no="110"?> 96 4.4.2.2 Short tags This category subsumes utterances consisting solely of an auxiliary-pronoun combination (17). (17) KDH 1681 Was it? 4.4.2.3 Invariant tags The invariant tags yeah, eh, okay, (al)right, no, really do not exhibit the internal structure of canonical tags. They have been categorized as invariant tags, regardless of whether they occurred in isolation (18) or at the end of a declarative utterance instead of a canonical tag (19) (invariant tags have also been called ‘fixed tags’, ‘informal tags’ and ‘response elicitors’; for more comprehensive discussions, see Columbus, 2010, and Stenström, 1984, Chapter 7). A particular feature of British English is the use of the invariant tag innit; originally a phonological reduction of isn’t it, it has become grammaticalized and can now be added to declaratives regardless of the pronoun and auxiliary involved, instances of which have also been found in this dataset (20) (for a more comprehensive discussion of innit, see Andersen, 2001, Chapter 4): (18) KC5 1729 Right? (19) KCV 3254 It's very hot no? (20) KB7 9795 No they're dissolvable plastic innit? Figure 4.3: The distribution of different types of tag-questions 62% 16% 22% Tag-questions (BNC) full short invariant <?page no="111"?> 97 Figure 4.3 shows the distribution within the class of tag-questions (n = 1413). Even though full tag-questions constitute the majority of all cases (62%), we still find a rather high number of short tags (16%) and a high percentage of invariant tags, suggesting that they play an important role in spontaneous conversation (Columbus, 2010). 4.4.3 Alternative questions Alternative questions contain the coordinating conjunction or and involve a choice between two or more alternatives (21). It has been suggested that yesno questions are ‘abbreviated’ versions of and thus constitute a special case of alternative questions, but this view has been largely abandoned (for a thorough discussion of this issue, see Bolinger, 1978); along similar lines, alternative questions are still sometimes subsumed under the heading yesno question because of the common formal properties (Estigarribia, 2007, p. 2). However, they are treated separately in some developmental studies and all reference grammars, mainly because of their semantics, which is the view adopted here. Since automatic extraction procedures are often applied, all utterances containing or are in most cases classified as alternative questions. If one goes through the data manually, it becomes obvious that not all structures containing or are in fact alternative questions. For example, I have grouped utterances containing the coordination tags or something? and or anything? under yes-no questions, since they function as a marker of vagueness rather than providing a real choice (22), (see Biber et al., 1999, pp. 116; 208). Along similar lines, I have classified utterances beginning with or as yes-no questions in cases where or is followed by a whole clause (in contrast to just a noun phrase, for example) and the utterance initiates a speaker’s turn - in these cases the or functions as a coordinating conjunction introducing a yes-no question, not as a connector between two or more choices (23). (21) KBU 86 You gonna take the car or get the bus? (22) KCT 7609 Yeah well he, didn’t go up the sides or something? (23) KPK 12 Or have we got an office in Hampton Court? I have made the simplifying assumption that utterances like (21) are indeed uttered and intended as alternative questions, even though it has been noted that or can also be used in yes-no questions; for example, (24) can be uttered either as an alternative question (presumably with a rise in intonation on Tuesday) or a yes-no question (with a final rise; Huddleston & Pullum, 2006, p. 869): (24) Are you free on Tuesday or Wednesday? Still, the whole group of alternative question only comprises about 1% of all questions found in the sample (n = 42). <?page no="112"?> 98 4.4.4 Yes-no questions This category has the most subcategories, which is not only due to the fact that it is the main concern of this study, it also exhibits the greatest formal heterogeneity. The coding scheme applied is based on the one developed by Estigarribia (2007, 2010), which in turn is a modified version of Bolinger (1957). The main difference to Estigarribia’s coding scheme is that I do not consider tag-questions and alternative questions to be a subtype of yes-no questions and that the coding is slightly more fine-grained. The coding of fragments is the one applied by Cameron-Faulkner et al. (Cameron-Faulkner et al., 2003, p. 849; the only exception are verb phrases, which were categorized as reduced questions). Even though there are a number of subcategories, the level of description applied here is relatively abstract and basically corresponds to what has been referred to as ‘utterance-level construction types’ (Cameron-Faulkner et al., 2003, p. 849); the developmental analyses in this book (Chapters 5-8) zoom in on lower-level constructional schemas (i.e., item-specific patterns in the sense of Cameron-Faulkner et al. 2003, p. 849). The (sub)categories coded for, along with examples, the respective numbers found in the BNC sample, and the labels used by Estigarribia (2007) are summarized in Table 4.2. In the vast majority of cases, the assignment of utterances in the dataset to the different categories was straightforward. There were, however, a few exceptions. The strategies used to deal with these cases are summarized first, before coming to the discussion of the results. First of all, utterances containing ‘possessive HAVE’ (25), which exhibits characteristics of both a (modal) auxiliary (since it inverts) and a full (transitive) verb (since it takes an object rather than a VP complement) (Trudgill, 1978, pp. 14-15) have been grouped with canonical questions containing MODAL/ HAVE/ BE. 82 (25) KDN 14 Have you no matches? Utterances containing a main verb which has an identical past participle and past tense form (26) have been coded as SPREDs (i.e., they have been considered to be reduced rather than assertive questions), since in the vast majority of cases it is clear from the context (and/ or the presence of question markers like any) that the past participle interpretation is intended. 82 For a more comprehensive discussion of the distribution of do you have, have you, and have you got, see Biber et al. (1999, pp. 213-214). <?page no="113"?> 99 Table 4.2: Coding scheme and results for yes-no questions in the BNC sample 83 Subcategories Example Results from BNC (n = 1729) Estigarribia (2007, pp. 33-35) Canonical 948 (55%) DO (untensed) Do you want some? 167 (10%) AUXF DO (tensed) Did you buy any? 139 (8%) MODAL/ HAVE/ BE Have you seen him? 368 (21%) copula BE Is he a writer? 274 (16%) Noncanonical 777 (45%) uninverted Strong assertive 156 (10%) w/ medial DO You do realize that? 15 (1%) SERTAUX w/ MODAL/ HAVE/ BE You can see him? 50 (3%) w/ BE (copula) You are a writer? 45 (3%) short uninverted You did? 15 (1%) without medial aux. You saw him? 31 (2%) SERT reduced 247 (14%) WSERT You want some? 95 (5%) WSERT SPRED They gone already? 56 (3%) SPRED APRED Don’t like this? 13 (1%) APRED PRED Want some? 83 (5%) PRED fragments 312 (18%) FRAG NP A man? 288 (17%) PP Under the table? 11 (1%) other one-word Now? 9 (0%) other multiword Seriously, though? 4 (0%) other subordinate Because of him? 66 (4%) errors No like it? n.a. ERRAUX 83 All percentages provided in this and the following tables (and figures) are rounded to the nearest whole number. <?page no="114"?> 100 (26) KBF 11445 You had any phone calls today? There are also other ambiguous cases: Consider example (27), which can be interpreted as either a prepositional phrase or a reduced question: (27) KCA 2234 like this? For disambiguation of these cases, the part of speech-tagging and/ or the discourse context provided by the BNC was appealed to. Noun phrases (including complex NPs) have been coded as fragments, even though they may be arguments of a missing main verb (Estigarribia, 2007, p. 79). On a more general note, I follow the common assumption that auxiliaries and subjects are either clearly present or absent in the respective utterances. However, there is phonetic variation in auxiliary-subject combinations and various intermediate forms are encountered (Estigarribia, 2007, p. 321). The problem of identifying these intermediate forms stems at least partly from the orthographic representation of spoken language. Transcriptions tend to use only standardized spellings, for example to ensure compatibility between corpora. Concerning the BNC, Heine (2011) has noted that it “does not make use of phonetic transcription and only marginally (and probably inconsistently) indicates assimilation phenomena in pronunciation” (Heine, 2011, p. 66). In my sample, there were only a total of seven cases in which such assimilation phenomena were made obvious in the transcription: In 5 cases, both the auxiliary do and the pronoun you were indicated to be phonetically reduced (28). In another two cases, only the auxiliary do was reduced (29). (28) KDF 308 D’ya want your slippers on? (29) KDE 3291 D’you wanna look at these? The same problem is encountered in the CHILDES-corpora: Even though the CHAT-format used for CHILDES offers the possibility to indicate pronunciations deviating from the standard form, it has been found that this option is underused (Peters & Menn, 1993, p. 753). 84 Since it is therefore likely that the transcription does not provide an adequate picture of the pronunciation variants which were used in the actual conversations, these cases were categorized along with the fuller (i.e., not phonetically reduced) forms. The application of these coding criteria led to the results summarized in Table 4.2. A simplified overview of the results is presented in Figure 4.4, 84 Along similar lines, one of the transcribers of the Thomas corpus has noted that the use of this option is probably rather inconsistent (B. Ambridge, personal communication, July 22, 2011). <?page no="115"?> 101 which shows the percentages of canonical vs. non-canonical yes-no questions as well as the internal distribution within non-canonical questions. Figure 4.4: The distribution of canonical and non-canonical yes-no questions in the BNC sample Only about half of all yes-no questions (55%) are ‘canonical’ in the sense that they contain an initial auxiliary-subject combination. Even though there is a lot of heterogeneity within the group of non-canonical questions, larger categories sharing some formal (and probably also functional) characteristics can be established. ‘Uninverted’ are all structures containing an auxiliary or an overtly tensed verb in medial position. They account for about 10% of yes-no questions overall. Secondly, ‘reduced’ questions are utterances lacking an auxiliary and/ or subject. With 14%, they constitute the secondbiggest group within non-canonical questions. Even more frequent are fragments (i.e., utterances without a main predicate; 18%). For the adult-toadult speech data, the category ‘other’ only contains subordinate clauses, which, even though they share the feature of being perceived as (structurally) incomplete, may contain a main predicate. With only about 4%, they are used rather infrequently. Even if one discarded ‘fragments’ and ‘other’ constructions, considering only utterances with a main predicate (i.e., ‘uninverted’ and ‘reduced’ questions) and canonical questions (n = 1351), non-canonical utterances would still account for about a third of all yes-no questions (30%). 55% 9% 14% 18% 4% 45% Distribution of different yes-no questions (BNC) canonical uninverted reduced fragments other <?page no="116"?> 102 4.5 Summary With about 17%, the overall frequency of questions in adult-to-adult discourse is somewhat lower than that found in studies on child-directed speech (21-44%). Within the category ‘question’, we find that apart from alternative questions, which are very rare (1% of all question) all question types are used with about equal frequency. Within the different question types, yes-no questions exhibit the greatest formal heterogeneity. Strikingly, canonical questions only make up about half of all yes-no questions, while the rest are ‘uninverted’, ‘reduced’ and fragmentary questions. These findings indicate that non-canonical yes-no questions form an integral part of the adult target system and are not to be dismissed as infrequent performance phenomena. Therefore, they should be considered as (potentially) wellformed structures when they occur in child speech and they should furthermore be integrated into accounts of question acquisition. Extending earlier work on non-canonical yes-no questions in child(-directed) speech which was based on low-density corpora (Dąbrowska, 2004, Chapter 9; Estigarribia, 2007, 2010), the next chapters provide an in-depth analysis of the acquisition and use of non-canonical questions in a high-density corpus of a boy learning English, addressing the roles of formal as well as functional variation in detail. <?page no="117"?> 103 5 Yes-No Questions in Child(-Directed) Speech In this chapter, I first introduce the data and methodology that is used as the basis for the developmental analyses presented in this book before discussing the first results. Since I focus on the high-density, longitudinal data of just one child, Section 5.1 addresses the role of case studies in developmental research. Section 5.2 summarizes the most important information on the corpus used, while Section 5.3 describes the extraction procedure and the refinement of the coding scheme made necessary by the special characteristics of child(-directed) speech. In 5.4, I summarize the findings on the overall frequency of questions and the distribution of different question types within this category. Finally, Section 5.5 focuses on yes-no questions, elaborating on the prevalence of non-canonical questions, child errors, and developmental patterns. This chapter thus provides the basis for the more finegrained analyses of particular types of non-canonical yes-no questions, which will be presented in Chapters 6 and 7, respectively. 5.1 The Importance of Case Studies One of the basic tenets of empirical research is that the findings obtained should be generalizable. In order to achieve this aim, the sample population needs to be (considered) representative of the population at large, usually requiring quite large sample sizes. This results from the assumption that while there may be some variance on an individual level, these differences will disappear once averages over large amounts of data points are calculated, usually yielding a normal distribution overall. Within developmental research, there are many studies using experimental designs or cross-sectional corpora in order to fulfill this criterion. While this strand of research is of utmost importance in evaluating competing hypothesis and for assessing overall developmental patterns, I would like to argue that certain issues can best be addressed by relying on (dense) data obtained from individual children. Within generative-transformational approaches, individual differences are largely marginalized, since Universal Grammar is thought to be innate and children’s language development is assumed to be rather uniform as a consequence. This is mirrored in the observation that language milestones are constant across individuals (Bellugi, 1965, p. 113). Concerning the outcome of the language learning process, Chomsky has noted “the striking uniformity of the resulting grammars” (Chomsky, 1965, p. 56), implying that individual differences in the linguistic competence of adults are negligible. <?page no="118"?> 104 In contrast to this, recent usage-based research on grammaticality judgment tasks in adults has revealed that there is much more individual variation than is commonly acknowledged (Dąbrowska, 2010, 2012; Street & Dąbrowska, 2010). This is not surprising, given the usage-based assumption that language structure emerges from language use, resulting in a conceptualization of ‘grammar’ that is based on each speaker’s individual experience with language rather than on the ideal speaker-listener proposed by Chomsky: Grammar is the cognitive organization of one’s experience with language. ... The result is a cognitive representation that can be called a grammar. This grammar, while it may be abstract, since all cognitive categories are, is strongly tied to the experience that a speaker has had with language. (Bybee, 2006, p. 711) Consequently, within usage-based approaches to language acquisition, the nativist continuity assumption is abandoned in favor of the developmental assumption, proposing “that whereas the processes working at different developmental stages are constant, the actual structures and representations are different at these different stages” (Tomasello, 2003, p. 324). Since children have not had much experience with language (yet), their grammars are expected (and have indeed been found) to exhibit even more variability and idiosyncrasies than the mental grammars of adults (Bannard, Lieven, & Tomasello, 2009). Therefore, it has repeatedly been pointed out that it is extremely important to investigate “changing representations within individual children (not just averages over different groups)” (Lieven & Behrens, 2012, p. 228). Along similar lines, it has been suggested that case studies based on large corpora are necessary in order to trace the gradual abstraction of item-based constructions (Dąbrowska, 2000, p. 87). Lastly, the exploration of the functional characteristics of constructions requires the analysis of naturalistic data, since only this kind of data offers insights into how the respective constructions are used in spontaneous conversational interactions. For these reasons, I have chosen to base my study on a large, highdensity, longitudinal corpus, whose main characteristics will be summarized below. 5.2 The Corpus The developmental part of this work is based on the Thomas corpus (Lieven et al., 2009), which is available online as part of the open-source Child Language Data Exchange System (CHILDES; MacWhinney, 2000). Like all corpora made available on CHILDES, the data has been transcribed following the CHAT conventions (MacWhinney, 2000), allowing for automated <?page no="119"?> 105 analyses using the CLAN program, which also includes options for the export of data to other applications such as MS Excel. It is a longitudinal naturalistic study of a boy who was born into an English middle class family and whose language development was recorded over a period of approximately three years. The name Thomas is the child’s real name - before consent was given to use this information, the child was referred to by his pseudonym Brian, which is used in research conducted on the data prior to its donation to CHILDES (Lieven & Behrens, 2012, p. 233). From 2; 0.12 to 3; 2.12, Thomas was recorded five times every week; from 3; 3.2 to 4; 11.20, recordings were made five times a week, one week in every month, resulting in a total number of 379 recordings which are usually about sixty minutes long. Audio data is available for all but two of these recording sessions, with a total duration of 408.3 hours. It is the densest CHILDES corpus for English available to date, with others using dense sampling regimes in preparation (Lieven & Behrens, 2012, pp. 233-234). It should be kept in mind that even during the densest period, only about 7-15% of the child’s linguistic experience was captured, even though this is still far more than the estimated 1-2% achieved in thinly sampled corpora (Lieven & Behrens, 2012, p. 226). 85 Dense sampling increases the probability that even infrequent structures are captured and also allows for a more realistic assessment of which structures are potentially rote-learned, since a higher percentage of the input is available for consideration. Furthermore, dense corpora offer the possibility to describe and analyze developmental processes on a very fine-grained level. In the present case, the dense sampling allows for a reliable quantitative assessment of the relative frequency with which particular canonical as well as non-canonical yes-no questions are found in the input and of how this relates to the child’s own speech. It is possible to evaluate the child’s reliance on item-based constructions and to trace the development of individual lowlevel constructions, re-evaluating and extending the findings obtained in previous studies which were based on much thinner sampling (particularly Dąbrowska, 2004, and Estigarribia, 2007, 2010; for comprehensive discussions of sampling issues, see Rowland et al., 2008, and Tomasello & Stahl, 2004). In order to exploit the richness of the whole dataset and to capture Thomas’ development of question constructions as comprehensively as possible, all transcripts were included in the analyses. It has to be kept in mind that the data collected from 3; 3.2 onwards is not as dense, i.e., both the amount of data points collected and the intervals between recordings are not directly comparable between the two sampling periods and overall developmental trends therefore must be interpreted with caution. The corpus contains 209.997 child utterances and over 368.098 utterances by adults, 329.772 of which are produced by the mother, who was the primary 85 So far, only the Human Speechome Project has made an attempt at ‘getting it all’ (Roy, 2009). <?page no="120"?> 106 caretaker. The rest of the utterances were produced by 11 other speakers, among them the child’s father (n = 657) and the investigators (n = 37.349). 86 All utterances by all speakers except Thomas were considered as the input (=INP) for subsequent analyses. While this conflation may mask interspeaker variation, it ensures the inclusion of all variation that Thomas was exposed to. In 137 (i.e., about 36% of all) sessions, two or more adults were present so it is possible that not all utterances considered as input were actually addressed to the child. Since capturing the child’s language development was the central aim of the recordings, it can be assumed that the adults are primarily interacting with the child. Furthermore, it has been shown that young children can learn words by overhearing adult-to-adult conversations (Akhtar, Jipson, & Callanan, 2001), so it is likely that Thomas also used utterances not addressed to him as a potential source for language learning. In order to gain a first impression of the linguistic complexity of child(directed) speech found in the corpus, the mean length of utterance (MLU; Brown, 1973), measured in morphemes, was assessed for both the input and the speech produced by Thomas, (Figure 5.1). 87 Although there is quite some session-to-session variability, the MLU of the input does not change overall. In contrast to this, Thomas’ MLU increases significantly over the course of the recording period (from around 1.25 at 2; 0 to about 4.0 at 4; 11). However, it is still lower than the MLU found in the input, suggesting that even at 4; 11, his utterances tend to be less complex than the ones addressed to him by his caretakers. Figure 5.2 shows the monthly MLU, calculated by averaging across all sessions in a given month. This way, the curve is smoothed, masking session-to-session variability, while still displaying the overall developmental trend. Since the same number of overall data points for developmental patterns is analyzed in the following sections, the same method (i.e., collapsing numbers from different recording-sessions in a given month) will be applied there. 88 86 The number of utterances was calculated by running the MLU command on all files, since this also gives information on the number of utterances. By default, the MLU program runs on the MOR-tier. Since file 2-03-19 does not include the MOR-tier, the number of utterances contained in this file was assessed by running the MLU program on the main speaker tiers. 87 Since it is not possible to collapse counts from different speakers within the MLUoption provided by CLAN, the average MLU of all adults was calculated separately for each session on the basis of the CLAN-output. 88 Note that this does not apply to the finer-grained analyses presented in Chapters 6 through 8, in which individual sessions and periods are considered. <?page no="121"?> 107 Figure 5.1: Mean Length of Utterance (MLU) in morphemes, calculated per session 89 Figure 5.2: Mean Length of Utterance (MLU) in morphemes, calculated per month 5.3 Data Extraction and Coding Analogous to the BNC-analysis, all utterances transcribed with a question mark as the utterance terminator were extracted from the corpus to assess 86 The data for file 2-03-19 is missing, because the transcript does not contain the MORtier. 1 3 5 7 9 2; 00.12 2; 01.08 2; 02.01 2; 02.27 2; 03.24 2; 04.17 2; 05.14 2; 06.16 2; 07.08 2; 08.04 2; 09.07 2; 10.01 2; 11.00 3; 00.01 3; 01.02 3; 02.03 3; 05.02 3; 08.05 4; 00.09 4; 04.04 4; 08.03 MLU in morphemes per session INP CHI Linear (INP) Linear (CHI) 1 2 3 4 5 6 7 8 2; 00 2; 02 2; 04 2; 06 2; 08 2; 10 3; 00 3; 02 3; 04 3; 06 3; 08 3; 10 4; 00 4; 02 4; 04 4; 06 4; 08 4; 10 MLU in morphemes per month CHI INP Linear (CHI) <?page no="122"?> 108 the overall frequency of questions; this was done using the KWAL-program provided by CLAN. According to the CHAT-transcription conventions, a question mark is used 1. for utterances that use “a wh-question-word, subject-verb inversion, or a tag question ending” and 2. for utterances “spoken with the rising intonation of a question” (MacWhinney, 2000, p. 57). Therefore, the extraction procedure yields all utterances belonging to the category question (as defined in Section 4.2). Table 5.1 shows the percentage of questions; the figures previously obtained from the BNC are included to allow for a direct comparison. Table 5.1: Overall frequency of questions in the Thomas corpus INP CHI BNC All utterances 368098 (100%) 209997 (100%) 610557 (100%) Transcribed with question mark 111635 (30%) 12466 (6%) 105430 (17%) The fact that almost a third of all utterances addressed to the child are questions is not surprising, given that previous studies on the composition of child-directed speech have found questions to account for 21 to 44% of the input (also see the more comprehensive discussion in Section 3.1). Questions in child-directed speech are thus more frequent than in the adult-to-adult speech sampled in the BNC, at least in terms of the overall frequency. Since the Thomas corpus contains longitudinal data, it is possible that the percentage of questions in the input decreases over time and, as Thomas gets older, slowly approaches the percentage found in the BNC - tentative analyses addressing this possibility will be presented below. 90 In contrast to the prevalence of questions in child-directed (and also adult-to-adult) speech, the child’s own speech only contains about 6% questions. Even though there are hardly any studies on the composition of children’s speech in terms of utterance types, it has been found that there are fewer questions asked by children than addressed to children, with questions accounting for 3.5-14% of children’s speech (Estigarribia, 2007, p. 138; MacCarthy, 1976; Smith, 1933, p. 201). Thomas’ data also falls within this range, indicating that children’s production does not necessarily mirror the input in terms of relative frequencies of abstract utterance types. 91 Therefore, it has to be kept in mind that while studies on question acquisition (includ- 90 The two corpora differ considerably in size, composition, and the coding system used. Because of this, all comparisons need to be interpreted with caution. 91 The fact that Thomas’ percentage of syntactic questions is particularly low was already noted by Dąbrowska and Lieven (2005, p. 144). <?page no="123"?> 109 ing the present one) take into account a high proportion of the input, they only account for a relatively small proportion of the child’s overall language production. However, previous usage-based research on question acquisition has shown that within the category of questions, there is a strong correspondence between particular item-based question constructions that children hear and the ones they produce (e.g., Dąbrowska & Lieven, 2005; Rowland et al., 2003; Steinkrauss, 2009); for non-canonical questions, this has only been shown on a rather abstract level so far (Estigarribia, 2007; 2010). The (formal as well as functional) relationship between particular noncanonical yes-no questions in the input and Thomas’ language is analyzed in more detail in Chapters 7 and 8. In contrast to the BNC-analysis, a dataset containing all extracted utterances was exported to MS Excel and then coded by hand. The basic coding procedure was the same as for the BNC-analysis, leading to the exclusion of unintelligible utterances, quotations, social routines (including counting, nursery rhymes and songs) and self-repetitions as well as imitations. Furthermore, utterances consisting solely of non-lexical conversational sounds (such as Mmh? ) were excluded. In contrast to the BNC, a larger context was available and self-repetitions as well as imitations were already coded as such on the main line of the transcript. Imitations were defined as “utterances where all components of the child’s speech occurred in one of another speaker’s preceding five utterances unless more than 10 seconds removed in time” and “self-repetitions were considered to be utterances where all components of the child’s speech occurred in one of his or her preceding five utterances unless more than 10 seconds removed in time” (MacWhinney, 2000b, p. 21). 92 Due to the specifics of the CHAT-format and in order to be suitable for the coding of child-directed speech and the child’s own productions, the following additional criteria were applied: 93 In utterances containing retracings involving the exact repetition of material, the material preceding the retracing was ignored (1). Material that was retraced with (2) or without correction (3) was also discarded (i.e., example (2) was coded as What do you not want to do? ): (1) (file 3-04-00) MOT: what [/ ] what are they doing? 94 92 I follow the general convention of excluding the child’s identical imitations of adult utterances. There are only 48 imitations of yes-no questions in the whole sample (mostly of fragments), so including them would not make a big difference anyway. 93 Since there were some cases where the application of the following criteria led to the exclusion or change of material within the respective utterance, a column with the original transcription was kept in order to be able to check the coded utterance against the original in all further analyses. <?page no="124"?> 110 (2) (file 2-03-10) MOT: what do you <want> [/ / ] not want to do? (3) (file 2-01-28) MOT: would you like that biscuit [/ / ] cheese biscuit? Whenever a reformulation was provided within one utterance (i.e., within one intonation contour), only the last part was kept (4): (4) (file 2-04-13) MOT: and what [/ / ] <what have they got on them> [/ / ] what have the pyjamas got on them? Along similar lines, false starts were removed (5): (5) (file 2-11-21) MOT: what [/ / ] where do you want to put your signs, Thomas? Like in all corpora involving early child speech, there are utterances which - even though they may not contain unintelligible material - are nevertheless unclassifiable. For example, Thomas uses a lot of schwas in his early language, as fillers in the place of prepositions and pronouns. Sometimes, it is not even possible to classify the respective utterances as (attempts at) either whor yes-no questions and these cases have consequently been excluded. Whenever the target form is clearly What’s this? , indicated by the transcription (6), the utterance has been grouped as a full wh-question (even though it probably functions as an unanalyzed holophrase for Thomas; however, this is also very likely to be the case for adult-like realizations of What’s this? ). In cases where the auxiliary is missing but the wh-word and the rest of the utterance are clearly articulated (7), it has been grouped as an auxiliary-less wh-question. (6) CHI: what’s this [= actually says wo’dis] (7) CHI: what 0are [*] you doing? Thomas also uses some non-target like expressions (=child-invented forms) when he wants to be given something: Wodar (which has also been transcribed as: a@sc there, &wo there, wada@c, a@sc dar@c). 95 It has been noted that “it is possible that wodar is used as different expression to @sc there or 94 Some of the codes used in the original transcripts (e.g., the hyphen used for the morphemization of complex word forms) have been removed from this and the following examples. 95 In CHAT transcription, @c is used to mark a child-invented form, @sc to mark schwas; the phonological form of an unintelligible string is preceded by the ampersand (MacWhinney, 2000). <?page no="125"?> 111 &wo there, but we do not yet have the contextual information to make any distinctions” (MacWhinney, 2000b, p. 32). These forms can also not be classified and have consequently been discarded. Lastly, there are structures which have been referred to as ‘ungrammatical parentese” (Estigarribia, 2007, p. 82), i.e., cases where the adult repeats the child’s ungrammatical utterance with question intonation. In total, the Thomas corpus contains 382 cases, with reactions to Thomas’ early attempts at expressing negation being the largest single group (example (8), n =78). 96 (8) (file 2-05-07) CHI: 0 [=! babble] &dis hot water. CHI: no like it. MOT: no like it? Since I went through many whole transcripts of the corpus during the course of my investigations, I discovered quite a few instances of utterances exhibiting the form of a canonical question which were not transcribed with a question mark (9): (9) (file 2-11-10) CHI: here you are, Dimitra. INV: thank you very much. INV: can I put it in my mouth. CHI: yes. Estigarribia (2007) did not include these cases in his dataset for two reasons: First of all, the speaker “may not have produced question prosody in that occasion, so the status of the utterance as an intended question could be put into doubt” (Estigarribia, 2007, p. 77). Second of all, “some non-canonical questions may not have been transcribed with a final question mark (due to transcriber error)” but there is no feasible way of recognizing those cases, so including canonical yes-no questions without a question mark could inflate the counts for canonical structures (Estigarribia, 2007, p. 77). While I do not find his first argument to be fully convincing (since it has been shown that by no means all canonical questions exhibit rising intonation), I regard his second point as justifying the decision not to include these cases in my dataset. 96 In their study on the development of negation in Thomas’s speech, Cameron-Faulkner et al. (2007) found that in the early stages, the “predominant multiword negator in [Thomas’] speech is no” (p. 261). Furthermore, they also identified examples like the one presented here, in which the mother repeats the child’s attempt at negation (Cameron-Faulkner, et al., 2007, pp. 258-259). As will be shown in Section 6.2, these repetitions are very often followed by paraphrases, thus providing Thomas with wellformed models for the expression of his communicative intention. <?page no="126"?> 112 For cases which are clearly wh-questions but which deviate so strongly from adult target patterns that they cannot be assigned to any of the existing categories, an additional category (wh-error) was created (10). (10) (file 2-02-28) CHI: what Nin+Nin there? MOT: it’s a piece of pineapple. For yes-no questions, there are very few classifiable errors, since ‘uninverted’ questions were not deemed incorrect. Along similar lines, the ‘omission’ of the auxiliary in cases like (11) was not seen as an error, contrary to what is indicated in the transcription. (11) (file 3-01-28) CHI: 0do [*] you want a train? The vast majority of yes-no questions which deviate from the adult target form contain errors (mostly of omission) which do not have anything to do with the formation of questions per se and which have consequently been coded as well-formed yes-no questions (12). (12) (file 3-08-05) CHI: can I hold 0the [*] water? Nevertheless, there are a few interesting cases of classifiable errors which have also been reported in the literature; they have been coded as ‘yes-no error’ and will be briefly discussed below (Section 5.5.1). Lastly, there were clear cases of erroneous transcription, which were corrected in the extracted dataset. First of all, there are (very few) cases in which an utterance was attributed to the wrong speaker (13). This was corrected after confirming the error by listening to the audio-tape. (13) (file 3-02-06) MOT: this is called “Small+Bear lost”. @Comment: Mother reads from the book. CHI: do you think you'll enjoy this? CHI: yes. Obvious typos were corrected (e.g., (14) and (15)): (14) (file 4-10-10) MOT: whose your girlfriend? who’s your girlfriend? (15) (file 2-07-26) MOT: is it tea of coffee, Thomas? is it tea or coffee, Thomas? <?page no="127"?> 113 The application of these criteria led to the exclusion of 7781 utterances. The remaining 128.786 utterances were assigned to the categories established in the preceding chapter with the addition of the categories ‘wh-error’ and ‘yes-no error’. To check reliability, a second coder who was blind to the hypotheses of the current study, coded a random subsample of 2000 utterances. A Cohen’s Kappa of 0.96 was obtained. 97 5.4 Overall Frequency and Distribution of Question Types Figures 5.3 and 5.4 show the overall distribution of question-types within the input and the child’s language, respectively. Like in the BNC-sample, alternative questions are very rare, in this case not even accounting for 1% of all questions. With 45%, yes-no questions constitute almost half of all questions, while tag-questions (31%) and wh-question are used with a slightly lower frequency (24%). The high percentage of tag-questions in the input is somewhat surprising, since most previous studies on child-directed speech have reported them to be rather rare (Broen, 1972; Estigarribia, 2007; but see Cross, 1977). In comparison with his caretakers, Thomas makes less use of tag-questions (7%) and yes-no questions, while wh-questions constitute the majority of all utterances in this category (60%). The prevalence of wh-questions (n = 5341) in Thomas’s speech compared to the input (n = 25487) is (at least partially) due to the percentage of oneword wh-questions he produces, which is much higher than the one found in the input (see Figures 5.5 and 5.6, respectively). Furthermore, Thomas’ use of wh-questions is highly stereotypical and since tokens rather than types were counted in the current analysis, formulas like What’s that? constitute a large proportion of the subsample. Lastly, 14% of Thomas wh-questions do not contain an (obligatory) auxiliary (e.g., *What happening now? ). Since auxiliary-less wh-questions are rarely found in the input (n = 90, displayed as 0% in Figure 5.5) and only involve cases of non-subject questions (e.g., ? What you doing? ), Thomas’ omission of auxiliaries in wh-questions cannot be the result of his taking over utterances from the input but must stem from non-adult like representations or performance limitations (see, e.g., Rowland et al., 2003). 97 In contrast to mere percentage agreement, Cohen’s Kappa corrects for the possibility of chance agreement (Cohen, 1960). <?page no="128"?> 114 Figure 5.3: The distribution of different question types in Thomas’s input Figure 5.4: The distribution of different question types in Thomas’s speech 24% 31% 0% 45% Question types (INP) wh-question tag-question alt. question yes-no question 60% 7% 0% 33% Question types (CHI) wh-question tag-question alt. question yes-no question <?page no="129"?> 115 Figure 5.5: The distribution of different types of wh-questions in Thomas’s input Figure 5.6: The distribution of different types of wh-questions in Thomas’s speech Lastly, Figures 5.7 and 5.8 show the distribution of different types of tagquestions (n = 33067 for the input, n = 591 for Thomas). The distributions are quite similar, apart from the fact that Thomas makes more use of short tags (such as Didn’t he? ). 91% 0% 4% 5% Wh-questions (INP) full without aux. echo/ quiz one-word 60% 14% 6% 17% 3% Wh-questions (CHI) full without aux. echo/ quiz one-word error <?page no="130"?> 116 Figure 5.7: The distribution of different kinds of tag-questions in Thomas’s input Figure 5.8: The distribution of different kinds of tag-questions in Thomas’s speech It has to be noted that all these numbers are based on all questions from the longitudinal set, thus obscuring developmental changes. They still give a first impression of the overall structure of the input and the child’s own productions and allow for a comparison with frequencies obtained in previous studies. Along similar lines, it is important to note that although rather abstract structural categories are used, this does not entail that I assume the child to operate with these abstract categories. Item-based structures are considered to be at the heart of (early) productivity and since similar 89% 11% 0% Tag-questions (INP) full short invariant 80% 18% 2% Tag-questions (CHI) full short invariant <?page no="131"?> 117 constructions will be found within the same broader structural class, this initial classification makes it possible to zoom in on lower-level patterns in subsequent analyses. This approach is similar to the one adopted in Cameron-Faulker et al.’s description of child-directed speech in which they presented the results “at two levels of analysis: (1) utterance level constructions, considered as abstract types; and (2) item-based utterance frames within these abstract construction types" (Cameron-Faulkner et al., 2003, p. 850). The next section provides a description of different types of yes-no questions on a rather abstract level (thus dealing with relatively abstract types), while Chapters 6-8 focus on item-based constructions within these categories. 5.5 Yes-No Questions Analogous to Table 4.2 in the preceding chapter, Table 5.2 provides an overview of the distribution of different types of yes-no questions in the Thomas corpus, this time without examples. Since the corpus is not directly comparable with the BNC, no statistics were performed but the figures obtained from the BNC sample are included to allow for a rough comparison of adult-to-adult and child-directed speech. Overall, the percentage of canonical questions is lowest in Thomas’ speech (45%) and highest in his input (63%), while the BNC occupies an intermediate position (55%). While this might lead to the conclusion that child-directed speech is more ‘grammatical’ than adult-to-adult speech, noncanonical questions still account for about a third of all yes-no questions addressed to Thomas. When discarding the categories FRAGs and ‘other’, non-canonical questions (n = 8928) still account for about a quarter (22.5%) of all yes-no questions addressed to Thomas (n = 30804). Again analogous to the presentation of the results obtained from the BNC sample, a simplified overview of the results is presented in Figures 5.9 and 5.10, which show the percentages of canonical vs. non-canonical yes-no questions as well as the internal distribution within non-canonical questions in the input and Thomas’ speech, respectively. <?page no="132"?> 118 Table 5.2: Distribution of different types of yes-no questions in the Thomas corpus Subcategories INP (n = 48849) CHI (n = 2979) BNC (n=1729) Estigarribia (2007, pp. 33-35) Canonical 30804 (63%) 1331 (45%) 948 (55%) DO (untensed) 5321 (11%) 152 (5%) 167 (10%) AUXF DO (tensed) 3109 (6%) 97 (3%) 139 (8%) MODAL/ HAVE/ BE 15378 (32%) 899 (30%) 368 (21%) copula BE 6996 (14%) 183 (6%) 274 (16%) Noncanonical 18045 (37%) 1648 (55%) 777 (45%) uninverted Strong assertive 4569 (9%) 91 (3%) 156 (10%) w/ DO 492 (1%) 8 (0%) 15 (1%) SERTAUX w/ MODAL/ HAVE/ BE 2228 (5%) 29 (1%) 50 (3%) w/ BE (copula) 983 (2%) 40 (1%) 45 (3%) short uninverted 133 (0%) 3 (0%) 15 (1%) without medial aux. 733 (1%) 11 (0%) 31 (2%) SERT reduced 4395 (9%) 657 (22%) 247 (14%) WSERT 1792 (4%) 324 (11%) 95 (5%) WSERT SPRED 876 (2%) 70 (2%) 56 (3%) SPRED APRED 35 (1%) 3 (0%) 13 (1%) APRED PRED 1692 (3%) 260 (9%) 83 (5%) PRED fragments 9081 (19%) 900 (30%) 312 (18%) FRAG NP 6939 (14%) 678 (23%) 288 (17%) PP 902 (2%) 79 (3%) 11 (1%) other one-word 509 (1%) 43 (1%) 9 (0%) other multiword 341 (1%) 62 (2%) 4 (0%) other subordinate 390 (1%) 28 (1%) 66 (4%) errors n.a. 10 (0%) n.a. ERRAUX <?page no="133"?> 119 Figure 5.9: The distribution of canonical and non-canonical yes-no questions in Thomas’ input Figure 5.10: The distribution of canonical and non-canonical yes-no questions in Thomas’ speech What is striking is that the caretakers make use of what have traditionally been termed ‘uninverted’ questions (i.e., ‘strong assertives’) with a higher relative frequency than Thomas, i.e., the rate of ‘uninversion’ found in childdirected speech is higher than that of the child; if one calculates the ratio of 63% 9% 9% 18% 1% 37% Distribution of different yes-no questions (INP) canonical uninverted reduced fragments other 45% 3% 22% 29% 1% 55% Distribution of different yes-no questions (CHI) canonical uninverted reduced fragments other <?page no="134"?> 120 uninverted questions vs. the sum of uninverted and inverted ones (as has been done in studies assessing the percentage of ‘inversion errors’; e.g., Erreich, 1984; Ingram & Tyack, 1979), the ‘uninversion rate’ for Thomas is only 6%, while the ‘uninversion rate’ for his caretakers is 13%. Since this can hardly be taken as evidence that Thomas’ caretakers are linguistically less advanced than Thomas, I would like to suggest that, in assessing children’s performance on yes-no questions, we need to re-evaluate how the target system actually looks like instead of relying on a priori assumptions about the well-formedness of particular structures. Examples (16) and (17) show that the mother produces an ‘uninverted’ yes-no question which, in very similar form is also used by Thomas. While the fact that similar utterances are found in the ambient language does not entail that Thomas’ use of this structure is contextually appropriate, it does suggest that these cases should not be regarded as grammatical errors. (16) (file 3-03-02) MOT: you can do funny tricks? (17) (file 3-05-02) CHI: you can do it? Interestingly, the fact that a priori assumptions exist in the evaluation of child speech is mirrored in the error coding found in the Thomas corpus. Examples (18) and (19) show a case in which the very same structure produced by the mother is marked as an error of omission when it is used by Thomas a few months later: (18) (file 2-09-08) MOT: you want some more? (19) (file 3-04-01) CHI: 0do [*] you want some more? Again, this is not to say that Thomas’ use of this structure is entirely adultlike. In order to evaluate this, it is necessary to address the discoursefunctional properties of the respective constructions, which will be done for weak assertives (18) in Chapter 6, and for strong assertives (16) in Chapter 7. First, ‘real’ errors, i.e., Thomas’ use of structures which are unlike any structures attested in the input, will be discussed briefly. 5.5.1 Thomas’ Yes-No Question Errors With only 10 tokens overall (constituting an ‘error rate’ of 0.8%), the number of classifiable errors is very low, which is in line with what has been found in many other studies on naturalistic samples (e.g., Estigarribia, 2007; Kuczaj, 1976). Since weak assertives (You want some? ) and strong assertives (You don’t want it? ) are not considered ungrammatical, they have not been <?page no="135"?> 121 classified as errors (of omission/ uninversion). Furthermore, as discussed in Section 5.3, some utterances deviated so strongly from potential target structures that they had to be discarded. Classifiable errors are discussed in the order in which they first occurred in the corpus. 98 First of all, there are two instances in which a (modal) auxiliary but no subject is produced: (20) (file 3-06-03) CHI: Mummy, can 0I [*] get some more? (21) (file 3-07-03) CHI: can 0you [*] see a big bang? Note that these cases could also be interpreted as strong assertives with subject-omission (<You> Can see a big bang? ). Next, there is one case where there is not only an auxiliary-verb mismatch but also a choice of the wrong auxiliary (22): (22) (file 3-08-04) CHI: Mummy, is [*] everybody eat [*] all their meals yet? In a similar case, only the wrong auxiliary is chosen: (23) (file 3-09-03) CHI: is [*] the crane got up? There are two errors in which the subject does not agree with the auxiliary chosen: (24) (file 4-00-07) CHI: is [*] there any pink quavers? (25) (file 4-03-06) CHI: are [*] your luggage on? There is only one auxiliary doubling error: (26) (file 4-01-05) CHI: would your little boy would * like a little treat? Lastly, there are three cases of auxiliary-verb mismatch errors, in which there is no (matching) tense marking on the verb: (27) (file 4-06-03) CHI: have you ate [*] your jelly? 98 There are no double tense errors (*Did you goed? ). <?page no="136"?> 122 (28) (file 4-07-10) CHI: have you ate [*] [= actually says et] it? (29) (file 4-10-06) CHI: have you finish0ed [*] yet? It is striking that - with the exception of subject-less errors and the auxiliaryverb mismatch errors in (22) and (29) - all of these errors could potentially result from the use of an unconstrained adjunction strategy which adds the wrong auxiliary to the left side of a strong assertive or a fragment (e.g., You ate your jelly? *Have you ate your jelly? ). However, no strong conclusions can be drawn from this data, since the low number of cases does not allow for the identification of potential underlying patterns. 5.5.2 Longitudinal changes in the relative frequency of noncanonical yes-no questions Up to now, the distribution of canonical vs. non-canonical yes-no questions in the Thomas corpus has only been assessed in an overall manner, i.e., without providing any temporal resolution. Despite the fact that developmental changes are expected for Thomas, it is also possible that the nature of his input changes as a reaction to his growing linguistic competence. Along similar lines, it has been suggested that early child-directed speech possesses specific prosodic characteristics, which might lead to a higher frequency of (particularly non-canonical) questions in the early vs. later sampling period. In order to address these issues, the frequency of yes-no questions was assessed for child(-directed speech (as mean percentage of yes-no questions to utterances, calculated per month). The resulting longitudinal curves are presented in Figures 5.11 and 5.12. Similar to the overall frequency of questions, yes-no questions are rather prevalent in child-directed speech, while they only account for a relatively small percentage for Thomas’ utterances overall (note that, for this reason, percentages are not equally scaled in the two figures). While there are only minor changes in percentage of yes-no questions provided in the input, Thomas starts using yes-no questions with a relative frequency of over 1% only at about 3; 0. Even though there is an overall increase of yes-no question use in Thomas, it never even comes close to the values found in his ambient language (Thomas’ highest percentage is 5.6% at 4; 08; the lowest value found in the input is 10.1% at 4; 04). <?page no="137"?> 123 Figure 5.11: Mean percentage of yes-no questions to utterances in Thomas’ input, calculated per month Figure 5.12: Mean percentage of yes-no questions to utterances in Thomas’s speech, calculated per month Child-directed speech (in English-speaking communities within a Western setting) has been found to exhibit exaggerated intonation, characterized by a higher median fundamental frequency than adult-to-adult speech (Remick, 1976). More specifically, an experimental study examining the fundamental frequency and frequency range of speech addressed to younger (mean age: 2; 3) and older children (mean age: 5; 4) found that of the sentences addressed 0.0 5.0 10.0 15.0 20.0 2; 00 2; 02 2; 04 2; 06 2; 08 2; 10 3; 00 3; 02 3; 04 3; 06 3; 08 3; 10 4; 00 4; 02 4; 04 4; 06 4; 08 4; 10 Frequency of yes-no questions (INP) INP 0.0 1.0 2.0 3.0 4.0 5.0 6.0 2; 00 2; 02 2; 04 2; 06 2; 08 2; 10 3; 00 3; 02 3; 04 3; 06 3; 08 3; 10 4; 00 4; 02 4; 04 4; 06 4; 08 4; 10 Frequency of yes-no questions (CHI) CHI <?page no="138"?> 124 to the younger children, 25% ended with a rising terminal, even though the sentences were declarative or imperative in form; this figure dropped to 9% in the speech addressed to the older children (Garnica, 1977, p. 76). Therefore, it is possible that non-canonical questions, particularly rising declaratives (and imperatives), are used predominantly in the early sampling period. Figures 5.13 and 5.14 show the frequency of non-canonical question in the input as well as Thomas’ speech (as mean percentage of non-canonical yes-no questions to all yes-no questions, calculated per month). For Figure 5.13, fragments were included as non-canonical questions while they were removed from the counts for Figure 5.14. Figure 5.13: Mean percentage of non-canonical yes-no questions in the Thomas corpus (including fragments), calculated per month As can be seen, the expectation that the frequency of non-canonical questions in the input decreases with age is not borne out by the data, regardless of whether or not fragments are included in the counts. This indicates that the presence of non-canonical questions in the input is not a transitory phenomenon, suggesting that they form an integral part of the adult target repertoire, thus corroborating the findings obtained from adult-to-adult speech (Chapter 4). As expected, the percentage of non-canonical questions in Thomas’ speech is rather high at the beginning and decreases as he gets older. From about 4; 0 onwards, Thomas’ percentage of non-canonical questions is about the same as that found in his ambient language. 0 20 40 60 80 100 2; 00 2; 02 2; 04 2; 06 2; 08 2; 10 3; 00 3; 02 3; 04 3; 06 3; 08 3; 10 4; 00 4; 02 4; 04 4; 06 4; 08 4; 10 Frequency of non-canonical yes-no questions (including fragments) CHI INP <?page no="139"?> 125 Figure 5.14: Mean percentage of non-canonical yes-no questions in the Thomas corpus (excluding fragments), calculated per month 99 With 37% (including fragments), the mean percentage of non-canonical yesno questions found in Thomas’ input is rather low compared to the values obtained by Estigarribia (which ranged between 43 and 67% Estigarribia, 2007, p. 116). He proposes that children whose input is predominantly canonical may follow different developmental paths than children who receive a high amount of non-canonical input. While the latter are assumed to gradually expand non-canonical questions by right-to-left elaboration (Type 1), the former are classified as ‘Type 2’, whose development “is based on an almost simultaneous emergence of all major classes of yes/ no questions, and proceeds (presumably) in a more holistic fashion” (Estigarribia, 2007, p. 328). There is a sudden rise in the overall percentage of yes-no questions in Thomas’s speech from 3; 05 to 3; 06 (from 1.6 to 3.6%), coinciding with a drop in the percentage of non-canonical questions (from 85 to 49%) thus suggesting that Thomas might be a ‘Type 2’. However, the low density of recordings from 3; 3.2 onward might mask more subtle developmental changes. Furthermore, Thomas does use both canonical as well as noncanonical questions before 3; 06, albeit with a lower overall frequency. Therefore, he might still make use of right-to-left elaboration; this possibility is addressed in the next chapter by analyzing the emergence and use of weak assertives (You want some? ) in more detail. 99 Apart from fragments, Thomas does not make use of any non-canonical questions in the period from 2; 0 to 2; 02, which is why it is not included in the figure. 0 20 40 60 80 100 2; 02 2; 04 2; 06 2; 08 2; 10 3; 00 3; 02 3; 04 3; 06 3; 08 3; 10 4; 00 4; 02 4; 04 4; 06 4; 08 4; 10 Frequency of non-canonical yes-no questions (excluding fragments) CHI INP <?page no="140"?> 126 5.6 Summary With about 30% and 6%, respectively, the overall percentage of questions found in Thomas’ input and his own speech falls within the range reported in the literature on child-directed and children’s use of questions, rendering questions a lot more prevalent in child-directed than child speech. Alternative questions are very infrequent (less than 1%); while whquestions are predominant in Thomas’ questions (60%), the caretakers primarily use yes-no questions (45%). Within the category of yes-no questions, the percentage of non-canonical questions in the input is lower than in the BNC but still accounts for about a third of all yes-no questions (37% as opposed to 45%), corroborating the claim that they are an integral part of the adult target repertoire. While non-canonical questions dominate Thomas’ use of yes-no questions overall (55%), he gradually approaches the levels found in his ambient speech. In contrast to this, the percentage of noncanonical questions in his input stays relatively stable over time, additionally indicating that their use is definitely not a transitory phenomenon. The next chapter zooms in on a particular type of non-canonical question, i.e., weak assertives (You want some? ), exploring their potential facilitative role in the learning of canonical questions and analyzing their discourse-functional properties. <?page no="141"?> 127 6 The Acquisition and Use of Weak Assertives This chapter addresses the emergence and use of a subset of yes-no questions (i.e., PREDs (1), weak assertives (2), and canonical questions with initial do (3)) in more detail, focusing on the level of weak assertives. The first aim of the chapter is to explore if and in how far Thomas’ learning of the three types of yes-no questions can be accounted for in terms of the gradual expansion of item-based constructions. Within Estigarribia’s right-to-left elaboration model, the developmental path is suggested to proceed as illustrated in examples (1) through (3) (Estigarribia, 2010, p. 175): (1) Want milk? (2) You want milk? (3) Do you want milk? Even though he provides this (hypothetical) and other (attested) examples, in which “the lexical identity of successive substrings [is conserved], this is not required by the [right-to-left elaboration] model” (Estigarribia, 2010, p. 85). His proposal that children might be operating with abstract categories from rather early on is an assumption made necessary by the low samplingdensity of the corpora used and he himself points out that “it is yet unclear at what level of granularity adjunction operates” and concludes that “the hypothesis of early item-based production is worthy of investigation with denser samples covering several months, given the protracted nature of YNQ [yes-no question] development” (Estigarribia, 2010, p. 88). Since there is strong evidence suggesting that a high degree of lexical specificity is involved in children’s learning of canonical questions (Dąbrowska, 2000, 2004; Dąbrowska & Lieven, 2005) as well as other constructions (Bannard, Lieven, & Tomasello, 2009; Lieven et al., 2009), it is highly likely that the same can be found for (early) non-canonical questions as well. As already discussed in some detail in Section 3.3, the fact that weak assertives may serve as the basis for more complex, canonical questions can be accounted for within a usage-based, constructivist framework by appealing to a refined version of the juxtaposition operation, which, like the related mechanism of superimposition, is likely to be constrained by semantic factors (Dąbrowska & Lieven, 2005). 100 Furthermore, since the concrete lexical material involved is assumed to play an important role, it is possible and indeed likely that children match possible outcomes of juxtaposition to (parts of) utterances attested in the input. 100 The operation of superimposition, in which one unit elaborates a schematic subpart of another unit, is constrained by “insisting on a semantic match between the items that create the slot and those that are inserted into it” (Dąbrowska and Lieven, 2005, p. 444). <?page no="142"?> 128 Besides lending overall support to the usage-based, constructivist approach, evidence for the gradual expansion of item-based constructions would also offer a (tentative) explanation for the fact that children only rarely choose the wrong auxiliary (Estigarribia, 2007; Kuczaj, 1976). It has been proposed that “children are probably tracking the semantics of each non-canonical question and learning which auxiliary goes with each intended meaning (even for zero-semantics DO)” (Estigarribia, 2010, p. 88), but to date there are no detailed studies exploring this possibility. The analyses addressing these issues are presented in Section 6.1. The second part of the chapter explores the potential facilitative role played by variation sets, or paraphrase relations, that the different types of yes-no questions might enter into. Focusing on a particular construction, i.e., weak assertives of the type You want (to) X? , the structural role of variation sets is analyzed in detail. The analysis of these constructions within their discourse context makes it possible to pursue a more general, mainly descriptive aim, which is to provide an account of the discourse functions fulfilled by (this particular type of) weak assertives. The results of this analysis is presented and discussed in Section 6.2. 6.1 Facilitation Paths in the Learning of Canonical Questions If Thomas’ production of increasingly complex yes-no questions relies on the gradual right-to-left expansion of (partially) lexically-specific material, we would expect to find many clear cases of the type exemplified in (1) to (3), with the respective PREDs being used first, followed by weak assertives and, eventually, canonical yes-no questions with initial do. This possibility was explored by first identifying the 5 most frequently occurring verbs (in the input as well as Thomas’ speech) in order to get an overview of distribution of different item-based constructions within each question type considered. Table 6.1 provides an overview of PREDs, weak assertives and canonical questions (starting with do) containing want, know, remember, think, and see, respectively. Apart from PREDs, which, by definition, do not contain a subject, only cases containing the second person pronoun you as subjects were considered, rendering the constructions even more item-specific (in fact, the pronoun you is the subject of 94% of all Doquestions and of 88% of all weak assertives). Each column presents the number of questions containing the respective (you+)verb combination found in Thomas’ speech, his input, and (for comparison), the BNC sample. The penultimate row (shaded in grey) indicates the percentage of utterances in the respective category which are based on frames involving the five most frequently used verbs. <?page no="143"?> 129 Table 6.1: Distribution of PREDs, weak assertives and canonical constructions 101 CHI INP BNC PRED weak assertive canonical PRED weak assertive canonical PRED weak assertive canonical want 36 (14%) 75 (23%) 70 (46%) 49 (3%) 690 (39%) 779 (15%) 10 (12 %) 11 (12%) 60 (36%) know 80 (31%) 152 (47%) 31 (20%) 2 (0%) 291 (16%) 871 (16%) 1 (1%) 26 (27%) 23 (14%) remember 3 (1%) - 3 (2%) 43 (3%) 13 (1%) 1050 (20%) 3 (4%) 1 (1%) 6 (4%) think 5 (2%) 3 (1%) 11 (7%) 33 (2%) 126 (7%) 1137 (21%) 1 (1%) 3 (3%) 15 (9%) see 14 (5%) 5 (2%) 8 (5%) 64 (4%) 51 (3%) 29 (1%) 10 (12 %) 5 (5%) 3 (2%) %age accounted for 138 (53%) 235 (73%) 123 (80%) 191 (11%) 1117 (62%) 3864 (73%) 25 (30%) 46 (48%) 107 (64%) total per category 260 (100%) 324 (100%) 152 (100%) 1692 (100%) 1792 (100%) 5320 (100%) 83 (100%) 95 (100%) 167 (100%) With the exception of the category PREDs in Thomas’ input and the BNC, which is very heterogeneous, the five most frequent verbs account for over half the attested instantiations of the respective item-based constructions. For example, 14% of all the PREDs used by Thomas are of the type Want (to) X? , 23% of all his weak assertives take the form You want (to) X? , and almost half (46%) of his canonical questions starting with do are instantiations of Do you want (to) X? . Overall, the degree of formulaicity (measured in terms of the percentage of utterances which is accounted for by the five most frequent item-based frames) is highest in Thomas’s speech, less pronounced in his input, and still (though only very slightly) lower in the BNC sample. Nevertheless, with 64% almost two thirds of all canonical questions with do in initial position found in the BNC sample involve only one of five verbs. This suggests that even though adult language use is more flexible, it is still centered around specific items or even multiword units (for more compre- 101 All percentages have been rounded to the nearest whole number. <?page no="144"?> 130 hensive discussions on the role of prefabricated units in adult speech, see Wray, 2002; Pawley & Syder, 1983). 102 In the next step, the order in which the different constructions emerged in Thomas speech was assessed by considering the first non-imitative uses attested in the dataset. The prediction that PREDs would be used before weak assertives, which in turn would emerge before the respective canonical questions, was indeed borne out. The first PRED to be used productively revolves around the verb see, with successive stages of complexity being attested in the predicted order (examples (4) to (6)): (4) (file 2-02-19) CHI: see the man gone? (5) (file 2-10-00) CHI: you see something? (6) (file 2-10-27) CHI: do you see my fire engine? A similar, and even slightly more protracted, development is attested for cases involving want (examples (7) to (9)): (7) (file 2-04-02) CHI: want a picnic? (8) (file 2-07-23) CHI: you want them, Mummy? (9) (file 3-00-24) CHI: do you want another photograph? For both think and know, the evidence is less clear (examples (10) to (12) and (13) to (15), respectively): think (10) (file 3-01-13) CHI: think (a)bout this one? (11) (file 3-02-09) CHI: you think? (12) (file 3-03-06) CHI: do you think he can cry and get his hat back? 102 This formulaicity basically stems from the Zipfian distribution found (not only) in language, i.e., the fact that the frequency of an item is inversely proportional to its frequency rank (this holds for single words as well as multi-word units; Bannard & Lieven, 2009). For a discussion of the influence of Zipfian distributions on language learning, see Ellis (2012). <?page no="145"?> 131 know (13) (file 3-02-06) CHI: know what I saw? (14) (file 3-02-04) CHI: you know what the train drivers did? (15) (file 3-02-11) CHI: do you know what this is? The reduced as well as the full forms appear within a rather short period of time, the different forms involving know even within only a week. This might be due to the fact that both are acquired rather late compared to see and want. At this stage, Thomas might already have developed more abstract and hence more flexible representations of the respective question constructions, enabling him to start using think and know in constructions of all three types almost simultaneously. However, since the constructions only emerge at the very end of the dense sampling period (after 3; 2.12, Thomas is only recorded for 5 consecutive days per month), it is difficult to assess the use and productivity of PREDs relative to weak assertives and full constructions with initial do. Lastly, consider the case of remember (16), which is rarely used by Thomas himself but which has been included because of its high frequency in the input and Thomas’ explicit use of an expansion strategy: (16) (file 2-11-18) CHI: remember? INV: pardon? CHI: yes you do. CHI: remember? [+ SR] INV: ah you asking me if I remember? CHI: do your remember? 103 In this case, Thomas produces the reduced question Remember? twice before making an attempt at the more complex, canonical form. This suggests that the two constructions are, at least in some contexts, formal competitors for the expression of the same discourse function. Further evidence for this suggestion, which will be elaborated on in the following section, is also provided by false starts found in the input (17): (17) (file 3-01-13) MOT: <you remember> [/ / ] do you remember going to see Monica and Paul in York? 103 In some of his questions (n = 24) between 2; 11.18 and 3; 2.2, Thomas realizes you as your. A possible (though at this point only very tentative) explanation might be that he has taken it over from his mother’s echo-questions beginning with the contracted subject-auxiliary combination you’re (e.g., You're undressing the farmer? , file 2-02-25). <?page no="146"?> 132 Overall, the observed pattern suggests that particular item-based PREDs and weak assertives emerge before the more complex, canonical questions, supporting the view that this elaboration operates on strings which are (partially) lexically specific. The more complex forms may be created by a refined version of the juxtaposition operation (Dąbrowska & Lieven, 2005, p. 442), which is constrained by 1. the semantics and 2. the concrete lexical material involved. For example, Thomas does not produce utterances like *Can you want X? , since there is a semantic mismatch between can and want in this particular construction. Furthermore, utterances of this type are not attested in the input, while there are many instances of the type Do you want (to) X? and Can you X? . Furthermore, positing that Thomas (initially) operates with item-based constructions (of the type (Do) you want X? , (Do) you remember(X)? and (Can) you remember(X)? ) learned as (partially specified) units from the input instead of more abstract constructional schemas like AUX SUB V? can also account for the fact that You remember? can be (and is indeed found to be) elaborated with both Do and Can, while only the first option is ever chosen for want. The order of emergence of the different variants also casts doubt on the formalization of weak assertives as ‘reduced’ or truncated versions of canonical questions with initial do (also see Estigarribia, 2010, p. 87). Rather, PREDs and weak assertives seem to become entrenched as constructions in their own right, potentially expressing the same (or at least a very similar) function as their canonical counterparts. A very similar observation has been made by Hopper (2001), who, in his study on elliptical and canonical pseudoclefts, concluded that “we should hesitate to privilege ‘complete’ constructions over ‘incomplete’ or ‘fragmentary’ ones, and should reject such notions as deviance and other terms implying a deficiency” (Hopper, 2001, p. 112). The overall frequencies and order of emergence in terms of the first nonimitative use by Thomas are in line with the predictions that the less complex forms emerge first and are expanded by adding elements to the beginning of the utterance. This gives a first impression of how less complex, noncanonical questions may be used as ‘stepping stones’ for the learning of particular, more complex, canonical questions. Still, first uses may be misleading since they give no indication of the frequency and productivity with which a structure is used from this point onwards, while overall frequency counts do not provide any temporal resolution and may therefore mask developmental patterns. Furthermore, the relationship between the different forms and the discourse function(s) they fulfill remain unspecified. The next sections address these issues by providing a fine-grained analysis of ((Do) you) want X? , focusing on the level of weak assertives. <?page no="147"?> 133 6.2 Variation Sets with Weak Assertives Even though want is only the second most frequent verb found in the subset (after know), constructions involving want still account for 39% of all weak assertives in the input and 23% of all weak assertives used by Thomas. Furthermore, since Thomas’ first uses of Want (to) X? , You want (to)X? , and Do you want (to)X? , respectively, occur with a gap of at least 4 months, and all first uses fall within the dense sampling period, it is possible to split the dataset without the data getting too sparse. The first period covers Thomas’s input up to 2; 4.2, at which point he uses Want X? for the first time. Period 2 covers all sessions in which Thomas uses Want X? but has not yet acquired You want (to) X? (and also makes no use of Want to X? ). Period 3 covers all sessions in which Thomas already uses both Want (to) X? and You want (to) X? . Finally, Period 4 covers all sessions from 3; 0.24 onwards, starting with the session in which he adds the most complex variant, i.e., Do you want (to) X? , to his repertoire. In the following analyses, constructions involving want + to-infinitival complements and those involving want + NP (i.e., ((Do) you) want to X? and ((Do) you) want X? ) have been treated separately. First of all, Thomas starts using Want X? before he starts using (You) want to X? . Overall, Want X? is used with a higher frequency and more productively, i.e., there is more variability in the NP-slot of Want X? than in the verb-slot of Want to X? , as measured in terms of the type-token frequency (TTR for Want X? = 0.87; TTR for Want to X? = 0.43). All this indicates that Thomas differentiates between Want X? and Want to X? . This ties in with evidence obtained from corpusbased priming studies on declarative structures with WANT-to vs. WANT-X in which the subjects (including Thomas) were suggested to operate with “at least two constructions for the main verb WANT” (Kirjavainen & Theakston, 2011, p. 648). In Thomas’ case, use of (You) want X? is attested relatively early in the corpus (2; 4.2; want a picnic? ), just like the corresponding (subjectless) declarative, which is first attested at 2; 0.27 (want more juice.). In contrast to this, Thomas’ starts using (You) want to X? only later, around the same time at which the corresponding declarative construction emerges ((You) want to X? at 2; 10.22; first use of Want to X in declaratives at 2; 8.13; (Kirjavainen & Theakston, 2011, p. 640). 104 104 Following Kirjavainen, Theakston, Lieven, and Tomasello (2009) and Kirjavainen and Theakston (2011), schwa realizations of infinitival to (e.g, you want ə hop on? ) were coded as instantiations of Want to X? . Instances where want to was realized as wanna were not included in the analysis (n = 8 for the input; n= 9 for Thomas), since it is not clear whether these cases should be regarded as instantiations of want + to-infinitival complement (Kirjavainen and Theakston, 2011, p. 639). <?page no="148"?> 134 Table 6.2: Distribution of PREDs, weak assertives and canonical constructions with want Period 1 <2; 04.02 Period 2 ≥2; 04.02- <2; 07.23 Period 3 ≥2; 07.23- <3: 00.24 Period 4 ≥3; 00.24 INP CHI INP CHI INP CHI INP CHI Want to X? 5 (3%) - 7 (4%) - 4 (1%) 4 (37%) 8 (2%) 7 (11%) You want to X? 58 (36%) - 114 (62%) - 148 (54%) 7 (63%) 90 (28%) 23 (37%) Do you want to X? 100 (74%) - 62 (34%) - 122 (45%) - 223 (69%) 33 (52%) Total 163 (100%) - 183 (100%) - 274 (100%) 11 (100%) 321 (100%) 63 (100%) Want X? 5 (6%) - 4 (4%) 5 (100%) 6 (4%) 2 (67%) 10 (4%) 17 (18%) You want X? 38 (47%) - 62 (64%) - 110 (65%) 4 (33%) 70 (30%) 42 (44%) Do you want X? 38 (47%) - 31 (32%) - 53 (31%) - 150 (65%) 37 (39%) Total 81 (100%) - 97 (100%) 5 (100%) 169 (100%) 6 (100%) 230 (100%) 96 (100%) As can be seen, the different constructions co-exist, both in the input and - as soon as they have been acquired in Thomas’ speech. Interestingly, weak assertives are used even more frequently than canonical questions in Period 2, i.e., shortly before Thomas himself starts using these constructions, while the canonical variants are dominant at the earliest and latest stages of the period covered (Periods 1 and 4, respectively). 105 This ties in with research suggesting that adults fine-tune their speech to the child’s developmental level by decreasing the complexity of particular structures in order to scaffold the learning process of things that have entered the child’s ‘Zone of Proximal Development’ (Vygotsky, 1978; see, e.g., Roy, 2009, who provides evidence for the fact that mother’s decrease the MLU containing particular words shortly before these words are learned by the child). The mere fact that all variants eventually co-exist in the speakers’ repertoire does not tell us anything about how these different constructions are 105 If one counted instances of You want to X? as ‘uninverted’ variants of Do you want X? , the inversion rate of the caretakers in the four different periods would range between only 35 - 71%. <?page no="149"?> 135 related to one another on the functional level, particularly if (and in which cases) they can be seen as formal competitors for the expression of the same (or at least a very similar) discourse function. In order to address this issue, it is necessary to go beyond mere frequency counts of decontextualized utterances by looking at how the different variants are used in conversation. Over four decades ago, Roger Brown suggested that “the changes produced in sentences as they move between persons in discourse may be the richest data for the discovery of grammar” (Brown, 1968, p. 288) and that “our very limited exploration of interaction patterns and the information they might convey persuades us that the empiricist position has possibilities that have not yet been explored” (Brown, 1968, p. 290). While this statement was made more than 45 years ago, there still seems to be a long road ahead of us, as evidenced by Lieven’s comment that „the analysis of discourse sequences to assess children’s representations could be more exploited“ (2014, p. 56). While there is some research (from a usage-based perspective) which takes into account the discourse function of constructions (Abbot- Smith & Behrens, 2006; Cameron-Faulkner et al., 2007; Cameron-Faulkner & Hickey, 2011; Steinkrauss, 2009, 2011), conversational exchanges are still often not considered in developmental studies. A rare exception is the work by Küntay and Slobin (1995, 1996, 2002) on the acquisition of Turkish, in which the authors explore the ways in which conversational exchanges may foster language development. They criticize the current treatment of corpus data, observing that “once the forms are picked out along with the propositional content surrounding them and made available for further statistical analysis, the real-time interactions that originally mediated these forms get entirely neglected” (Küntay & Slobin, 2002, p. 6). They analyze a special form of verbal interaction, termed variation set, which is defined as “a sequence of utterances with a constant intention but varying form” (Küntay & Slobin, 2002, p. 7). Since child-directed speech is often very redundant and caretakers tend to produce multiple utterances within one speaking turn, reformulations and (partial) repetitions are frequently found in caretakerchild interactions (e.g., Furrow et al., 1979; Hoff-Ginsberg, 1990; Olsen- Fulero & Conforti, 1983, pp. 510-511; Snow, 1972). By making relationships between similar structures salient for the child, these variation sets may have a facilitative role in language learning (and - on a more general note they might also provide insights into the psychological reality of links between constructions posited within Construction Grammar; Fischer, 2008). 106 There are three types of phenomena observed for variation sets in Turkish, “1) lexical substitution and rephrasing, (2) addition and deletion of specific referential terms, and (3) reordering of constituents” (Küntay & Slobin, 2002, p. 7). Of these, only the first is focused in the subsequent 106 Note that variation sets may also involve adult reformulations of children‘s (nontargetlike) utterances. For a study suggesting that these reformulations may help children recover from errors, see Chouinard & Clark (2003). <?page no="150"?> 136 analysis for two reasons. First of all, English, in contrast to Turkish, is a highly analytical language with much less possible variation in word order. Second of all, the mother’s use of rephrasing is the operation most relevant to the exploration of potential paraphrase relations between the different question forms. In the following analysis, the discourse context of all weak assertives of the type You want (to) X? is considered in order to explore the functional characteristics of weak assertives and the paraphrase relations between different variants (and related constructions). The following predictions are made: 1. Since weak assertives are formally ambiguous between ‘reduced’ and ‘declarative’ questions, they not only function as requests for information, but also as requests for clarification, particularly in the sense of echo uses, for which ‘declarative questions’ have been reported to be the unmarked option (Quirk et al., 1985, pp. 814; 835-837). This use will be found with a high frequency in the dataset, since early child speech is often erroneous and partially unintelligible. 2. Weak assertives enter into paraphrase relations. Particularly, weak assertives functioning as requests for information are offered as paraphrases of the ‘full’, canonical questions. This may provide the child with evidence that the different forms can be used interchangeably (in some contexts) while at the same time offering examples of how these forms are related structurally (i.e., the child is given models for possible expansions by juxtaposition). In contrast to this, echo-uses recast and rephrase/ expand the child’s (ungrammatical and/ or unintelligible) utterances, thus providing adult-like models for the expression of the child’s communicative intention. First, for all cases of You want to X? (n = 410) and You want X? (n = 280) in the dataset, based on a discourse context of 20 speaker turns (10 preceding, 10 following the target construction), the respective discourse function was a assessed and each utterance was assigned to one of the following two categories: 107 6.2.1 You want (to) X? as an information-seeking question All cases that were used as sincere requests for information were assigned to this category. Importantly, these questions - in contrast to echo-uses - can be used to initiate a conversational exchange, i.e., without recourse to an interlocutor’s prior utterance (Bishop et al., 2004, p. 421). The exchanges in (18) 107 In her analysis on form-function mappings of questions in child-directed speech, Shatz (1979) noticed that (Do (you)) want/ wanna X? can also function as a directive (p. 1097). However, there were no (clear) cases of this use in this dataset. <?page no="151"?> 137 and (19) contain examples of this use of You want X? and You want to X? , respectively. (18) (file 2-00-19) MOT: and when the postman comes three dogs from that house and one dog from this house all go woof+woof+woof@o, don’t they? CHI: oh. MOT: is that Dobbin? MOT: you want a ride on Dobbin? (19) (file 2-00-26) CHI: what’s this [= actually says wo'dis]. MOT: it’s a toggle. MOT: you want to throw it? 6.2.2 You want (to) X? as a request for clarification Requests for clarification are also commonly referred to as echo-questions or reprise questions, since they are used to elicit a clarification or repetition of what has just been said, repeating (part of) the preceding speaker turn (Quirk et al., 1985, pp. 814; 835-837). Since it has been suggested that “morphologically and syntactically, echo questions seem to have more in common with the sentence type they are used to echo than with the corresponding nonecho questions” (Radford, 1988, p. 464), the unmarked option for an echo-question as a reaction to a declarative utterance (20) is a rising declarative (21) rather than a canonical question (22): (20) I bought a car. (21) You bought a car? (22) ? Did you buy a car? Examples of echo-uses found in the dataset are presented in (23) and (24): (23) (file 2-06-25) INV: did you bump your foot? CHI: yeah. [+ SR] INV: ouch. INV: does it hurt? CHI: mmhm. [+ SR] INV: <shall we put all these> [>] . CHI: xxx [<] find something play. [+ PI] INV: pardon? CHI: find something play. INV: you want to find something to play? <?page no="152"?> 138 (24) (file 2-09-30) MOT: these are sultanas. MOT: the light ones. MOT: and the currants are the dark ones. MOT: currants are black. MOT: sultanas are brown . CHI: want a (sul)tana now. MOT: you want a (sul)tana now? Overall, 235 of all 410 You want to X? and 161 of the 280 You want X? questions were found to express a request for clarification, i.e., these uses accounted for 57% of all uses within both construction types, confirming the predictions about the dual function of weak assertives and the prevalence of echo-uses in child-directed speech. In a next step, all cases in which the respective utterance entered into a paraphrase relation were identified. Table 6.3 provides an overview of the functional distribution within You want (to) X? and the number of variation sets found. Table 6.3: Distribution of discourse functions expressed by You want (to) X? (INP) You want to X? You want X? Function Variation sets Function Variation sets Request for Information 175 (43%) 13 119 (43%) 9 Request for clarification 235 (57%) 22 161 (57%) 16 Total 410 (100%) 280 (100%) Overall, about 9% of all utterances are part of a variation set. In the presentation of the results, You want to X? and You want X? will not be discussed separately, since variation sets with both constructions are found with about equal proportions. Variation sets with utterances used as requests for information will be addressed first. 6.2.3 Variation sets with You want (to) X? as a request for information As expected, we do find cases (n = 8) in which the caretaker expands a weak assertive, resulting in a canonical question. <?page no="153"?> 139 (25) (file 02-01-17) CHI: ah the bo@c. [+ SR] INV: you want to have the gop back? INV: you want to have gop back? INV: hmm? INV: do you want gop to come back? These paraphrase relations may help Thomas learn the more complex forms, since utterances of varying degrees of complexity are embedded within one discourse situation. However, as has already been noted by Estigarribia (2007, p. 324), it is not necessarily the case that weak assertives are expanded (in order to make the question explicit). There are also cases (n = 11) in which the canonical form is followed by the ‘reduced’ form (26). (26) (file 2-03-10) MOT: well you keep picking up the boxes and squealing and screaming and being silly, aren’t you? MOT: what’s the matter? MOT: do you want a cuddle? MOT: are you just feeling a bit out of sorts? MOT: are you feeling out of sorts? MOT: you want a cuddle with Mummy? MOT: Thomas? MOT: do you want to go to bed? MOT: Thomas do you want to go to bed? In this example, we can see a case of both reduction and expansion, since the initial Do is omitted while a prepositional phrase (with Mummy) is added, providing information on additional ways in which utterances may be expanded. Lastly, we also find variation sets (n = 3) in which the caretaker does not alternate between different stages of complexity with respect to question forms, but makes use of formally distinct constructions which fulfill a similar discourse function. (27) (file 2-06-02) MOT: you want the bridge? MOT: would you like the bridge bringing over? CHI: yeah. (28) (file 2-03-06) MOT: there. MOT: would you like to get in there? CHI: yes [/ / ] no. MOT: you want to get in there? <?page no="154"?> 140 MOT: have a big splash? MOT: you're going to sit in it? Overall, the patterns found in the data suggest that the paraphrase relations offered in variation sets are a rich source of information concerning the relationships between (formally and/ or functionally) related constructions. 6.2.4 Variation sets with You want (to) X? as a request for clarification As expected, requests for clarification do not enter into the same paraphrase relations as requests for information, i.e., there are no cases in which a request for clarification is expanded to a canonical question or results from a reduction of a canonical question. Instead, we find reformulations and expansions of Thomas’ (declarative) utterances (n = 38). Especially in the early recordings, these are often preceded by a repetition of Thomas’ (ungrammatical and/ or partially unintelligible) productions (29) (the fact that mothers tend to (re)produce fragments has also been noted by Snow, 1972, p. 555). (29) (file 2-01-11) MOT: and since you've been home you've been drawing on [/ ] on your hands with your crayons, haven’t you? CHI: no. CHI: Nin chair. CHI: Nin+Nin chair. MOT: Nin+Nin chair? MOT: you want me to sit down again? CHI: xxx. MOT: well before I sit down let me just wash your hands. Later on, as Thomas’ language skills become more sophisticated, the echouses take the form of expansions of Thomas’ utterances rather than mere repetitions (with a change in intonation) followed by paraphrases (30). (30) (file 2-06-12) MOT: are you tired now? CHI: no. CHI: talk Mummy a moment. MOT: you want to talk to Mummy a moment? Another particularly striking example (31) is a case in which the mother summarizes all events mentioned by Thomas in the final constituent of the variation set, proposing a linear temporal ordering and linking the events by <?page no="155"?> 141 means of the conjunction and (then) (a similar strategy found in Turkish child-directed speech is discussed by Küntay & Slobin, 2002, pp. 8-9). (31) (file 2-04-17) CHI: a window. MOT: that’s [/ ] that’s not a window, Thomas. MOT: that’s the mirror. CHI: mirror. [+ I] CHI: mirror. [+ SR] CHI: juice a@sc bed. MOT: you want to look in the mirror and then have some juice and go to bed? In most of these examples, the paraphrase relations do not hold between different adjacent utterances by the caretaker. Instead, there is a paraphrase relation between an utterance by Thomas and the caretaker’s reply, which provides the child with adult-like models for the expression of his current communicative intention. 6.2.5 Thomas’ use of You want (to) X? 108 It has already been shown that Thomas’ own productions mirror the input on a formal level, since he picks up both You want to X? and You want X? (albeit with different relative frequencies than found in the input). The last analyses presented in this chapter focus on the functional properties of You want (to) X? in Thomas’ speech, and the ways in which he makes use of variation sets himself. Table 6.4: Distribution of discourse functions expressed by You want (to) X? (INP; CHI) You want to X? You want X? INP CHI INP CHI Request for Information 175 (43%) 30 (100%) 119 (43%) 46 (100%) Request for clarification 235 (57%) - 161 (57%) - Total 410 (100%) 30 (100%) 280 (100%) 46 (100%) 108 The (potentially simplified) assumption has been made that You want (to) X? as a question and a request for clarification, respectively, are formally indistinguishable so that the child must rely solely on the context for a correct interpretation. It is, of course, possible that there is a difference in prosody which might aid the disambiguation process. <?page no="156"?> 142 There are no uses of You want (to) X? in Thomas’ speech functioning as requests for clarification. This is not unexpected, since the speech provided by caretakers is usually clearly enunciated and redundant. Other pragmatic functions of echo-uses (such as expressing surprise or disbelief) are probably only mastered later on and/ or associated with other constructions. Since Thomas’ utterances are usually reacted to right away with a (verbal or nonverbal) response, variation sets in the narrow sense are not expected to occur often in Thomas’ speech and no clear examples could be found. There are, however, some interesting cases involving structural variation. First of all, consider examples (32) and (33). (32) (file 3-01-28, line 16) CHI: want some newspapers? (33) (file 3-01-28, line 187) CHI: your want a newspaper? In this case, providing evidence for the potential functional equivalence of PREDs and weak assertives (and tentatively also for his operating with expansion strategies), the utterances are further removed in time, but still occur within the same discourse situation. This suggests that the consideration of larger discourse units may be worthy of further investigation. Like his caretakers, Thomas also alternates between weak assertives and their canonical counterparts (34): (34) (file 3-10-01) CHI: do you want to throw something in the rubbish? MOT: no I don’t think we want anything in the rubbish at the moment, do we? CHI: Mummy, you want a tissue throwing in the rubbish? MOT: no thank you. Lastly, Thomas also makes use of other expansion strategies, elaborating a weak assertive by adding a direct object (35): (35) (file 3-00-07) CHI: hello, Daddy. CHI: hello xxx a@sc dustbin+men’s [*] come-ing. [+ PI] CHI: you want to@sc see? CHI: you want to@sc see a@sc dustbin+men? FAT: yes. FAT: I've come to see them too. While these analyses offer a quite comprehensive picture, a few issues remain unexplored: <?page no="157"?> 143 Due to the low number of Want (to) X? -PREDs in the input (n = 49), it is not surprising that there is only one corresponding variation set (36), in which the PRED is expanded to a canonical question. Therefore, the possible role played by PREDs in terms of structural expansion strategies as well as their functional characteristics are still largely unclear. An analysis of cases like (Do (you)) see X? , for which PREDs are more frequent than the more complex variants, may provide further insights in this respect. (36) (file 2-03-26) MOT: do you not want me to cut them? MOT: I thought you liked watching your nails being cut. MOT: want me to do that one? MOT: shall we do fingernails? MOT: or toenails? MOT: do you not want me to do that? Along similar lines, the variation sets which the other weak assertives enter into have not been analyzed. Since cases similar to the ones found for (Do)(you))want (to) X? can easily be detected even upon a superficial inspection of the dataset (37), it seems likely that these cases would provide further evidence for the facilitative role of structural variation in the input. (37) (file 2-05-06) INV: mmm this is very nice. INV: and you know what we’re going to (.) be doing next? INV: do you know what we'll do next? INV: we’re going to be watching a video. Lastly, looking at conversational exchanges between caretaker and child may shed further light on the relationship between formally related constructions such as non-canonical questions and declaratives. In (38), the child imitates the mother’s utterance, but changes the intonational contour, so that his mother’s question serves as the basis for the expression of Thomas’ communicative intention in declarative form (note that the subject is missing in both the mother’s and Thomas’ utterance). 109 (38) (file 2-01-21) MOT: this is a bright green card and it’s got a heart and a flower and a star. CHI: blue star. MOT: a blue star. MOT: it’s a very bright card, isn’t it? 109 This may also indicate that subject omission in PREDs is related to the omission of subjects in declaratives. <?page no="158"?> 144 CHI: uhm. MOT: want some more? CHI: want some more. [+ I] 6.3 Summary This chapter has provided an in-depth analysis of the emergence and use of a subset of yes-no questions (i.e., PREDs, weak assertives, and canonical questions with initial do), with a particular focus on the role played by weak assertives. It has been found that the three types of yes-no questions emerge in the order predicted by Estigarribia’s right-to-left elaboration model (Estigarribia, 2010, p. 175) and that this pattern can be accounted for in terms of the gradual expansion of item-based constructions by a modified version of the juxtaposition operation proposed by Dąbrowska and Lieven (Dąbrowska & Lieven, 2005). Additionally, an investigation of different types of paraphrase relations, particularly those found in variation sets (Küntay & Slobin, 2002), revealed that the weak assertive You want (to) X? enters into formal competition with (PREDs like Want X? and) the full, canonical counterpart Do you want (to) X? . Furthermore, the analysis of these constructions within their discourse context showed that they can either function as requests for information or as echo-questions, which corroborates the claim that their structural (and hence functional) status is ambiguous between ‘reduced’ question and rising declarative (Fodor & Crain, 1987, p. 40). <?page no="159"?> 145 7 The Acquisition and Use of Strong Assertives This chapter considers the acquisition and use of another type of noncanonical question, namely strong assertives. In particular, strong assertives with a medial auxiliary (e.g., You can see this? ) will be focused on. Estigarribia (2007, 2010) does not address them in detail because they cannot enter into right-to-left elaboration due to their ‘uninverted’ word order, i.e., (1) cannot be a result of adding an initial element to (2): 110 (1) Can you see this? (2) You can see this? In most developmental studies, they are treated as ‘uninversion errors’, whose prevalence is assessed by calculating children’s ‘(un)inversion rates’ (e.g., Erreich 1984; Ingram & Tyack, 1979; Nakayama, 1986). To my knowledge, the only study comparing the child’s ‘inversion’ rate with that found in the input on a rather fine-grained level is offered by Dąbrowska (2004, Chapter 9). Interestingly, the rate of ‘inversion’ in the input hardly ever reaches 100% and it is particularly low for questions with do/ does (62% note that weak assertives were counted as ‘uninverted’; Dąbrowska, 2004, pp. 187-188). Her overview of ‘uninverted’ questions found in the data shows that the vast majority of ‘uninverted’ questions are indeed weak assertives (Dąbrowska, 2004, p. 192), which - as has been shown - are not to be considered ‘uninverted’ (nor reduced) versions of their canonical counterparts, while still sharing some of their discourse-functional characteristics. The focus of her study is on the mastery of canonical questions, thus pursuing a different aim than the analyses presented here. Still, her study provides a good starting point for the investigation of strong assertives (with medial auxiliary), since her approach takes into account the structural variation found in the input instead of just evaluating the children’s performance against an inversion rate of 100% presumably found for all syntactic questions in the ambient language. The first analysis, presented in Section 7.1, focuses on the distribution of ‘uninverted’ vs. ‘inverted’ auxiliary-subject combinations, considering only strong assertives with medial auxiliaries as ‘uninverted’. The aim is to apply Dąbrowska’s method to a high-density sample, thus allowing for a fine-grained analysis of particular auxiliary-subject combinations in different periods of Thomas’ speech. This way, it is possible to assess if (and in how 110 They are only considered as potential sources for auxiliary doubling errors, which might result from the erroneous assumption that they can also be elaborated by adding an initial auxiliary (You’re a cactus? *Are you’re a cactus? (Estigarribia, 2007, p. 250). <?page no="160"?> 146 far) the distribution of canonical questions vs. strong assertives (with medial auxiliary) in Thomas’ speech mirrors the one found in the input. Proceeding from the assumption that strong assertives are not (necessarily) to be regarded as ‘uninversion errors’ (since they are found in the ambient language), Section 7.2 addresses the use of these constructions by both the caretakers and Thomas, focusing on the two structures That’s X? and It’s X? . First of all, this is done to address the question of whether (and in how far) Thomas’ use of these structures is related to the input not merely on the formal but also on the functional level. Secondly, it allows for an empirical evaluation of claims found in the literature regarding the discourse functions of strong assertives. 7.1 Inversion Rates Even though the Thomas corpus has a high density (for the period from 2; 0.12 to 3; 2.12), it is not possible to obtain ‘inversion’ rates for every attested auxiliary + subject combination at different stages of development, since the numbers for less frequently used auxiliaries are too low. Still, the larger sample allows for a fairly fine-grained analysis of the most frequently used combinations. For this, the Thomas corpus has been split in three parts: Thomas1 contains all transcripts up to (and including) 3; 05; Thomas2 those up to 4; 04, and Thomas3 the transcripts covering the remaining period from 4; 05 to 4; 11. For the input, all adult utterances from the transcripts covering the period from 2; 0.12 to 2; 05.29 (i.e., the first 119 transcripts) have been considered, since Thomas starts using his first auxiliary + subject combination productively at about 2; 06 (also recall that the overall distribution of canonical vs. non-canonical forms in the input does not change over time, rendering this sample representative of the overall input Thomas receives). All specific auxiliary-subject combinations that are produced by Thomas at least eight times (inverted or uninverted) in at least two out of the three periods were treated separately. In all other cases, figures have been collapsed across syntactic subjects (labeled e.g., can X); in cases where this did not result in a category fulfilling the frequency criterion of at least 8 instances in at least 2 periods, figures were further collapsed across different forms of an auxiliary (e.g., HAVE X) or even across different auxiliaries (in the case of could/ may/ might/ must/ should X). I follow Dąbrowska (2004) in only providing percentages in cases where there are at least eight instances (inverted or uninverted) of the respective auxiliary + subject combination. Questions containing the auxiliary DO have not been included because I do not consider weak assertives (You want some? ) to be ‘uninverted’ questions, and the number of strong assertives (You don’t want some? ) containing a medial auxiliary (for emphasis or in negated form) is very low (n = 492 for the input, and 8 for Thomas; in contrast to this, strong assertives with medial <?page no="161"?> 147 auxiliaries other than DO are much more frequent: n = 2211 for the input and 69 for Thomas). The results are summarized in Table 7.1. Table 7.1: Inverted yes-no questions according to auxiliary + subject Auxiliary Thomas1 2; 00 - 3; 05 Thomas2 3; 06 - 4; 04 Thomas3 4; 05 - 4; 11 Thomas’ input (2; 0 - 2; 05) N % N % N % N % can I 32 100 143 100 105 100 49 98 can you 13 62 53 100 51 100 739 99 can we 8 100 11 100 13 100 15 73 can X 1 - - - 5 - 35 100 can’t X 4 - 1 - 6 could/ may/ might must/ should X 11 100 7 - 5 - 71 100 shall I 3 - 53 100 65 100 122 99 shall X 6 - 16 100 7 - 528 100 will X 3 - 16 94 28 93 65 98 would you 2 - 31 100 47 98 825 95 would X - - 1 - 5 - 30 100 are you (aux) 13 77 13 100 16 100 2024 93 BE X (aux) 6 - 7 - 16 88 684 92 have you 5 - 42 95 37 100 563 91 HAVE X 4 - 9 89 7 - 415 93 are you (cop) 8 88 12 100 18 100 239 91 is it (cop) 14 40 16 81 20 90 362 92 is that (cop) 13 46 12 83 13 92 778 98 BE X (cop) 23 62 37 92 33 94 706 92 Overall, the ‘inversion rates’ (provided in the % column) found in both the input and Thomas’ speech are rather high. Still, the ‘inversion rates’ in the ambient language rarely reach 100%, casting doubt on the assumption that ‘uninverted’ questions are not part of the target repertoire. The only auxiliary + subject combinations for which clear developmental patterns can be found are the ones involving copula BE (in particular, Is it X? , and Is that X? ), with Thomas’ ‘inversion rates’ increasing over time. <?page no="162"?> 148 Overall, the sparsity of data points is still a problem, which due to the fact that Thomas does not produce that many questions overall (also see Dąbrowska, 2004, for a comprehensive analysis of formulaicity in Thomas’ canonical questions). Furthermore, only Thomas1 falls within the dense sampling period, suggesting that denser sampling regimes should be used for longer periods in order to allow for more fine-grained analyses of developments occurring after about 3; 0. Before proceeding to the functional analysis of the strong assertives It’s X? and That’s X? , two issues regarding their formal characteristics need to be addressed. First of all, the structures can potentially be realized with or without contraction of the copula (i.e., It’s X? vs. It is X? ) and many children have been found to “go through a stage when they prefer the full forms, sometimes after a period of fairly consistent use of contracted forms” (Dąbrowska, 2004, p. 191); in informal adult speech, the contracted variants are usually preferred (Dąbrowska, 2004, p. 191). The sample contains only two uncontracted cases overall, one by the mother (1), and one by Thomas (2), involving you are and it is, respectively. (1) (file 2-04-30) CHI: ta@d much (in)deed. MOT: thanks. MOT: ta@d much indeed. MOT: oh you are polite? MOT: do you like this? (2) (file 2-10-06) MOT: it made a sound, didn’t it? MOT: like a bang. CHI: yes. CHI: it is bell? CHI: like a ding+a+ling. CHI: that-'is a bell. CHI: see ding+a+ling+a+ling. This finding indicates that the mother (as well as the other caretakers) have a strong preference for the contracted forms. Furthermore, Thomas does not go through a stage in which an analysis of it’s and that’s into the respective component parts leads to an increase in the use of uncontracted forms. Secondly, the fact that It’s and That’s have been analyzed with respect to their assumed canonical counterparts is not meant to imply that these constructions are directly related to one another in the speakers’ mind. Recall that in contrast to generative-transformational approaches, question formation is not thought to involve any kind of movement operation within the usage-based, constructionist paradigm (note that a less direct relationship is not only possible, but also posited to exist, for example via associative links <?page no="163"?> 149 between the component parts of the constructions). Along similar lines, the fact that ‘uninverted’ questions are attested in the input does not necessarily entail that all cases of uninversion in the child should be regarded as wellformed (nor that ‘inversion’ in yes-no questions should be regarded as optional). Rather, it is suggested that since these structures occur in the input, it is simply not tenable to judge them as ‘ungrammatical’ whenever they occur in child speech. It is possible, however, to judge in how far the child’s use of these constructions can be considered pragmatically appropriate. To this end, it is necessary to analyze the discourse-functional properties of the strong assertives It’s X? and That’s X? used by the caretakers. The next section presents such an analysis, taking as a starting point the functional characteristics of strong assertives as proposed in the literature. 7.2 The Discourse Function of Strong Assertives Recall that echo-questions structurally mirror the utterance they are responding to. Therefore, strong assertives should be found to echo declaratives with medial auxiliaries, potentially embedded within the same variation sets (i.e., rephrasings and expansions) that have been found with echouses of the weak assertive You want (to) X? , as discussed in the preceding chapter. Additionally, strong assertives have been proposed to be used when “a fact is assumed, pending confirmation” (Bolinger, 1957, p. 59). Consider the following examples ((3) and (4), taken from Bolinger, 1957, p. 59): (3) A: It’s too hot. B: It’s too hot? (4) A: He hasn’t any money left. B: He spent all of it? In (3), speaker B takes up the assumption established by speaker A and echoes his original utterance. In contrast to this, the assertive question in (4) asks for the confirmation of an assumption speaker B has established on the basis of the preceding utterance by speaker A, without echoing (a part of) his utterance. Furthermore, in contrast to echo-uses, assertive questions asking for the confirmation of an assumption can also be used to initiate a conversational exchange, especially in order to bias the interlocutor towards an answer “with the same propositional content as the question” (Huddleston & Pullum, 2006, p. 881); see examples (5) and (6). Even though some canonical, i.e., ‘inverted’ yes-no questions can also be biased, this only goes for negative questions with preposed negation (7) (Huddleston & Pullum, 2006, p. 881). <?page no="164"?> 150 (5) You’ve finished the paper on time? (bias towards positive answer) (6) You haven’t finished the paper on time? (bias towards negative answer) (7) Doesn’t he drink a lot? (bias towards a negative answer) Therefore, it is to be expected that (at least in the input), strong assertives will not be found to be used interchangeably with their canonical, ‘inverted’ counterparts. First, for all cases of It’s X? (n = 28) and That’s X? (n = 14) in Thomas’ input, based on a discourse context of 20 speaker turns (10 preceding, 10 following the target construction), the respective discourse function was a assessed and each utterance was coded as either ‘Request for confirmation (echo-use)’ or ‘Initiating (biased)’. Even though the coding scheme included a category ‘Other’ to be able to account for cases fulfilling other functions (e.g., initiating (non-biased)), no such uses were found in the input. The results are summarized in Table 7.2. As can be seen, both the caretakers and Thomas use both constructions for the expression of the posited discourse functions, albeit to slightly different degrees. However, due to the small sample size, it cannot be reliably assessed whether these differences are statistically significant. Nevertheless, analyzing these patterns in a qualitative manner also yields interesting results, which will be presented and briefly discussed below. Table 7.2: Distribution of discourse functions expressed by It’s X? and That’s X? It’s X? That’s X? INP CHI INP CHI Request for clarification (echo-use) 17 (61%) 10 (67%) 10 (71%) 4 (40%) Initiating (biased) 11 (39%) 5 (33%) 4 (29%) 6 (60%) Total 28 (100%) 15 (100%) 14 (100%) 10 (100%) 7.2.1 That’s X? and It’s X? as requests for clarification (echo-uses) Analogous to You want (to) X? , which is also primarily used as a request for clarification in the early sessions (Section 6.2), there are many examples (n = 27) of such uses for That’s X? and It’s X? in Thomas’ input (see (8) and (9), respectively): (8) (file 2-03-19) CHI: oh <Nin+Nin and@sc> [? ] Dimitra. CHI: oh what’s this? <?page no="165"?> 151 CHI: wash it. MOT: wash it? CHI: yeah a@sc Mitra. INV: that’s for me? In this example, Thomas wants to offer a toy to Dimitra (one of the investigators visiting Thomas and his mother on a weekly basis). However, he does not clearly indicate this linguistically, leading to an echo-question by the adult. (9) (file 2-03-23) MOT: you dropped the talcum powder and the lid wasn’t shut, was it? CHI: no. MOT: powder on the floor. CHI: it’s [? ] a big one coming. MOT: what did you say? MOT: it’s a big one coming? Even though there is an intervening utterance, this is a clear case of an echoquestion, since the mother explicitly indicates that she has not understood Thomas. The fact that his utterance may have been hard to understand is also mirrored in the transcript, which contains a question mark indicating the transcriber’s ‘best guess’ (MacWhinney, 2000, p. 71). 7.2.2 That’s X? and It’s X? as initiating questions (biased) Even though this function is less prevalent than the echo-uses discussed so far, there are examples of initiating That’s X? and It’s X? -questions (n = 15) in Thomas’ input (see (10) and (11), respectively): (10) (file 2-04-06) MOT: yes. MOT: we’re still trying to work out how to do the +/ / . [+ IN] CHI: <shoping> [? ]. MOT: upstairs. MOT: there must be a little knack to it that Mummy can’t work out. MOT: perhaps that’s it? CHI: throw the [? ] one tree. MOT: oh no. MOT: don’t throw the trees. This use of That’s X? is clearly different from the echo-uses in that the mother does not refer to something that Thomas has said. Instead, she makes a tentative suggestion, which is also indicated by her use of the pragmatic marker perhaps. <?page no="166"?> 152 (11) (file 2-03-02) MOT: it-'is a nice morning, isn’t it? CHI: nice sunny morning. MOT: a nice sunny morning. CHI: chip shops. MOT: chip shops. MOT: oh it’s too nice to go to the chip shop today, Thomas? CHI: no. MOT: it’s too warm. CHI: too warm. [+ I] MOT: too warm. MOT: we’ll go and see Jo and Becky. This example is an even clearer case, since it is obvious that the mother does not ask Thomas for his opinion in a neutral way but wants him to confirm her assumption that is too nice to spend the day inside. 7.2.3 That’s X? and It’s X? as requests for clarification (echo-uses) in Thomas’ speech Analogous to his caretakers, Thomas makes use of the strong assertives That’s X? and It’s X? as requests for clarification (n = 14, examples (12) and 13)): (12) (file 2-11-28) CHI: oh! CHI: what’s that? MOT: rice. MOT: I’ve just realised, Thomas, I need to boil the kettle. MOT: so we’ll just +... [+ IN] CHI: that’s rice? MOT: that’s rice. MOT: I just need to boil the kettle so we'll have a little break [>]. CHI: that's [<] black [? ] rice. CHI: <I see> [=? that's it] [>]. MOT: <black rice> [<]? Listening to the audio sample reveals that Thomas seems really surprised to see black rice, indicating that he echoes his mother’s statement out of incredulity and not because he has not understood what she has said. It is interesting to note that the mother responds to his question with an ‘echodeclarative’ exhibiting a falling intonation. Exchanges like this might make the role of prosodic factors salient for the child, even though this is just a very tentative suggestion at this point. (13) (file 3-10-02) <?page no="167"?> 153 MOT: we have the dustbin man here on a Monday. MOT: he was called Mr. Fowler and he used to wear a big hat. CHI: like a cook? MOT: like a cook. MOT: it wasn’t a cook’s hat but it was a big hat. CHI: it’s like a cook? MOT: <and underneath> [<] +... [+ IN] MOT: yes (.) perhaps. MOT: and underneath his hat (.) he had no hair! MOT: he was bald. MOT: I think that’s why he used to wear a hat. In this example, Thomas expands an earlier fragmentary question (Like a cook? ) in order to get a verification of his assumption that Mr. Fowler’s hat is like one usually worn by cooks. Interestingly, it shows that strong assertives may in fact be the result of ‘right-to-left’-elaboration of a less complex structure. Cases like this suggest that the study of conversational exchanges may be a rich source for gaining further insights into potential relationships between constructions and the ways in which they are assembled during language production. 7.2.4 That’s X? and It’s X? as initiating questions (biased) in Thomas’ speech Finally, there are 11 cases in which Thomas uses That’s X? and It’s X? to offer tentative suggestions awaiting verification by the caretaker. (14) (file 3-02-10, talking about pictures on a margarine box ) CHI: what’s that hole [? ]. CHI: that’s a kite? MOT: pardon. CHI: that’s a kite? [+ SR] MOT: well it [/ ] well it is. MOT: this margarine is flora and that is a red heart shape because it's trying to tell you it's good for your heart. MOT: but that’s right. MOT: they’ve put a little string on it. MOT: it’s a kite. In this example, Thomas notices that the heart displayed on the margarine box resembles a kite, since it appears to be up in the air and has a string attached to it. After closer examination of the picture, the mother concludes that the object may indeed be construed as a kite. This is also an example of a paraphrase relation, with the mother replacing the demonstrative pronoun that in Thomas’ utterance with it in her confirmation of Thomas’ tentative proposition. <?page no="168"?> 154 (15) (file 2-11-26, talking about a parcel) CHI: big post lorry? MOT: no [/ ] no. MOT: it’s not a lorry but it’s in a parcel. MOT: it’s all wrapped up. CHI: it’s a ginger cat for me? MOT: hmm [=! laugh]. MOT: a ginger cat for you? MOT: I don’t think so. MOT: and even if it was I don’t think we’d leave it wrapped up in a parcel for you. Even though in this example the use of a strong assertive rather than a canonical question may have been primed by the mother’s previous declarative utterances starting with it’s, it still fulfills the criteria for being classified as a non-echo use, since Thomas obviously wants to bias his mother towards a positive reply. Overall, the results indicate that Thomas’ use of strong assertives may largely be due to the fact that he pursues particular communicative goals rather than to missing linguistic knowledge (about inversion or - from a usage-based, constructivist perspective - about item-based frames with particular initial auxiliary-subject combinations). Of course, the coding of discourse function is always subjective to a certain degree so it might be that some cases could also be regarded as pragmatically odd, or as entirely inappropriate uses of strong assertives by Thomas. Still, this will not be the case for all examples discussed, indicating that strong assertives in child speech are not to be regarded as pragmatic errors ‘per se’. In addition to this, the fact that strong assertives are found in the target repertoire indicates that they are certainly not grammatical errors ‘per se’. Furthermore, it is obvious that strong assertives can be used for a variety of different functions, particularly as echo-questions (asking for clarification or expressing incredulity) and initiating questions requesting the verification of a tentative assumption or suggestion. 7.3 Summary This chapter has explored the learning and use of another sub-class of noncanonical yes-no questions, i.e., strong assertives containing a medial auxiliary. First, a modified version of Dąbrowska’s (2004) method for assessing the distribution of strong assertives (with medial auxiliary) vs. canonical questions (with auxiliaries other than DO) was used to assess the ‘uninversion rate’ of constructions with particular aux-subject combinations in both the input and Thomas’s speech. It was shown that while the overall inver- <?page no="169"?> 155 sion rate in the input is rather high, it rarely reaches 100%, suggesting that strong assertives are part of the adult language repertoire and thus should not be classified as ‘uninversion errors’ when they are encountered in child speech. In order to address the question of whether Thomas’ uses of strong assertives can also be regarded as pragmatically appropriate, the discourse function of That’s X? and It’s X? in child(-directed) speech was analyzed in more detail. It was found that these constructions are indeed used as suggested in the literature, i.e., as echo-questions or assertive questions biasing the interlocutor towards a positive reply (Bolinger, 1957, p. 59), both by Thomas and his caretakers. Overall, the results from this and the preceding chapter indicate that non-canonical questions are worthy of investigation in their own right and that much can be gained from the investigation of larger discourse patterns, thus moving beyond the level of analysis provided by focusing only on the formal characteristics of decontextualized utterances. In the following chapter, it is suggested that the study of canonical questions can also profit from the inclusion of discourse-functional aspects. Focusing on Can-PERSON-PROCESS? , the analyses presented illuminate our understanding of the discourse functions fulfilled by questions and the role played by these functional factors in the acquisition process. <?page no="170"?> 156 8 The Role of Functional Factors in the Acquisition of Canonical Yes-No Questions 111 Up to now, the main focus of this work has been on the acquisition and use of non-canonical questions. However, as has already been pointed out in Section 3.4, there are many open issues pertaining to canonical questions as well, particularly concerning their discourse-functional properties. The lack of research on form-function mappings in canonical yes-no questions and the impact of functional factors on the acquisition process is striking, since, from a truly usage-based perspective, the communicative intentions underlying the use of constructions, i.e., their functional components, are just as important as their formal characteristics and it has been stressed repeatedly that meaning is “probably the single most important characteristic of language from the child’s point of view” (Lieven, 2010, p. 100; also see Behrens, 2009c). The lack of research focusing on or at least taking into account functional factors can be attributed to a number of factors: First of all, recent technological advances have made it possible to work with large corpora such as the ones made available on CHILDES (Panther & Thornburg, 2003b) and features like automated part-of-speech-tagging and powerful search mechanisms enable us to perform advanced statistical analyses on huge amounts of data. Along similar lines, the resources needed to develop and implement very complex computational models in order to simulate learning processes are readily available. Of course, these advances are to be welcomed, since they enable us to construct and test emergentist accounts using quantitative methods. As a consequence, however, the functional side of constructions has been neglected almost entirely, which is also due to practical considerations: While the formal coding of utterances is (relatively) straightforward and can in many cases even be done (semi)automatically, functional coding procedures pose a number of problems. First of all, in contrast to (different aspects of) linguistic form, the notion of meaning or function is very hard (if not impossible) to operationalize. Secondly, functional coding cannot be done automatically, since aspects like discourse context and shared knowledge have to be taken into account, making it a very time-consuming process. Lastly, the judgments involved are inevitably subjective to a certain degree. 111 This chapter is a revised version of Kania (2013). <?page no="171"?> 157 Despite these difficulties, we should be careful not to get “trapped in a numerical maze by the irresistible lure of masses of data” (Schmid, 2010, p. 102). One of the basic tenets of the usage-based model is that “real language use, in all its complexity, must be the basis of linguistic research” (Glynn, 2010, p. 10), so in order to stay true to and fully exploit the possibilities offered by the usage-based, constructivist framework, we also need to consider how individual constructions are used in actual conversations. While Chapters 6 and 7 have provided a largely descriptive account of the ways in which particular non-canonical questions are used, functional factors also possess an explanatory power. Besides the fact that functional characteristics are of interest in their own right, there are already some studies suggesting that formal frequency alone cannot completely account for the observed patterns. Evidence for the importance of function was found in a case study on negation showing that the child’s linguistic development was driven by both formal and functional frequencies in the input (Lieven et al., 2009). Furthermore, a case study by Cameron-Faulkner (2012) revealed that the child’s use of verbs in early multiword utterances is closely tied to their use in the input. Another example is a study on passiveand future-constructions in German by Abbot- Smith and Behrens (2006), in which they found that the acquisition of a new construction was blocked by a functionally very similar construction which had already been learned. What is more, it has been shown that predictions concerning the order of acquisition which are based on formal frequencies alone are not always borne out by the data (see, e.g., Steinkrauss 2009, 2011, on wh-questions in German). Tying in with this previous research, this chapter focuses on one case in which Thomas’ order of acquisition and frequency of use runs counter to what would be expected based on the input he receives. Analyzing the emergence and use of different Can-PERSON-PROCESS? -constructions, this chapter thus provides an in-depth analysis of the role played by functional factors. 112 Section 8.1 presents a correlation analysis exploring the relationship between the relative frequencies of specific auxiliary-subject combinations found in the input and Thomas’ age of acquisition and relative frequency of use of the same initial clusters. Based on the assumption that outliers may provide particularly interesting cases for functional analyses (Steinkrauss, 112 In line with previous research (e.g., Dąbrowska and Lieven, 2005, p. 444), semantic labels (i.e. PERSON, PROCESS) rather than traditional labels like ‘subject’ are used in order to stress that the slots are not entirely syntactically defined (however; they are not defined solely on a semantic basis either because they are “formed by ‘functionallybased distributional analysis’ (Tomasello, 2003), i.e., by abstracting across items that have similar distributions AND share similar communicative functions” (Ambridge and Rowland, 2009, p. 230). <?page no="172"?> 158 2009, 2011), constructions which are used by Thomas with either a considerably higher or lower relative frequency than that found in the input are identified. Section 8.2 presents a short review of the functional characteristics of canonical questions (for a more comprehensive discussion, see Section 3.4). In Section 8.3, the emergence and frequency of use Can-I-PROCESS? is discussed in more detail, while Section 8.4 provides an in-depth analysis of its functional characteristics, with Section 8.4.1 zooming in on the functional frequency of Can-I-PROCESS? as a permission-seeking construction. 8.1 Correlation Analysis The notion of frequency plays a very important role in Construction Grammar accounts, since the absolute as well as relative frequencies of occurrence of particular linguistic items is thought to be closely linked to the strength of their mental representation, i.e., their degree of entrenchment (e.g., Barlow & Kemmer, 2000; Bybee, 1995; Goldberg, 2006; Langacker, 1987). Within language acquisition research from a usage-based, constructivist perspective, it has often been noted that there is a relationship between the frequency of particular structures in the input and their order of emergence and frequency of use in the child’s language. For example, studies on wh-questions have revealed that relative input frequency is the best predictor for the order of acquisition of particular wh-auxiliary-combinations (e.g., Rowland & Pine, 2000; Rowland et al., 2003; also see Clancy, 1989, for initial wh-words in Korean). Similar findings were obtained for English declaratives containing the auxiliaries HAVE and BE, concluding that “the order in which children acquire specific subject-auxiliary combinations ... will reflect their relative frequencies in the language children hear” (Theakston et al., 2005, p. 255). The focus on particular utterance-initial combinations (wh-word + auxiliary, and subject + auxiliary) in these structures is based on the rationale that “a combination of high token frequency and low type frequency is said to lead to the entrenchment of word combinations” (Rowland, 2007, p. 113). The low type frequency in these cases stems from the fact that auxiliaries as well as wh-words are closed-class items and subjects in child(-directed) speech are “filled by a limited range of pronominal and nominal forms” (Rowland, 2007, p. 113). Moreover, research on child-directed speech has shown that mothers use only a limited range of auxiliary + subject-combinations (Cameron-Faulkner et al., 2003). 113 113 The high degree of entrenchment of particular utterance-initial combinations has also been suggested to lead to a higher percentage of correctness in children’s production of wh-questions, indicating that well-entrenched patterns are protected from error (Ambridge & Rowland, 2009; Rowland, 2007). <?page no="173"?> 159 This indicates that the learning of these structures is closely tied to knowledge about item-specific patterns (also see Dąbrowska & Lieven, 2005, for a study revealing that Thomas’ early questions on the whole are highly formulaic). Moreover, input frequency seems to be a driving factor in the acquisition process, since the relative frequency of use of particular structures has been found to be a good determinant of their order of acquisition by the child. However, the correspondences between input and child language are not entirely straightforward: The study on subject+auxiliarycombinations involving HAVE and BE by Theakston et al. (2005) found the correlation between the frequencies in the input and the age of acquisition found in the children’s speech only to be significant once the outliers you’re and you’ve were removed and they conclude that “clearly, these forms depend on something in addition to input frequency for their acquisition” (p. 254). That (part of) this ‘something’ could be discourse-function was indicated by Steinkrauss (2009) in his high-density corpus-study on the acquisition of wh-questions by Leo, a boy learning German as his first language (Behrens, 2006). Referring to the exclusion of outliers from correlation analyses, he states that “the downside of this method is that the excluded constructions might just be particularly interesting to look at” (Steinkrauss 2009, p. 65). On the basis of the assumption that these interesting cases might shed further light on influential factors other than input frequency, the following correlation analyses were used to identify outliers, one of which will be focused on in the subsequent functional analyses. Included were all canonical questions containing HAVE, BE, and modal auxiliaries, focusing on the initial auxiliary+subject-combination. Like in Chapter 7, counts for the input included all utterances addressed to Thomas up to 2; 06 (i.e., the first 119 transcripts in the corpus), since it is around this time that Thomas acquires the first auxiliary-subject-frames and starts using them productively. These structures were chosen not only because these initial combinations have been found to exhibit a high token, but a low type-frequency (e.g., Theakston et al., 2005), but also because this group of constructions is the most frequent for the input as well as for Thomas both within the group of canonical-questions and within yes-no questions overall (32% of all questions in the input, 30% of all questions asked by Thomas). Analogous to Theakston et al. (2005), I compared the input frequency and the order of emergence (i.e., age of acquisition) of particular auxiliary-subject frames, “defined as the second use of a given combination” (Theakston et al., 2005, p. 263). 114 Additionally, relative input frequencies and relative frequencies of use in Thomas’s speech were compared. 114 Following Theakston et al. (2005), combinations that were not acquired by Thomas during the recoding-period “were assigned an age of acquisition seven days after the last recording” ( p. 264), i.e., at 4; 11.27 (=1810 days). <?page no="174"?> 160 In the vast majority of cases, there was no material preceding the auxiliary+subject-combination in the respective utterance. If there was other material in initial position, it was often separated from the rest of the utterance by a pause, a conjunction, or a syntactic boundary (see example (1); this observation is similar to the one made by Steinkrauss 2009, p. 23, on material preceding wh-word+subject-combinations in German). This material was consequently discarded (subordinate clauses; conversational sounds like whoops, erm, etc.; proper names; coordinating conjunctions) (1) (file 2-01-00) MOT: Thomas, would you like to sit in your chair and watch the Teletubbys? All auxiliary+subject-combinations that were used at least 5 times by either the adults or the child were included. Spearman rank order correlations were then calculated for the relative frequencies in the input and Thomas’ speech, including all subject+auxiliary-combinations which were found at least five times in the input the relative input frequencies and Thomas’ age of acquisition, including all subject+auxiliary-combinations which were found at least five times in the input the relative frequencies in the input and Thomas’ speech, including all subject+auxiliary-combinations which were used at least five times by Thomas the relative input frequencies and Thomas’ age of acquisition, including all subject+auxiliary-combinations which were used at least five times by Thomas 115 When all the subject-auxiliary pairs that were used five or more times in the input were included, there was a moderate positive correlation between input and production frequency (Spearman’s rho = 0.49, N = 44, p = 0.002) and a moderate negative correlation between input frequency and age of acquisition (Spearman’s rho = -0.52, N = 44, p <0.001). When only those subject-auxiliary pairs that were used five or more times by the child were included, there was again a moderate positive correlation between input frequency and frequency in Thomas’ production (Spearman’s rho = 0.58, N = 17, p = 0.015) and a moderate negative correlation between input frequency and age of acquisition, i.e., the more frequent a subject-auxiliary pairing was 115 Spearman rank correlations rather than Pearson correlations have been used because the former not only takes into account the non-normal distribution of the data but is also less sensitive to outliers. <?page no="175"?> 161 in the input, the earlier it was acquired (Spearman’s rho = -0.63, N = 17, p = 0.007). The fact that the correlations are only moderate suggests that there are structures which are quite frequent in the input which are acquired rather late and which are thus ‘underused’ by the child. Secondly, there are also structures which are acquired earlier and used with a higher frequency than would be predicted on the basis of their relative input frequency alone (i.e., they are ‘overused’). These outliers were identified by calculating the absolute differences between relative input and production frequency and identifying all subject+auxiliary-combinations which had an input-output difference higher than the double standard deviation of all input-production differences (this is based on the method used by Steinkrauss, 2009, p. 68). In total, there are five outliers, with three constructions being ‘underused’ and two being ‘overused’ by Thomas (see Table 8.1). Table 8.1: Subject+auxiliary-combinations ‘overused’ or ‘underused’ by Thomas Absolute (and relative) frequency of use ‘underused’ INP CHI are you 1891 (33%) 39 (4%) shall we 490 (8%) 28 (3%) would you 783 (14%) 79 (9%) ‘overused’ shall I 122 (2%) 121 (14%) can I 49 (1%) 279 (32%) It is striking that the two ‘overused’ constructions contain the first-person subject I, and two of the three underused constructions the second-person subject you. However, this cannot be (entirely) due to a general ‘I/ youasymmetry’ since there are quite a few I+auxiliary-combinations in the input which are not overused by Thomas (in fact, Am I, May I, and Could I are hardly ever used by Thomas). Can I is the construction most ‘overused’ by Thomas and its analysis offers the opportunity to account for a large proportion of Thomas’ questions. The following sections explore the possibility that this overuse is due to functional factors by exploring form-function mappings within the larger paradigm of Can-PERSON-PROCESS? constructions in more detail. <?page no="176"?> 162 8.2 Form-function Mappings in Canonical Yes-No Questions Recall (from Section 3.4.2) that it is assumed that syntactic questions are prototypically used in cases where “the speaker lacks but wants certain information, believes that the hearer possesses it, and directs the hearer to supply it” (Langacker, 1991, p. 505; also see Quirk et al., 1985, p. 804). Of course, the fact that syntactic questions can be used for a variety of other functions such as requesting an action or directing attention is usually acknowledged (e.g., Biber et al., 1999, p. 202; Huddleston and Pullum, 2006, p. 853; Langacker, 1991, p. 506; Quirk et al., 1985, pp. 805-806). However, these other functions are mostly seen as deviations from the more prototypical information-seeking question. For example, syntactic questions which are used as requests for action are classified as indirect speech acts within traditional frameworks (e.g., Searle, 1975). More recent cognitive approaches conceptualize this indirectness differently by proposing metonymic links between indirect speech acts such as Can you pass the salt? and e.g., the cognitive model for ‘giving’ (Panther & Thornburg, 1998, 2003a, 2004; Thornburg & Panther, 1997). Nevertheless, these cases are still seen as being in some way based on the functional prototype, with the correct interpretation of ‘indirect’ uses being usually computed ad-hoc. Interestingly, Panther and Thornburg address the possibility that indirect uses “can also, through frequency of use, become a conventionalized meaning, stored separately in the lexicon” (Panther & Thornburg, 2004, p. 97). Exploring this possibility in more detail, Stefanowitsch (2003) proposed that conventionalized indirect speech acts can be viewed as constructions whose “illocutionary force is directly linked to their form (and [which] are thus not really ‘indirect’ speech acts at all)” (Stefanowitsch, 2003, p. 106). 116 This chapter evaluates this possibility from a developmental perspective by analyzing the form-function mappings of different Can-PERSON-PROCESS? constructions in more detail. Evidence for the assumption that conventionalized indirect speech acts may indeed be learned as form-function pairings from the input would also offer an explanation for the fact that these structures are found to be comprehended and produced by rather young children, both in naturalistic and experimental settings (e.g., Bara & Bucciarelli, 116 Concerning syntactic questions as a whole, it has even been suggested that “situations in which an interrogative merely has its ‘literal’, compositional value are probably the exception rather than the rule” (Taylor, 2012, p. 94), implying that they do not prototypically request information. Conversely, other forms (e.g., declaratives with a neutral or falling intonation) often function as requests for information: In her study of adult-toadult conversations, Weber (1993) notes that 41% of all questions were noninterrogatives, calling into doubt the assumption that requests for information are prototypically expressed by interrogatives (Weber, 1993, p. 220). <?page no="177"?> 163 1998; Bernicot & Legros, 1987; Bucciarelli, Colle, & Bara, 2003; Garvey, 1975; Horgan, 1977; Shatz, 1978). Unfortunately, apart from very few exceptions (Shatz, 1979; Steinkrauss, 2009, 2011), previous developmental research on indirect speech acts and question use in general has only been concerned with mappings of form and function on a very abstract level, usually only differentiating between yesno-, tag-, and wh-questions. As a consequence, we do not know much about the functions expressed by specific forms and about different forms that can be used to express the same or at least very similar functions. From a usage-based, constructivist perspective, this is a crucial issue, since constructions are formalized (and learned) as pairings of forms with particular functions and language users are expected to make use of several lower-level constructions instead of operating with only one abstract constructional schema for all yes-no questions, for example. In order to address these issues, the next sections focus on one such lower-level construction (Can-PERSON-PROCESS? ), providing insights on form-function mappings on a very fine-grained level. 8.3 Formal Coding and Analyses In contrast to the correlation analysis presented in Section 8.1, all utterances containing Can + subject combinations found in Thomas’ input were considered here. Therefore, the dataset included all turns which either begin with or contain a Can-PERSON-string and a verb, do not contain the negator not in its full or contracted form, and are produced by either an adult (=INP) or the target child (=CHI). Applying these criteria resulted in 2864 utterances - since there are only very few instances of proper names or pronouns other than I and you, these cases have been subsumed under the category Canother for the present analysis whose results are summarized in Table 8.2. 117 Table 8.2: Formal frequencies of Can-PERSON-PROCESS? 118 Can I Can you Can other Total INP 291 (12%) 2000 (83%) 131 (5%) 2422 (100%) CHI 293 (66%) 110 (25%) 39 (9%) 442 (100%) 117 Besides utterances containing proper names and pronouns referring to people other than the child or the caretaker, there were 25 cases in which the mother used Mummy instead of I to refer to herself and 10 cases where she or the investigator used Thomas rather than you to refer to the child. Just like Thomas’ only me-for-I error (And can me do that as well? ) these cases were subsumed under the category Can other. 118 All percentages provided in this and the following tables are rounded to the nearest whole number. <?page no="178"?> 164 Based on formal frequency alone, we would expect constructions with Can you to emerge before Can-I-PROCESS? -constructions and to be used more frequently as well. However, the exact opposite is the case: Can-I-PROCESS? is used much more frequently than the relative input frequency would suggest, while the opposite is true for the relationship between relative input frequency and the child’s productions of Can-you-PROCESS? . Furthermore, applying the criterion of at least two different, non-imitative, non-repetitive tokens of the respective Can-PERSON-frame, Can-I-PROCESS? is acquired at 2; 10.07, while Can-you-PROCESS? is acquired only later, at 3; 0. 119 In order to exclude the possibility that the observed pattern is due to changes in the adults’ linguistic behavior in the course of the sampling period (e.g., initially making extensive use of Can-I-PROCESS? and increasing the frequency of Can-you-PROCESS? only gradually), the dataset was split in two parts: The first part includes the input utterances before Thomas’ acquisition of Can-I-PROCESS? while the second part captures the period from 2; 10.07 onwards (see Table 8.3). Table 8.3: Formal frequencies of Can-PERSON-PROCESS? in the input Can I Can you Can other Total INP < 2; 10.07 95 (7%) 1177(88%) 68 (5%) 1340 (100 %) INP ≥ 2; 10.07 196 (18%) 823 (76%) 63 (6%) 1082 (100 %) It can be seen that the relative frequency of Can-you-PROCESS? is even somewhat higher in the first part of the sampling period (88% compared to 76% of all Can-PERSON-PROCESS? -constructions) and that the overall pattern stays constant, with Can-you-PROCESS? exhibiting by far the highest relative frequency throughout. Of course, it might be that there is still a formal match between the input and Thomas’ own use of Can-PERSON-PROCESS? on a more fine-grained level. Since the dense sampling allows for an assessment of the degree of lexical specificity beyond the initial auxiliary-subject combination, the four verbs used most frequently in the PROCESS-slot were determined for both Thomas’s input in the first period (i.e., up to 2; 10.17) and Thomas’ (overall) use of Can-I-PROCESS? and Can-you-PROCESS? , respectively. The results are visualized in Figure 8.1 (for Can-you-PROCESS? ) and 8.2 (for Can-I- PROCESS? ), respectively. 119 Even if a stricter view is adopted and productive mastery of the respective Can-I- PROCESS? -constructions is defined as at least three different, non-imitative, nonrepetitive tokens of the respective Can-PERSON-frame, the earlier learning of Can-I- PROCESS? persists and the gap gets even bigger (with Can-I-PROCESS? acquired at 3; 01.23 and Can-you-PROCESS? acquired at 3; 06.00), suggesting that Can-I-PROCESS? is also used with greater flexibility than Can-you-PROCESS? . <?page no="179"?> 165 For both Can-I-PROCESS? and Can-you-PROCESS? , there are only two ‘matches’, i.e., verbs which are used with a high relative frequency by both Thomas and his caretakers. Still, within these matches, Can you see X? can be said to be ‘underused’ by Thomas, since it only accounts for about 15% of all Can-you-PROCESS? constructions. In contrast to this, 66% of all Can-you- PROCESS? constructions used by the caretakers take the form Can you see X? . Since Can-I-PROCESS? is the construction that has been found to be ‘overused’ by Thomas in the correlation analysis, it is interesting to note that within the group of Can-I-PROCESS? , the relative frequency of the particular item-based formula Can I have X? is very similar (39% in the input vs. 45% in Thomas’ speech). It seems that certain three-word frames are particularly salient for Thomas, especially Can I have X? , which may be due to its discourse-functional properties. It is clear, then, that the earlier acquisition and more frequent use of Can- I-PROCESS? cannot be explained entirely on the basis of its relative frequency in child-directed speech, even if one takes into account the degree of lexical specificity found in the PROCESS-slot. Therefore, the aim of the analyses presented below is to find out in how far the observed pattern in the child’s language development can be accounted for by appealing to functional factors. First, in order to establish which discourse functions the respective constructions actually fulfill, all Can-PERSON-PROCESS? constructions were coded for discourse function (Section 8.4). Since a very consistent mapping of form-to-function was found for Can-I-PROCESS? , Figure 8.2: Percentages of Can you X? accounted for by the four most frequent verbs Figure 8.1: Percentages of Can I X? accounted for by the four most frequent verbs <?page no="180"?> 166 additional analyses were carried out to establish the frequency of Can-I- PROCESS? relative to its functional rather than formal competitors (Section 8.5). 8.4 Functional Coding and Analyses In the analyses presented in Chapters 6 and 7, the concept of discourse function was mainly appealed to on the descriptive level. Besides, cases like echo-uses could be identified unambiguously on the basis of the preceding utterance (A: That’s it. B: That’s it? ). In contrast to this, the current analysis makes claims concerning the explanatory level, positing discourse function to be a driving factor in acquisition. Therefore, this section first discusses the notion of discourse function in more detail, before presenting the results of the functional analyses. A comprehensive discussion of meaning or discourse function is beyond the scope of this (and probably any) work; I will therefore confine myself to describing the reasoning behind the choices made in the current study. Within the usage-based constructivist approach to language development, the ability to understand communicative intentions is seen as a crucial prerequisite for the acquisition of linguistic constructions. Children are able to arrive at the intended meaning of an utterance because they see other persons (and themselves) as intentional agents who are trying to manipulate the mental or intentional states of others by using various linguistic expressions, mostly within joint-attentional frames (Tomasello, 2003). Importantly, the notion of communicative intention is correlated with function, since if “someone uses a piece of language with a certain communicative intention, ... we may say that that piece of language has a certain function” (Tomasello, 2003, p. 3). Along similar lines, (discourse) function may in turn be used interchangeably with meaning, which is indeed what we find in the literature, for example in definitions of ‘construction’ within Construction Grammar frameworks (e.g., Goldberg, 2006, p. 5). It is thus the intended meaning or function of specific utterances in specific contexts (i.e., what has traditionally been called ‘illocutionary force’) that children focus on, rather than ‘literal’ meanings that might be posited for any given expression (i.e., the ‘locutionary force’, Villiers & Villiers, 1985). This also entails that when we assign a specific function to a specific utterance, we are always describing not only the utterance itself but also certain aspects of the context in which it was used (Shatz, 1979; Sinclair & van Gessel, 1990). Second of all, (not only) children probably make use of rather concrete, lowlevel functional categories, analogous to their reliance on item-based constructions and low-level abstractions on a formal level. In order to capture similarities on a functional level while staying as close to the data as possible, the categories were established after examination of the dataset, adapt- <?page no="181"?> 167 ing coding schemes developed in the context of earlier studies on syntactic questions (Holzman, 1972; Shatz, 1979) rather than relying on more comprehensive but also more abstract coding schemes developed for the classification of all utterances regardless of sentence type (Austin, 1962). In most cases, the chosen discourse window, which consisted of the target utterance plus the three preceding and following speaker turns, was sufficient to identify the functional category. Whenever this was not the case, additional parts of the discourse context were considered until the corresponding function could be established or the utterance was judged to be ‘unclassifiable’. Furthermore, since the audio data is available for all but two of the transcripts, the material could be listened to for the confirmation (or reevaluation) of a judgment. In order to make quantitative analyses possible, each utterance was assigned to one functional category only, even though utterances can of course fulfill more than just one communicative function. Interestingly, there were almost no cases in which the identified intended meaning did not (immediately) lead to the expected reaction by the child, in which case the interlocutor would choose an alternative form: 120 (2) (file 2-05-06) INV: can you open it? CHI: yes. INV: yeah? INV: open it then. In order to ensure reliability, 200 randomly selected utterances were judged by a second coder, who was blind to the aims of the current study, obtaining a Cohen’s Kappa of 0.86. In total, there were 6 categories: 1. Sincere requests for information This is assumed to be the primary function of interrogatives and includes all cases in which the speaker wants to elicit from the hearer information the speaker does not have: (3) (file 2-07-19, mother is looking for a toy) MOT: it would be helpful if you could tell me. MOT: can you remember? CHI: no. MOT: oh dear. 120 Note that the ‘failure’ to comply immediately to this request for action does not entail that Thomas has not understood the investigators communicative intention, since “failure to comply or even to acknowledge a directive may be deceptive because the child may not want to comply” (Ervin-Tripp, 1977, p. 179; also see Garvey, 1975). <?page no="182"?> 168 2. Seeking Permission All cases in which the speaker wants to obtain permission (usually to perform an action) are grouped in this category: (4) (file 3-07-01) CHI: oh can I have a go first? It could be argued that obtaining permission is very similar to requesting information from the hearer; however, instances of Can-PERSON-PROCESS? asking for permission can be used with please while this is not possible for real requests for information (Huddleston & Pullum, 2006, p. 865). 3. Polite requests for actions/ directives Traditionally defined as indirect speech acts, this category subsumes all cases in which the speaker (politely) asks the hearer to perform an action: (5) (file 4-08-06, Thomas is playing with his cassette recorder) MOT: can you turn the music down please? Like permission-seeking utterances, directives can also take please (Biber et al., 1999, p. 207). However, even within the group of utterances generally considered to be ‘indirect requests’, not all constructions can be used with please (for example, consider *Don’t you think you should please take out the garbage? , Sadock, 1974a, p. 90). 121 4. Test questions Test questions are used to ask for information obviously already known to the speaker. They are used frequently in child-directed speech (at least in most Western cultures; see, e.g., Bishop, 1974; Holzman, 1972; Olsen-Fulero & Conforti, 1983; Sinclair & van Gessel, 1990). In fact, they have been found to be even more common than sincere requests for missing information (see, e.g., Shatz, 1979, for English). Since test questions are not overtly different in form from genuine questions, early research suggested that children are not aware of the fact that their knowledge is being tested (Garvey, 1975; Shatz, 1979). It has been shown, however, that children from at least 2 year onwards interpret the same form as either a test 121 Both Thomas and his caretakers use please quite frequently with both permissionseeking formulas and directives, but never with constructions expressing any of the other functions (n = 59 for Thomas and 103 for his input; in each case please is present in 15% of all uses of Can-PERSON-PROCESS? expressing a request for permission or a directive). <?page no="183"?> 169 or a genuine question, depending on the context (interactive vs. onlooking situation; Grosse & Tomasello, 2012). (6) (file 3-09-02) MOT: erm do you remember when we watched that programme I told you (.) what men who chop down trees are called? MOT: can you remember what they’re called? CHI: timber. MOT: no. MOT: that is what they shout. 5. Directing attention (establishing/ maintaining joint attention) Establishing and maintaining joint attention is a crucial aspect of the interaction between caretaker and child and it has often been noted that many syntactic questions in child-directed speech are used to that end (e.g., Shatz, 1979; Sinclair & van Gessel, 1990; Werker & McLeod, 1989). In many cases, they are redundant to an ongoing activity and are embedded in a number of adjacent utterances produced by the caretaker (i.e., they are embedded in variation sets in the sense of Küntay & Slobin, 2002): (7) (file 02-01-17) MOT: it is rolling, isn’t it? MOT: can you see the pot rolling? MOT: look. 6. Other/ unclassifiable There are only three cases which could not be assigned to any of the five categories, either because the respective utterance fulfilled a different function or could not be classified on the basis of the discourse context available in both the transcript and the corresponding audio recording. One time, the mother rephrased a question produced by Thomas to ensure that this is what he wanted to ask her, i.e., the utterance was an echo-question to the child’s production: (8) (file 2-11-13) MOT: two daffodils? MOT: I didn’t hear her say anything like that. CHI: your see anything red green? MOT: can I see anything red or green? MOT: yes. MOT: I can see your bicycle, Thomas. <?page no="184"?> 170 One of the two cases which were categorized as unclassifiable is provided in example (9). (9) (file 4-11-00) MOT: don’t you think? MOT: anyway you did it thank you very much. CHI: can you hear Maria &barea [? ] Louise? 122 CHI: now you’re not playing. MOT: yes I’m just finishing off this. Table 8.4: Form-function mappings of Can-PERSON-PROCESS? -constructions information permission directives test directing attention other/ unclass. Total Can I INP 13 (4%) 241 (83%) 36 (12%) 1 (0%) 291 (100%) CHI 7 (2%) 285 (98%) 292 (100%) Can you INP 530 (27%) 345 (17%) 118 (6%) 1006 (50%) 1 (0%) 2000 (100%) CHI 27 (25%) 80 (73%) 1 (1%) 2 (2%) 110 (100%) Can other INP 35 (27%) 52 (40%) 25 (19%) 18 (14%) 1 (0%) 131 (100%) CHI 7 (18%) 22 (56%) 10 (26%) 39 (100%) Total INP 591 (24%) 293 (12%) 406 (17%) 136 (6%) 1007 (41%) 1 (0%) 2434 (100%) CHI 41 (9%) 307 (70%) 90 (20%) 1 (0%) 2 (0%) 441 (100%) The results of the functional coding procedure are summarized in Table 8.4. When comparing the different functions served by Can-I-PROCESS? and Can-you-PROCESS? , respectively, an interesting pattern emerges: While Canyou-PROCESS? is used by both mother and child to express a wider variety of functions, Can-I-PROCESS? (and also Can-other) is used almost exclusively as a permission-seeking formula. In other words, Can-I-PROCESS? constructions exhibit a rather straightforward and consistent mapping between form and function, leading to a higher salience of this particular mapping, which might in turn make learning this construction easier (the 122 Recall that in CHAT transcription, the phonological form of an unintelligible string is preceded by the ampersand and unclear cases are followed by [? ] (MacWhinney, 2000, p. 21). <?page no="185"?> 171 facilitative role of consistent form-function mappings has often been observed; see, e.g., Shatz, 1979; Slobin, 1973; Tomasello & Lieven, 2008). In particular, a functional analysis on a lexically even more specific level reveals that all instances of Can I have X? in the child’s (n = 132) and as well as the caretakers’ speech (n = 113) are used as ‘indirect’ speech acts, with permission-seeking again being the predominant function (e.g., Can I have another go, please? as opposed to directive uses such as Can I have my book back now, please? ). In this case, the very consistent mapping of the form Can I have X? to the function ‘request permission’ may also guide the child’s comprehension and use of this (and potentially also similar) construction as an ‘indirect’ speech act, suggesting that Can I have X? as a conventionalized indirect speech act can indeed be understood as a construction (as proposed by Stefanowitsch, 2003). Another reason related to functional aspects which may explain the prevalence of Can-I-PROCESS? in the child’s speech, besides the consistent form-function mapping, is its function proper. It has been noted that of course “mothers and children will select particular forms for production depending on their communicative goals” (Clancy, 1989, p. 345). Obviously, the ability to ask the caretaker for permission to do something is of great importance for the child. This is also mirrored in the fact that most previous studies mentioning different functions served by interrogatives in children’s speech explicitly state permission-seeking as one of them (and exclusively give examples of the type Can-I-PROCESS? , e.g., Dąbrowska, 2000; Grosse & Tomasello, 2012; Shatz, 1979). However, in order for this fact to be more than just a mere ad-hoc explanation of the observed pattern, it is necessary to elaborate on the role of functional salience within the usage-based model by relating it to the wellestablished notions of frequency and entrenchment. According to Langacker (1987, p. 59) “the degree of entrenchment of a given linguistic phenomenon, which indicates its cognitive significance, depends on its frequency [emphasis added]”. The problem with this view is that it focuses on (absolute) formal frequencies only, although it appears likely that it is “not frequency of use as such that determines entrenchment, but frequency of use with regard to a specific meaning or function, in comparison with alternative expressions of that meaning or function” (Stefanowitsch, 2003, p. 116). Along similar lines, it has been noted that certain prefabricated sequences (e.g., experience delays) are perceived as conventional ways of expressing a certain notion, even though their absolute frequency is not very high (Bybee, 2008, p. 231). The notion of what might be called the functional frequency of a construction has, so far, mainly been dealt with on the level of lexical categories. However, based on the definition of constructions as linguistic signs, we can expect the same kinds of relationships between lexemes as between more complex constructions and it has been shown that - analogous to seman- <?page no="186"?> 172 tic/ associative priming between words (e.g., Carrell, 1981) - there are also priming effects between semantically related constructions (Rouveret, 2012). Within lexicology, the relative frequency with which a lexical item is chosen in comparison with potential paradigmatic competitors has been suggested to be indicative of its onomasiological salience, which in turn influences its degree of entrenchment (Geeraerts, Grondelaers, & Bakema, 1994). Along similar lines, Hoffmann (2004) proposes the term ‘conceptual frequency’ for the “frequency with which a particular concept or speech event occurs” (Hoffmann, 2004, p. 190). Despite difficulties stemming from the fact that “the concept can be realized in a great number of different and formally unrelated ways” (Hoffmann, 2004, pp. 189-190), an attempt was made in the following analysis to establish the relative functional frequency of Can-I- PROCESS? as a permission-seeking construction. 8.4.1 Functional Frequency of Can-I as a Permission-Seeking Formula Concerning the present case, it was not deemed feasible to go through all 379 transcripts in order to find all other constructions which might be used to request permission. However, as a first step towards the assessment of the conceptual/ functional frequency of Can-I-PROCESS? as a permissionseeking formula, all yes-no questions were manually searched for constructions potentially expressing a request for permission and all these cases were subsequently evaluated on the basis of the discourse context. Of course, it cannot be ruled out that other structures might also be used to request permission and that consequently the relative functional frequencies of the constructions considered have been overestimated. However, based on both examples for permission-seeking constructions found in the literature (e.g,. Dąbrowska, 2000; Grosse & Tomasello, 2012; Shatz, 1979) and functional alternatives found in the corpus, it can reasonably be assumed that most cases of permission-seeking have been identified with the current procedure. Functional alternatives are most evident in false starts without retracing (10) and reformulations (in two subsequent utterances) (11), since this linguistic behavior suggests that the speaker considered the corresponding constructions to be alternative ways of expressing the same (or a very similar) communicative intention. (10) file 2-01-17 MOT: <ooh can I have> [/ / ] am I allowed a grape now I’m wearing a bib? (11) file 2-01-15 MOT: can I sit next to you? MOT: is that alright? <?page no="187"?> 173 Table 8.5: Functional frequencies of all permission-seeking constructions 123 Can I Can other Could I/ other May I/ other Is it/ that alright (if) Am I/ Is it allowed Other Total permission INP 241 (54%) 52 (12%) 41 (9%) 58 (13%) 23 (5%) 9 (2%) 25 (6%) 336 (100%) CHI 285 (85%) 25 (7%) 5 (1%) 7 (2%) 12 (4%) 334 (100%) The findings summarized in Table 8.5 clearly show that the mother makes use of quite a number of different constructions in order to ask for permission. In contrast to this, the child’s repertoire is still more limited. However, in both child-directed speech and Thomas’ own productions, Can I exhibits by far the highest relative frequency within all yes-no questions with the same (or a similar) discourse function (54% in the caretakers’ speech and 85% in the child’s production, respectively). We may therefore conclude that while Can-I-PROCESS? does not exhibit the highest formal frequency of all Can-PERSON-PROCESS? -constructions in the input (as was summarized in Tables 8.2 and 8.3), it does exhibit the highest functional frequency within the paradigm of all permission-seeking constructions, making it the most salient and thus likely option to be chosen for the expression of this function. Along similar lines, the functional salience and high degree of entrenchment of Can-I-PROCESS? (and possibly the lower-level formula Can I have X? ) may also preempt the acquisition (and use) of its functional competitors: While Can-I-PROCESS? is acquired at 2; 10.07 and used quite frequently from then onwards, both May-I-PROCESS? and Could-I-PROCESS? are acquired only later (at 2; 11.06, and 3; 08.05, respectively) and stay relatively infrequent. This ties in with the suggestion that “if the child has a very wellentrenched way of saying something, this may preempt the learning of new constructions” (Tomasello & Lieven, 2008, p. 171) and it has indeed be shown that an acquired structure may hinder the learning of a functionally very similar construction (Abbot-Smith & Behrens, 2006). 123 Analogous to Tables 8.2 - 8.4, the one me-for-I error produced by Thomas (And can me do that as well? ) was subsumed under Can other, just like the 25 cases in which the mother used Mummy instead of I to refer to herself. Additionally, three subjectless utterances including Can which were produced by Thomas were included in this category. The cases in which the mother used Mummy instead of I to refer to herself in utterances with Could and May are included in Could I/ other (2 cases) and May I/ other (11 cases), respectively. 9 out of the 12 utterances produced by Thomas in the category Other were of the type Please-I-PROCESS? , in 8 cases of which the modal can was indicated to be omitted by the transcriber. This again points towards the fact that Can-I- PROCESS? is taken to be the most likely way of asking for permission. <?page no="188"?> 174 8.5 Summary This chapter has explored the impact of functional factors on the learning of canonical questions. First, a correlation analysis exploring the relationship between 1. the relative frequencies of specific auxiliary-subject combinations found in the input and 2. Thomas’ age of acquisition and relative frequency of use of the same initial clusters was carried out. Based on the assumption that outliers of such analyses may provide particularly interesting cases for the exploration of functional factors (Steinkrauss, 2009, 2011), the subsequent study has focused on the formal and functional characteristics of a construction that is heavily overused by Thomas (Can-I-PROCESS? ). The functional analyses have revealed a number of potential explanations based on functional characteristics: First of all, Can-I-PROCESS? is used almost exclusively to ask for permission, while Can-you-PROCESS? is used for a wider range of functions. Can-I-PROCESS? thus exhibits a more consistent mapping of form and function, which might make its acquisition easier. In particular, all formulas of the type Can I have X? are used by both Thomas and his caretakers for this ‘indirect’ purpose, corroborating the claim that conventionalized indirect speech acts are indeed learned as fixed form-function pairings. Furthermore, the function expressed by Can-I-PROCESS? is very salient from the child’s point of view. Lastly, functional competitors (i.e., other yes-no questions used as requests for permission) are found far less frequently in the input, probably leading to a higher degree of entrenchment of Can-I-PROCESS? constructions relative to these functionally similar constructions. <?page no="189"?> 175 9 Summary and General Discussion This chapter reviews the major findings of the corpus-based studies presented in the empirical part of this book (Chapters 4-8). After a short summary of the results (Section 9.1), I will discuss how they relate to more general issues within usage-based, constructivist approaches to language development and use (Section 9.2). Finally, suggestions for further research (Section 9.3) and a brief, general conclusion will be offered (Section 9.4). 9.1 Summary Chapter 4 assessed the overall frequency of questions as well as the distribution of canonical vs. non-canonical yes-no questions in adult-to-adult speech on the basis of a subsample of the BNC containing spontaneous, spoken dialogue. While the overall frequency of questions was found to be not as high as that reported for child-directed speech (17% in comparison to 21- 44%), non-canonical questions were found to be very prevalent (constituting 45% of all yes-no questions). These findings were interpreted as lending support to the claim that non-canonical questions are part of the adult target repertoire and that they must consequently be learned by children acquiring English. In Chapter 5, the first of the four studies conducted on the basis of the high-density Thomas corpus (Lieven et al., 2009) was presented. Questions were found to be a lot more prevalent in the caretakers’ than Thomas’ speech (30% for the input and 6% for Thomas). Within the category of yes-no questions, non-canonical questions predominate in Thomas’s speech (55%), while their relative frequency found in the input is lower than that in the BNC. Still, over a third of all yes-no questions addressed to Thomas are noncanonical (37%). Subsequently, the issue of developmental changes in the relative frequency of non-canonical questions was addressed. It was shown that the relative frequency of non-canonical questions in the input stays relatively constant, additionally indicating that their use in child-directed speech is not a transitory phenomenon. In Thomas’ speech, non-canonical questions dominate initially. Their relative frequency gradually decreases over time, approaching the level found in his input at about 4; 0. Overall, the findings from Chapters 4 and 5 corroborate the claim made in previous research (Estigarribia 2007, 2010) that non-canonical yes-no questions must be regarded as (potentially) well-formed, and it is suggested that their emergence in child speech should therefore be investigated in more detail. <?page no="190"?> 176 Chapter 6 explored the potential facilitative role played by PREDs (Want some? ) and weak assertives (You want some? ) in the learning of canonical questions, thus extending earlier work by Estigarribia (2007, 2010). Moreover, the functional characteristics of weak assertives were analyzed in detail, with the aim of evaluating previous claims concerning their use in discourse. It was shown that Thomas uses item-specific PREDs first and that the corresponding weak assertives and canonical questions emerge only later on. It was proposed that the more complex, canonical questions may be learned by using an adjunction strategy which adds material to the left side of PREDs and weak assertives. This operation was formalized as a modified version of the juxtaposition operation suggested by Dąbrowska & Lieven (2005). Further evidence for this possibility was supplied by showing that weak assertives enter into paraphrase relations, i.e., they are often used alongside their more complex counterparts in child-directed speech, providing Thomas with explicit models for the adjunction operation. In the second part of the chapter, the discourse-functional properties of weak assertives were investigated in more detail. Besides being used as functional equivalents to canonical questions, weak assertive can also serve as echo-questions (A: I want some. B: You want some? ), lending support to the claim that their status is ambiguous between ‘reduced’ questions and rising declaratives. Chapter 7 focused on the acquisition and use of another class of noncanonical yes-no questions, so-called strong assertives (That’s a ball? ). Reevaluating and expanding findings from a study by Dąbrowska (2004, Chapter 9), the distribution of strong assertives vs. their canonical, ‘inverted’ counterparts was assessed. It was shown that the ‘inversion rate’ found in the input does hardly ever reach 100%, casting doubt on the assumption that strong assertives in child speech (necessarily) result from a ‘failure to invert’. Further analyses on the strong assertives It’s X? and That’s X? addressed the issue of whether (and in how far) the child’s use of these constructions mirrors the input not only on the formal but also on the functional level. Strong assertives are used by both the caretakers and Thomas to either echo a previous utterance (A: That’s rice. B: That’s rice? ) or to bias the interlocutor towards a positive reply to a tentative assumption (I can have that? ), suggesting that Thomas uses the construction in pragmatically appropriate ways. It was proposed that the discourse context should be considered before classifying children’s productions as ‘errors’. Chapter 8 addressed the role of discourse-functional factors in the acquisition of canonical yes-no questions by providing an in-depth analysis of different Can-PERSON-PROCESS? -constructions. First, it was shown that their emergence and frequency of use in Thomas’s speech cannot be (fully) accounted for by the relative frequency of these structures in his input. A detailed analysis of form-function mappings revealed that Thomas is influenced by the functional rather than formal frequencies found in the ambient <?page no="191"?> 177 language. Since the item-based construction Can I have X? was found to be consistently used to express requests (for permission or action), it was proposed that these ‘indirect’ uses are learned as form-function pairings from the input, corroborating the claim that they can be formalized as constructions whose (presumably ‘indirect’) function is directly associated with their form (as suggested by Stefanowitsch, 2003). 9.2 General Discussion On the basis of the BNC and a high-density corpus of a child learning English, the studies presented in this monograph have analyzed a wide range of formal as well as functional characteristics of a family of constructions (i.e., canonical as well as non-canonical yes-no questions), offering evidence on how these constructions are learned and how they are used in discourse. Additionally, suggestions on how non-canonical questions (particularly weak assertives; You want some? ) and conventionalized indirect speech acts (Can I have X? ) may be formalized within a usage-based, constructivist framework have been provided. In the following sections, I will discuss how these findings tie in with current debates within usage-based linguistics, focusing on the notion of grammaticality/ acceptability, its relationship to frequency and entrenchment, and the importance of discourse-functional properties of utterances and larger units. 9.2.1 The notion of grammaticality One of the basic tenets of the usage-based, constructivist approach is that “language structure emerges from language use” (Tomasello, 2003, p. 5). This entails that authentic usage-data should serve as the basis for all claims regarding characteristics of linguistic structures, including judgments concerning their acceptability/ grammaticality. Nevertheless, much research is still strongly influenced by rather traditional notions of well-formedness. For example, as has been shown, noncanonical yes-no questions are often discarded or treated as errors in usagebased studies on question acquisition (e.g., Dąbrowska, 2000). Since these structures occur frequently in adult-to-adult as well as child-directed speech (e.g., Estigarribia, 2007, 2010; Weber, 1993), it is highly likely that they are perceived as well-formed by the majority of speakers. 124 This assessment derives from the usage-based claim that the degree of acceptability of a 124 Even though no strong claims can be made on the basis of data from only one child, the fact that Thomas acquires the less complex weak assertives (You want some? ) earlier than their more complex, canonical counterparts casts doubt on the assumption that reduced questions (necessarily) result from a phonological reduction of the fuller forms (also see Estigarribia, 2010, p. 87, and Hopper, 2001, p. 112). <?page no="192"?> 178 structure is directly related to the degree to which native speakers perceive it as familiar (Bybee, 2007, p. 695), which, in turn, is a direct consequence of the frequency of occurrence of a particular linguistic item. In Cognitive Linguistics, this familiarity through repeated use has been conceptualized as entrenchment, i.e., the representational strength in the speaker’s mind (Langacker, 1987, pp. 59-60). Since this is not an all-or-nothing matter, acceptability is conceptualized as a graded phenomenon. What is more, acceptability judgments will also vary from speaker to speaker, since each human’s experience with language is different, resulting in individual mental grammars which exhibit a certain degree of heterogeneity (Langacker, 1987, p. 376). This claim is corroborated by experimental research which indicates that there is a great amount of variability within (different groups of) native speakers (e.g., Dąbrowska, 2008, 2010; Street & Dąbrowska, 2010). This variability is even larger in small children, since they have not had that much experience with language (yet) (Bannard, Lieven, & Tomasello, 2009). For this reason, it has been suggested that “changing representations within individual children (not just averages over different groups)” should be investigated (Lieven & Behrens, 2012, p. 228). By providing an in-depth case study based on high-density samples, the current study has made an important contribution in this respect. Still, even dense databases do not capture everything the child (over)hears. And even if we did ‘get it all’ (Roy, 2009), important issues concerning the assessment of entrenchment as a function of frequency of occurrence would still remain unsolved, since the relationship between frequency of occurrence and entrenchment is not entirely straightforward. 9.2.2 Frequency and entrenchment It has been claimed that even fully compositional patterns will be stored as independent constructions “as long as they occur with sufficient [emphasis added] frequency” (Goldberg, 2006, p. 5), but to date no estimate of what constitutes a sufficient number of occurrences has been made. In most developmental studies (including the present one), structures occurring with a relatively high frequency are focused on, in order to increase the likelihood that the respective constructions have indeed become entrenched in the child’s mental grammar. If ‘cutoff-points’ are proposed concerning which constructions are assumed to have unit status, rather conservative criteria are usually applied. For example, in her study on the acquisition of canonical questions, Dąbrowska defined a formula as an utterance (with or without a slot) that occurred at least five times, stating that “about 11 percent of the words in Naomi’s corpus have a frequency of four or less, and yet no one would argue that these rarer words are not retrieved from memory” (Dąbrowska 2000, p. 89). Therefore, the degree of formulaicity in child <?page no="193"?> 179 speech, which is often reported to be very high anyways, may still be underestimated. It is, however, highly unlikely that frequency of occurrence per se is the only factor determining the strength of entrenchment of a(n assumed) linguistic unit. Amongst others, “construction type, lexical and construction frequencies, sentence length and complexity” (Gurevich, Johnson, & Goldberg, 2010, pp. 72-73) have been suggested to play a role, but to date there are, to my knowledge, no studies explicitly addressing this possibility. Most importantly, it has been argued that since linguistic units are tightly associated with the usage-events they occur in, the strongest determinant of entrenchment will be “frequency of use with regard to a specific meaning or function, in comparison with alternative expressions of that meaning or function” (Schmid, 2010, p. 116). Up to now, the notion of ‘functional frequency’ has mainly been explored on the level of lexical items in adult language (Geeraerts, Grondelaers, & Bakema, 1994; Hoffmann, 2004). For the purposes of the current study, the concept was applied to the utterance-level (Can-PERSON-PROCESS? ), thus providing evidence for the fact that functional frequency may also be relevant for the description of larger linguistic units. Furthermore, the role played by discourse-function proper in the learning of constructions was explored, since developmental studies taking into account functional factors at all are surprisingly rare. 9.2.3 The importance of discourse (function) With a few notable exceptions (Abbot-Smith & Behrens, 2006; Cameron- Faulkner, 2012; Lieven et al.; 2009; Steinkrauss 2009, 2011), previous studies investigating language development from a usage-based, constructivist perspective have focused exclusively on form, in spite of the fact that both formal and functional characteristics of the linguistic constructions involved are thought to be of equal importance. This may be a result of methodological considerations: Discourse function is hard (if not impossible) to operationalize and coding cannot be done automatically, making the analysis of large amounts of data a very time-consuming endeavor. The studies presented here suggest that we should face these difficulties, since functional factors can provide valuable insights on a number of different levels. First of all, the functional frequency of particular constructions may account for cases in which the observed order of acquisition deviates from what would be expected based solely on relative input frequency (Chapter 8; also see Steinkrauss, 2009, 2011). In the case of Can-PERSON-PROCESS? , the consistency of the form-function mapping in Can-I-PROCESS? , the functional salience of permission-seeking constructions, and the lack of frequent functional competitors for Can-I-PROCESS? have also been found to play a <?page no="194"?> 180 role, suggesting that not only formal, but also functional characteristics of constructions are a driving force in language development. Secondly, results of the functional analyses indicate that it is important not to rely on a priori assumptions about the discourse functions expressed by particular (families of) constructions. For example, the assumed prototypical function of syntactic questions i.e., to elicit information not known to the speaker (e.g., Langacker, 1991, p. 505) was found only in the minority of cases within the subset of Can-PERSON-PROCESS? -constructions (Chapter 8). Thirdly, based on the form-function mappings found for Can-you- PROCESS? constructions, it seems very likely that certain ‘indirect’ speech acts are highly conventionalized and may thus indeed be formalized as constructions whose “illocutionary force is directly linked to their form” (Stefanowitsch, 2003, p. 106), again underlining the importance of studies investigating form-function mappings on a fine-grained level. Lastly, the analyses in Chapters 6 and 7 show that paraphrase relations (within so-called variation-sets) are a rich source of information for the child concerning the formal and functional relationships between different constructions and their constituents. In particular, they may aid the learning of more complex constructions and also make the relationships between different constructions salient. Due to theoretical and methodological considerations, most previous research has not taken this into account (but see, e.g., Fischer, 2008; Küntay & Slobin, 2002). 9.3 Suggestions for further research On a general level, I would like to suggest that future studies should aim at combining large quantitative analyses with more fine-grained (functional) analyses of utterances (as well as larger discourse units) in a meaningful way in order to shed further light on the emergence and use of formally and/ or functionally related constructions. Based on the findings of the current work, I now offer concrete suggestions for further research before providing a general conclusion. 9.3.1 Conventionalized indirect speech acts In Chapter 8, it was proposed that the ‘indirect’ meaning associated with the item-based construction Can I have X? is directly linked to its form and that these cases may thus be conceptualized as ‘direct’ after all (as proposed by Stefanowitsch, 2003). It remains to be seen, however, if/ in how far this also applies to other constructions in child(-directed) speech which have been regarded as conventionalized indirect speech acts, such as Would you mind <?page no="195"?> 181 X? , and Why don’t you X? . Fine-grained analyses of the learning and use of these constructions are needed in order to evaluate this possibility. 125 9.3.2 The role of prosody Concerning the majority of corpus-based investigations within Cognitive Linguistics, Glynn (2010) has noted that “the semantics encoded by phonological structures, such as prosody, are entirely absent” (Glynn, 2010, p. 25). The current work is no exception in this respect, since certain prosodic characteristics of the investigated constructions were assumed rather than assessed on the basis of the empirical data. For example, based on the transcription guidelines for both the BNC and the CHILDES-corpora, structures identified as non-canonical yes-no questions were described as being uttered with a final rise. Apart from the fact that the acquisition of prosody by children is of interest in its own right, analyses of the prosodic characteristics of non-canonical questions in child(-directed) speech and how these map to the different functions expressed may reveal in how far prosody can be regarded as an integral part of the formal side of (these) constructions (for a discussion of this option, see Imo, 2009). Future research should therefore address the contribution of prosodic factors. 9.3.3 Sampling Children’s Language Despite the fact that the developmental analyses in this book are based on a high-density corpus, the data obtained within the densest sampling period (2; 0.12 - 3; 2.12) still only comprise about 7-15% of Thomas’ speech (Lieven & Behrens, 2012, p. 227). For Thomas’ later language development from 3; 3.2 to 4; 11.20, this percentage is even lower (about 1-2%; Lieven & Behrens, 2012, p. 227). This amount of data was found to be sufficient for the fine-grained analysis of structures used with a high frequency and/ or from early on (e.g., weak assertives with want, the strong assertives That’s X? and It’s X? , and the canonical formula Can-PERSON-PROCESS? ). However, a detailed investigation of structures which Thomas only acquired later was not possible (for example, weak assertives with think and know). Even though it certainly is not feasible to aim at capturing 100% of a child’s linguistic experience (but see Roy, 2009), obtaining denser samples for (certain periods of) the later stages of language development (i.e., from 3; 3 onwards) may provide the opportunity to investigate the emergence of particular structures in more detail (for an overview of denser samples which are in preparation, see Lieven & Behrens, 2012, pp. 232-234). 125 For a first step in this direction, see Kania (2016). <?page no="196"?> 182 9.3.4 Representativeness of child-language corpora The Thomas corpus used here mainly captures interactions of the child with his primary caretaker, which is characteristic of most child-language corpora. However, even rather young children usually engage also with other interlocutors, for example in the context of daycare-centers or preschools. Moreover, due to technical considerations, the recordings are often confined to situations taking place inside the house during activities without much background noise, such as playing, book reading and eating, which again is true of most longitudinal corpora (Lieven & Behrens, 2012, p. 234). It has been suggested that these biases might make the child’s language skills seem more limited and stereotypical than they actually are (Eisenbeiss, 2009, p. 3). In order to arrive at a more realistic view of children’s communicative behavior, language samples obtained in all these different settings should be taken into account. This certainly is an aspect requiring further investigation. 9.4 Conclusion The studies presented here show how illuminating it can be to look at the richness of authentic usage data without making a priori assumptions about the acceptability of linguistic structures and the discourse function they fulfill. In particular, I have offered evidence for the fact that non-canonical yes-no questions are part of the adult target repertoire and have analyzed how they may be learned by children. Furthermore, I have shown that both non-canonical and canonical yes-no questions express a variety of discourse functions and have presented analyses exploring how these functional factors may influence the acquisition process. Taken together, the studies presented in this book offer a proposal of how non-canonical constructions and the discourse-functional properties of linguistic structures may be integrated into a usage-based, constructivist account of language acquisition and use. As has become evident in this last chapter, there are a lot of open issues awaiting further investigation. Overall, we are still far away from a complete understanding of how humans build and make use of their ‘constructicon’, i.e., their structured inventory of formally and functionally related constructions. <?page no="197"?> 183 References Abbot-Smith, K., & Behrens, H. (2006). How Known Constructions Influence the Acquisition of Other Constructions: The German Passive and Future Constructions. Cognitive Science, 30(6), 995-1026. doi: 10.1207/ s15516709cog0000_61 Abbot-Smith, K., & Tomasello, M. (2006). Exemplar-learning and schematization in a usage-based account of syntactic acquisition. The Linguistic Review, 23(3), 275-290. doi: 10.1515/ TLR.2006.011 Akhtar, N., Jipson, J., & Callanan, M. A. (2001). Learning Words through Overhearing. Child Development, 72(2), 416-430. doi: 10.1111/ 1467-8624.00287 Akhtar, N., & Tomasello, M. (1996). Two-year-olds learn words for absent objects and actions. British Journal of Developmental Psychology, 14(1), 79-93. doi: 10.1111/ j.2044- 835X.1996.tb00695.x Akmajian, A., Steele, S. M., & Wasow, T. (1979). The Category AUX in Universal Grammar. Linguistic Inquiry, 10(1), 1-64. Ambridge, B., Kidd, E., Rowland, C. F., & Theakston, A. L. (2015). The ubiquity of frequency effects in first language acquisition. Journal of child language, 42(2), 239- 273. doi: 10.1017/ S030500091400049X Ambridge, B., & Lieven, E. V. M. (2011). Child language acquisition: Contrasting theoretical approaches. Cambridge: Cambridge University Press. Ambridge, B., & Rowland, C. F. (2009). Predicting children's errors with negative questions: testing a schema-combination account. Cognitive Linguistics, 20(2), 225- 266. Ambridge, B., Rowland, C. F., & Pine, J. M. (2008). Is Structure Dependence an Innate Constraint? New Experimental Evidence From Children's Complex-Question Production. Cognitive Science: A Multidisciplinary Journal, 32(1), 222-255. doi: 10.1080/ 03640210701703766 Ambridge, B., Rowland, C. F., Theakston, A. L., & Tomasello, M. (2006). Comparing different accounts of inversion errors in children’s non-subject wh-questions: ‘What experimental data can tell us? ’. Journal of Child Language, 33, 519-557. Ambridge, B., Theakston, A. L., Lieven, E. V. M., & Tomasello, M. (2006). The distributed learning effect for children's acquisition of an abstract syntactic construction. Cognitive Development, 21(2), 174-193. doi: 10.1016/ j.cogdev.2005.09.003 Andersen, G. (2001). Pragmatic markers and sociolinguistic variation: A relevancetheoretic approach to the language of adolescents. Pragmatics & beyond: Vol. 84. Amsterdam, Great Britain: John Benjamins Publishing Company. Anderson, J. R. (1991). The adaptive nature of human categorization. Psychological Review, 98(3), 409-429. doi: 10.1037/ 0033-295X.98.3.409 Arnon, I., & Snider, N. (2010). More than words: Frequency effects for multi-word phrases. Journal of Memory and Language, 62(1), 67-82. doi: 10.1016/ j.jml.2009.09.005 Auer, P., & Pfänder, S. (Eds.). (2011). Linguae et litterae: Vol. 6. Constructions: emerging and emergent. Berlin: De Gruyter Mouton. Austin, J. L. (1962). How to do things with words: The William James lectures delivered at Harvard University in 1955. Cambridge, Mass.: Harvard University Press. <?page no="198"?> 184 Babyonyshev, M., Ganger, J., Pesetsky, D., & Wexler, K. (2001). The maturation of grammatical principles: Evidence from Russian unaccusatives. Linguistic Inquiry, 32, 1-44. Bakeman, R., & Adamson, L. B. (1984). Coordinating attention to people and objects in mother-infant and peer-infant interaction. Child Development, (55), 1278-1289. Bannard, C., & Lieven, E. V. M. (2009). Repetition and reuse in child language learning. In R. L. Corrigan, E. A. Moravcsik, H. Ouali, & K. M. Wheatley (Eds.), Typological studies in language: Vol. 83. Formulaic language. Volume 2. Acquisition, loss, psychological reality, and functional explanations: Acquisition, Loss, Psychological Reality (pp. 297-321). Amsterdam: Benjamins. Bannard, C., Lieven, E. V. M., & Tomasello, M. (2009). Modeling children's early grammatical knowledge. Proceedings of the National Academy of Sciences, 106(41), 17284-17289. doi: 10.1073/ pnas.0905638106 Bannard, C., & Matthews, D. (2008). Stored Word Sequences in Language Learning: The Effect of Familiarity on Children's Repetition of Four-Word Combinations. Psychological Science, 19(3), 241-248. doi: 10.1111/ j.1467-9280.2008.02075.x Bara, B. G., & Bucciarelli, M. (1998). Language in context: the emergence of pragmatic competence. In A. C. Quelhas & F. Pereira (Eds.), Cognition and context (pp. 317- 345). Lisboa: Instituto Superior de Psicologia Aplicada. Barcelona, A., & Valenzuela, J. (2011). An overview of cognitive linguistics. In M. Brdar, S. T. Gries, & M. Žic Fuchs (Eds.), Human cognitive processing: Vol. 32. Cognitive linguistics. Convergence and expansion (pp. 17-44). Amsterdam: John Benjamins Publishing Company. Barlow, M., & Kemmer, S. (Eds.). (2000). Usage Based Models of Grammar. Stanford, Calif.: CSLI Publications. Bartels, C. (1999). The intonation of English statements and questions, New York: Routledge. Bates, E. (1976). Language and context: The acquisition of pragmatics. New York: Academic Press. Bates, E. (1979). The Emergence of Symbols. New York: Academic Press. Bates, E. (1998). Construction grammar and its implications for child language research. Journal of Child Language, 25, 462-466. Behrens, H. (2006). The input-output relationship in first language acquisition. Language and Cognitive Processes, 21(1/ 2/ 3), 2-24. Behrens, H. (2009a). Konstruktionen im Spracherwerb. Zeitschrift für Germanistische Linguistik, 20, 427-444. Behrens, H. (2009c). Usage-based and emergentist approaches to language acquisition. Linguistics, 47(2), 383-411. Bellugi, U. (1965). The Development of Interrogative Structures in Children's Speech. In K. F. Riegel (Ed.), In the development of language functions (pp. 103-137). Ann Arbor, MI: University of Michigan. Bellugi, U. (1971). Simplification in children's language. In R. Huxley & E. Ingram (Eds.), Developmental sciences series. Language acquisition. Models and methods (pp. 95-119). London, New York: Academic Press. Bergen, B., & Chang, N. (2005). Embodied Construction Grammar in Simulation- Based Language Understanding. In J.-O. Östman & M. Fried (Eds.), Construction grammars: cognitive grounding and theoretical extensions (pp. 147-190). John Benjamins Publishing Company. <?page no="199"?> 185 Berglund Prytz, Y., Hoffmann, S., Lee, D., & Smith, N. (2002). BNCweb Manual: Glossary. Retrieved from http: / / rdues.bcu.ac.uk/ bncweb/ manual/ bncwebmanglossary.htm Berman, R. A. (1993). Marking verb transitivity in Hebrew-speaking children. Journal of Child Language, 20, 641-670. Bernicot, J., & Legros, S. (1987). Direct and indirect directives: What do young children understand? Journal of Experimental Child Psychology, 43(3), 346-358. doi: 10.1016/ 0022-0965(87)90012-9 Biber, D., Johansson, S., Leech, G., Conrad, S., & Finegan, E. (1999). Longman grammar of spoken and written English. Harlow, England: Longman. Bishop, D. V. M., Chan, J., Hartley, J., & Weir, F. (2004). When a nod is as good as a word: Form-function relationships between questions and their responses. Applied Psycholinguistics, 19(3), 415-432. Bloom, L., Merkin, S., & Wootten, J. (1982). Wh-Questions: Linguistic Factors that Contribute to the Sequence of Acquisition. Child Development, 53(4), 1084-1092. Bloom, L., Rocissano, L., & Hood, L. (1976). Adult-child discourse: Developmental interaction between information processing and linguistic knowledge. Cognitive Psychology, 8(4), 521-552. Bod, R. (2001). Sentence memory: storage vs. computation of frequent sentences. CUNY, Philadelphia. Retrieved from http: / / staff.science.uva.nl/ ~rens/ cuny2001.pdf Bolinger, D. (1957). Interrogative structures of American English: The direct question. Publication of the American Dialect Society: Vol. 28. Alabama: University of Alabama. Bolinger, D. (1978). Yes-no questions are not alternative questions. In H. Hiż (Ed.), Synthese language library: Vol. 1. Questions (pp. 87-105). Dordrecht: Reidel. Borer, H., & Wexler, K. (1987). The maturation of syntax. In T. Roeper & E. Williams (Eds.), Parameter Setting (pp. 23-172). Dordrecht: Reidel. Borer, H., & Wexler, K. (1992). Bi-unique relation and the maturation of grammatical principles. Natural Language & Linguistic Theory, 10, 147-189. Borsley, R. D., & Newmeyer, F. J. (2009). On Subject-Auxiliary Inversion and the notion “purely formal generalization”. Cognitive Linguistics, 20(1), 135-143. doi: 10.1515/ COGL.2009.007 Boyd, J. K., & Goldberg, A. E. (2012). Young children fail to fully generalize a novel argument structure construction when exposed to the same input as older learners. Journal of Child Language, 39(03), 457-481. doi: 10.1017/ S0305000911000079 Braine, M. D. S. (1963). The ontogeny of English phrase structure: the first phrase. Language, 39, 1-14. Braine, M. D. S. (1976). Children's first word combinations. Chicago: University of Chicago Press. Braine, M. D. S. (1988). Modeling the acquisition of linguistic structure. In Y. Levy, I. M. Schlesinger, & M. D. S. Braine (Eds.), Categories and processes in language acquisition (pp. 217-259). Hillsdale, NJ: Lawrence Erlbaum Associates. Braine, M. D. S. (1994). Is nativism sufficient? Journal of Child Language, 21, 9-31. Braine, M. D. S., & Brooks, P. J. (1995). Verb argument strucure and the problem of avoiding an overgeneral grammar. In M. Tomasello & W. E. Merriman (Eds.), Beyond names for things: young children's acquisition of verbs (pp. 352-376). Hillsdale, NJ: Lawrence Erlbaum Associates. <?page no="200"?> 186 Bresnan, J. (1978). A realistic transformational grammar. In M. Halle, J. Bresnan, & G. A. Miller (Eds.), MIT bicentennial studies: Vol. 3. Linguistic theory and psychological reality (2nd ed., pp. 1-59). Cambridge, Mass: MIT Press. Broen, P. A. (1972). The verbal environment of the Language-learning child. ASHA Monographs: Vol. 17. Washington, D.C.: American Speech and Hearing Association. Brooks, P. J., & Tomasello, M. (1999). How Children Constrain Their Argument Structure Constructions. Language, 75(4), 720-738. doi: 10.2307/ 417731 Brown, R. (1968). The Development of Wh Questions in Child Speech. Journal of Verbal Learning and Verbal Behavior, 7(2), 279-290. Brown, R. (1973). A first language: The early stages. Cambridge, MA: Harvard University Press. Brown, R. (1977). Introduction. In C. E. Snow & C. A. Ferguson (Eds.), Talking to children. Language input and acquisition (1st ed., pp. 1-27). Cambrigde: Cambridge University Press. Brown, R., & Hayes, J. R. (1970). Derivational Complexity and Order of Acquisition in Child Speech. In J. R. Hayes (Ed.), Cognition and the development of language (pp. 11-53). New York: Wiley. Bruner, J. S. (1983). Child's talk: Learning to use language (1st ed.). New York: W.W. Norton. Bucciarelli, M., Colle, L., & Bara, B. G. (2003). How children comprehend speech acts and communicative gestures. Journal of Pragmatics, 35(2), 207-241. doi: 10.1016/ S0378-2166(02)00099-1 Buium, N. (1976). Interrogative Types in Parental Speech to Language-Learning Children: a Linguistic Universal? Journal of Psycholinguistic Research, 5(2), 135-142. Burnard, L. (2007). Reference Guide for the British National Corpus (XML Edition). Retrieved from http: / / www.natcorp.ox.ac.uk/ docs/ URG/ cdifsp.html#cdifspu Butterworth, G., & Cochran, E. (1980). Towards a mechanism of joint visual attention in human infancy. International Journal of Behavioral Development, 3(3), 253-272. Bybee, J. L. (1985). Morphology: A study of the relation between meaning and form. Typological studies in language: Vol. 9. Amsterdam: John Benjamins Publishing Company. Bybee, J. L. (1995). Regular Morphology and the Lexicon. Language and Cognitive Processes, 10, 425-455. Bybee, J. L. (2004). Mechanisms of Change in Grammaticization: The role of frequency. In B. D. Joseph & R. D. Janda (Eds.), The Handbook of Historical Linguistics (pp. 603-623). Oxford, UK: Blackwell Publishers. Bybee, J. L. (2006). From Usage to Grammar: The Mind's Response to Repetition. Language, 82(4), 711-733. doi: 10.1353/ lan.2006.0186 Bybee, J. L. (2007). Adele Goldberg, Constructions at work: The nature of generalization in language. Oxford: Oxford University Press, 2006. Pp. 280. ISBN 0-19-9- 268517 and 0-19-9-268525 (pbk). Journal of Child Language, 34(03), 692-697. doi: 10.1017/ S0305000906007987 Bybee, J. L. (2008). Usage-based grammar and second language acquisition. In P. Robinson & N. C. Ellis (Eds.), Handbook of Cognitive Linguistics and Second Language Acquisition (pp. 216-236). Routledge. Bybee, J. L., & Hopper, P. J. (2001). Frequency and the emergence of linguistic structure. Typological studies in language: Vol. 45. Amsterdam, Great Britain: John Benjamins Publishing Company. <?page no="201"?> 187 Bybee, J. L., & Scheibman, J. (1999). The effect of usage on degrees of constituency: the reduction of don't in English. Linguistics, 37(4), 575-596. doi: 10.1515/ ling.37.4.575 Caines, A. (2012). You talking to me? Corpus and experimental data on the zeroauxiliary interrogative in British English. In S. T. Gries & D. Divjak (Eds.), Trends in linguistics : studies and monographs: 244: 1. Frequency effects in language learning and processing (pp. 177-206). Berlin, Boston: De Gruyter Mouton. Cairns, H. S., & Ryan Hsu, J. (1978). Who, why, when, and how: a developmental study. Journal of Child Language, 5(3), 477-488. Cameron-Faulkner, T. (2012). A functional account of verb use in the early stages of English multiword development. Journal of Child Language, 39(04), 885-897. doi: 10.1017/ S0305000911000328 Cameron-Faulkner, T., & Hickey, T. (2011). Form and function in Irish child directed speech. Cognitive Linguistics, 22(3), 569-594. doi: 10.1515/ cogl.2011.022 Cameron-Faulkner, T., Lieven, E. V. M., & Theakston, A. L. (2007). What part of no do children not understand? A usage-based account of multiword negation. Journal of Child Language, 34(02), 251-282. doi: 10.1017/ S0305000906007884 Cameron-Faulkner, T., Lieven, E. V. M., & Tomasello, M. (2003). A construction based analysis of child directed speech. Cognitive Science, 27(6), 843-873. Carpenter, M., Akhtar, N., & Tomasello, M. (1998). Fourteenthrough 18-month-old infants differentially imitate intentional and accidental actions. Infant Behavior and Development, 21(2), 315-330. doi: 10.1016/ S0163-6383(98)90009-1 Carpenter, M., Nagell, K., & Tomasello, M. (1998). Social cognition, joint attention, and communicative competence from 9 to 15 months of age. Monographs of the Society for Research in Child Development: 255 = 63,4. Chicago, Ill: University of Chicago Press. Carr, E. B. (1972). Da Kine Talk: From pidgin to standard english in Hawaii. Hawaii: University Press of Hawaii. Carrell, P. L. (1981). Children's understanding of indirect requests: comparing child and adult comprehension. Journal of Child Language, 8(02), 329-345. doi: 10.1017/ S0305000900003226 Casenhiser, D. M., & Goldberg, A. E. (2005). Fast mapping of a phrasal form and meaning. Developmental Science, 8(6), 500-508. Cazden, C. B. (1968). The acquisition of noun and verb inflections. Child Development, 39(2), 433-448. Chafe, W. (1998). Language and the flow of thought. In M. Tomasello (Ed.), The New Psychology of Language: Cognitive and Functional Approaches to Language Structure (pp. 93-111). Hillsdale, NJ: Lawrence Erlbaum Associates. Chen, R. (2013). Subject auxiliary inversion and linguistic generalization: Evidence for functional/ cognitive motivation in language. Cognitive Linguistics, 24(1), 1-32. doi: 10.1515/ cog-2013-0001 Chipere, N. (2003). Understanding complex sentences: Native speaker variation in syntactic competence. New York: Palgrave Macmillan. Chomsky, N. (1957). Syntactic Structures. London: Mouton. Chomsky, N. (1959). A Review of B. F. Skinner's Verbal Behavior. Language, 35(1), 26- 58. Chomsky, N. (1964). Current Issues in Linguistic Theory. The Hague: Mouton. Chomsky, N. (1965). Aspects of the Theory of Syntax. Cambridge: MIT Press. Chomsky, N. (1970). Remarks on nominalization. In R. Jacobs & P. Rosenbaum (Eds.), Reading in English Transformational Grammar (pp. 184-221). Waltham: Ginn. <?page no="202"?> 188 Chomsky, N. (1971). Problems of knowledge and freedom. New York: Vintage Books. Chomsky, N. (1973). Conditions on Transformations. In S. R. Anderson & P. Kiparsky (Eds.), A Festschrift for Morris Halle (pp. 232-286). New York: Holt Rinehart and Winston. Chomsky, N. (1980). Rules and representations. New York: Columbia University Press. Chomsky, N. (1981). Lectures on Government and Binding: The Pisa Lectures. Dordrecht: Foris. Chomsky, N. (1986). Knowledge of language. Berlin: Praeger. Chomsky, N. (1988). Language and problems of knowledge: the Managua lectures: MIT Press. Chomsky, N. (1989). Some notes on economy of derivation and representation. In I. Laka & A. Mahajan (Eds.), MIT Working Papers in Linguistics: Vol. 10. Functional Heads and Clause Structure (pp. 43-74). Cambridge, Mass.: MIT Press. Chomsky, N. (1995). The minimalist program. Cambridge, MA: MIT Press. Chomsky, N. (2000). New Horizons in the Study of Language and Mind. Cambridge, MA: Cambridge University Press. Chouinard, Michelle M.: Clark, Eve. (2003). Adult reformulations of child errors as negative evidence. Journal of Child Language, 30(3), 637-669. doi: 10.1017/ S0305000903005701 Clahsen, H. (1990). Constraints on parameter setting: A grammatical analysis of some acquisition stages in German child language. Language Acquisition, (1), 361-391. Clancy, P. M. (1989). Form and Function in the acquisition of Korean wh-questions. Journal of Child Language, 16(2), 323-347. Clark, E. V. (1978). Awareness of language: some evidence from what children say and do. In A. Sinclair, R. J. Jarvella, & W. J. M. Levelt (Eds.), The child's conception of language (pp. 17-43). Berlin: Springer. Clark, H. H. (1979). Responding to indirect speech acts. Cognitive Psychology, 11(4), 430-477. doi: 10.1016/ 0010-0285(79)90020-3 Clark, H. H., & Lucy, P. (1975). Understanding what is meant from what is said: A study in conversationally conveyed requests. Journal of Verbal Learning and Verbal Behavior, 14(1), 56-72. doi: 10.1016/ S0022-5371(75)80006-5 Clark, H. H., & Schunk, D. H. (1980). Polite responses to polite requests. Cognition, 8(2), 111-143. Clark, R. (1974). Performing without competence. Journal of Child Language, 1(1). doi: 10.1017/ S0305000900000040 Cohen, J. (1960). A Coefficient of Agreement for Nominal Scales. Educational and Psychological Measurement, 20(1), 37-46. doi: 10.1177/ 001316446002000104 Columbus, G. (2010). "Ah lovely stuff, eh? : invariant tag meanings and usage across three varieties of English. In S. T. Gries, S. Wulff, & M. Davies (Eds.), Language and computers: Vol. 71. Corpus-linguistic applications. Current studies, new directions (pp. 85-102). Amsterdam: Rodopi. Corkum, V., & Moore, C. (1995). Development of joint visual attention in infants. In C. Moore & P. J. Dunham (Eds.), Joint attention. Its origins and role in development (pp. 61-83). Hillsdale, NJ: Lawrence Erlbaum Associates. Cowper, E. A. (1992). A concise introduction to syntactic theory: The governmentbinding approach. Chicago: University of Chicago Press. Crain, S., & Pietroski, P. (2001). Nature, nurture and Universal Grammar. Linguistics and Philosophy, 24, 139-185. <?page no="203"?> 189 Crain, S., Thornton, R., & Khlentzos, D. (2009). The case of the missing generalizations. Cognitive Linguistics, 20(1), 145-155. doi: 10.1515/ COGL.2009.008 Croft, W. (2001). Radical Construction Grammar: Syntactic Theory in Typological Perspective. Chicago: Chicago University Press. Croft, W. (2005). Logical and typological arguments for Radical Construction Grammar. In J.-O. Östman & M. Fried (Eds.), Construction grammars: cognitive grounding and theoretical extensions (pp. 273-314). John Benjamins Publishing Company. Croft, W. (2009). Constructions and generalizations. Cognitive Linguistics, 20(1), 157- 165. doi: 10.1515/ COGL.2009.009 Cross, T. G. (1977). Mothers' speech adjustments: the contributions of selected child listener variables. In C. E. Snow & C. A. Ferguson (Eds.), Talking to children. Language input and acquisition (1st ed., pp. 151-188). Cambridge: Cambridge University Press. Crowdy, S. (1995). The BNC spoken corpus. In G. Leech, G. Myers, & J. Thomas (Eds.), Spoken English on computer. Transcription, mark-up and application (pp. 224- 234). Harlow: Longman. Culicover, P. W. (1976). Syntax. New York: Academic Press. Dąbrowska, E. (2000). From formula to schema: The acquisition of English questions. Cognitive Linguistics, 11(1-2), 83-102. Dąbrowska, E. (2004). Language, mind and brain: Some psychological and neurological constraints on theories of grammar. Edinburgh: Edinburgh University Press. Dąbrowska, E. (2008). Questions with long distance dependencies: A usage-based perspective. Cognitive Linguistics, 19(3), 391-425. Dąbrowska, E. (2010). Naive vs. expert intuitions: An empirical study of acceptability judgments. The Linguistic Review, 27, 1-23. Dąbrowska, E. (2012). Different speakers, different grammars: Individual differences in native language attainment. Linguistic Approaches to Bilingualism, 2(3), 219-253. doi: 10.1075/ lab.2.3.01dab Dąbrowska, E., & Lieven, E. V. M. (2005). Towards a lexically specific grammar of children's question constructions. Cognitive Linguistics, 16(3), 437-474. Davis, E. A. (1932). The Form and Function of Children's Questions. Child Development, 3(1), 57-74. Dennis, M., Sugar, J., & Whitaker, H. A. (1982). The Acquisition of Tag Questions. Child Development, 53(5), 1254. doi: 10.2307/ 1129014 Derwing, T. M., & Smyth, R. H. (1988). Auxiliary placement revisited. Journal of the Atlantic Provinces Linguistic Association, 10, 1-21. Diessel, H. (1997). Verb-first constructions in German. In M. H. Verspoor, K.-D. Lee, & E. Sweetser (Eds.), Amsterdam studies in the theory and history of linguistic science Ser. 4, Current issues in linguistic theory: Vol. 150. Lexical and syntactical constructions and the construction of meaning (pp. 51-68). Amsterdam: John Benjamins Publishing Company. Diessel, H. (2004). The acquisition of complex sentences. Cambridge: Cambridge University Press. Diessel, H. (2015). Usage-based construction grammar. In E. Dąbrowska & D. Divjak (Eds.), Handbook of Cognitive Linguistics (pp. 296-322). Berlin: De Gruyter Mouton. Dore, J. (1974). A Pragmatic Description of Early Language Development. Journal of Psycholinguistic Research, 3(4), 343-350. Eimas, P. D., Siqueland, E. R., Juszcyk, P., & Vigoroto, J. (1971). Speech perception in infants. Science, (209), 1140-1141. <?page no="204"?> 190 Eisenbeiss, S. (2009). Contrast is the Name of the Game: Contrast-Based Semi- Structured Elicitation Techniques for Studies on Children’s Language Acquisition (Essex Research Reports in Linguistics No. 57). Retrieved from http: / / www.essex.ac.uk/ linguistics/ publications/ errl/ errl57-7.pdf Ellis, N. C. (2012). What can we count in language, and what counts in language acquisition, cognition, and use? In S. T. Gries & D. Divjak (Eds.), Trends in linguistics : studies and monographs: 244: 1. Frequency effects in language learning and processing (pp. 7-34). Berlin, Boston: De Gruyter Mouton. Emonds, J. E. (1976). A transformational approach to English syntax: Root, structurepreserving, and local transformations. New York: Academic Press. Erreich, A. (1984). Learning how to ask: patterns of inversion in yes-no and whquestions. Journal of Child Language, 11(3), 579-592. Ervin-Tripp, S. M. (1970). Discourse agreement: how children answer questions. In J. R. Hayes (Ed.), Cognition and the development of language (pp. 79-107). New York: Wiley. Ervin-Tripp, S. M. (1977). Wait for me, roller-skate! In S. M. Ervin-Tripp & C. Mitchell-Kernan (Eds.), Child Discourse (pp. 165-188). New York: Academic Press. Ervin-Tripp, S. M., & Miller, W. (1977). Early discourse: Some questions about questions. In M. M. Lewis & L. A. Rosenblum (Eds.), Interaction, conversation, and the development of language (pp. 9-25). New York: Wiley. Estigarribia, B. (2007). Asking Questions. Language Variation and Language Acquisition (Dissertation). Stanford University, Stanford, Calif. Estigarribia, B. (2010). Facilitation by Variation: Right-to-Left Learning of English Yes/ No Questions. Cognitive Science, 34(1), 68-93. doi: 10.1111/ j.1551- 6709.2009. 01053.x Evans, V., & Green, M. (2006). Cognitive linguistics: An introduction. Edinburgh: Edinburgh University Press. Everett, D. (2005). Cultural Constraints on Grammar and Cognition in Pirah-: Another Look at the Design Features of Human Language. Current Anthropology, 46, 621-646. Everett, D. (2007). Cultural Constraints on Grammar in Pirah-: A Reply to Nevins, Pesetsky, and Rodrigues, Ms., Illinois State Univ. Ewing, G. (1982). Word order invariance and variability in 5 children’s three word utterances. In C. E. Johnson & C. L. Thew (Eds.), Proceedings of the Second International Congress for the Study of Child Language (p. 316). Washington, D.C: University Press of America. Fay, D. (1978). Transformations as mental operations: a reply to Kuczaj. Journal of Child Language, 5(01), 143-149. doi: 10.1017/ S0305000900001999 Fillmore, C. J. (1988). The mechanisms of construction grammar. Berkeley Linguistics Society, 14, 35-55. Fillmore, C. J. (1999). Inversion and constructional inheritance. In G. Webelhuth, J.-P. Koenig, & A. Kathol (Eds.), Studies in constraint-based lexicalism. Lexical and constructional aspects of linguistic explanation (pp. 113-128). Stanford, CA: CSLI Publications. Fillmore, C. J., Kay, P., & O'Connor, C. (1988). Regularity and idiomaticity in grammatical constructions: the case of let alone. Language, 64, 501-538. Fischer, K. (2008, September). Deriving relations between grammatical constructions from child-directed speech. Dritte Internationale Konferenz der Deutschen Gesellschaft <?page no="205"?> 191 für Kognitive Linguistik, Leipzig. Retrieved from http: / / www.uni-leipzig.de/ ~gcla08/ upload/ abstr13.pdf Fodor, J. A. (1983). The modularity of mind: An essay on faculty psychology. A Bradford book. Cambridge, Mass: MIT Press. Fodor, J. A., Bever, T. G., & Garrett, M. F. (1974). The psychology of language: An introduction to psycholinguistics and generative grammar. McGraw-Hill series in psychology. New York, NY: McGraw-Hill. Fodor, J. D. (2001). Setting syntactic parameters. In M. Baltin & C. Collins (Eds.), The Handbook of Contemporary Syntactic Theory (pp. 730-767). Oxford: Blackwell Publishers. Fodor, J. D. (2003). Evaluating models of parameter setting. Handout at LSA Institute Workshop on UG Principles and Input. Fodor, J. D., & Crain, S. (1987). Simplicity and Generality of Rules in Language Acquisition. In B. MacWhinney (Ed.), Mechanisms of Language Acquisition (pp. 35- 63). Hillsdale, NJ: Lawrence Erlbaum Associates. Freudenthal, D., Pine, J. M., & Gobet, F. (2002). Modelling the development of Dutch optional infinitives in MOSAIC. In W. D. Gray & C. D. Schunn (Eds.), Proceedings of the 24th Meeting of the Cognitive Science Society (pp. 328-333). Mahwah, NJ: Lawrence Erlbaum Associates. Fried, M., & Östman, J.-O. (2004). Construction Grammar: A thumbnail sketch. In M. Fried & J.-O. Östman (Eds.), Constructional approaches to language: Vol. 2. Construction grammar in a cross-language perspective (pp. 11-86). Amsterdam: John Benjamins Publishing Company. Furrow, D., Nelson, K., & Benedict, H. (1979). Mothers' speech to children and syntactic development: some simple relationships. Journal of Child Language, 6(3). doi: 10.1017/ S0305000900002464 Garnica, O. K. (1977). Some prosodic and paralinguistic features of speech to young children. In C. E. Snow & C. A. Ferguson (Eds.), Talking to children. Language input and acquisition ; papers from a conference sponsored by the Committee on Sociolinguistics of the Social Science Research Council (USA) [held 6-8 September 1974 … in Boston, Massachusetts] (1st ed., pp. 63-88). Cambrigde: Cambridge University Press. Garvey, C. (1975). Requests and responses in children's speech. Journal of Child Language, 2(01), 41-63. doi: 10.1017/ S030500090000088X Geeraerts, D., Grondelaers, S., & Bakema, P. (1994). The structure of lexical variation: Meaning, naming, and context. Cognitive linguistics research: Vol. 5. Berlin: De Gruyter Mouton. Geluykens, R. (1987). Intonation and Speech Act type: An Experimental Approach to Rising Intonation in Queclaratives. Journal of Pragmatics, 11, 483-494. Geluykens, R. (1988). On the Myth of rising intonation in polar questions. Journal of Pragmatics, 12, 467-485. Gerhand, S., & Barry, C. (1999). Age of acquisition, word frequency, and the role of phonology in the lexical decision task. Memory & cognition, 27(4), 592-602. Gibbs, R. W. (1980). Spilling the beans on understanding and memory for idioms in conversation. Memory & cognition, 8(2), 149-156. Gibbs, R. W. (1994). The poetics of mind: Figurative thought, language, and understanding. Cambridge: Cambridge University Press. Gibson, E., & Wexler, K. (1994). Triggers. Linguistic Inquiry, 25(3), 407-454. Ginzburg, J., & Sag, I. A. (2001). Interrogative investigations: The form, meaning, and use of English interrogatives. Stanford: CSLI Publications. <?page no="206"?> 192 Glynn, D. (2010). Corpus-driven Cognitive Semantics: An Introduction to the field. In D. Glynn & K. Fischer (Eds.), Cognitive linguistics research: Vol. 46. Quantitative methods in cognitive semantics. Corpus-driven approaches (pp. 1-42). Berlin: De Gruyter Mouton. Goldberg, A. E. (1991). It can't go down the chimney up: Paths and the English Resultative. In Proceedings of the Seventeenth Annual Meeting of the Berkeley Linguistics Society. General Session and Parasession on The Grammar of Event Structure (pp. 368-378). Goldberg, A. E. (1995). Constructions: A construction grammar approach to argument structure. Chicago: University of Chicago Press. Goldberg, A. E. (1998). Patterns of Experience in Patterns of Language. In M. Tomasello (Ed.), The New Psychology of Language: Cognitive and Functional Approaches to Language Structure (pp. 203-219). Hillsdale, NJ: Lawrence Erlbaum Associates. Goldberg, A. E. (2003). Constructions: a new theoretical approach to language. Trends in Cognitive Sciences, 7(5), 219-224. doi: 10.1016/ S1364-6613(03)00080-9 Goldberg, A. E. (2006). Constructions at work: the nature of generalization in language: Oxford University Press. Goldberg, A. E. (2009a). Constructions work. Cognitive Linguistics, 20(1), 201-224. doi: 10.1515/ COGL.2009.013 Goldberg, A. E. (2009b). The nature of generalization in language. Cognitive Linguistics, 20(1), 93-127. doi: 10.1515/ COGL.2009.005 Goldberg, A. E., Casenhiser, D. M., & Sethuraman, N. (2004). Learning argument structure generalizations. Cognitive Linguistics, 15, 289-316. Goldberg, A. E., & Del Guidice, A. (2005). Subject-auxiliary inversion: A natural category. The Linguistic Review, 22(2-4), 411-428. Gomez, R. L., & Gerken, L. (1999). Artificial grammar learning by 1-year-olds leads to specific and abstract knowledge. Cognition, (70), 109-135. Green, G. M. (1985). The Description of Inversions in Generalized Phrase Structure Grammar. Proceedings of the Annual Meeting of the Berkeley Linguistics Society, 117- 145. Gries, S. T. (2008). Corpus-based methods in alayses of second language acquisition data. In P. Robinson & N. C. Ellis (Eds.), Handbook of Cognitive Linguistics and Second Language Acquisition (pp. 406-431). Routledge. Gries, S. T., & Stefanowitsch, A. (2004). Extending collostructional analysis: A corpusbased perspective on `alternations'. International Journal of Corpus Linguistics, 9(1), 97-129. doi: 10.1075/ ijcl.9.1.06gri Grosse, G., & Tomasello, M. (2012). Two-year-old children differentiate test questions from genuine questions. Journal of Child Language, 39(01), 192-204. doi: 10.1017/ S0305000910000760 Guasti, M. T. (2002). Language Acquisition: The Growth of Grammar. Cambridge, MA: MIT Press. Guasti, M. T., Thornton, R., & Wexler, K. (1995). Negation in children's questions: The case of English. In D. MacLaughlin & S. McEwen (Eds.), Proceedings of the 19th boston university conference on language development (pp. 228-239). Boston, MA: Cascadilla Press. Gunlogson, C. (2003). True to form: Rising and falling declaratives as questions in English. Outstanding dissertations in linguistics. New York: Routledge. Gunter, R. (1963). Elliptical sentences in American English. Lingua, 12, 137-150. <?page no="207"?> 193 Gurevich, O., Johnson, M. A., & Goldberg, A. E. (2010). Incidental verbatim memory for language. Language and Cognition, 2(1), 45-78. doi: 10.1515/ langcog.2010.003 Haan, G. de. (1987). A theory-bound approach to the acquisition of verb placement in Dutch. In G. de Haan & W. Zonneveld (Eds.), Yearbook: Vol. 111. Formal parameters of generative grammar (pp. 15-30). Utrecht: University of Utrecht. Hare, M. L., & Goldberg, A. E. (1999). Structural priming: Purely syntactic? In M. Hahn & S. C. Stoness (Eds.), Proceedings of the twenty first annual conference of the Cognitive Science Society. August 19 - 21, 1999, Simon Fraser University, Vancouver, British Columbia (pp. 208-211). Mahwah, NJ: Lawrence Erlbaum Associates. Hart, B., & Risley, T. R. (1995). Meaningful differences in the everyday experience of young American children. Baltimore: Paul Brookes. Hart, B., & Risley, T. R. (1999). The social world of children learning to talk. Baltimore: Paul Brookes. Hauser, M. D., Chomsky, N., & Fitch, W. T. (2002). The Faculty of Language: What is it, who has it, and how did it evolve? Science, 298(5598), 1569-1579. Hedberg, N., & Sosa, J. M. (2002). The Prosody of Questions in Natural Discourse. In Proceedings of Speech Prosody 2002 (the First International Conference on Speech Prosody) (Ed.) (pp. 375-378). Aix-enProvence, France. Heine, L. (2011). Non-coordination-based ellipsis from a Construction Grammar perspective: The case of the coffee construction. Cognitive Linguistics, 22(1), 55-80. Hendrick, R. (1982). Reduced Questions and Their Theoretical Implications. Language, 58(4), 800-819. Hickey, T. (1993). Identifying formulas in first language acquisition. Journal of Child Language, 20(01), 27-41. doi: 10.1017/ S0305000900009107 Hoff-Ginsberg, E. (1990). Maternal speech and the child's development of syntax: a further look. Journal of child language, 17(1), 85-99. Hoffmann, S. (2004). Are low-frequency complex prepositions grammaticalized? On the limits of corpus data and the importance of intuition. In H. Lindquist & C. Mair (Eds.), Studies in corpus linguistics: Vol. 13. Corpus approaches to grammaticalization in English (pp. 171-210). Philadelphia: John Benjamins Publishing Company. Hoffmann, S. & Evert, S. (2007). Errors, Omissions and Inconsistencies in the XML- Version of the BNC. Retrieved from http: / / corpora.lancs.ac.uk/ BNCweb/ BNCerrors_and_inconsistencies.pdf Hoffmann, S., Evert, S., Smith, N., Lee, D., & Berglund Prytz, Y. (2008). Corpus linguistics with BNCweb a practical guide. English corpus linguistics: Vol. 6. Frankfurt am Main: Lang. Holzman, M. (1972). The Use of Interrogative Forms in the Verbal Interaction of Three Mothers and Their Children. Journal of Psycholinguistic Research, 1(4), 311-336. Hopper, P. J. (1998). Emergent Grammar. In M. Tomasello (Ed.), The New Psychology of Language: Cognitive and Functional Approaches to Language Structure (pp. 155-175). Hillsdale, NJ: Lawrence Erlbaum Associates. Hopper, P. J. (2001). Grammatical constructions and their discourse origins: prototype or family resemblance? In M. Pütz, S. Niemeier, & R. Dirven (Eds.), Cognitive linguistics research: Vol. 19,1. Theory and language acquisition (pp. 109-130). Berlin: De Gruyter Mouton. Horgan, D. (1977). How to answer questions when you've got nothing to say. Journal of Child Language, 5(1), 159-165. <?page no="208"?> 194 Huddleston, R. (1994). The contrast between interrogatives and questions. Journal of Linguistics, 30(2), 411-439. Huddleston, R., & Pullum, G. K. (Eds.). (2006). The Cambridge grammar of the English language (1st ed., 4th pr.). Cambridge: Cambridge University Press. Hudson, R. (1990). English Word Grammar. Oxford: Blackwell Publishers. Hurford, J. R. (1975). A child and the English question formation rule. Journal of Child Language, 2(02), 299-301. doi: 10.1017/ S0305000900001148 Huttenlocher, J., Haight, W., Bryk, A., Seltzer, M., & Lyons, T. (1991). Early vocabulary growth: Relation to language input and gender. Developmental Psychology, 27(2), 236-248. doi: 10.1037/ 0012-1649.27.2.236 Hyams, N. (1986). Language Acquisition and the Theory of Parameters. Dordrecht: Reidel. Hyams, N., Mateu, V., Putnam, M., Orfitelli, R., Rothman, J., & Sánchez, L. (2015). Parameters in language acquisition and language contact. In A. Fábregas, J. Mateu, & M. T. Putnam (Eds.), Contemporary linguistic parameters (pp. 353-374). New York: Bloomsbury. Imo, W. (2009). Die Grenzen von Konstruktionen: Versuch einer granularen Neubestimmung des Konstruktionsbegriffs der Construction Grammar. gidi Arbeitspapierreihe, (24). Ingram, D., & Tyack, D. L. (1979). Inversion of Subject NP and Aux in Children's Questions. Journal of Psycholinguistic Research, 8(4), 333-341. Jackendoff, R. (1977). X-bar-Syntax: A Study of Phrase Structure. Linguistic Inquiry Monograph 2. Cambridge, MA: MIT Press. James, S. L., & Seebach, M. A. (1982). The Pragmatic Function of Children's Questions. Journal of speech and hearing research, 25(1), 2-11. Jordens, P. (1990). The acquisition of verb placement in Dutch and German. Linguistics, 28, 1407-1448. Kania, U. (2013). ‘Can I ask you something? ': The influence of functional factors on the L1-acquisition of yes-no questions in English. Yearbook of the German Cognitive Linguistics Association, 1(1). doi: 10.1515/ gcla-2013-0007 Kania, U. (2016). Why Don’t You Just Learn it from the Input? : A Usage-based Corpus Study on the Acquisition of Conventionalized Indirect Speech Acts in English and German. In L. Ortega, A. Tyler, H. I. Park, & M. Uno (Eds.), The usage-based study of language learning and multilingualism (pp. 37-53). Washington D.C.: Georgetown University Press. Kidd, E. (2012). Individual differences in syntactic priming in language acquisition. Applied Psycholinguistics, 33(02), 393-418. doi: 10.1017/ S0142716411000415 Kirjavainen, M., & Theakston, A. (2011). Are infinitival to omission errors primed by prior discourse? The case of WANT constructions. Cognitive Linguistics, 22(4), 629- 657. doi: 10.1515/ COGL.2011.024 Kirjavainen, M., Theakston, A., Lieven, E., & Tomasello, M. (2009). `I want hold Postman Pat': An investigation into the acquisition of infinitival marker `to'. First Language, 29(3), 313-339. doi: 10.1177/ 0142723709105312 Klee, T. (1985). Role of inversion in children's question development. Journal of speech and hearing research, 28(2), 225-232. Klima, E. S., & Bellugi, U. (1966). Syntactic regularities in the speech of children. In J. Lyons & R. J. Wales (Eds.), Psycholinguistics papers. The proceedings of the 1966 Edinburgh Conference (pp. 183-219). Edinburgh: Edinburgh Univ. Press. Kuczaj, S. A. (1976). Arguments against Hurford's ‘Aux Copying Rule’. Journal of Child Language, 3(03), 423-427. doi: 10.1017/ S0305000900007297 <?page no="209"?> 195 Kuczaj, S. A., & Brannick, N. (1979). Children's use of the wh question modal auxiliary placement rule. Journal of Experimental Child Psychology, 28(1), 43-67. doi: 10.1016/ 0022-0965(79)90101-2 Kuczaj, S. A., & Maratsos, M. P. (1975). What children can say before they will. Merrill-Palmer Quarterly, 21(89-111). Kuczaj, S. A., & Maratsos, M. P. (1983). Initial verbs of yes-no questions: A different kind of general grammatical category. Developmental Psychology, 19(3), 440-444. doi: 10.1037/ 0012-1649.19.3.440 Kuhl, P. K. (1987). The Special-Mechanisms Debate in Speech Perception: Nonhuman Species and Nonspeech Signals. In S. Harnad (Ed.), Categorical perception: The groundwork of Cognition. New York: Cambridge University Press. Küntay, A. C., & Slobin, D. I. (1995). Nouns and verbs in Turkish child directed speech. In D. MacLaughlin & S. McEwen (Eds.), Proceedings of the 19th boston university conference on language development (pp. 323-334). Boston, MA: Cascadilla Press. Küntay, A. C., & Slobin, D. I. (1996). Listening to a Turkish mother: Some puzzles for acquisition. In D. I. Slobin, J. Gerhardt, A. Kyratzis, & J. Guo (Eds.), Social interaction, social context, and language. Essays in honor of Susan Ervin-Tripp (pp. 265-286). Hillsdale, NJ: Lawrence Erlbaum Associates. Küntay, A. C., & Slobin, D. I. (2002). Putting interaction back into child language: Examples from Turkish. Psychology of Language and Communication, 6(1), 5- 14. Labov, W., & Labov, T. (1978). Learning the syntax of questions. In R. N. Campbell (Ed.), NATO conference series 3, Human factors: Vol. 4. Recent advances in the psychology of language. [proceedings of the … Stirling Psychology of Language Conference, held at the Univ. of Stirling, Scotland, June 21 - 26, 1976]. New York, NY: Plenum Press. Laguna, G. M. A. de. (1927). Speech: its functions and development. New Haven: Yale University Press. Lakoff, G. (1971). On generative semantics. In D. D. Steinberg & L. A. Jakobovits (Eds.), Semantics. An Interdisciplinary Reader in Philosophy, Linguistics and Psychology (pp. 232-296). Cambridge: Cambridge University Press. Lakoff, G. (1976). Toward generative semantics. In J. D. MacCawley (Ed.), Syntax and semantics: Vol. 7. Notes from the linguistic underground (pp. 43-61). New York: Academic Press. Lakoff, G. (1987). Women, fire, and dangerous things: What categories reveal about the mind. Chicago: University of Chicago Press. Lakoff, G. (1993). The contemporary theory of metaphor. In A. Ortony (Ed.), Metaphor and thought (2nd ed., pp. 202-251). Cambridge: Cambridge University Press. Lakoff, G., & Brugman, C. (1987). The semantics of aux-inversion and anaphora constraints. 13th Meeting of the Berkeley Linguistics Society, Berkeley. Lakoff, G., & Johnson, M. (1981). Metaphors we live by. Chicago: University of Chicago Press. Lambrecht, K. (1990). 'What me worry? ' Mad magazine sentences revisited. In K. Hall, J.-P. Koenig, M. Meacham, S. Reinman, & L. Sutton (Eds.), Proceedings of the Sixteenth Annual Meeting of the Berkeley Linguistics Society 1989-1990 (pp. 215-228). Berkeley. <?page no="210"?> 196 Lambrecht, K. (1994). Information structure and sentence form: Topic, focus and the mental representations of discourse referents. Cambridge studies in linguistics: Vol. 71. Cambridge: Cambridge University Press. Langacker, R. W. (1987). Foundations of Cognitive Grammar, Volume 1. Stanford, Calif.: Stanford University Press. Langacker, R. W. (1991). Foundations of Cognitive Grammar, Volume 2. Stanford, Calif.: Stanford University Press. Langacker, R. W. (2000). A dynamic usage-based model. In M. Barlow & S. Kemmer (Eds.), Usage Based Models of Grammar (pp. 1-63). Stanford, Calif.: CSLI Publications. Langacker, R. W. (2008). Cognitive grammar: a basic introduction: Oxford University Press. Langacker, R. W. (2009). Cognitive (Construction) Grammar. Cognitive Linguistics, 20(1), 167-176. doi: 10.1515/ COGL.2009.010 Lenneberg, E. H. (1967). Biological foundations of language. New York: Wiley. Lewis, J. D., & Elman, J. L. (2002). Learnability and the statistical structure of language: Poverty of stimulus arguments revisited. In B. Skarabela, S. Fish, & A. H.-J. Do (Eds.), Proceedings of the 26th Boston University Conference on Language Development (pp. 359-370). Cascadilla Press. Liebal, K., Behne, T., Carpenter, M., & Tomasello, M. (2009). Infants use shared experience to interpret pointing gestures. Developmental Science, 12(2), 264-271. doi: 10.1111/ j.1467-7687.2008.00758.x Lieven, E. V. M. (1994). Crosslinguistic and crosscultural aspects of language addressed to children. In C. Gallaway & B. J. Richards (Eds.), Input and interaction in language acquisition (pp. 56-73). Cambridge: Cambridge University Press. Lieven, E. V. M. (2008). Learning the English auxiliary: A usage-based approach. In H. Behrens (Ed.), Corpora in Language Acquisition Research. History, methods, perspectives (pp. 61-98). Amsterdam: John Benjamins Publishing Company. Lieven, E. V. M. (2009). Developing Constructions. Cognitive Linguistics, 20(1), 191- 199. Lieven, E. V. M. (2010). Language development in a cross-linguistic context. In M. Kail & M. Hickmann (Eds.), Language acquisition across linguistic and cognitive systems (pp. 91-108). Amsterdam ; Philadelphia: John Benjamins Publishing Company. Lieven, E. (2014). First language development: a usage-based perspective on past and current research. Journal of child language, 41 Suppl 1, 48-63. doi: 10.1017/ S0305000914000282 Lieven, E. V. M., & Behrens, H. (2012). Dense Sampling. In E. Hoff (Ed.), Guides to research methods in language and linguistics. Research methods in child language. A practical guide (pp. 226-239). Malden, Mass: Wiley-Blackwell. Lieven, E. V. M., Pine, J. M., & Baldwin, G. (1997). Lexically-based learning and early grammatical development. Journal of Child Language, 24(1), 187-219. Lieven, E. V. M., Salomo, D., & Tomasello, M. (2009). Two-year-old children's production of multiword utterances: A usage-based analysis. Cognitive Linguistics, 20(3), 481-507. doi: 10.1515/ COGL.2009.022 Lightfoot, D. (1989). The child's trigger experience: Degree-0 learnability. Behavioral and Brain Sciences, 12(02), 321-375. doi: 10.1017/ S0140525X00048883 <?page no="211"?> 197 MacCarthy, D. A. (1976). The language development of the preschool child (2. reprint [1930]). Monograph series / Institute of Child Welfare: Vol. 4. Westport, Conn: Greenwood Press. MacWhinney, B. (1978). The Acquisition of Morphophonology. Monographs of the Society for Research in Child Development: Vol. 43. MacWhinney, B. (1987). The competition model. In B. MacWhinney (Ed.), Mechanisms of Language Acquisition (pp. 249-308). Hillsdale, NJ: Lawrence Erlbaum Associates. MacWhinney, B. (2000). The CHILDES Project: Tools for Analyzing Talk (3rd edn.). Mahwah, NJ: Lawrence Erlbaum Associates. MacWhinney, B. (2000b). The CHILDES Project: The database (3rd edn.). Mahwah, NJ: Lawrence Erlbaum Associates. MacWhinney, B. (2008). Enriching CHILDES for morphosyntactic analysis. In H. Behrens (Ed.), Corpora in Language Acquisition Research. History, methods, perspectives (pp. 165-198). Amsterdam: John Benjamins Publishing Company. MacWhinney, B. (2010). A tale of two paradigms. In M. Kail & M. Hickmann (Eds.), Language acquisition across linguistic and cognitive systems (pp. 17-32). Amsterdam ; Philadelphia: John Benjamins Publishing Company. Maratsos, M. P. (1978). New models in linguistics and language acquisition. In M. Halle, J. Bresnan, & G. A. Miller (Eds.), MIT bicentennial studies: Vol. 3. Linguistic theory and psychological reality (2nd ed., pp. 247-263). Cambridge, Mass: MIT Press. Maratsos, M. P., & Kuczaj, S. A. (1978). Against the transformationalist account: a simpler analysis of auxiliary overmarkings. Journal of Child Language, 5(02), 337- 345. doi: 10.1017/ S0305000900007510 Marcus, G. F. (1999). Do infants learn grammar with algebra or statistics? Science, (284), 436-437. Mayer, J. W., Erreich, A., & Valian, V. V. (1978). Transformations, basic operations and language acquisition. Cognition, 6(1), 1-13. doi: 10.1016/ 0010-0277(78)90006-9 Mazuka, R. (1996). How can a grammatical parameter be set before the first word? In J. L. Morgan & K. Demuth (Eds.), Signal to Syntax: Bootstrapping from Speech to Grammar in Early Acquisition (pp. 24-25). Mahwah, NJ: Lawrence Erlbaum Associates. Meltzoff, A. N. (1995). Understanding the intentions of others: Re-enactment of intended acts by 18-month-old children. Developmental Psychology, 31(5), 838-850. doi: 10.1037/ 0012-1649.31.5.838 Meltzoff, A. N., & Brooks, R. (2006) Eyes wide shut: The importance of eyes in infant gaze following and understanding other minds. In R. Flom, K. Lee, & D. Muir (Eds.), Gaze following: Its development and significance (pp. 217-241). Mahwah, NJ: Lawrence Erlbaum Associates. Meyer, W. J., & Shane, J. (1973). The form and function of children's questions. Journal of Genetic Psychology, 123(2), 285-296. Michaelis, L. A., & Lambrecht, K. (1996). The exclamative sentence type in English. In A. E. Goldberg (Ed.), Conceptual structure, discourse and language (pp. 375-398). Cambridge: Cambridge University Press. Miller, G. A. (1956). The magical number seven, plus or minus two: some limits on our capacity for processing information. Psychological Review, 63(2), 81-97. doi: 10.1037/ h0043158 Miller, G. A., & McKean, K. O. (1964). A chronometric study of some relations between sentences. Quarterly Journal of Experimental Psychology, 16(4), 297-308. doi: 10.1080/ 17470216408416385 <?page no="212"?> 198 Morgan, J. L. (1978). Two types of convention in indirect speech acts. In P. Cole (Ed.), Syntax and semantics: Vol. 9. Pragmatics (pp. 261-280). New York: Academic Press. Munro, A. (1979). Indirect Speech Acts are not strictly conventional. Linguistic Inquiry, 10(2), 353-356. Murphy, G. L. (2002). The big book of concepts. A Bradford book. Cambridge, Mass: MIT Press. Nakayama, M. (1987). Performance factors in subject-auxiliary inversion by children. Journal of Child Language, 14(1), 113-125. Nelson, K. (1985). Making sense: The acquisition of shared meaning. New York: Academic Press. Newmeyer, F. J. (1998). Language Form and Language Function Cambridge. Cambridge, MA: MIT Press. Newmeyer, F. J. (2000). Language form and language function (2nd ed.). Cambridge, Mass: MIT Press. Newport, E. L. (1977). Motherese: the speech of mothers to young children. In N. J. Castellan, D. B. Pisoni, & G. R. Potts (Eds.), Cognitive theory. vol. II (pp. 177-217). Erlbaum: Hillsdale. Newport, E. L., Gleitman, H., & Gleitman, L. R. (1977). Mother, I'd rather do it myself: some effects and non-effects of maternal speech style. In C. E. Snow & C. A. Ferguson (Eds.), Talking to children. Language input and acquisition (1st ed., pp. 109- 149). Cambrigde: Cambridge University Press. O'Connor, J. D., & Arnold, G. F. (1973). Intonation of colloquial English (2nd ed.). London: Longman. Olsen-Fulero, L., & Conforti, J. (1983). Child responsiveness to mother questions of varying type and presentation. Journal of Child Language, 10(3), 495-520. Panther, K.-U., & Thornburg, L. (1998). A cognitive approach to inferencing in conversation. Journal of Pragmatics, 30(6), 755-769. doi: 10.1016/ S0378- 2166(98)00028-9 Panther, K.-U., & Thornburg, L. L. (2003a). Introduction: On the nature of conceptual metonymy. In K.-U. Panther & L. L. Thornburg (Eds.), Pragmatics & beyond: Vol. 113. Metonymy and pragmatic inferencing (pp. 1-2). Philadelphia: John Benjamins Publishing Company. Panther, K.-U., & Thornburg, L. L. (Eds.). (2003b). Pragmatics & beyond: Vol. 113. Metonymy and pragmatic inferencing. Philadelphia: John Benjamins Publishing Company. Panther, K.-U., & Thornburg, L. L. (2004). The Role of Conceptual Metonymy in Meaning Construction. metaphorik.de, 6, 91-116. Parnell, M. M., Patterson, S. S., & Harding, M. A. (1984). Answers to Wh-questions: A Developmental Study. Journal of speech and hearing research, 27(2), 297-305. Pavlov, I. P. (1927). Conditioned Reflexes: An Investigation of the Physiological Activity of the Cerebral Cortex. London: Oxford University Press. Pawley, A., & Syder, F. H. (1983). Two puzzles for linguistic theory: nativelike selection and nativelike fluency. In J. C. Richards & R. W. Schmidt (Eds.), Applied linguistics and language study. Language and communication (pp. 191-226). London: Longman. Peters, A. M., & Menn, L. (1993). False starts and filler syllables: ways to learn grammatical morphemes. Language, 69, 742-777. Piaget, J. (1928). Judgment and reasoning in the child. Abingdon, UK: Taylor & Francis. Piaget, J. (1935/ 1952). The origins of intelligence in children. New York: Norton. <?page no="213"?> 199 Piaget, J. (1937/ 1954). The construction of reality in the child. New York: Basic Books. Pickering, M. J., & Branigan, H. P. (1998). The Representation of Verbs: Evidence from Syntactic Priming in Language Production. Journal of Memory and Language, 39(4), 633-651. doi: 10.1006/ jmla.1998.2592 Pine, J. M. (1994). The language of primary caregivers. In C. Gallaway & B. J. Richards (Eds.), Input and interaction in language acquisition (pp. 15-37). Cambridge: Cambridge University Press. Pine, J. M., & Lieven, E. V. M. (1993). Reanalysing rote-learned phrases: individual differences in the transition to multi-word speech. Journal of Child Language, 20(3), 551-571. doi: 10.1017/ S0305000900008473 Pine, J. M., Lieven, E. V. M., & Rowland, C. F. (1998). Comparing different models of the development of the English verb category. Linguistics, 36, 807-830. Pinker, S. (1984). Language Learnability and Language Development. Cambridge, MA: Harvard University Press. Pinker, S. (1989). Learnability and cognition: The acquisition of verb-argument structure. Cambridge, MA: Harvard University Press. Pinker, S. (1994). The Language Instinct. New York: Willam Morrow and Co. Pinker, S. (1999). Words and rules: The ingredients of language. New York: Perennial. Pizutto, E., & Caselli, M. C. (1992). The acquisition of Italian morphology. Journal of Child Language, 19, 491-557. Pollard, C. J., & Sag, I. A. (1994). Head-driven phrase structure grammar. Studies in contemporary linguistics. Stanford: Center for the Study of Language and Information. Pullum, G. K., & Scholz, B. C. (2002). Empirical assessment of stimulus poverty arguments. The Linguistic Review, 19(9-50). Quirk, R., Greenbaum, S., Leech, G., & Svartvik, J. (1985). Comprehensive grammar of the English language. London: Longman. Radford, A. (1988). Transformational grammar: a first course. Cambridge: Cambridge University Press. Radford, A. (1997). Syntactic theory and the structure of English: A minimalist approach. Cambridge textbooks in linguistics. Cambridge: Cambridge University Press. Reeder, K. (1980). The emergence of illocutionary skills. Journal of Child Language, 7(01), 13-28. doi: 10.1017/ S0305000900007005 Remick, H. (1976). Maternal speech to children during language acquisition. In W. von Raffler-Engel (Ed.), Neurolinguistics: Vol. 5. Baby talk and infant speech (pp. 223- 233). Amsterdam: Swets & Zeitlinger. Richards, B. J. (1990). Language development and individual differences: A study of auxiliary verb learning. Cambridge: Cambridge University Press. Richards, B. J. (1994). Child-directed speech and influences on language acquisition: Methodology and interpretation. In C. Gallaway & B. J. Richards (Eds.), Input and interaction in language acquisition (pp. 74-106). Cambridge: Cambridge University Press. Robinson, P., & Ellis, N. C. (Eds.). (2008). Handbook of Cognitive Linguistics and Second Language Acquisition: Routledge. Rosch, E. (1975). Cognitive representations of semantic categories. Journal of Experimental Psychology: General, (104), 192-233. <?page no="214"?> 200 Ross, B. H., & Makin, V. S. (1999). Prototype versus exemplar models. In R. J. Sternberg (Ed.), A Bradford book. The nature of cognition (pp. 205-241). Cambridge: MIT Press. Rouveret, A. (2012). VP ellipsis, phases and the syntax of morphology. Natural Language & Linguistic Theory, 30(3), 897-963. doi: 10.1007/ s11049-011-9151-3 Rowland, C. F. (2007). Explaining errors in children's questions. Cognition, 104(1), 106-134. Rowland, C. F., & Fletcher, S. L. (2006). The effect of sampling on estimates of lexical specificity and error rates. Journal of Child Language, 33(04), 859-877. doi: 10.1017/ S0305000906007537 Rowland, C. F., Fletcher, S. L., & Freudenthal, D. (2008). How big is enough? In H. Behrens (Ed.), Corpora in Language Acquisition Research. History, methods, perspectives (pp. 1-24). Amsterdam: John Benjamins Publishing Company. Rowland, C. F., & Pine, J. M. (2000). Subject-auxiliary inversion errors and whquestion acquisition: 'what children do know? '. Journal of Child Language, 27(1), 157-181. Rowland, C. F., Pine, J. M., Lieven, E. V. M., & Theakston, A. L. (2003). Determinants of acqusition order in wh-questions: re-evaluating the role of caregiver speech. Journal of Child Language, 30(3), 609-635. Rowland, C. F., Pine, J. M., Lieven, E. V. M., & Theakston, A. L. (2005). The Incidence of Error in Young Children's Wh-Questions. Journal of speech, language, and hearing research, 48(2), 384-404. Roy, D. (2009). New Horizons in the Study of Child Language Acquisition. Proceedings of Interspeech 2009. Retrieved from http: / / www.media.mit.edu/ cogmac/ publications/ Roy_interspeech_keynote.pdf Sachs, J. (1983). Talking about the there and the then: the emergence of displaced reference in parent-child discourse. In K. E. Nelson (Ed.), Children's language (Vol. 4). Hillsdale, NJ: Lawrence Erlbaum Associates. Sachs, J., Brown, R., & Salerno, R. (1976). Adult speech to children. In W. von Raffler- Engel (Ed.), Neurolinguistics: Vol. 5. Baby talk and infant speech (pp. 240-245). Amsterdam: Swets & Zeitlinger. Sadock, J. M. (1970). Whimperatives. In J. M. Sadock & A. Vanek (Eds.), Studies presented to Robert B. Lees by his students (pp. 223-238). Edmonton: Linguistic Research. Sadock, J. M. (1974a). Toward a linguistic theory of speech acts. New York: Academic Press. Sadock, J. M. (1974b). Towards a Linguistic Theory of Speech Acts. New York: Academic Press. Saffran, J. R., Aslin, R. N., & Newport, E. L. (1996). Statistical Learning by 8-Month- Old Infants. Science, (274), 1926-1928. Sampson, G. (1997). Educating Eve: The language instinct debate. London: Casell Academic. Sampson, G. (2005). The 'Language Instinct' Debate: (Revised Edition). London: Continuum International Publishing. Sandra, D., & Rice, S. (1995). Network analyses of prepositional meaning: Mirroring whose mind—the linguist’s or the language user’s? Cognitive Linguistics, 6(1), 89- 130. doi: 10.1515/ cogl.1995.6.1.89 <?page no="215"?> 201 Santelmann, L., Berk, S., Austin, J., Somashekar, S., & Lust, B. (2002). Continuity and development in the acquisition of inversion in yes/ no questions: dissociating movement and inflection. Journal of Child Language, 29, 813-842. Savage, C., Lieven, E. V. M., Theakston, A. L., & Tomasello, M. (2003). Testing the abstractness of children's linguistic representations: lexical and structural priming of syntactic constructions in young children. Developmental Science, 6(5), 557-567. doi: 10.1111/ 1467-7687.00312 Savic, S. (1975). Aspects of adult-child communication: the problem of question acquisition. Journal of Child Language, 2(2), 251-260. Scaife, M., & Bruner, J. S. (1975). The capacity for joint visual attention in the infant. Nature, 253(5489), 265-266. doi: 10.1038/ 253265a0 Schegloff, E. A. (1968). Sequencing in conversational openings. American Anthropologist, 70(6), 1075-1095. Schmid, H.-J. (2010). Does frequency in text instantiate entrenchment in the cognitive system? In D. Glynn & K. Fischer (Eds.), Cognitive linguistics research: Vol. 46. Quantitative methods in cognitive semantics. Corpus-driven approaches (pp. 101-133). Berlin: De Gruyter Mouton. Schmitt, N. (Ed.). (2004). Language learning and language teaching: Vol. 9. Formulaic sequences: Acquisition, processing and use. Amsterdam: John Benjamins Publishing Company. Schneider, W., & Bjorklund, D. F. (1998). Memory. In D. Kuhn & W. Damon (Eds.), Handbook of child psychology (5th ed., pp. 467-522). New York: Wiley. Searle, J. R. (1969). Speech Acts: An Essay in the Philosophy of Language. Cambridge: Cambridge University Press. Searle, J. R. (1975). Indirect speech acts. In P. Cole & J. L. Morgan (Eds.), Syntax and semantics: Vol. 3. Speech acts (pp. 59-82). New York: Academic Press. Searle, J. R. (1979). Metaphor. In A. Ortony (Ed.), Metaphor and Thought (1st ed., pp. 92-123). New York: Cambridge University Press. Shatz, M. (1978). Children's comprehension of their mothers' question-directives. Journal of Child Language, 5(1), 39-46. Shatz, M. (1979). How to Do Things by Asking: Form-Function Pairings in Mothers' Questions and Their Relation to Children's Responses. Child Development, 50(4), 1093-1099. Sinclair, A., & van Gessel, R. (1990). The form and function of questions in children's conversations. Journal of Pragmatics, 14(6), 923-944. Singleton, D. M. (2005). The Critical Period Hypothesis: A coat of many colours. International Review of Applied Linguistics, 43(4), 269-285. Skinner, B. F. (1957). Verbal Behavior. New York: Appleton-Century-Crofts. Slobin, D. I. (1973). Cognitive prerequisites for the development of grammar. In C. A. Ferguson & D. I. Slobin (Eds.), Studies in child language development (pp. 175-208). New York: Holt, Rinehart and Winston. Slobin, D. I. (1979). Psycholinguistics (2nd ed.). Glenview, Ill: Scott Foresman. Smith, M. E. (1933). The Influence of Age, Sex, and Situation on the Frequency, Form and Function of Questions Asked by Preschool Children. Child Development, 4(3), 201-213. doi: 10.2307/ 1125682 Smith, N., Hoffmann, S., & Rayson, P. (2007). Corpus Tools and Methods, Today and Tomorrow: Incorporating Linguists' Manual Annotations. Literary and Linguistic Computing, 23(2), 163-180. doi: 10.1093/ llc/ fqn004 <?page no="216"?> 202 Snow, C. E. (1972). Mothers' Speech to Children Learning Language. Child Development, 43(2), 549-565. Snyder, W. (2007). Child language: The parametric approach. Oxford ; New York: Oxford University Press. Sobin, N. (2010). Syntactic analysis. Malden Mass: Wiley-Blackwell. Stefanowitsch, A. (2003). A construction-based approach to indirect speech acts. In K.- U. Panther & L. L. Thornburg (Eds.), Pragmatics & beyond: Vol. 113. Metonymy and pragmatic inferencing (pp. 105-126). Philadelphia: John Benjamins Publishing Company. Stefanowitsch, A., & Gries, S. T. (2003). Collostructions: Investigating the interaction of words and constructions. International Journal of Corpus Linguistics, 8(2), 209- 243. doi: 10.1075/ ijcl.8.2.03ste Steinkrauss, R. G.-A. (2009). Frequency and function in WH question acquisition: A usage-based case study of German L1 acquisition. Groningen dissertation in linguistics: Vol. 75. Groningen: Selbstverl. Steinkrauss, R. G.-A. (2011). The interaction of function and input frequency in L1acquisition: The case of was…fur 'what kind of…' questions in German. In D. Schönefeld (Ed.), Converging evidence. Methodological and theoretical issues for linguistic research (pp. 249-272). Amsterdam: John Benjamins Publishing Company. Stemmer, N. (1981). A note on empiricism and structure-dependence. Journal of Child Language, 8(03), 649-663. doi: 10.1017/ S0305000900003494 Stenström, A.-B. (1984). Questions and responses in English conversation. Lund studies in English: Vol. 68. Malmö: CWK Gleerup. Stevenson, A. (1893). The speech of children. Science, (21), 118-120. Stirling, L., & Huddleston, R. (2006). Deixis and Anaphora. In R. Huddleston & G. K. Pullum (Eds.), The Cambridge grammar of the English language (1st ed., pp. 1449- 1564). Cambridge: Cambridge University Press. Street, J., & Dąbrowska, E. (2010). More individual differences in Language Attainment: How much do adult native speakers of English know about passives and quantifiers? Lingua, 120, 2080-2094. Stromswold, K. (1990). Learnability and the acquisition of auxiliaries. Unpublished Ph.D. dissertation. Sundberg, M. L., & Michael, J. (2001). The Benefits of Skinners Analysis of Verbal Behavior for Children with Autism. Behavior Modification, 25(5), 698-724. Talmy, L. (1988). Force Dynamics in Language and Cognition: Cognitive Science. Cognitive Science, 12, 49-100. Tamir, L. (1980). Interrogatives in Dialogue: Case Study of Mother and Child 16-19 Months. Journal of Psycholinguistic Research, 9(4), 407-424. Taylor, J. R. (2012). The mental corpus: How language is represented in the mind. Oxford: Oxford University Press. Theakston, A. L. (2004). The role of entrenchment in children’s and adults’ performance on grammaticality judgment tasks. Cognitive Development, 19(1), 15-34. doi: 10.1016/ j.cogdev.2003.08.001 Theakston, A. L., Lieven, E. V. M., Pine, J. M., & Rowland, C. F. (2001). The Role of Performance Limitations in the Acquisition of Verb-Argument Structure. Journal of Child Language, 28(1), 127-152. Theakston, A. L., Lieven, E. V. M., Pine, J. M., & Rowland, C. F. (2004). Semantic generality, input frequency and the acquisition of syntax. Journal of Child Language, 31(1), 61-99. doi: 10.1017/ S0305000903005956 <?page no="217"?> 203 Theakston, A. L., Lieven, E. V. M., Pine, J. M., & Rowland, C. F. (2005). The acquisition of auxiliary syntax: BE and HAVE. Cognitive Linguistics, 16(1), 247-277. doi: 10.1515/ cogl.2005.16.1.247 Thornburg, L. L., & Panther, K.-U. (1997). Speech act metonymies. In W.-A. Liebert, G. Redeker, & L. R. Waugh (Eds.), Current issues in Linguistic Theory: Vol. 151. Discourse and perspective in cognitive linguistics (pp. 205-219). Amsterdam; Philadelphia: John Benjamins Publishing Company. Thornton, R. (2008). Why continuity. Natural Language & Linguistic Theory, 26(1), 107- 146. Todd, P. (1982). Tagging after red herrings: evidence against the processing capacity explanation in child language. Journal of Child Language, 9(01), 99-114. doi: 10.1017/ S0305000900003640 Tomasello, M. (1988). : The role of joint attentional process in early language development. Language Sciences, 10(69-88). Tomasello, M. (1992). First Verbs: A Case Study of Early Grammatical Development. Cambridge: Cambridge University Press. Tomasello, M. (1995). Joint attention as social cognition. In C. Moore & P. J. Dunham (Eds.), Joint attention. Its origins and role in development (pp. 103-130). Hillsdale, NJ: Lawrence Erlbaum Associates. Tomasello, M. (1998a). Introduction: A cognitive-functional perspective on language structure. In M. Tomasello (Ed.), The New Psychology of Language: Cognitive and Functional Approaches to Language Structure (pp. vii-xxiii). Hillsdale, NJ: Lawrence Erlbaum Associates. Tomasello, M. (Ed.). (1998b). The New Psychology of Language: Cognitive and Functional Approaches to Language Structure. Hillsdale, NJ: Lawrence Erlbaum Associates. Tomasello, M. (1999). The cultural origins of human cognition. Cambridge, MA: Harvard University Press. Tomasello, M. (2000). First steps toward a usage-based theory of language acquisition. Cognitive Linguistics, 11(1/ 2), 61-82. Tomasello, M. (2003). Constructing a language: A usage-based theory of language acquisition. Cambridge, MA: Harvard University Press. Tomasello, M. (2005). Beyond formalities: The case of language acquisition. The Linguistic Review, 22, 183-197. Tomasello, M. (2006). Acquiring Linguistic Constructions. In D. Kuhn & R. S. Siegler (Eds.), Handbook of Child Psychology, vol. 2. (6th ed., pp. 255-298). New York: Wiley. Tomasello, M. (2007). Cognitive linguistics and first language acquisition. In D. Geeraerts & H. Cuyckens (Eds.), The Oxford handbook of cognitive linguistics (pp. 1092-1112). Oxford ; New York: Oxford University Press. Tomasello, M., & Barton, M. E. (1994). Learning words in nonostensive contexts. Developmental Psychology, 30(5), 639-650. doi: 10.1037/ 0012-1649.30.5.639 Tomasello, M., Carpenter, M., Call, J., Behne, T., & Moll, H. (2005). Understanding and sharing intentions: The origins of cultural cognition. Behavioral and Brain Sciences, 28(05), 675-691. doi: 10.1017/ S0140525X05000129 Tomasello, M., Kruger, A. C., & Ratner, H. H. (1993). Cultural learning. Behavioral and Brain Sciences, 16(03), 495-511. doi: 10.1017/ S0140525X0003123X <?page no="218"?> 204 Tomasello, M., & Lieven, E. V. M. (2008). Children's first language acquisition from a usage-based perspective. In P. Robinson & N. C. Ellis (Eds.), Handbook of Cognitive Linguistics and Second Language Acquisition (pp. 168-196). Routledge. Tomasello, M., & Stahl, D. (2004). Sampling children's spontaneous speech: how much is enough? Journal of Child Language, 31, 101-121. Tomasello, M., & Todd, J. (1983). Joint attention and lexical acquisition style. First Language, 4, 197-212. Trudgill, P. (1978). Introduction: sociolinguistics and sociolinguistics. In P. Trudgill (Ed.), Sociolinguistic patterns in British English (pp. 1-18). London: Edward Arnold. Tsui, A. (1992). A functional description of questions. In M. Coulthard (Ed.), Advances in spoken discourse analysis (pp. 89-110). London: Routledge. Tyack, D. L., & Ingram, D. (1977). Children's production and comprehension of questions. Journal of Child Language, 4(2), 211-224. Vaidyanathan, R. (1988). Development of forms and functions of interrogatives in children: a longitudinal study in Tamil. Journal of Child Language, 15(3), 533-549. Valian, V. V., & Casey, L. (2003). Young children's acquisition of wh-questions: the role of structured input. Journal of Child Language, 30(1), 117-143. Valian, V. V., Lasser, I., & Mandelbaum, D. (1992). Children's early questions. 17th Annual Boston University Conference on Language Development, Boston, MA. Valin, R. D. van. (1998). The Acquisition of WH-Questions and the Mechanisms of Language Acquisition. In M. Tomasello (Ed.), The New Psychology of Language: Cognitive and Functional Approaches to Language Structure (pp. 221-249). Hillsdale, NJ: Lawrence Erlbaum Associates. Valin, R. D. van. (2002). The development of subject-auxiliary inversion in English wh-questions: an alternative analysis. Journal of Child Language, 29(1), 161-175. Villiers, J. G. de. (1991). Why question? In T. L. Maxfield & B. Plunkett (Eds.), Papers in the acquisition of wh. proceedings of the UMASS Roundtable, May 1990. (pp. 155- 173). Amherst, MA: University of Massachusetts Occasional Papers. Villiers, J. G. de, & Villiers, P. A. de. (1985). The Acquisition of English. In D. I. Slobin (Ed.), Vol. 1: The Data. The crosslinguistic study of language acquisition (pp. 27-140). Hillsdale, NJ: Lawrence Erlbaum Associates. Vygotsky, L. S. (1978). Mind in society: The development of higher psychological processes. Cambridge: Harvard University Press. Watson, J. B., & Rayner, R. (1920). Conditioned emotional responses. Journal of Experimental Psychology, 3(1), 1-14. Weber, E. G. (1993). Varieties of questions in English conversation. Studies in discourse and grammar: Vol. 3. Amsterdam: John Benjamins Publishing Company. Wells, C. G. (1981). Learning through interaction: The study of language development. Cambridge: Cambridge University Press. Wells, G., & Ferrier, L. (1976). A study of child speech in its conversational context. In W. von Raffler-Engel (Ed.), Neurolinguistics: Vol. 5. Baby talk and infant speech (pp. 323-345). Amsterdam: Swets & Zeitlinger. Werker, J. F., & McLeod, P. J. (1989). Infant preference for both male and female infant-directed talk: a developmental study of attentional and affective responsiveness. Canadian journal of psychology, 43(2), 230-246. Westergaard, M. (2009). The acquisition of word order: Micro-cues, information structure, and economy. Linguistik aktuell / linguistics today: Vol. 145. Philadelphia, PA: John Benjamins Publishing Company. <?page no="219"?> 205 Wray, A. (2002). Formulaic language and the lexicon. Cambridge: Cambridge University Press. Wray, A. (2012). What Do We (Think We) Know About Formulaic Language? : An Evaluation of the Current State of Play. Annual Review of Applied Linguistics, 32, 231-254. doi: 10.1017/ S026719051200013X Zeschel, A. (2009). What's (in) a construction? : Complete inheritance vs. full-entry models. In V. Evans & S. Pourcel (Eds.), Human cognitive processing: Vol. 24. New directions in cognitive linguistics (pp. 185-200). Amsterdam, Philadelphia: John Benjamins Publishing Company. Zwicky, A. M., & Pullum, G. K. (1983). Deleting named morphemes. Lingua, 59(2-3), 155-175. doi: 10.1016/ 0024-3841(83)90061-X <?page no="220"?> Nancy Grimm, Michael Meyer, Laurenz Volkmann Teaching English bachelor-wissen 2015, XII, 341 Seiten €[D] 24,99 ISBN 978-3-8233-6831-1 Teaching English covers all of the major issues and current trends in language learning and teaching, such as the trends toward empiricism, constructivism, differentiation, learnerand output-orientation, intercultural learning, and the use of multimedia. This book bridges the gap between the suggestions of theoretical approaches to foreign language teaching and the practical needs of both the educators (regardless of the institutions they are teaching and the experiences they have gathered) as well as the students. It will help readers pro t from the materials and re ected practices for use in their own classrooms. And lastly, the book offers optimal preparation for exams in university courses and in teacher-training seminars. JETZT BESTELLEN! Narr Francke Attempto Verlag GmbH+Co. KG \ Dischingerweg 5 \ 72070 Tübingen \ Germany Tel. +49 (07071) 9797-0 \ Fax +49 (07071) 97 97-11 \ info@narr.de \ www.narr.de <?page no="221"?> This monograph offers a comprehensive account of the L1-acquisition and use of yes-no questions in English from a usage-based, construction grammar perspective. On the basis of the BNC and a high-density, longitudinal CHILDES corpus, the book explores two issues which have largely been neglected in previous research: 1. the prevalence of non-canonical questions (such as elliptical and declarative questions) in adult-to-adult as well as child(-directed) speech and the L1-acquisition of these structures. 2. The discourse-functional properties of both canonical and non-canonical yes-no questions, especially with regard to their in uence on the acquisition process. 36 Language Development ISBN 978-3-8233-8068-9
