Tokar Stress Variation in English Alexander Tokar Stress Variation in English This monograph is concerned with the question of why some English words have more than one stress pattern. E.g., 'overt vs. o'vert, 'pulsate vs. pul'sate, etc. It is argued that cases such as these are due to the fact that the morphological structure of one and the same English word can sometimes be analyzed in more than one way. Thus, 'overt is the stress pattern of the suffixation analysis over + -t, whereas o'vert is due to the prefixation analysis o- + -vert (cf. covert). Similarly, pulsate is simultaneously pulse + -ate (i.e., a suffixed derivative) and a back-derivative from pul'satance. „Tokar᾽ s approach in the use of both dictionary and corpus data holds promise of a scholarly breakthrough on the vital linguistic prosodic topic of English stress assignment of doublets and of stress assignment in general.“ (Irmengard Rauch, Professor of Germanic Linguistics at the University of California, Berkeley) Language in Performance LiP ISBN 978-3-8233-8180-8 Stress Variation in English 50 Edited by Rainer Schulze Advisory Board: Anita Fetzer (Augsburg), Thomas Herbst (Erlangen), Andreas Jucker (Zürich), Manfred Krug (Bamberg), Christian Mair (Freiburg i.Br.), Ute Römer (Atlanta, GA, USA), Andrea Sand (Trier), Hans-Jörg Schmid (München), Josef Schmied (Chemnitz) and Edgar W. Schneider (Regensburg) Alexander Tokar Stress Variation in English Bibliografische Information der Deutschen Nationalbibliothek. Die Deutsche Nationalbibliothek verzeichnet diese Publikation in der Deutschen Nationalbibliografie; detaillierte bibliografische Daten sind im Internet über http: / / dnb.dnb.de abrufbar. © 2017 · Narr Francke Attempto Verlag GmbH + Co. KG Dischingerweg 5 · D-72070 Tübingen Das Werk einschließlich aller seiner Teile ist urheberrechtlich geschützt. Jede Verwertung außerhalb der engen Grenzen des Urheberrechtsgesetzes ist ohne Zustimmung des Verlages unzulässig und strafbar. Das gilt insbesondere für Vervielfältigungen, Übersetzungen, Mikroverfilmungen und die Einspeicherung und Verarbeitung in elektronischen Systemen. Gedruckt auf säurefreiem und alterungsbeständigem Werkdruckpapier. Internet: www.narr.de E-Mail: info@narr.de Printed in Germany Satz: pagina GmbH, Tübingen ISSN 0939-9399 ISBN 978-3-8233-9180-7 V Contents Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 Abbreviations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 1.1 Scope of the study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 1.2 Previous studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 1.3 Structure of the book . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24 2 Theoretical preliminaries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 2.1 Stress vs. accent . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 2.2 Primary stress vs. secondary stress . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28 2.3 Stress / accent vs. no stress . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31 2.4 Stress placement across the globe . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37 3 Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44 3.1 Dictionary-based study: OED . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44 3.2 Other resources and tools . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47 3.3 YouTube-based study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58 4 Stress assignment in English . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68 4.1 Monosyllables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69 4.2 Disyllables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70 4.2.1 Stress preservation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71 4.2.2 Segmental length . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80 4.2.3 Suffix-like endings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87 4.2.4 Hiatus resolution and / or vowel elision . . . . . . . . . . . . . . . . . . . . . . . 89 4.2.5 Disyllables as parts of longer words . . . . . . . . . . . . . . . . . . . . . . . . . . 90 4.2.6 Rhythm, emphasis, and semantics . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91 4.2.7 Compounds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96 4.2.8 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97 4.3 Three and more syllables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97 4.3.1 Suffixed and back-derivatives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98 4.3.2 Prefixation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104 VI Contents 4.3.3 More on penultimate stress . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107 4.3.4 Stress non-preservation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111 4.3.5 Stress shifts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115 4.3.6 Segmental length in trisyllables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117 4.3.7 Suffix-like endings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120 4.3.8 Heavy ults in trisyllables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122 4.3.9 Foreignness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123 4. 3. 10 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125 4.4 Secondary stress . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126 4.5 Concatenations of words . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128 4.6 Across-varietal differences? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133 5 Case studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134 5.1 Stress variation in the OED . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134 5.1.1 Overall results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134 5.1.2 Degree-of-stress variation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139 5.1.3 Location-of-stress variation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145 5.1.3.1 Penultimate vs. antepenultimate stress . . . . . . . . . . . . . . . . . . . . 145 5.1.3.2 Final vs. penultimate stress . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152 5.1.3.3 Final stress vs. other stresses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158 5.1.3.4 Other categories . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160 5.1.4 Left-/ right-prominence vs. word stress . . . . . . . . . . . . . . . . . . . . . . 163 5.1.5 Stresslessness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 166 5.1.6 Hiatus resolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167 5.2 Stress variation in YouTube . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170 5.2.1 Overall results: YouTube vs. OED . . . . . . . . . . . . . . . . . . . . . . . . . . . 173 5.2.2 Adjacent words . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176 5.2.3 Vowel effect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181 5.2.4 More on stress non-preservation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184 5.2.5 More on disyllables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190 5.2.6 Final stress in trisyllables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193 5.2.7 Hiatus resolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195 5.2.8 Within-speaker variation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195 5.2.9 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197 VII 6 Concluding remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199 6.1 English as a Germanic language . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199 6.2 Future work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204 7 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211 7.1 Dictionaries / databases and corpora . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211 7.2 Software and online tools . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211 7.3 Literature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212 8 Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219 9 Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 235 1 Acknowledgements I am indebted to Professors Irmengard Rauch (University of California at Berkeley) and Jean-Michel Fournier (l’Université François-Rabelais, Tours), who reviewed this monograph for the Language in Performance series. It is difficult to say how many mistakes, inadequacies, inconsistencies, etc., this book would contain if not for, in particular, Prof. Fournier’s thorough critique of an earlier version of this manuscript. All remaining errors are, of course, entirely my own responsibility! My sincere gratitude goes also to the editor of the Language in Performance series, Professor Rainer Schulze (Leibniz Universität Hannover), whose comments on the structure of the book have made it more reader-friendly. To Prof. Schulze, my thanks are also due for adding this title to the Language in Performance series! I also thank 1) Professors Ulrich Busse and Alexander Brock (Martin-Luther-Universität Halle-Wittenberg), for inviting me to present parts of this research in their Forschungskolloquium in October 2016; 2) the UC Berkeley Library (represented by Lydia Petersen), for providing access to its vast resources, without which I would have been unable to write this book; 3) Samantha Tanner, whose native-speaker competence was of great help during the earlier stages of this project back in 2014; and 4) a number of people—colleagues, family, and friends—with whom I had the pleasure of discussing this topic during the many years I spent studying English stress. Some of these people include, in alphabetical order, Denis Balagurov, Sergei Danilov, Yuri Dyomin, Tibaut Houzanme, Natalia Peters, Karen Sullivan, and Viktor Treshchev. For their moral support and much more, I thank my parents Boris and Natalia Tokar. The book is dedicated to them! 2 Abbreviations Abbreviations 4 stressed pre-antepenult or pre-antepenultimate stress 5 stressed pre-pre-antepenult or pre-pre-antepenultimate stress 6 stressed pre-pre-pre-antepenult or pre-pre-pre-antepenultimate stress a stressed antepenult or antepenultimate stress BNC British National Corpus C consonant CDO Cambridge Dictionaries Online COCA Corpus of Contemporary American English d double-prominence/ double-prominent accentuation H heavy syllable H stressed heavy syllable L left-prominence/ left-prominent accentuation L light syllable L stressed light syllable LDOCE Longman Dictionary of Contemporary English MRC MRC Psycholinguistic Database MWO Merriam-Webster Online OD Oxford Dictionaries OED Oxford English Dictionary p stressed penult or penultimate stress r right-prominence/ right-prominent accentuation u stressed ult or final stress V short vowel V: long vowel/ diphthong/ triphthong χ 2 chi-squared test statistic 1.1 Scope of the study 3 1 Introduction This monograph is about English words with stress doublets, i.e., words in which stress is interchangeably placed upon more than one syllable without changing the meaning. For example, in the YouTube video whose unique identifier is -AnsohxXn QU (17. 09. 2016; to retrieve the video the identifier should be preceded by https: / / www.youtube.com/ watch? v=), an American English speaker can be heard stressing applicable initially in This name change will be more functionally ˈ applicable (00: 01: 25.899 --> 00: 01: 33.979), but in we’ve made the name change to make it more applicable (00: 03: 01.980 --> 00: 03: 04.880), the very same American English speaker can be heard using the stress pattern ap ˈ plicable . (According to Longman Dictionary of Contemporary English (henceforth LDOCE ), the variation / əˈplɪkəbəl/ vs. / ˈæplɪkəbəl/ occurs in both British and American English, but observe that in YouTube videos featuring the voices of British English speakers, the author heard only the stress pattern / əˈplɪ-/ .) A very similar case is adult , which in British and American English is interchangeably stressed / ˈædʌlt/ and / əˈdʌlt/ ( LDOCE ) (with, however, initial stress being the preferred stress pattern in British and Australian English vs. final stress being more frequently used (than initial stress) in American English); a You- Tube video in which a British English speaker can be heard vacillating between / ˈædʌlt/ and / əˈdʌlt/ is S5hXrgfwK8o (10. 04. 2017). 1.1 Scope of the study In addition to the obvious questions raised by variably-stressed words such as applicable and adult —1) What are the causes of stress variation exhibited by these and other English words with stress doublets? 2) Why do only some English words have stress doublets? 3) Do words with stress doublets prefer particular stress patterns and if so, why? —the present monograph will attempt to give a more precise answer to the general question of why English words (either with or without stress doublets) are stressed the way they are stressed. According to a popular view, Present-day English is a Latin-like language as far as its stress system is concerned (Hayes 1995: 181). Thus, since “[i]n the three hundred years that intervened between the Norman Conquest and Chaucer, the [English] language was inundated by Romance words” (Halle & Keyser 1971: 97), the stress rule of contemporary English is essentially the stress rule of Clas- 4 1 Introduction sical Latin: “Stress falls on the penultimate syllable if it contains a long vowel or is closed. […] Else stress is antepenultimate” (van der Hulst 2010a: 459). An illustration is the word paprika , which in British English, according to LDOCE , is interchangeably stressed / ˈpæprɪkə/ and / pəˈpriːkə/ . As one can notice, when stress in paprika is penultimate, the vowel in the penultimate syllable / ˈpriː/ , which bears stress, is long. When, by contrast, paprika is pronounced / ˈpæprɪkə/ , the vowel in the unstressed penult / prɪ/ is short. The stress patterns / ˈpæprɪkə/ and / pəˈpriːkə/ can thus both be said to be in accordance with the Latin Stress Rule. At the same time, however, in the case of the stress variation / ˈɛkskwɪzɪt/ vs. / ɪkˈskwɪzɪt/ (Oxford English Dictionary, henceforth OED ), the vowel in the penultimate syllable remains short irrespective of whether stress in exquisite is penultimate or antepenultimate. Similarly, it does not matter whether stalactite and stalagmite are stressed penor antepenultimately. In both / ˈstælæktaɪt/ vs. / stəˈlæktaɪt/ ( OED ) and / ˈstælæɡmaɪt/ vs. / stəˈlæɡmaɪt/ ( OED ), there is a short vowel in the penult that is followed by a coda consonant, i.e., / læk/ and / læɡ/ ; stress in the trisyllables stalactite and stalagmite is thus supposed to be penultimate in accordance with the Latin Stress Rule. A fairly similar case is necropsy , for which the OED gives segmentally identical British English transcriptions / ˈnɛkrɒpsi/ and / nɛˈkrɒpsi/ : As one can notice, these transcriptions differ from each other only with regard to the location of the stress symbol (ˈ). Similar examples can be found among disyllabic English words, which are not covered by the Latin Stress Rule, stating that “[i]n words with 2 or fewer syllables, primary stress occurs on the initial syllable” (StressTyp2 database), i.e., in contrast to a disyllabic English word, in which stress is either initial or final (e.g., / ˈædʌlt/ vs. / əˈdʌlt/ of adult ), in a disyllabic Latin word, stress can only be initial, i.e., Words with a heavy penultimate syllable receive penultimate stress, words with a light penult receive antepenultimate stress, and in all other cases where a word is too short to obey these laws, stress falls as far as possible to the left. (Hayes 1995: 50) According to van der Hulst (2010a: 445), in the English language “[p]rimary stress falls on the final syllable in nouns if the vowel is long, in verbs if the vowel is long or there are two closing consonants.” The diachronic basis of this assertion is the view, expressed in Halle & Keyser (1971: 99-101), that the stress system of contemporary English was shaped not only by Latin but also by (Old) French: The nonnative vocabulary of Chaucer consisted of two types of words, namely learned words largely of Latin origin and everyday words borrowed from Old French or Anglo-Norman. These two classes had different stress patterns. (Halle & Keyser 1971: 99) 1.1 Scope of the study 5 A convincing critique of this view can be found in Fournier (2007: 232), who argues that: French stress is not a central component of English stress, an analysis confirmed by history: most words of all lengths stressed on the final are relatively late borrowings, from the 17th century onwards. (Fournier 2007: 232; author’s italics) From a purely synchronic perspective, the view that final stress in English crucially depends upon the length of the vowel in the final syllable / the number of closing consonants when the vowel is short cannot be accepted because especially words with stress doublets provide too many counterexamples. For instance, for the verb migrate the OED gives segmentally identical British English transcriptions / mʌɪˈɡreɪt/ vs. / ˈmʌɪɡreɪt/ and the American English transcription / ˈmaɪˌɡreɪt/ , i.e., migrate is always pronounced with a diphthong in the ult (which counts as a long vowel), but stress in migrate is not always final in Present-day English. A similar case is the adjective overt , for which LDOCE gives segmentally identical British English transcriptions / ˈəʊvɜːt/ vs. / əʊˈvɜːt/ and American English transcriptions / ˈoʊvɜːrt/ vs. / oʊˈvɜːrt/ . The adjective under consideration is thus also always pronounced with a long vowel in the ult, even when stress in overt is non-final. The noun decade is interchangeably stressed / ˈdekeɪd/ and / deˈkeɪd/ ( LDOCE ), with both the stressed ult / ˈkeɪd/ of the latter and the unstressed ult / keɪd/ of the former containing a diphthong. Of the 268 variably-stressed disyllables in LDOCE (including words such as caffeine , which, according to the dictionary, is stressed only / ˈkæfiːn/ in British English vs. only / kæˈfiːn/ in American English), 173 (~64.55 %) are disyllables such as caffeine , in which stress differences are not accompanied by segmental differences involving the quality of the vowel in the ult. Thus, if the ult bearing stress contains a long vowel or diphthong, then also the unstressed ult of an initially-stressed alternative pronunciation likewise contains a long vowel or diphthong (e.g., / kæˈfiːn/ vs. / ˈkæfiːn/ of caffeine ). Similarly, if the ult bearing stress contains a short vowel, then (a qualitatively unreduced) short vowel also occurs in the ult of an initially-stressed alternative pronunciation. E.g., address is, according to LDOCE , only / əˈdres/ in British English, but it is / əˈdres/ vs. / ˈædres/ in American English. The ult / res/ , which contains a short vowel (being followed by only one coda consonant), thus occurs in both the finally-stressed pronunciation / əˈdres/ and the initially-stressed alternative / ˈædres/ . (Notice, however, that in YouTube videos featuring the spoken occurrences of address , initial stress was heard by the author only in environments such as IP address or street address , whereas in contexts such as State of the Union Address , Gettysburg Address , inaugural address , commencement address , etc., where address expresses the meaning “public speech,” stress in address was exclusively final. The varia- 6 1 Introduction tion / əˈdres/ vs. / ˈædres/ is thus, at least in part, a matter of semantics.) Anyway, the fact that disyllables such as caffeine and address constitute the majority of variably-stressed disyllables in English strongly suggests that the quality of the vowel in the ult plays a relatively insignificant role in the assignment of stress in disyllabic English words. Another important number is that in Oxford Dictionaries (henceforth OD ), there are 48,428 solidly-spelled polysyllables whose only stress pattern (given in the dictionary) is non-initial stress. E.g., inhibit is stressed only / ɪnˈhɪbɪt/ ( OD ), with stress in the word not falling upon its first syllable / ɪn/ . Of the 48,428 polysyllabic words in the OD such as inhibit , 28,944 (~59.77 %) have righthand strings such as, e.g., / -ˈhɪbɪt/ of / ɪnˈhɪbɪt/ , which occur in at least one other English word. E.g., apart from occurring in the transcription / ɪnˈhɪbɪt/ of inhibit , the righthand string / -ˈhɪbɪt/ , which begins with the primary stress symbol (ˈ), also occurs in the transcription / prə(ʊ)ˈhɪbɪt/ of prohibit ( OD ). (The percentage of such cases rises to ~66.55 % if, apart from counting only exact matches such as / -ˈhɪbɪt/ of / ɪnˈhɪbɪt/ and / prə(ʊ)ˈhɪbɪt/ , we ignore the onset of the primary-stressed syllable. E.g., / kəˈlɒtəmi/ of colotomy is the only transcription in the OD that contains the righthand string / -ˈlɒtəmi/ . At the same time, however, apart from / kəˈlɒtəmi/ of colotomy , the dictionary also has the transcription / ləˈbɒtəmi/ of lobotomy , which shares with / kəˈlɒtəmi/ the righthand string / -ˈ-ɒtəmi/ .) Notice further that of the 48,428 polysyllabic words such as inhibit , 35,525 (~73.36 %) have transcriptions such as / ɪnˈhɪbɪt/ , in which the righthand string that begins with the primary stress symbol (ˈ) is segmentally longer (i.e., contains more phonetic segments, with the duration symbol (: ) not counting as one of them) than the remaining lefthand string preceding it. Thus, for instance, the righthand string / -hɪbɪt/ of / ɪnˈhɪbɪt/ consists of four phonetic segments while the immediately preceding lefthand string / ɪn/ has only two. (If also the duration symbol (: ) is taken into consideration (e.g., in / ʃɪˈkɑːɡəʊ/ of Chicago ( OD ), the righthand string / -kɑːɡəʊ/ consists of six rather than five phonetic segments), 36,017 (~74.37 %) non-initially-stressed words in the OD dictionary can be said to have segmentally longer righthand strings such as / -kɑːɡəʊ/ of / ʃɪˈkɑːɡəʊ/ .) The connection between these numbers is the Old English Stress Rule: “Primary stress falls on the first syllable (of the root / stem)” (van der Hulst 2010a: 446). Assuming that this rule is still alive in contemporary English, we can argue that inhibit is stressed / ɪnˈhɪbɪt/ because the righthand string hibit counts morphologically as the root of the word. Thus, because apart from occurring in inhibit this righthand string also occurs in, e.g., prohibit , both inhibit and prohibit count for an English speaker as morphologically complex words consisting of the components in -/ pro and hibit . These components do not have meanings of their own, but as many authors suggest, components of which morphologically 1.1 Scope of the study 7 complex English words are made up are not necessarily traditional morphemes (i.e., meaning-bearing units). As, e.g., Aronoff (1976: 12) aptly observes in this connection: “What even vague sense does prefer share with confer and transfer ? or commit with remit and submit ? or receive with conceive and perceive ? or consume with presume and assume ? or reduce with induce and deduce ? None.” The (purely formal) segmentation in -/ pro - + hibit raises, however, the question of which of these components should count morphologically as the root of inhibit and prohibit , for the root is the only obligatory element in a word (e.g., Mel’čuk 2001: 69), i.e., a word may be affixless, but it may not be rootless: Any word contains at least one root. As Mel’čuk (2001: 69-79) explains, roots are different from affixes both semantically and formally. With regard to the former, roots are associated with more concrete meanings than those usually expressed by affixes. With regard to the latter, however, roots usually contain more segments than affixes (Mel’čuk 2001: 69). Needless to say, since the components in -/ pro and hibit are not associated with discernible meanings of their own, it is only the formal length-criterion that can be relied upon to answer the question of which of these two components counts in inhibit and prohibit as the root. Since the righthand component hibit contains more segments than the lefthand components in -/ pro -, the former is clearly more like a root whereas the latter are more like prefixes. Inhibit and prohibit are, therefore, both stressed / -ˈhɪbɪt/ , with the location of stress in these words being the root-prefix boundary location (and precisely because of this fact, it does not matter that the penult / ˈhɪ/ , which receives stress, ends in a short vowel and is thus not supposed to be stressed in accordance with the Latin Stress Rule). Likewise, given the numbers presented above, it can be conjectured that similar analyses apply to the majority of non-initially-stressed English words. E.g., colotomy is stressed / kəˈlɒtəmi/ not (or at least, not only) because the penult / tə/ ends in a short vowel—stress in / kəˈlɒtəmi/ should thus be antepenultimate in accordance with the Latin Stress Rule—but (also) because the segmentally longer righthand string otomy , which colotomy shares with lobotomy , counts morphologically as the root. Consider also the stress variation / ˈɒksᵻdʒᵻneɪt/ vs. / ɒkˈsɪdʒəneɪt/ of oxygenate ( OED ; only British English). To begin with, observe that in the view of many authors, suffixes in English have intrinsic properties with regard to stress; e.g., Kettemann (1988: 290) speaks of a diacritic feature that is contained in the lexicon entry dedicated to a particular English suffix. This diacritic feature is what triggers, in a particular suffixed derivative, a particular stress-related behavior. Thus, for instance, it can be stated that Anglo-Saxon suffixes never affect the stress of the words to which they are added […]. Of the others, some, like esce , take a stress themselves in addition 8 1 Introduction to the stress on the root-word […]; others, like ion , invariably throw the stress on to the syllable preceding them […]; while others again throw it on to the second syllable before them. (Kingdon 1949: 148) A challenge to views similar to this is, however, posed by “mixed suffixes,” which “have two or more distinct modes of operation” (Fudge 1984: 45). E.g., the suffix ate of / ˈɒksᵻdʒᵻneɪt/ is a stress-neutral suffix: Stress in the derived verb oxygenate falls upon the same syllable as in the base noun oxygen : / ˈɒksɪdʒ(ə)n/ (OD). Similarly, vaccinate is stressed / ˈvaksɪneɪt/ (OD) because vaccine is stressed / ˈvaksiːn/ ( OD ) whereas adsorbate is / adˈsɔːbeɪt/ or / adˈzɔːbeɪt/ ( OD ) because adsorb is / ədˈzɔːb/ or / ədˈsɔːb/ ( OD ); the suffix ate thus again acts as a stress-neutral suffix. In / ɒkˈsɪdʒəneɪt/ , by contrast, the very same suffix ate acts as a stress-shifting suffix, causing the derived verb oxygenate to have a different stress pattern than the base noun oxygen . What is interesting about the case of oxygenate is that the preferred stress pattern of this verb is not the preferred stress pattern of a similar variably-stressed ate -derivative hydrogenate , which is / ˈhaɪdrəʊdʒəneɪt/ vs. / haɪˈdrɒdʒəneɪt/ according to the OED. Of 69 native English speakers who were found to have pronounced oxygenate in YouTube videos, everybody (100 %) used initial stress in this verb, i.e., / ˈɒk-/ , but of 14 native English speakers who were found to have pronounced the verb hydrogenate , 10 (~71.43 %) used the antepenultimately-stressed version / haɪˈdrɒ-/ . In agreement with these findings, the OD gives for oxygenate only the initially-stressed transcription / ˈɒksɪdʒəneɪt/ , whereas in the case of hydrogenate the transcription / haɪˈdrɒdʒəneɪt/ is placed before the transcription / ˈhʌɪdrədʒəneɪt/ . This stress difference can only be accounted for if we assume that (for the majority of contemporary English speakers) hydrogenate is not a suffixed but prefixed derivative, with the segmentally longer righthand string rogenate , which hydrogenate shares with nitrogenate (vacillating, according to the OD , between the stress patterns / ˈnʌɪtrədʒᵻneɪt/ and / nʌɪˈtrɒdʒᵻneɪt/ ), counting morphologically as the root and hence receiving stress on its first syllable. Oxygenate is, by contrast, the only ygenate -word in the OD dictionary; English speakers have therefore no reasons to morphologically segment oxygenate into the prefix ox and the root ygenate (or into the prefix o and the root xygenate ). The morphological structure of oxygenate is thus oxygen + ate , i.e., the verb is a genuine ate -derivative and therefore, like other genuine ate -derivatives, preserves the stress of its base oxygen . Another English suffix that has more than one distinct mode of operation with respect to stress is al . Thus, for instance, while both the trisyllabic derivative personal and the disyllabic base person are stressed initially, / ˈpəːs(ə)n(ə)l/ and / ˈpəːs(ə)n/ ( OD ), in the derivative-base pair parental vs. parent only the latter is 1.1 Scope of the study 9 stressed on the first syllable: / pəˈrɛnt(ə)l/ vs. / ˈpɛːr(ə)nt/ ( OD ). It is tempting to say that the stress patterns of personal and parental are both in accordance with the Latin Stress Rule: In / ˈpəːs(ə)n(ə)l/ , stress is antepenultimate because the penult / sə/ ends in a short vowel, whereas in / pəˈrɛnt(ə)l/ , stress is penultimate because the penult / ˈrɛn/ is closed. Note, however, that also in the trisyllabic derivatives parentage and parenthood , the penultimate syllable is closed, but they are nonetheless stressed / ˈpɛːr(ə)ntɪdʒ/ and / ˈpɛːr(ə)nthʊd/ ( OD ), preserving the initial stress of the disyllabic base parent . An alternative explanation is that the morphological structure of parental is not the “correct” (i.e., semantics-based) parent + al but pa - + rental , i.e., the disyllabic righthand string rental , which occurs in English as a separate word, counts morphologically as the root and therefore attracts stress on to its first syllable. (Cf. / satɪsˈfakt(ə)ri/ of satisfactory and / ɒˈkʌlt/ of occult , whose righthand strings / -ˈfakt(ə)ri/ and / -ˈkʌlt/ also occur in the OD as the transcription of the initially-stressed trisyllable factory / the monosyllable cult . Cases such as satisfactory - factory or occult - cult constitute ~29.82 % of the 48,428 non-initially-stressed polysyllabic words in the OD dictionary, i.e., in 14,440 non-initially-stressed English words, the righthand string that begins with the primary-stressed syllable also occurs in the OD as the transcription of either an initially-stressed polysyllable or a monosyllable.) Likewise, simply because there is the word oral , mayoral is in American English stressed not only / ˈmeɪər(ə)l/ ( OED ), preserving the stress of mayor , but also / meɪˈɔr(ə)l/ ( OED ); pastoral is (also in American English) stressed not only / ˈpæstər(ə)l/ , preserving the stress of pastor , but also / pæsˈtɔr(ə)l/ (OED); electoral is, according to Merriam-Webster Online, not only e ˈ lectoral , preserving the stress of elect(or) , but also elecˈtoral . In the YouTube video 1oTFB9wdGl4 (14. 06. 2016), the stress pattern cliˈtoral is used by an American English speaker whereas in InJCUD0K2co (14. 06. 2016) a British English speaker can be heard saying cliˈtoral ; a YouTube video in which clitoral is interchangeably stressed ˈ clitoral and cliˈtoral by one and the same English speaker is _U_wKwVj8i8 (14. 06. 2016). These stress patterns strongly suggest that some Present-day English speakers analyze the ( t ) oral of these words as the root, i.e., these al -derivatives are for them not suffixed but prefixed derivatives. Another similar case is the variation / ˈkant(ə)n(ə)l/ vs. / kanˈtɒn(ə)l/ of cantonal (OD). In addition to regarding it as the inherited variation / ˈkantɒn/ vs. / kanˈtɒn/ of canton ( OD ), the stress pattern / kanˈtɒn(ə)l/ can also be seen as the preserved initial stress of the disyllabic tonal , which occurs in English as a separate word and may therefore (from a purely formal point of view) be perceived as the root of the trisyllabic cantonal . To reiterate, words with stress doublets challenge the popular assumption that stress in a polysyllabic English word should fall upon a particular syllable 10 1 Introduction when it exhibits a particular segmental makeup (e.g., a long vowel in the penult). Additionally, cases of stress variation among suffixed derivatives raise the question of why in the case of some English suffixes, there is apparently more than one diacritic feature triggering a particular stress-related behavior. It is thus the author’s hope that a systematic account of the phenomenon of stress variation, which will be presented in the following chapters of this monograph, will contribute to a better understanding of the general stress assignment principles by which contemporary English speakers abide when deciding where to place stress in a polysyllabic English word. 1.2 Previous studies With the exception of Berg’s (1999) study of stress differences between British and American English, none of the previous publications (on English stress) can be referred to as a systematic analysis of English words with stress doublets; the phenomenon of stress variation is only briefly touched upon in these publications as a side aspect of the more general issues pertaining to English stress. For example, Chomsky & Halle (1968: 73) observe that in the word umbilicus , which is interchangeably pronounced / ʌmbɪˈlaɪkəs/ and / ʌmˈbɪlɪkəs/ (OED), “we have penultimate stress if the penultimate vowel is taken to be tense in the underlying representation, and antepenultimate stress if the penultimate vowel is taken to be lax.” (Note that Chomsky & Halle (1968: 73) do not speak of long vs. short vowels because “[i]n few cases in American English as a whole is time length, or duration, of vowels significant—that is, used to distinguish from each other words otherwise alike” (Kenyon & Knott 1953[1944]: xxvi). E.g., the phonetic contrast between the “long” / iː/ of meal and the “short” / ɪ/ of mill is in American English by and large a matter of vowel quality rather than of duration.) A problem with this explanation is, however, the above mentioned fact that apart from cases such as / ʌmbɪˈlaɪkəs/ vs. / ʌmˈbɪlɪkəs/ , with stress being penultimate when the penult / ˈlaɪ/ is heavy vs. antepenultimate when the penult / lɪ/ is light (the Latin Stress Rule is thus in both cases abided by), there are also cases such as / ˈɛkskwɪzɪt/ vs. / ɪkˈskwɪzɪt/ , where stress differences do not correlate with segmental differences, i.e., stress in exquisite can be penultimate even when the vowel in the penult is phonetically realized as the short / ɪ/ . Another highly problematic explanation for the phenomenon of stress variation in English is Cruttenden’s (2008: 245) notion of rhythmic pressure. (Cruttenden (2008) is, however, the seventh edition of Gimson’s Pronunciation of English . An earlier edition of the same book (Gimson 1970: 232) also mentions rhythmic pressures as one of the causes of stress variation in contemporary 1.2 Previous studies 11 English.) What is meant by this is that “[i]n some words containing more than two syllables there appears to be a tendency to avoid a succession of weak syllables, especially if these have / ə/ or / ɪ/ ” (Cruttenden 2008: 245). For example, deficit is / ˈdɛfɪsɪt/ vs. / diˈfɪsɪt/ (Cruttenden 2008: 246), with the latter pronunciation representing in the view of the author a more rhythmic alternative to the former, i.e., the initially-stressed / ˈdɛfɪsɪt/ contains a sequence of two unstressed syllables both of which have the qualitatively reduced vowel / ɪ/ . Notice, however, that the OD dictionary has 3,466 initially-stressed words in which the first syllable, bearing stress, is followed by two unstressed syllables that have / ə/ or / ɪ/ in the nucleus position. E.g., in both / ˈaldʒɪbrə/ ( OD ) and / ˈanɪm(ə)l/ ( OD ), the stressed antepenults / ˈal/ and / ˈa/ are followed by the unstressed penult-ult sequences / dʒɪbrə/ and / nɪm(ə)l/ , which contain the vowels / ɪ/ and / ə/ . Of the 3,466 antepenultimately-stressed trisyllables such as algebra and animal , only 58 (~1.67 %) are, according to the OD , also pronounced by contemporary English speakers with penultimate stress. E.g., in addition to the above mentioned / ˈpaprɪkə/ vs. / pəˈpriːkə/ of paprika ( OD ), also the adjective integral is interchangeably stressed / ˈɪntɪɡr(ə)l/ and / ɪnˈtɛɡr(ə)l/ ( OD ). Observe now that apart from initially-stressed trisyllables such as algebra and animal , the OD also has 234 initially-stressed trisyllables such as, e.g., / ˈɔːtɑːki/ ( OD ). The unstressed penult-ult string in trisyllables such as autarky contains any long or short vowel with the exception of / ə/ or / ɪ/ . Of the 234 initially-stressed trisyllables such as autarky , five (~2.14 %) are, according to the OD , also pronounced by contemporary English speakers with penultimate stress. E.g., in addition to the above mentioned / ˈnɛkrɒpsi/ vs. / nɛˈkrɒpsi/ of necropsy ( OD ), also autopsy is stressed both / ˈɔːtɒpsi/ and / ɔːˈtɒpsi/ ( OD ), conversely is interchangeably / ˈkɒnvəːsli/ and / kənˈvəːsli/ ( OD ), covertly is both / ˈkəʊvəːtli/ and / kəʊˈvəːtli/ ( OD ), and patchouli is / ˈpatʃʊli/ and / pəˈtʃuːli/ ( OD ). Since the difference of 58 / 3,466 vs. five / 234 does not count as statistically significant— χ 2 (1) = 0.281, p = 0.5959—we can claim that trisyllables such as algebra and animal are in contemporary English not more frequently interchangeably pronounced with antepenultimate and penultimate stress than trisyllables such as autarky . Notice also that among the examples provided by Cruttenden (2008: 246) to substantiate his claim that “[h]esitancy and variation of accentual pattern occurring at the present time are the result of rhythmic […] pressures […]” (Cruttenden 2008: 245) is also the word acumen , which, according to LDOCE , is / ˈækjəmən/ vs. / əˈkjuːmən/ (cf. the OED , where acumen is only / ˈakjᵿmən/ as far as the British variety is concerned). As one can see, when stress in acumen is antepenultimate, the unstressed vowels in the ult and the penult may undergo qualitative reduction, yielding thereby the rhythmically unfortunate pronunciation / ˈækjəmən/ : This pronunciation contains a sequence of two weak 12 1 Introduction syllables, both of which have schwas. Note, however, that according to the OED, “[p]ronunciation with stress on the first syllable was first noted in the mid 20th cent.”; the original pronunciation of acumen was the penultimately-stressed / əˈkjuːmɪn/ , which etymologically is due to the Latin acūmen , in which stress is penultimate because the vowel in the penultimate syllable is long. Proceeding from Cruttenden’s (2008: 245) notion of rhythmic pressure, the pronunciation / əˈkjuːmɪn/ should be seen as a more fortunate pronunciation of acumen with respect to rhythm (compared to pronunciations in which the stress is antepenultimate), but it has nonetheless been abandoned in British English in favor of the presumably less rhythmic antepenultimately-stressed pronunciation / ˈakjᵿmən/ . Likewise, of 84 General American English speakers whose voices could be heard in YouTube videos containing the spoken occurrences of acumen , only four (~4.76 %) pronounced the word with a penultimate stress. Thus, we can say that not only in British but also in American English, the stress in acumen is close to becoming exclusively antepenultimate. It is fairly obvious, then, that if the variation between penultimate and antepenultimate stress in acumen did indeed have anything to do with rhythm in the sense of Cruttenden (2008: 245), we would now be observing a different tendency: More rhythmic pronunciations with stress on the middle syllable would be supplanting less rhythmic pronunciations with stress on the first syllable. Interestingly, in stark contrast to Cruttenden (2008: 245), Friederich (1967: 25) notes that “[i]n dreisilbigen Wörtern die Mittelsilbe zu betonen, ist ein Rhythmus, der dem Engländer nicht sonderlich liegt,” i.e., placing stress upon the middle syllable in a trisyllabic word is a rhythm that an English speaker does not particularly like. To substantiate this claim, Friederich (1967: 25) refers to stress shifts such as / kəmˈpɛnseɪt/ → / ˈkɒmpənseɪt/ ( OED ), which were undergone by many ate -trisyllables. E.g., apart from compensate , also concentrate , confiscate , contemplate , demonstrate , illustrate , infiltrate , inundate , etc. were originally pronounced with penultimate stress but have over the course of time abandoned this stress pattern in favor of antepenultimate stress ( OED ). At the same time, however, some ate -trisyllables still prefer penultimate to antepenultimate stress. E.g., for demarcate , elongate , and impregnate , LDOCE gives only penultimately-stressed American English transcriptions / dɪˈmɑːrkeɪt/ , / ɪˈlɒːŋɡeɪt/ , and / ɪmˈpreɡneɪt/ (whereas in British English, these verbs are stressed / ˈdiː-/ , / ˈiː-/ , and / ˈɪm-/ ). More generally, Present-day English would have relatively few penultimately-stressed trisyllables if this stress pattern were indeed dispreferred by English speakers (in trisyllabic words). LDOCE has, however, (no less than) 2,479 penultimately-stressed trisyllables (vs. 4,979 antepenultimately-stressed ones). Penultimate stress in a trisyllabic English word is thus without a doubt not a marginal stress pattern. 1.2 Previous studies 13 Another aspect pertaining to rhythm is the English Rhythm Rule. Because “adjacent stressed syllables make speech sound jerky” (Kingdon 1949: 149) and because some English words vacillate between final and non-final stress, English is believed to have “a rule that shifts a stress leftward when a stronger stress follows,” resulting in alternations such as thir ˈ teen ~ ˈ thirteen men (Hayes 1995: 18). It is acknowledged, though, that “the Rhythm Rule is optional, at least in certain contexts” (Hayes 1995: 18). For example, Langendoen (1975: 207), who is a native speaker of English, reports that for him, final stress in Detroit in the combination Detroit Lions is as acceptable as initial stress; similarly, in the combination Marlene Deetz , final stress in the modifier Marlene is interchangeable with initial stress. Halle & Vergnaud (1987: 271, f. 29) also mention that the word-internal application of the Rhythm Rule […] is restricted to lexical compounds. This explains why retraction is (almost) obligatory in Marcel Proust , but almost impossible in Marcel left : the former is a lexical compound, whereas the latter is formed in the syntax. The optionality of stress retraction in Marcel’s book would then be attributed to the ambiguous nature of the construction: it can be analyzed either as a lexical compound or as a syntactic construction. In the former case retraction is obligatory; in the latter it is impossible. The difference is brought out more clearly by such examples as We know about Marcel’s book, but not yet about Mary’s . In this sentence Marcel’s book must be pronounced without retraction because the pronominal relation that holds between book and the empty noun following Mary’s forces the syntactic analysis of the collocation. (Halle & Vergnaud 1987: 271, f. 29) Sometimes it is also added that the English Rhythm Rule “tends to apply in frequent words, e.g., ántìque bóok , and not in rare ones, e.g., àrcáne sort ” (Kraska-Szlenk & Żygis 2012: 327). Thus, of the finally-stressed modifiers antique and arcane , the former, which is more frequent than the latter, is likely to be initially-stressed as / ˈæntɪk/ (OED) rather than finally-stressed as / ænˈtiːk/ (OED) when the immediately following head word is either a monosyllable or an initially-stressed polysyllable; by contrast, in the less frequent arcane , stress is likely to stay final in a similar environment. Unfortunately, statements such as these, which describe when the English Rhythm Rule should apply and when it should not, are as a rule based not upon extensive corpus or experimental studies but upon authors’ own introspective judgments of how particular English words in which stress is normally final are supposed to be stressed when in an immediately following head word stress is initial. A fortunate exception is Mompéan (2014), who has recently studied the stress variation exhibited by the English cardinal numerals ending in teen and the corresponding ordinal numerals ending in teenth . For this purpose the 14 1 Introduction author has drawn a corpus of 1,263 newscasts from the BBC World Service Web site. These newscasts were recorded between 1999 and 2009 and feature mainly RP accents, i.e., British English speakers whose accent is Received Pronunciation. As the author reports: The analysis of the corpus provides empirical evidence to suggest an informed answer to the research question of the present study, namely how variable stress shift is in expressions involving teen numbers in a corpus of spoken RP speech. The analysis shows that out of the 343 potential cases of stress shift identified, 329 actually involved stress shift (95.9 %) whereas stress shift did not apply in the remaining 14 potential cases (4.1.%). This shows that stress shift is the rule, rather than the exception, in potential cases involving teen numbers as the first constituent of a compound or as the modifier of a head noun in a noun phrase. (Mompéan 2014: 155) Note, however, that these results do not automatically corroborate the reality of the rhythm rule in English. Thus, Mompéan (2014: 156) also reports that in his corpus “head nouns often designate fractions of 100 such as per cent (14 cases) or multiples of ten such as hundred , thousand , million and billion (31 potential cases), with stress shift applying in all potential cases except six.” If the so-called English Rhythm Rule is essentially nothing more than a stress clash avoidance strategy, why is stress also retracted in combinations such as thirteen per cent , in which a teen -numeral modifies a non-initially-stressed head? That is, since the pronunciation thir ˈ teen per ˈ cent would not involve a stress clash, why is thirteen per cent nonetheless as a rule pronounced ˈ thirteen per ˈ cent ? (A possible answer to this question is that also a one-syllable distance between two syllables bearing stress is not entirely unproblematic with regard to rhythm. Thus, “adjacent stresses are strongly avoided; stresses that are close but not adjacent are less strictly avoided; and at a certain distance (perhaps four syllables) the spacing becomes fully acceptable” (Hayes 1995: 372). E.g., retraction of stress in thirteen per cent occurs because in thir ˈ teen per ˈ cent only one unstressed syllable would separate two stressed ones. A similar example, discussed by Cruttenden (2008: 296), is the alternation West ˈ minster ~ ˈ Westminster Abbey , where there is also no more than one unstressed syllable separating two syllables bearing stress. An interesting case is the finally-stressed modifier Barack , / bəˈrɑːk/ (Dictionary.com), which in the combination Barack Obama is often initially-stressed as / ˈbæræk/ ( LDOCE ) even though the stressed syllables in / bəˈrɑk oʊˈbɑmə/ (Dictionary.com) are, again, separated from each other by one unstressed one. Retraction of stress in the finally-stressed modifier / bəˈrɑːk/ occurs even though the penult / bə/ contains a schwa and can thus be said to lack stress altogether, i.e., “a syllable of English is completely stressless if its vowel is schwa” (Hayes 1995: 12). The point here is that the Rhythm Rule should, according to Hayes 1.2 Previous studies 15 (1995: 19), be “unable to retract stress onto a completely stressless syllable,” i.e., the finally-stressed / bəˈrɑːk/ is supposed to be the only stress pattern of Barack , even when it modifies Obama . The alternation Ba ˈ rack ~ ˈ Barack Obama thus challenges Hayes’ (1995: 19) assertion.) Another interesting fact regarding the stress of the teen -cardinals is that “the pattern with primary stress on the first syllable is more common when counting” (Mompéan 2014: 152). That is, when an English speaker counts from 13 to 19, she usually says ˈ thirteen , ˈ fourteen , ˈ fifteen , ˈ sixteen , ˈ seventeen , ˈ eighteen , ˈ nineteen rather than ? thir ˈ teen , four ˈ teen , fifˈteen , six ˈ teen , seven ˈ teen , eigh ˈ teen , nine ˈ teen . Since both the former and the latter pronunciations have a one-syllable distance between the syllables bearing primary stress, we cannot say that the initially-stressed version ˈ thirteen , ˈ fourteen , ˈ fifteen , ˈ sixteen , ˈ seventeen , ˈ eighteen , ˈ nineteen is from the point of view of rhythm better than the finally-stressed thir ˈ teen , four ˈ teen , fifˈteen , six ˈ teen , seven ˈ teen , eigh ˈ teen , nine ˈ teen . Consider also compound dates involving the teen -cardinals. When these occur in the beginning of a compound date (for instance, 1821 or 1882 ), stress in them is almost exclusively penultimate / initial. For instance, in the YouTube videos containing the spoken occurrences of eighteen , this numeral was found to occur in the beginning of a compound date such as 1821 and 1882 87 times. Stress in eighteen was initial in each of these spoken occurrences. By contrast, in compound dates where the teen -cardinals occur in the end, there is considerable variation. For example, compound dates that end in eighteen (e.g., 1618 ) were found to occur in YouTube videos 27 times. With final stress eighteen was found to have been pronounced in such compound dates 19 times (~70.37 %), the initially-stressed ˈ eighteen was by contrast attested only eight times: ~29.63 %. However, if we specifically consider the compound dates 1913 , 1914 , 1915 , 1916 , 1917 , 1918 , and 1919 , the results are different. These dates were found to occur in YouTube videos 367 times. With final stress, the components 13 , 14 , 15 , 16 , 17 , 18 , and 19 occurring at the end of these dates were found to have been pronounced only 66 times (~17.98 %), whereas initially-stressed pronunciations could be heard by the author 301 times: ~82.02 %. Needless to repeat, the English Rhythm Rule, for which alternations such as thir ˈ teen ~ ˈ thirteen men are nothing more than a stress clash resolution strategy, cannot adequately account for the differences in the stress behavior that the numerals ending in teen exhibit in the beginning vs. the end of compound dates such as 1821 and 1918 . A less problematic explanation for why some English words have more than one stress pattern is emphasis. Emphatic stress shifts usually occur in prefixed words. Prefixes in English are generally stress-neutral (e.g., Cruttenden 2008: 241), but “[w]hen there is contrast or when the idea expressed by the prefix is given special prominence, the prefix bears the primary stress and the base a 16 1 Introduction secondary stress […]” (Poldauf 1984: 24). For example, triangle is supposed to be pronounced / traɪˈæŋɡ(ə)l/ ( OED ), with the location of stress in it coinciding with the location of stress in the disyllabic base angle , in which stress is initial. However, since a triangle is a geometrical figure that has three angles, in contrast to, e.g., a rectangle , which contains four right angles, stress in triangle is also frequently placed upon the semantically important prefix tri -: / ˈtraɪæŋɡ(ə)l/ ( OED ; cf. the more synchronically oriented OD , which gives for triangle only the prefix stress / ˈtrʌɪaŋɡ(ə)l/ ; likewise, in YouTube videos featuring more than 3,000 spoken occurrences of this prefixed derivative, the root stress / -ˈæŋɡ(ə)l/ was heard by the author only once. The prefix stress / ˈtrʌɪ-/ can thus be said to have over the course of time become the word’s default stress pattern). A similar case is / ˈsʌbməriːn/ vs. / ˌsʌbməˈriːn/ of submarine ( LDOCE ), with the latter pronunciation preserving the stress of the disyllabic base marine : / məˈriːn/ ( LDOCE ) vs. the former pronunciation emphasizing via stress the prefix sub -, which distinguishes submarine from other English marine -formations (cf. aquamarine , supermarine , transmarine , ultramarine , etc.); note also that especially the meaning of supermarine , “[t]hat is situated, takes place, or operates above or on the surface of the sea” ( OED ; boldface mine), is the spatial opposite of the meaning of submarine , “[t]hat exists or occurs under the surface of the sea” ( OED ; boldface mine). An interesting example is defense vs. offense . In YouTube videos featuring the spoken occurrences of defense , prefix stress could be attested only in sports-related contexts, namely, when defense is used to refer to actions of a team in a game of sports (especially basketball or American football) when the team does not possess the ball (used in the game). Similarly, when offense is pronounced / ˈɔˌfɛns/ or / ˈɑˌfɛns/ ( OED ), it invariably refers to actions taken by the team while it possesses the ball. In other contexts, stress in both defense and offense stays final. Examples include self-defense , secretary or ministry of defense , defense forces , legal or criminal defense , defense attorney , defense mechanism , Ph.D. defense , in defense of , etc. Regarding sports, pronunciations with stress falling upon the ult were heard only in chess-related contexts (e.g., Alekhine’s Defense ). The obvious explanation for this difference is that it is only ball games such as basketball or American football where we find such a prominent semantic opposition between defense (actions of a team without the ball) and offense (actions while possessing it). In stark contrast: A Ph.D. thesis can only be defended but not offended ; there is no such thing as Ph.D. offense understood as the semantic opposite of Ph.D. defense . Similarly, there are secretaries of defense , but there are no secretaries of offense . As for the more difficult chess-example, there is no doubt that just like actions of a team in a game of basketball or American football, chess-players’ actions can also be classified into defensive and offensive 1.2 Previous studies 17 actions. Note, however, that terms such as Alekhine’s Defense , Sicilian Defense , etc. denote instances of chess openings, which are usually classified into defenses and attacks (rather than offenses ). Thus, there can only be Alekhine’s Defense vs. Alekhine’s Attack , but not * Alekhine’s Offense . Hence, also in the case of chess-opening names such as Alekhine’s Defense , the meaning of defense is not the opposite of the meaning of offense . Stress in defense in names such as Alekhine’s Defense , Sicilian Defense , etc. therefore also stays final. The stress patterns exhibited by the words defense and offense in different semantic environments thus corroborate Friederich’s (1967: 62) assertion that “[a]m häufigsten findet sich emphatische Betonung natürlich zum Ausdruck des Gegensatzes oder Kontrastes,” i.e., since emphatic stress usually serves the expression of opposition or contrast, initial stress in defense and offense occurs only in those environments where the meanings of these two fense -words do indeed semantically contrast with each other. Emphatic stress shifts occur not only in words beginning with a prefix but also in words ending in a suffix. For example, in mortgagor - mortgagee it is the suffixes or and ee that bring about a contrast between borrowers and lenders in a mortgage. Therefore, in addition to stressing, e.g., mortgagor / ˈmɔːɡᵻdʒə/ (OED; both British and American English), preserving the stress of the base noun ˈ mortgage , this derivative can also be pronounced / ˌmɔːɡᵻˈdʒɔː/ ( OED ; British and American English), with the semantically and formally important suffix or receiving stress. A similar case is registrar , which is stressed / -ˈstrɑː/ vs. / ˈrɛ-/ in British English and only / ˈrɛ-/ in American English ( OED ). Given a contrast between registrars (those who keep registers) and registrees (those who register or are registered by others), the stress / -ˈstrɑː/ seems to represent an emphatic alternative to the base stress / ˈrɛ-/ . (Etymologically, final stress in the shorter English word registrar is due to antepenultimate stress in the longer Latin word registrarius . Since, however, contemporary English speakers—who by and large do not have any command of Latin—no longer remember this historical etymology, the stress pattern of registrarius in the source language Latin should not be a factor determining the stress pattern of registrar in the source language English.) The same seems to be true of / ˈfiːtjʊərət/ vs. / fiːtjʊəˈrɛt/ of featurette ( OED ) as well as / brɪˈkɛt/ vs. / ˈbrɪkɪt/ of briquette ( OED ). Because diminutiveness and artificiality are important aspects of the meanings of featurette and briquette —“A short feature film or programme” ( OD ) and “A block or slab of artificial stone” ( OED )—the suffix ette , which expresses these meanings in featurette and briquette , is on some occasions emphasized via stress. Notice also that several English suffixes capable of attracting stress (on to themselves; hence “autostressed” in Fudge’s (1984: 40) terminology) have a negative connotation (which is why—as, e.g., Lutz (2009: 290) observes—“[i]n Mod- 18 1 Introduction ern English, suffix-oriented accent is only rarely final”). This is especially true of the suffix ese in formations such as academese , “[t]he language or writing style of academic scholarship, especially when considered dry or over-complicated” ( OD ; boldface mine) and Brooklynese , “[a]n uncultivated form of New York speech associated especially with the borough of Brooklyn” ( OD ; boldface mine). Similarly, “English nouns in which the suffix ette designates a feminine role or identity have been perceived by many people as implying inferiority or insignificance: bachelorette ; drum majorette ; farmerette ; suffragette ; usherette .” According to Dictionary.com, “[o]f these terms, only drum majorette —or sometimes just majorette —is still widely used.” Likewise: Nouns in ess denoting occupation or profession are rapidly disappearing from American English. Airlines now refer to cabin personnel as flight attendants , not stewards and stewardesses . In the arts, authoress , editress , poetess , sculptress , and similar terms are either rejected or discouraged and almost always replaced by author , editor , poet , sculptor . Nouns in ess designating the holder of public office are hardly ever encountered in modern American usage. Women holding the office of ambassador, mayor, or governor are referred to by those titles rather than by the older, sex-marked ambassadress , mayoress , or governess . ( Governess has developed a special sense in relation to childcare; this use is less common in the U. S. than in Britain.) Among other terms almost never used in modern American English are ancestress , directress , instructress , manageress , oratress , postmistress , and proprietress . […] Jewess and Negress are usually considered offensive today. (http: / / www.dictionary.com/ browse/ -ess, 30. 03. 2017) Note also that “women are throughout history and across cultures defined by some kind of sexualization […]”, whereas “men have historically figured as non-gendered, non-sexualized, neutered subjects, those who have no sex (despite the attribution of sexual desire to men only) […]” (Baer 2008: 351). Because of this fact, many terms that are used to refer to women “have sexual meanings. If women are referred to as ‘chicks’, […] it is not just a reference to animals, but it is also a reference to a sexualized image” (Baer 2008: 352). An ess -formation, which emphasizes that the person referred to is female, is thus a priori emphatic in nature. That is why, some of the ess -formations that have not, as of today, disappeared from the English language have an emphatic alternative to base stress. In contrast to, e.g., actress , which is stressed only / ˈaktrəs/ ( OD ), with the suffix ess acting as a stress-neutral suffix (Cruttenden 2008: 240), princess is, according to the OED , stressed / -ˈsɛs/ vs. / ˈprɪn-/ in British English and / ˈprɪn-/ vs. / -ˈsɛs/ in American English. (The YouTube data has revealed, however, that also in British English, the base stress / ˈprɪn-/ is more frequently used than the suffix stress / -ˈsɛs/ .) 1.2 Previous studies 19 Related to emphasis is what Torsuev called (1960: 5, 6) the semantic factor of stress in the English language. This factor plays a role in, e.g., the stress variation / ˈju-ˈes-ˈeɪ/ vs. / ˈju-es-ˈeɪ/ of the alphabetism USA (Torsuev 1960: 5, 6). The semantic factor of English stress manifests itself in the variant / ˈju-ˈes-ˈeɪ/ . Since the abbreviated components U , S , and A all contribute to the meaning “ USA ,” primary stress in / ˈju-ˈes-ˈeɪ/ is placed upon each of the three syllables constituting the alphabetism USA . The alternative pronunciation / ˈju-es-ˈeɪ/ is, by contrast, a manifestation of the rhythmic factor. Since, as pointed out above, “adjacent stressed syllables make speech sound jerky” (Kingdon 1949: 149), the middle syllable of the alphabetism USA is either completely destressed or its level of stress is demoted to secondary. Noteworthy are also (low) frequency of use and polysyllabicity, which in contrast to emphasis, are, however, not immediate causes of stress variation but rather its prerequisites. A comparison of the pronunciations of all 75,000 entries in Wells’s dictionary yields 932 stress-divergent words. This set is found to differ from the entire lexicon in three respects of which the first appears to be more important than the others. The average frequency of the words is lower and their average length is higher. Furthermore, proper nouns are overrepresented. Generally, main stress falls further to the left in British than in American English, with the latter variety accommodating stress more in the word edges than the former. (Berg 1999: 123) With regard to polysyllabicity, Berg (1999: 127) observes that while disyllabic items occur less often in the stress-divergent set than in the general set, the reverse is true of longer words. The comparison of disyllabic vs. longer words produces a statistically significant difference […]. This allows us to conclude that words with a stress difference are longer than those without. (Berg 1999: 127) An explanation for the fact that words with stress doublets are more frequently found among “longer words” (i.e., words whose syllabic length is at least three) is that a disyllabic word cannot be derived from a variably-stressed monosyllabic base (in which there is only one syllable to place stress upon) whereas a word of three and more syllables can be derived from a variably-stressed disyllabic base. Thus, since the bases of disyllabic derivatives are as a rule monosyllables ( active ← act + ive ), stress in the former can only be placed upon the only syllable constituting the latter, i.e., e.g., in the disyllabic derivative active stress can only fall upon the only syllable constituting the monosyllabic base act . Stress in disyllabic derivatives such as active is thus almost exclusively initial in contemporary English. (With final stress, falling upon the suffix, a disyllabic derivative whose base is monosyllabic is pronounced only when the meaning inherent in 20 1 Introduction the suffix has a strong emphatic potential, such as, e.g., the meaning “female sex” of the suffix ess of the above mentioned princess .) As for trisyllabic derivatives, consider, for example, adeptness , adulthood , adversely , Nobelist , overtly , etc. Each of these derivatives is, according to the OED, interchangeably pronounced with penultimate and antepenultimate stress because also their disyllabic bases adept , adult , adverse , Nobel , overt , etc. are interchangeably pronounced with final and initial stress. (Note also that because, as pointed out above, / ˈædʌlt/ is the preferred stress pattern of adult in British English whereas American English speakers usually stress it / əˈdʌlt/ , also adulthood is more frequently stressed / ˈædʌlthʊd/ in British English vs. / əˈdʌlthʊd/ in American English.) Similarly, among variably-stressed words whose syllabic length is at least four are many derivatives whose bases are variably-stressed trisyllables. For example, since the trisyllabic aberrant is / ˈæbərənt/ vs. / əˈbɛrənt/ and aberrance is / ˈæbərəns/ vs. / əˈbɛrəns/ , the tetrasyllabic derivative aberrancy is / ˈæbərənsi/ vs. / əˈbɛrənsi/ ( OED ; only American English); since the trisyllabic consummate is / ˈkɒns(j)ᵿmət/ vs. / kənˈsʌmət/ , the tetrasyllabic consummately is / ˈkɒns(j)ᵿmətli/ vs. / kənˈsʌmətli/ ( OED ); since the trisyllabic controvert is / ˈkɒntrəvɜːt/ vs. / kɒntrəˈvɜːrt/ , the tetrasyllabic controvertist is / ˈkɒntrəvɜːtɪst/ vs. / kɒntrəˈvɜːtɪst/ ( OED ). Aspirant is interchangeably / əˈspaɪərənt/ and / ˈæspərənt/ ( LDOCE ; both in British and American English), with the former pronunciation being derived from / əˈspaɪə/ of aspire ( LDOCE ) and the latter being due to / ˈæspəreɪt/ of aspirate ( LDOCE ). That there is a connection between the meanings of the verbs aspire and aspirate is best illustrated by the popular Latin saying Dum spiro spero , which means “I hope as long as I breathe.” Similarly, from a semantic point of view it is not clear whether impactive , “[h]aving a strong effect or influence” ( OD ), should be regarded as a derivative from an ˈ impact , “[a] marked effect or influence” ( OD ), or from to im ˈ pact , “[h]ave a strong effect on someone or something” ( OD ). The longer derivative impactive is therefore interchangeably stressed / ˈɪmpæktɪv/ and / ɪmˈpæktɪv/ ( OED ). The morphological structure of advertisement that leads to the stress pattern / ədˈvəːtɪzm(ə)nt/ (which, according to the OED , is the only stress pattern in British English while American English speakers prefer the stress pattern / ˈædvərˌtaɪzm(ə)nt/ ) is advert + ise + ment , i.e., the etymon of the English word advertisement is, according to the OD , the Latin verb advertere , “turn towards.” Because the purpose of advertising is attracting other people’s attention (to what is being advertised), the semantic connection between the finally-stressed verb advert , whose original sense was “turn one’s attention to,” later “bring to someone’s attention” (OD), and the noun advertisement is still alive in Present-day English (which is corroborated by the fact that an advert occurs in British English as an informal shortening of adver- 1.2 Previous studies 21 tisement ). A natural alternative to pronouncing advertisement / ədˈvəːtɪzm(ə)nt/ , preserving the stress of the disyllabic base / ədˈvəːt/ (OD), is to pronounce it with the stress of the semantically related trisyllabic verb advertise : / ˈadvətʌɪz/ ( OD ). Thus, the stress of advertisement is either the preserved final stress of to advert (British English) or the preserved antepenultimate stress of advertise (American English). Cases such as these, when more than one word can synchronically be regarded as the base form are, again, much more typical of trisyllables and longer words (i.e., one of the few similar cases that involve a variably-stressed disyllable is solute , which means “[t]he minor component in a solution, dissolved in the solvent” ( OD ; boldface mine). In addition to counting as a back-derivative from solution , which leads to the stress pattern / sɒˈljuːt/ ( OD ), solute is, just like solvent , also initially-stressed as / ˈsɒljuːt/ ( OD ), with the location of stress being the monosyllabic base solve .) To conclude, in contrast to Berg (1999: 127), who argues that “[l]ength thus appears to erode the integrity of a word’s stress pattern. In other words, it encourages variability in lexical-stress placement,” the present monograph argues that the fact that words with stress doublets are more frequently found among words whose syllabic length is at least three is a relatively insignificant, fairly accidental consequence of the impossibility of deriving disyllabic words from variably-stressed monosyllabic bases. Apart from this, however, there is no intrinsic connection between syllabic length and stress variation. Both disyllables and longer words are stressed variably by English speakers only when there are reasons (e.g., emphasis) for this. As for (low) frequency of use, an obvious explanation, not requiring a detailed elaboration, is that “infrequency […] weakens the memory trace (or prevents it from growing strong in the first place) and thereby destabilizes the information to be remembered” (Berg 1999: 137). An example might be albumenize , which, according to the OED, is pronounced only / ælˈbjuməˌnaɪz/ in American English, but it can be pronounced either alˈbumenize or ˈ albumenize in British English. To find out whether stress in this verb in British English more frequently falls upon the first or the second syllable, the orthographic form albumenize was, together with the word-forms albumenizes , albumenized , and albumenizing , searched for on the YouTube Web site. Additionally, similar searches were performed with respect to the orthographic alternatives albumenise , albuminise , and albuminize , which, according to the OED , also occur in Present-day English. Unfortunately, however, neither the former nor the latter searches have returned any results, i.e., on the YouTube Web site there seem to be no videos in which the verb albumenize is pronounced by an English speaker at least one time. The non-existence of YouTube videos featuring the spoken occurrences of albumenize thus 22 1 Introduction clearly points to the status of the verb under analysis as a low frequency verb. Given this finding, however, it is important to emphasize that the infrequency of the verb albumenize is not the cause of the stress variation in it. The cause is that the base form albumen is pronounced both / ˈælbjʊmɪn/ and / ælˈbjuːmɪn/ (LDOCE). The derived verb albumenize thus merely inherits this stress variation from its base. At the same time, however, it is clear that the infrequency of the verb under consideration contributes to the fact that neither the pronunciation alˈbumenize nor the pronunciation ˈ albumenize becomes the institutionalized pronunciation of albumenize , i.e., “infrequency […] weakens the memory trace (or prevents it from growing strong in the first place) and thereby destabilizes the information to be remembered” (Berg 1999: 137). Do observe, however, that infrequency is not a conditio sine qua non of stress variation. To be stressed variably, an English word is not required to be a low frequency word. Thus of the variably-stressed English words discussed thus far, many are, according to LDOCE , high frequency words. E.g., address , adult , decade , princess , etc. A more recent study that just like Berg’s (1999), exclusively deals with variably-stressed English words is Henderson (2010). The latter is, however, a smallscale study aimed at establishing the preferred stress patterns of the words complex , create , economic , individual , Japanese , necessarily , and research , which are known to be stressed variably in Present-day English. What Henderson (2010) did was analyze how these variably-stressed words are pronounced by American English speakers whose voices can be heard in the videos available on the TED Web site (https: / / www.ted.com/ talks, 10. 06. 2015); this Web site hosts transcribed conference talks dealing with “Technology, Entertainment and Design” (hence the abbreviation TED ). The findings of Henderson’s (2010) study are as follows: The adjectival lexeme complex was more frequently stressed ˈ complex than com ˈ plex ; create was only stressed cre ˈ ate and never ˈ create ; economic was more frequently stressed eco ˈ nomic than ˈ economic , individual was only stressed indiˈvidual and never ˈ individual , Japanese vacillated between Japa ˈ nese and ˈ Japanese , but this variation did not always constitute a stress clash avoidance strategy, abiding by the English Rhythm Rule (e.g., Japa ˈ nese ~ ˈ Japanese ˈ language ); necessarily was more frequently stressed neces ˈ sarily than ˈ necessarily ; research was more frequently stressed ˈ research than re ˈ search (Henderson 2010: 106-110). Needless to say, since only a handful of variably-stressed English words were object of Henderson’s (2010) study, the findings reported by her do not justify any meaningful generalizations regarding the phenomenon of stress variation in Present-day English. To reiterate, several authors working on English stress have noticed that some English words have stress doublets, i.e., more than one stress pattern 1.2 Previous studies 23 corresponding to one and the same meaning. However, of the publications cited above, only Berg (1999) can be regarded as a systematic study, i.e., the author has manually analyzed the stress patterns of all English words for which phonetic transcriptions are given in the 1990’s edition of Longman Pronouncing Dictionary (Wells 1990). Note, however, that Berg’s (1999) study is exclusively concerned with words that have different stress patterns in British vs. American English (e.g., combatant is, according to LDOCE , only / ˈkɒmbətənt/ in British English vs. only / kəmˈbætnt/ in American English). Similar studies focusing upon across-varietal stress differences are Peng & Ann (2001), Van Rooy (2002), Wiltshire & Moon (2003), Simo Bobda (2010), Krivokapić (2013), Altmann & Kabak (2015), and Tan (2015). These articles deal with either stress location differences or phonetic properties of stress, which (among other things) make varieties such as Nigerian English, Singapore English, Tswana English, Indian English, and Cameroon English phonetically different from British and American English (the two major contemporary English reference accents). Recall, however, that some English words have different stress patterns within one and the same Present-day English variety (and, as observed above, sometimes even identical English speakers can be heard pronouncing one and the same English word with more than one stress pattern). Recall also that the findings reported by Berg (1999) are based exclusively upon dictionary data: the 1990’s edition of Longman Pronouncing Dictionary. However, even renowned lexicographic resources such as Longman or the OED do not always include all actually occurring stress patterns. E.g., locate is, according to LDOCE , only / ləʊˈkeɪt/ in British English vs. only / ˈloʊkeɪt/ in American English. In disagreement with this, however, in the YouTube videos a CMA 9Llb QH 4 (25. 05. 2017) and ql IG v0CtqXs (25. 05. 2017), British English speakers can be heard pronouncing locate with initial stress; to be more precise, in the former video locate is ˈ locate in I am drawing a map of Britain, just so I can ˈ locate where I have been (00: 02: 14.160 --> 00: 02: 26.590), but the very same British English speaker stresses locate finally in I am just going to roughly lo ˈ cate these sketches (00: 02: 36.200 --> 00: 02: 41.560). Similarly, in YouTube videos where locate is pronounced by American English speakers, the author could hear not only initiallybut also finally-stressed pronunciations (with, however, the initially-stressed / ˈloʊkeɪt/ being the more frequently-used stress pattern). A fairly similar case is the above mentioned paprika . In both LDOCE and the OED , this word is said to vacillate between the pronunciations / ˈpæprɪkə/ and / pəˈpriːkə/ only in British English, whereas in American English, stress in paprika is exclusively penultimate: / pəˈpriːkə/ . However, in YouTube videos in which paprika is pronounced by native English speakers, both British and American English speakers were heard to have stressed paprika penand antepenultimately, with, 24 1 Introduction however, the former stress pattern being more frequent than the latter both in British and American English. 1.3 Structure of the book The present monograph thus aims to give a systematic account of the phenomenon of within-varietal stress assignment instability in Present-day English using both dictionary and corpus data (i.e., YouTube videos in which English words with stress doublets are actually pronounced by contemporary English speakers.) The monograph is organized in the following way. Chapter 2 first discusses the phonetic properties of stress (i.e., differences between syllables bearing stress and syllables lacking it) and then summarizes the inventory of the stress assignment principles in the languages of the world (e.g., differences between fixed-stress languages and variable-stress languages). Chapter 3 elaborates upon methodological issues, such as, for example, the advantages and disadvantages of using YouTube as a corpus of spoken English. Using predominantly dictionary data ( LDOCE , OD , etc.), Chapter 4 will discuss the stress patterns of all English words, whereas the first section of Chapter 5 will be exclusively concerned with English words with stress doublets (i.e., the question of why some English words have more than one stress pattern). Finally, using YouTube data, the second section of Chapter 5 will clarify why, in the case of words with stress doublets, one stress pattern is as a rule more frequently used by contemporary English speakers than an alternative stress pattern, e.g., why Euclidean is considerably more frequently pronounced / juːˈklɪdɪən/ than / juːklɪˈdiːən/ , with both these pronunciations counting, according to the OED , as the legitimate pronunciations of Euclidean . 2.1 Stress vs. accent 25 2 Theoretical preliminaries Before proceeding to a more systematic analysis of the contemporary English stress system (i.e., principles governing the incidence of stress in English words), it makes sense to look at the physical nature of the phenomenon of linguistic stress. This chapter will therefore discuss phonetic differences between 1) syllables bearing stress and syllables bearing accent, 2) syllables bearing primary stress and syllables bearing secondary stress, and, more generally, between 3) syllables bearing stress and / or accent and unstressed syllables. Additionally, the chapter will address stress placement from a cross-linguistic perspective, i.e., on the basis of which principles languages of the world, including English, stress words occurring in them. 2.1 Stress vs. accent Following authors such as Turk & Sawusch (1997: 25), Cho et al. (2007: 218), Scarborough et al. (2009: 137), Plag et al. (2011), Mücke & Grice (2014: 48), etc., the present monograph will use the term accent in the sense “pitch accent as a phonetic realization of phrasal stress,” whereas the term stress will be used in connection with other phonetic means (mainly duration) that English speakers rely upon in out-of-focus environments, in which a particular syllable bears only word but no phrasal stress. Whereas word stress mainly performs a constitutive function, which means that it merely “arranges syllables in words” (Sokolova et al. 1997: 138-139), phrasal stress “indicates that a constituent contains new or important information: it is focused ” (Scarborough et al. 2009: 136; authors’ italics). Regarding the phonetic realization of phrasal stress, consider, for instance, the combination working theory , which depending upon the meaning it expresses is pronounced working ˈ theory or ˈ working theory . As Poldauf (1984: 118) observes, when phrasal stress in this combination falls upon the righthand component theory , it means “a theory that works or functions well.” In the other case, i.e., when phrasal stress is placed upon the lefthand component working , the meaning is “a theory that can be used as a point of departure for some work / a theory that is intended for work, working, or functioning.” Phonetically, however, the difference is that while in ˈ working theory the vowel of the stressed syllable of the lefthand component working has a considerably higher pitch peak 26 2 Theoretical preliminaries than the vowel of the stressed syllable of the righthand component theory , in working ˈ theory , by contrast, the stressed vowel of working and that of theory exhibit near-identical pitch peaks (Farnetani et al. 1988). In other words, from the point of view of the speaker, working ˈ theory is actually ˈ working ˈ theory , i.e., “both words are accented,” as Farnetani et al. (1988: 171) observe. Nonetheless, listeners “incorrectly” perceive the pronunciation ˈ working ˈ theory as working ˈ theory , which is most likely due to the unfulfilled expectation of the declination of pitch, normally occurring at the end of an intonation group (Farnetani et al. 1988: 170). More recently, similar results have been obtained by Kunter (2011), whose dissertation is concerned with the accentuation of Noun + Noun combinations, such as budget deficit : in right-prominent compounds, left and right elements have nearly indistinguishable pitch values, while the pitch of left elements is clearly higher than that of right elements if the whole compound is left-prominent (Kunter 2011: 89). In other words, when budget deficit is pronounced ˈ budget deficit , the stressed syllable of the lefthand component budget is pitch-accented. When, however, budget deficit is pronounced budget ˈ deficit , both the stressed syllable of the lefthand component budget and the stressed syllable of the righthand component deficit are pitch-accented, i.e., budget deficit is actually pronounced ˈ budget ˈ deficit but perceived to have been pronounced budget ˈ deficit . Even more recently, Kösling’s (2013) dissertation has studied the accentuation of triconstituent Noun + Noun + Noun combinations. Her main finding is that while in so-called right-branching combinations such as Boston [gang members] (= “gang members who are from Boston or commit crimes in Boston”) there is usually “a high pitch on both constituent N1 and constituent N2 and a clearly lower pitch on constituent N3,” left-branching combinations such as [credit card] companies (“companies that issue credit cards”) are normally characterized by “a high pitch on constituent N1 but clearly lower pitches on both constituent N2 and constituent N3” (Kösling 2013: 54). In other words, Boston gang members is accented ˈ Boston ˈ gang members , but credit card companies is accented ˈ credit card companies . As for word stress, it is unfortunate that many authors do not distinguish between its phonetic realizations in accented and non-accented positions. A classic case is Fry (1958: 126), who described “the all-or-none effect” caused by changes in the fundamental frequency of a syllable, to which the impression that a particular syllable has a particular pitch level is intrinsically connected. What is meant by this all-or-none effect is that in a situation when a listener is required to make a subjective judgment as to which syllable in a word bears 2.1 Stress vs. accent 27 stress, a higher pitch peak of a particular syllable is capable of overriding the effect of a reduced loudness and shorter duration of the same syllable. In other words, a particular syllable is likely to be perceived as stressed even when that syllable only has a higher pitch peak, but is shorter and quieter than the other syllables co-occurring in the same word. Notice that Fry’s (1958) experiments involved artificially-synthesized manipulations of the fundamental frequency, duration, and intensity of the syllables in the disyllabic words subject , object , digest , contract , and permit . His subjects listened to these manipulated pronunciations and were then asked to indicate whether the words they had just heard were nouns or verbs. As the nouns subject , object , digest , contract , and permit are stressed initially, whereas in the corresponding orthographically identical verbs the stress is final, the participants’ word-class membership judgments served simultaneously as stress-location judgments. Similar results were later obtained by Morton & Jassem (1965), who studied how test persons’ stress-location judgments were affected by manipulations of the pitch, duration, and intensity levels of the syllables occurring in the synthesized pronunciations of the nonsense words * Soso , * Sisi , and * Sasa . Variations in fundamental frequency produced far greater effects than variations in either intensity or duration, a syllable being marked as stressed if it differed from the ‘context’ fundamental. A raised fundamental was more efficient than a lowered one. (Morton & Jassem 1965: 159) A serious problem with the treatment of pitch as a cue of word stress is, however, the well-known fact that phrasal stress usually falls on the same syllable as lexical stress: When a word bears a pitch accent, that accent is usually attached to the syllable bearing lexical stress […] As a result, a single syllable can be simultaneously lexically stressed and pitch accented, for example, the syllable par in We drove to the ‘PARty . For a word in isolation, then, lexical stress is necessarily confounded with phrasal pitch accent, and vice versa […] (Scarborough et al. 2009: 137) The same does not, however, apply to out-of-focus positions, in which words do not constitute new or important information and thus do not receive phrasal stress. The determination of the phonetics of ‘stress’ (whatever kind) has turned out to be notoriously difficult […] Earlier assessments put duration, fundamental frequency and intensity central, in addition to other factors involving the relative ‘strength’ of phonemes in ‘stressed’ syllables. […] Some claimed that fundamental frequency takes the lead in all of this […] It was then shown that the big role of (changes in) fundamental 28 2 Theoretical preliminaries frequency results from the fact that the pitch properties of ‘stressed’ syllables are due to the fact that the examined words are uttered in isolation. This causes their ‘stressed’ syllable to be bearers of intonational pitch movement. When words are examined in positions where their syllables do not attract intonational pitch movement (in ‘out-offocus’ positions) it turns out that pitch is not a major cue at all, but rather duration, spectral tilt and other effects of articulatory force (which may include a small pitch rise). (van der Hulst 2012: 1508) Indeed, much earlier than van der Hulst (2012: 1508), Huss (1978: 86) had reported that Stress oppositions between words like im port ≠ im port are audible in the nucleus of intonation contours where they are distinguished by fundamental frequency. In the post-nuclear position the opposition is neutralized: the two stresses are identical in terms of fundamental frequency, but measurably different in terms of intensity and duration. (Huss 1978: 86) Similarly, an experimental study by Adams & Munro (1978), which was concerned with word stress cues relied upon in naturally occurring connected speech, has demonstrated that “duration was by far the most frequently used cue” (Adams & Munro 1978: 125). Furthermore, even as far as in-focus positions are concerned, the all-or-none role of fundamental frequency was relativized by the findings of Mcclean & Tiffany (1973: 283), who had discovered that pitch level contrasts between stressed and unstressed syllables predominate only in initially-stressed (nonsense) disyllables— Say * / ˈsasa/ please —whereas in disyllables in which stress occurs finally— Say * / saˈsa/ please —the same contrast is mainly cued by duration. 2.2 Primary stress vs. secondary stress The importance of distinguishing between accented and non-accented positions is also exemplified by the (absence of the) phonetic contrast between primaryand secondary-stressed syllables (in English). According to Plag et al. (2011: 362), There is a large body of literature available on the acoustic correlates of stress in English (probably starting with Fry, 1955, 1958), and there is a host of parameters that have been suggested to be acoustic correlates of stress. However, previous research has almost entirely focused on the question of how stressed and unstressed syllables differ from each other in terms of their acoustic properties, whereas the distinction between primary and secondary stress has hardly received any attention by phoneticians. (Plag et al. 2011: 362) 2.2 Primary stress vs. secondary stress 29 To fill in the research gap, Plag et al. (2011) have studied the phonetic properties of primary and secondary stress in both leftand right-prominent English words. Left-prominent words are words such as, for example, illustrate , in which primary stress precedes secondary stress, i.e., / ˈɪl·əˌstreɪt/ (Cambridge Dictionaries Online, henceforth CDO ). In right-prominent words, such as, for example, illustration , the order of stresses is reversed: Primary stress is followed by secondary stress, i.e., / ˌɪl·əˈstreɪ·ʃən/ ( CDO ). The most interesting findings of Plag et al.’s (2011) study are as follows. In accented positions, in which words bear phrasal stress, which in English is phonetically realized as pitch accent, the differences between left-prominent words such as illustrate and right-prominent words such as illustration resemble the above mentioned difference between ˈ working theory and working ˈ theory (Plag et al. 2011: 372). That is, in left-prominent words occurring in accented positions there is only one pitch accent, which falls on the primary-stressed syllable: Illustrate is pronounced ˈ illustrate , with pitch accent being placed only on the primary-stressed initial syllable / ˈɪ/ , but not on the secondary-stressed final syllable / ˌstreɪt/ . (Note also that both the OD and LDOCE give the transcription / ˈɪləstreɪt/ , where there is only the primary stress symbol (ˈ). A well-known fact is that ate -words in English sometimes occur not only as verbs but also as adjectives. E.g., what distinguishes to subordinate from the adjective subordinate is that the former has a diphthong in the ult—/ səˈbɔːdɪneɪt/ ( OD )—whereas the latter is pronounced with a reduced vowel: / səˈbɔːdɪnət/ ( OD ). Given that both the transcription / səˈbɔːdɪneɪt/ and the transcription / səˈbɔːdɪnət/ contain only the primary stress symbol (ˈ), the present monograph argues that with regard to stress, to subordinate is not different from the adjective subordinate , i.e., both the ult / neɪt/ of the former and the ult / nət/ of the latter are unstressed (but because the ult / nət/ of the adjective subordinate contains a reduced vowel, it is of course less prominent than the ult / neɪt/ of to subordinate , which contains a diphthong). Similarly, a graduate and to graduate both have antepenultimate stress—/ ˈɡradʒʊət/ and / ˈɡradʒʊeɪt/ ( OD )—but while the former is pronounced with a reduced vowel in the ult, the latter pronunciation contains a diphthong.) In right-prominent words, by contrast, there are two (equal) pitch accents. That is, illustration is pronounced ˈ illu ˈ stration , with pitch accent being placed both upon the secondary-stressed first syllable / ˌɪ/ and upon the primary-stressed third syllable / ˈstreɪ/ . (In agreement with this finding, the OD gives the transcriptions / ᵻˈməʊʃn̩ (ə)lʌɪˈzeɪʃn/ for emotionalization , / ˈdʒɒlɪfɪˈkeɪʃ(ə)n/ for jollification , / fəˈrɪŋɡəlʌɪˈzeɪʃn/ for pharyngalization , etc.; similarly, agreeability is / əˈɡriːəˈbɪlᵻti/ , allowability is / əˈlaʊᵻˈbɪlᵻti/ , materiality is / məˈtɪərɪˈalɪti/ , etc., with more than one syllable in these transcriptions counting, according to the OD , as syllables bearing primary stress). 30 2 Theoretical preliminaries With regard to non-accented positions, in which words do not receive phrasal pitch accent, the finding of Plag et al.’s (2011) investigation is that the phonetic contrast between primaryand secondary-stressed syllables is almost completely neutralized. As the authors report: […] primary and secondary stress syllables are stressed syllables that are different from unstressed syllables, but not from each other, unless the word is accented. In this case the target of a nuclear accent corresponds to what is usually labeled the primary stress syllable. If the word occurs in an environment in which no accents are present (e.g. in post-nuclear position), there is no phonological difference between the first and third syllable in words such as i.so.late and i.so.la.tion . These syllables are simply stressed (or strong) syllables (which, of course, differentiates them from the second syllable .so. , which is unstressed, or weak). (Plag et al. 2011: 373) Accordingly, “if we disregard accentuation, there is no difference between secondary and primary stress” (Plag et al. 2011: 373). (This conclusion poses a major empirical challenge to Chomsky & Halle’s (1968) highly counterintuitive claim that non-primary stresses in English are not only secondary, but also tertiary, quaternary, quinary, senary, etc., i.e., as Halle & Vergnaud (1987: 37-38) explain, since especially the derivation of a phrase might, depending upon its syntactic complexity, involve a fairly large number of cycles, there is no maximum number of degrees of stress in English.) The finding that (if accentuation is left out of consideration) primary stress is phonetically not different from secondary stress also has implications for the study presented in this monograph. Consider, for instance, the adjective meningococcic , for which the OED gives the American English transcription / məˈˌnɪŋɡoʊˈkɑk(s)ɪk/ . This transcription stands for the variation between the pronunciation / məˈnɪŋɡoʊˈkɑk(s)ɪk/ , where there are two primary stresses, and / məˌnɪŋɡoʊˈkɑk(s)ɪk/ , where there is only one primary stress. Given, however, Plag et al.’s (2011) findings, we can argue that these pronunciations are identical. Since in non-accented positions the contrast between primary stress and secondary stress is neutralized, the secondary-stressed syllable / ˌnɪŋ/ of the pronunciation / məˌnɪŋɡoʊˈkɑk(s)ɪk/ is not different from the primary-stressed syllable / ˈnɪŋ/ of the pronunciation / məˈnɪŋɡoʊˈkɑk(s)ɪk/ . Likewise, also in accented positions the two pronunciations cannot be different from each other because in right-prominent words such as / məˌnɪŋɡoʊˈkɑk(s)ɪk/ both primaryand secondary-stressed syllables are pitch-accented. That is, the acoustic form that the listener will perceive as meningo ˈ coccic will actually be pronounced me ˈ ningo ˈ coccic by the speaker, with two pitch accents being assigned to both the secondary-stressed syllable / ˌnɪŋ/ and the primary-stressed syllable / ˈkɑk/ . The only possible variation that meningococcic can exhibit in an accented po- 2.3 Stress / accent vs. no stress 31 sition is that between the left-prominent pronunciation me ˈ ningococcic and the right-prominent alternative meningo ˈ coccic (which, as just said, is phonetically equivalent to the double-prominent me ˈ ningo ˈ coccic ), but this type of variation is not stipulated by the OED ’s transcription / məˈˌnɪŋɡoʊˈkɑk(s)ɪk/ , where the stressed syllable / ˈkɑk/ is said to be always pronounced with primary stress, i.e., it is only the stressed syllable / ˈˌnɪŋ/ where, according to the OED , the level of stress vacillates between primary and secondary. In summary, the adjective meningococcic , which, according to the OED , vacillates between the pronunciations / məˈnɪŋɡoʊˈkɑk(s)ɪk/ and / məˌnɪŋɡoʊˈkɑk(s)ɪk/ (as far as the American variety is concerned), cannot exhibit this variation because the pronunciation / məˈnɪŋɡoʊˈkɑk(s)ɪk/ , which involves two primary stresses, is in no way different from the pronunciation / məˌnɪŋɡoʊˈkɑk(s)ɪk/ , where the primary-stressed syllable / ˈkɑk/ is preceded by the secondary-stressed syllable / ˌnɪŋ/ . 2.3 Stress / accent vs. no stress Finally, we briefly discuss differences between stressed and unstressed syllables without specifically considering whether the former bear not only word but also phrasal stress. According to van der Hulst (2010b: 9), the prototypical characteristics of a stressed syllable are 1) greater duration, 2) a balanced spectral tilt, 3) a higher fundamental frequency/ pitch, 4) greater precision or extra phonetic traits (such as full vowel quality, aspiration), 5) extra phonotactic possibilities / greater complexity, 6) marking sites for morphological processes, and 7) serving as an anchor for intonational tones. Indeed, as already mentioned in the previous sections, syllables bearing stress are longer and louder (than unstressed ones) and, if they bear not only word but also phrasal stress, also higher-pitched. What is meant by “a balanced spectral tilt” is “a more even intensity distribution across the frequency spectrum for accented vowels” (van der Hulst 2010b: 4). This is because “the vowel of an accented syllable is more precisely articulated (less-centralized) and / or longer” (van der Hulst 2010b: 4). E.g., Kent & Netsell (1971) studied the production of the words convoy , commune , compact , escort , impact , increase , insight ( incite ), insult , protest , and suspect , which, depending upon the location of stress in them, function as instances of different morphosyntactic categories. The central finding is that the production of stressed syllables is characterized by “a relative lowering of the jaw and a greater opening of the lips” as well as the “[displacement of] the tongue points in the direction of the presumed articulatory target” (Kent & Netsell 1971: 33, 36). With regard to the latter, Kent & Netsell (1971: 36) note that, for example, “the tongue-marker 32 2 Theoretical preliminaries positions during stressed / i/ are displaced upward and forward relative to the tongue-marker positions during unstressed / i/ .” In other words, the stressed / i/ is higher and fronter than / i/ that occurs in non-stressed positions. In a more recent article Gay (1978: 348) also mentioned a greater electromyographic activity (i.e., the activity of the muscles) that underlies the production of stressed vowels, and even more recently, Mücke & Grice (2014: 49) have observed that “[a]rticulation is more distinct in prominent positions, resulting in temporal and / or spatial expansion of articulatory movements.” The two strategies employed are sonority expansion and hyperarticulation: The Sonority Expansion Hypothesis claims that the intrinsic sonority of a vowel is enhanced to strengthen syntagmatic contrasts, i.e. the contrast between vowels and consonants. Under accent, the speaker intends to produce a louder vowel by opening the mouth wider over a longer time. A more open oral cavity allows for greater radiation from the lips, leading to an increase of the overall acoustic energy. […] The Strategy of Hyperarticulation involves the enhancement of contrastive features as well as of sonority. A low vowel is produced with a lower tongue position, a front vowel with a more fronted tongue position […] and a back vowel with a more retracted tongue position (Mücke & Grice 2014: 49) Of note is also an investigation by Scarborough et al. (2009), which deals with the so-called “optical phonetics” of both word and phrasal stress in English. Similar to earlier contributions, Scarborough et al. (2009) have studied the production of forms that, depending upon their stress contours, function as instances of different morphosyntactic categories: discharge , discount , pervert , and subject . Scarborough et al. (2009) have addressed the question of whether the production of syllables receiving either word or phrasal stress differs from the production of unstressed / unaccented syllables with regard to the following parameters: eyebrow displacement, head displacement, lip distance, lip displacement in opening and closing gestures, chin displacement in opening and closing gestures, lip opening and lip closing velocities, and chin opening and chin closing velocities (Scarborough et al. 2009: 145). As far as word stress is concerned, syllables bearing it have turned out to be different from unstressed ones with regard to the parameters of head displacement, lip distance, lip opening displacement, lip opening velocity, chin opening displacement, and chin opening velocity: These measures are all larger and faster in stressed syllables (Scarborough et al. 2009: 146). The parameters of lip closing displacement, lip closing velocity, chin closing displacement, and chin closing velocity exhibit the same pattern, but as the authors report, they fail to achieve statistical significance (Scarborough et al. 2009: 146-147). The only parameter that apparently has nothing to do with the difference between stressed 2.3 Stress / accent vs. no stress 33 and unstressed syllables is eyebrow displacement: As Scarborough et al. (2009: 147) observe, “the eyebrows showed no movement at all.” As for phrasal stress, each of the parameters named above has achieved statistical significance, i.e., the measures for them are larger and faster in accented syllables. Of particular interest are the results for eyebrow and head displacement. According to Scarborough et al. (2009: 151-152), these parameters differ from the others in that they are not directly associated with segmental articulation. Therefore, their movements could be purely prosodic, free to cue stress without being constrained by segmental context. As noted above, the typical movement of the head associated with stress was in a downward direction, while brow movements were typically upward. From our qualitative observation of the tokens, these movements were coordinated such that the eyebrows (or at least one of them) rose just before the stressed word and fell slowly during the word, and the head fell, or nodded, beginning during the closure of the initial stop in stressed words […] (Scarborough et al. 2009: 151-152) To conclude, Scarborough et al.’s (2009) results indicate that not only is the optical phonetics different for stressed and unstressed syllables as well as for accented and non-accented words. In addition, stressed and accented syllables differ from each other with regard to eyebrow movements. These play an important role in the production of accented syllables, but they play no role at all when a stressed syllable is produced. The claim that stressed units are generally produced more intensively than unstressed ones seems to be true not only of spoken but also of sign languages. For example, a study by Wilbur & Schick (1987) addressed the articulatory nature of stress in American Sign Language. Since it is primarily a monosyllabic language, Wilbur & Schick (1987: 303) were concerned with the production of stress at the sentence level. According to the authors, signers who participated in their study […] frequently used strong facial features, such as brow furrowing, eyebrow raising, puffed cheeks, labial plosions, tongue thrusts, or clenched teeth, to indicate that a particular sign or phrase was stressed. […] In addition to facial behavior, in many stressed signs, subjects would use body shifting, head movement and tilting, and body and head jerks (body shifts) to add emphasis to a sign or its phrasal group. (Wilbur & Schick 1987: 310) Note also that reduction of vowels in unstressed positions, which is so typical of the English language, can likewise be attributed to production differences between stressed and unstressed syllables. Because stressed syllables are longer than unstressed ones, 34 2 Theoretical preliminaries […] for destressed (or fast) speech, commands for successive articulatory movements are issued at a rate that is too fast for the targets to be attained. The result is articulatory (and acoustic) undershoot, the extent of which is directly proportional to the duration of the vowel. In other words, spectral reduction is a natural consequence of a decrease in vowel duration. (Gay 1978: 348) With regard to marking sites for morphological processes, consider the following three transcriptions: / ˈti: ni/ , / ˈmɑrɾi/ , and / mɑr/ . According to Carter & Clopper (2002: 333), these are three possible American English shortenings of the noun martini . The obvious structural differences between them are as follows. In the variant / ˈti: ni/ the stressed syllable / ˈti: / is the preserved stressed syllable / ˈtiː/ of the base form martini . In the variant / ˈmɑrɾi/ the original stressed syllable appears in a reduced, unstressed form / ɾi/ , with the original / t/ being replaced by a flap. Finally, the shortening / mɑr/ does not contain the original stressed syllable / ˈtiː/ in any form. As the production experiment by Carter & Clopper (2002: 334) has demonstrated, of these three clipping possibilities, the first possibility, exemplified by / ˈti: ni/ , has more often been made use of than the other two possibilities, exemplified by the forms / ˈmɑrɾi/ and / mɑr/ . In other words, when English speakers shorten words, they tend to preserve the original stressed syllables of longer input forms as stressed syllables of shorter output forms. Further examples from Carter & Clopper (2002: 321-322) include Bécca for Rebécca , cáuse for becáuse , ráffe for giráffe , nána for banána , etc. Examples like these provide a good illustration of the fact that stress can be an important factor determining the outcome of a morphological process such as shortening, or as van der Hulst (2010b: 9) notes, stress can “mark sites for morphological processes.” As for the last characteristic—serving as an anchor for intonational tones— recall once again that the big role of (changes in) fundamental frequency results from the fact that the pitch properties of ‘stressed’ syllables are due to the fact that the examined words are uttered in isolation. This causes their ‘stressed’ syllable to be bearers of intonational pitch movement. (van der Hulst 2012: 1508) Thus, because in the English language, a falling tone is usually “associated with finality, completeness, and definiteness” (Gut 2009: 122), an individual English word cited in isolation (e.g., in online dictionaries, in which a standard pronunciation of a word can often be listened to) is as a rule pronounced with a high-falling tone (Cruttenden 2008: 282); the stressed syllable of such a word is therefore higher-pitched than the other syllables occurring in the same word. In addition to the aforementioned prototypical features of stressedness, provided by van der Hulst (2010b: 9), for purely practical purposes of determining 2.3 Stress / accent vs. no stress 35 which syllable in a word (that occurs in a corpus of spoken data) bears stress, any researcher can also rely upon his or her own auditory impressions of some syllable in the word under analysis being more prominent than other syllables constituting the same word. At first glance this may seem fairly subjective, but as Hayes (1995: 22) observes: The relevant facts of segmental phonetics and pitch contours are clear to anyone with a reasonably good ear and a little practice. Since the auditory data involve discrete patterns rather than physical quantities, they are more reliable and easier to interpret. (Hayes 1995: 22) The assertion that “the auditory data […] are more reliable and easier to interpret” is, among other things, justified by the parasitic nature of stress, which means that “it invokes phonetic resources that serve other phonological ends” (Hayes 1995: 7). Consider, for instance, the role of vowel duration as a cue of word stress. As Torsuev (1960) demonstrates, durational differences between English vowels can, apart from the presence / absence of stress, occur due to several other reasons. For example, the fact that in pairs of words such as bead - bid , pool - pull , naught - not , etc., the phonetically very similar vowels / iː/ and / ɪ/ , / uː/ and / ʊ/ , / ɔː/ and / ɒ/ exhibit durational differences has obviously nothing to do with the location of stress in these monosyllabic words. It is simply a peculiarity of the English language that it has vowel contrasts such as / iː/ vs. / ɪ/ , / uː/ vs. / ʊ/ , / ɔː/ vs. / ɒ/ , accounting thereby for a semantic difference between the words bead and bid , pool and pull , and naught and not ; according to Torsuev (1960: 45), in cases like these we are dealing with durational differences due to the historical length of English vowels. Another well-known fact is that vowels in English that are immediately followed by voiced consonants are longer than those that are followed by voiceless ones (which is also completely unrelated to stress): E.g., the length of the short vowel / ɪ/ of bid , where the following sound is the voiced / b/ , can, according to Torsuev (1960: 45), be identical to or even exceed the length of the long vowel / iː/ of beat , where it is followed by the voiceless / t/ ; similarly, the long vowel / iː/ of see is longer than the same vowel in seat and seen because / siː/ is an open syllable, whereas both / siːt/ and / siːn/ end in consonants (Torsuev 1960: 45). Likewise, reliance of stress on loudness as one of its cues is parasitic in Hayes’ (1995: 7) sense because in phonology the primary function of loudness is the organization of segments into syllables. Thus there is the famous Sonority Sequencing Principle (see, e.g., Gandour 1998: 210; Hall 2006: 330; Yavaş 2011: 142; Colina 2012: 137), which explains why English has words such as plate , blame , clue , glue , fly , sly but not * lpate , * lbame , * lkue , * lgue , * lfy , * lsy . The latter are illicit because the Sonority Sequencing Principle requires that onsets that are made up 36 2 Theoretical preliminaries of multiple segments be characterized by an increasing sonority. That is, the following sound must be more sonorous (i.e., louder) than the preceding one. The words plate , blame , clue , glue , fly , sly are in compliance with this requirement because the following sound / l/ is more sonorous than the preceding sounds / p/ , / b/ , / k/ , / g/ , / f/ , / s/ . In the asterisked words, by contrast, the more sonorous sound / l/ is followed by the less sonorous / p/ , / b/ , / k/ , / g/ , / f/ , / s/ , which violates the Sonority Sequencing Principle. What matters, however, in connection with stress is that just like vowels may exhibit a particular duration irrespective of whether they occur in stressed syllables or not, so is every sound characterized by a particular degree of loudness irrespective of whether the syllable in which that sound occurs bears stress. According to Yavaş (2011: 135), the two main factors that determine the intrinsic sonority of a sound are the vibration of the vocal folds (voicing) and the degree of opening of the vocal tract during the articulation of that sound. The loudest sounds are obviously low vowels, such as, e.g., / æ/ , / ɑ: / , / ɒ/ (Yavaş 2011: 135). Their articulation involves the maximum opening of the vocal tract and like all vowels, they are voiced. The least sonorous sounds are voiceless plosives, such as / p/ , / t/ , / k/ (Yavaş 2011: 136). Their articulation involves a complete obstruction to the airflow, and, additionally, the vocal folds do not vibrate. Finally, the parasitic nature of reliance on pitch as a cue of stress is perhaps even more obvious because apart from its association with stress, pitch is to a much larger extent associated with intonation, one of whose aspects is tone. Traditionally, tones are classified into falling tones, rising tones, and combinations of falling and rising tones. (A more refined classification will also distinguish between low and high falling and rising tones). As pointed out above, a falling tone in English is “associated with finality, completeness, and definiteness” (Gut 2009: 122), and this is precisely the reason why such tones occur in words cited in isolation (Cruttenden 2008: 282). To summarize, every syllable is characterized by a particular pitch height, a particular duration, and a particular degree of loudness. These can be measured relatively easily with a speech analysis software, such as Praat (Boersma & Weenink 2014). But as the values that syllables exhibit with respect to these parameters might be caused by things other than stress, stress contours, as Hayes (1995: 8) argues, cannot be “read off” from the phonetic record. That is, when we measure syllables’ fundamental frequencies, durations, and intensities, the values we obtain reflect not only the syllables’ stress contours, but also things such as the historical / combinatorial / syllabic length of English vowels, the intrinsic loudness of segments constituting a syllable, or the intonational patterns employed by a speaker when uttering a particular word containing a particular syllable, etc. The latter factors do of course also contribute to the acoustic effect 2.4 Stress placement across the globe 37 of the stressedness / stresslessness of a syllable, but they are, strictly speaking, not directly related to stress. 2.4 Stress placement across the globe Enlarging on what was stated in the previous sections, we note that with regard to phonetic realization of stress, languages are traditionally classified into pitch-accent languages and stress-accent languages. Pitch-accent languages rely on (typically a high) pitch as their primary cue, whereas in stress-accent languages other cues such as duration, intensity, and vowel quality play an important role as well (see, e.g., van der Hulst 2010b: 12). According to Hyman (1977: 37-38), Japanese is a prototypical example of a pitch-accent language, whereas English exemplifies a stress-accent language. As Gussenhoven (2006: 216) observes in this connection: Whereas the most prominent syllable in a word in English is longer than other syllables and has a less reduced vowel, in addition to being marked with tone in focused pronunciation, […] the accented syllable of a Japanese word stands out virtually exclusively because it is associated with pitch features. (Gussenhoven 2006: 216) As regards stress placement, world’s languages are often classified into fixedstress languages and variable-stress languages (van der Hulst 2010b: 33). An example of the former is Dyirbal, in which primary stress falls on the first syllable in a word (van der Hulst et al. 2010: 672). The latter category (in addition to being exemplified by contemporary English) is also exemplified by Modern Greek, in which stress falls upon one of the last three syllables in a word (van der Hulst et al. 2010: 803). The category of fixed-stress languages can be further divided into languages in which stress assignment rules count from the beginning vs. the end of the word. For example, in Dyirbal, in which stress falls on the first syllable in a word, stress assignment can be said to be left-bound, counting from the beginning of the word (its left edge), whereas, e.g., Macedonian is a right-edge fixed-stress language, i.e., stress in Macedonian falls upon the antepenultimate syllable (van der Hulst et al. 2010: 807), which is the third syllable counting from the end of the word. As van der Hulst (2010b: 33) observes, so far typological studies on stress have discovered only three stress location possibilities for both leftand right-edge languages with fixed stress. That is, in left-edge languages, stress can fall only on the first three syllables (stress in such languages can thus be said to be initial, pen-initial, or post-pen-initial). Similarly, in right-edge languages stress can fall 38 2 Theoretical preliminaries only on the last three syllables (i.e., the final syllable or the ult, the pre-final syllable or the penult, or the pre-pre-final syllable or the antepenult). There seem to be no left-edge fixed-stress languages in which stress is placed on the fourth syllable counting from the beginning of the word. Likewise, there seem to be no right-edge fixed-stress languages in which stress is placed on the pre-antepenult (i.e., the fourth syllable from the end of the word). Just like fixed-stress languages, languages with variable stress can be further classified into several subcategories. On the one hand, there are languages such as the above mentioned Modern Greek, where the location of stress is “completely unpredictable and thus lexically marked” (van der Hulst 2010b: 42), i.e., as pointed out above, stress in a word of Greek falls upon one of the last three syllables in a word. On the other hand, there are languages with variable stress in which variability can be attributed to what has traditionally been called the weight of a syllable. With regard to this parameter, syllables are usually classified into heavy and light syllables, but note that an answer to the question of which syllables count as heavy and stressable and which as light and unstressable varies from language to language and, what is more, typically within one and the same language, syllables are in different positions required to satisfy different conditions in order to be eligible to receive stress. A case in point is Modern High German, in which the placement of primary stress is said to obey the following rules: 1) final stress if the vowel in the ult is long or there are two coda consonants, 2) penultimate stress if Condition 1 is not fulfilled and if the penult contains a coda consonant, and 3) antepenultimate stress in all other cases (van der Hulst et al. 2010: 803). It is not difficult to see that word-finally, Modern High German only treats syllables whose rhymes exhibit the structures V: or VCC , where C stands for a consonant, V: for a long vowel, and V for a short vowel, as stressable. Thus, “a word-final syllable must have more consonants to be counted as heavy, since word-final consonants are often extrametrical” (Hayes 1995: 59) i.e., invisible to metrical construction rules so that the rhyme structure VC is at the end of a word equivalent to the rhyme structure V and counts therefore as light rather than heavy. (By contrast, the rhyme structure VCC is word-finally equivalent to the rhyme structure VC and counts therefore as heavy.) In the penultimate position, by contrast, only syllables with a coda count in German as stressable, i.e., the length of the penultimate vowel does not play a role (with regard to which Modern High German is different from Present-day English, where a codaless penult also counts as heavy when it contains a long vowel). Finally, in the antepenultimate position, syllables in German are (just as in English) not required to exhibit a particular structure in order to be able to attract stress. Just like Present-day English, Modern High German is therefore also often referred to as a language with 2.4 Stress placement across the globe 39 Latin-like stress (Hayes 1995: 181; cf. Domahs et al. 2014: 80, who conclude that “German, Dutch, and English must be considered quantity-sensitive languages, with the three languages showing very similar patterns overall,” i.e., stress is penultimate when the penult is heavy and antepenultimate when the penult is light.) In some quantity-sensitive languages, the weight of the stressed syllable depends on factors other than the length of the vocalic nucleus and the presence / number of consonants in the coda position, which is typical of Modern High German. For example, in Kera stress is normally placed on a syllable that contains a long vowel, but if no such syllables occur, stress falls on the first syllable that is pronounced with a high tone (van der Hulst et al. 2010: 840). Similarly, in Serbo-Croatian stress is normally placed on the first syllable that is pronounced with a tone (van der Hulst et al. 2010: 814). In Shipibo-Conibo stress is placed on a second syllable if it ends in a consonant or a nasal vowel (van der Hulst et al. 2010: 782). Similarly, in Apurin- stress is placed on the final syllable when the vowel occurring in it is characterized by a nasal manner-of-articulation (van der Hulst et al. 2010: 775). Observe also that in Mam the stress placement rules normally assign stress to the last long vowel in a word, but when no such vowels occur, stress is placed on the last syllable that ends in a glottal stop (van der Hulst et al. 2010: 772). Controversies surround the issue of whether consonants occurring in the onset position have any effect on syllable weight. A popular, often repeated view is that this is not the case. According to Hayes (1995: 51; author’s boldface), “prevocalic segments in the syllables (i.e. onset segments) are prosodically inert: […] VC is prosodically equivalent to CVC and CCVC , V: [is equivalent] to CV : and CCV : , and so on.” As Hayes (1995: 51) adds, “[w]hile this claim is not fully valid at the observational level […], it is so well supported across languages that it serves as the central observation for formal theories of syllable weight.” The present monograph suggests, however, that the question of whether onset segments play a role in stress assignment should be split into two separate questions: 1) Is the presence of an onset consonant a necessary condition for a syllable to receive stress and 2) does the presence of an onset consonant increase the probability of a syllable receiving stress? As for Question 1, the answer to it varies from language to language. For instance, in Thaypan and Mbabaram, stress is placed on the first syllable with an onset, whereas Lamu-lamu, Western Aranda, and Alyawarr place stress on the first syllable if it contains an onset, otherwise stress in these languages is pen-initial (van der Hulst et al. 2010: 685, 680, 677, 669). Similarly, “[i]n Madimadi, an Australian language of New South Wales […], stress on words of three or more syllables is sensitive to the presence of onsets that have coronal consonants” (Davis 1988: 4). 40 2 Theoretical preliminaries By contrast, in the case of, e.g., the English language, the answer to Question 1 is a clear no, which is strongly supported by the following fact. In the entries of Merriam-Webster Online in which boundaries between syllables are marked by means of (-) (e.g., for animal , the dictionary gives the syllabified transcription / ˈa-nə-məl/ ), there are 51,090 onsetless syllables, i.e., those that begin with a vowel. Since 22,179 of these syllables (~43.41 %) bear either primary or secondary stress (e.g., the first syllable in animal ), we can say that a syllable beginning with a vowel is in the English language almost as frequently stressed as it is unstressed; the presence of an onset consonant is thus clearly not a conditio sine qua non of stress in the English language. At the same time, however, with regard to Question 2, it must be observed that of the 311,074 English syllables that (in the entries of the same dictionary that contain syllabified phonetic transcriptions) have at least one consonant in the onset position, 169,910 (~54.62 %) are syllables that bear either primary or secondary stress (e.g., / ˈka-pə-təl/ of capital ). Syllables with onsets are thus more frequently (~54.62 %) stressed in the English language than onsetless ones (~43.41 %), with this difference counting as statistically (hugely) significant: χ 2 (1) = 2,214, p < 0.000001. Similar results are reported by Kelly (2004) with regard to disyllabic English words. His finding is that percentages of initially-stressed disyllables “increase monotonically with the number of consonants beginning a word” (Kelly 2004: 233). Thus, of the disyllables lacking onset consonants in the penult, only 35 % are stressed initially (Kelly 2004: 233). By contrast, in the case of one, two, and three onset consonants (beginning the penult of a disyllabic English word), the corresponding percentages are 69 %, 83 %, and 98 % respectively (Kelly 2004: 233). Notice in this connection that in the English language, “prefixes are relatively common on words beginning with few or no consonants” (Kelly 2004: 234). E.g., inflame . However, as Kelly (2004: 235) reports, “the relationship between word onset and stress is not eliminated when prefixation is factored out” (author’s italics). The fact that, e.g., the verb excel is stressed / ɪkˈsɛl/ (OD) whereas the verb cancel is stressed / ˈkans(ə)l/ ( OD ) cannot then be attributed to the prefix status of ex in formations such as, for instance, ex-president or external (vs. internal ). What accounts for the stress difference between these two disyllabic verbs is that while the first syllable in excel lacks an onset consonant, the first syllable in cancel contains it; stress is therefore final in the former, while it is initial in the latter. As Kelly (2004: 237) concludes, statistical analyses of a large sample of disyllabic English words document a significant relationship between word stress and word onset characteristics. This relationship is both large in magnitude and broad in scope, as it permeates the English lexicon 2.4 Stress placement across the globe 41 and cannot be localized to a small set of atypical words. As such, it conflicts with many theories of English phonology which do not consider onset structure as relevant to stress assignment. (Kelly 2004: 237) Kelly (2004: 237-239) also studied the assignment of stress in pseudowords (also commonly referred to as “nonsense words” or “nonwords”), i.e., potentially possible but thus far non-existent words in the English language. Thus 20 participants expressed their judgments as to which syllable stress should fall upon in pairs of pseudowords such as bontoon - brontoon differing only with regard to the number of consonants occurring in the initial syllable’s onset position. The results obtained also “supported the hypothesis that onset patterns affect English stress assignments. The mean proportion of [initial-stress] assignments was .78 for CC pseudowords but only .61 for C pseudowords” (Kelly 2004: 238). That is, pseudowords such as brontoon are more likely to receive initial stress than pseudowords such as bontoon . Consider also the placement of stress in Russian, which is typologically a language with unpredictable stress: “Stress falls on the first syllable lexically marked for [stress], else on the first” (van der Hulst et al. 2010: 813). Notice, however, that of the total number of the 157,877 words (of any syllabic length) for which stress markings are provided in the online version of the Russian Orthographic Dictionary (http: / / tinyurl.com/ q6x5jj6, 11. 08. 2015), 1,028 are initially-stressed words beginning with a vowel, 18,189 are initially-stressed words beginning with a consonant, and 370 are initially-stressed words beginning with a / j/ -glide, which in the Russian orthography is associated with the letters е , ё , й , ю , and я . The / j/ -glide is often referred to as a semi-vowel because phonetically it is indeed more like a vowel in that its articulation does not involve a major obstruction to the airflow; phonologically, however, the / j/ -glide is more like a consonant in that it does not function as the nucleus of a syllable (see, e.g., Stadnik-Holzer 2009: 26). What is important here is that the 1,028 initially-stressed words beginning with a vowel constitute only ~3.82 % of the total number of the 26,907 Russian words beginning with a vowel. By contrast, the 18,189 initially-stressed words beginning with a consonant constitute ~14.04 % of the total number of the 129,582 Russian words beginning with a consonant. And the 370 initially-stressed words beginning with a / j/ -glide constitute ~26.64 % of the total number of the 1,389 Russian words beginning with a / j/ -glide. These differences all count as statistically significant: For a vowel vs. a consonant, χ 2 (1) = 2,159, p < 0.000001; for a vowel vs. a semi-vowel, χ 2 (1) = 1,464, p < 0.000001; for a consonant vs. a semi-vowel, χ 2 (1) = 179, p < 0.000001. Note also that 11,923 initially-stressed Russian words whose stressed syllables have only one consonant in the onset position constitute ~12.1 % of the total 42 2 Theoretical preliminaries number of the 98,538 Russian words beginning with a single consonant. However, 5,297 initially-stressed Russian words whose stressed syllables have two consonants in the onset position constitute ~18.57 % of the total number of the 28,529 Russian words beginning with two consonants and 881 initially-stressed Russian words whose stressed syllables have three consonants in the onset position constitute ~37.75 % of the total number of the 2,334 Russian words beginning with three consonants. Finally, 88 initially-stressed Russian words that have four consonants in the onset position (such consonant clusters are impossible in English, but they are possible in Russian: e.g., vstroennyj , “inbuilt”) constitute ~48.89 % of the total number of the 180 Russian words beginning with four consonants. These differences are, again, all statistically significant. For one vs. two consonants, χ 2 (1) = 790, p < 0.000001; for two vs. three consonants, χ 2 (1) = 496, p < 0.000001; for three vs. four consonants, χ 2 (1) = 8.76, p = 0.003. Thus, very similar to what Kelly (2004) could establish with regard to initial stress in English (disyllabic words), it appears that the probability of initial stress in Russian likewise increases with an increase in the number of consonants in the (initial syllable’s) onset position. At the end of this section, it must be also mentioned that in contrast to the languages discussed thus far, few world’s languages can be referred to as stressless or at least partially stressless languages. Partial stresslessness is characteristic of, e.g., Chinanteco, which has minimal pairs whose members differ from each other only with regard to the presence / absence of stress. As van der Hulst et al. (2010: 774) report, the form fi means “handle” when it is stressed, but it means “road” when pronounced without stress. Similarly, in Lak-Dargwa, “there are stressless words such as turlu ‘cloud’” (van der Hulst 2010a: 482). As for languages that completely lack stress, it must be observed that stresslessness is usually one of the possible analyses that can be found in the literature alongside analyses that do not deny the existence of stress. A classic case is modern French, whose stress system is analyzed differently. Sometimes it is argued that word stress in French is final unless the ult ends in a schwa, which leads to penultimate stress (e.g., Carpenter 2010: 367). The latter, however, does not occur very often because “word-final schwa in French is only pronounced under very specific circumstances nowadays” (van der Hulst 2010a: 462). Stress in Present-day French is therefore predominantly final. At the same time, however, it is also suggested that French lacks stress at the word level altogether (see, e.g., Niebuhr 2007: 175). Thus: when words appear on the edge of larger prosodic units, their edge syllables may be the anchor point for intonational tones […] or boundary phenomena (tonal or segmental) which create the perceptual sensation of these syllables being prominent. 2.4 Stress placement across the globe 43 The linking of intonational pitch movement or other properties to edges of phrases, and thus edges of words that are peripheral in phrases, may lead to the illusion of the words having primary ‘stress’, while, synchronically speaking, there is no primary word stress at all (which is one way of analyzing French ‘word’ stress […] (van der Hulst 2012: 1515; author’s italics) The analysis of French as a stressless language is much more intuitive given that French is often said to exemplify a syllable-timed language, “in which all syllables are of approximately equal length” (Marks 1999: 191). Although some authors question the often taken-for-granted syllable-timed nature of the French language—according to Wenk (1985: 158), “those syllables occurring in final position in rhythmic groups (or “phonological words”) in French are on average about twice as long as non-final syllables”—there is little doubt that syllables constituting words in French are considerably different than those that make up words in a language such as English: In French, unlike English, all the syllables are fully realized. The accented syllable is always the word’s last syllable and the differences in duration, energy, and pitch, are small between accented and nonaccented syllables. Finally, there is a high proportion of open to closed syllables in French […]. These parameters give an impression of syllable regularity for French, that can be distinguished from the stress-timed rhythmic pattern of English […] (Dehaene-Lambertz & Houston: 1998: 23) To conclude, because in a language such as French, “vowel reduction is much less evident” than it is in a language such as English (Payne et al. 2012: 204), the former can from the perspective of the latter be regarded as a stressless language, i.e., as, e.g., Jones (1917: xl) observes, “French and Hindustani […] make no use of lexically significant word-stress.” 44 3 Methodology 3 Methodology This chapter elaborates upon the methodological steps undertaken by the author to answer the research questions formulated in Chapter 1. 3.1 Dictionary-based study: OED As Berg (1999: 125) points out, “[i]n view of the fact that comprehensiveness is top priority, the most appropriate method [for studying stress variation in English] is a dictionary analysis.” Thus, between October 20, 2013 and December 20, 2013, the online version of the Oxford English Dictionary / OED comprised 273,404 entries. Each of these entries was manually opened by the author and those entries that were found to contain at least two within-varietal transcriptions differing from each other with regard to the location of the primary stress symbol (ˈ) were saved to the hard drive on the authors’ PC . The entries saved were then manually classified with regard to the parameter “variety of English in which the words to which these entries refer occur with variable stress.” Several .docx-files were created to accommodate the words with stress doublets in different varieties of Present-day English, including American English, British English, New Zealand English, South African English, Australian English, Scottish English, Canadian English, Irish English, and Caribbean English. The information that was copied and pasted into these .docx-files included the words themselves, the word classes to which they belong, and the relevant transcriptions, which, as pointed out above, were found to differ from each other with regard to the location of the primary stress symbol (ˈ). Finally, the words in each separate .docx-file were manually classified with regard to the syllables upon which stress interchangeably falls in these words (i.e., e.g., ult-penult, ult-antepenult, penult-antepenult, etc.). Studying stress placement variation in English using the OED as the source of the information of how words in English are pronounced has both advantages and disadvantages. As for the former, the OED defines itself as “the definitive record of the English language,” where the adjective definitive is used in the sense “exhaustive.” That is, the OED is a dictionary that aims at an exhaustive coverage of the English language, which is supposed to comprise all words that have ever existed in the language under consideration. With regard to pronunciation, especially the entries that in the online version of the dictionary 3.1 Dictionary-based study: OED 45 have been updated since the year 2000 always contain both British and American English phonetic transcriptions (and in the case of some words, it is even indicated how they are pronounced in other varieties alongside British and American English. Thus, as mentioned above, a manual search for words with stress doublets has yielded instances of within-varietal stress variation not only in British and American English, but also in Canadian, Australian, New Zealand, South African, Scottish, Irish, and Caribbean English varieties): The pronunciations given are those in use among educated urban speakers of standard English in Britain and the United States. While avoiding strongly regionally or socially marked forms, they are intended to include the most common variants for each word. The keywords given are to be understood as pronounced in such speech. (http: / / public.oed.com/ how-to-use-the-oed/ key-to-pronunciation/ , 30. 06. 2015) Of particular importance is also the fact that in the case of words with stress doublets the OED often indicates which stress pattern had appeared in the English language prior to another variant. Thus, for example, it is thanks to the OED we know that applicable was originally pronounced with initial stress as / ˈaplᵻkəbl/ , whereas the antepenultimately-stressed alternative / əˈplɪkəbl/ has existed in British English only since the late 19th century. Likewise, we know, thanks to the OED , that the penultimately-stressed / kənˈtɛmpleɪt/ had been the default pronunciation of contemplate “down to third quarter of [the] 19th cent[ury].” Sometimes the OED even attempts to explain why a particular English word has stress doublets in Present-day English. For instance, the word miscellany was in British English, similar to applicable , originally pronounced with initial stress but over the course of time came to be associated with the antepenultimately-stressed pronunciation / mᵻˈsɛləni/ ( OED ). In American English, by contrast, the word still vacillates between the stress patterns / ˈmɪsəˌleɪni/ and / məˈsɛləni/ , with the former stress pattern being, however, more frequent than the latter ( OED ). With regard to the pronunciation / mᵻˈsɛləni/ , the OED states that [t]he pronunciation with stress on the second syllable arises by analogy with the stress pattern of other four-syllable words in any (e.g. accompany v. , E piphany n. […], mahogany n ., etc.). Although it is not recorded in dictionaries until the early 20th cent., there are traces of some earlier occurrences […] It has never gained great currency in North America, but by the end of the 20th cent. had become the dominant variant throughout the rest of the English-speaking world. ( OED ) In summary, the OED is a very detailed English language dictionary, which lists multiple stress patterns (taking both British and American English into account) and sometimes even offers explanations for how a particular stress pattern could historically be arrived at (e.g., the factor of analogy). Additionally, it is some- 46 3 Methodology times pointed out which of the stress variants had historically been anterior to another variant. The OED is thus an indispensable research tool for anyone interested in the topic of stress variation in English from both synchronic and diachronic perspectives. As for disadvantages, it was already observed that some actually existing stress patterns are not given in the OED. E.g., similar to applicable , comparable is not only / ˈkɒmpərəb(ə)l/ ( OED ) but also / kəmˈpeərəbəl/ (Cruttenden 2008: 246); an example of within-speaker variation involving these two stress patterns is gc_ IZ -F8C14 (17. 11. 2016): The first comparable (00: 00: 20.190 --> 00: 00: 25.210) is stressed by the speaker / kəmˈpeərəbəl/ , while at 00: 00: 29.360 --> 00: 00: 33.039 and 00: 01: 02.690 --> 00: 01: 08.480, the stress pattern used is / ˈkɒmpərəb(ə)l/ . Similarly, it was mentioned in Chapter 1 that in the noun research , stress in Present-day English falls either upon the prefix re or upon the root search . According to the OED , however, this variation is exhibited by research only in the combination market research , which is / ˌmɑːkɪt rᵻˈsəːtʃ/ vs. / ˌmɑːkɪt ˈriːsəːtʃ/ in British English and / ˈˌmɑrkət ˈrisərtʃ/ vs. / ˈˌmɑrkət rəˈsərtʃ/ in American English. A fairly similar case is hotel and motel , which are both interchangeably pronounced with final and initial stress, but in the OED , this variation is said to be exhibited by motel only in the combination no-tell motel , which is / ˈnəʊˌtɛl ˈməʊˌtɛl/ vs. / ˈnɛʊˌtɛl məˈtɛl/ in British English and / ˈnoʊˌtɛl ˈmoʊˌtɛl/ vs. / ˈnoʊˌtɛl moʊˈtɛl/ in American English. The reverse situation is possible as well. In an attempt to serve as “the definitive record of the English language,” the OED sometimes gives stress contours that are no longer in use in contemporary English (cf. Gimson 1972: v, lamenting that “[t]oo often, English dictionaries persist in showing pronunciations which are seriously archaic”). For example, desiccate is / dɪˈsɪkeɪt/ vs. / ˈdɛsɪkeɪt/ in the OED , but of 31 native English speakers whose voices could be heard in YouTube videos containing the spoken occurrences of this verb, not a single one used penultimate stress in it; in full agreement with this finding, the OD gives for desiccate only the antepenultimately-stressed transcription / ˈdɛsɪkeɪt/ . A fairly similar case is decadent , which is / ˈdɛkədənt/ vs. / dɪˈkeɪdənt/ in the OED , but it is exclusively / ˈdɛkəd(ə)nt/ in the OD as well as in 171 YouTube videos in which decadent was found to have been pronounced by native English speakers 212 times. (Note, however, that if one stress pattern of a word is no longer used by English speakers, this fact must be accounted for in a comprehensive study of English stress, i.e., e.g., it is important to understand why desiccate and decadent are no longer stressed / dɪˈsɪkeɪt/ and / dɪˈkeɪdənt/ . The inclusion of archaic stress patterns in the OED is thus actually an advantage of this dictionary in comparison with more specialized pronouncing dictionaries, focusing as a rule on contemporary English pronunciations.) 3.2 Other resources and tools 47 Note finally that in the phonetic transcriptions given in the OED , there is unfortunately no indication of boundaries between syllables. It is therefore not possible to (automatically) retrieve words exhibiting a particular phonological structure, e.g., trisyllables in which stress is penultimate or disyllables in which stress is final. Since the object of the present study is all English words (rather than only words with stress doublets), the absence of this technical possibility has necessitated the use of other lexicographic resources alongside the OED . 3.2 Other resources and tools The Medical Research Council Psycholinguistic Database (henceforth MRC or MRC database) (http: / / ota.ox.ac.uk/ headers/ 1054.xml, 07. 02. 2016) contains “150,837 words and up 26 linguistic and psycholinguistic attributes for each” (Wilson 1988: 6). Among these attributes are syllabic length and stress pattern, and what is particularly important, for some of the items in the database, phonetic transcriptions are given in which boundaries between syllables are explicitly marked by means of the syllable boundary marker (/ ). Thus, knowing which phonetic symbols are used in the MRC to represent the sounds of English, the researcher can easily establish, e.g., how many English words whose syllabic length is at least three do indeed exhibit segmental structures that are in accordance with the Latin Stress Rule, i.e., a heavy penult, whose rhyme contains either a long vowel or a coda consonant, when stress is penultimate vs. a light penult, which ends in a short vowel, when stress is antepenultimate. For example, for casino , in which stress is penultimate, the MRC gives the syllabified phonetic transcription k@/ si/ n@U, whereas for algebra , in which stress is antepenultimate, the transcription given is &l/ d ZI / br@. Since the symbol (i), which we find in the penult / si/ of casino , is used in the MRC to represent the long / iː/ of, e.g., bead , whereas the symbol (I), which we find in the penult / d ZI / of algebra , is used in the database to represent the short / ɪ/ of, e.g., bid , we are justified in claiming (proceeding from the assumption that the assignment of stress in English abides by the Latin Stress Rule) that in the trisyllabic word casino stress is regularly penultimate (i.e., the penultimate syllable is heavy) whereas in the trisyllabic algebra stress is regularly antepenultimate (i.e., the penultimate syllable is light). For a full list of phonetic symbols used in the MRC to represent English sounds, see the Web page http: / / websites.psychology.uwa. edu.au/ school/ MRCDatabase/ uwa_mrc.htm (13. 03. 2016). Alternatively, to study stress pattern-segmental structure correspondences, the researcher can also use Merriam-Webster Online (henceforth MWO ) (http: / / www.merriam-webster.com/ , 07. 02. 2016), which is based upon the 11th edi- 48 3 Methodology tion of Merriam-Webster’s Collegiate Dictionary (2003). Similar to the phonetic transcriptions in the MRC , some of the transcriptions in the MWO likewise contain a syllable boundary marker, (-), and just like in the MRC , the phonetic symbols used in the MWO to represent the sounds of the English language are (sometimes) not the corresponding, usually used symbols of the International Phonetic Alphabet. For example, for casino the MWO gives the transcription / kə-ˈsē-(ˌ)nō/ , whereas for algebra the transcription given is / ˈal-jə-brə/ (for an explanation of the symbols used, see http: / / www.merriam-webster.com/ pronsymbols.html, 13. 03. 2016). Of particular importance is also the already mentioned Oxford Dictionaries / OD (http: / / www.oxforddictionaries.com/ , 07. 02. 2016), which comprises 201,079 entries (as of 31. 12. 2015). Since “Oxford Dictionaries focuses on current language and practical usage” (boldface mine)—the OD is thus essentially the same thing as the OED without, however, obsolete words and archaic stress patterns—the use of these two lexicographic resources has nicely complemented each other. Thus, for instance, for the verb concentrate , the OED gives the transcriptions / ˈkɒnsəntreɪt/ and / kənˈsɛntreɪt/ and states that “[t]he first-mentioned pronunciation, now prevalent, is recent.” In the OD , by contrast, we find only the transcription / ˈkɒns(ə)ntreɪt/ , which, coupled with the historical information provided in the OED, allows us to conclude that in the case of the stress patterns / ˈkɒnsəntreɪt/ and / kənˈsɛntreɪt/ , we are dealing not with a synchronic stress variation but with a diachronic stress shift, i.e., the stress pattern / kənˈsɛntreɪt/ has been (recently) abandoned by English speakers in favor of the stress pattern / ˈkɒnsəntreɪt/ . To distinguish between actually occurring and archaic stress patterns, also Longman Dictionary of Contemporary English (http: / / www.ldoceonline.com/ ), consisting (as of 09. 11. 2016) of 69,132 entries, can be relied upon. A major advantage of LDOCE in comparison with the other lexicographic resources named above is that 1,236 of its phonetic transcriptions contain the stress shift symbol (◀). E.g., academic is / ˌækəˈdemɪk◀/ ( LDOCE ), which means that the two possible stress patterns of this adjective are / ˌækəˈdemɪk/ (with, however, the secondary-stressed syllable / ˌæ/ being phonetically not different from the primary-stressed syllable / ˈde/ ) and / ˈækədemɪk/ , with the latter being used especially in combinations such as academic year or academic study , in which the head noun, modified by academic , is either a monosyllable or an initially-stressed polysyllable. Note also that although transcriptions in LDOCE do not contain a boundary symbol, the dictionary does give orthographic hyphenations, such as, e.g., a∙ble for able . Using these, the researcher can easily retrieve words exhibiting a particular syllabic length. Thus, for instance, if the boundary symbol (∙) occurs only 3.2 Other resources and tools 49 once, the word under consideration is a disyllable whereas in a trisyllabic word the same symbol would occur two times, which applies to, e.g., Ko∙re∙a of Korea . The 26,441 hyphenations given in LDOCE for 26,334 orthographically non-identical solidly-spelled items fall into 12,461 (~47.13 %) disyllabic items (i.e., hyphenations in which the boundary symbol occurs one time), 8,334 (~31.52 %) trisyllables, 4,003 (~15.14 %) tetrasyllables, 1,351 (~5.11 %) pentasyllables, 251 (~0.95 %) hexasyllables, 34 (~0.13 %) heptasyllables, six (~0.02 %) octasyllables, and one (very long! ) word of 18 syllables: Llan∙fair∙pwll∙gwyn∙gyll∙go∙ger∙y∙chwyrn∙dro∙bwll∙llan∙ty∙si∙lio∙go∙go∙goch, which is “a small village on Anglesey in North Wales, famous for being the place with the longest name in the UK ” ( LDOCE ). To find out how frequent a particular word is in contemporary English, the author used the British National Corpus / BNC . To be more precise, the author used 1) the now-defunct BNC Simple Search, which had been available at http: / / www.natcorp.ox.ac.uk/ , 2) the BYU - BNC , which is currently available at http: / / corpus.byu.edu/ bnc/ , and 3) the BNC XML edition, which can be downloaded for free at http: / / ota.ox.ac.uk/ desc/ 2554 (07. 02. 2016). Note also that 7,879 items in LDOCE are classified into high-, medium-, and lower-frequency words. Thus, for instance, of the 1,542 solidly-spelled high-frequency polysyllables, 55 (~3.57 %) have more than one stress pattern. E.g., the high-frequency word cigarette is / ˌsɪɡəˈret/ in British English (with the diminutive suffix ette —a cigarette is a small cigar —being emphasized via stress) and / ˈsɪɡəˌret/ vs. / ˌsɪɡəˈret/ in American English ( LDOCE ). By contrast, in the case of mediumand lower-frequency polysyllables, the corresponding percentages are ~5.44 % (111 / 2,042) and ~7.52 % (162 / 2,153). E.g., the medium-frequency word controversy is, according to LDOCE , in British English interchangeably stressed / ˈkɒntrəvɜːsi/ and / kənˈtrɒvəsi/ and the lower-frequency word barricade is both in British and American English interchangeably stressed / ˈbærəkeɪd/ and / ˌbærəˈkeɪd/ ( LDOCE ). For highvs. medium-frequency words, χ 2 (1) = 6.948, p = 0.008; for mediumvs. lower-frequency words, χ 2 (1) = 7.514, p = 0.006; for highvs. lower-frequency words, χ 2 (1) = 25, p < 0.000001. These differences are all statistically significant, which supports Berg’s (1999: 137) assertion that infrequency is a prerequisite of stress instability in English. Note, however, that if cases such as / ˌækəˈdemɪk◀/ (i.e., transcriptions in LDOCE that contain the stress shift symbol) are not counted as instances of stress variation, the corresponding χ 2 -statistics are as follows. For highvs. medium-frequency words (22 / 1,542 vs. 29 / 2,042), χ 2 (1) = 0.0003, p = 0.987; for mediumvs. lower-frequency words (29 / 2,042 vs. 48 / 2,153), χ 2 (1) = 3.809, p = 0.051; for highvs. lower-frequency words (22 / 1,542 vs. 48 / 2,153), χ 2 (1) = 3.115, p = 0.078. These differences are not statistically significant. 50 3 Methodology The point here is that the average syllabic length of a high-frequency polysyllabic English word is ~2.47, but the average syllabic lengths of mediumand lower-frequency polysyllables are ~2.76 and ~2.94 respectively. High-frequency polysyllables thus contain on average fewer syllables than medium-frequency polysyllables, which in turn are as a rule shorter than lower-frequency polysyllables. (This is because derived forms are as a rule used less frequently than their base words (Plag 2003: 111). E.g., the base government is, according to LDOCE , a high-frequency word, but the derivative governmental is a lower-frequency one. Derived forms have on average more syllables than base forms / morphologically simple words.) Accordingly, because mediumand lower-frequency words are on average longer than high-frequency words, the former have more chances of being stressed not only primary but also secondary. Of the 1,542 high-frequency polysyllables, only 102 (~6.61 %) are words such as cigarette , which have both primary and secondary stress. By contrast, in the case of the 4,195 mediumand lower-frequency polysyllables, the number of secondary-stressed words is 601: ~14.33 %. This difference is statistically hugely significant: χ 2 (1) = 62, p < 0.000001. To conclude, because secondary stress is more typical of mediumand lower-frequency words, stress shifts such as / ˌɡʌvəˈmentl◀/ ( LDOCE ), which involves the promotion of the secondary-stressed syllable / ˌɡʌ/ to the primary-stressed syllable (i.e., / ˈɡʌvəmentl/ ), are also more typical of mediumand lower-frequency words. The connection between stress variation and frequency of use is thus a very indirect one! For searches involving regular expressions, the software TextCrawler, Version 2.5.0.0 (DigitalVolcano Software 2013), and GNU grep, Version 2.5.4 (Free Software Foundation, Inc.: 2009) were used. For simpler searches (e.g., finding and eliminating duplicates, extracting strings beginning with a particular symbol), Microsoft Excel 2007 was relied upon. Regular expression or simply regex is essentially an advanced search option that is available in many computer programs, such as the above mentioned TextCrawler. One of the most powerful regular expressions is (|), which in combination with the parentheses matches alternatives, e.g., the search query (a|e|i|o|u|y) matches any of the vowel symbols used in the English alphabet, whereas the search query (b|c|d|f|g|h|j|k|l|m|n|p|q|r|s|t|v|w|x|z) will, by contrast, match any consonantal symbol. Another important regex is the curly braces {}, using which the researcher can indicate the number of occurrences of the string to be matched. For example, (b|c|d|f|g|h|j|k|l|m|n|p|q|r|s|t|v|w|x|z)(a|e|i|o|u|y) {2} will match a consonant being followed by two vowels (e.g., the string bea of bead ), whereas (b|c|d|f|g|h|j|k|l|m|n|p|q|r|s|t|v|w|x|z)(a|e|i|o|u|y){1,} will, apart from matching bea of bead , also match bi of bid , where there is only one vowel occurring immediately after a consonant. 3.2 Other resources and tools 51 Using these two simple regular expressions, it was established by the author that items such as / kəˈsiːnəʊ/ of casino ( OD ), in which stress is regularly penultimate (i.e., the penultimate syllable / ˈsiː/ contains a long vowel), and items such as / ˈaldʒɪbrə/ of algebra , in which stress is regularly antepenultimate (i.e., the penultimate syllable / dʒɪ/ ends in a short vowel), constitute ~67.15 % of the total number of the 19,545 items of three and more syllables in the case of which stress patterns and syllabified phonetic transcriptions are available in the MRC . Similarly, in the MWO dictionary, of the total number of the 22,653 antepenults bearing stress, 17,032 (~75.19 %) co-occur with light penults (i.e., cases such as / ˈaldʒɪbrə/ of algebra ), whereas in the case of the 14,961 stressed penults occurring in words of three and more syllables, only 4,110 (~27.47 %) can be referred to as light syllables; e.g., in / prə(ʊ)ˈhɪbɪt/ of prohibit , stress falls upon the light penult / ˈhɪ/ . English words of three and more syllables are thus by and large (segmentally) compliant with the provisions of the Latin Stress Rule: Penultimately-stressed trisyllables and longer words have as a rule heavy penults, whereas light penults normally occur when stress is antepenultimate. The question that remains to be answered is, however, whether the stress patterns that these words exhibit are indeed due to their segmental structures. Thus, it was pointed out in Chapter 1 that of the non-initially-stressed words in the OD dictionary, 67 % have segmentally longer righthand strings such as / -ˈhɪbɪt/ of prohibit , which occur in at least one other English word ( inhibit ). The location of stress in these words can thus also be seen as their root-prefix boundary location. To automatically identify words exhibiting a particular morphological structure (e.g., words beginning with a prefix / ending in a suffix), the online tool Morphological Analysis (https: / / open.xerox.com/ Services/ fst-nlp-tools/ Consume/ Morphological%20Analysis-176, 07. 02. 2016) was used. E.g., in the case of the derivative abbreviator the analysis returned by the tool is &lt; abbreviate&gt; or}+ Noun + Sg , i.e., the word is correctly analyzed as the product of suffixation of the verbal base abbreviate by means of the suffix or . Morphological Analysis even copes with derivatives in which the addition of a suffix is accompanied by segmental changes. For example, the adjective corrosive is correctly analyzed as the product of suffixation of the verbal base corrode by means of the suffix ive , i.e., &lt; corrode&gt; ive}+ Adj ; similarly, for secrecy the segmentation returned is &lt; secret&gt; cy}+ Noun + Sg . Relying upon base-suffix segmentations similar to these, we can automatically count, e.g., the percentage of suffixed derivatives in English whose stress pattern is the stress pattern of the corresponding base form (counting from left to right, i.e., from the beginning of the word). In the case of, e.g., 14,222 solidly-spelled items from the OD that the tool Morphological Analysis considers to 52 3 Methodology be suffixed derivatives (i.e., for these items, the tool has returned segmentations that contain the suffix symbol (})), this is true of no less than 9,833 (~69.14 %) derivatives. These include the following cases (which have been identified by the author with the help of the software Excel): 1) both the derivative and the base are stressed initially (e.g., / ˈabəsi/ of abbacy and / ˈabət/ of abbot ), 2) the derivative is stressed initially and the base is a monosyllable (e.g., / ˈfaktʃʊəl/ of factual and / fakt/ of fact ), 3) in the transcriptions under comparison, the primary stress symbol (ˈ) is followed by three identical symbols (e.g., / ˈɒlə/ of / bʌɪˈɒlədʒɪst/ of biologist and / ˈɒlə/ of / bʌɪˈɒlədʒi/ of biology ), which helps us find identically-stressed derivatives and bases in which stress is non-initial; additionally, with the help of this strategy, we do not miss cases such as / bʌɪˈaksɪəl/ of biaxial vs. / ˈaksɪs/ of axis , with the former being both a suffixed and prefixed derivative (the more identical symbols in the transcriptions under comparison occur after the primary stress symbol, the more likely it is that we are dealing with a genuine case of stress preservation. E.g., / ˌrɛfəˈrɛnʃ(ə)l/ of referential and / ˈrɛf(ə)r(ə)nt/ of referent share the string / ˈrɛ/ , which comprises two symbols occurring after the primary stress symbol. The strings under comparison do not, however, count as identical if the threshold is raised to three identical symbols (following the primary stress symbol): / ˈrɛn/ vs. / ˈrɛf/ ; the case of referential vs. referent is therefore not treated in the same way as, e.g., the pair biologist - biology ), 4) the location of the primary stress symbol (relative to the left transcription boundary) is identical in the transcriptions under comparison (e.g., in the transcriptions / prʌɪˈɒrətʌɪz/ of prioritize and / prʌɪˈɒrɪti/ of priority the primary stress symbol is not followed by three identical symbols (i.e., / ˈɒrə/ vs. / ˈɒrɪ/ ), but simply because the position in which this symbol occurs is the same in the transcriptions under comparison (i.e., in both / prʌɪˈɒrətʌɪz/ and / prʌɪˈɒrɪti/ the stress symbol (ˈ) is the sixth symbol relative to the lefthand (/ )), we have good reasons to assume that stress in the derivative prioritize falls upon the same syllable as in the base priority , and 5) if the position of the primary stress symbol in the base is subtracted from the position of the primary stress symbol in the derivative, the result is either one or minus one (e.g., in / vəːˈbəʊs/ of verbose the symbol (ˈ) occupies the fifth position (relative to the left transcription boundary) whereas in / vəˈbɒsɪti/ of verbosity the corresponding number is four). An obvious alternative to these strategies (especially 3-5) is to count the Levenshtein distance (Levenshtein 1966), i.e., the number of edits required to transform one string of symbols to another string. E.g., the Levenshtein distance between / ˈban/ of / əˈbandənm(ə)nt/ and / ˈban/ of / əˈband(ə)n/ is zero (because these strings are identical), but the Levenshtein distance between / ˈtəː/ of / dɪˈtəːmɪn/ and / ˈtɜː/ of / dɪˈtɜːmɪnəbl/ is one (because the nucleus of the stressed syllable is, according to the OD, non-identical in the transcriptions under comparison, even 3.2 Other resources and tools 53 though the stress pattern is the same (note that phonetically, / ˈtəː/ of / dɪˈtəːmɪn/ might be very similar to / ˈtɜː/ of / dɪˈtɜːmɪnəbl/ , but the only thing that matters now is that the symbol / ə/ of / ˈtəː/ is not identical to the symbol / ɜ/ of / ˈtɜː/ ); to obtain the stressed syllable / ˈtɜː/ of / dɪˈtɜːmɪnəbl/ out of the stressed syllable / ˈtəː/ of / dɪˈtəːmɪn/ , we should thus replace / ə/ through / ɜ/ ). If the Levenshtein distance between the strings under comparison (which should include the primary stress symbol being followed by at least three further symbols) is either zero or one (i.e., no or only one edit is required to obtain the primary-stressed syllable of the derivative out of the primary-stressed syllable of the base), the case under consideration is most likely an instance of stress preservation. Using these strategies, we can also establish whether a derivative, in addition to preserving the stress, also preserves the segmental structure of the base. Thus, for instance, because the Levenshtein distance between the strings / ˈbɒs/ of / vəˈbɒsɪti/ and / ˈbəʊ/ of / vəːˈbəʊs/ is two (i.e., the symbols / ə/ and / ʊ/ should be replaced through the symbols / ɒ/ and / s/ ) and because the location of the primary stress symbol relative to the left transcription boundary is not identical in the transcriptions under comparison, the pronunciation of the base form verbose can be said to be only partially preserved in the derived form verbosity . The transcription / ˈadm(ə)r(ə)lti/ of the derivative admiralty can, by contrast, be regarded as an instance of “agglutinative stress,” which means that “suffixes are simply hooked on, glued on, or ‘agglutinated’ to a word without influencing its structure” (Poldauf 1984: 50-51). Of the 14,222 suffixed derivatives in the OD, no less than 7,486 (~52.64 %) can be regarded as instances of agglutinative suffixation, which means that the transcriptions of these words’ base forms are part of the transcriptions of the corresponding derived forms; the usual case is when the transcription of a base form occurs in the beginning of the transcription of a corresponding derived form. E.g., the transcription / ˈadm(ə)r(ə)l/ of the base admiral occurs (in the very same form) in the beginning of the transcription / ˈadm(ə)r(ə)lti/ of the derivative admiralty . In the case of prefixed suffixed derivatives, however, the transcription of the base form occurs in the middle of the transcription of the corresponding derivative, which is true of, e.g., / ˈsteɪtsmən/ of / ʌnˈsteɪtsmənlʌɪk/ . In addition to attributing, e.g., the stress pattern / prʌɪˈɒrətʌɪz/ of the derivative prioritize to the stress pattern / prʌɪˈɒrɪti/ of the base priority , it can be argued that stress in the derivative is antepenultimate simply because the penultimate syllable / rə/ is light. Likewise, in the case of / ˈabəsi/ of abbacy , / bʌɪˈɒlədʒɪst/ of biologist , and / vəˈbɒsɪti/ of verbosity , the Latin Stress Rule also requires antepenultimate stress because the corresponding penultimate syllables / bə/ , / lə/ , and / sɪ/ are light (and, from a purely diachronic point of view, stress in the English word verbosity is antepenultimate because it is antepenultimate in the Latin 54 3 Methodology etymon word verbōsitās (Dictionary.com), which has a short vowel in the penult.) Indeed, of all items in the MRC database, 12,866 are, according to the tool Morphological Analysis, suffixed derivatives, i.e., for these items, the tool has returned segmentations that contain the suffix symbol (}). Stress indications and syllabified phonetic transcriptions are available in the MRC for 4,990 of these items whose syllabic length is no less than three. Of these 4,990 items, 1,329 have regular penultimate stress (e.g., / əˈbjuːzə/ of abuser ( OD ), where stress falls on the heavy penult / ˈbjuː/ ) and in 1,909 items, stress is regularly antepenultimate (e.g., / ˈaktɪvɪst/ of activist ( OD ), where the penult / tɪ/ is light). In other words, of the 4,990 suffixed derivatives in English whose syllabic length is at least three, 3,238 (~64.89 %) have stress patterns that are superficially in accordance with the Latin Stress Rule: Stress is penultimate when the penult is heavy and antepenultimate when the penult is light. Notice, however, that abidance by the Latin Stress Rule is more typical of trisyllabic suffixed derivatives: Of the 2,439 trisyllabic suffixed derivatives for which syllabified phonetic transcriptions and stress indications are given in the MRC , 1,081 are items such as / əˈbjuːzə/ of abuser , in which stress is regularly penultimate (i.e., the penultimate syllable is heavy), and 789 are items such as / ˈaktɪvɪst/ of activist , in which stress is regularly antepenultimate (i.e., the penultimate syllable is light). Thus, in the case of the 1,870 trisyllabic suffixed derivatives (~76.67 %), the location of stress can be said to be in compliance with the provisions of the Latin Stress Rule. By contrast, in the case of the 2,551 suffixed derivatives whose syllabic length is no less than four and for which syllabified phonetic transcriptions and stress indications are available in the MRC , the same is true of only 1,368 (~53.63 %) items: 1,120 are derivatives such as / əˈbɪlɪti/ of ability ( OD ), in which stress is regularly antepenultimate (i.e., the penultimate syllable / lɪ/ is light), and 248 are derivatives such as / ˌanɪkˈdəʊtl/ of anecdotal ( OD ), in which stress is regularly penultimate (i.e., the penultimate syllable / ˈdəʊ/ is heavy). The difference of 1,368 / 2,551 vs. 1,870 / 2,439 is statistically hugely significant: χ 2 (1) = 291, p < 0.000001. As for disyllables, the segmentations returned by Morphological Analysis allow us to conclude that the location of stress in a disyllabic English word (among other things) crucially depends upon whether it begins with a prefix vs. ends in a suffix. As for prefixed derivatives (i.e., those for which the tool Morphological Analysis has returned segmentations that contain the prefix symbol ({); e.g., {a&lt; live&gt; + Adj , for alive ), the proportions in the OD dictionary among the 23,147 initiallyvs. 4,585 finally-stressed disyllables are (55 / 23,147=)~0.24 % vs. (538 / 4,585=)~11.73, χ 2 (1) = 2,417, p < 0.000001, which counts as a statistically hugely significant difference. As for suffixed derivatives, the corresponding proportions are (2,800 / 23,147=)~12.1 % vs. (321 / 4,585=)~7 %, χ 2 (1) = 99, p < 0.000001, which also counts as an extremely significant difference. Final stress is 3.2 Other resources and tools 55 thus the preferred stress pattern of disyllables beginning with a prefix whereas disyllables ending in a suffix are, by contrast, more frequently pronounced in English with initial stress. (The latter claim is also supported by the fact that of the 1,918 disyllables in the MRC that the tool Morphological Analysis considers to be suffixed derivatives, 1,797 (~93.69 %) are stressed on their first syllables. E.g., lover is stressed / ˈlʌvə/ ( OD ). Final stress in disyllabic suffixed derivatives is thus virtually non-existent in the English language. It can be found in, e.g., / əˈmʌŋst/ of amongst ( OD ), whose base is the finally-stressed / əˈmʌŋ/ of among ( OD ), to which the suffix st , which on its own does not constitute a syllable, was added. Additionally, because some English suffixes bring about important semantic distinctions (e.g., lessor - lessee , standing for granters vs. holders of a lease), some English disyllables have final stress as an emphatic alternative to initial stress. In all other cases when the base of a disyllabic suffixed derivative is monosyllabic, stress in the derived form is always placed upon the only syllable constituting the base form, i.e., e.g., stress in lover falls upon the syllable constituting its monosyllabic base love .) Interestingly, initial stress is in English the preferred stress pattern not only of actual suffixed derivatives but also of disyllables such as, e.g., music , which is not segmentable into the base muse and the suffix ic the way, e.g., cubic is segmentable into the base cube and the suffix ic . Thus, of the 23,147 initially-stressed disyllables in the OD , 259 (~1.12 %) end orthographically in ic , and, what is particularly important, of these, only 12 (~4.63 %) are, according to the tool Morphological Analysis, actual ic -derivatives: calcic , centric , cubic , cyclic , cystic , fistic , metric , mythic , rhythmic , scenic , spheric , and splenic . By contrast, of the 4,585 finally-stressed disyllables, only five (~0.11 %) end orthographically in ic . E.g., the shorter word ridic is stressed / rɪˈdɪk/ ( OD ), preserving the stress of the longer word ridiculous : / rɪˈdɪkjʊləs/ ( OD ). Since the difference of 259 / 23,147 vs. five / 4,585 is statistically hugely significant— χ 2 (1) = 41, p < 0.000001—we are justified in claiming that initial stress is the preferred stress of an ic -disyllable irrespective of whether it is a genuine ic -derivative like cubic or a word like music ( comic , magic , panic , topic , etc.). The same is true of the orthographic strings a , al , an , and , ard , cy , dom , er , ess , et , ful , ia , ian , ie , ing , ion , is , ish , ite , less , let , ling , ly , ness , o , oid , ous , ry , ship , some , ty , ule , um , ward , way , and y , which, according to the tool Morphological Analysis, also occur in English as suffixes. The median χ 2 -statistic of these strings is 18.71 (with the minimum being ~3.961 (ule ) and the maximum ~296 (er )), which means that the median p -value is ~0.000015. Notice also that the total number of initially-stressed disyllables that end orthographically in these 36 strings is 10,748 and the total number of initially-stressed disyllabic suffixed derivatives (according to the tool Morphological Analysis) is only 2,800. 56 3 Methodology Thus, disyllables ending in these strings (e.g., any , finish , medal , river , topic , etc.) are (as a rule) stressed initially even when these strings are not actual suffixes. As for prefixed derivatives, notice also that with the help of the tool Morphological Analysis we can find emphatically-stressed English words. An example is subcategory , whose morphological structure is correctly analyzed by the tool as {sub&lt; category&gt. The stress pattern of this prefixed derivative is, according to the OD , / ˈsʌbkatɪɡ(ə)ri/ , which is emphatic stress falling upon the semantically important prefix sub -, i.e., a subcategory is a “secondary or subordinate category” ( OD ). Another similar case is the stress pattern / ˈkaʊntəprəˌpəʊz(ə)l/ of counterproposal ( OD ). Since also the counter -words countercheck , counterclaim , counterdemonstration , countermarch , countermove , counteroffensive , counterplot , counterpunch , counterspy , and countertenor are, according to the OD , all pronounced with primary stress falling upon the prefix counter -, we argue that the meaning inherent in this prefix, “movement or effect in the opposite direction” ( OD ), is emphasized in English via stress. Another semantic structure capable of giving rise to emphatic prefix stress is “considerably more / less than the base.” E.g., a megaton , which corresponds to one million tons, is stressed / ˈmɛɡətʌn/ ( OD ), and, similarly, a millisecond , which is one thousandth of a second, is stressed / ˈmɪlɪsɛk(ə)nd/ ( OD ). Similarly, the fact that archangel is, according to the OD , more frequently stressed / ˈɑːkeɪndʒ(ə)l/ than / ɑːkˈeɪndʒ(ə)l/ is neatly attributable to the fact that an archangel is “[a]n angel of greater than ordinary rank” ( OD ). Unfortunately, the tool Morphological Analysis, which copes well with identifying suffixed and prefixed derivatives, returns identical analyses for, e.g., the morphologically simple algebra (&lt; algebra&gt; + Noun + Sg ) and the compound battlefield (&lt; battlefield&gt; + Noun + Sg ). Given this fact, compounds were, when necessary, identified in a different manner. Using the formulae = LEFT , = RIGHT , and = LENGTH in Microsoft Excel 2007, an orthographic string constituting a word of the English language was cut into all possible substrings; e.g., battlefield was cut into the substrings b and attlefield , ba and ttlefield , bat and tlefield , batt and lefield , battl and efield , battle and field , battlef and ield , battlefi and eld , battlefie and ld , and battlefiel and d . Using the software grep, these resultant combinations were then searched for in the BNC . Since compounds in English are often spelled interchangeably (e.g., English Word Stress (Poldauf 1984) and English Word-Stress (Fudge 1984) are the titles of two different monographs on English word stress, which were both published in 1984), a solidly-spelled compound can be expected to have a separately-spelled alternative. Since of the possible segmentations of the orthographic string battlefield , only battle field makes sense from a semantic point of view, it can be expected to occur in a representative corpus of the English language, such as the BNC : Despite her 3.2 Other resources and tools 57 many affairs - one of her immortal duties being to entertain heroes slain on the battle field - she loved Odin dearly and wept tears of gold when she lost him . This approach produces more precise results than a simple orthographic analysis. Thus, for instance, also the word question happens to be orthographically segmentable into the strings quest and ion , which both exist in English as nouns. However, in contrast to the combination battle field of battlefield , the combination quest ion of question does not make sense from a semantic point of view and thus does not occur in the BNC . The word question is therefore not wrongly analyzed as a compound (but note that, e.g., the derived adjective notable , which is also not a compound, happens to be orthographically segmentable into the strings not and able , which occur in the BNC in the form not able 451 times. This method is thus not entirely unproblematic either). Naturally, to quickly find out whether the word under consideration is a compound, a lexicographer’s intuition can also be taken into consideration. For example, in Dictionary.com (which as of 07. 02. 2017 consisted of 237,057 entries), a compound analysis, such as, e.g., blue + print for blueprint , is given for 5,707 solidly-spelled English words. Thus, with the help of these two strategies (i.e., searches for separate occurrences in the BNC and compound analyses given in Dictionary.com), it was established by the author that of the 23,147 initially-stressed disyllables in the OD, no less than 2,477 (~10.7 %) are compounds. E.g., the disyllabic compound airline is stressed / ˈɛːlʌɪn/ ( OD ). By contrast, in the case of the 4,585 finally-stressed disyllables, this is true of only 145 (~3.16 %) words. E.g., the disyllabic compound backstage is stressed / bakˈsteɪdʒ/ ( OD ). Since the difference of 2,477 / 23,147 vs. 145 / 4,585 is statistically very significant— χ 2 (1) = 254, p < 0.000001—we are justified in claiming that compounds in English are indeed, as it is argued by many authors, more frequently leftthan right-prominent (e.g., Bell & Plag 2013: 130, reporting that “[i]n running speech, about one third of NN compounds […] are stressed on N2, while two thirds are stressed on N1 […]”). Finally, observe that the 6,227 monosyllables in the OD contain on average ~4.39 orthographic symbols. Given this number, it comes as no surprise that of the 6,198 initially-stressed disyllables (in the same dictionary) that contain eight and more orthographic symbols, 1,724 (~27.82 %) count morphologically as compounds (in accordance with the strategies described above). By contrast, of the 16,949 initially-stressed disyllables that contain no more than seven orthographic symbols, only 753 (~4.44 %) are compounds. The difference of 1,724 / 6,198 vs. 753 / 16,949 is statistically hugely significant— χ 2 (1) = 2,594, p < 0.000001—which means that disyllabic compounds should be mainly sought among disyllables that contain at least eight orthographic symbols. (A related finding is that disyllables containing eight and more orthographic symbols are more frequently 58 3 Methodology stressed initially than finally: 6,198 / 23,147(=~26.78 %) vs. 999 / 4,585(=~21.79 %), χ 2 (1) = 50, p < 0.000001. Because 1) disyllabic compounds are made up of two monosyllables, which contain on average ~4.39 orthographic symbols, and 2) compounds in English are more frequently leftthan right-prominent, words that contain eight and more orthographic symbols are more frequently stressed on their first syllables.) To establish whether a difference between words exhibiting different stress patterns counts as statistically significant, Pearson’s (1900) chi-squared test ( χ 2 ) was performed in the majority of the situations requiring such a comparison. Calculations were performed with the help of the software Minitab, Version 15. 1. 30.0 (Minitab Inc 2007) and Microsoft Excel 2007. Statistical significance is (regarded as) achieved when p < 0.05, but notice that the lower the p -value, the more significant the difference under consideration. Thus, especially in connection with the result p < 0.000001, which means that the χ -statistic should exceed the critical value of 23.92812699, this monograph will sometimes use the modifying adverb hugely (as well as highly / extremely / very ), i.e., e.g., This difference is statistically hugely significant . 3.3 YouTube-based study To determine the preferred stress patterns of words for which stress doublets are given in the OED , i.e., e.g., whether stress in paprika is more frequently penultimate than it is antepenultimate, the Web site YouTube was between May 2014 and May 2015 searched for the presence of videos in which words with variable stress are used by native English speakers (who pronounce these words in these videos). (With regard to this, the present monograph is different from pronouncing dictionaries, in which this question is as a rule answered on the basis of opinion polls conducted (either electronically or via traditional post) among a panel of native English speakers, who are requested to express their pronunciation preferences (e.g., Wells 2008: xviii). An obvious limitation of this method is that in the case of such opinion polls, the compiler of a pronouncing dictionary analyzes not actual English pronunciations but only impressions of people of how they pronounce English words. A corpus study taking actual language usage into account is thus undeniably more reliable.) As of today, it is still not possible to directly search for words occurring in YouTube videos (for especially automatic speech recognition, which is a prerequisite for this, is by no means a trivial task). However, it is possible to achieve the desired result indirectly. Many YouTube videos contain so-called “closed captions,” which, in the ideal case, represent the full transcripts of videos, i.e., 3.3 YouTube-based study 59 they document which time point in a video corresponds to which part of an utterance produced by (a) speaker(s) in that video. E.g., the following is an excerpt from the closed captions appended to the YouTube video ho VRH 670 BO c (01. 07. 2015): 1 00: 00: 00,820 --> 00: 00: 06,610 Okay, so this tutorial is an introduction to the blood supply to the abdominal organs. 2 00: 00: 06,610 --> 00: 00: 11,719 So I’m going to talk about the arterial supply and the venous drainage of the abdominal viscera. 3 00: 00: 11,719 --> 00: 00: 18,719 So we’re looking here at a model of the circulatory system and we’re looking at the abdominal 4 00: 00: 19,060 --> 00: 00: 20,160 region. 5 00: 00: 20,160 --> 00: 00: 27,160 So here you can see the abdominal aorta and to the right of it, you can see the inferior 6 00: 00: 27,470 --> 00: 00: 34,470 vena cava. So first, let’s just talk about the arterial supply to the abdominal viscera. Searching for words occurring in the closed captions appended to YouTube videos thus represents an indirect means of searching for words occurring in corresponding YouTube videos themselves. In light of this fact, it was first established by the author which YouTube videos with (English language) closed captions have occurrences of a particular English word with stress doublets. This was done using the command-line downloader youtube-dl (Gonzalez et al. 2006-), which can be ordered to execute multiple commands; e.g., when ordered --all-subs --skip-download “ytsearchall: paprika,cc”, youtube-dl downloads only subtitles (but not the corresponding videos files! ) of all captioned YouTube videos that have something to do with the search term “paprika.” It is very fortunate that captions appended to YouTube videos always have language-specific endings (e.g., .de for German, .fr for French, .ru for Russian, 60 3 Methodology etc.) and in the case of subtitles whose language is English, the corresponding endings are .en , .en- GB , .en- US , .en- CA , .en- IE , etc., i.e., the string . en is always part of the name of an English language subtitle file. Note, though, that the fact that the name of a particular subtitle file ends in . en does not ensure that the language of the corresponding YouTube video is (native) English and, similarly, the fact that the name of a particular English language subtitle file has a variety-specific ending, such as . en- GB , does not necessarily mean that the speaker in a corresponding YouTube video has a British English accent. Subtitles that were found to contain particular search items were then opened and it was manually checked by the author whether the senses in which these items are used there are identical to the senses given for these items in the OED . E.g., a YouTube search for invalid , whose pronunciation vacillates between the finally-stressed / ɪnvəˈliːd/ and the antepenultimately-stressed / ˈɪnvəliːd/ only when used in the sense “[i]nfirm from sickness or disease; enfeebled or disabled by illness or injury” (OED), yields many YouTube captions where invalid expresses the meaning “the opposite of valid .” This sense is, however, exclusively associated with the penultimately-stressed pronunciation / ɪnˈvalɪd/ ( OD ). Similarly, to find out whether Abba , “[a]n invocation to God as father” ( OED ), is in American English more frequently stressed / ˈɑbə/ or / əˈbɑ/ , the occurrences of Abba as the name of a popular Swedish pop band must be left out of consideration. Once youtube-dl finished downloading corresponding video files, i.e., those that are accompanied by English language closed captions found to contain a particular English word with stress doublets, it was checked by the author via playing random parts of those videos whether the speakers whose voices can be heard in them are indeed native English speakers. In addition to deleting videos featuring languages other than native English, the author also deleted English language videos featuring 1) computer-synthesized voices (which sound fairly unnatural and are therefore easily distinguishable from the voices of human speakers) and 2) the voices of native English speakers occurring in song videos. With regard to the latter, the reason for their exclusion is that singers are known to occasionally alter the usual location of stress in words (which is usually done for rhyming purposes). Hence a researcher cannot be entirely sure whether an occurrence of a particular stress pattern in a song does indeed reflect the singer’s genuine linguistic preferences. For example, in the 2009 song Do You Wanna Date My Avatar? (urNyg1ft MIU , 01. 07. 2015), stress in avatar is final. This is most likely because avatar in the preceding line Do you wanna date my avatar? rhymes with star in the following line She’s a star , i.e., the speaker normally pronounces avatar with antepenultimate stress, but rhyming with star necessitates exceptional final stressing in the song. Note, however, that since English speakers in general vacillate between stressing avatar finally and ante- 3.3 YouTube-based study 61 penultimately, i.e., / ɑːvəˈtɑː(r)/ vs. / ˈævətɑː(r)/ ( OED ), it is theoretically possible that the singer in the video urNyg1ft MIU pronounces avatar with final stress in other environments as well. The question of whether the former or the latter explanation holds true for the situation at hand cannot be answered only on the basis of the video urNyg1ft MIU (unless a video like this contains occurrences of avatar in which it is pronounced by the same speaker in other environments with antepenultimate stress). Another problem is posed by multiple videos featuring identical speakers. E.g., among captioned YouTube videos containing the spoken occurrences of antique are 25 videos in which this adjective is pronounced (with initial stress) by an identical British English speaker. Similarly, among captioned YouTube videos containing the spoken occurrences of perfume are 41 videos in which this adjective is pronounced (with initial stress) by an(other) identical British English speaker. Cases such as these might skew the finding of one stress pattern being more frequently used in contemporary English than an alternative stress pattern. To solve the problem, the author used the command --write-info-json in youtube-dl. This command writes video metadata (in the form of a .json-file), containing, among other things, the URL of a YouTube channel from which a particular video stems. Extracting such URL s from corresponding json-files, one can see how many YouTube videos have been contributed by a particular uploader. The solution practiced by the author was thus to delete all but one downloaded videos stemming from one and the same YouTube channel, reducing thereby the risk of counting identical speakers more than once. When all irrelevant videos were removed, a batch-file was prepared by the author ordering the command-line player ffplay ( FF mpeg developers 2000-) to bulk-play portions of the downloaded videos in which, in accordance with the timeframes (e.g., 00: 00: 06,610 --> 00: 00: 11,719) given in the corresponding subtitle file, a particular English word with stress doublets is supposed to be pronounced. While ffplay was playing the portions specified in the batch-file, the number of speakers using a particular stress pattern was counted by the author on the basis of his auditory impressions. Studying stress variation in English using YouTube as the source of the information of how English words are stressed has both advantages and disadvantages. As for the former, note that “300 hours of video are uploaded to YouTube every minute” (https: / / www.youtube.com/ yt/ press/ statistics.html, 01. 07. 2015). The platform is thus a very large (freely accessible) online database of spoken language, containing, among other things, 1) excerpts as well as sometimes full-episodes of TV shows, 2) movies, documentaries, and cartoon films (likewise as excerpts or sometimes as their full-length versions), 3) commercials, 4) music clips, 5) recordings of all sorts of public speeches, including, for instance, 62 3 Methodology university lectures and conference presentations, 6) audio books, 7) amateur videos featuring speakers in natural private settings, etc. As for disadvantages, YouTube is not a genuine linguistic corpus (because it has never been intended to be used that way; conducting phonetic research is clearly not the primary YouTube use the founders of the Web site had in mind when launching it in the early days of Web 2.0). Thus a researcher who will decide to utilize YouTube in a corpus-like manner will most likely end up using other pieces of software (i.e., those that go beyond their Internet browser, where YouTube is running) such as, for example, the aforementioned youtube-dl and ffplay. However, even with these programs, researching how particular words are pronounced by speakers in YouTube videos is an extremely time-consuming activity, which, as pointed out above, took a year of the author’s life. Another shortcoming concerns the inaccessibility of some of the actually available video content. As was mentioned in the beginning of this section, it is at present not possible to directly search for words occurring in YouTube videos (because we still do not have an effective automatic speech-recognition technology, which would convert in a matter of seconds millions of spoken words to searchable text). Thus only YouTube videos containing closed captions, which document what is spoken in a video during which timeframe, are accessible for phonetic research. By contrast, captionless videos, which might be as relevant as captioned ones, are thus completely left out of consideration (for the only means to access such videos is to listen to them from the beginning to the end. In light of the above mentioned fact that “300 hours of video are uploaded to YouTube every minute,” this is clearly an unrealistic task). In addition to this, many Internet users have noticed that the number of search results that is reported by Google in the first results page once a particular search has been run is not the actual number of search results that can be physically accessed by means of Google. As, e.g., Jack Schofield points out in a 2012 blog post: Bragging is a long-standing feature of Google’s search results, even though it has no practical benefits for users. The fact is, no matter how many results Google claims it has found, it is never ever going to show more than a thousand. (http: / / www.zdnet.com/ hello-world-google-search-claims-522-million-results-serves-up-934-7000008360/ , 01. 07. 2015) The same is also true of the YouTube Web site, which is a subsidiary of the Google company. For example, a YouTube search for “absolute” (with quotation marks) yields “about 2,690,000 results.” This, however, is only the number of results YouTube claims it has found in its database when “absolute” is searched for. In reality, though, when the user reaches results page Nr. 33, the fantastic 3.3 YouTube-based study 63 number of “about 2,690,000 results” miraculously changes into only “about 658 results,” with no further video results being loaded beyond the videos listed in the first 33 pages loaded. The same can be said for any other search term no matter how popular and frequent it might seem at first glance. For example, there are in reality no more than 661 accessible YouTube videos that have something to do with Barack Obama (or any other individual or thing named Obama ), even though the first results page of the search for “Obama” (with quotation marks) says that there are “about 11,700,000 results.” Given these facts, the following question must be answered: If, e.g., of 2,690,000 YouTube videos that have something to do with the search term “absolute,” only 658 are physically accessible (and of these, only 358 videos were found to contain actual spoken occurrences of the word under consideration), can research that takes only these accessible YouTube videos into account also produce reliable results? The answer to this question (argued for in this monograph) is yes. Observe, first of all, that this problem is relevant only for high frequency words, such as absolute . Recall, however, that words with stress doublets are also often represented by low frequency words. Thus a fairly large number of these words did not in total yield more than 1,000 search results, which is the maximum possible results number. As for high frequency words, observe, for instance, that the spoken occurrences of the word absolute were attested 2,945 times in the above mentioned 358 captioned YouTube videos that were made accessible by YouTube at the time when a corresponding YouTube search was run by the author. Notice now that the very same word absolute occurs only 2,055 times in the spoken component of the Corpus of Contemporary American English (henceforth COCA ), which its creators describe as “the only large and balanced corpus of American English.” What this means is that if it were possible to use the spoken component of the COCA corpus the way YouTube data can be used (i.e., being able to actually listen to the underlying sound files), we would end up with a comparable number of the total occurrences of absolute . Two more issues must be dwelled upon in this section: 1) legality of downloading YouTube videos by means of youtube-dl and 2) reliability of the results of this study, which, as explained above, are based upon the author’s auditory impressions of a particular English word being pronounced in a particular captioned YouTube video with a particular stress pattern. The first issue might arise because in accordance with YouTube’s Terms of Use: You may access Content for your information and personal use solely as intended through the provided functionality of the Service and as permitted under these Terms 64 3 Methodology of Service. You shall not download any Content unless you see a “download” or similar link displayed by YouTube on the Service for that Content. (https: / / www.youtube. com/ t/ terms? gl=US, 01. 07. 2015) In disagreement with this, however, German media lawyer Christian Solmecke argues in a video (http: / / praxistipps.chip.de/ download-von-youtube-videos-legal-oder-nicht_9496, 01. 07. 2015) that YouTube’s Terms of Use could only be violated in the legal sense if visitors of the YouTube Web site were required to explicitly accept these terms before being allowed to access content on the YouTube Web site. This, however, does not as a rule take place. To simply play a YouTube video, it is not necessary to accept (or even read) these terms of use. Downloading YouTube videos using third-party applications, such as youtube-dl, is thus completely legal notwithstanding what is written with regard to this matter in the corresponding paragraph of the Web site’s Terms of Use. As regards reliability, an obvious fact is that a stress placement study in which the location of stress in a sample of spoken data is determined exclusively by the author’s own auditory senses is somewhat subjective, especially if the author is not a native speaker of the language the placement of stress in which he is describing in that study. A good illustration of this claim is the stress system of the Indonesian language, which is analyzed differently by Western and Indonesian linguists. As van Zanten et al. (2010: 101) report, the former prefer the penultimate-stress analysis (i.e., Indonesian is a language with fixed stress on the penult), whereas the latter vacillate between fixed final stress and stresslessness, i.e., stress in Indonesian is, as Indonesian linguists think, either final or the language lacks stress altogether. Similarly: We must also acknowledge the fact that linguists may claim to perceive rhythmic beats that objectively are simply not present in the signal. If such perceptual beats are, or can be, biased by a linguist’s own language, this may result in attributing a rhythmic signature to the language being described that native speakers of that language would not perceive. (van der Hulst 2010b: 30) In light of such facts, in contemporary phonetic studies dealing with stress (e.g., Mompéan 2014), it is usually not the author of the study himor herself but rather two or three independent judges (preferably, native speakers of the language under study) who do the actual task of deciding which syllables in the words under analyses bear stress. Or, alternatively, author’s stress-location conclusions are Praat-supported, i.e., corresponding sound files are opened with a speech analysis software, such as Praat, which then determines the characteristics of syllables that are traditionally seen as the phonetic correlates of linguistic stress (pitch, duration, intensity, vowel quality, etc.). 3.3 YouTube-based study 65 It is clear, though, that these methodologies are suitable only for small-scale phonetic studies, similar to those of Mompéan (2014) and Henderson (2010), who studied the preferred stress patterns of only a handful of English words. The present monograph analyzed, however, the preferred stress patterns of 1,232 English words with stress doublets in the OED that were found to have been pronounced in captioned YouTube videos by at least one native English speaker. Among these words is, for instance, the noun content , which, according to the OED , is both / ˈkɒntɛnt/ and / kənˈtɛnt/ —“ con ˈ tent is historical, and still common among the educated, but ˈ content is now used by many, especially by young people; some make a difference, saying con ˈ tents , but cubic ˈ content ; and printers often use ˈ contents technically, while saying con ˈ tents generally” (OED)— but as the subsequent analysis of the corresponding YouTube data has revealed, is almost exclusively / ˈkɒntɛnt/ in Present-day English. The spoken occurrences of the noun content were attested 7,127 times (only one of which was found to be an occurrence of the stress pattern / kənˈtɛnt/ ) in 524 captioned YouTube videos. Analyzing the location of stress in the noun content in each of these spoken occurrences using either Praat or relying upon the acoustic impressions of several independent judges would have inhibited the already extremely time-consuming work described above and, in the case of independent judges, would have additionally cost the author thousands of Euros or U. S. dollars (which the author did not have), for no one would without a serious financial incentive have agreed to spend hours or even days listening to how content is pronounced in 524 YouTube videos, in which it occurs 7,127 times. Given this fact, an important question arises: How likely is it that the author could misinterpret the native English YouTube data in which the words the placement of stress in which he wished to analyze occur? To begin with, observe that the author’s native language is Russian, in which just like in English, word stress is mainly cued by duration, vowel quality, and loudness. There are, however, important differences. According to Sokolova et al. (1996: 123), The nature of word stress in Russian seems to differ from that in English. The quantitative component plays a greater role in Russian accentual structure than in English word accent. In the Russian language we never pronounce vowels of full formation and full length in unstressed positions, they are always reduced. Therefore the vowels of full length are unmistakably perceived as stressed. In English the quantitative component of word stress is not of primary importance because of the nonreduced vowels in the unstressed syllables which sometimes occur in English words […] Russian phoneticians […] insist on the quantitative character of the Russian word stress as its principal feature, though other components of word stress in Russian are not denied. (Sokolova et al. 1996: 123; cf. Pettersson 1982: 68, where quality rather than quantity is termed the decisive feature of a stressed syllable in Russian) 66 3 Methodology The two types of vowel reduction in Russian are the so-called akan’e and ikan’e , which essentially mean pronouncing unstressed vowels in particular positions as / ă/ and / ĭ/ (Byun et al. 2014: no pagination in the online-first version of the article). An example of akan’e is okno , “window,” which is normally pronounced in Russian / ăkˈno/ rather than / okˈno/ ; an example of ikan’e is časy , “clock,” whose standard Russian pronunciation is / čĭˈsy/ rather than / čăˈsy/ (from Byun et al. 2014: no pagination). Similarly, Chrabaszcz et al. (2014: 1468), who have recently compared how word stress is cued in English, Mandarin, and Russian, report that although the vowel quality cue was the strongest cue for all groups of listeners, pitch was the second strongest cue for the English and the Mandarin listeners but was virtually disregarded by the Russian listeners. Duration and intensity cues were used by the Russian listeners to a significantly greater extent compared with the English and Mandarin participants (Chrabaszcz et al. 2014: 1468). In other words, native speakers of Russian pay much more attention to duration and loudness than they do to pitch, which is an important cue (in accented positions) in English. Notice, however, that Chrabaszcz et al.’s (2014: 1468) test persons “performed a stress identification task on nonce disyllabic words” (emphasis mine). In contrast, the words studied in this monograph are not nonce, but established words, which had been used by English speakers more than once and were therefore included into one of the most famous English language dictionaries: The Oxford English Dictionary. For each of these words, phonetic transcriptions are given in the dictionary and it is these phonetic transcriptions (e.g., / ˈæbərənt/ vs. / əˈbɛrənt/ for aberrant , / ˈdɒktrɪnəl/ vs. / dɒkˈtraɪnəl/ for doctrinal , / ˈkɒntɛnt/ vs. / kənˈtɛnt/ for content , etc., where different stress patterns sometimes correlate with differences in the segmental makeup) that the author has used as guidelines for deciding which syllable in a word occurring in a YouTube video bears stress. In other words, the author’s task was not to determine the location of the stress in unfamiliar words, which he had never heard before, but only to establish a transcription-sound correspondence, i.e., to identify a pronunciation of a word in a particular YouTube video and to determine which pronunciation of the transcription given in the OED for a particular word with stress doublets it most resembles. Thus, even though it is of course possible that the location of the stress in some of the occurrences of these words could indeed be misinterpreted by the author due to his native language bias (i.e., paying more attention to duration and loudness than to pitch), it is unlikely that the results reported in this monograph misrepresent the tendencies of particular stress patterns as to occurring more / less frequently than corresponding alternative stress patterns 3.3 YouTube-based study 67 (e.g., the fact that paprika is more often pronounced with penultimate than with antepenultimate stress). This claim is strongly supported by the fact that in the case of 556 (~69.33 %) solidly-spelled words from the OED that also have an entry in the more synchronically-oriented OD , the first or the only stress pattern given in the latter dictionary is also the stress pattern that according to the author, was more frequently / exclusively used in captioned YouTube videos. E.g., Archimedean , which is / ɑːkɪˈmiːdɪən/ vs. / -miːˈdiːən/ in the OED , is only / ˌɑːkɪˈmiːdiən/ in the OD . In agreement with this, penultimate stress in Archimedean was heard by the author in only one (31r DP h_gqsg, 13. 07. 2017, at 00: 38: 06.889 --> 00: 38: 08.829) of the 10 captioned YouTube videos in which Archimedean is pronounced by native English speakers 14 times. 68 4 Stress assignment in English 4 Stress assignment in English Given the inventory of the stress assignment principles, which were summarized in 2.4, the question of why, in a particular English word, only a particular syllable counts as eligible to receive stress, can be answered in two different ways. First of all, it can be conjectured that the location of stress in English words is simply stored in English speakers’ mental lexica, i.e., English speakers simply learn words together with particular stress patterns (see, e.g., Domahs et al. 2014, who provide a critical discussion of this possibility). This view can be found in, e.g., Kettemann (1988: 329), who makes precisely this claim not only with respect to morphologically simple words but also even with regard to derived forms. Similarly, in more practically-oriented (i.e., applied) descriptions of contemporary English pronunciation, especially learners of English as a second or foreign language are frequently warned that “[i]n most cases there is no rule to the incidence of the stress […] It is therefore necessary for the foreign student to learn the stress of every word individually” ( Jones 1956; cited in Friederich 1967: 5). Similarly: “I know of no easy solution to the problem of stress in individual words; it must simply be learned, and dictionaries consulted” (Halverson 1967: 214). However, already Friederich (1967: 24-25) could establish that the assignment of stress in English words is largely in accordance with what he called “die Grundschemata des englischen Akzents” (i.e., the basic principles of English stress). Thus disyllabic English words are normally stressed on the first syllable, whereas in words of three and more syllables stress is normally antepenultimate (Friederich 1967: 24-25; in the work of Lionel Guierre and that of his followers (see, e.g., Trevian 2007: 427), these empirical facts are referred to as the Normal Stress Rule of the English language). Similarly, as Field (2003: 328) reports, “some 85.6 % of all content words in running speech are either monosyllabic or stressed on the first syllable.” Given these facts, the claim that stress assignment in English is entirely unpredictable (and thus lexically listed) does not seem convincing. Another approach to English stress is thus a rule-based approach, which is associated with a computational model of stress assignment. As, e.g., Halle & Keyser (1971: 11) observe in this connection, in the English lexicon, that is, in the list of words of the language that every speaker must at least in part commit to memory, there are no indications of stress. Instead, 4.1 Monosyllables 69 stress contours are computed as required with the help of rules […] (Halle & Keyser 1971: 11) Note, however, that some authors who speak of stress assignment rules in English do not deny the possibility of stress in English words also being lexically stored. This view is characteristic of Hayes (1995: 112), who mentions the lexicalized character of the English stress system […]: since English stress is lexically listed, the stress rules serve essentially as redundancy rules, describing the set of possible (as well as unmarked) patterns. A word has a well-formed stress contour if it is stressable by some derivation provided by the rules. (Hayes 1995: 112) Leaving aside the question of whether stress assignment in English is purely computational or stress rules are redundancy rules because the location of stress in English words is stored in English speakers’ mental lexica, let us now enlarge on the exact nature of the English stress rules (or stress principles). What are their provisions? The answer to this question will in the following be given with regard to English words exhibiting different syllabic and / or morphological structures: monosyllables, disyllables, words of three and more syllables, and concatenations of words (which are usually classified into compounds and phrases even though in the case of especially sequences of two nouns, an answer to the question of whether a combination such as, e.g., budget deficit is a compound or phrase is far from being obvious). 4.1 Monosyllables Since stress is a relational concept—i.e., if one syllable in a word is considered stressed, than the neighboring syllables co-occurring in the same word are considered either unstressed or stressed to a lesser degree than the syllable bearing primary stress—it is difficult to speak of stress in a monosyllabic English word, which consists of only one syllable (see, e.g., Poldauf 1984: 14 and Kenyon & Knott 1953[1944]: xxiv, who argue that monosyllables do not have word stress). At the same time, however, from a phonological point of view, there are good reasons to regard syllables of which monosyllabic English words are made up as syllables bearing stress. (For example, for the monosyllabic word cat , the MWO gives the phonetic transcription / ˈkat/ , where the word’s only syllable / ˈkat/ is said to bear primary stress.) A reason why a word such as cat , where there is only one syllable, can nonetheless be regarded as a word bearing stress is that the vowel in it is a qualita- 70 4 Stress assignment in English tively full vowel rather than a schwa, which is an important characteristic of stresslessness in English: Of the 6,229 phonetic transcriptions given in the OD for monosyllabic items, only 18 (~0.29 %) are transcriptions such as, e.g., / tə/ of to ( OD ), where the nucleus is a schwa. Notice also that with the exception of function words (e.g., a , the , to ), English monosyllables are never light (Hayes 1995: 88), i.e., of the 6,229 monosyllabic transcriptions in the OD , only 16 (~0.26 %) have codaless rhymes ending in a short vowel. English monosyllables are thus segmentally similar to finally-stressed English polysyllables, whose final syllables, which bear stress, are virtually never light: A light ult bearing stress occurs in the OD in only 23 finally-stressed (solidly-spelled) polysyllables, which are almost exclusively represented by poorly institutionalized foreign words. E.g., Moskva is stressed / maˈskva/ (OD), imitating how the word, which is the Russian name for Moscow , is pronounced in the source language Russian (which is a quantity-insensitive language). Finally, observe that 1,763 (~28.3 %) monosyllables in the OD exhibit the rhyme structure VC , i.e., e.g., / kat/ of cat , where a short vowel is followed by only one coda consonant; if monosyllables whose rhymes contain a short vowel that is followed by an affricate, i.e., e.g., / fɛtʃ/ of fetch ( OD ), are left out of consideration, this number drops to 1,686 items: ~27.07 %. The point here is that if word-finally, consonants in English were indeed extrametrical, as it is argued by some authors (e.g., Hayes 1995: 59), English monosyllables, in which stress falls upon the last syllable (i.e., the syllable that is not followed by any other syllable belonging to the same word), would only exceptionally exhibit the rhyme structure VC , which is exhibited by the monosyllabic word cat . Monosyllables such as cat constitute, however, almost 30 per cent of the English monosyllables. They are by no means exceptional in contemporary English, which is one of the reasons why the present monograph is inclined to reject the notion of consonant extrametricality (at least, as far as the English language is concerned), i.e., if the nucleus of the last syllable in an English word is a short vowel, that syllable should not have two coda consonants to count as eligible to receive stress. 4.2 Disyllables In the MRC database, syllabified phonetic transcriptions and stress indications are given for 12,882 disyllabic items. For 2,270 of these items the MRC gives the stress pattern 02, which means that the penult is unstressed whereas the ult bears stress—e.g., / əˈləːt/ of alert ( OD ) or / pəˈliːs/ of police ( OD )—whereas for 4.2 Disyllables 71 10,119 items the stress pattern given in the MRC is 20, which is the reverse of 02, i.e., the penult bears stress, whereas the ult is unstressed; e.g., / ˈeɪb(ə)l/ of able ( OD ) or / ˈrɪvə/ of river ( OD ). As for the remaining 493 items, the majority of them (483) are said to have the stress pattern 22, which means that both the ult and the penult bear some degree of stress. This is mainly true of disyllabic compounds (e.g., armchair , backside , checkmate , farewell , guildhall , etc.) as well as prefixed formations such as, for instance, distaste , inbreathe , misplaced , reload , undone , etc. Similarly, as pointed out in 3.2, the OD dictionary has 23,147 initiallyand only 4,585 finally-stressed disyllables. The former are thus considerably more numerous in contemporary English than the latter. Final stress is, as Poldauf (1984: 77) observes, “a stress pattern foreign to English,” which, in addition to the numbers just given, is also reflected in the fact that 398 (~8.68 %) finally-stressed disyllables from the OD have an entry in (the online version of the) Oxford Essential Dictionary of Foreign Terms in English (Speake & LaFlaur 2002). E.g., imam , which is a borrowing from Arabic, is stressed / ɪˈmɑːm/ ( OD ). By contrast, of the 23,147 initially-stressed disyllables, only 1,292 (~5.58 %) count (for Speake & LaFlaur 2002) as essential foreign terms in English. E.g., bureau , which is a borrowing from French, is stressed / ˈbjʊərəʊ/ (OD; cf. the OED, where bureau is said to be both / bjuːˈrəʊ/ and / ˈbjʊərəʊ/ , but in more than 1,000 spoken occurrences of this word in captioned YouTube videos, the author heard only initial stress). The difference of 398 / 4,585 vs. 1,292 / 23,147 is statistically highly significant— χ 2 (1) = 64, p < 0.000001—which allows us to say that a disyllabic word that is perceived by an English speaker as a foreign word is likely to receive final rather than initial stress. The fact that bureau is no longer stressed / bjuːˈrəʊ/ thus seems to suggest that this word no longer counts for English speakers as a foreign word. Indeed, of the 673 finally-stressed disyllables from the OD that according to Dictionary.com, were borrowed into English from (Old / Middle / Anglo-)French, 106 (~15.75 %) occur in Speake & LaFlaur’s (2002) dictionary. E.g., burlesque , caprice , finesse , gouache , naïve , etc. By contrast, in the case of the 1,280 initially-stressed importations from (Old / Middle / Anglo-)French, this is true of only 113 (~8.83 %) words, such as the above mentioned bureau . The difference of 106 / 673 vs. 113 / 1,280 is also statistically highly significant: χ 2 (1) = 21, p = 0.0000041. 4.2.1 Stress preservation According to the tool Morphological Analysis, 873 finallyand 1,332 initially-stressed disyllables from the MRC serve in English as bases of suffixed derivatives. E.g., the finally-stressed disyllable employ is, according to the tool 72 4 Stress assignment in English Morphological Analysis, the base of the suffixed derivative employment and the initially-stressed disyllable beauty is the base of the suffixed derivative beautiful . A difference between the 873 finallyand 1,332 initially-stressed disyllabic bases is that in the former group, 416 bases (~47.65 %) have fewer occurrences in the British National Corpus / BNC than at least one suffixed derivative containing them. E.g., the derivative employment has 10,684 attestations in the corpus while the base employ occurs only 1,706 times; the former is thus considerably more frequent than the latter. By contrast, in the case of the 1,332 initially-stressed disyllabic bases, the same is true of only 238 bases: ~17.87 %. E.g., the base beauty occurs in the BNC only 4,108 times whereas the derivative beautiful has 8,387 attestations. The difference of 416 / 873 vs. 238 / 1,332 is statistically hugely significant: χ 2 (1) = 224, p < 0.000001. Similarly, of disyllables that according to LDOCE, are in British English either exclusively or preferably pronounced with final stress, 629 occur in English as the base of at least one suffixed derivative (according to the tool Morphological Analysis). 24 (~3.82 %) of these bases are, according to LDOCE , less frequently used by contemporary English speakers than at least one suffixed derivative containing them. E.g., election is, according to LDOCE , a high-frequency word whereas elect is a medium-frequency one; assignment is a high-frequency word whereas assign is a lower-frequency one; frustrated , frustrating , and frustration are medium-frequency words whereas frustrate is a lower-frequency one; etc. (Cf. invest vs. investment , which occur in the BNC 1,572 vs. 10,853 times respectively (the longer noun investment can thus be said to be considerably more frequent in the corpus than the shorter verb invest ). In spite of this fact, however, in LDOCE both invest and investment are said to be medium-frequency words. This pair is therefore not analyzed in the same way as the pairs elect vs. election , assign vs. assignment , and frustrate vs. frustrated , frustrating , and frustration .) By contrast, in the case of (in British English either exclusively or preferably) initially-stressed disyllables from LDOCE , the corresponding proportion is 13 / 907(=~1.43 %). E.g., punishment is, according to LDOCE , a high-frequency word whereas punish is a medium-frequency one. The difference of 24 / 629 vs. 13 / 907 is also statistically significant: χ 2 (1) = 8.967, p = 0.003. The importance of these findings stems from the well-known fact that “[i]n general, there is a strong tendency for derived words being less frequently used than their base words” (Plag 2003: 111). (As Plag (2003: 111) explains, the reason for this is that “[b]eing semantically more complex, derived words tend to have a narrower range of meaning, to the effect that they cannot be used in as many contexts as their base words.”) Of the 12,866 items in the MRC that the tool Morphological Analysis considers to be suffixed derivatives, 10,179 (~79.12 %) have fewer occurrences in the BNC than the corresponding base form (according to 4.2 Disyllables 73 the tool Morphological Analysis); the mean of these base vs. derivative frequency differences is ~4,948.11 and the median is 744 (e.g., torture has 848 attestations, but torturous occurs only 10 times). The base of a suffixed derivative is thus (in the English language) as a rule considerably more frequent than the derivative. (Similarly, in LDOCE there are 1,776 suffixed derivatives (according to the tool Morphological Analysis) in the case of which a frequency-related characteristic—high-frequency, medium-frequency, lower-frequency—is given for both the base and the derivative. Of these derivatives, only 111 (6.25 %) are derivatives such as amazing , equipment , eraser , instruction , permission , etc., which, according to LDOCE , are in contemporary English more frequently used than the corresponding base forms amaze , equip , erase , instruct , permit , etc. E.g., instruction is a high-frequency word whereas instruct is a medium-frequency one). Accordingly, since unlike the majority of suffixed derivatives in English, e.g., the semantically related longer noun employment has considerably more attestations in the BNC than the shorter verb employ , the latter should from a synchronic point of view be seen as a back-derivative from the former: to employ = to give employment to somebody. What is important in this connection is that, as pointed out in 2.3, when English speakers shorten longer words, the stressed syllable of a longer input form is as a rule preserved as the stressed syllable of a shorter output form. Thus, for example, the verbs liaise and secrete are stressed / lɪˈeɪz/ and / sɪˈkriːt/ (OD) because they were back-derived from the nouns liaison and secretion , which are stressed / lɪˈeɪz(ə)n/ and / sɪˈkriːʃ(ə)n/ ( OD ). (By contrast, the verbs edit and injure are stressed / ˈɛdɪt/ and / ˈɪndʒə/ ( OD ) because they were back-derived from the nouns editor and injury , which are stressed / ˈɛdɪtə/ and / ˈɪn(d)ʒ(ə)ri/ ( OD ), i.e., the stress pattern of the longer input words is thus also preserved in the shorter output words. In general, however, back-derivation is, as suggested above, more typical in English of finally-stressed disyllables, especially verbs.) Similarly, we argue that because the shorter verb employ is synchronically secondary to the longer noun employment , final stress in the former can be regarded as the preserved penultimate stress of the latter, i.e., employ is pronounced / ɪmˈplɔɪ/ or / ɛmˈplɔɪ/ ( OD ) because employment is pronounced / ɪmˈplɔɪm(ə)nt/ or / ɛmˈplɔɪm(ə)nt/ ( OD ). (Penultimate stress in the base employment can be attributed either to the Latin Stress Rule (i.e., the penult / ˈplɔɪ/ is heavy) or to the prefixation analysis (i.e., the segmentally longer string ployment , which employment shares with deployment , is like a root and therefore attracts stress on to its first syllable).) Note also that [o]ne important semantic consideration is that the meaning of the derived form depends on the meaning of the base. This manifests itself on the level of the underlying 74 4 Stress assignment in English sentence […] and all later levels of analysis in the fact that the base appears in the paraphrase of the derivation. (Schmid 2011: 192) Given this, observe that for 99(/ 873=~11.34 %) finally-stressed disyllabic bases from the MRC , WordNet (semantic database) gives meaning definitions such as “be abundant” (= abound ), “make an announcement” (= announce ), “be attractive to” (= attract ), “have a beginning” (= begin ), “show a certain behavior” (= behave ), “have a discussion” (= discuss ), “become inflated” (= inflate ), “follow a procedure” (= proceed ), “make a proposal” (= propose ), “be or become joined or united” (= unite ), etc. Paraphrases such as these also support the analyses of the (less frequently used) disyllabic verbs abound , announce , attract , begin , behave , discuss , inflate , proceed , propose , and unite as back-derivatives from the (more frequently used) longer words abundant , announcement , attractive , beginning , behavior , discussion , inflated , procedure , proposal , and united . Final stress in the former can therefore be regarded as the preserved stress of the latter. E.g., the derived verb begin is stressed / bɪˈɡɪn/ ( OD ) because the base noun beginning is stressed / bɪˈɡɪnɪŋ/ ( OD ). (By contrast, in the case of the initially-stressed disyllabic bases, the corresponding percentage is (79 / 1,332=)~5.93 %. E.g., “have an argument about something” (= argue ), “frequently encountered” (= frequent ), “provide or equip with furniture” (= furnish ), “knowingly tell an untruth in a legal court and render oneself guilty of perjury” (= perjure ), “inflict punishment on” (= punish ), etc. The difference of 99 / 873 vs. 79 / 1,332 is statistically highly significant— χ 2 (1) = 21, p = 0.000005—which lends further support to the claim that back-derivation is more typical in English of finally-stressed disyllabic words.) That final stress in disyllabic English verbs is often the preserved stress of longer English words is also a reason why in the case of some English disyllables, stress differences accompany word class differences, i.e., an identically-spelled disyllable is stressed finally when occurring as a verb vs. initially when functioning morphosyntactically as a noun and / or adjective. For example, the string object occurs in the BNC 6,208 times, of which, however, only 969 are occurrences of object as a verb. To object is thus less frequent than objection (1,273) and can therefore be regarded as its back-derivative: to object = “express or raise an objection” (WordNet). The derived form to object is therefore stressed / əbˈdʒɛkt/ ( OD ) because the base form objection is stressed / əbˈdʒɛkʃ(ə)n/ ( OD ). The more frequently used an object is, by contrast, stressed initially: / ˈɒbdʒɛkt/ ( OD ). Similarly, the string project occurs in the BNC 15,265 times, of which, however, only 498 are occurrences of project as a verb. To project is thus less frequent than projection (603) and can therefore be regarded as its back-derivative: to project = “draw a projection of,” as defined by WordNet. The derived form to 4.2 Disyllables 75 project is therefore stressed / prəˈdʒɛkt/ ( OD ) because the base form projection is stressed / prəˈdʒɛkʃ(ə)n/ ( OD ). The more frequently used a project is, by contrast, pronounced with initial stress: / ˈprɒdʒɛkt/ ( OD ). As a verb, contract occurs in the BNC only 494 times (with the total number of occurrences of this string in the corpus being 11,884) and is thus less frequent than contractual (876), contracted (785), contractor (708), and contraction (514). The back-derivative to contract —“enter into a contractual arrangement” (WordNet)—is therefore pronounced with final stress, / kənˈtrakt/ ( OD ), whereas in the more frequently used a contract , stress is initial: / ˈkɒntrakt/ ( OD ). Another similar case is to second , which occurs in the BNC only 56 times and can thus be regarded as a back-derivative from secondment , which occurs in the corpus 184 times. The derived form to second is therefore stressed / sɪˈkɒnd/ ( OD ), preserving the stress of the base form / səˈkɒndm(ə)nt/ ( OD ). The considerably more frequently used noun / ordinal numeral second is, by contrast, stressed / ˈsɛk(ə)nd/ ( OD ). Absent themselves / himself / herself / oneself / myself has only 22 attestations in the BNC and is thus less frequent than absentia , which occurs in the corpus 41 times. The former is therefore pronounced / əbˈsent, æb-/ ( LDOCE ) because the latter is pronounced / æbˈsentiə/ ( LDOCE ). The considerably more frequently used adjective absent has, by contrast, regular initial stress: / ˈæbsənt/ ( LDOCE ). Note also that contemporary English does not have the words * limission and * vomission , but it has the words admission , commission / commitment , emission , omission , permission , remission , submission , and transmission , which are as a rule (~62.5 %) more frequent in the BNC than the finally-stressed mit -verbs admit , commit , emit , omit , permit , remit , submit , and transmit . Thus we argue that, e.g., transmit is stressed / trænzˈmɪt/ ( LDOCE ) because it is a back-derivative from transmission (which occurs in the BNC 1,463 times while transmit has only 322 attestations), but limit and vomit , which are not back-derivatives, have regular initial stress: / ˈlɪmɪt/ and / ˈvɒmɪt/ ( LDOCE ). In some cases, however, a less frequently used verb vacillates between final and initial stress. E.g., for to accent , the OED gives the British English transcription / -ˈsɛnt/ and the American English / ˈæk-/ vs. / -ˈsɛnt/ . In COCA , which is a corpus of contemporary American English, to accent is less frequent than accented (which, according to the OED , is stressed / -ˈsɛn-/ vs. / ˈak-/ in British English and / ˈæk-/ vs. / -ˈsɛn-/ in American English). The respective frequency numbers are 289 ( accent as a verb) vs. 894 ( accented ). These numbers suggest that to accent is a back-derivative from accented : to accent = to make something accented. At the same time, however, accent as a verb is considerably less frequent in COCA than accent as a noun: 289 vs. 6,881. These numbers suggest, by contrast, that to accent is a product of conversion of an accent , i.e., to accent = “utter with an accent” (WordNet). Stress in the derived form to accent should thus be the 76 4 Stress assignment in English preserved stress of the base form an accent . In this way, we obtain the variation between final and initial stress in to accent (and since an accent is more frequent not only than to accent but also than accented , the latter is likewise stressed interchangeably). According to the OED , this variation currently occurs only in American English, but since the corresponding frequency numbers from the BNC are fairly similar to those of the COCA corpus, we can be sure that it is only a matter of time before also British English speakers start stressing to accent / ˈak-/ , preserving thereby the stress of the more frequently used an accent . E.g., to progress is in British English interchangeably stressed / prə(ʊ)ˈɡrɛs/ and / ˈprəʊɡrɛs/ ( OED ). The former stress pattern is due to the analysis of to progress as a back-derivative from longer English words, i.e., the string progress occurs in the BNC 8,081 times, of which, however, only 613 are occurrences of progress as a verb; to progress is thus less frequent than progression (735) and especially progressive (1,763). At the same time, however, because one of the meanings of to progress is “[t]o make a progress” (OED), to progress also counts as a converted verb. There are therefore good reasons to pronounce to progress with the stress of the more frequently used a progress : / ˈprəʊɡrɛs/ . (Cf. to object and to project , which semantically do not have much to do with an object and a project . The stress pattern of the former is therefore not destabilized by the stress pattern of the latter. The same is true of second as a verb vs. second as a noun / ordinal numeral, but notice also that apart from the finally-stressed to second , which means “transfer an employee to a different, temporary assignment” (WordNet), there is also the initially-stressed to second , which means “give support or one’s approval to” (WordNet). Since the connection between these senses is far from being obvious, we can assume that the stress difference between these two identically-spelled verbs is only semantically-conditioned.) Similar cases of stress preservation can be easily found among disyllabic instances of other morphosyntactic categories (but note that in contemporary English, “the most productive process within [back-formation] is the formation of verbs from action nouns” (Stašková 2012: 104); e.g., gentrification → gentrify , “to attempt or accomplish the gentrification of ” (Stašková 2012: 104). Back-derived verbs are therefore more numerous in English than back-derived instances of other morphosyntactic categories). For example, the adjectives abrupt , adroit , exact , extreme , precise , profuse , sincere , and succinct are less frequent in the BNC than the corresponding adverbs abruptly , adroitly , exactly , extremely , precisely , profusely (as well as profusion ), sincerely , and succinctly (481 vs. 1,154, 34 vs. 42, 2,237 vs. 10,305, 3,626 vs. 6,632, 57 vs. 92 and 166, 457 vs. 1,114, 104 vs. 186; the mean of these differences is 1,579.75 and the median is 383). Similarly, in the BNC , the shorter word austere is less frequent than the longer word austerity (292 vs. 423), the shorter averse is less frequent than the longer aversion (142 4.2 Disyllables 77 vs. 196), the shorter conjunct is less frequent than the longer conjunction (six vs. 1,488), the shorter discrete is less frequent than the longer discretion (489 vs. 1,870), the shorter distinct is less frequent than the longer distinction (3,158 vs. 4,088), the shorter diverse is less frequent than the longer diversity (1,311 vs. 1,394), the shorter humane is less frequent than the longer humanity (355 vs. 1,199), the shorter inert is less frequent than the longer inertia (292 vs. 345), the shorter intent is less frequent than the longer intention (1,847 vs. 4,653), the shorter rotund is less frequent than the longer rotunda (43 vs. 53), etc. Notice also that it is only the adjective august , “of or befitting a lord / profoundly honored” (WordNet), that is stressed / ɔːˈɡʌst/ ( OD ), preserving the penultimate stress of / ɔːˈɡʌstəs/ of Augustus ( OD ), whereas the noun august (month of the year) is stressed / ˈɔːɡəst/ ( OD ). Etymologically, both adjectival and nominal uses of august go back to the first Roman emperor Augustus ( OD ), but what matters from a synchronic perspective is that it is only the adjective august that is less frequent in the BNC than the proper noun Augustus : 94 vs. 369. A fairly similar case is / ˈhjuːmən/ of human vs. / hjʊˈmeɪn/ of humane ( OD ), of which only the latter has fewer attestations in the BNC than humanity , which is stressed / hjʊˈmanɪti/ ( OD ). The analysis of humane as a back-derivative from humanity is also supported by the definition “pertaining to or concerned with the humanities,” given for one of the senses of humane in the WordNet database. An example of a finally-stressed back-derived noun is dement —“A person suffering from dementia” (OD)—which is stressed / dɪˈmɛnt/ (OD), preserving the stress of its base / dɪˈmɛnʃə/ ( OD ). Similarly, the noun caprice , which is stressed / kəˈpriːs/ (OD), occurs in the BNC only 36 times whereas the adjective capricious , which is stressed / kəˈprɪʃəs/ ( OD ), has 114 attestations. An objection that can be raised against the back-derivation analysis is that of the aforementioned finally-stressed disyllables, some do not preserve the segmental structure of the corresponding more frequently used longer word(s). E.g., we have a short vowel in the stressed antepenult of / hjʊˈmanɪti/ of humanity , but the stressed ult of / hjʊˈmeɪn/ of humane contains a diphthong. To account for this, recall that, as we established in 3.2, of the 14,222 suffixed derivatives in the OD , 7,486 (~52.64 %) exemplify agglutinative suffixation, which means that the derived form preserves not only the stress but also the segmental structure of the base. At the same time, however, we also established that in 9,833 (~69.14 %) suffixed derivatives in the OD , stress falls upon the same syllable (counting from left to right) as in the corresponding base form. Since these two numbers—7,486 and 9,833—are not identical, it is clear that stress preservation cannot be equated with agglutinative suffixation: A derived form may preserve the stress of its base form, without, however, preserving each and every phonetic segment constituting it. E.g., the shorter word cadaver is, 78 4 Stress assignment in English according to the tool Morphological Analysis, the base of the longer word cadaverous . This analysis is strongly supported by the fact that the former occurs in the BNC 43 times whereas the latter has only 13 attestations in the corpus. In spite of this fact, however, the base cadaver is, according to the OD , pronounced / kəˈdɑːvə/ or / kəˈdeɪvə/ , with the nucleus of the stressed syllable being filled by either a long vowel or diphthong, but the derivative cadaverous is pronounced only / kəˈdav(ə)rəs/ ( OD ), with the nucleus of the stressed syllable containing a short vowel; to account for cases similar to this, it is usually assumed that “[i]n stressed antepenultimate syllables followed by a stressless penult, long vowels are subject to so-called ‘trisyllabic shortening’: e.g. sāne ~ sănity , serēne ~ serěnity ” (Bermúdez-Otero & McMahon 2006: 386). As for the question of why the stressed syllable of / hjʊˈmeɪn/ of humane contains a diphthong in the nucleus, the answer to it is that there simply exists an association between the orthographic form ane (occurring at the end of a word) and the phonetic realization (of this string) as / eɪn/ : Of the 201 words in the OD that end orthographically in ane , 174 (~86.57 %) also have / -eɪn/ at the end of the corresponding phonetic transcription(s) given in the dictionary. At the same time, however, of these 174 ane -words, only 19 (~10.92 %) are finally-stressed words such as humane , in which the phonetic string / -eɪn/ , in which these words end, receives stress. In the majority of the English ane -words, this string does not receive stress even when it is phonetically realized as / -eɪn/ . E.g., membrane is initially-stressed as / ˈmɛmbreɪn/ ( OD ) even though the ult in it, just like the ult in / hjʊˈmeɪn/ of humane , also contains the diphthong / eɪ/ ; the 57 initially-stressed ane -disyllables in the OD are mainly represented by chemistry-related ane -formations such as, e.g., methane (← methyl + ane ): Just like in other disyllabic suffixed derivatives in English, stress in them is supposed to be initial, irrespective of the phonetic realization of their orthographic final string ane as / -eɪn/ . Notice also that both finallyand initially-stressed disyllables sometimes preserve not primary but (as explained in Section 2.2, phonetically not different) secondary stress of a longer English word (Kreidler 1979: 29). E.g., the stress pattern / ɪɡˈzam/ of exam ( OD ) is the preserved secondary stress of / ɪɡˌzamɪˈneɪʃ(ə)n/ of examination ( OD ) and similarly, because the longer word Alexander is stressed / ˌalɪɡˈzɑːndə/ (OD), the shorter word Alex is stressed / ˈalɪks/ , with the secondary stress of the former becoming primary stress in the latter. The shorter words exam and Alex are not back-derivatives but shortenings of the longer words examination and Alexander . The difference between back-derivation and shortening is that the former process gives rise to (a semantically more complex) new lexeme (i.e., e.g., while the base form dementia means “dementia,” the derived form dement means “a person suffering from dementia”; 4.2 Disyllables 79 the meaning of the latter is thus more complex than the meaning of the former), the latter process yields an allolex of an already existing lexeme. That is, Alex is simply a shorter alternative to Alexander and, similarly, the meaning of the shorter word exam is one of the meanings of the longer word examination : “a set of questions or exercises evaluating skill or knowledge” (WordNet). Note also that etymologically, many finally-stressed disyllables appeared in English not via back-derivation but rather as shortenings (of especially Latin words). E.g., the above mentioned sincere is historically not back-derivative from sincerely but shortening of the Latin sincērus (Dictionary.com), in which stress is penultimate because the vowel in the penult is long; likewise, antepenultimate stress in the Latin permittere , which has a short vowel in the penult, is the diachronic reason why to permit in English has final stress. Cf. exist , for which different etymological analyses are given in Dictionary.com vs. the OD: The former regards the verb under consideration as a shortening of the Latin ex(s)istere (where, just as in permittere , stress is antepenultimate because the vowel in the penult is short), whereas the latter analyzes exist as “probably a back-formation from existence ” (which also makes sense from a synchronic point of view: The shorter exist occurs in the BNC 5,356 times whereas the longer existence has 6,472 attestations; additionally, the meaning of exist is, according to WordNet, “have an existence”). The last number that must be presented in connection with stress preservation is that of the 4,585 finally-stressed disyllables in the OD , 50 (~1.09 %) lack orthographic vowels. E.g., RP is stressed / ɑːˈpiː/ ( OD ). By contrast, in the case of the 23,147 initially-stressed disyllables, the same can be said about only nine (~0.04 %) words. E.g., Jr is stressed / ˈdʒuːnjə/ ( OD ). The difference of 50 / 4,585 vs. nine / 23,147 is statistically hugely significant: χ 2 (1) = 194, p < 0.000001 (with Yates’ correction, because one of the expected values is smaller than 10). The point here is that stronger stress in an English phrase falls as a rule upon its rightmost primary-stressed syllable. Thus, as, e.g., Hayes (1995: 368) observes, “[a] common but not invariant pattern across languages is for syntactic phrases to receive final prominence.” E.g., Re ˌ ceived Pronunciˈation ( LDOCE ). Accordingly, when a bi-constituent phrase such as Received Pronunciation is abbreviated to an initialism such as RP , the latter is pronounced with final stress: / ɑːˈpiː/ ; words such as RP , which do not contain orthographic vowels, are therefore more numerous in English among finally-stressed disyllables. (The orthographic abbreviation Jr is, by contrast, phonetically in no way different from junior .) 80 4 Stress assignment in English 4.2.2 Segmental length Apart from shortening of/ back-derivation from longer English words, stress in a disyllabic English word may be final due to the already mentioned principle of not placing primary stress upon an element that counts morphologically as a prefix (unless there are reasons to emphasize the meaning inherent in it). This stress principle continues the Old English Stress Rule, which, as observed in Chapter 1, places primary stress in a polysyllabic word upon the first syllable of its root or stem. Recall also that, as argued in Chapter 1, prefixed derivatives in English are not necessarily formations such as, e.g., regain , whose segmentation into the prefix re and the base gain is fully transparent both formally and semantically; additionally, gain is considerably more frequent in the BNC than regain : 12,300 vs. 598. Stress in the disyllabic derivative regain is therefore placed upon the only syllable constituting its monosyllabic base gain , i.e., / rɪˈɡeɪn/ ( OD ). Consider, however, the nature of final stress in the verb / noun approach , i.e., / əˈprəʊtʃ/ ( OD ), which, just like, e.g., approve , seems to be made up of the prefix ap and the root proach . Note, however, that whereas the righthand component prove of approve occurs in English as a separate word and, in addition to this, is in the BNC considerably more frequent than approve —14,329 vs. 1,037—the righthand component proach of approach does not occur in the BNC and, similarly, in the corresponding OD entry referring to it, it is said to be “[n]ow rare (archaic and poetic in later use).” The justification for the segmentation of approach into the bound components ap and proach is thus the existence of the formally similar word reproach , which shares with approach the ult . proach . The segmentation ap - + proach necessitates, however, the treatment of either the former or the latter as the root of approach , for, as pointed out in Chapter 1, the root is the only obligatory element in a word; a word may be affixless, but it may not be rootless. Since the components ap and proach of approach are not associated with discernible meanings of their own, it is only the segmental length principle (which states that the root should contain more phonetic segments than the affix) that can be relied upon to answer the question of which of these two components counts in approach as the root. Since the ult / ˈprəʊtʃ/ contains more phonetic segments than the penult / ə/ , the former is clearly more like a root whereas the latter is more like a prefix. Stress in approach is therefore placed upon the only syllable constituting the bound root proach . This analysis is strongly supported by the following numbers. Of the 2,270 finally-stressed disyllables in the MRC , 872 (~38.41 %) are HH-disyllables, i.e., not only the ult bearing stress but also the penultimate syllable, which is unstressed, are heavy. E.g., / əbˈsəːd/ of absurd ( OD ), with both the ult / ˈsəːd/ and 4.2 Disyllables 81 the penult / əb/ counting as heavy. At the same time, however, of the 10,119 initially-stressed disyllabic transcriptions, 4,141 (~40.92 %) exhibit the structure HH, i.e., e.g., / ˈrɛskjuː/ of rescue ( OD ), with both the penult / ˈrɛs/ , bearing stress, and the unstressed ult / kjuː/ counting as heavy. Although this difference does achieve statistical significance— χ 2 (1) = 4.845, p = 0.0277 (with the p -value being, however, not far away from the threshold of 0.05)—it is nonetheless clear that in general, the phonological weight is a very poor predictor of the location of stress in a disyllabic English word: In disyllabic English words in which both syllables are heavy, final stress is almost as frequent as initial stress: ~38.41 % vs. ~40.92 %. What distinguishes the 872 HH -disyllables such as absurd from the 4,141 HH -disyllables such as rescue is the segmental length of the ult in comparison with the corresponding length of the neighboring penult. Of the 872 HH-disyllables, 624 (~71.56 %) have segmentally longer ults (i.e., those that contain more segments than the corresponding penults), whereas in the case of the 4,141 HH-disyllables, the same can be said about only 1,996 items: ~48.2 %. This difference is, in contrast to that of 872 / 2,270 vs. 4,141 / 10,119, statistically hugely significant: χ 2 (1) = 158, p < 0.000001. Applying the notion of weighted arithmetic mean (see, e.g., Ventcel’ 1969: 87-88), i.e., taking into account the number of occurrences of a particular value (e.g., the number of English disyllables in which the ult consists of, e.g., four segments), this length difference between HHand HH-disyllables can also be presented in the following way: In the case of the 872 HH-disyllables, the average segmental length of the ult is ~3.7 (symbols used in a corresponding MRC transcription; e.g., for absurd the MRC gives the phonetic transcription @b/ s3d, in which the ult / s3d consists of three symbols. MRC transcriptions do not employ the duration symbol (: ) and are therefore particularly suitable for analyzing the segmental length: One symbol in an MRC transcription normally corresponds to one sound, with two symbols being only used to represent diphthongs and affricates, which can be regarded as bi-segmental sounds) and the average segmental length of the penult is ~2.6; the difference between the segmental length of the former and that of the latter is thus 1.1. As for the 4,141 HH-disyllables, the average segmental length of the ult is ~3.5 and the average segmental length of the penult is ~2.99; the length difference is thus only 0.51. The ult of a finally-stressed disyllable in English thus contains on average more phonetic segments than the ult of an initially-stressed disyllable (3.7 vs. 3.5), whereas the penult of an initially-stressed disyllable has, by contrast, more segments than the penult of a finally-stressed disyllable: 2.99 vs. 2.6. This is the reason why Kelly’s (2004) study has demonstrated that in disyllabic English words, an increase in the number of onset consonants in the penult correlates statistically with an increase in the incidence of initial stress: The 82 4 Stress assignment in English fewer consonants in the onset of the penult, the shorter the penult in comparison with the ult, the more reasons English speakers have to regard the latter as root and the former as prefix. By contrast, the more consonants in the onset of the penult, the longer the penult in comparison with the ult, the fewer reasons English speakers have to regard the latter as root and the former as prefix. Note also that of the 2,270 finally-stressed transcriptions in the MRC , only 341 (~15.02 %)—the minority—are transcriptions such as / bɪˈɡɪn/ of begin and / ədˈmɪt/ of admit ( OD ), in the case of which the rhyme of the ult, bearing stress, is made up of a short vowel being followed by only one consonant. Similarly, in the OD dictionary, the number of such words among the 4,625 finally-stressed disyllables is 925 (20 %). The fact that among finally-stressed disyllabic English words, disyllables such as / bɪˈɡɪn/ and / ədˈmɪt/ are outnumbered by disyllables such as / əˈdapt/ ( OD ), in which there are at least two consonants following a short vowel in the rhyme of the ult, creates the illusion of word-final consonants being extrametrical in English, i.e., it is argued by some authors that the rhyme structure VC counts word-finally as light rather than heavy; to count as heavy and stressable, a final syllable must therefore either contain a long vowel or be closed by at least two consonants. As pointed out in the previous section, the present monograph rejects this claim. The rhyme structure VC is word-finally as heavy and stressable as it is in the penultimate position; e.g., the final syllables / ˈɡɪn/ of begin and / ˈmɪt/ of admit are no less heavy than the final syllable / ˈdapt/ of adapt . That finally-stressed words such as begin and admit , whose ults exhibit the rhyme structure VC , are outnumbered in English by finally-stressed disyllables such as adapt , in which the structure of the rhyme of the ult is VCC , is a consequence of the fact that the assignment of stress in disyllabic English words is by and large conditioned morphologically (rather than phonologically): The more consonants in the coda of the ult, the longer the ult relative to the penult, the more reasons English speakers have to regard the latter as root and the former as prefix. By contrast, the fewer consonants in the coda of the ult, the shorter the ult in relation to the penult, the fewer reasons English speakers have to regard the former as root and the latter as prefix. This is also the reason why in contrast to the structure HH , among the instances of which final stress is, as reported above, almost as frequent as initial stress, in disyllables exhibiting the structure LH (light penult + heavy ult), final stress is considerably more frequent than initial stress. Of the 2,270 syllabified phonetic transcriptions that are given in the MRC for finally-stressed disyllables, 1,209 (~53.26 %) are transcriptions such as / əˈləːt/ of alert and / pəˈliːs/ of police , which exhibit the structure LH . By contrast, in the case of the 10,119 transcriptions whose stress pattern in the MRC is 20, the same can be said 4.2 Disyllables 83 about only 1,588 transcriptions: ~15.69 %. For example, in / ˈʃadəʊ/ of shadow ( OD ), the penult / ˈʃa/ , which bears stress, is light, whereas the ult / dəʊ/ , which is unstressed, is heavy. This difference (i.e., 1,209 / 2,270 vs. 1,588 / 10,119) is statistically hugely significant: χ 2 (1) = 1,497, p < 0.000001. (In metrical phonology, the structure LH is referred to as the canonical iamb, which “is quantitatively unbalanced, and contains a light syllable followed by a heavy syllable” (Kager 1995: 438). Iambs are different from trochees in that an iambic foot ends in a stressed syllable, whereas a trochee begins with it (see, e.g., Alber 2006: 541) and the standard definition of the foot is that of “a grouping of syllables that contains exactly one stressed syllable” ( Jensen 1993: 77). Notice, however, that as Domahs et al. (2013: 272) point out, “[…] it remains still unknown whether feet are cognitively valid concepts.”) An explanation for the finding that the stress pattern LH is more frequent in English than LH is (thus not that the former gives rise to a canonical iamb but) that light syllables contain on average fewer phonetic segments than heavy syllables, i.e., a light syllable ends in a short vowel, but a heavy syllable contains a long vowel / diphthong / triphthong or / and a coda consonant. That is why, of the 1,209 LH-transcriptions in the MRC , 1,113 (~92.06 %) have (heavy) ults consisting of more phonetic segments than the corresponding (light) penult: The weighted arithmetic means are ~3.62 segments in the ult vs. ~1.78 segments in the penult. Stress in disyllables such as these should naturally be final rather than initial. Due to the very same reason, final stress is, as established in 4.1, in general non-existent in English when the ult of a polysyllabic word is light. Because 1) roots are segmentally longer than affixes and 2) light syllables are segmentally shorter than heavy ones, a light syllable occurring at the end of a polysyllabic English word is segmentally too short to count morphologically as its root. Notice also that of the 10,119 initially-stressed transcriptions in the MRC , only 2,954 (~29.19 %) are transcriptions such as the above mentioned / ˈʃadəʊ/ of shadow , in which the penult / ˈʃa/ , which bears stress, is light. Further similar examples include / ˈɛni/ of any ( OD ) and / ˈrɪvə/ of river ( OD ), with stress in these disyllabic words likewise falling upon the light penults / ˈɛ/ and / ˈrɪ/ . What we see is that, on the one hand, in stark contrast to final stress, initial stress in a disyllabic English word does occur even when the syllable upon which it falls is a light one (i.e., given that almost 30 per cent of initially-stressed disyllables in English have light penults, it would be counterintuitive to regard stress patterns such as / ˈʃadəʊ/ of shadow as exceptions.) At the same time, however, heavy penults among initially-stressed disyllables outnumber in English light penults—70 % vs. 30 %—because, again, as pointed out above, the longer the penult, the fewer reasons there are to regard the ult as the root. 84 4 Stress assignment in English An even more precise predictor of the location of stress in a disyllabic English word seems to be the orthographic length of the ult in comparison with the corresponding length of the neighboring penult. Consider, e.g., the stress difference between the verbs differ and defer : / ˈdɪfə/ vs. / dɪˈfəː/ (OD). This difference cannot be accounted for if we consider the segmental length of the ult / ˈfəː/ of defer in comparison with the segmental length of the penult / dɪ/ : With regard to this parameter, the finally-stressed / dɪˈfəː/ is not different from the initially-stressed / ˈdɪfə/ (i.e., since we are only concerned with the segmental length, we ignore the duration symbol (: ) in the transcription / dɪˈfəː/ .) This difference can, however, be accounted for if, instead of considering the phonetic transcription / dɪˈfəː/ , we consider the orthographic form defer . Assuming that defer is underlyingly de.fer while differ is underlyingly dif . fer (cf. Chomsky & Halle 1968: 83, arguing that umbrella is stressed / ʌmˈbrɛlə/ ( OD ) because under the orthographic syllabification um . brel . la , the penult, which contains a short vowel, is closed and counts therefore as heavy), we note that it is only defer where the ult is orthographically longer than the penult. Stress is therefore final in defer while it is initial in differ . (Initial stress in differ can, however, also be seen as the preserved initial stress of / ˈdɪf(ə)r(ə)ns/ of difference ( OD ), which has considerably more attestations in the BNC than differ : 11,167 vs. 1,806. To differ = to exhibit a difference, cf. “be different,” “be of different opinions,” which is how to differ is defined in the WordNet semantic database.) A similar case is the stress pattern / ˈʃadəʊ/ of shadow . Under the orthographic syllabification sha . dow , the ult and the penult in this word both consist of three orthographic segments; stress in shadow should therefore be initial. Indeed, of the 2,147 hyphenations given in LDOCE for disyllables that in British English are either exclusively or preferably pronounced with final stress, 1,866 (~86.91 %) are hyphenations such as, e.g., ab∙surd of absurd , whose righthand strings that follow the boundary symbol (∙) contain more orthographic segments than the corresponding lefthand string preceding it; the weighted arithmetic means are ~4.3 (ult) vs. ~2.4 (penult). By contrast, of the 10,312 hyphenations given in LDOCE for (in British English either exclusively or more frequently) initially-stressed disyllables, only 3,020 (~29.29 %) have segmentally longer ults (e.g., colleague , whose hyphenation in LDOCE is col∙league, is stressed / ˈkɒliːɡ/ ( LDOCE ) even though the ult in it contains more orthographic segments than the penult); the average values are ~3.24 (ult) vs. ~3.55 (~penult), which means that prototypically, the ult and the penult in an initially-stressed disyllabic English word consist of approximately the same number of orthographic symbols (which is what we find in hyphenations such as, e.g., res∙cue of rescue ). The difference of 1,866 / 2,147 vs. 3,020 / 10,312 is statistically hugely significant: χ 2 (1) = 2,476, p < 0.000001. 4.2 Disyllables 85 Whereas LDOCE gives hyphenations, such as ab∙surd and res∙cue, the OD provides line breaks, such as ab¦surd and res¦cue. The difference is that for, e.g., the trisyllabic artery the OD gives the line break ar¦tery, where the boundary symbol (¦) occurs only once, whereas in the hyphenation ar∙te∙ry, given in LDOCE , the boundary symbol (∙) occurs twice. Hyphenations in LDOCE are therefore more suitable for analyzing trisyllables and longer words. Of the 3,195 line breaks given in the OD for finally-stressed disyllables, 2,548 (~79.75 %) have orthographically longer ults; the average values are ~2.63 (penult) vs. ~4.35 (ult). As for the initially-stressed disyllables, the corresponding proportion is 4,140 / 13,219 (~31.32 %) and the corresponding average values are ~3.64 (penult) vs. ~3.62 (ult). The difference of 2,548 / 3,195 vs. 4,140 / 13,219 is also statistically hugely significant: χ 2 (1) = 2,500, p < 0.000001. Thus, we can reiterate that in finally-stressed disyllabic English words, the ult is orthographically longer than the penult, with the length difference between them approaching two orthographic symbols (which is what we observe in cases such as ab¦surd of absurd ). In initially-stressed disyllables, by contrast, the ult has as a rule the same orthographic length as the penult, which is what we observe in cases such as res¦cue of rescue . Notice also that in LDOCE , 2,092 initiallyand 770 finally-stressed disyllables are classified into high-, medium-, and lower-frequency words. Of the 262 high-frequency finally-stressed disyllables (e.g., about ), 232 (~88.55 %) have orthographically longer ults, but in the case of the 740 high-frequency initially-stressed disyllables (e.g., colleague ), this is true of only 227 (~30.68 %) words. This difference is statistically hugely significant: χ 2 (1) = 261, p < 0.000001. Likewise, of the 236 medium-frequency finally-stressed disyllables (e.g., extinct ), 213 (~90.25 %) have orthographically longer ults, but in the case of the 759 medium-frequency initially-stressed disyllables (e.g., evil ), this is true of only 223 (~29.38 %) words. This difference is also statistically hugely significant: χ 2 (1) = 271, p < 0.000001. Finally, of the 272 lower-frequency finally-stressed disyllables (e.g., eclipse ), 256 (~94.12 %) have orthographically longer ults, but in the case of the 593 lower-frequency initially-stressed disyllables (e.g., archive ), this is true of only 180 (~30.35 %) words. This difference is also statistically hugely significant: χ 2 (1) = 303, p < 0.000001. Abidance by the segmental length principle can thus be said to be equally characteristic of high-, medium-, and low-frequency disyllabic English words. Finally, observe that in contrast to the MRC , where any is e/ nI, finish is fI/ n IS , medal is me/ dl, river is rI/ v@, topic is t0/ pIk, etc. (i.e., in the syllabified transcriptions given in the MRC database, the medial consonants / n/ , / d/ , / v/ , and / p/ are in accordance with the Maximal Onset Principle (e.g., Yavaş 2011: 137) regarded as onset of the ult), LDOCE gives for the same words the hyphenations an∙y, fin∙ish, 86 4 Stress assignment in English med∙al, riv∙er, and top∙ic, where the same medial consonants are analyzed as coda of the penult. Disagreements such as these are due to the fact that apart from the Maximal Onset Principle, there is also, e.g., the morphemic principle, which requires that syllable boundaries be aligned with morpheme boundaries (Cruttenden 2008: 50); additionally, Cruttenden (2008: 50, 258) also mentions the phonotactic principle (“align syllable boundaries to parallel syllable codas and onsets at the ends and beginnings of words”) and the allophonic principle (“align syllable boundaries to best predict allophonic variation, e.g. the voicing of / r/ following / t/ ”). “Unfortunately such principles often conflict with one another” (Cruttenden 2008: 50), which is why we sometimes find syllabification disagreements such as fi . nish in the MRC vs. fin . ish in LDOCE . Given cases such as this, a question arises as to what extent compliance with the segmental length principle (in especially initially-stressed disyllabic English words) is due to the fact that the syllabification principles abided by the compilers of LDOCE as well as the OD are different from those of, e.g., the MRC database. The answer to this question is that of the 2,270 syllabified phonetic transcriptions given in the MRC for 1,935 orthographically non-identical finally-stressed disyllables, 1,882 (~82.91 %) have segmentally longer ults. By contrast, in the case of the 10,119 transcriptions given in the database for 8,879 orthographically non-identical initially-stressed disyllables, the same can be said about no more than 3,881 (~38.35 %) transcriptions. This percentage is, just as expected, higher than the corresponding percentages in both LDOCE (~29.29 %) and the OD (~31.32 %), but the difference of 1,882 / 2,270 vs. 3,881 / 10,119 still counts as statistically hugely significant: χ 2 (1) = 1,479, p < 0.000001. The segmental length principle is thus also in the MRC by and large abided by. Notice further that the 2,270 finally-stressed transcriptions in the MRC database fall into two (~0.09 %), 149 (~6.56 %), 857 (~37.75 %), 926 (~40.79 %), 286 (~12.6 %), 45 (~1.98 %), and five (~0.22 %) transcriptions in which the ult contains one-seven phonetic segments. The 10,119 initially-stressed transcriptions in the database fall, by contrast, into 104 (~1.03 %), 3,574 (~35.32 %), 3,742 (~36.98 %), 2,120 (~20.95 %), 528 (~5.22 %), and 51 (~0.5 %) transcriptions in which the ult contains one-six phonetic segments. For one segment (two / 2,270 vs. 104 / 10,119), χ 2 (1) = 19.3, p < 0.0001; for two segments (149 / 2,270 vs. 3,574 / 10,119), χ 2 (1) = 729, p < 0.000001; for three segments (857 / 2,270 vs. 3,742 / 10,119), χ 2 (1) = 0.475, p = 0.4907; for four segments (926 / 2,270 vs. 2,120 / 10,119), χ 2 (1) = 394, p < 0.000001; for five segments (286 / 2,270 vs. 528 / 10,119), χ 2 (1) = 165, p < 0.000001; for six segments (45 / 2,270 vs. 51 / 10,119), χ 2 (1) = 53, p < 0.000001; for seven segments (five / 2,270 vs. zero / 10,119), χ 2 (1) = 22, p < 0.00001. What we see, then, is that final stress is the preferred stress pattern of a disyllabic English word only when its ult contains no less than four phonetic segments, whereas initial 4.2 Disyllables 87 stress is the preferred stress pattern when the segmental length of the ult is one or two and in the case of disyllables whose ults contain three segments, initial stress is as frequent as final stress; there is no preferred stress pattern. Thus, for instance, analyzing the stress patterns / ˈarəb/ of Arab ( OD ) and / ˈvɪzɪt/ of visit ( OD ), we note that in contrast to LDOCE , which gives the hyphenations Ar∙ab and vis∙it (stress in these words is thus in accordance with the segmental length principle initial), the MRC gives the syllabified transcriptions &/ r@b and vI/ zIt, where the medial consonants / r/ and / z/ are analyzed not as coda of the penult but as onset of the ult. (In the MWO , visit is, just like in the MRC , only / ˈvi-zət/ , but for Arab , the dictionary gives both the syllabifications / ˈa-rəb/ and / ˈer-əb/ .) Note, however, that even if we proceed from the syllabifications A . rab and vi . sit rather than from Ar . ab and vis . it , the ults in these words still fail to achieve the segmental length of four symbols, which is prototypically exhibited in English by the ult of a finally-stressed disyllable. The disyllables Arab and visit are therefore stressed initially even though the ult in these words is segmentally longer than the penult. 4.2.3 Suffix-like endings Further deviations from the segmental length principle can be found among disyllables whose ults occur in English as suffixes. Thus it was reported in 3.2 that not only actual suffixed derivatives (such as active and lover ), which are almost exclusively (94 %) pronounced in English with initial stress, but also disyllables such as any , finish , medal , river , topic , etc., whose final syllables end orthographically in strings such as y , ish , al , er , ic , etc., which occur in English as suffixes, clearly prefer initial to final stress. Accordingly, irrespective of whether we analyze the medial consonants / n/ , / d/ , / v/ , and / p/ as onset of the ult vs. coda of the penult, stress in any , finish , medal , river , and topic is supposed to be initial simply because the strings y , ish , al , er , and ic , in which these disyllables end, occur in English as suffixes and therefore repel stress. Note also that virtually any English ate -disyllable (e.g., frustrate , migrate , stagnate , truncate , etc.) vacillates between final and initial stress (with the former being the preferred stress pattern among British English speakers vs. the latter being more frequently used by American English speakers). With final stress an ate -disyllable is supposed to be pronounced not only because of the segmental length principle but also because of the fact that of the 795 ate -words with an ion -counterpart in the MRC , 495 (~62.26 %) have fewer occurrences in the BNC than a corresponding ion -word. E.g., while the shorter verb locate occurs in the corpus 843 times, the longer noun location has 3,894 attestations (cf. LDOCE , where location is said to be a high-frequency word and locate a medi- 88 4 Stress assignment in English um-frequency one); the mean of the frequency differences between the 495 ate and corresponding ion -words is ~512.06 and the median is 56. The analysis of locate as a back-derivative from location , which is also in accordance with the meaning definitions “assign a location to / discover the location of,” given in the WordNet semantic database, results in the finally-stressed pronunciation / ləʊˈkeɪt/ ( OED ), which is the preserved stress of the base / lə(ʊ)ˈkeɪʃ(ə)n/ ( OD ). At the same time, however, because the string ate occurs in English as a suffix (e.g., pulsate ← pulse + ate ) and should therefore be stress-repellent, locate is also (especially by Americans as a rule stressed) / ˈləʊkeɪt/ ( OED ). A fairly similar case is ose -verbs, which appeared in the English language as back-derivatives from osis -nouns ( OD ). For example, because the longer base noun necrosis is stressed / nɛˈkrəʊsɪs/ ( OD ), one of the stress patterns of the shorter derived verb necrose is / nɛˈkrəʊs/ (OD). At the same time, however, because the string ose , in which necrose ends, occurs in English as a suffix (e.g., comatose ), necrose is also sometimes pronounced / ˈnɛkrəʊs/ , with the ult in it repelling stress. Finance is, according to LDOCE , / ˈfaɪnæns/ vs. / fəˈnæns/ in British English and / fəˈnæns/ vs. / ˈfaɪnæns/ in American English. Because the shorter finance occurs in the BNC only 7,553 times while the longer financial has 16,559 attestations (cf. 16,163 vs. 66,815, which are the corresponding frequency numbers from the COCA corpus), the former is supposed to be pronounced / fəˈnæns/ , preserving the stress of the more frequently used longer / fəˈnænʃəl/ ( LDOCE ). Because, however, the string ance , in which finance ends, occurs in English as a suffix (e.g., vengeance ), stress in finance can also be initial. Likewise, romance occurs in the BNC only 991 times, but the penultimately-stressed trisyllabic romantic has 1,987 attestations; similarly, in COCA , the corresponding frequency numbers are 8,240 vs. 12,073. The back-derivation analysis (coupled with the segmental length principle: ro∙mance) results in the finally-stressed pronunciation / rəʊˈmæns/ ( LDOCE ) whereas the ending ance is responsible for the initially-stressed alternative / ˈrəʊmæns/ ( LDOCE ). In Here’s how it is with the police (at 00: 15: 42,023 --> 00: 15: 45,288) of Kill Switch (2008 action film), American actor Steven Seagal can be heard pronouncing police with initial stress. The pronunciation / ˈpoʊlis/ is, according to Dictionary.com, especially characteristic of South Midland and Midland U. S. speakers; similarly, according to the OD , Scottish and Irish English speakers pronounce police / ˈpəʊlɪs/ or / ˈpɒlɪs/ (and spell it polis ). Because the string ice occurs in English as a suffix (e.g., justice , service ), many ice -disyllables have initial stress even when this string, in which these disyllables end, is not an actual suffix. E.g., office is stressed / ˈɒfɪs/ ( OD ) even though it cannot be segmented into the base off and the suffix ice the way, e.g., justice is segmentable into the base just and the suffix ice . 4.2 Disyllables 89 4.2.4 Hiatus resolution and / or vowel elision In the case of no less than 273 initially-stressed disyllabic transcriptions in the MRC , we are dealing with a syllabification variation, involving especially hiatus resolution. Hiatus, which means that a codaless syllable is followed by an onsetless one, is “dispreferred cross-linguistically” (Chitoran & Hualde 2007: 61; cf. Hieke 1984: 346, who speaks of “the avoidance of hiatus where possible”). That is why, pronunciation of, e.g., the word radiant vacillates between the trisyllabic version / ˈrā-dē-ənt/ (MWO), which involves a hiatus (i.e., ra . di . ant ), and the disyllabic alternative reI/ dj@nt ( MRC ), where the hiatus i . a is resolved via replacing the preceding vowel i through the phonetically similar glide / j/ : “Glide Formation is a subrule of the general resyllabification device which is responsible for the avoidance of vocalic hiatus in a variety of languages” (de Haas 1988: 244). Another popular strategy serving the same purpose is diphthongization. Thus, as, for example, Jones (1917: xxx) points out, the adjective glorious vacillates between the trisyllabic pronunciation glo . ri . ous , which involves a hiatus (i.e., / rɪ-əs/ ), and the disyllabic alternative glo . rious , with the hiatus sequence / ɪ-ə/ being pronounced “in the gliding (diphthongal) manner.” According to Jones (1917: xxx), hiatus-involving pronunciations such as glo . ri . ous are mainly typical of “slow utterance of certain classes of words.” (Notice, however, that in contrast to the MRC , where glorious is glO/ rI@s, the MWO gives for glorious only the trisyllabic hiatus-involving pronunciation / ˈglȯr-ē-əs/ ; similarly, also LDOCE regards glorious as a trisyllable: glo∙ri∙ous.) Noteworthy is also vowel deletion as a hiatus resolution strategy (which, according to, e.g., Wetzels (1995: 291), “is [likewise] a very common phonological process”). E.g., stygian is / ˈsti-j(ē-)ən/ according to the MWO ; the syllabic structure of this word thus vacillates between the hiatus-involving trisyllabic / ˈstɪ-dʒɪ-ən/ and the hiatus-free disyllabic / ˈstɪ-dʒən/ , where the preceding vowel of the hiatus i . a is deleted. Apart from this, it should be added that vowel elision is in general characteristic in English of sequences of unstressed syllables that involve especially / ə/ and / ɪ/ (Cruttenden 2008: 250). E.g., (the hiatus-free) interest is / ˈɪnt(ə)rɪst/ ( OD ), with elision taking place in the unstressed sequence / tərɪst/ , which contains the vowels / ə/ and / ɪ/ . Cases similar to these (i.e., those exhibiting a syllabification variation due to hiatus resolution / vowel elision) can be also found among finally-stressed disyllables (e.g., idea is both / ī-ˈdē-ə/ and / ī-ˈdēə/ in the MWO (and only aI/ dI@ in the MRC / i∙dea in LDOCE ); the word is thus either a penultimately-stressed trisyllable or a finally-stressed disyllable, with the hiatus e . a being resolved via diphthongization), but the proportion of such cases among initially-stressed disyllabic transcriptions in the MRC is higher than that among finally-stressed 90 4 Stress assignment in English disyllables: ~2.7 % (273 / 10,119) vs. ~0.4 % (nine / 2,270). This difference is statistically significant: χ 2 (1) = 44, p < 0.000001. Accordingly, analyzing the nature of initial stress in, e.g., radiant , we note that this word is not an irregularly-stressed disyllable (in which stress falls upon a segmentally shorter penult) but a regularly-stressed trisyllable: Because the string ant occurs in English as a suffix (e.g., claimant ) and is therefore stress-repellent, stress in radiant should in accordance with the segmental length principle be initial in the remaining lefthand disyllabic string ra . di -. Similarly, in interest 1) the ult . rest cannot take stress because the string est occurs in English as a suffix (e.g., cutest ) and 2) the remaining lefthand disyllabic string in . te should receive initial stress in accordance with the segmental length principle. Note also that leopard is phonetically only / ˈlepəd/ ( LDOCE ), but because in a number of English words, the orthographic vowel sequence eo is phonetically realized as hiatus (e.g., Be∙o∙wulf, Cle∙o∙pat∙ra, cre∙ole, de∙o∙do∙rant, etc.), we can suggest that underlyingly, leopard is le . o . pard , with stress being (regularly) initial in the lefthand disyllabic string le . o - (i.e., the righthand monosyllabic string pard cannot take stress because the string ard occurs in English as a suffix (e.g., drunkard ) and is thus stress-repellent). These claims are strongly supported by the fact that of the 273 transcriptions in the MWO such as / ˈrā-dē-ənt/ (i.e., those that correspond to transcriptions such as reI/ dj@nt in the MRC ), 228 (~83.52 %) are transcriptions such as / ˈrā-dē-ənt/ , in which the antepenult is segmentally not shorter than the immediately following penult; the average values are ~2.62 (antepenult) vs. ~1.97 (penult), which means that the first syllable in a word such as radiant is as a rule segmentally longer than the second syllable. 4.2.5 Disyllables as parts of longer words Of the 1,499 highand medium-frequency (in British English either exclusively or preferably) initially-stressed disyllables in LDOCE , 35 (~2.33 %) also occur in the dictionary as the penult-ult string of at least one highor medium-frequency penultimately-stressed trisyllable. E.g., the phonetic transcription / ˈtrɪbjuːt/ of tribute constitutes in LDOCE the penult-ult string of the transcription / kənˈtrɪbjuːt/ of contribute . By contrast, of the 770 highand medium-frequency (in British English either exclusively or preferably) finally-stressed disyllables in LDOCE , only two (~0.26 %)— cement and rely —occur orthographically at the end of the penultimately-stressed trisyllables announcement , replacement , entirely , and severely ; the phonetic realization of cement and rely is, however, different when they occur as separate words vs. parts of the trisyllables announcement / replacement and entirely / severely . 4.2 Disyllables 91 The difference of 35 / 1,499 vs. two / 770 achieves statistical significance— χ 2 (1) = 13.656, p = 0.0002—which allows us to say that a disyllable such as tribute , which occurs in English not only separately but also as a righthand part of a trisyllable such as contribute , is more likely to receive initial stress than a disyllable such as, e.g., dispute , which occurs in English only as a separate word. (According to LDOCE , dispute is in Present-day English more frequently stressed / dɪˈspjuːt/ than / ˈdɪspjuːt/ .) The explanation for this seems to be as follows. On the one hand, because there is the disyllabic word tribute , the trisyllabic word contribute can be segmented into the prefix con and the root tribute ; this segmentation results in the stress pattern / kənˈtrɪbjuːt/ . At the same time, however, because the disyllabic string tribute is root of the trisyllabic word contribute , the former can also be seen as the root of the disyllabic word tribute itself, which is then not segmented into the prefix tri and the (segmentally longer) root bute . The stress pattern of tribute is therefore only / ˈtrɪbjuːt/ , whereas dispute is in accordance with the segmental length principle usually stressed / dɪˈspjuːt/ . 4.2.6 Rhythm, emphasis, and semantics Deviations from the segmental length principle may also have rhythm-, emphasis-, or semantics-related causes (which sometimes simultaneously apply to one and the same disyllabic English word). An example is the adjective complex , which, as mentioned in Chapter 1, is in contemporary English more frequently stressed initially than finally. In fact, the only transcription given for complex in the OD dictionary is the initially-stressed / ˈkɒmplɛks/ and, similarly, in the OED the negative incomplex is said to be “/ ɪnˈkɒmplɛks/ formerly / ɪnkəmˈplɛks/ .” Of all the 9,384 occurrences of complex in the BNC , 1,971 (~21 %) are occurrences in the modifier position, with complex being adjacent to a noun (e.g., complex process ). Of the 723 nouns immediately preceded by complex in the BNC for which phonetic transcriptions are available in the OD , 485 (~67.08 %) are either monosyllables or initially-stressed polysyllables. From the point of view of rhythm, stress in complex should therefore be initial rather than final, i.e., e.g., the stress pattern / ˈkɒmplɛks ˈprəʊsɛs/ is better than / kəmˈplɛks ˈprəʊsɛs/ , which involves a stress clash. Similarly, unlike to rebel , which is stressed / rɪˈbɛl/ ( OD ), preserving the stress of the more frequently used longer noun rebellion (972 vs. 90 occurrences in the BNC ; note also that the meaning of to rebel is defined in the WordNet database as “take part in a rebellion”), a rebel is stressed / ˈrɛb(ə)l/ ( OD ) because of all the 1,106 occurrences of the string rebel in the BNC , 440 (~39.78 %) are occurrences such as rebel forces , rebel leader , rebel group , rebel movement , rebel soldiers , rebel 92 4 Stress assignment in English army , rebel troops , etc., i.e., occurrences of rebel as a modifier of another (usually, initially-stressed) noun. In combinations such as these, the stress pattern / ˈrɛb(ə)l/ is from the point of view of rhythm better than the stress pattern / rɪˈbɛl/ (which, however, is unproblematic when rebel occurs as a verb because to rebel is as a rule immediately followed by the preposition against , in which stress is final). Stress clash avoidance is thus undeniably also a reason why disyllabic instances of different morphosyntactic categories prefer different stress patterns. Thus: disyllabic verbs in English are typically preceded by a stressed and followed by an unstressed syllable whereas nouns are typically preceded by an unstressed syllable and followed by a stressed one. Hence, disyllabic verbs are more rhythmical when they have an iambic (i.e. weak-strong) stress pattern while disyllabic nouns are more rhythmical when they have a trochaic (i.e. strong-weak) stress pattern. (Berg 1999: 136) Recall also the fact, mentioned in Chapter 1, that while in American English adult is more frequently stressed finally, British English speakers prefer the initially-stressed pronunciation / ˈædʌlt/ . A possible explanation for this fact is that of all the 5,035 occurrences of adult in the BNC , 2,318 (~46.04 %) are occurrences in combinations such as adult life , adult world , adult women , adult literacy , adult children , etc., in which adult is immediately followed by a noun (which is often either a monosyllable or an initially-stressed polysyllable: Of the 494 nouns modified by adult in the BNC for which phonetic transcriptions are given in the OD, 370 (~74.9 %) are either monosyllables or initially-stressed polysyllables). By contrast, in the case of the 29,431 total occurrences of adult in COCA , which is a corpus of contemporary American English, the same can be said about only 10,257 occurrences: ~34.85 %. Since this difference is statistically hugely significant— χ 2 (1) = 232, p < 0.000001—we can say that combinations such as adult life , adult world , adult women , adult literacy , adult children , etc. (which encourage initial stress in adult , i.e., / ˈædʌlt ˈlʌɪf/ rather than / əˈdʌlt ˈlʌɪf/ , which involves a stress clash) are more frequent in British than in American English. Notice also that in addition to occurring in the modifier position, adult also often occurs in the head position, in which it itself is modified by other adjectives. The most frequent modifier of adult in the COCA corpus is the monosyllabic young , i.e., the combination young adult occurs in the corpus 1,114 times (~3.79 %). In the BNC , by contrast, the same combination occurs only 60 times (~1.19 %) and is thus even less frequent than the combination universal adult , which has 84 attestations. The difference of 1,114 / 29,431 vs. 60 / 5,035 is also statistically hugely significant: χ 2 (1) = 88, p < 0.000001. We can therefore say that the combination young adult , in which final stress is from the point of view of 4.2 Disyllables 93 rhythm better than initial stress, i.e., / ˈjʌŋ əˈdʌlt/ rather than / ˈjʌŋ ˈædʌlt/ , which involves a stress clash, is more frequent in American than in British English. For an audio illustration, the reader is referred to the already mentioned YouTube video S5hXrgfwK8o (10. 04. 2017), in which one and the same British English speaker pronounces the word adult in three different environments. At 00: 01: 16.014 --> 00: 01: 18.063, we hear but it explores some very ˈ adult themes , where adult , followed by the monosyllabic head themes , is stressed initially. At 00: 03: 24.610 --> 00: 03: 26.659, we hear The humor in it is really a ˈ dult , where adult , preceded by the modifier really , / ˈrɪəli/ ( OD ), is stressed finally (thus, we can argue that also a one-syllable distance between two primary-stressed syllables, / ˈrɪəli ˈædʌlt/ , counts for the speaker in the video as rhythmically unfortunate). Finally, at 00: 06: 04.526 --> 00: 06: 06.531, we hear It’s one of those ˈ cartoons for ˈ adults , where retraction of stress occurs in the preceding noun cartoons : / ˈkɑːtuːnz fɔː ˈædʌlts/ (which, again, suggests that for the speaker in the video, there should be at least two unstressed syllables between two stressed ones). Also in other videos featuring the voice of the same British English speaker, she seems to be very consistent about pronouncing adult with final stress in environments such as since becoming an a ˈ dult and moving out (9LkG43iZ4u4, 10. 04. 2017), It’s basically your classic young a ˈ dult high school novel (9wh7wEx- Tna0, 10. 04. 2017), I think I’m still quite an entitled a ˈ dult ( EK hQnn FL 6 ZM , 10. 04. 2017), etc. By contrast, in environments such as It felt more like an ˈ adult fable (rVa TT zdZ0Ck, 11. 04. 2017), she pronounces adult with initial stress. As for emphasis, recall that prefixes in English are stress-neutral (e.g., Cruttenden 2008: 241), but “[w]hen there is contrast or when the idea expressed by the prefix is given special prominence, the prefix bears the primary stress and the base a secondary stress […]” (Poldauf 1984: 24). For example, to restyle was (in captioned YouTube videos) found to have been pronounced by six native English speakers with initial stress, i.e., ˈ restyle , while only five speakers were heard to have used the finally-stressed version re ˈ style . Likewise, to resell is stressed ˈ resell in the videos oO-A1apjPo0 (04. 04. 2017) and szeNiRsjriQ (04. 04. 2017). In His Last Vow ( BBC ’s Sherlock , third episode of Season Three), we hear Probably want to ˈ restart your morphine at 00: 43: 22,960 --> 00: 43: 25,040. Initial stress in these prefixed verbs seems to be of a purely emphatic nature, falling upon the semantically transparent prefix re -, which expresses the meaning “repetition of an action, doing something again.” A similar case is uphold , for which LDOCE incorrectly gives only the base stress / ʌpˈhəʊld/ (for both British and American English). An initially-stressed version of this verb can, however, be heard in, e.g., When a man joins the bureau, he takes an oath to ˈ uphold certain values , uttered by FBI Special Agent Dale Cooper (played by actor Kyle MacLachlan) at 00: 02: 01,609 --> 00: 02: 06,615 of the 94 4 Stress assignment in English seventh episode of the first season of Twin Peaks . Similarly, at 00: 06: 43,487 --> 00: 06: 48,824 of Predestination (2014), we hear Do you solemnly swear to ˈ uphold the rules and regulations set forth by T. B. R. Code 7286? Note also that according to the OD , only the noun download is stressed / ˈdaʊnləʊd/ whereas the verb download is stressed only / daʊnˈləʊd/ . (The same is true of many other disyllabic formations such as uphold and download , which are said to exhibit initial stress only when they occur as nouns and / or adjectives but not as verbs). It is obvious, however, that a claim such as this cannot be true (even if it is made by a lexicographer working for a renowned academic publisher) because the only reason why, e.g., download is stressed / ˈdaʊnləʊd/ is that its lefthand component down is what distinguishes download from its directional opposite upload . The emphatic potential inherent in the down of download equally applies to a download and to download , which is why also the latter is more frequently pronounced with initial stress. E.g., in the video pmtcFhz22 MQ (27. 05. 2017), we can hear British Prime-Minister Theresa May saying at 00: 04: 06.569 --> 00: 04: 10.230 encourage everybody to ˈ download the single ; of six British English speakers who were found to have pronounced to download in captioned YouTube videos, five were heard to have used the initially-stressed pronunciation / ˈdaʊnləʊd/ (with one of the speakers, however, vacillating between / ˈdaʊn-/ and / -ˈləʊd/ in the videos l0x2PxNcEPs (27. 05. 2017) and Yu8KpLpkrPs (27. 05. 2017)). (Thus we can say that not only the OD, but also LDOCE , where to download is / ˌdaʊnˈləʊd/ in British English vs. / ˈdaʊnloʊd/ in American English, fails to recognize that initial stress is the preferred stress of to download in both British and American English). Finally, observe that according to the OED , to surname is stressed both / ˈsɜːneɪm/ and / sɜːˈneɪm/ , but in captioned YouTube videos featuring the spoken occurrences of this verb, only the prefix stress / ˈsɜːneɪm/ was heard by the author; in agreement with this finding, also the OD and LDOCE give for the form surname only initially-stressed transcriptions (i.e., it is not stated in these dictionaries that surname is stressed differently when functioning morphosyntactically as a noun vs. verb.) Just as in formations such as uphold and download , the lefthand element sur modifies the righthand name in an important way irrespective of whether surname is a noun or verb. Both a surname and to surname are therefore pronounced with initial stress. As for semantics, consider the (actually occurring) stress difference between / pəˈmɪt/ of to permit and / ˈpəːmɪt/ of a permit ( OD ). What is interesting about this stress difference is that in English, there is also the verb to remit , which, just like to permit , is only finally-stressed as / rɪˈmɪt/ ( OD ), but unlike the derived noun a permit , in which stress is only initial, the derived noun a remit is both initially-stressed as / ˈriːmɪt/ and finally-stressed as / rɪˈmɪt/ ( OD ). The reason why 4.2 Disyllables 95 the stress variation occurs only in the case of a remit but not of a permit seems to be the fact that the meaning of the derived noun a permit represents a fairly specialized extension of the meaning of the base verb to permit . This is reflected in the fact that in contrast to a remit , which is defined in the MWO dictionary as “an act of remitting” and “something remitted to another person or authority,” a permit cannot be defined either as an act of permitting or as something that was permitted to another person. A permit is “a written warrant or license granted by one having authority” ( MWO ). This noun can often be heard in collocations such as gun permit , residence permit , work permit , etc. Similarly, a variety of things in English can be affixed , collected , combined , compressed , produced , etc., but an affix stands only for a particular thing that undergoes affixing, i.e., “a linguistic element added to a word to produce an inflected or derived form” (WordNet); a collect stands only for “a short prayer generally preceding the lesson in the Church of Rome or the Church of England” (WordNet); a combine denotes either “a consortium of independent organizations formed to limit competition by controlling the production and distribution of a product or service” or “harvester that heads and threshes and cleans grain while moving across the field” (WordNet); a compress is defined in WordNet as “a cloth pad or dressing (with or without medication) applied firmly to some part of the body (to relieve discomfort or reduce fever”; produce as a noun only means “fresh fruits and vegetable grown for the market” (WordNet), etc. The adjective minute , which is stressed / mʌɪˈnjuːt/ ( OD ), expresses the meaning “[e]xtremely small” ( OD ). The noun minute , which is stressed / ˈmɪnɪt/ ( OD ), is, by contrast, associated with the sense “[a] period of time equal to sixty seconds” ( OD ), which can be regarded as a particular instance of being minute . What can be seen in the case of these initially-stressed derived nouns is that their meanings represent fairly specialized extensions of what the identically-spelled words signify as verbs / adjective. We thus argue that stress differences such as / pəˈmɪt/ of to permit vs. / ˈpəːmɪt/ of a permit are conditioned only semantically, i.e., these stress differences are deliberately employed by English speakers to make the semantically specialized derived noun formally different from the base verb / adjective. When, by contrast, the meaning of the derived form represents a fairly general extension of the meaning of the base form (e.g., a remit = an act of remitting), the tendency in contemporary English is to pronounce the derived form with the stress of the base form. This claim is strongly supported by cases such as a control and to control , which are both finally-stressed as / kənˈtrəʊl/ ( OD ), a concern and to concern , which are both finally-stressed as / kənˈsəːn/ ( OD ), complete as a verb and complete as an adjective, which are both finally-stressed as / kəmˈpliːt/ ( OD ), etc. In the case of these words, word-class differences do not give rise to stress 96 4 Stress assignment in English differences because the derived forms in these pairs mean approximately the same thing as the corresponding base forms: For example, a control is “power to direct or determine” (WordNet) and to control is “exercise authoritative control or power over” (WordNet). There is thus no need to have different stress patterns for a control and to control , both have therefore regular final stress falling upon the segmentally longer ult . trol . Likewise, a surmise , which is / sɜːˈmaɪz/ vs. / ˈsɜːmaɪz/ according to the OED , is in contemporary English more frequently pronounced with final stress because the meaning of a surmise , “a message expressing an opinion based on incomplete evidence” (WordNet), represents a fairly general extension of the meaning of to surmise , “infer from incomplete evidence” (WordNet). Both the base verb and the derived noun have therefore regular final stress, abiding by the segmental length principle. 4.2.7 Compounds Of the 2,477 left-prominent disyllabic compounds in the OD dictionary (which were automatically detected by the author in the manner described in 3.2), 907 (~36.62 %) have orthographically longer righthand components. E.g., the disyllabic compound boyfriend is pronounced / ˈbɔɪfrɛnd/ ( OD ) even though its righthand component friend is both phonetically and orthographically longer than its lefthand component boy . By contrast, of the 145 right-prominent disyllabic compounds, 80 (~55.17 %) are compounds such as, e.g., himself , whose righthand components contain more orthographic symbols than the corresponding lefthand component. Since the difference of 907 / 2,477 vs. 80 / 145 is statistically highly significant— χ 2 (1) = 20, p = 0.000007—it is tempting to say that also in disyllabic compounds, the choice between leftand right-prominence is (at least in part) in accordance with the segmental length principle. E.g., himself is stressed / hɪmˈsɛlf/ ( OD ) because its righthand component self is segmentally longer than the lefthand component him . Notice, however, that of the 145 right-prominent disyllabic compounds (which, as observed above, were automatically detected by the author), 42 (~28.97 %) are verbs such as uphold and download , which are fairly frequently left-prominent even though the OD gives for them only the right-prominent transcriptions / ʌpˈhəʊld/ and / daʊnˈləʊd/ ; a similar case is update , which, according to the OD , is / ʌpˈdeɪt/ as a verb vs. / ˈʌpdeɪt/ as a noun. Just as in the case of uphold and download , the lefthand element up modifies the righthand date in an important way in both an update and to update . Just like the former, the latter is therefore also frequently pronounced / ˈʌpdeɪt/ . 4.3 Three and more syllables 97 If items such as update are left out of consideration, the proportion of right-prominent disyllabic compounds that abide by the segmental length principle drops from ~55.17 % to ~47.57 %. The difference of 907 / 2,477 vs. 80 / 145 still counts as statistically significant— χ 2 (1) = 5.089, p = 0.024—but as one can see, the new result is not far away from the threshold of p = 0.05. Given this fact and given that compounds in English are in general considerably more frequently leftthan right-prominent, we can conclude that in the case of disyllabic English compounds, abidance by the segmental length principle is far less pronounced than it is in the case of other disyllabic English words. 4.2.8 Summary Final stress in a disyllabic English word is prototypically either the preserved stress of a longer English word (synchronic back-derivation/ shortening) or root stress, falling upon a segmentally longer ult. When these conditions do not hold or when the segmentally longer ult occurs in English as a suffix, stress in a disyllabic English word is as a rule initial. 4.3 Three and more syllables As already reported in Section 3.2, syllabified phonetic transcriptions and stress indications are given in the MRC for 19,545 items (of any morphological structure) whose syllabic length is at least three. Of these items, 5,383 are words in which stress can be said to be regularly penultimate (i.e., the penultimate syllable is heavy) and 7,741 are words in which stress is regularly antepenultimate (i.e., the penultimate syllable is light). In other words, of the 19,545 items of three and more syllables, 13,124 (~67.15 %), i.e., the absolute majority, have stress patterns that respect the Latin Stress Rule. Examples include, for instance, / əˈdʒɛndə/ of agenda ( OD ) and / eɪˈɔːtə/ of aorta ( OD ), in which stress falls upon the heavy penults / ˈdʒɛn/ and / ˈɔː/ , vs. / ˈanɪm(ə)l/ of animal ( OD ) and / əˈmerɪkə/ of America ( OD ), where the penultimate syllables / nɪ/ and / rɪ/ are light. Related to this finding is the fact that of the 7,741 items in which stress is regularly antepenultimate (i.e., the penultimate syllable is light), 4,114 (~53.15 %) are words such as animal and America , in which stress falls on a light antepenult, i.e., in / ˈanɪm(ə)l/ and / əˈmerɪkə/ the antepenults / ˈa/ and / ˈme/ , which bear stress, are light. Since the Latin Stress Rule does not require that the antepenult itself be heavy in order to count as eligible to receive stress (i.e., the only requirement for antepenultimate stress is that the penult be light), stressed antepenultimate syllables are in the English language almost as frequently light as they are heavy. 98 4 Stress assignment in English Note, however, that just as in the case of disyllabic English words, the stress of an English word of three and more syllables can likewise often be analyzed as the stress of another English word (which challenges the analysis of Present-day English as a Latin-like language). 4.3.1 Suffixed and back-derivatives Of the 4,794 trisyllables for which the MRC gives the stress pattern 200 (i.e., antepenultimate stress), 1,153 (~24.05 %) are, according to the tool Morphological Analysis, actual suffixed derivatives. Of the 992 bases of these derivatives (according to the tool Morphological Analysis), 936 (~94.35 %) are, according to the OD , either initially-stressed disyllables (e.g., canon of canonize ) or monosyllables (e.g., drink of drinkable ). Thus, e.g., canonize is stressed / ˈkanənʌɪz/ ( OD ), preserving the stress of / ˈkanən/ , and in / ˈdrɪŋkəbl/ ( OD ), stress falls upon the only syllable constituting the monosyllabic base drink . Similarly, of the 3,685 trisyllables for which the MRC gives the stress pattern 20 (i.e., penultimate stress), 1,340 (~36.36 %) are, according to the tool Morphological Analysis, actual suffixed derivatives. Of the 995 bases of these derivatives (according to the tool Morphological Analysis), 727 (~73.07 %) are, according to the OD , either finally-stressed disyllables (e.g., abuse of abuser ) or penultimately-stressed trisyllables (e.g., dependent of dependence ). E.g., abuser (which is considerably less frequent in the BNC than abuse : 43 vs. 3,578) is stressed / əˈbjuːzə/ , preserving the stress of / əˈbjuːz/ , and dependence is stressed / dɪˈpɛnd(ə)ns/ ( OD ), preserving the stress of / dɪˈpɛnd(ə)nt/ ( OD ). As for the 2,551 tetrasyllables and longer words from the MRC that the tool Morphological Analysis considers to be suffixed derivatives, 1,319 are, according to the MRC , stressed antepenultimately. Of the 1,056 bases of these derivatives (according to the tool Morphological Analysis), no less than 471 (~44.6 %) are, according to the OD , either finally-stressed disyllables (e.g., adapt of adaptable ) or penultimately-stressed trisyllables (e.g., accomplish of accomplishment ). That is, e.g., adaptable is pronounced / əˈdaptəb(ə)l/ ( OD ), preserving the stress of / əˈdapt/ ( OD ), and accomplishment is pronounced / əˈkʌmplɪʃm(ə)nt/ ( OD ), preserving the stress of / əˈkʌmplɪʃ/ ( OD ). Apart from derivatives such as adaptable and accomplishment , 893 (~35.01 %) are derivatives such as presidency and indicator , in which stress falls upon a syllable farther than the antepenult (for which the Latin Stress Rule does not have a provision). That is, presidency is stressed / ˈprɛzɪd(ə)nsi/ ( OD ), preserving the stress of / ˈprɛzɪd(ə)nt/ ( OD ), and indicator is stressed / ˈɪndɪkeɪtə/ ( OD ), preserving the stress of / ˈɪndɪkeɪt/ ( OD ). 4.3 Three and more syllables 99 Note also that of the 817 antepenultimately-stressed trisyllables from the MRC that the tool Morphological Analysis considers to be the base of at least one suffixed derivative, 332 (~40.64 %) have fewer occurrences in the BNC than at least one corresponding suffixed derivative. A case in point is an ate -verb, which, as mentioned in the previous section, is as a rule (~62.26 %) less frequent in the BNC than a corresponding ion -noun. Thus, for instance, 1) since in the BNC , the shorter verb educate is considerably less frequent than the longer noun education : 365 vs. 26,113, and 2) since in the WordNet semantic database, the meaning of educate is defined as “give an education to,” it is not surprising that on some occasions, this verb is pronounced by English speakers with final stress, i.e., / ɛdjʊˈke: t/ is the stress pattern used by Irish English speakers (Hickey 2007: 139) and, similarly, also in other English varieties, “we hear forms like propagáte rather than the expected própagate , presumably influenced by propagátion ” (Bauer 2014: no pagination in the online-first version of the article). At the same time, however, because the string ate , in which educate ends, occurs in English as a suffix and therefore repels stress, the more frequently used stress pattern of educate in both British and American English is not / ˌedjʊˈkeɪt/ but / ˈɛdjʊkeɪt/ ( OD ), with the (phonetically not different) secondary stress of the longer base / ˌedjʊˈkeɪʃən/ ( LDOCE ) becoming primary stress in the shorter derivative. A similar example involving an adjective is / ˈanɪmət/ of animate ( OD ), which is the promoted secondary stress of / ˌænəˈmeɪʃən/ of animation ( LDOCE ), which occurs in the BNC 322 times while the shorter animate has only 120 attestations. (Because the adjective animate is stressed / ˈanɪmət/ , also the semantically related noun animal is stressed / ˈanɪm(ə)l/ , i.e., animals are animate creatures.) The same applies to 34 ute -words in the MRC , 23 of which (~67.65 %) are less frequent in the BNC than a corresponding ion -word. E.g., while execute occurs 402 times, execution has 1,353 attestations; the mean of the frequency differences between the 23 ute and corresponding ion -words is ~1,795.91 and the median is 455. The stress pattern / ˈɛksɪkjuːt/ of the derived form execute ( OD ) can thus be seen as the promoted secondary stress of the base form / ˌɛksɪˈkjuːʃ(ə)n/ ( LDOCE ), i.e., we assume that English speakers are reluctant to pronounce execute / ˌeksəˈkjuːt/ because the string ute , in which execute ends, is like a suffix and therefore repels stress (but do observe that according to Hickey (2007: 139), Irish English speakers stress distribute / dɪstrɪˈbjuːt/ , preserving the (primary) stress of the longer base noun distribution , which occurs in the BNC 10,323 times while the shorter verb distribute has only 602 attestations. Similarly, in the Corpus of Global Web-based English (Davies 2013), using which researches can “carry out comparisons between different varieties of English,” distribution and education are used by Ireland-based English-speaking Internet users 3,101 and 100 4 Stress assignment in English 28,938 times respectively, whereas distribute and educate are used by them only 820 and 1,350 times. Thus, also from the perspective of an Irish English speaker, the latter can be seen as back-derivatives from the former.) Similarly, of the 25 ite -words with an ion -counterpart in the MRC , 15 (60 %) are less frequent in the BNC than a corresponding ion -word. E.g., while the shorter verb extradite (which, according to the OD , did indeed come into existence in English (in mid 19th century) via back-formation from extradition ) occurs in the corpus 54 times, the longer noun extradition has 315 attestations (the latter is thus not only diachronically but also synchronically anterior to the former); the mean of the frequency differences between the 15 ite and corresponding ion -words is ~1,646.67 and the median is 966. The stress pattern / ˈɛkstrədʌɪt/ of extradite ( OD ) can thus be seen as the promoted secondary stress of / ˌekstrəˈdɪʃən/ of extradition ( LDOCE ), i.e., the derived verb is not pronounced with final stress because the string ite occurs in English as a suffix (e.g., Israelite ) and is therefore stress-repellent. (As in the above mentioned case of / hjʊˈmanɪti/ vs. / hjʊˈmeɪn/ , we ignore the fact that the segmental structure of the more frequently used / ˌekstrəˈdɪʃən/ is not completely preserved in the less frequently used / ˈɛkstrədʌɪt/ .) A slightly different case is the 80 ify -words with an ification -counterpart in the MRC , 39 of which (48.75 %) have fewer occurrences in the BNC than a corresponding ification -word. E.g., amplify occurs in the corpus 108 times, but amplification has 299 attestations; the mean of the frequency differences between the 39 ify and corresponding ification -words is 119 and the median is 37. Accordingly, because the longer base amplification is stressed / ˌamplɪfɪˈkeɪʃ(ə)n/ ( OD ), the only stress that can be preserved in its back-derivative amplify is obviously the secondary stress, i.e., amplify is stressed / ˈamplɪfʌɪ/ ( OD ), with the secondary stress of the longer base form thus becoming primary stress in the shorter derived form. Likewise, antepenultimate stress in the back-derivative satisfy —/ ˈsætəsfaɪ/ ( LDOCE )—is the promoted secondary stress of / ˌsætəsˈfækʃən/ of satisfaction ( LDOCE ), which occurs in the BNC 2,761 times, whereas the shorter verb satisfy has only 1,925 attestations. Similarly, of the 115 ize -words with an ization -counterpart in the MRC , 62 (~53.91 %) have fewer occurrences in the BNC than a corresponding ization -word. E.g., while the shorter verb organize occurs in the corpus 817 times, the longer noun organization has 6,297 attestations; the mean of the frequency differences between the 62 ize and corresponding ization -words is ~154.16 and the median is ~25.5. Accordingly, because the longer base is stressed / ˌɔːɡənaɪˈzeɪʃən/ ( LDOCE ), the shorter derivative can only be stressed / ˈɔːɡ(ə)nʌɪz/ ( OD ). (Note that the initially-stressed disyllable organ occurs in the BNC 1,326 times and is thus also more frequent than organize , which, as just 4.3 Three and more syllables 101 pointed out, has only 817 attestations. Semantically, however, the derivation of organize from organization is much more intuitive than the suffixation analysis organ + ize . Thus one of the paraphrases of organize given in WordNet is “bring order and organization to.”) Noteworthy are also tetrasyllabic ic -words such as diplomatic , which are responsible for antepenultimate stress in trisyllables such as diplomat . Since the longer adjective diplomatic occurs in the BNC 1,968 times while the shorter noun diplomat has only 360 attestations, the latter should from a synchronic point of view be seen as a back-derivative from the former. (Similarly, from a purely diachronic perspective, hieroglyph , megalith , onomast , photogene , and polygene are, according to the OD , actual back-derivatives from ic -words hieroglyphic , megalithic , onomastic , photogenic , and polygenic . As for diplomat , the word is, according to the OD , etymologically also a back-derivative, but it was French speakers who back-derived it from the longer diplomatique , “diplomatic.”) As pointed out in 3.2, a tetrasyllabic ic -word vacillates between penultimate and initial stress, with the latter occurring especially in the modifier position. E.g., diplomatic is, according to LDOCE , / ˌdɪpləˈmætɪk◀/ , which means that in environments such as diplomatic incident , the modifier is stressed / ˈdɪpləmætɪk/ rather than / ˌdɪpləˈmætɪk/ ; because of the 1,968 occurrences of diplomatic in the BNC , 1,289 (~65.5 %) are occurrences in combinations such as diplomatic service , diplomatic mission , diplomatic sources , diplomatic skills , etc., with diplomatic being adjacent to a noun, the stress pattern / ˈdɪpləmætɪk/ should be used more frequently than the stress pattern / ˌdɪpləˈmætɪk/ . Accordingly, because the base form diplomatic is (usually) stressed / ˈdɪpləmætɪk/ , the derived form diplomat is also stressed / ˈdɪpləmæt/ ( LDOCE ). Similarly, in the BNC , acrobat is less frequent than acrobatic (34 vs. 48), democrat is less frequent than democratic (988 vs. 5,878), patriot is less frequent than patriotic (187 vs. 556), pessimism is less frequent than pessimistic (186 vs. 378), vitriol is less frequent than vitriolic (25 vs. 73), etc. Accordingly, because, e.g., the more frequently used democratic is, according to LDOCE, stressed / ˌdeməˈkrætɪk◀/ , which means that the actual stress pattern of democratic is often / ˈdeməkrætɪk/ , the less frequently used democrat (which is defined in WordNet as “an advocate of democratic principles”) should also be stressed / ˈdeməkræt/ ( LDOCE ). A very similar case is residential , which has 2,918 attestations in the BNC and is thus more frequent than resident and residence , which occur 2,114 and 1,709 times respectively. For residential , LDOCE gives the transcription / ˌrezɪˈdenʃəl◀/ , which means that the actual stress pattern of this adjective is often / ˈrezɪdenʃəl/ : Of the 2,918 attestations of residential in the BNC , 2,280 (~78.14 %) are attestations in combinations such as residential care , residential homes , residential area , etc., with residential being adjacent to a noun. The stress pattern / ˈrezɪdenʃəl/ of 102 4 Stress assignment in English the more frequently used longer adjective residential is therefore preserved in the less frequently used shorter nouns resident and residence . As for penultimately-stressed trisyllables, note that of the 370 items from the MRC that the tool Morphological Analysis considers to be the base of at least one suffixed derivative, 115 (~31.08 %) have fewer occurrences in the BNC than at least one corresponding suffixed derivative. (Since the difference of 332 / 817 vs. 115 / 370 achieves statistical significance— χ 2 (1) = 9.904, p = 0.0016—we can say that back-derivation is more typical in English of anterather than of penultimately-stressed trisyllables.) For example, because the longer noun development has considerably more attestations in the BNC than the shorter verb develop (32,078 vs. 8,540), the stress pattern of the latter can be regarded as the preserved stress of the former, i.e., develop is stressed / dɪˈvɛləp/ (OD) because development is stressed / dɪˈvɛləpm(ə)nt/ ( OD ). The same can be said about / dɪˈlɪvə/ of deliver ( OD ), where stress is penultimate because it is antepenultimate in the more frequently used longer noun delivery : / dɪˈlɪv(ə)ri/ ( OD ). The latter occurs in the BNC 3,533 times, while the former is attested in the corpus only 2,072 times. The verb deliver is thus synchronically a back-formation from the noun delivery , i.e., the meaning of deliver is, according to WordNet, “bring to a destination, make a delivery.” Similarly, because the meaning of interpret is “give an interpretation or explanation to” (WordNet) and because the shorter form interpret is less frequent in the BNC than the longer form interpretation (1,313 vs. 4,315), penultimate stress in the former—/ ɪnˈtəːprɪt/ ( OD )—appears to be the promoted secondary stress of the latter: / ɪnˌtɜːprɪˈteɪʃən/ ( LDOCE ). The highest proportion of trisyllabic back-derivatives can, quite expectedly, be found among finally-stressed trisyllables (which, according to the tool Morphological Analysis, function as the base of at least one suffixed derivative): 64 / 102(=~62.75 %). Thus, for instance, in the BNC , the shorter word acquiesce is less frequent than the longer word acquiescence (65 vs. 171), appertain is less frequent than appertaining (15 vs. 38), apprehend is less frequent than apprehension , apprehensive , and apprehended (92 vs. 507, 278, and 128), comprehend is less frequent than comprehensive and comprehension (344 vs. 3,582 and 601), condescend is less frequent than condescending and condescended (14 vs. 88 and 17), contradict is less frequent than contradiction and contradictory (268 vs. 768 and 691), convalesce is less frequent than convalescent and convalescence (21 vs. 101 and 77), correspond is less frequent than corresponding , correspondent , and correspondence (754 vs. 2,145, 2,004, and 1,522), effervesce is less frequent than effervescent and effervescence (five vs. 40 and 24), evanesce is less frequent than evanescent and evanescence (zero vs. 28 and five), intermit is less frequent than intermittent , intermittently , and intermittency (two vs. 339, 198, and 19), introduce is less frequent than introduction (3,456 vs. 6,621), etc. 4.3 Three and more syllables 103 Similarly, interject , interrupt , intervene , introspect , resurrect , and vivisect are less frequent in the BNC than the corresponding ion -nouns interjection , interruption , intervention , introspection , resurrection , and vivisection (18 vs. 29, 458 vs. 616, 977 vs. 3,171, five vs. 126, 80 vs. 465, zero vs. 28), domineer , engineer , mountaineer , orienteer , profiteer , racketeer , and sloganeer are less frequent in the BNC than the corresponding ing -words domineering , engineering , mountaineering , orienteering , profiteering , racketeering , and sloganeering (zero vs. 81, 2,199 vs. 4,910, 79 vs. 214, one vs. 36, nine vs. 43, seven vs. 56, one vs. seven), countermand and intercept are less frequent in the corpus than the corresponding ed -forms countermanded and intercepted (10 vs. 34, 218 vs. 233), interfere and persevere are less frequent than the corresponding ence / ance -nouns interference and perseverance (1,019 vs. 1,408, 90 vs. 139), entertain is less frequent than entertainment (638 vs. 1,995), commissar is less frequent than commissariat (25 vs. 48), etc. As in the case of finally-stressed disyllables, the analyses of finally-stressed trisyllables as back-derivatives from longer English words are also often supported by meaning definitions given in the WordNet semantic database: Condescend = “behave in a patronizing and condescending manner,” contradict = “be in contradiction with,” entertain = “provide entertainment for,” phosphoresce = “to exhibit phosphorescence,” represent = “be representative or typical for,” etc. The shorter words condescend , contradict , entertain , phosphoresce , and represent are thus semantically secondary to the longer words condescending , contradiction , entertainment , phosphorescence , and representative . The stress of the former is thus the preserved stress of the latter. Note, however, that in addition to counting as back-derivatives from longer words, some of these finally-stressed trisyllables can also be analyzed as prefixed disyllables. E.g., comprehend is com - + prehend (cf. apprehend , reprehend ); stress in the disyllabic base prehend should be final in accordance with the segmental length principle, i.e., the ult . hend , which is segmentally longer than the penult pre ., is more like a root and should therefore receive stress. Similarly, because the disyllabic base respond is in accordance with the segmental length principle stressed / rɪˈspɒnd/ (OD), the trisyllabic derivative correspond is stressed / kɒrɪˈspɒnd/ ( OD ). Another similar case is the tetrasyllabic noun entrepreneur . On the one hand, it can be suggested that entrepreneur is / ˌɒntrəprəˈnəː/ ( OD ) because entrepreneurial is / ˌɒntrəprəˈnəːrɪəl/ ( OD ). That is, since the longer adjective entrepreneurial is more frequent in the BNC than the shorter noun entrepreneur , i.e., the respective numbers are 385 vs. 282, the latter is synchronically a back-derivative from the former and thus merely preserves its stress. Additionally, however, it can be argued that stress in entrepreneur is final simply because its morphological structure is entre - + preneur (cf. similar formations infopreneur , intrapreneur , 104 4 Stress assignment in English solopreneur , technopreneur , tenderpreneur ). Assuming that the syllabification of the bound root preneur , which is the string eligible to receive primary stress in entrepreneur , is pre . neur , i.e., the ult is orthographically longer than the penult, we naturally obtain the finally-stressed pronunciation / ˌɒntrəprəˈnəː/ , with the prefix entre receiving secondary stress. 4.3.2 Prefixation Of 2,479 penultimately-stressed trisyllables extracted by the author from LDOCE , 55 (~2.22 %) have penult-ult strings such as / -ˈrentl/ of parental , which also occur in the dictionary as initially-stressed disyllables (i.e., e.g., / ˈrentl/ of rental ), and, what is particularly important, the latter are, according to LDOCE , more frequent than the former. E.g., rental is, according to the dictionary, a medium-frequency word whereas parental is a lower-frequency one; habit is a high-frequency word but inhabit is a medium-frequency one; version is a high-frequency word but conversion / diversion are lower-frequency ones. Additionally, in the case of 34 penultimately-stressed trisyllables, LDOCE gives identical characteristics as far as frequency of use is concerned. E.g., both remember and member are high-frequency words; both contribute and tribute are medium-frequency words; both contender and tender are lower-frequency words; etc. By contrast, of the 4,979 antepenultimately-stressed trisyllables, only 36 (~0.72 %) have orthographic penult-ult strings such as, e.g., famous of infamous , which, according to LDOCE , also occur in English as a more frequently used initially-stressed disyllable. (The stress pattern / ˈɪnfəməs/ ( LDOCE ) is most likely due to emphasis/ semantics: Being infamous means being famous in a particular way. Similarly, bicycle is stressed / ˈbaɪsɪkəl/ ( LDOCE ) rather than ? / baɪˈsaɪkəl/ because apart from bicycles there are also tricycles . Cranberry is stressed / ˈkrænbəri/ ( LDOCE ) because apart from the berry -word cranberry , there are also the berry -words blackberry , gooseberry , raspberry , etc.) The difference of 55 / 2,479 vs. 36 / 4,979 is statistically highly significant— χ 2 (1) = 31, p < 0.000001—which allows us to say that penultimately-stressed trisyllables are more frequently than antepenultimately-stressed ones represented in English by prefixed disyllables (and, as in the case of prefixed monosyllables (e.g., become ), it does not matter whether prEfix + basE segmentations are semantically transparent or not. Thus, for instance, although apparent and transparent are from a semantic point of view not segmentable into the prefixes ap -/ trans and the base parent the way, e.g., dishonest is segmentable into the prefix dis and the base honest , both apparent / transparent and dishonest count for an English speaker as prefixed derivatives and are therefore stressed pen-initially: / əˈpærənt/ , / trænˈspærənt/ , and / dɪsˈɒnɪst/ (OD), with stress in the disyllabic bases 4.3 Three and more syllables 105 parent and honest being initial because the strings ent and est occur in English as suffixes and therefore repel stress). This finding is particularly relevant for a synchronic understanding of why stress is penultimate in the majority of the English ic -words (which have been extensively discussed in many publications on English stress. For a summary of different approaches, see the article Hill 1974 or more recently, Kettemann 1988: 73-75). The etymological answer to the question of why, e.g., atomic is stressed / əˈtɒmɪk/ ( OD ) is that this word came into existence in English not via suffixation ( atom + ic ) but as a shortening of the Latin word atomicus ( OD ), in which stress is antepenultimate because the vowel in the penultimate syllable is short and not followed by a coda consonant. Cf. the modern Italian adjective atomico , which is stressed / aˈtɔ: miko/ ( PONS ). From a synchronic perspective, however, many English ic -words can be analyzed as prefixed disyllables. Of the 4,043 ic -words in the OD dictionary, 1,883 (~46.57 %) are words such as republic , whose righthand substrings are initially-stressed disyllabic ic -words, such as public . The reason why the trisyllabic derivative republic is stressed / rɪˈpʌblɪk/ ( OD ) is thus that its disyllabic base public is stressed / ˈpʌblɪk/ ( OD ). Indeed, the mean BNC frequency of an ic word such as republic is ~44.43, but the median frequency is zero (i.e., of the 1,883 ic -words such as republic , 1,298 (~68.93 %) lack attestations in the BNC ). By contrast, the mean BNC frequency of an ic -word such as public is ~640.05 and the median frequency is 18.5 (i.e., of the 146 ic -words such as public , 73 (50 %) occur in the BNC no less than 20 times). A longer word such as republic is thus on average less frequent than a shorter word such as public . Additionally, some authors analyze penultimate stress in an ic -word as the preserved antepenultimate stress of a corresponding longer ical -word. E.g., while historic occurs in the BNC 2,300 times, historical has 5,513 attestations; note also that the meaning of historic is, according to the OD , “famous or important in history” (e.g., historic occasion ), but historical is said to mean “concerning history or historical events” (e.g., historical evidence ). The meaning of historic thus appears to be more specialized than the meaning of historical : When something is historic , it is historical (i.e., concerns history) in a particular way: by being important in history. Similarly, as Bryson (1987: 42; italics mine) points out, “[s]omething that is comic is intended to be funny. Something that is comical is funny whether or not that is the intension” (cf. Hamawand 2011: 145, arguing that the suffixes ic and ical differ from each other with regard to semantic range: “The suffix ic is reserved to emphasize the core feature of the noun expressed in the root. By contrast, the suffix ical is reserved to merely stress the relationship between the noun and the field of knowledge or subject of study expressed in the root.”) 106 4 Stress assignment in English Do observe, however, that of the 379 ic -words with an ical -counterpart in the MRC , only 132 (~34.83 %) have fewer occurrences in the BNC than a corresponding ical -word. Many ical -words cannot thus be regarded as the base of a corresponding ic -word. E.g., democratical , which lacks attestations in the BNC , is most certainly not the base of democratic , which occurs in the corpus 5,878 times. In addition to this, even if we analyze, e.g., the more frequently used biological as the base of the less frequently used biologic —the former has 1,971 attestations in the BNC while the latter occurs in the corpus only five times—there still remains the question of why biological is stressed / ˌbaɪəˈlɒdʒɪkəl/ ( LDOCE ). The present monograph attributes this stress pattern to the prefixation analysis bio - + logical (which is in accordance with the frequency of use criterion: logical has 2,271 attestations in the BNC and is thus more frequent than biological ). Similarly, because the disyllable logic has 2,236 attestations, also biologic is easily segmentable into the prefix bio and the root logic . To obtain penultimate stress in biologic we thus do not need to back-derive it from biological . Likewise, of the 3,136 ion -words in the OD dictionary, 1,038 (~33.1 %) are words such as obsession , whose righthand substrings are words such as session , i.e., initially-stressed ion -words that contain no less than three orthographic vowels. The shorter word session is stressed / ˈsɛʃ(ə)n/ ( OD ) because the string ion , which occurs in English as a suffix (e.g., action ), repels stress. The longer session -word obsession merely preserves this stress pattern: / əbˈsɛʃ(ə)n/ ( OD ). Similar pairs include admission and mission , adoption and option , compassion and passion , diffusion and fusion , division and vision , emotion and motion , intension and tension , proportion and portion , reaction and action , refraction and fraction , etc. Stress in the prefixed derivatives admission , adoption , compassion , diffusion , division , emotion , intension , proportion , reaction , refraction , etc. is pen-initial because it is initial in the corresponding bases mission , option , passion , fusion , vision , motion , tension , portion , action , fraction , etc. The mean BNC frequency of an ion -word such as obsession is ~375.6, but the median frequency is only two. The mean BNC frequency of an ion -word such as session is, however, ~2,724.92 and the median frequency is 186. A longer ion word such as obsession is thus on average less frequent than a shorter ion -word such as session . Proceeding from these numbers, we can claim that, e.g., devastation is stressed / ˌdɛvəˈsteɪʃn/ ( OD ) because from a purely formal point of view, there are more reasons to analyze it deva - + station than devastate + ion : While the shorter verb devastate is less frequent in the BNC than the longer noun devastation (44 vs. 266), the shorter noun station is, by contrast, considerably more frequent than it: 9,929 vs. 266. Note also that the OD has 42 mental -words (excluding the word mental itself) that are pronounced / -ˈmɛnt(ə)l/ . E.g., adjustmental , compartmental , departmen- 4.3 Three and more syllables 107 tal , detrimental , developmental , elemental , environmental , experimental , fundamental , governmental , incremental , instrumental , etc. all exhibit this stress pattern. Of these longer mental -words, only environmental has more attestations in the BNC than mental : 8,319 vs. 5,735 (with the third most frequent mental -word fundamental occurring in the BNC only 4,480 times). Thus, we have good reasons to assume that the mental of these words counts morphologically as the root and hence receives stress upon its first syllable. E.g., segmental is stressed / sɛɡˈmɛnt(ə)l/ (OD) because the morphological structure of this derivative is not the “correct” (i.e., semantics-based) segment + al , but the purely formal seg - + mental . Needless to say, in some English words of three and more syllables, the root, which receives stress, is not free but bound, not existing in the language as a separate word. Thus, for instance, in contrast to the free root mission of the prefixed derivatives admission , commission , emission , intermission , intromission , omission , permission , remission , submission , and transmission , the bound root dition of the prefixed derivatives addition , audition , condition , edition , erudition , expedition , extradition , perdition , reddition , rendition , sedition , and tradition does not occur in English as a separate word. Likewise, the string matic occurs in English only at the end of 155 longer ic -words such as aromatic , asthmatic , automatic , climatic , dalmatic , diplomatic , emblematic , enigmatic , idiomatic , pragmatic , rheumatic , symptomatic , etc. As for al -words, consider, for example, the bound root ternal of eternal , external , fraternal , internal , maternal , paternal , etc. 4.3.3 More on penultimate stress In the MRC database, the proportions of penultimately-stressed items among words of three and more syllables are ~38.16 % (trisyllables) or ~40.77 % if also the stress pattern 220 (e.g., archbishop ) is taken into consideration, ~35.71 % (tetrasyllables), ~28.41 % (pentasyllables), ~20.93 % (hexasyllables), ~9.26 % (heptasyllables), and 10 % (octasyllables). For three vs. four syllables (3,685 / 9,657 or 3,937 / 9,657 vs. 2,056 / 5,758), χ 2 (1) = 9.279, p = 0.0023 or χ 2 (1) = 39, p < 0.000001; for four vs. five syllables (2,056 / 5,758 vs. 644 / 2,267), χ 2 (1) = 39, p < 0.000001; for five vs. six syllables (644 / 2,267 vs. 112 / 535), χ 2 (1) = 12.27, p = 0.0005; for six vs. seven syllables (112 / 535 vs. 10 / 108), χ 2 (1) = 7.968, p = 0.0048. The incidence of penultimate stress in English words can thus be said to decrease with an increase from three to four, four to five, five to six, and six to seven syllables. Note also that of the items that the MRC regards as penultimately-stressed tetrasyllables, 736 have hyphenations in LDOCE in which the boundary symbol (∙) occurs three times (e.g., ton∙sil∙li∙tis of tonsillitis ). Of these items, 364 108 4 Stress assignment in English (~49.46 %) have the vowel sequences io , ia , ie , ue , and eo occurring in these words’ orthographic ults; the bulk is constituted by 309 ion -words, such as, e.g., / ˌfedəˈreɪʃən/ of federation ( LDOCE ). The point here is that according to LDOCE , in contemporary English there is at least one word in which the vowel sequences io , ia , ie , ue , and eo are phonetically realized as hiatus. E.g., champion is / ˈcham-pē-ən/ ( MWO ), with the phonetic realization of io being the hiatus / ɪ-ə/ (cf. the MRC , where champion is said to be tS&m/ pj@n, with the hiatus i . o being resolved via replacing the preceding vowel i through the phonetically similar glide / j/ ; likewise, accordion is / ə-ˈkȯr-dē-ən/ in the MWO vs. @/ kO/ dj@n in the MRC ). Given words like champion , where the word-final orthographic string ion is (at least on some occasions) phonetically realized as hiatus, we might assume that also the word federation is underlyingly not -ˈ ra . tion but -ˈ ra . ti . on , with stress being antepenultimate rather than penultimate. The proportion of penultimately-stressed tetrasyllables in the MRC thus drops dramatically from ~29.99 % to ~15.16 % if items such as federation are left out of consideration. Note also that 35 penultimately-stressed tetrasyllables from the MRC occur in Speake & LaFlaur’s (2002) dictionary of foreign terms in English. 15 of these words— ballerina , campanile , chipolata , dilettante , influenza , macaroni , manifesto , maraschino , moderato , ocarina , operetta , pizzicato , propaganda , tarantella , and virtuoso —were borrowed into English from Italian, where stress is predominantly penultimate: Of the 25,925 syllabified Italian transcriptions in PONS ’ Italian-English dictionary, 20,714 (~79.9 %) exhibit penultimate stress. Given these facts, penultimate stress in English tetrasyllables such as ballerina , campanile , chipolata , dilettante , etc. may be regarded as an exceptional stress pattern, arising from the fact that these words still count for English speakers as Italian words / Italianisms. Likewise, of the 839 items from the MRC whose hyphenations in LDOCE have four occurrences of the boundary symbol (e.g., ab∙bre∙vi∙a∙tion), 289 (~34.45 %) are, according to the MRC, penultimately-stressed. 119 (~41.18 %) of these words are, however, ion -words such as abbreviation , which, similar to items such as federation , could be regarded as antepenultimately-stressed hexasyllables rather than as penultimately-stressed pentasyllables. Finally, as for hexasyllables and longer words, we can say that genuine penultimate stress does not exist in English altogether, which follows from the fact that of the 37 items that the MRC considers to be penultimately-stressed, 36 (~97.3 %) end orthographically in the strings ion (e.g., identification ), ian (e.g., humanitarian ), ia (e.g., megalomania ), eous (e.g., contemporaneous ), and ial (e.g., extraterritorial ). Stress in these words is, again, not penultimate but antepenultimate: -ˈ ca . ti . on , -ˈ ta . ri . an , -ˈ ma . ni . a , -ˈ ra . ne . ous , and -ˈ to . ri . al . 4.3 Three and more syllables 109 In summary, in tetraand pentasyllables, penultimate stress is a very rare stress pattern and among hexasyllables and longer English words this stress pattern does not exist at all. Because roots contain more phonetic segments than affixes, the analysis of a disyllabic ult-penult string as the root makes sense mainly in the case of trisyllabic English words, among which penultimate stress is almost as frequent as antepenultimate stress. E.g., the analysis of the bound string hibit as the root of the trisyllabic inhibit is entirely unproblematic not only because apart from occurring in inhibit , this string also occurs in exhibit and prohibit , but also because the disyllabic string hibit is segmentally longer than the monosyllabic string in -; the stress pattern of this word is therefore / ɪnˈhɪbɪt/ , with the location of stress being the root-prefix boundary location. Similar analyses are not entirely counterintuitive in the case of tetrasyllabic English words, which can often be segmented into a disyllabic prefix and a disyllabic root, with the latter being segmentally at least not shorter than the former. E.g., laryngitis and meningitis share the penult-ult string gitis , which receives stress upon its first syllable: / -ˈdʒʌɪtɪs/ ( OD ). The analysis of this righthand string as the root of laryngitis and meningitis makes sense because the remaining lefthand strings laryn and menin are segmentally not longer than it. (Whenever possible, a tetrasyllable is, however, segmented into a monosyllabic prefix and a (segmentally longer) trisyllabic root. E.g., analysis , catalysis , dialysis , neutralysis , and paralysis share the trisyllabic righthand string alysis and are therefore all stressed / -ˈ(-)alɪsɪs/ ( OD ). The same is true of the stress pattern / -ˈ(-)abəlɪz(ə)m/ ( OD ) of anabolism , catabolism , diabolism , and metabolism , which share the (orthographically) trisyllabic righthand string abolism .) Beginning, however, with pentasyllabic English words, the disyllabic ult-penult string is segmentally too short to count morphologically as the root. Thus, for example, given the stress pattern / ˌanθrəˈpɒlədʒi/ of anthropology ( OD ), a question can be raised as to why stress in this word does not fall upon the combining form logy , which denotes “a subject of study or interest” ( OD ). Although formally combining forms are more like affixes in that they as a rule do not occur as separate words, semantically they are more like roots in that the meanings they express are fairly concrete. E.g., the formation dramaturgy is stressed / ˈdraməˌtəːdʒi/ ( OD ), with the combining form urgy , which means “technique or art of dealing or working with (such) a product, matter, or tool” ( MWO ), receiving secondary stress. What accounts for antepenultimate stress in anthropology is that of the 542 logy -words in the OD dictionary, 523 (~96.49 %) are ology -words, i.e., an English word that ends in logy thus almost always ends in ology , i.e., e.g., 110 4 Stress assignment in English astrology , cardiology , dialectology , ethnology , futurology , geology , etc. Given this fact and given that a logy -word in the OD dictionary contains on average ~5.5 orthographic vowels (which means that many of these words are pentasyllables), we can argue that it is the trisyllabic righthand string ology (rather than the disyllabic logy ) that in a word such as anthropology counts morphologically as the root and hence receives stress upon its first syllable. Likewise, since 1) the 3,136 ion -words in the OD dictionary contain on average ~5.44 orthographic vowels and 2) of the 3,136 ion -words in the OD dictionary, 1,965 (~62.66 %) are ation -words, there are good reasons to regard this recurrent righthand trisyllabic string (occurring at the end of the majority of the English ion -words) as the root. E.g., situation is stressed / ˌsɪtʃuˈeɪʃən/ ( LDOCE ) because its morphological structure is not situate + ion (which cannot be the case because situate is considerably less frequent in the BNC than situation : 48 vs. 15,708; situate is thus a back-derivative from situation ) but situ - + ation , of which the righthand bound string ation , which contains three orthographic vowels, is more like a root, whereas the lefthand bound string situ -, which contains only two orthographic vowels, is more like a prefix. Note also that the 1,220 ity -words in the OD dictionary, which contain on average ~5.34 orthographic vowels, fall into 286 ility -words ( ability ), 224 ality -words ( sexuality ), 126 icity -words ( heroicity ), 93 ivity -words ( activity ), 64 osity -words ( curiosity ), 49 arity -words ( familiarity ), 27 inity -words ( affinity ), 25 anity -words ( Christianity ), 22 idity -words ( fluidity ), 22 acity -words ( loquacity ), 17 erity -words ( austerity ), 16 neity -words ( spontaneity ), etc. The bulk (~80 %) of ity -words in English is thus constituted by words ending in the twelve righthand trisyllabic strings ility , ality , icity , ivity , osity , arity , inity , anity , idity , acity , erity , and neity . Given that the average syllabic length of an ity -word in English is (close to) five, these twelve recurrent righthand strings can be regarded as the roots of the corresponding longer ity -words. Stress in these ity -words should therefore be antepenultimate. Among the 1,274 ian -words in the OD , which contain on average ~4.53 orthographic vowels, are 154 arian -words ( libertarian ), 90 onian -words ( Amazonian ), 77 ician -words ( electrician ), 57 erian -words ( Oliverian ), 42 inian -words ( Carolinian ), 34 anian -words ( Lithuanian ), etc. Stress in words such as these should likewise be antepenultimate. To reiterate: In the majority of non-initially-stressed polysyllabic English words, the righthand string that receives stress (in addition to occurring as a separate word / bound string of at least one other English word) is also segmentally longer than the remaining lefthand string, which is not stressed or secondary-stressed. Because of this fact, penultimate stress is a frequent stress pattern only among trisyllabic English words. Among tetraand pentasyllables, 4.3 Three and more syllables 111 this stress pattern is very rare and among words of six and more syllables, this stress pattern does not exist altogether. 4.3.4 Stress non-preservation In Section 3.2 (as well as in 4.3.1) it was reported that stress preservation represents the default outcome of English suffixation: 70 %. At the same time, however, as observed in 3.2, it is not always clear whether the stress pattern of a derivative is due to the stress pattern of its base or due to abidance by the Latin Stress Rule. Is, for example, prioritize stressed / prʌɪˈɒrətʌɪz/ because priority is stressed / prʌɪˈɒrɪti/ or because the penult / rə/ is light? A different case is, however, derivatives such as presidency and indicator , in which stress falls upon a syllable farther than the antepenult. Since the Latin Stress Rule only has provisions for penultimate and antepenultimate stress, we can be sure that the stress patterns / ˈprɛzɪd(ə)nsi/ and / ˈɪndɪkeɪtə/ are due to the stress patterns / ˈprɛzɪd(ə)nt/ and / ˈɪndɪkeɪt/ . Therefore, derivatives such as presidency and indicator (i.e., these derivatives are initially-stressed tetrasyllables and longer words whose bases are also initially-stressed or monosyllables) will in the following be contrasted with derivatives such as, e.g., maniacal , which similar to presidency and indicator was derived from an initially-stressed trisyllabic base ( maniac ), but in spite of this fact, is stressed non-initially: / məˈnaɪəkəl/ ( LDOCE ). Why does the derived form maniacal not preserve the stress of the base / ˈmeɪniæk/ ( LDOCE )? Relying upon morphological segmentations returned by the tool Morphological Analysis, phonetic transcriptions/ hyphenations in LDOCE , and frequency numbers from the BNC , English suffixes were classified by the author into the following two groups: 1) able (e.g., actionable ), ary (e.g., budgetary ), ate (e.g., oxygenate ), cy (e.g., advocacy ), ed (e.g., concentrated ), er (e.g., advertiser ), ery (e.g., imagery ), ing (e.g., elevating ), ism (e.g., structuralism ), ist (e.g., parachutist ), ite (e.g., meteorite ), ive (e.g., educative ), ize (e.g., actualize ), ly (e.g., adequately ), ness (e.g., seriousness ), or (e.g., operator ), ry (e.g., citizenry ), ship (e.g., citizenship ), ty (e.g., admiralty ), ure (e.g., candidature ), y (e.g., orthodoxy ) vs. 2) al (e.g., accidental ), an (e.g., diocesan ), ant (e.g., luxuriant ), ar (e.g., molecular ), arian (e.g., disciplinarian ), at (e.g., secretariat ), ent (e.g., beneficent ), etic (e.g., sympathetic ), ia (e.g., suburbia ), ible (e.g., accessible ), ic (e.g., alcoholic ), ical (e.g., liturgical ), icity (e.g., multiplicity ), iferous (e.g., carboniferous ), ific (e.g., scientific ), ify (e.g., humidify ), ion / ation (e.g., deforestation ), istic (e.g., euphemistic ), ity (e.g., ability ), ography (e.g., iconography ), ology (e.g., methodology ), ory (e.g., supervisory ). (Note that these are not all English suffixes but only those that according to the tool Morphological Analysis, occur in 112 4 Stress assignment in English tetrasyllabic and longer suffixed derivatives whose bases are initially-stressed or monosyllabic. For an encyclopedic survey of the stress behavior of different English suffixes, see the monograph Trevian 2015.) The 21 suffixes in the first group are either exclusively or more frequently stress-neutral. For instance, although the suffix ist does sometimes act as a stress-shifting suffix—e.g., saxophonist is / sækˈsɒfənɪst/ in British English ( LDOCE ), non-preserving the stress of the base / ˈsæksəfəʊn/ ( LDOCE )—the usual outcome of an ist -derivation is stress preservation. Thus aerialist , capitalist , caricaturist , generalist , manicurist , miniaturist , nationalist , naturalist , parachutist , rationalist , sensualist , and separatist are all stressed initially, preserving the stress of the corresponding base forms aerial , capital , caricature , general , manicure , miniature , national , natural , parachute , rational , sensual , and separate ; note also that in American English saxophonist is, in contrast to British English, stressed / ˈsæksəfoʊnɪst/ ( LDOCE ), preserving the stress of the base saxophone . The 23 suffixes in the second group are, by contrast, either exclusively or more frequently stress-shifting. Thus the examples given above— accidental , diocesan , luxuriant , molecular , disciplinarian , secretariat , beneficent , sympathetic , suburbia , accessible , alcoholic , liturgical , multiplicity , carboniferous , scientific , humidify , deforestation , euphemistic , ability , iconography , methodology , supervisory —are all stressed non-initially even though the bases of these derivatives are, according to the tool Morphological Analysis, the initially-stressed accident , diocese , luxury , molecule , discipline , secretary , benefit , sympathy , suburb , access , alcohol , liturgy , multiple , carbon , science , humid , forest , euphemism , able , icon , method , and supervise . The 789 suffixed derivatives in LDOCE that end in these 23 suffixes contain on average ~4.43 orthographic vowels. Of these words, 501 (~63.5 %) are, according to LDOCE , words of four and more syllables. By contrast, the 3,505 suffixed derivatives that end in the 21 suffixes from the first group contain on average only ~3.24 orthographic vowels. Of these words, only 844 (~24.08 %) are, according to LDOCE , words of four and more syllables. The difference of 501 / 789 vs. 844 / 3,505 is statistically hugely significant— χ 2 (1) = 465, p < 0.000001—which allows us to say that stress-shifting suffixes occur in longer words than stress-neutral suffixes. Given this finding, we reiterate that deviations from stress preservation occur in English when a word that ends in a suffix is (from a purely formal point of view) analyzed by English speakers as a word that begins with a prefix. E.g., the morphological structure of suburbia that results in the stress pattern / səˈbɜːbiə/ ( LDOCE ) is not suburb + ia (which is the segmentation returned by the tool Morphological Analysis) but sub - + urbia (cf. the stress pattern / ɛkˈsəːbɪə/ of exurbia ( OD ), with stress in this prefixed derivative likewise falling upon the 4.3 Three and more syllables 113 bound root urbia ); labyrinthian is / lab(ə)ˈrɪnθɪən/ ( OD ) because its morphological structure is not labyrinth + ian but laby - + rinthian (cf. / kəˈrɪnθɪən/ of Corinthian ( OD ), with stress in this prefixed derivative likewise falling upon the bound root rinthian ); artificial , beneficial , maleficial , official , orificial , pontificial , sacrificial , superficial , and surficial (some of which can from a semantic point of view be seen as suffixed derivatives; e.g., sacrificial ← sacrifice + ial ) share the bound root ficial and are therefore all stressed / -ˈfɪʃ(ə)l/ ( OD ); advantageous is not advantage + ous but advant - + ageous and is therefore stressed / -ˈ-eɪdʒəs/ ( LDOCE ), which is also the stress pattern of courageous , outrageous , and rampageous ; although accidental is from a semantic point of view not segmentable into acci and dental the way, e.g., interdental is segmentable into inter and dental , the reason why accidental is stressed / aksɪˈdɛnt(ə)l/ ( OD ) is identical to that why interdental is stressed / ɪntəˈdɛnt(ə)l/ ( OD ). Because the string dental occurs in English as a separate word, there are good reasons to regard this string as the root of both accidental and interdental (even though, as just said, it is only the latter where this analysis makes sense also from a semantic point of view). Cases such as these usually take place when the syllabic length of a suffixed derivative is no less than four because, as explained in 4.3.3, tetrasyllabic words in English are often segmentable into a monosyllabic prefix and a trisyllabic root (e.g., sub - + urbia of suburbia ) or into a disyllabic prefix and a disyllabic root (e.g., acci - + dental of accidental ). Similarly, pentasyllables in English are usually segmentable into a disyllabic prefix and a trisyllabic root (e.g., bene -/ male - + ficial of beneficial / maleficial , which is in part semantically transparent). These claims are also strongly supported by variably-stressed derivatives such as, e.g., the ly -derivatives ordinarily / extraordinarily , momentarily , necessarily , primarily , and voluntarily , which all share the trisyllabic righthand string arily . E.g., voluntarily is / ˈvɒləntərəli/ vs. / ˌvɒlənˈterəli/ in British English and only / ˌvɑːlənˈterəli/ in American English ( LDOCE ) because in addition to being “correctly” segmented into the initially-stressed base voluntary and the suffix ly , the very same word voluntarily can also be “incorrectly” segmented into the disyllabic prefix volun and the trisyllabic root tarily . Condemnatory is / kənˈdemnətəri/ vs. / ˌkɒndemˈneɪtəri/ in British English and only / kənˈdemnətɔːri/ in American English (LDOCE). On the one hand, the stress pattern / -ˈneɪtəri/ is the preserved stress of / kɒndəmˈneɪʃən/ of condemnation ( LDOCE ), which has considerably more attestations in the BNC than condemnatory : 442 vs. 13. On the other hand, it is most likely not a coincidence that similar deviations from stress preservation are also exhibited (in British English) by compensatory , participatory , regulatory , circulatory , and celebratory , which all share the trisyllabic righthand string atory . Accordingly, to account for antepenultimate stress in these words, not only back-derivation but also prefix- 114 4 Stress assignment in English ation analyses can be invoked. A similar case is / ˌsætəsˈfæktəri◀/ of satisfactory ( LDOCE ), which means that stronger stress in the word under consideration can occur not only antepenultimately but also initially. On the one hand, both these stress patterns can be attributed to the stress pattern / ˌsætəsˈfækʃən/ of the more frequently used word satisfaction , which, as pointed out in 4.3.1, occurs in the BNC 2,761 times whereas satisfactory has only 2,160 attestations. Note, however, that the initially-stressed trisyllabic word factory occurs in the BNC 4,475 times and is thus even more frequent than satisfaction . The prefixation analysis satis - + factory , which results in the stress pattern / ˌsætəsˈfæktəri/ , should thus make more sense than the back-derivation from satisfaction . The same is true of manufactory and ratifactory , which also end in factory and are therefore stressed / -ˈfakt(ə)ri/ ( OD ). Finally, it seems that in some cases, deviations from stress preservation have purely diachronic causes. E.g., while the base contemplate has abandoned the historical stress pattern / kənˈtɛmpleɪt/ in favor of antepenultimate stress, the derivative contemplative is still stressed not only / ˈkɒntəmpleɪtɪv/ but also / kənˈtemplətɪv/ ( LDOCE ). Similarly, demonstrate is no longer / dɪˈmɒnstreɪt/ , but demonstrable is still / dɪˈmɒnstrəbəl/ alongside / ˈdemənstrəbəl/ ( LDOCE ). A possible explanation for these facts is that because tetrasyllables in English are usually made up of a monosyllabic prefix and a trisyllabic root, the antepenultimately-stressed pronunciations / kənˈtemplətɪv/ and / dɪˈmɒnstrəbəl/ are not perceived by contemporary English speakers as “wrong” pronunciations, even though they are as a matter of fact “wrong.” The suffixes able and ive are, as pointed out above, stress-neutral suffixes. E.g., educative is only / ˈedjʊkətɪv/ in British English and only / ˈedʒəkeɪtɪv/ in American English (LDOCE), preserving the stress (and in the case of the American pronunciation, also the segmental structure) of the base educate . Likewise, educable is stressed only / ˈedjʊkəbəl/ (LDOCE), preserving the stress of the base educate . (Note also that apart from the pseudo-derivative demonstrable , which was imported into English from Latin, the OD also has the genuine able -derivative demonstratable , which is stressed only / ˈdɛmənstreɪtəbl/ , preserving not only the stress but also the segmental structure of the base demonstrate . Likewise, the pronunciations / ˈeksəkjuːtəbəl/ and / ˈekstrədaɪtəbəl/ ( LDOCE ) of the genuine able -derivatives executable and extraditable also exemplify agglutinative suffixation.) Accordingly, also contemplative and demonstrable are supposed to be stressed / ˈkɒntəmpleɪtɪv/ and / ˈdemənstrəbəl/ rather than / kənˈtemplətɪv/ and / dɪˈmɒnstrəbəl/ . 4.3 Three and more syllables 115 4.3.5 Stress shifts For 929 solidly-spelled trisyllables and longer words, LDOCE gives phonetic transcriptions that contain the stress shift symbol (◀). E.g., European is / ˌjʊərəˈpiːən◀/ , which means that the actual stress pattern of European is on some occasions / ˈjʊərəpiːən/ . Thus: When a word (simple or compound) pattern consists in isolation of a primary accent preceded by a secondary accent, the primary accent may be lost, if, in connected speech, a strong accent follows closely. (Cruttenden 2008: 296) E.g., European often occurs in the modifier position (e.g., European Union ), which is why the stress pattern / ˈjʊərəpiːən/ is from the point of view of rhythm better than the stress pattern / ˌjʊərəˈpiːən/ . Note, however, that stress shifts also occur in antepenultimately-stressed words, such as elementary : / ˌeləˈmentəri◀/ ( LDOCE ), in which the stressed syllable / ˈmen/ is separated from the right word boundary by the unstressed sequence / təri/ . Accordingly, even when elementary occurs in combinations such as elementary school , there are no less than two unstressed syllables separating two stressed ones. Is a two-syllable distance between syllables bearing stress also rhythmically unfortunate? Note also that of the 331 occurrences of Pennsylvania in the BNC , only 65 (~19.64 %) are occurrences in combinations such as Pennsylvania avenue , Pennsylvania state , Pennsylvania law , etc., with Pennsylvania being adjacent to a noun. Occurrences in the modifier position are thus relatively rare in the case of Pennsylvania , but the stress pattern of this word is, according to LDOCE , / ˌpensəlˈveɪniə◀/ ; the actual pronunciation of Pennsylvania is thus on some occasions the initially-stressed / ˈpensəlveɪniə/ . Given variably-stressed words such as elementary and Pennsylvania , the present monograph argues that stress shifts such as / ˌjʊərəˈpiːən◀/ are conditioned not only rhythmically but also morphologically. Of the 929 stress shifts that according to LDOCE , occur in trisyllables and longer English words, 406 (~43.7 %) occur in words that end orthographically in al , an , ic , or ous . As pointed out in 4.3.4, these endings are associated with suffixed derivatives that do not preserve the base stress. E.g., territorial is stressed / ˌtɛrɪˈtɔːrɪəl/ ( OD ) because it was derived not from the English base territory , in which stress is initial, but from the Latin base territōrium (Dictionary.com), in which stress is antepenultimate because the vowel in the penultimate syllable is short. From a synchronic point of view, the stress pattern / ˌtɛrɪˈtɔːrɪəl/ still makes sense because apart from occurring in territorial , the trisyllabic string torial also occurs in, e.g., editorial . 116 4 Stress assignment in English At the same time, however, because in LDOCE , territorial is said to be / ˌterəˈtɔːriəl◀/ , we know that the actually occurring stress pattern of territorial is on some occasions the initially-stressed / ˈterətɔːriəl/ . Assuming that this stress pattern is not only a matter of rhythm (for even when territorial occurs in combinations such as territorial gains , the primary-stressed syllable / ˈtɔː/ is separated from the right word boundary by the unstressed sequence / riəl/ ), we argue that the stress pattern / ˈterətɔːriəl/ is simply the preserved stress of the (English) base territory , which occurs in the BNC 3,065 times while territorial has only 1,192 attestations. Similarly, the above mentioned accidental , alcoholic , carboniferous , euphemistic , scientific , and sympathetic do as a matter of fact sometimes preserve the stress of the corresponding base forms accident , alcohol , carbon , euphemism , science , and sympathy . E.g., scientific is, according to LDOCE , stressed / ˌsaɪənˈtɪfɪk◀/ , with the initially-stressed pronunciation / ˈsaɪəntɪfɪk/ occurring in, e.g., there are those who believe in a ˈ scientific basis for attraction ( Twin Peaks , Season Two, Episode 20). The same is true of so-called stress-attracting suffixes. Given the transcriptions / ˌæbsənˈtiː◀/ of absentee , / ˌbrɪɡəˈdɪə◀/ of brigadier , / ˌdɒktrəˈneə◀/ of doctrinaire , / ˌleməˈneɪd◀/ of lemonade , / ˌpɪktʃəˈresk◀/ of picturesque , etc. ( LDOCE ), it is clear that the stress-attracting suffixes ee , ier , aire , ade , and esque are as a matter of fact, just like other English suffixes, at least on some occasions stress-neutral. (Once again, it is fairly obvious that the stress shifts exhibited by these words cannot be exclusively attributed to stress clash avoidance. E.g., 16 occurrences of lemonade in environments such as lemonade bottle constitute only ~7.51 % of the 213 total occurrences of lemonade in the BNC . The reason why lemonade is stressed / ˈleməneɪd/ is that from a synchronic point of view, there are absolutely no reasons to stress it / leməˈneɪd/ , i.e., the suffix ade does not have a strong emphatic potential and should therefore be stress-repellent.) As for Pennsylvania , which literally stands for “Penn’s Woods” (Dictionary. com), observe that etymologically, the word is “a hybrid formed from the surname Penn (Welsh, literally “head”) + Latin sylvania ” (cf. Transylvania ). Although contemporary English speakers may not know that Pennsylvania was named after the father of William Penn (Dictionary.com), they can nonetheless folk-etymologically guess that the name Pennsylvania has something to do with a person named Penn . Stress in Pennsylvania should therefore be initial, falling upon the only syllable constituting the monosyllabic base Penn . Finally, we note that also emphatic stress shifts “involve an interchange of primary and secondary stress” (Berg 2008: 165; see also Friederich 1967: 62). Thus, for instance, according to Kenyon & Knott (1953[1944]: xxv): 4.3 Three and more syllables 117 When such adverbs as ˈ abso ˌ lutely , ˈ neces ˌ sarily , ˈ ordi ˌ narily are made emphatic, they are likely to be accented ˌ abso ˈ lutely , ˌ neces ˈ sarily , ˌ ordiˈnarily . Most adverbs of this type show this tendency in American English, and in some cases this accentuation has become generalized; i.e., many speakers say ˌ ordiˈnarily , ˌ neces ˈ sarily under all circumstances, even when these are not emphatic. 4.3.6 Segmental length in trisyllables Of the 4,000 syllabified phonetic transcriptions given in the MRC for 3,685 orthographically non-identical penultimately-stressed trisyllables, 1,969 (49.225 %) have penults that contain more phonetic segments than the corresponding neighboring antepenult (e.g., I/ lek/ tr0n of e ˈ lectron ); the average values / weighted arithmetic means are ~2.76 (penult) vs. ~2.29 (antepenult). By contrast, in the case of the 5,164 transcriptions given in the database for 4,794 orthographically non-identical trisyllables in which stress is antepenultimate, the same is true of only 1,195 (~23.14 %) transcriptions (e.g., e/ lI/ m@nt of ˈ element ); the average values are ~2.25 (penult) vs. ~2.39 (antepenult). This difference is statistically hugely significant: χ 2 (1) = 678, p < 0.000001. Note also that in the case of 849 antepenultimately-stressed trisyllables from the MRC database and 335 penultimately-stressed ones, the OD gives line breaks such as, e.g., ac¦ci|dent and ac¦com|plish, where “[a] bar(‘|’) indicates a preferred or primary division point, at which a word can be divided under almost any circumstances,” whereas “[a] broken bar indicates a secondary division point, at which a word is best divided only if absolutely necessary” (http: / / tinyurl.com/ h46ys8d, 27. 03. 2016). In both the line break ac¦ci|dent and the line break ac¦com|plish, the primary division point is the ult-penult boundary, whereas the secondary division point coincides with the penult-antepenult boundary. Since the righthand monosyllabic strings dent and plish cannot take stress, because ent and ish occur in English as suffixes (e.g., urgent , banish ), the lefthand disyllabic string ac . ci of accident , in which both syllables consist of two orthographic symbols, receives stress on its first syllable, / ˈaksɪd(ə)nt/ ( OD ), while in ac . com of accomplish , stress falls upon the segmentally longer ult . com : / əˈkʌmplɪʃ/ ( OD ). (Note that accomplish has more attestations in the BNC than accomplishment : 257 vs. 131. The back-derivation analysis is thus not invoked to account for penultimate stress in accomplish .) Of the 849 line breaks such as ac¦ci|dent, 680 (~80.09 %) are line breaks such as al¦co|hol or cal¦cu|late, in which the number of orthographic symbols constituting the antepenult is identical to / exceeds that of the neighboring penult; the 118 4 Stress assignment in English weighted arithmetic means are ~2.77 (antepenult) vs. ~2.47 (penult). By contrast, in the case of the 335 line breaks such as ac¦com|plish, the same can be said about only 67 (20 %) line breaks (e.g., com¦mit|ment), with the average values being 2.29 (antepenult) vs. ~3.34 (penult). The difference of 680 / 849 vs. 67 / 335 is also statistically hugely significant: χ 2 (1) = 373, p < 0.000001. Some English trisyllables vacillate between penultimate and antepenultimate stress because the latter is required by the segmental length principle whereas the former is in accordance with the analysis of a word as a prefixed derivative. E.g., composter is, according to the OED , in British English either / ˈkɒmpɒstə/ or / kəmˈpɒstə/ ( OED ), with antepenultimate stress being in accordance with the segmental length principle ( com . po - + ster / com . pos - + ter ; it does not matter whether the medial / s/ belongs to onset of the ult or to coda of the penult) vs. penultimate stress abiding by the segmentation into the prefix com and the root poster , which occurs in English as a separate word. Similarly, the OD has four sequent -adjectives: Consequent and subsequent have only the prefix stress / ˈkɒn-/ and / ˈsʌb-/ while obsequent and resequent vacillate between root and prefix stress, i.e., / ˈɒbsᵻkw(ə)nt/ vs. / ɒbˈsiːkw(ə)nt/ and / ˈriːsᵻkw(ə)nt/ vs. / riˈsiːkw(ə)nt/ . As was pointed out in Chapter 1, in a word beginning with a prefix, primary stress falls upon it when the prefix is semantically important (e.g., triangle ) or, if the prefix is semantically opaque, when it makes the word containing it formally different from another semantically related word (e.g., defense and offense ). Accordingly, since one of the meanings of subsequence is “[a] sequence contained in or forming part of another sequence” ( OD ), stress in this prefixed derivative can be emphatically placed upon the prefix sub -. Nonetheless, it is clear that emphasis cannot be the sole explanation why consequent and subsequent are exclusively stressed / ˈkɒnsɪkw(ə)nt/ and / ˈsʌbsɪkw(ə)nt/ , for, just as in the case of the prefixed derivative subsequence , also the prefix sub of the prefixed derivative substandard , “[b]elow the usual or required standard” ( OD ), modifies the meaning of the base standard in an important way. In spite of this fact, however, substandard is stressed / sʌbˈstandəd/ ( OD ). What accounts for antepenultimate stress in the trisyllabic sequent -derivatives is that in con . se of consequent and sub . se of subsequent the first syllables are segmentally longer than the second syllables and in ob . se of obsequent and re . se of resequent the segmental length of the first syllables is identical to that of the second syllables. Notice also that since the ult . men of abdomen contains the string en , which occurs in English as a suffix (e.g., blacken ), and since the remaining disyllabic string ab . do should have initial stress in accordance with the segmental length principle, there is no reason why Present-day English should still use the historical pronunciation / æbˈdəʊmən/ , preserving the stress of the Latin etymon 4.3 Three and more syllables 119 abdōmen (OED), in which the vowel in the penult is long. Thus, of 58 native (predominantly American) English speakers who were found to have pronounced the word abdomen in captioned YouTube videos, everyone (100 %) used initial stress in this word, i.e., abdomen is in contemporary American English exclusively stressed / ˈæbdəmən/ . Likewise, of the 377 syllabified phonetic transcriptions given in the MRC for antepenultimately-stressed ate -trisyllables, 241 (~63.93 %) have antepenults whose segmental length exceeds / is identical to that of the corresponding neighboring penult. Accordingly, even if, e.g., the shorter verb compensate (which occurs in the BNC only 862 times) did not count synchronically as a back-derivative from the longer noun compensation (which has 3,128 attestations), the former would nonetheless be stressed / ˈkɒmpənseɪt/ rather than / kəmˈpɛnseɪt/ simply because the segmental length principle requires initial stress in its lefthand disyllabic string com . pen -. Another similar case is the variation between final and antepenultimate stress in the prefixed derivative premature . Unlike the OED , which still gives the finally-stressed pronunciations / ˌprɛməˈtʃɔː/ and / ˌprɛməˈtjɔː/ , i.e., stress in the trisyllabic prefixed derivative premature is supposed to be final because it is final in the disyllabic base mature (= “exhibiting maturity,” which is how it is defined in WordNet), the OD gives for premature only the initially-stressed transcriptions / ˈprɛmətʃə/ and / ˈprɛmətjʊə/ . The reason why stress in premature more frequently falls upon the prefix pre is clearly not only emphasis (cf. / priːdɪˈtəːmɪn/ of predetermine ( OD ), with primary stress in this prefixed derivative not falling upon the prefix pre -) but the fact that in the disyllabic string pre . ma -, the first syllable contains more segments than the second syllable, whereas the final syllable . ture contains the string ure , which occurs in English as a suffix (cf. closure ), and should therefore be stress-repellent. (Apart from this, it should be added that while the shorter adjective premature occurs in the BNC 802 times, the longer noun prematurity has only 24 attestations. The former is thus most likely not a back-derivative from the latter. Additionally, the American English transcription / ˌpriːməˈtʃʊr◀/ ( LDOCE ) contains the stress shift symbol, which suggests that antepenultimate stress in premature is also a matter of rhythm.) Non-abidance by the segmental length principle is also a reason why in the case of some trisyllabic suffixed derivatives, the stress of the base form is not preserved in the derived form. A case in point is antepenultimate stress in the words afference , conference , deference , difference , efference , inference , preference , reference , and transference , some of which can from a synchronic point of view be seen as ence -suffixations derived from finally-stressed disyllabic bases (e.g., preference ← prefer + ence ). The reason why preference is stressed / ˈprɛf(ə)r(ə)ns/ ( OD ), non-preserving the stress of / prɪˈfəː/ of prefer ( OD ), is that the segmental 120 4 Stress assignment in English length principle requires initial stress in the lefthand disyllabic string pre . fe - (and the righthand monosyllabic string rence , which contains the suffix ence , repels stress). Similarly, in the lefthand disyllabic strings af . fe- , con . fe- , de . fe- , dif . fe- , ef . fe- , in . fe- , re . fe- , and tran . sfe -, stress is also supposed to be initial in accordance with the segmental length principle, i.e., the ult in these disyllabic strings is segmentally not longer than the corresponding penult. Another similar case is vaginal , vacillating between the stress patterns / vəˈdʒʌɪnl/ and / ˈvadʒᵻnl/ ( OD ). The latter stress pattern respects the segmental length principle (i.e., the segmentation of vaginal into the initially-stressed disyllabic string va . gi and the stress-repellent monosyllabic string nal ) more than the stress pattern / vəˈdʒʌɪnə/ of the base vagina ( OD ). Finally, notice that according to the OED , the stress variation / rᵻˈfəːrəbl/ vs. / ˈrɛf(ə)rəbl/ could originally correlate with the orthographic variation referrable vs. referable . Indeed, to preserve the stress of the base refer in the derivative referable , the former should be spelled referrable rather than referable , i.e., in the latter case, the segmental length principle requires initial stress in the remaining lefthand disyllabic string re . fe -. 4.3.7 Suffix-like endings Just like in disyllabic English words, the ult of a word of three and more syllables is stress-repellent if it contains a string of letters that occurs in English as a suffix. For instance, although the shorter verb automate has relatively recently (1950s, according to the OD ) appeared in the English language as a back-formation from the longer noun automation , it is pronounced not with final but with antepenultimate stress, / ˈɔːtəmeɪt/ ( OD ), because the string ate occurs in English as a suffix and therefore repels stress. Cf. the stress pattern / rɛzəˈrɛkt/ of the shorter verb resurrect ( OD ), which appeared in the English language as a back-derivative from the longer noun resurrection , / rɛzəˈrɛkʃ(ə)n/ ( OD ), in the late 18th century ( OD ). Because (according to both the OD dictionary and the tool Morphological Analysis) the righthand strings ect / rect do not occur in English as suffixes, final stress in the back-derived verb resurrect is a possible stress pattern. Similarly, because contemporary English does not have the suffixes ict / dict , the verb contradict can be stressed / kɒntrəˈdɪkt/ ( OD ), preserving the stress of the more frequently used noun contradiction . Notice also that temporary is stressed / ˈtɛmp(ə)rəri/ ( OD ) because the string ary , which occurs in English as a suffix (cf. primary ), cannot take stress; stress in temporary can therefore, in accordance with the segmental length principle, only fall upon the first syllable of its lefthand disyllabic string tem . po -. Similarly, in ordinary and military stress can only fall upon the initial syllables of 4.3 Three and more syllables 121 their lefthand disyllabic strings or . di and mi . li -, i.e., / ˈɔːdɪn(ə)ri/ and / ˈmɪlɪt(ə)ri/ ( OD ). Of the 27 CVCVCVCV -words in the OD that end orthographically in ary , 23 (~85.19 %) are initially-stressed words such as military ( coronary , culinary , funerary , honorary , lapidary , literary , luminary , monetary , sanitary , etc.). A slightly different case is / ˈdɪkʃ(ə)n(ə)ri/ of dictionary ( OD ), where stress is initial because both the strings ion and ary occur in English as suffixes (e.g., action , primary ) and therefore repel stress. The same can be said about / ˈlæbrətɔːri/ of laboratory in American English ( LDOCE ), where stress is, again, initial because the strings or and atory occur in English as suffixes (e.g., actor , accusatory ) and are therefore stress-repellent. (Note that in contrast to LDOCE , the OED gives the American transcription / ˈlæbrəˌtɔri/ , where the penultimate syllable is marked as bearing secondary stress. The point here is that the etymon of the English word laboratory is the Latin labōrātōrium (Dictionary.com), which has a long vowel in the stressed antepenult . tō . This long vowel is (still) preserved in the American pronunciation / ˈlæbrətɔːri/ , which is why its penultimate syllable / tɔː/ is perceived as bearing secondary stress. Similarly, the above mentioned territory is / ˈterətəri/ in British English vs. / ˈterətɔːri/ in American English ( LDOCE ), with the latter pronunciation preserving the long vowel ō of the Latin etymon word territōrium .) A related finding is that of the 801 trisyllables in LDOCE that end orthographically in y , 709 (~88.51 %) have antepenultimate stress, and, if only the orthographic structure CVCVCV (e.g., luxury ) is taken into consideration, the proportion of initially-stressed y -trisyllables rises to (79 / 82=)~96.34 % (i.e., canary , jalopy , and milady are the only penultimately-stressed CVCVC y -trisyllables in LDOCE ). By contrast, of the 97 CVCVCV -trisyllables that end orthographically in a / e / i / o / u (e.g., banana , finale , bikini , kimono , Sudoku ), only 36 (~37.11 %) have antepenultimate stress; e.g., / ˈkæmərə/ of camera ( LDOCE ). The difference of 79 / 82 vs. 36 / 97 is statistically hugely significant: χ 2 (1) = 68, p < 0.000001. Given the suffixes cy (e.g., bankruptcy ), -( i ) fy (e.g., amplify ), ly (e.g., quarterly ), ry (e.g., rivalry ), sy (e.g., folksy ), ty (e.g., royalty ), and the suffix y itself (e.g., girly ), the finding that among trisyllables such as luxury , which end orthographically in y , stress in English is almost exclusively antepenultimate does not seem surprising. Thus, although luxury cannot be segmented into the base form luxu and the suffix ry the way, e.g., rivalry is segmentable into the base form rival and the suffix ry , the fact that the string ry occurs in English as a suffix means that in accordance with the segmental length principle, stress in luxury can only fall upon the initial syllable of its lefthand disyllabic string lu . xu -: / ˈlʌkʃ(ə)ri/ ( OD ). Similarly, although the ult . ly of family is not an actual suffix, the fact that it occurs in English as a suffix (e.g., merely ) means that stress in family can only fall upon the initial syllable of its lefthand disyllabic string 122 4 Stress assignment in English fa . mi -, i.e., / ˈfamɪli/ ( OD ). Since similar analyses apply to all other CVCVCV words whose final letter is y (e.g., legacy , ratify , salary , deputy , etc.), stress patterns other than initial stress are virtually non-existent in English among words such as deputy , family , legacy , luxury , ratify , salary , etc. 4.3.8 Heavy ults in trisyllables According to Domahs et al. (2014: 76), who have studied how trisyllabic nonsense words ( serdanuls , huzaldomp , folmadoth , fekomof , etc.) are stressed in German, English, and Dutch, “an increase in heaviness of the final syllable increases the chances of antepenultimate stress in all three languages.” This finding is in accordance with what Chomsky & Halle (1968: 77-79) had labeled the Alternating Stress Rule. This rule is believed to retract stress from the ult that contains a long vowel / diphthong / triphthong to the antepenult. E.g., anecdote is stressed / ˈænɪkdəʊt/ ( LDOCE ) because the ult / dəʊt/ contains a diphthong. Recall, however, that, as observed in 4.1, monosyllabic English words are with the exception of function words never light. Accordingly, a trisyllabic English word whose ult is heavy can (from a purely formal point of view) be seen as a combination of two words: a disyllable and a monosyllable. Because compounds in English are more frequently leftthan right-prominent, stronger stress in trisyllables such as these will naturally occur initially rather than finally. This claim is strongly supported by the following numbers. Of the 4,794 antepenultimately-stressed trisyllables in the MRC, 1,401 (~29.22 %) end orthographically in a string that according to the OD , exists in English as a monosyllable of at least four letters (the median orthographic length of a monosyllabic English word). Examples include lute of absolute , lope of antelope , fact of artifact , comb of catacomb , poop of nincompoop , etc. By contrast, in the case of the 3,685 penultimately-stressed trisyllables in the MRC , the number of such words is only 317: ~8.6 %. E.g., rode of electrode . The difference of 1,401 / 4,794 vs. 317 / 3,685 is statistically hugely significant: χ 2 (1) = 548, p < 0.000001. Likewise, if also three-letter monosyllables (e.g., fit of benefit and bit of exhibit ) are taken into consideration, the corresponding proportions are 1,650 / 4,794(=~34.42 %) vs. 712 / 3,685(=~19.3 %), which also counts as a statistically hugely significant difference: χ 2 (1) = 236, p < 0.000001. Thus, if the ult of a trisyllabic English word occurs in English as a separate (monosyllabic) word, stress in the former is more likely to fall on the first rather than on the second syllable. 4.3 Three and more syllables 123 4.3.9 Foreignness Of the 817 finally-stressed trisyllables and longer words from the MRC database, 71 (~8.69 %) have an entry in Speake & LaFlaur’s (2002) dictionary of foreign terms in English. E.g., accoucheur , aquarelle , arabesque , bassinet , chevalier , comedienne , commissar , commissionaire , enfilade , etc. By contrast, of the other 17,550 trisyllables and longer words in the MRC (i.e., words in which primary stress is non-final), only 701 (~3.99 %) also occur in Speake & LaFlaur’s (2002) dictionary. E.g., avalanche , which is a borrowing from French, is stressed / ˈavəlɑːnʃ/ ( OD ). Since the difference of 71 / 817 vs. 701 / 17,550 is statistically highly significant— χ 2 (1) = 43, p < 0.000001—we can reiterate Poldauf ’s (1984: 77) claim that final stress is “a stress pattern foreign to English.” That is, just like in disyllabic English words, this stress pattern is likely to occur in those words of three and more syllables that are perceived by English speakers as foreign words. Recall, however, that sometimes also penultimate stress can be regarded as a stress pattern foreign to English: Of the 2,479 penultimately-stressed trisyllables in LDOCE , 348 (~14.04 %) end orthographically in a (e.g., cantata ), i (e.g., Ferrari ), or o (e.g., libretto ). By contrast, in the case of the 4,979 antepenultimately-stressed trisyllables in the same dictionary, the same is true of only 172 (~3.45 %) words—e.g., algebra , Medici , piccolo —which counts as a statistically hugely significant difference: χ 2 (1) = 286, p < 0.000001. Note also that of the 348 penultimately-stressed trisyllables such as cantata , Ferrari , or libretto , 133 (~38.22 %) have an entry in Speake & LaFlaur’s (2002) dictionary, counting thereby as foreign terms in English. By contrast, of the 172 antepenultimately-stressed trisyllables such as algebra , Medici , or piccolo , only 44 (~25.58 %) count in the view of Speake & LaFlaur (2002) as foreign terms. This difference is also statistically significant: χ 2 (1) = 8.187, p = 0.004. The simple point here is that words such as cantata , Ferrari , or libretto are in general atypical English words: Of the 109,834 solidly-spelled words in the OD dictionary, only 8,940 (~8.14 %) end orthographically in these vowel letters, with 1,437 (~16.07 %) of these words counting, according to Speake & LaFlaur (2002), as foreign words. Words such as cantata , Ferrari , or libretto are, by contrast, very typical of the Italian language, where, as pointed out in 4.3.3, stress is predominantly penultimate: Of the 24,961 solidly-spelled Italian words in PONS’ Italian-English dictionary, 15,163 (~60.75 %) end orthographically in a / i / o . Thus we conclude that trisyllables such as cantata , Ferrari , or libretto are pronounced in English with penultimate stress simply because these words still count for English speakers as Italian words whereas trisyllables such as armada , burrito , fiesta , potato , quinoa , savanna , tobacco , veranda , etc. also receive penultimate stress because these words still count for English speakers as Spanish words, in 124 4 Stress assignment in English which, just like in Italian, stress is predominantly penultimate (van der Hulst 2010a: 460). Note also that Latina and Latino are in accordance with the Spanish / Italian Stress Rule stressed / laˈtiːnə/ and / ləˈtiːnəʊ/ (OD), non-preserving the stress of / ˈlatɪn/ of Latin ( OD ), which occurs in the BNC 2,729 times and is thus considerably more frequent than both Latina (19) and Latino (36). Similarly, analyzing pronunciations such as / fɪˈnɑːli/ of finale , / fjʊˈrɔːri/ of furore , / vɪˈvɑːtʃi/ of vivace , etc. ( LDOCE ), where the orthographic word-final e is phonetically not empty (which is very typical of the Italian language but extremely untypical of contemporary English; for example, / in.fan.ˈti: .le/ , / ke.ro.ˈzɛ: .ne/ , / si.li.ˈko: .ne/ , / sta.lag.ˈmi: .te/ , / vo.ˈlu: .me/ ( PONS ) are the Italian pronunciations of the words infantile , kerosene , silicone , stalagmite , volume ), it seems that the English pronunciations / fɪˈnɑːli/ , / fjʊˈrɔːri/ , / vɪˈvɑːtʃi/ , etc. are nothing more than imitations of how the words finale , furore , and vivace are pronounced in the source language Italian: / fi.ˈna: .le/ , / fu.ˈro: .re/ , / vi.ˈva: .tʃe/ ( PONS ). Note also that maestro is / ˈmaɪstrəʊ/ in English (LDOCE), preserving the stress of the Italian / maˈɛstro/ ( PONS ) but resolving the hiatus a . e via diphthongization, as a result of which in the English disyllabic word maestro stress occurs initially. A very similar case is al Qaeda , which is interchangeably pronounced / alˈkʌɪdə/ and / ˌalkɑːˈiːdə/ ( OD ). Penultimate stress in al Qaeda is due to the last orthographic vowel a , in which al Qaeda ends. Pronouncing al Qaeda / ˌalkɑːˈiːdə/ is, however, unfortunate from a phonetic point of view as this pronunciation involves a hiatus sequence occurring at this word’s penult-antepenult juncture: / ɑːˈiː/ . To resolve the hiatus, the sequence / ɑːˈiː/ is pronounced as a diphthong, as a result of which the trisyllabic Qa . e . da changes into the disyllabic Qae . da , in which stress occurs initially. The Spanish/ Italian Stress Rule does not, however, apply in English to ica words, which can often be analyzed as suffixed derivatives whose bases are penultimately-stressed ic -words. Of the 21 ica -words in LDOCE whose syllabic length is at least three, 19 (~90.48 %) are stressed antepenultimately. E.g., exotica , which occurs in the BNC 26 times, is stressed / ɪɡˈzɒtɪkə/ ( OD ), preserving the stress of / ɪɡˈzɒtɪk/ of exotic (OD), which occurs in the BNC 1,119 times. Similarly, because Antarctic , which occurs in the BNC 497 times, is stressed / anˈtɑːktɪk/ ( OD ), also Antarctica , which occurs in the corpus only 274 times, is stressed / antˈɑːktɪkə/ ( OD ). (We can suggest that also the ica -words Africa and America have antepenultimate stress because these words are analogized to ica -words such as exotica and Antarctica , which have a penultimately-stressed ic -counterpart.) Another deviation from the Spanish/ Italian Stress Rule is the stress pattern / ˈsɪnəmə/ of cinema ( LDOCE ), which is the promoted secondary stress of the longer word cinematograph : / ˌsɪnɪˈmatəɡrɑːf/ ( OD ). Note, however, that for the 4.3 Three and more syllables 125 less frequently used word kinema , which is “variant of cinema ,” the OED gives both the stress patterns / ˈkɪnɪmə/ and / kaɪˈniːmə/ , with the latter pronunciation abiding by the Spanish/ Italian Stress Rule. 4. 3. 10 Summary Just like in disyllabic English words, the assignment of stress in English words of three and more syllables is also by and large conditioned morphologically (rather than phonologically). Thus, for example, looking at the phonetic transcription / prɒɡˈnəʊsɪs/ of prognosis ( OD ), we cannot say that stress in this word is in accordance with the Latin Stress Rule penultimate because the penultimate syllable / ˈnəʊ/ contains a diphthong and counts therefore as heavy. This explanation does not explain anything because it is not clear why the penultimate syllable in prognosis should necessarily contain a diphthong. Cf. apotheosis , which is pronounced not only / -ˈəʊsɪs/ but also / -ˈθiːəsɪs/ ( OED ), with the phonetic realization of the same orthographic string osis vacillating between a diphthong and a short monophthong. The only reason why prognosis is stressed / prɒɡˈnəʊsɪs/ is the morphological segmentation of this word into the prefix pro and the root gnosis (cf. diagnosis ), with the location of stress in / prɒɡˈnəʊsɪs/ being thus the root-prefix boundary location, and, similarly, the only reason why in addition to being stressed / əˌpɒθiːˈəʊsɪs/ , apotheosis is also stressed / æpəʊˈθiːəsɪs/ is that apart from the osis -words acidosis , alkalosis , brucellosis , cyanosis , diagnosis , endocytosis , hypnosis , kurtosis , leucosis , mitosis , necrosis , prognosis , etc., English also has the eosis -words homeosis , mononucleosis , and rubeosis , i.e., words that share with apotheosis not the string osis but the string eosis . The location of stress in the alternative pronunciation / æpəʊˈθiːəsɪs/ is thus the alternative root-prefix boundary location in the same word apotheosis . Additionally, whereas the segmentation of the trisyllabic prognosis into the monosyllabic prefix pro and the disyllabic root gnosis is in accordance with the segmental length principle, which requires more phonetic segments in the root than in the affix, the segmentation of the pentasyllabic apotheosis into the trisyllabic prefix apothe and the disyllabic root osis violates this principle and might therefore be perceived as counterintuitive. Therefore, of the osis -words prognosis and apotheosis , only the latter but not the former has more than one stress pattern. In general, we can conclude that stress assignment in trisyllables and longer words abides by essentially the same principles that in 4.2 were said to influence the location of stress in disyllabic English words (i.e., stress preservation in suffixation and back-derivation, respecting root-prefix boundaries, reluctance to place stress upon actual suffixes and suffix-like strings, segmental length principle, and stress clash avoidance). 126 4 Stress assignment in English 4.4 Secondary stress Of the 26,334 solidly-spelled polysyllables in LDOCE , only 4,551 (~17.28 %) are, according to the dictionary, pronounced with secondary stress, i.e., in the corresponding transcriptions given for these words in LDOCE, the secondary stress symbol (ˌ) occurs at least once. In 3,280 (~72.07 %) secondary-stressed transcriptions in LDOCE, the secondary stress symbol (ˌ) occurs initially, with the majority being constituted by 1,690 (~51.52 %) penultimately-stressed tetrasyllables (e.g., / ˌevəˈnesənt/ of evanescent ) and antepenultimately-stressed pentasyllables (e.g., / ˌkriːeɪˈtɪvəti/ of creativity ). Similarly, the OD has 94,699 solidly-spelled polysyllables whose transcriptions contain the primary stress symbol (ˈ), but it has only 22,832 (~24.11 %) solidly-spelled polysyllables whose transcriptions contain the secondary stress symbol (ˌ). In 17,519 (~76.73 %) secondary-stressed transcriptions, the secondary stress symbol (ˌ) occurs initially and in 21,994 (~96.33 %) secondary-stressed transcriptions, the secondary stress symbol (ˌ) occurs only one time. Thus, we can say that secondary-stressed English words almost always have only one secondary-stressed syllable, which is as a rule their first syllable. The marginality of secondary stress in English stems from the fact that polysyllabic English words are mainly (~78.65 %) diand trisyllables, which (as a rule) cannot have secondary stress due to reasons of rhythm (see, e.g., Pater 2000: 237; Cruttenden 2008: 242) and / or morphology. Indeed, of the 12,430 disyllables in LDOCE , only 489 (~3.93 %) are, according to the dictionary, secondary-stressed. (By and large, these are either compounds or semantically transparent prefixed derivatives, such as / ˌɔːlˈraɪt/ of alright and / ˌmɪsˈpleɪs/ of misplace .) Because “adjacent stressed syllables make speech sound jerky” (Kingdon 1949: 149), disyllabic English words usually have only one stressed syllable: the ult or the penult. Similarly, of the 8,298 trisyllables in LDOCE, only 1,290 (~15.55 %) have secondary stress. (Once again, these are mainly compounds and semantically transparent prefixed derivatives, such as, e.g., / ˌɑːtʃˈbɪʃəp/ of archbishop and / ˌdiːˈflaʊə/ of deflower ). In a penultimately-stressed trisyllable such as, e.g., inhibit , neither the first nor the last syllable can have secondary stress, for this would give rise to a rhythmically unfortunate sequence of two stressed syllables. An antepenultimately-stressed trisyllable such as, e.g., animal does not, by contrast, receive secondary stress upon its ult (which would not involve a rhythmically unfortunate sequence of two stressed syllables) due to the same reason why primary final stress is in general “a stress pattern foreign to English” (Poldauf 1984: 77). Of the 40,296 trisyllables and longer words in the MRC database, 36,056 (~89.48 %) end orthographically in a string (containing at least one orthographic vowel) that according to the OD dictionary, occurs in English 4.4 Secondary stress 127 as a suffix, i.e., an entry exists in the OD dictionary in which that string is said to function in English as a suffix. The ult of a word such as animal is thus either an actual suffix or like a suffix and therefore repels stress. Therefore, of the 4,979 antepenultimately-stressed trisyllables in LDOCE , only 521 (~10.46 %) are, according to the dictionary, secondary-stressed. (These are mainly compounds such as, e.g., speechwriter : / ˈspiːtʃˌraɪtə/ ). In a word such as animal secondary stress cannot (by contrast) occur either upon its second syllable (due to reasons of rhythm) or upon its third syllable, which is as a rule stress-repellent. Due to the same reason, in the case of the 1,770 antepenultimately-stressed tetrasyllables in LDOCE , a transcription containing the secondary stress symbol (ˌ) is given only for 39 (~2.2 %) words. E.g., / ˌɑːtʃˈenəmi/ of archenemy , / ˌdiːˈsentrəlaɪz/ of decentralize , etc. By contrast, in, e.g., / əˈkædəmi/ of academy ( LDOCE ), secondary stress cannot occur upon either its first / third syllables (due to reasons of rhythm) or upon its last syllable, which is like a suffix and therefore repels stress. Accordingly, given the oft-mentioned fact that in English, “[n]o word can begin with two unstressed syllables” (Fournier 2007: 222), we are left with two main location possibilities for secondary stress in polysyllabic English words: 1) the first syllable of a penultimately-stressed tetrasyllable (e.g., / ˌevəˈnesənt/ ) and 2) the first syllable of an antepenultimately-stressed pentasyllable (e.g., / ˌkriːeɪˈtɪvəti/ ). Another possibility is the first syllable of a finally-stressed trisyllable (e.g., / ˌæfrɪˈkɑːns/ of Afrikaans ), but since final stress is “a stress pattern foreign to English” (Poldauf 1984: 77), secondary-stressed trisyllables such as Afrikaans are quite rare in the language. A different case is / ˌæbrəkəˈdæbrə/ of abracadabra ( LDOCE ), where initial secondary stress is conditioned only morphologically. Because there is the disyllabic (lefthand) string abra -, the trisyllabic (righthand) string cadabra is morphologically segmentable into the prefix cad and the root abra . At the same time, however, because the disyllabic string abra counts morphologically as the root of the trisyllabic string cadabra , it can also count as the root of the disyllabic string abra itself; stress in abra should therefore be initial. (Another reason that contributes to the stress pattern / ˌæbrəkəˈdæbrə/ is that both the righthand string cadabra and the lefthand string abra end orthographically in a and should therefore receive penultimate stress.) An example of a non-initially secondary-stressed word is / səˌtɪfɪˈkeɪʃən/ ( LDOCE ; cf. Cutler 2015: 113, who observes that certification is sometimes stressed / səˈtɪfɪkeɪʃən/ , with the secondary stress being promoted to primary stress). Because certificate has 2,852 attestations in the BNC while certification occurs only 359 times, it is not surprising that the stress pattern of the former, / səˈtɪfɪkət/ ( LDOCE ), is preserved in the latter in the form of either second- 128 4 Stress assignment in English ary or primary stress. Likewise, extravaganza has non-initial secondary stress, / ɪkˌstrævəˈɡænzə/ (LDOCE), because the more frequently used word extravagant is stressed / ɪkˈstrævəɡənt/ ( LDOCE ) (and primary stress in extravaganza is penultimate because the last orthographic vowel in which extravaganza ends is a ). Finally, observe that in the case of 1,107 (~24.32 %) secondary-stressed words in LDOCE , a corresponding transcription in the dictionary contains the stress shift symbol (◀). E.g., aboriginal is / ˌæbəˈrɪdʒɪnəl◀/ , which means that the actual stress pattern of this adjective is on some occasions / ˈæbərɪdʒɪnəl/ . The secondary stress of an English word is thus often its primary stress. 4.5 Concatenations of words Although the focus of this monograph is on word stress (variation), a brief discussion of how combinations of words are accented by English speakers (i.e., cases such as ˈ working theory vs. working ˈ theory , which is phonetically equivalent to ˈ working ˈ theory ) would not be superfluous here. Following Chomsky & Halle (1968: 89-94), many authors have asserted that right-prominent accentuations (e.g., working ˈ theory ) are in accordance with the Nuclear Stress Rule, which applies to syntactic concatenations (i.e., phrases, clauses, and sentences), whereas left-prominent pronunciations such as ˈ working theory abide by the Compound Rule, which accounts for the accentuation of compounds (which are believed to originate in the lexicon). The Nuclear Stress Rule requires that accent be placed on the rightmost primary-stressed syllable, whereas the Compound Rule places accent on the leftmost primary-stressed syllable (Chomsky & Halle 1968: 94). Thus the biconstituent noun phrase elementary proposal , the adjective phrase very difficult , the verb phrase explain theorems , and the clause Jesus wept are in accordance with the Nuclear Stress Rule accented elementary pro ˈ posal , very ˈ difficult , explain ˈ theorems , and Jesus ˈ wept , with accent being placed on the primary-stressed syllables of their righthand components proposal , difficult , theorems , and wept (Halle & Keyser 1971: 23). As for the primary-stressed syllables of their lefthand components elementary , very , explain , and Jesus , they do not become completely unstressed as a consequence of the application of the Nuclear Stress Rule. Rather, they undergo stress weakening in accordance with the Stress Adjustment Rule (Chomsky & Halle 1968: 90), becoming thereby secondary-stressed. In contrast, the biconstituent compounds Christmas party , mealy-mouthed , and (to) trouble-shoot are in accordance with the Compound Rule accented ˈ Christmas party , ˈ mealy-mouthed , and (to) ˈ trouble-shoot , with the primary-stressed syllables of their righthand components party , mouthed , and shoot 4.5 Concatenations of words 129 being demoted to secondary-stressed. Sometimes, however, the secondary stress of a compound’s righthand component disappears over the course of time. Thus, for instance, cupboard and vineyard are pronounced / ˈkʌbəd/ and / ˈvɪnjəd / (OD), with their final syllables being completely unstressed (which follows from the fact that the vowel in them is a schwa). According to Poldauf (1984: 16), this is “a signal of monolithicity,” i.e., the words cupboard and vineyard , which can be regarded as compounds from a diachronic point of view, are synchronically no longer perceived as such. An important question, which, however, is not answered by Chomsky & Halle (1968: 89-94), is why different combinations of especially two nouns should be differently analyzed by English speakers. Thus, for instance, while the street name Downing Street is pronounced / ˈdaʊnɪŋ striːt/ ( LDOCE ), counting thereby as a compound, the street name Portobello Road is pronounced / ˌpɔːtəbeləʊ ˈrəʊd/ (LDOCE), counting thereby as a phrase (an exceptionally-accented compound? ); similarly, the street name Rodeo Drive is accented / rəʊˌdeɪəʊ ˈdraɪv/ ( LDOCE ), meaning that it also counts for an English speaker as a phrase. What are the differences between the left-prominent compound Downing Street and the right-prominent phrases (exceptionally-accented compounds) Portobello Road and Rodeo Drive ? Of little theoretical interest are, by contrast, cases such as ˈ working theory vs. working ˈ theory , in which a change in the location of accent brings about a semantic difference, i.e., as pointed out in Section 2.1, the right-prominent working ˈ theory means “a theory that works,” whereas the left-prominent ˈ working theory is “a theory that can be used as a point of departure for some work.” A similar case is mankind , which, according to the OED , is accented / ˈmankʌɪnd/ when used to express the meaning “[t]he male sex; men or male people in general,” but it is accented / ˌmanˈkʌɪnd/ when the intended meaning is “[t]he human species. As a collective noun: human beings in general.” According to Bell & Plag (2013: 131), “informativity is […] the most important determinant of compound stress, underlying most other known predictors.” Before explaining what the term “informativity” stands for, we reiterate that while left-prominent combinations are pitch-accented only on their lefthand components, right-prominent combinations are pitch-accented on both their leftand righthand components. The only question that must be answered in connection with left-prominent combinations such as Downing Street vs. right-prominent combinations such as Portobello Road / Rodeo Drive is thus under which circumstance(s) their righthand components are pitch-accented, for their lefthand components always receive a pitch accent. Having said this, we can now return to the notion of informativity. “[I]n general in language, uninformative elements tend to be unaccented, while more 130 4 Stress assignment in English informative and unexpected information is accented” (Bell & Plag 2012: 486). Informativity factors include, according to Bell & Plag (2012: Section 2.2; 2013: 135), semantic specificity, the positional size of N2, and the conditional probability of N2. The criterion of semantic specificity refers to the number of senses a lexical item possesses. According to Bell & Plag (2013: 135), “the greater the number of senses a word has […], the less semantically specific it is, and hence the less informative.” For example, the WordNet database gives only two senses for road —“an open way (generally public) for travel or transportation” and “a way or means to achieve something”—but it mentions five senses in the case of street : “a thoroughfare (usually including sidewalks) that is lined with buildings,” “the part of a thoroughfare between the sidewalks; the part of the thoroughfare on which vehicles travel,” “the streets of a city viewed as a depressed environment in which there is poverty and crime and prostitution and dereliction,” “a situation offering opportunities,” and “people living or working on the same street.” The word road is thus obviously more specific and hence more informative than the word street . The positional size of N2 essentially means the likelihood of a particular noun to appear as the righthand component of a biconstituent combination. According to Bell & Plag (2013: 135), “[t]he greater the positional family size of N2, the greater the number of compound types in which it occurs as the righthand constituent, hence the more expected it is following a noun, and the less informative it is in that position.” Thus, for instance, because in the BNC , there are 1,672 types of combinations such as Downing Street (i.e., combinations in which the righthand component Street is preceded by a proper noun), but there are only 245 types of combinations such as Rodeo Drive , with the righthand Drive being preceded by a proper noun, the righthand Drive is undeniably more informative at the end of a street name than the righthand Street . (And in general, it seems that simply because the combinations Downing Street , Portobello Road , and Rodeo Drive are all street names, we have more reasons to expect Street at the end of such a combination than both Road and Drive . The latter are thus more informative than the former.) Finally, the conditional probability of N2 is essentially the positional size of N1. That is, “[t]he greater the positional family size of N1, the greater the number of nouns that might potentially follow it, hence the less expected and more informative is N2 relative to N1” (Bell & Plag 2013: 135). For example, apart from the combination Downing Street , the BNC only has the Proper noun + Noun combinations Downing College , Downing Drive , Downing Lane , and Downing Square , with Downing occurring in the modifier position. The righthand component Street is thus a very expected and hence a fairly uninformative continu- 4.5 Concatenations of words 131 ation of the lefthand component Downing , which also follows from the fact that of the 712 occurrences of Downing in the BNC , 631 (~88.62 %) are occurrences in the combination Downing Street . This combination is therefore pronounced / ˈdaʊnɪŋ striːt/ , with pitch accent falling only upon the lefthand Downing (because, as pointed out above, a lefthand component always receives a pitch accent) but not upon the highly uninformative righthand Street . While informativity is most likely “the most important determinant of compound stress” (Bell & Plag 2013: 131), it is not the only determinant. According to the same authors, other factors that might play a role include semantic properties of / semantic relations between the components of a compound, lexicalization, and length. With regard to semantics, a combination of two nouns is likely to be right-prominent when its components are characterized by the following semantic properties: “N1 refers to a period or point in time” (e.g., morning edition ), “N2 is a geographical term” (e.g., Boston area ), “N1 and N2 form a proper noun” (e.g., Tufts University ), “N1 is a proper noun” (e.g., Dukakis administration ), and “N1 and N2 form a left-headed compound” (e.g., attorney-general ) (Bell & Plag 2013: 133; Plag et al. 2008: 779). Additionally, right-prominence is likely to be triggered by the following semantic relations holding between the components of a Noun + Noun combination: “N1 has N2” (e.g., state officials ), “N2 is made of N1” (e.g., ivory towers ), “N1 is N2” (e.g., exhibition games ), “N2 located at N1” (e.g., Boston offices ), “N2 during N1” (e.g., summer vacations ), and “N2 is named after N1” (e.g., Rockefeller Center ) (Bell & Plag 2013: 133; Plag et al. 2008: 780). The factor of lexicalization refers to the degree of establishment of a compound. Establishment can be defined as the feeling that a particular word unconditionally belongs to the wordstock of a particular language (for a detailed discussion of the notion of establishment, see the article Schmid 2008). As Bell & Plag (2013: 133) point out, (more) established compounds are likely to have a higher frequency of use than less established ones; the former are likely to be spelled either solidly or with a hyphen, whereas in the latter, the component nouns are likely to be separated from each other by means of a blank space; finally, (more) established compounds are more likely to be listed in dictionaries. An example illustrating these points is the combinations health care vs. success theatre , which is defined by Word Spy (McFedries 1996-) as “[p]osting images and stories designed to make others believe you are more successful than you really are.” The combination health care has 32,533 attestations in the COCA corpus, while the combination success theater does not occur in the corpus at least one time. In addition to being spelled health care , this combination is also often spelled healthcare (6,202 times) and health-care (4,769 times), while neither success-theater nor successtheater occurs in the COCA corpus. (A Google search 132 4 Stress assignment in English for successtheater does, however, return 178 results.) For the combination health care , but not for success theatre , there is an entry in the MWO as well as the OD dictionary, whereas Word Spy (McFedries 1996-), which contains an entry for success theater , is in fact an online-based collection of neologisms, which have only recently appeared in the English language. To conclude, the combination health care is with regard to each of the three aforementioned criteria much more established in the English language than the neologism success theater . Of these two combinations, it is therefore the considerably less established success theater that is likely to be accented success ˈ theater (Bell & Plag 2013: 133). Finally, with regard to length, Bell & Plag (2013: 141) point out that “[t]he longer the compound, the higher the probability of it receiving an accent on N2.” For example, a combination such as deputy superintendent is more likely to be right-prominent than success theater simply because the former is longer than the latter. As for triconstituent combinations, Kösling (2013) and Kösling et al. (2013) have recently demonstrated that their accent contours largely depend on the accent contours of the biconstituent combinations that are embedded into them. (This could perhaps be analogized to stress preservation among suffixed derivatives.) Compare, for example, the accentuation of the combinations [hay fever] treatment , [science fiction] book , business [credit card] , and family [Christmas dinner] (from Kösling et al. 2013: 549). The combination hay fever treatment is accented ˈ hay fever treatment because the embedded biconstituent hay fever is left-prominent: ˈ hay fever . The combination science fiction book is accented ˈ science ˈ fiction book because the embedded biconstituent combination science fiction is right-prominent, i.e., in it, both the lefthand component science and the righthand component fiction are accented: ˈ science ˈ fiction . The combination business credit card is accented ˈ business ˈ credit card because the embedded biconstituent combination credit card is left-prominent (i.e., ˈ credit card ) and the first noun business is accented simply because in both leftand right-prominent combinations the first noun is always accented. Finally, the combination family Christmas dinner is accented ˈ family ˈ Christmas ˈ dinner (i.e., there are three pitch-accents in this combination) because the embedded biconstituent combination Christmas dinner is right-prominent (i.e., ˈ Christmas ˈ dinner ) and the noun family again represents the first constituent noun of a combination, which is always accented, regardless of whether the combination is leftor right-prominent. 4.6 Across-varietal differences? 133 4.6 Across-varietal differences? According to Berg (1999: 123), of the 75,000 entries constituting the 1990’s edition of the Longman Pronouncing Dictionary (Wells 1990), only 932 (~1.24 %) have different stress patterns in British and American English. (An oft-mentioned systematic difference between British and American English is that the latter “tends to preserve secondary stress, and thus has more fully realized vowels than British English in words like secretary , laboratory ” (Kretzschmar 2010 108). With regard to this characteristic, American English can be said to be more archaic than British English because, as pointed out in 4.3.7, apart from preserving the segmental structure of the Latin etymon word territōrium , there is no justification for pronouncing territory / ˈterətɔːri/ rather than / ˈterətəri/ : The string ory , which occurs in English as a suffix, should not receive either primary or secondary stress. The British pronunciation / ˈterətəri/ , which has a reduced vowel in the unstressed penult, is thus undeniably more correct than the American version / ˈterətɔːri/ .) Note also that the typological database StressTyp2 (http: / / st2.ullet.net/ , 23. 11. 2016), which contains “information on stress and accent patterns in over 750 of the world’s languages with nearly every language family represented,” provides stress-related information with reference to lects rather than languages. Thus, for instance, a search in the database for “Arabic” returns results for 18 lects: Gulf Arabic, Lebanese Arabic, Egyptian Radio Arabic, Beirut Arabic, Palestinian Arabic, Negev Bedouin Arabic, etc., which have non-identical stress systems. By contrast, a search in the database for “English” returns only two lects (which, according to the compilers of the database, have different stress systems): English and Old English. Finally, observe that according to Foulkes & Docherty (2007: 71), “[a]lthough word stress in English is not wholly fixed […], dialects seem to differ relatively little in the location of main stress placement. This is true both for word and sentence stress. Thus, words like acquisition , deliver , telephone will almost invariably take stress on the indicated syllable, no matter what the dialect or style of speech.” Given these three facts, we can repeat, e.g., Turcsan & Herment’s (2015: 38) claim that “[t]he general mechanisms behind stress assignment are largely the same in all inner-circle varieties of English, with some lexical exceptions,” i.e., British, American, Australian, Canadian, and New Zealand English speakers abide by the same principles when deciding where to place stress in a polysyllabic English word. 134 5 Case studies 5 Case studies Having discussed the factors influencing the location of stress in a polysyllabic English word, we can now proceed to a more systematic analysis of the phenomenon of stress assignment instability, i.e., situations when one and the same English word has more than one stress pattern. 5.1 Stress variation in the OED The first section of the chapter approaches the phenomenon under consideration using the OED (phonetic transcriptions) as the primary source of information of how words in English are stressed. 5.1.1 presents the overall results (i.e., the total number of OED entries with stress doublets, the distribution of variably-stressed words across different morphosyntactic categories as well as across different varieties of English, types of stress variation), while 5.1.2-5.1.6 elaborate upon the causes of stress variation in English words exhibiting a particular syllabic and / or morphological structure (i.e., disyllables, words of three and more syllables, suffixed derivatives, etc.). 5.1.1 Overall results Of the 273,404 entries that between October 20, 2013 and December 20, 2013 were available in the online version of the Oxford English Dictionary / OED , 7,628 were found to contain at least two phonetic transcriptions differing from each other with regard to the location of the primary stress symbol (ˈ). Notice, however, that in the OED , multiple entries might refer to identical forms, when, e.g., they function as instances of different morphosyntactic categories. For example, there is the adjectival traverse -entry, in which traverse is said to vacillate between the pronunciations / ˈtrævəs/ and / trəˈvɜːs/ , but there are also the nominal and verbal traverse -entries, in which traverse is also said to vacillate between the initially-stressed / ˈtrævəs/ and the finally-stressed / trəˈvɜːs/ . Because of this fact, the 7,628 OED entries, in which multiple stress patterns are given for one and the same English word, actually correspond to only 7,149 orthographically non-identical strings, which, according to the OED , can be stressed in English in at least two different ways. 5.1 Stress variation in the OED 135 The finding that of the 273,404 OED entries, only 7,628 (~2.79 %) have more than one stress pattern creates the illusion of stress variation being a fairly marginal aspect of contemporary English phonology. Do observe, however, that of the 26,334 solidly-spelled polysyllables in LDOCE , 1,671 (~6.35 %) have more than one stress pattern. The reason why LDOCE has more variably-stressed items than the OED is that the latter does not regard words such as, e.g., academic as variably-stressed. That is, academic is in the OED only / ˌakəˈdɛmɪk/ in British English and only / ˌækəˈdɛmɪk/ in American English, but, as pointed out in 3.2, in especially the modifier position (e.g., academic discipline ), academic is usually stressed / ˈækədemɪk/ . Note also that among the variably-stressed items in LDOCE are the ion -words coronation , imitation , and reproduction , for which the dictionary gives the transcriptions / ˌkɒrəˈneɪʃən◀/ , / ˌɪmɪˈteɪʃən◀/ , and / ˌriːprəˈdʌkʃən◀/ . The actual stress patterns of these ion -words are thus at least on some occasions / ˈkɒrəneɪʃən/ , / ˈɪmɪteɪʃən/ , and / ˈriːprədʌkʃən/ . Given this fact, a good guess would be that other English ion -tetrasyllables are also at least on some occasions pronounced with initial rather than with penultimate stress. E.g., constitution is / ˌkɒnstɪˈtjuːʃən/ in LDOCE , but in, e.g., the video xh_ HA nzXn3Y (07. 08. 2017), we can hear U. S. President Donald Trump stressing constitution / ˈkɒnstɪtjuːʃən/ . (For constitutional , LDOCE does, however, give the transcription / ˌkɒnstəˈtjuːʃənəl◀/ , which contains the stress shift symbol. The actual stress pattern of constitutional is thus on some occasions / ˈkɒnstətjuːʃənəl/ , with stronger stress occurring initially rather than antepenultimately.) Similarly, in LDOCE there are two variably-stressed ity -words: Intercity is / ˌɪntəˈsɪti◀/ and university is / ˌjuːnəˈvɜːsəti◀/ , which means that the actual stress patterns of these words are at least on some occasions / ˈɪntəsɪti/ and / ˈjuːnəvɜːsəti/ . Recall also the above mentioned fact that in the OD dictionary, agreeability is / əˈɡriːəˈbɪlᵻti/ , allowability is / əˈlaʊᵻˈbɪlᵻti/ , materiality is / məˈtɪərɪˈalɪti/ , etc., with more than one syllable in these transcriptions being marked as syllables bearing primary stress. Transcriptions such as these strongly suggest that antepenultimate stress is not the only stress pattern of many other English ity -words, especially pentasyllables. E.g., for productivity , LDOCE gives the transcription / ˌprɒdʌkˈtɪvəti/ , which lacks the stress shift symbol. The author’s auditory impression is, however, that the actually occurring stress pattern of this pentasyllable is (very often) / ˈprɒdʌktɪvəti/ . Thus, we can conclude that many polysyllabic English words for which only one stress pattern is given in dictionaries are as a matter of fact pronounced with more than one stress pattern. The 7,628 OED entries with stress doublets are represented by instances of the following morphosyntactic categories: 136 5 Case studies Category Number Per cent noun 5,762 (5,233) ~59.05 (58.89) adjective 3,065 (2,797) ~31.41 (31.48) verb 406 (380) ~4.16 (4.28) adverb 344 (326) ~3.53 (3.67) combining form 91 (67) ~0.93 (0.75) interjection 58 (54) ~0.59 (0.61) preposition 16 ~0.16 (0.18) phrase 9 (8) ~0.09 (0.09) prefix 3 (2) ~0.03 (0.02) suffix 2 (1) ~0.02 (0.01) conjunction 2 ~0.02 (0.02) Tab. 1: Morphosyntactic categories as identified in the OED That Table 1 contains two numbers for some categories is connected to the fact that one and the same OED entry sometimes counts simultaneously as an instance of within-varietal stress instability in more than one Present-day English variety (e.g., for adult the OED gives the British English transcriptions / ˈadʌlt/ and / əˈdʌlt/ and the American English transcriptions / əˈdəlt/ and / ˈæˌdəlt/ ). The fact that, for example, both the numbers 5,762 and 5,233 are given in Table 1 for the category “noun” thus means that there are 5,762 instances of within-varietal stress instability of nouns in English, but there are only 5,233 nouns with stress doublets in the OED , 529 of which exhibit a stress variation in more than one Present-day English variety (usually British and American English). Notice also that the sum of the instances of the morphosyntactic categories given in Table 1-9,758 or 8,886—is not equal to the total number of the OED entries with multiple stress patterns, given above: 7,628. This is because one and the same item to which an OED entry refers sometimes functions as an instance of more than one morphosyntactic category. For example, adult belongs in Table 1 to both nouns and adjectives. Across-varietal differences with regard to within-varietal stress variation are summarized in Table 2. 5.1 Stress variation in the OED 137 Variety Number of items Per cent American English 5,386 ~62.66 British English 1,927 ~22.42 without indication 1,245 ~14.49 New Zealand English 11 ~0.13 South African English 8 ~0.09 Scottish English 6 ~0.07 Australian English 5 ~0.06 Canadian English 4 ~0.05 Irish English 2 ~0.02 Caribbean English 1 ~0.01 Tab. 2: Stress variation across different varieties of Present-day English As for “without indication,” note that, for example, the OED entry dealing with the adjective compact only states that this adjective can be pronounced either / kəmˈpækt/ or / ˈkɒmpækt/ , but it is not indicated whether it is American or / and / vs. British English speakers who stress compact interchangeably. (In LDOCE , the adjective compact is / kəmˈpækt/ vs. / ˈkɒmpækt/ in British English and only / kəmˈpækt/ in American English, but as the subsequent analysis of the YouTube data has revealed, also American English speakers stress compact initially when it functions morphosyntactically as an adjective). Similarly, in the case of archival , it is not specified in the OED in which variety or varieties of English this adjective is stressed not only / ɑːˈkaɪvəl/ but also / ˈɑːkɪvəl/ . As the numbers in Table 2 might suggest, words with stress doublets are more typical of American English than they are of British English let alone of any other variety mentioned in Table 2. However, as the following explanation of the types of stress variation will demonstrate, this is not the case. Type Occurrences Per cent degrees of stress 5,012 ~57.42 location of stress 3,186 ~36.50 stresslessness 306 ~3.51 degree vs. location 161 ~1.84 identical syllables 63 ~0.72 Tab. 3: Types of stress variation in the OED 138 5 Case studies A degree-of-stress variation is exhibited by, e.g., aforementioned , which is / əˈfɔːˌmɛnʃnd/ vs. / əˌfɔːˈmɛnʃnd/ in British English and / əˈfɔrˌmɛn(t)ʃənd/ vs. / əˌfɔrˈmɛn(t)ʃənd/ in American English (OED). Stronger stress in aforementioned is thus the stress of either the lefthand component afore or of the righthand component mentioned . A location-of-stress variation is exhibited by the aforementioned / ɑːˈkaɪvəl/ vs. / ˈɑːkɪvəl/ of archival , which (from especially a semantic point of view) is not segmentable into ar and chival the way aforementioned is segmentable into afore and mentioned . Stresslessness refers to words such as bizarre , which, according to the OED , is pronounced by English speakers either “/ bɪˈzɑː(r)/ , or as French / bizar/ .” Because, as explained in 2.4, French does not have word stress, words borrowed into English from French are sometimes pronounced by English speakers in a stressless manner, imitating how they are pronounced in the source language French. Degreevs. location-of-stress variation refers to cases such as, e.g., / ˈkɪləmiːtə(r)/ vs. / kɪˈlɒmɪtə(r)/ of kilometer ( OED ). Although the segmentation kilo - + meter is completely transparent both formally and semantically, kilometer is in contemporary English more frequently stressed / kɪˈlɒmɪtə(r)/ , which is not in accordance with this morphological structure. Cf. centimeter , which, according to LDOCE , has only the “correct” stress pattern / ˈsentəˌmiːtə/ , with the lefthand component centi -, which distinguishes centimeter from kilometer , being made more prominent than the righthand component meter . Finally, there are also cases such as / ɒˈfjuːkəs/ vs. / ˌɒfɪˈuːkəs/ (OED; only British English), which means that stress in Ophiuchus is always penultimate, but because the hiatus i . u is either resolved or unresolved, the penultimate syllable, which bears stress, has a different segmental makeup: / ˈfjuː/ vs. / ˈuː/ . Notice now that of the 5,012 instances of degree-of-stress variation in the OED, 4,447 (~88.73 %) occur in American English and only 417 (~8.32 %) in British English. (This is the reason why in Table 2, American English is said to have the largest number of words with stress doublets.) If, however, location-of-stress variation is taken into consideration, variably-stressed items are, by contrast, more numerous in British English: 1,451 vs. 880. Notice also that the fact that only 37 entries are marked in the OED as having variable stress in varieties other than British and American English does not mean that these varieties exhibit stress variability to a considerably lesser extent than British and American English. This fact is primarily a reflection of the bias toward Standard American and British English accents as two major Present-day English reference accents. Thus, with the exception of the entries belonging to the category “without indication,” in the case of the other OED en- 5.1 Stress variation in the OED 139 tries it is always indicated how they are pronounced both in British and American English. But only a tiny fraction of these entries also give, e.g., Australian or New Zealand English transcriptions, i.e., it is specifically indicated in these entries how a word is pronounced in varieties other than British and American English. This is mainly true of words denoting culture-specific concepts playing a particularly prominent role in the life of neither American nor British English speakers. E.g., mulgara , “[a] rat-sized carnivorous marsupial with a pointed snout, […] native to arid regions of central Australia” ( OED ), which is stressed either / mʌlˈɡarə/ or / ˈmʌlɡərə/ in Australian English, or Oamaru , “[d]esignating a division of Tertiary rock formations found in the region of Oamaru” ( OED ), which in New Zealand English is pronounced either with antepenultimate stress as / ˈɒməru/ or with final stress as / ɒməˈru/ . 5.1.2 Degree-of-stress variation We begin with morphologically complex words / combinations of words whose components occur with varying degrees of stress. For biconstituent items, the four theoretically possible variability scenarios are 1) left-prominence vs. right-prominence (“l/ r”), 2) left-prominence vs. double-prominence (“l/ d”), 3) right-prominence vs. double-prominence (“r/ d”), and 4) left-prominence vs. right-prominence vs. double-prominence (“l/ r/ d”). Each of these variability scenarios could be attested in the OED , with, however, the scenario “r/ d” being found considerably more often than “l/ r,” “l/ r/ d,” and “l/ d.” Type Occurrences Per cent r/ d 3,949 ~78.79 l/ r 581 ~11.59 l/ r/ d 460 ~9.18 l/ d 22 ~0.44 Tab. 4: Degree-of-stress variation in the OED The type “r/ d” is the most frequent variation type, which, according to the OED , occurs particularly often in American English (3,915). An example of r/ d-variation is the already discussed adjective meningococcic , which is pronounced / məˈˌnɪŋɡoʊˈkɑk(s)ɪk/ in American English. That is, the word vacillates between the right-prominent pronunciation / məˌnɪŋɡoʊˈkɑk(s)ɪk/ and the double-prominent alternative / məˈnɪŋɡoʊˈkɑk(s)ɪk/ . Similar examples include agronomic , adjutant general , air-to-air , machine-washable , macro-econ- 140 5 Case studies omist , etc., for which the OED gives the American English transcriptions / ˈˌæɡrəˈnɑmɪk/ , / ˈˌædʒədənt ˈdʒɛn(ə)rəl/ , / ˈˌɛrtəˈɛ(ə)r/ , / məˈˌʃinˈwɔʃəbəl/ , and / ˈˌmækroʊiˈkɑnəməst/ . What can be seen in these transcriptions is that, again, these items’ lefthand components agro -, adjutant , air , machine , and macro are said to be (ˈˌ)-stressed, which means that the level of stress borne by one of these component’s syllables vacillates between primary and secondary. The type “r/ d” is the most controversial variation category because, as explained in 2.2, primary-stressed syllables are phonetically not different from secondary-stressed ones (Plag et al. 2011). Thus meningococcic cannot vacillate between the stress contour / məˌnɪŋɡoʊˈkɑk(s)ɪk/ , where / ˌnɪŋ/ bears secondary stress, and / məˈnɪŋɡoʊˈkɑk(s)ɪk/ , where the same syllable bears primary stress, simply because / məˌnɪŋɡoʊˈkɑk(s)ɪk/ is phonetically equivalent to / məˈnɪŋɡoʊˈkɑk(s)ɪk/ . Similarly, as Farnetani et al. (1988) and Kunter (2011) have demonstrated, concatenations of words such as adjutant general are perceptually either leftor right-prominent, with the latter, however, being double-prominent from a speaker’s point of view. That is, when the listener perceives adjutant general to be pronounced / ˌædʒədənt ˈdʒɛn(ə)rəl/ , it is actually pronounced / ˈædʒədənt ˈdʒɛn(ə)rəl/ by the speaker, with both (the stressed syllables of) the components adjutant and general receiving pitch accents. An example of the category “l/ r,” which stands for left-prominence vs. right-prominence, is alongside . According to the OED, it is pronounced in British English either / əˈlɒŋsʌɪd/ or / əˌlɒŋˈsʌɪd/ ( OED ). In contrast to the phonetically non-existent variation type “r/ d,” the category “l/ r” is phonetically real. As was mentioned on numerous occasions, left-prominent words have only one pitch accent, which is placed upon these words’ lefthand components, while right-prominent words are double-accented (but perceived to have been accented only on their righthand components). That is, alongside can be a ˈ longside vs. a ˈ long ˈ side , which is perceptually along ˈ side . Just like the category “r/ d,” the combined type “l/ r/ d” is problematic because right-prominent and double-prominent pronunciations are phonetically identical. An example is alike-minded , which is said to be pronounced / əˈˌlaɪkˈˌmaɪndᵻd/ ( OED ) in American English, i.e., its pronunciation is supposed to vacillate between the left-prominent / əˈlaɪkˌmaɪndᵻd/ , the right-prominent / əˌlaɪkˈmaɪndᵻd/ , and the double-prominent / əˈlaɪkˈmaɪndᵻd/ . But the right-prominent pronunciation / əˌlaɪkˈmaɪndᵻd/ is phonetically equivalent to the double-prominent / əˈlaɪkˈmaɪndᵻd/ . Hence the only variation alike-minded can exhibit is that between the left-prominent a ˈ like-minded and the perceptually rightbut phonetically double-prominent a ˈ like- ˈ minded . Instances of the variation category “l/ r/ d” can thus be regarded as instances of the variation category “l/ r.” 5.1 Stress variation in the OED 141 Finally, the category “l/ d” can be illustrated with mother’s milk , which is said to be pronounced / ˈmʌðəz ˈˌmɪlk/ in British English ( OED ), i.e., the item vacillates between the left-prominent pronunciation / ˈmʌðəz ˌmɪlk/ and the double-prominent alternative / ˈmʌðəz ˈmɪlk/ . Again, from a perceptual point of view, there is no such thing as double-prominence as far as accentuation realized by means of pitch accents is concerned: When combinations of two nouns, such as, for example, mother’s milk , receive two pitch accents, they are due to the non-occurrence of the expected declination of pitch (at the end of such a combination) perceived to be right-prominent. Instances of the category “l/ d” can thus also be regarded as instances of the category “l/ r,” i.e., mother’s milk can be accented either ˈ mother’s milk or mother’s ˈ milk , which is phonetically equivalent to ˈ mother’s ˈ milk . As for the causes of degree-of-stress variability, recall that the main predictor of compound stress in English is the degree of informativity of its righthand component (Bell & Plag 2012, 2013), i.e., the (stressed syllable of the) lefthand component in a concatenation such as adjutant general always bears a pitch accent, but the (stressed syllable of the) righthand component is pitch-accented only when it is perceived as sufficiently informative by the speaker. This is as a rule the case when a particular righthand component is not easily predictable as the righthand component in a concatenation beginning with a particular lefthand component: “The greater the positional family size of N2, the greater the number of compound types in which it occurs as the right-hand constituent, hence the more expected it is following a noun, and the less informative it is in that position” (Bell & Plag 2013: 135). Consider, for example, the combination apple pie , which is stressed / ˌapl ˈpʌɪ/ in British English and / ˈæpəl ˌpaɪ/ in American English ( OED ). In the COCA corpus, there are 487 Noun + Noun combinations such as pumpkin pie , pecan pie , blueberry pie , etc., whereas in the BNC the number of similar combinations with pie as the righthand component is only 99. Note that this difference is in part due to the fact that the COCA is, as of 06. 02. 2015, 5.2 times bigger in terms of the number of words than the BNC , but even if we use, e.g., Time Magazine Corpus (Davies 2007-), whose size is comparable to that of the BNC (100 million words), it still remains the case that American English has more Noun + Noun combinations with pie as the righthand component than British English: 132 vs. 99. The positional size of pie as N2 is thus greater in American than in British English and it is therefore American English speakers for whom the righthand component pie does not count as a sufficiently informative continuation of the lefthand component apple . Consider also the combination ice cream , for which the OED gives the British English transcription / ˈˌʌɪs ˈˌkriːm/ and the American English transcription 142 5 Case studies / ˈˌaɪs ˈˌkrim/ . The combination ice cream is thus interchangeably accented ˈ ice cream and ˈ ice ˈ cream (which perceptually corresponds to ice ˈ cream ) irrespective of whether it is used by British or American English speakers. An example of within-speaker variation involving this combination is the YouTube video _ BI 4 LB a NE nQ (23. 06. 2015), where the accentuations ˈ ice cream and ice ˈ cream are interchangeably used by one and the same American English speaker. Commenting upon cases similar to these, Bell & Plag (2012: 517) observe that “[t]he family sizes of constituents will vary across the mental lexicons of different speakers, and unconscious perceptions of informativeness might also therefore differ.” Drawing upon this, we can then suggest that different perceptions of how informative cream is as the righthand component in a combination such as ice cream might arise due to the polysemic nature of this righthand component. The point here is that in the COCA corpus, cream occurs as the righthand component of 518 Noun + Noun combinations mainly in the sense “medical / cosmetic cream” (e.g., hand cream , skin cream , day cream , etc.) and in the sense “a sweat eatable thing that is creamy” (e.g., coconut cream , chocolate cream , vanilla cream , etc.), but the sense “milk’s cream” is only represented by the single combination ice cream . Thus if we do not distinguish between the different senses of cream , this righthand component should count as fairly uninformative; ice cream should therefore be accented ˈ ice cream , with only the lefthand component ice receiving a pitch accent. If, however, we consider only the sense “milk’s cream,” which is expressed by cream in the single combination ice cream , the righthand cream should, by contrast, count as hugely informative, leading thereby to the right-prominent accentuation ice ˈ cream . (What also contributes to this accentuation is the fact that COCA has 1,299 combinations such as ice cream , ice age , ice water , ice cubes , ice hockey , ice storm , etc., i.e., Noun + Noun combinations with ice as their lefthand component. Righthand nouns following ice in such combinations are therefore potentially very informative.) It is very much possible that English speakers on some occasions “switch off” the irrelevant senses of cream when determining how to accent the combination ice cream (in this case ice cream is accented ice ˈ cream ), while on other occasions these irrelevant senses do flow to the overall count of the positional size of cream in ice cream , resulting thereby in the left-prominent pronunciation ˈ ice cream . The very same informativity principle also seems to be in part responsible for the fact that especially the cardinal numerals ending in teen are in different environments stressed differently (i.e., it is not denied, however, that from the point of view of rhythm, the stress pattern ˈ thirteen men is better than thir ˈ teen men , while in, e.g., Chapter Fourteen the stress pattern four ˈ teen is better than 5.1 Stress variation in the OED 143 ˈ fourteen ). To begin with, observe that the righthand component teen of, e.g., fourteen exists in the English language as an adjective meaning “[r]elating to teenagers,” i.e., e.g., a teen idol . This meaning is related to the meaning that teen expresses in fourteen . The numeral fourteen is thus a compound and must therefore be accented in accordance with the same principles as combinations such as apple pie and ice cream . The compound analysis is strongly supported by dictionary transcriptions in which the lefthand component of a numeral such as fourteen is often said to be secondary-stressed. E.g., thirteen is / ˌθɜːˈtiːn/ in the OED . Recall in this connection that normally, “[w]hen there is only one syllable before the primary [stress], this is usually un[stressed]” (Cruttenden 2008: 242), i.e., in a finally-stressed disyllabic English word (e.g., apply , concern , deceive ), the penult does not bear secondary stress (for this would create a rhythmically unfortunate sequence of two stressed syllables). Words such as thirteen and fourteen are, however, not “normal” disyllabic words. These are compounds and must therefore be either left-prominent as ˈ thirteen and ˈ fourteen , with only the lefthand components thir and four receiving pitch accents, or double-prominent as ˈ thir ˈ teen and ˈ four ˈ teen , with pitch accents being placed upon both the lefthand components thir and four and the righthand component teen . (Cf. the numerals thirty , forty , fifty , sixty , seventy , eighty , and ninety , which are stressed only / ˈθəːti/ , / ˈfɔːti/ , / ˈfɪfti/ , / ˈsɪksti/ , / ˈsɛv(ə)nti/ , / ˈeɪti/ , and / ˈnʌɪnti/ (OD). Because the string ty , in which these numerals end, does not occur in English as a separate word, the only possible analysis of this string is the suffix analysis. Just like other English suffixes, the suffix ty of these numerals repels stress.) In summary, in numerals ending in teen and teenth , the righthand components teen and teenth must be accented only when they are perceived as sufficiently informative (i.e., not particularly expected) in the context of a combination beginning with a lefthand component such as, e.g., four . Thus, for example, if an English speaker counts consecutively from, say, 10 to 20, the component teen is naturally going to be fairly expected once the speaker has reached 12. The component teen is thus in the counting-context extremely uninformative and is therefore normally unaccented. That is why, we usually hear the left-prominent accentuations ˈ thirteen , ˈ fourteen , ˈ fifteen , ˈ sixteen , ˈ seventeen , ˈ eighteen , ˈ nineteen rather than the right-prominent thir ˈ teen , four ˈ teen , fifˈteen , six ˈ teen , seven ˈ teen , eigh ˈ teen , nine ˈ teen ; an audio illustration of this is the YouTube video gOc FL cL7dik (07. 07. 2015). Consider now compound dates such as 1821 . Apart from teen , eight can, at least in theory, be also followed in the context of a date such as 1821 by either ty ( 8021 , i.e., eighty twenty-one ) or hundred ( 821 , i.e., eight hundred twenty-one ). However, since the year 8021 represents very distant future and the year 821 lies in the very distant past, the probability of someone talking about the years 144 5 Case studies 8021 / 821 instead of 1821 is extremely low. Hence in the context of a compound date such as 1821 the component teen is also fairly expected and thus extremely uninformative. Hence eighteen is exclusively left-prominent in such combinations. By contrast, in the context of a compound date such as 1918 , the component teen of eighteen is fairly informative because instead of talking about the year of 1918 we can also refer to events that took place in 1980 or even 1908. Notice, however, that since different speakers have differently-sized mental lexica, non-identical speakers can be expected to have non-identical perceptions of how informative the component teen is in eighteen in 1918 . For a professional historian, this component should not count as particularly informative because the year of 1918 is the year when World War I came to an end, whereas no event of a comparable historic scope took place in 1980 or 1908. Thus, historians are likely to use the left-prominent pronunciation ˈ eighteen in 1918 . (And indeed, that in the YouTube videos listened to by the author, the teen -cardinals at the end of the dates 1913 , 1914 , 1915 , 1916 , 1917 , 1918 , and 1919 were found to be more frequently leftthan right-prominent is most likely due to the fact that the speakers in these videos are either professional historians or at least individuals with keen interest in important events that took place during this time span.) By contrast, for an individual who is not particularly interested in history but who, for instance, got married in the year of 1980 (or any other important event happened in the life of that speaker in 1980), the component ty in eighty is going to be less informative than teen . Accordingly, if on some rare occasion that speaker will be required to say 1918 , he or she will most likely use the right-prominent pronunciation ˈ nineteen eigh ˈ teen . Notice also that in YouTube videos containing the spoken occurrences of 2013 , thirteen was, as expected by the author, more frequently pronounced ˈ thirteen than thir ˈ teen . The latter pronunciation was mainly attested in contexts involving especially comparisons of products released in 2013 as opposed to earlier releases / releases of the following years. For example, since as of today, we do not know yet whether Microsoft will release a new version of the software Excel in 2030, the component teen of thirteen of Excel 2013 is extremely uninformative and therefore does not as a rule receive accent, i.e., thirteen in this combination is usually pronounced ˈ thirteen ; the right-prominent pronunciation thir ˈ teen occurs, however, when an English speaker specifically refers to unique features of Excel 2013, distinguishing it from, e.g., Excel 2010. 5.1 Stress variation in the OED 145 5.1.3 Location-of-stress variation The six syllabic positions in which stress was found to occur in variably-stressed English words include the ult (u), the penult (p), the antepenult (a), the pre-antepenult (4), the pre-pre-antepenult (5), and the pre-pre-pre-antepenult (6). The overall results are as follows: Type Occurrences Per cent p/ a 1,468 ~46.08 u/ p 841 ~26.4 a/ 4 312 ~9.79 u/ a 294 ~9.23 p/ 4 111 ~3.48 a/ 5 60 ~1.88 4/ 5 45 ~1.41 u/ 4 17 ~0.53 u/ p/ a 14 ~0.44 4/ 6 7 ~0.22 p/ a/ 4 6 ~0.19 p/ 5 5 ~0.16 a/ 6 3 ~0.09 5/ 6 1 ~0.03 a/ 4/ 5 1 ~0.03 p/ 6 1 ~0.03 Tab. 5: Location-of-stress variation in the OED 5.1.3.1 Penultimate vs. antepenultimate stress The type “p/ a” stands for variability involving the placement of stress on the penult vs. stressing the antepenult. This variation category has the largest number of instances, but do observe that of the 1,096 orthographically non-identical p/ a-strings (i.e., this number is not identical to the corresponding number in Table 5, because, as explained in 5.1.1, different entries in the OED sometimes refer to orthographically identical items, when, e.g., they function as instances of different morphosyntactic categories), 416 items (~37.96 %) replicate the stress 146 5 Case studies variation of the type “u/ p” (i.e., placing stress on the ult vs. stressing the penultimate syllable) inherent in their disyllabic bases or they replicate the stress variation of the type “p/ a” inherent in their bases, whose syllabic length is no less than three. An example of the former is the aforementioned adulthood , which is ˈ adulthood vs. a ˈ dulthood only because adult is ˈ adult vs. a ˈ dult . An example of the latter is the longer word microglial , vacillating between the stress patterns / -ˈkrɒɡlɪəl/ and / -ˈɡlʌɪəl/ ( OED ) because also the shorter word microglia vacillates between the stress patterns / -ˈkrɒɡlɪə/ and / -ˈɡlʌɪə/ ( OED ; both in British and American English). If items such as adulthood are left out of consideration, because they are not genuine instances of variability between penultimate and antepenultimate stress, and if cases such as microglia - microglial , which exemplify identical types of p/ a-variability, are counted only once, we are left with only 679 orthographically non-identical p/ a-instances, with the second largest variation category “u/ p” comprising 653 items. As suggested in the previous chapter, a variably-stressed English word is one in the case of which different stress-assignment factors favor different stress-assignment outcomes. The usual variability scenario involves more than one morphological analysis making sense in the case of one and the same English word. For example, one natural way of looking at the verbs contribute , distribute , and retribute is to regard them as prefixed derivatives sharing the free root tribute , which occurs in English as a separate word. This morphological analysis results in the stress pattern / -ˈtrɪbjuːt/ , which is the preserved stress of the disyllabic word tribute . At the same time, however, these ute -verbs all have fewer occurrences in the BNC than the corresponding longer ion -noun: 2,643 vs. 5,307 ( contribute vs. contribution ), 602 vs. 10,323 ( distribute vs. distribution ), no occurrences vs. 297 ( retribute vs. retribution ). The back-derivation analysis of these verbs should result in final stress in them (and, as observed in Chapter 4, Irish English speakers do indeed stress distribute / dɪstrɪˈbjuːt/ ), but because the string ute is like a suffix, these three verbs are (outside Irish English) stressed not finally but antepenultimately, promoting thereby the secondary (initial) stress of the nouns contribution , distribution , and retribution to primary stress. That is, e.g., / ˌkɒntrəˈbjuːʃən/ ( LDOCE ) → / ˈkɒntrɪbjuːt/ ( OD ). (Of the stress patterns / kənˈtrɪbjuːt/ and / ˈkɒntrɪbjuːt/ , it is, however, the penultimately-stressed version that according to the OD dictionary, “is held to be the standard, correct pronunciation” of contribute .) A slightly different case is substantive , which, according to the OED , is / ˈsʌbst(ə)ntɪv/ vs. / səbˈstantɪv/ in British English and only / ˈsəbstən(t)ɪv/ in American English. Antepenultimate stress in the trisyllabic substantive is the preserved initial stress of the disyllabic substance (in which the string ance is stress-repellent; cf. the aforementioned vengeance ), whereas penultimate stress 5.1 Stress variation in the OED 147 is due to the stress pattern / səbˈstænʃəl/ of substantial ( LDOCE ), which occurs in the BNC 6,163 times and is thus even more frequent than substance (2,152) let alone substantive (950). (Because one of the stress patterns of finance is, as pointed out in 4.2.3, / fəˈnæns/ , which is the preserved stress of the more frequently used / fəˈnænʃəl/ of financial , it is possible that at some point in the future, also the less frequently used disyllable substance will be stressed not (only) initially but (also) finally. From a synchronic point of view, it is difficult to explain why finance is both / ˈfaɪnæns/ and / fəˈnæns/ , but substance is only / ˈsʌbstəns/ . Perhaps the back-derivation analysis is more obvious in the case of finance because the more frequently used financial has 16,559 attestations in the BNC , whereas the more frequently used substantial occurs only 6,163 times.) Another frequent variability scenario, which was already mentioned in 4.3.6, involves a trisyllabic prefixed derivative that, however, in accordance with the segmental length principle, should be stressed not penultimately but antepenultimately. E.g., posthumous is in American English stressed either / ˈpɑstʃəməs/ or / pɑˈst(j)uməs/ ( OED ) and similarly, the adjective posthumate and the verb inhumate vacillate, according to the OED , between root and prefix stress: / ˈpɑstʃəˌmeɪt/ vs. / pɑˈst(j)uˌmeɪt/ (only in American English) and / ɪnˈhjuːmeɪt/ vs. / ˈɪnhjuːmeɪt/ . In spite of the semantic transparency (and hence importance) of the prefix post in both posthumous and posthumate , it would be counterintuitive to attribute antepenultimate stress in these words to emphasis alone: Cf. postorbital , which in spite of the semantic transparency of the prefix post -, has only the root stress / pəʊstˈɔːbɪt(ə)l/ (OD) and, similarly, insignificant is normally stressed / ɪnsɪɡˈnɪfɪk(ə)nt/ ( OD ) even though the prefix in modifies the meaning of the base significant in an important way. The reason why the trisyllabic prefixed derivatives posthumous , posthumate , and inhumate are often (and in British English, exclusively) pronounced with antepenultimate stress is that in their lefthand strings post . hu and in . hu -, stress should be initial in accordance with the segmental length principle (and their righthand strings mous and mate cannot take stress because ous and ate occur in English as suffixes and are therefore stress-repellent). (In the case of inhumate , which does not have attestations in the BNC whereas inhumation occurs 18 times, also the back-derivation analysis contributes to antepenultimate stress in it.) A related scenario is when the segmental length principle is in conflict with the analysis of a trisyllabic word as a suffixed derivative whose base is a finally-stressed disyllable / a penultimately-stressed trisyllable. E.g., Abbasid is in both British and American English interchangeably stressed / əˈbasɪd/ and / ˈabəsɪd/ (OED). Penultimate stress in the suffixed derivative Abbasid , “[a] member of a dynasty of caliphs who ruled in Baghdad from 750 to 1258, founded by the descendants of al-῾Abbās” ( OED ), is the preserved final stress of the disyllab- 148 5 Case studies ic base Abbas : / əˈbɑːs/ (OD), which is more frequent in the BNC than Abbasid : 88 vs. 10. Antepenultimate stress is, by contrast, in accordance with the segmental length principle: Ab . ba - + sid . A similar example is / kɒˈrəʊnəl/ vs. / ˈkɒrənəl/ of coronal ( OED ). On the one hand, since the shorter word corona is more frequent in the BNC than the longer word coronal (70 vs. 34), the stress pattern of the former is supposed to be preserved in the latter, i.e., coronal should be stressed / kɒˈrəʊnəl/ because corona is stressed / kəˈrəʊnə/ ( OD ). Simultaneously, penultimate stress in coronal can also be due to the analysis of this derivative as a prefixed rather than suffixed derivative, i.e., co - + ronal (cf. neuronal , for which the OD , in contrast to the OED , gives only the transcription / njʊəˈrəʊn(ə)l/ , with stress in this prefixed derivative likewise falling upon the root ronal ). Pronouncing coronal / kɒˈrəʊnəl/ is, however, not in accordance with the segmental length principle, which requires initial stress in its lefthand disyllabic string co . ro -. The ative -trisyllables allative , constative , dictative , mandative , mutative , notative , optative , placative , privative , and prolative vacillate between penultimate and antepenultimate stress because on the one hand, the string ative occurs in English as a suffix (e.g., talkative ) and therefore repels stress, while on the other hand, the segmentally longer disyllabic string ative can in accordance with the segmental length principle be regarded as the root of the above mentioned trisyllabic ative -words. (In constative and mandative , the prefixation analyses, which result in the pronunciations / kənˈsteɪtɪv/ and / manˈdeɪtɪv/ ( OED ), are reinforced by the existence of the disyllabic words stative and dative ; a very similar case is / ɔːˈneɪtɪv/ of ornative ( OD ), which is the preserved stress of the disyllabic base native . Of the 960 ive -words in the OD dictionary, 158 (~16.46 %) are words such as abstractive , whose righthand substrings are words such as active , which contain three orthographic vowels. At the same time, however, 203 (~21.15 %) ive -words are words such as computative , whose righthand substrings are words such as putative , which contain four orthographic vowels. Thus the trisyllabic ive -word abstractive is stressed / əbˈstraktɪv/ ( OD ) because its disyllabic righthand string active is stressed / ˈaktɪv/ ( OD ) (and because apart from the tractive -word abstractive , there are also the tractive -words attractive , extractive , and subtractive ), but the tetrasyllabic ive -word computative is stressed / kəmˈpjuːtətɪv/ ( OD ) because its trisyllabic righthand string putative is stressed / ˈpjuːtətɪv/ ( OD ). The stress of an ive -word in English is thus often simply the stress of another, shorter ive -word.) The stress pattern / ɔːˈfiːən/ (OED; only British English) is the etymological penultimate stress of the Latin base Orphēus , which has a long vowel in the penult. The English base Orpheus is, however, stressed / ˈɔːfɪəs/ ( OD ), which is why also the derivative Orphean is supposed to be stressed / ˈɔːfɪən/ (the only stress pattern 5.1 Stress variation in the OED 149 in American English). The same applies to the variation / mɔːˈfiːən/ vs. / ˈmɔːfɪən/ of Morphean ( OED ), with stress in the derived form being either the preserved penultimate stress of the Latin base Morphēus or the preserved antepenultimate stress of the English base Morpheus : / ˈmɔːfɪəs/ ( OD ). Recall also that, as pointed out in 4.2.1, English has both the finally-stressed disyllabic adjective / ɔːˈɡʌst/ and the initially-stressed disyllabic noun / ˈɔːɡəst/ . The trisyllabic derivative Augustin is therefore interchangeably stressed / ɔːˈɡʌstɪn/ and / ˈɔːɡəstɪn/ ( OED ). 167 (~15.24 %) p/ a-instances end orthographically in the vowels a , i , o , which, as established in 4.3.9, are one of the predictors of penultimate stress in a trisyllabic English word. E.g., for pavlova (originally Pavlova cake ), the OED gives the British English transcriptions / pavˈləʊvə/ and / ˈpavləvə/ and the American English transcription / pɑvˈloʊvə/ . Just like in the source language Russian, stress in the derivative Pavlova (the surname of the ballerina Anna Pavlova; according to the OED , “[t]he name was probably given to the dessert to commemorate [her] tour of Australia and New Zealand in 1926”) is supposed to be antepenultimate, i.e., / ˈpavləvə/ , preserving the initial stress of the disyllabic base Pavlov , i.e., / ˈpavlɒf/ ( OD ), which is considerably more frequent in the BNC than Pavlova (111 vs. 34). At the same time, however, because of the last orthographic vowel a , in which pavlova ends, stress in it is in contemporary English predominantly (and in American English exclusively) penultimate. A very similar case is / -dəːˈmɑːtə/ vs. / -ˈdɜːmətə/ of Echinodermata (OED; only British English). Since morphologically, dermata is derm + ata , stress in this suffixed derivative is supposed to be placed upon the only syllable constituting the monosyllabic base derm : / dəːm/ ( OD ). Dermata is thus supposed to be pronounced / ˈdɜːmətə/ , with the suffix ata repelling stress. Because, however, the last orthographic vowel in which dermata ends is a , this suffixed derivative is in Present-day English also associated with the penultimately-stressed pronunciation / dəːˈmɑːtə/ . The same is true of the combining forms aceto -, adelo -, amino -, crystallo -, echino -, gameto -, Germano -, immuno -, membrano -, metallo -, myceto -, myrmeco -, nemato -, organo -, ornitho -, oscheo -, oscillo -, oxalo -, pallado -, piezo -, pneumato -, poikilo -, prismato -, pterido -, pyrito -, sidero -, Theatro -, tracheo -, and trigono -, which all end in o and have therefore come to be associated with penultimate stress even though—from a morphological point of view—many of these combining forms were derived from initially-stressed disyllabic bases and should therefore be stressed antepenultimately. E.g., because the base metal is stressed / ˈmetl/ (LDOCE), also the derivative metallo is supposed to be stressed / ˈmɛtələʊ/ ( OED ). Because, however, the last orthographic vowel in which this combining form ends is o , metallo is stressed not only / ˈmɛtələʊ/ but also / mɛˈtaləʊ/ ( OED ; both in British and American English). 150 5 Case studies Note also that of the p/ a-instances in the OED , 20 are according to LDOCE stressed only penultimately. E.g., the above mentioned pavlova is, according to LDOCE , only / pævˈləʊvə/ in British English and only / pɑːvˈloʊvə/ in American English; likewise, angina is / ˈændʒɪnə/ vs. / ænˈdʒaɪnə/ in the OED , but LDOCE gives only the penultimately-stressed transcription / ænˈdʒaɪnə/ . Of the 20 words such as pavlova and angina , eight (40 %) are the words angina , Attila , incognito , Miskito , patchouli , pavlova , piano , and Satsuma , which end orthographically in the vowels a / i / o . By contrast, of 27 p/ a-instances that according to LDOCE are stressed only antepenultimately (e.g., enema is / ˈɛnɪmə/ vs. / ɛˈniːmə/ in the OED , but LDOCE gives only the transcription / ˈenəmə/ ), only three (~11.11 %) are the words enema , Medici , and vertigo , which end orthographically in a / i / o . The difference of eight / 20 vs. three / 27 is statistically significant: χ 2 (1) = 3.859, p = 0.0495 (with Yates’ correction). Noteworthy are also emphatically-conditioned cases of p/ a-instability among trisyllabic English words. E.g., in the tuplet -words octuplet , quintuplet , sextuplet , stress falls not only upon the root tuplet but also upon the semantically important prefixes oc -, quin -, sex -. The same applies to majuscule and its counterpart minuscule . A similar case is / ˈjuːdʒɪnɪst/ vs. / juːˈdʒiːnɪst/ of eugenist , / juːˈpætrɪd/ vs. / ˈjuːpətrɪd/ of eupatrid , and / ˈjuːpɛpsɪ/ vs. / juːˈpɛpsɪ/ of eupepsy ( OED ). The meanings of these derivatives are all related to the meaning “well,” which is the meaning of the Greek word eu , to which the English prefix eu etymologically goes back ( OD ). E.g., eupepsy means “[h]ealthy action of the digestive organs; good digestion” (OED). Therefore, in addition to having the root stress / -ˈdʒiːnɪst/ , / -ˈpætrɪd/ , and / -ˈpɛpsɪ/ , these derivatives also have the prefix stress / ˈjuː-/ . Recall also that stalactite is / ˈstælæktaɪt/ vs. / stəˈlæktaɪt/ and, similarly, stalagmite , whose meaning can be regarded as the directional opposite of the meaning of stalactite , is / ˈstælæɡmaɪt/ vs. / stəˈlæɡmaɪt/ . On the one hand, because in the lefthand disyllabic strings sta . lac and sta . lag -, stress should be initial in accordance with the segmental length principle and the righthand monosyllabic string mite cannot take stress (cf. the ite of meteorite ), both stalactite and stalagmite are supposed to be stressed antepenultimately: / ˈstæ-/ . At the same time, however, since these two words are often contrasted with each other and since the formal contrast between the sta . lac of stalactite and the sta . lag of stalagmite is due to the coda consonant in which these strings end, stalactite and stalagmite are also sometimes stressed / -ˈlæk-/ and / -ˈlæɡ-/ . So far we have been concerned with variably-stressed trisyllables, which constitute the bulk of the variation category “p/ a” (740 / 1,096=~67.52 %). Tetrasyllables and longer words are, however, also interchangeably pronounced with penultimate and antepenultimate stress when more than one syllabic boundary, 5.1 Stress variation in the OED 151 namely penult-antepenult and antepenult-pre-antepenult, can simultaneously qualify as the root-prefix boundary. A good example is the numerous variably-stressed words ending in the strings poda , podal , or podan : Actinopoda , Arthropoda , gastropoda , octopodal / octopodan , pseudopodal , Tetrapoda , etc. The “correct” morphological segmentation of, e.g., Arthropoda is into the prefix Arthro and the root poda , which etymologically goes back to the Ancient Greek ποδ -, πούς , meaning “foot” ( OED ). This segmentation results in the stress pattern / -ˈpəʊdə/ ( OED ; both in British and American English). At the same time, however, since the common element that Arthropoda shares with, e.g., Actinopoda is not poda but opoda and since the common element that Arthropoda shares with gastropoda is ropoda , the alternative segmentations of Arthropoda are Arthr - + opoda or Arth - + ropoda . These segmentations result in the pronunciations / ɑːˈθrɒpədə/ in British English and / ɑrˈθrɑpədə/ in American English ( OED ), with the location of stress in these pronunciations being the alternative root-prefix boundary location in the same word Arthropoda . Similarly, the words macroglia , microglia , and neuroglia all vacillate between the stress patterns / -ˈrɒɡlɪə/ and / -ˈɡlʌɪə/ ( OED ; both in British and American English). The “correct” morphological segmentation of these words is obviously macro -/ micro -/ neuro - + glia ; stress in these words is thus supposed to be penultimate, i.e., / -ˈɡlʌɪə/ . However, since the common string, shared by macroglia , microglia , and neuroglia , is not glia but roglia (or croglia in the case of macroglia and microglia ), the very same words can also be segmented into the prefixes ma -/ mi -/ neu and the root roglia / croglia . This morphological segmentation results in the stress pattern / -ˈ(k)rɒɡlɪə/ , with the location of stress being the alternative root-prefix boundary location in the same words under consideration. A very similar case is centrifugal and febrifugal , which both vacillate between / -ˈfjuːɡəl/ and / -ˈrɪfjʊɡəl/ (OED) because either the shared disyllabic string fugal , which occurs in English as a separate word, or the shared bound trisyllabic string rifugal can count morphologically as the root of these words. Gladiolus is both / ɡlæˈdaɪələs/ and / ɡlædɪˈəʊləs/ ( OED ) because the root of this word is either the trisyllabic bound string diolus , which gladiolus shares with modiolus , or the disyllabic bound string olus , which gladiolus shares with, e.g., alveolus . A different example is Caribbean , which in both British and American English is interchangeably stressed / ˌkarᵻˈbiːən/ and / kəˈrɪbɪən/ ( OED ) because on the one hand, there is the finally-stressed word Caribbee , / karᵻˈbiː/ ( OD ), while on the other hand, there is also the penultimately-stressed word caribe : / kəˈribi/ (Dictionary.com). (Etymologically, penultimate stress in the English word Caribbean is most likely due to penultimate stress in the Spanish word caribeño , whereas antepenultimate stress appears to preserve the stress pattern / ka.ˈri: .bi.ko/ ( PONS ) of the Italian word caribico .) Similarly, the stress pattern 152 5 Case studies / əkɪˈliən/ of Achillean ( OED ); both in British and American English) is the etymological penultimate stress of the Latin base Achillēus , which has a long vowel in the penult. In the target language English, the base Achilles is, however, stressed / əˈkɪliːz/ ( OD ), which is why also the derivative Achillean is as a rule stressed / əˈkɪlɪən/ ( OED ). 5.1.3.2 Final vs. penultimate stress The second largest variation category “u/ p” is different from the type “p/ a” in that final stress is in general “a stress pattern foreign to English” (Poldauf 1984: 77). Therefore, it comes as no surprise that of the 653 orthographically non-identical u/ p-instances, 118 (~18.07 %) have an entry in Speake & LaFlaur’s (2002) Oxford Essential Dictionary of Foreign Terms in English, which was mentioned in the previous chapter. Accordingly, the variation between final and penultimate stress exhibited by (many of) these items can often be seen as foreign vs. native stress pattern. Thus, for instance, analyzing cases such as / ˈɡʊruː/ vs. / ɡʊˈruː/ of guru ( OED ) and / ˈtælməd/ vs. / tælˈmuːd/ of Talmud ( OED ), we note that in both CVCV and CVCCVC -disyllables (i.e., disyllables such as guru and Talmud ), stress in English is (in accordance with the segmental length principle) almost exclusively initial: Of the 744 CVCVand 2,444 CVCCVC-disyllables in the OD dictionary, only 34 (~4.57 %) and 158 (~6.46 %) have final stress. E.g., defy is stressed / dɪˈfʌɪ/ ( OD ), violating the segmental length principle from an orthographic point of view but abiding by it phonetically (i.e., the ult / ˈfʌɪ/ is segmentally longer than the penult / dɪ/ ); additionally, final stress in defy can be regarded as the preserved penultimate stress of the more frequently used / dɪˈfʌɪəns/ and / dɪˈfʌɪənt/ ( OD ), i.e., the corresponding frequency numbers from the BNC are 515 ( defiance ) vs. 339 ( defiant ) vs. 304 ( defy ). (Given these numbers, note that in contrast to the OED , the OD gives for guru and Talmud only the initially-stressed transcriptions / ˈɡʊruː/ , / ˈtalmʊd/ or / ˈtalməd/ .) Thus, it seems that in words such as guru and Talmud , in which stress should be initial in accordance with the segmental length principle, final stress serves essentially as a foreignness marker, i.e., stress in the words guru and Talmud can be final because 1) final stress is “a stress pattern foreign to English” (Poldauf 1984: 77) and 2) neither Jews nor Hindus constitute the majority of Present-day British and American societies; the concepts associated with the words guru and Talmud are thus inherently foreign to a prototypical native English speaker. Likewise, the variation / baɪˈrɑːm/ vs. / ˈbaɪrəm/ of Bairam , / ˈbæntuː/ vs. / bænˈtuː/ of Bantu , / ˈbɜːlɪn/ vs. / bɜːˈlɪn/ of Berlin , / ˈkældiː/ vs. / kælˈdiː/ of Chaldee , / ˈhɑːlə/ vs. / xɑːˈlɑː/ of challah (only British English), / ˈfɑːsiː/ vs. / ˌfɑːˈsi/ of Farsi (only British English), / ˈhælvə/ vs. / halˈvaː/ of halva , / ʃɔˈɑː/ vs. / ˈʃəʊə/ of Shoah , 5.1 Stress variation in the OED 153 / ˈtʊŋʊs/ vs. / tʊŋˈuːs/ of Tungus ( OED ), etc. is closely connected to the fact that the things to which the words Bairam , Bantu , Berlin , Chaldee , challah , Farsi , halva , Shoah , Tungus , etc. refer are inherently non-English. Note also that in contrast to the type “p/ a,” which is mainly constituted by trisyllables, it is disyllables (532 / 653=~81.47 %) that form the bulk of the variation type “u/ p.” (An example of a trisyllabic word vacillating between final and penultimate stress is accouchement , which is / ˌɑkuʃˈmɑnt/ vs. / əˈkuʃmənt/ in American English ( OED ); similarly, Megillah is in British English / mᵻˈɡɪlə/ vs. / məɡiːˈlɑː/ ( OED ). Penultimate stress in these trisyllables is in accordance with the segmental length principle— ac . couche - + ment and Me . gil - + lah (with the latter being like a suffix (cf. Abdullah , Ayatollah , Hezbollah , Ramallah , etc.) and thus repelling stress)—whereas final stress is due to the perceived foreignness of the words accouchement and Megillah , both of which occur in Speake & LaFlaur’s (2002) dictionary.) For 225 variably-stressed disyllables from the OED , syllabified phonetic transcriptions are given in the MRC database. 159 (~70.67 %) of these disyllables have segmentally longer ults (e.g., @/ dres of address ); the average values are ~3.43 (ult) vs. ~2.41 (penult). The only issue that must be further established in connection with such items is thus why in addition to being stressed finally (which is how these words are supposed to be stressed in accordance with the segmental length principle), these disyllabic English words can also be stressed initially. The usual scenario involves a conflict between prefixation and suffixation analyses, both of which make sense in the case of one and the same disyllabic English word. For example, in addition to the prefixation analyses o - + vert for overt and co - + vert for covert , also the suffixation analyses over / cover + t can be easily arrived at by an English speaker (because especially in the case of covert , the connection between the meanings “covert” and “cover” is more than obvious). Indeed, covert is, according to both LDOCE and the OD , exclusively initially-stressed as / ˈkʌvət/ or / ˈkəʊvəːt/ , and similarly, in the MWO , the initially-stressed transcription / ˈkō-(ˌ)vərt/ is placed before the finally-stressed alternative / kō-ˈvərt/ . The analysis cover + t can thus be said to make more sense for a contemporary English speaker than co - + vert . Observe also that among variably-stressed words such as address , in which the ult is segmentally longer than the penult, are 11 age -words homage , mirage , orage , Osage , paysage , photomontage , portage , potage , presage , propage , rampage and five ard -words Jacquard , Picard , Pommard , Reynard , toward . (Note also that in the YouTube video K2l0ngf0 PG 8 (26. 05. 2017), a British English speaker can be heard saying at 00: 00: 29.748 --> 00: 00: 32.745 that she will never be able to figure out how to stress Barnard of Sara Barnard ). Each of these disyllables should have final stress in accordance with the segmental length principle / be- 154 5 Case studies cause of the foreignness of what (some of) these words stand for (e.g., “[a] native or inhabitant of the northern French region of Picardy” ( OED ), which is a meaning of Picard ), but because the strings age and ard occur in English as suffixes (e.g., linkage , drunkard ) and therefore repel stress, each of these words also has an initially-stressed alternative. E.g., Picard is / ˈpɪkɑːd/ vs. / pᵻˈkɑːd/ in British English ( OED ). The same applies to the our -words contour , detour , notour , and tambour . Because the word-final our is an orthographic alternative to or , which occurs in English as a suffix, both our and or -disyllables are almost exclusively—35 / 43=~81.4 % and 107 / 115=~93.04 %—pronounced in English with initial stress. E.g., colour / color is / ˈkʌlə/ ( OD ) even though it is not segmentable into the base col and the suffix or the way, e.g., the genuine suffixed derivative actor is segmentable into the base act and the suffix or . This (i.e., the stress-repellent nature of the string our ) is essentially the reason why English speakers are reluctant to pronounce the words contour , detour , notour , and tambour with final stress, even though in the case of contour , detour , and notour , the ult . tour should count morphologically as the root not only because it is segmentally longer than the penults co -/ de -/ no but also because it occurs in English as a separate word. Noteworthy are also 17 ate -words such as innate , irate , prolate , etc. and eight ose -words such as globose , plumose , rugose , etc., many of which are supposed to have final stress not only due to the segmental length principle but also due to the analyses of these words as back-derivatives from longer English words. E.g., the longer word migration occurs in the BNC 1,317 times while there are only 320 attestations in the corpus of the shorter word migrate . Accordingly, because the base noun migration is stressed / mʌɪˈɡreɪʃ(ə)n/ ( OD ), the derived verb migrate is supposed to be stressed / mʌɪˈɡreɪt/ ( OED ; only British English). At the same time, however, because ate occurs in English as a suffix (cf. pulsate ) and therefore repels stress, migrate is also frequently stressed / ˈmʌɪɡreɪt/ ( OED , both British and American English). Cf. / pʌlˈseɪt/ vs. / ˈpʌlseɪt/ of pulsate ( OD ), whose segmentation into the base pulse and the suffix ate is completely transparent both formally and semantically; this segmentation results in the initially-stressed pronunciation / ˈpʌlseɪt/ , with stress in it falling upon the only syllable constituting the monosyllabic base pulse . At the same time, however, similar to the verb migrate , also the verb pulsate is less frequent in the BNC than the corresponding ion -noun pulsation (14 vs. 18) and especially pulsatance , which occurs in the corpus 51 times. Thus because the more frequently used longer nouns pulsation and pulsatance are stressed / pʌlˈseɪʃ(ə)n/ and / pʌlˈseɪt(ə)ns/ ( OD ), the less frequently used shorter verb pulsate is, apart from being stressed / ˈpʌlseɪt/ , also stressed / pʌlˈseɪt/ . 5.1 Stress variation in the OED 155 A similar case is the verbs diagnose , metamorphose , and necrose , which have fewer occurrences in the BNC than the corresponding osis -nouns diagnosis , metamorphosis , and necrosis : 172 vs. 1,707, 17 vs. 109, no attestations vs. 114. The former can thus be regarded as back-derivatives from the latter (which is also true from a diachronic point of view; see the OD / OED for details) and are therefore pronounced / -ˈnəʊz/ , / -ˈfəʊz/ or / -ˈfəʊs/ , and / -ˈkrəʊs/ ( OED ), i.e., e.g., since the base diagnosis is stressed / ˌdaɪəɡˈnəʊsɪs/ ( LDOCE ), the derivative diagnose is supposed to be stressed / dʌɪəɡˈnəʊz/ . (As for metamorphosis , observe that in British English this noun vacillates between the stress patterns / -ˈmɔːfəsɪs/ and / -ˈfəʊsɪs/ ( OED ), with antepenultimate stress being in accordance with the segmental length principle ( mor . pho - + sis ) vs. penultimate stress being due to the analysis of the string osis as the root. The variation between final and initial stress exhibited by the morphose of the back-derivative metamorphose can thus be seen as the inherited variation between penultimate and antepenultimate stress exhibited by the morphosis of the base form metamorphosis .) At the same time, however, since each of these derived verbs ends in the string ose , which occurs in English as a suffix (e.g., comatose ), they are also often pronounced / ˈdʌɪəɡnəʊz/ , / -ˈmɔːfəʊz/ , and / ˈnɛkrəʊs/ ( OED ), i.e., in the disyllabic morphose and necrose , stress is placed upon the syllables constituting the stems morph and necr and in the trisyllabic di . ag . nose , stress in the lefthand disyllabic string di . ag is initial because the ult in it has the same segmental length as the penult. Second, in Section 4.2.2 it was observed that lexicographers (and theoretical linguists) sometimes give more than one answer to the question of whether a medial consonant in a disyllabic English word is onset of the ult or coda of the penult. E.g., adult is &/ dVlt in the MRC vs. ad∙ult in LDOCE . Assuming that syllabification uncertainties such as a . dult vs. ad . ult are also characteristic of laypersons, we can suggest that final stress in adult is the stress pattern of the syllabification a . dult (because prototypically, the ult of a finally-stressed disyllable contains four orthographic segments) whereas initial stress is due to the syllabification ad . ult , where the ult is composed of only three segments; recall also that the ult of a finally-stressed disyllable usually contains two more orthographic segments than the penult, and in the case of the syllabification ad . ult , the segmental length difference is only one. A very similar case is adept , which is &/ dept in the MRC vs. ad∙ept in LDOCE . Proceeding from the latter, where the ult consists of only three segments, we obtain the initially-stressed pronunciation / ˈadɛpt/ , which, according to the OED , occurs only in British English. The syllabification a . dept results, by contrast, in the finally-stressed pronunciation / əˈdɛpt/ ( OED ), which is the only stress pattern of adept in American English. 156 5 Case studies Synapse is / ˈsi-ˌnaps/ vs. / sə-ˈnaps/ in the MWO , which means that the medial consonant / n/ is analyzed only as onset of the ult. LDOCE gives, however, the hyphenation syn∙apse, in which the same medial consonant / n/ is said to be coda of the penult. If synapse is syn . apse rather than sy . napse , the ult is still longer than the penult, but the difference between them is not two—which is usually exhibited by a finally-stressed disyllable in English—but one. Stress in synapse is therefore initial (which, according to LDOCE, is the only stress pattern of synapse in British English). The other syllabification, sy . napse , in which the length difference is three, results, by contrast, in final stress. Malay is back-derivative from Malaysia (93 vs. 664) and should therefore be stressed / məˈleɪ/ ( OED ), preserving the stress of / məˈleɪziə/ ( LDOCE ). However, even if we proceed from the syllabification Ma . lay , which is given in the MRC (m@/ leI), the ult in Malay still consists of only three segments. Stress in Malay can therefore also be initial: / ˈmeɪˌleɪ/ ( OED ; only American English). Obit (= obituary ) is 0/ bIt in the MRC and o∙bit in LDOCE (pronouncing obit / əˈbɪt/ ( OED ), preserving the stress of / əˈbɪtʃuəri/ ( LDOCE ), is thus in accordance with the segmental length principle), but obituary is, according to the OD , ob¦itu|ary. Accordingly, assuming that also the shorter obit is sometimes ob . it rather than o . bit , we naturally arrive at the initially-stressed alternative / ˈəʊbɪt/ , which is the only pronunciation given in LDOCE . Third, the reader is reminded of the fact that prefix stress in a disyllabic English word is often a matter of emphasis. For example, the noun increase is according to the OED “/ ˈɪnkriːs/ formerly / ɪnˈkriːs/ ” and similarly, the noun import is “/ ˈɪmpɔːt/ , formerly / ɪmˈpɔət/ ” (whereas to export is still stressed both / ɛkˈspɔət/ and / ˈɛkspɔət/ and similarly, for a decrease the OED gives both the finally-stressed transcription / dɪˈkriːs/ and the initially-stressed alternative / ˈdiːkriːs/ ). Since the prefixes in of increase and im of import make these prefixed derivatives formally different from the semantically opposite prefixed derivatives decrease and export , it is natural that the prefix stress has over the course of time become these nouns’ default stress pattern (and similarly, in the case of the corresponding verbs to increase and to import , this stress pattern can be heard as well. Note also that in the case of to export , the prefix stress ˈ export was in captioned YouTube videos more frequent than the root stress ex ˈ port ). Similarly, to discount is, according to the OED , stressed not only / dɪˈskaʊnt/ but also / ˈdɪskaʊnt/ and to refund is in British English interchangeably stressed / (ˌ)riːˈfʌnd/ and / ˈriːfʌnd/ . Given that discounting can be seen as negative counting and refunding as funding again, i.e., refilling someone’s funds, we have reasons to think that initial stress in these verbs is of a purely (or mainly) emphatic nature. 5.1 Stress variation in the OED 157 Further examples among nouns are / rɪˈsəːtʃ/ vs. / ˈriːsəːtʃ/ of research ( OD ) and / ˈri-/ vs. / -ˈsɔ(ə)rs/ of resource (OED; only American English), in which the prefix re carries the meaning “repetition of an action”: Research is a repeated search (for scientific truth) and a resource is a source that can be used repeatedly. Occult is in British English stressed not only upon its root cult , i.e., / əˈkʌlt/ ( OED ), but also upon the prefix oc -: / ˈɒkʌlt/ ( OED ). The likeliest explanation seems to be the perception of the occult as a special form of cult. In contrast to the only initially-stressed ant -adjectives constant , distant , and instant , the ant -adjective extant is stressed both initially and finally: / ˈɛkstənt/ vs. / ɛkˈstænt/ ( OED ). Since one of the meanings of extant is “[i]n existence; existing” ( OD ), which makes the adjective extant the semantic opposite of the adjective extinct , the finally-stressed version / ɛkˈstænt/ stills remains an alternative to the initially-stressed pronunciation / ˈɛkstənt/ , i.e., in contrast to cases such as increase vs. decrease , the formal difference between the semantically opposite terms extant and extinct is due to their ults . tant and . tinct , which are therefore emphasized via stress. (This claim is strongly supported by the fact that in YouTube videos that were found to contain the spoken occurrences of extant , stress in it was final even in combinations such as extant species , extant creature , extant mammals , etc., in which extant modifies an initially-stressed head. Because extant species , extant creature , extant mammals , etc. contrast semantically with extinct species , extinct creature , extinct mammals , etc., the usual stress pattern of extant in these combinations is / ɛkˈstænt/ even though this is unfortunate from the point of view of rhythm.) Fourth, as established in 4.2.5, a disyllable such as tribute , which occurs in English not only separately but also as a righthand part of a trisyllable such as contribute , is more likely to receive initial stress than a disyllable such as, e.g., dispute , which occurs in English only as a separate word. Accordingly, analyzing the variation / pɪˈnʌlt/ vs. / ˈpɛnʌlt/ of the linguistic term penult ( OD ), we note that on the one hand, since there is the shorter word ult , the morphological structure of the longer word penult is pen - + ult ; stress in penult is therefore final. (Additionally, because the shorter word penult does not have attestations in the BNC whereas the longer word penultimate occurs in the corpus 163 times, the former is secondary to the latter. Thus, because the base form penultimate is stressed / pɪˈnʌltɪmət/ ( OD ), the derived form penult is supposed to be stressed / pɪˈnʌlt/ .) At the same time, however, since apart from occurring in the shorter word penult , this string also occurs at the end of the longer word antepenult , whose morphological structure is thus ante - + penult , the string penult can also be regarded as the root of the shorter word penult itself; this results in the initially-stressed pronunciation / ˈpɛnʌlt/ . (Additionally, this stress pattern is in accordance with the syllabification pen . ult , which follows from the line break 158 5 Case studies pen¦ult, given in the OD .) A similar case seems to be manteau , which apart from occurring as a separate word, also occurs as a righthand part of the trisyllabic word portmanteau . Manteau is therefore stressed not only / mænˈtoʊ/ but also / ˈmæntoʊ/ , which, according to the OED , is the only stress pattern in British English. As for u/ p-instances in which stress can be final even when the ult is segmentally not longer than the penult, the usual case involves a disyllable whose ult occurs in English as a separate word. E.g., the verb combat can be stressed / kɒmˈbæt/ ( OED ) because the string bat , in which this verb ends, occurs in English as a separate word; the analysis of combat as a prefixed derivative ( com - + bat ) can thus be arrived at by an English speaker even though the ult in this word contains the same number of both phonetic and orthographic segments as the penult. The usual stress pattern of combat is, however, the initially-stressed / ˈkɒmbæt/ ( OED ), which abides by the segmental length principle. A fairly similar case is / səˈfɪks/ vs. / ˈsʌfɪks/ of to suffix ( OED ). Final stress in this verb can be arrived at because, again, the righthand string fix , which consists of the same number of orthographic segments as the lefthand string suf -, occurs in English as a separate word. A different case is / ˈpriːlɪm/ vs. / prᵻˈlɪm/ of prelim (OED; only British English), which is a shortening of preliminary . Thus because the longer word preliminary is stressed / prɪˈlɪmɪn(ə)ri/ ( OD ), its shortening prelim can be stressed / prᵻˈlɪm/ even though the segmental length principle requires initial stress in this disyllabic string. 5.1.3.3 Final stress vs. other stresses Similar analyses also seem to apply to the variation types “u/ a,” “u/ 4,” and “u/ p/ a,” which involve final stress in trisyllables and longer English words. This stress pattern is either the stress pattern of a foreign language serving for English speakers as the source language or / and the stress pattern of a related longer English word serving for a corresponding shorter word as the base (back-formation/ shortening). For example, the shorter academe and Yugoslav have fewer occurrences in the BNC than the longer academia and Yugoslavia : 40 vs. 102 and 616 vs. 1,337. Accordingly, because academia is stressed / ˌakəˈdiːmɪə/ ( OD ) and Yugoslavia / ˌjuːɡə(ʊ)ˈslɑːvɪə/ ( OD ), academe and Yugoslav are supposed to be stressed / -ˈdiːm/ and / -ˈslɑːv/ ( OED ). However, because the eme of academe occurs in English as a suffix (e.g., phoneme ) and the Slav of Yugoslav exists as a separate word, academe and Yugoslav are usually pronounced / ˈakədiːm/ and / ˈjuːɡəʊslɑːv/ ( OED ), with the secondary stress of academia and Yugoslavia being promoted to primary stress. 5.1 Stress variation in the OED 159 Similarly, final stress in the shorter English word avatar (which, however, is not given in the OD / LDOCE ) is the etymological penultimate stress of the longer Sanskrit word avatāra , “descent,” consisting of the components ava , “down,” and tar -, “to pass over” ( OED ). Because, however, the string ar , in which avatar ends, occurs in English as a suffix (e.g., liar ), avatar is in contemporary English stressed / ˈavətɑː/ ( OD ), with its original secondary stress having over the course of time become the word’s primary stress. The shortening of the Latin tetrasyllables absolūtus , obsolētus , and opportūnus , in which stress is penultimate because the vowel in the penult is long, gave rise to the finally-stressed trisyllabic English adjectives abso ˈ lute , obso ˈ lete , and oppor ˈ tune . Note also that while the shorter word opportune occurs in the BNC only 82 times, the longer word opportunity has 10,096 attestations. The former is thus a back-derivative from the latter and should therefore preserve its stress, i.e., because opportunity is stressed / ˌɒpəˈtjuːnəti/ ( LDOCE ), opportune is supposed to be stressed / ˌɒpəˈtjuːn/ ( OED ). Similarly, because one of the stress patterns of the longer word absolutely , which is more frequent in the BNC than the shorter word absolute (5,670 vs. 3,428), is / ˌabsəˈluːtli/ ( OED ), the stress pattern / ˌabsəˈluːt/ of absolute ( OED ) can be arrived at by a contemporary English speaker without the knowledge of the stress pattern of the Latin etymon word absolūtus . Final stress in the adjectives absolute , obsolete , and opportune is, however, unfortunate from the point of view of rhythm as each of these adjectives often occurs in the modifier position. E.g., of all the 3,428 occurrences of absolute in the BNC , 1,579 (~46.06 %) are occurrences in combinations such as absolute terms , absolute zero , absolute value , absolute minimum , absolute numbers , absolute truth , etc., in which the modifier absolute is often followed by either a monosyllabic or initially-stressed polysyllabic head: Of the 646 nouns preceded by absolute in the BNC (in the case of which phonetic transcriptions are given in the OD ), 434 (~67.18 %) are either monosyllables or initially-stressed polysyllables. Accordingly, instead of / ˌabsəˈluːt/ , / ˌɒbsəˈliːt/ , and / ˌɒpəˈtjuːn/ , these adjectives are as a rule stressed / ˈabsəluːt/ , / ˈɒbsəliːt/ (in British English), and / ˈɒpətjuːn/ ( OD ), with the secondary stress being promoted to primary stress. (Among the factors contributing to antepenultimate stress in absolute and opportune is also the existence of the monosyllabic words lute and tune .) An example of the category “u/ 4” is admiraless , in which stress can be final, / ˌadm(ə)rəˈlɛs/ ( OED ; only British English), due to the emphatic potential inherent in the suffix ess . An example of the category “u/ p/ a” is marcottage , which is / mɑːkɒˈtɑːʒ/ , / ˈmɑːkɒtɪdʒ/ , or / mɑːˈkɒtɪdʒ/ ( OED ; both in British and American English). The pronunciation / mɑːkɒˈtɑːʒ/ , which contains only full vowels, is an imitation of how this word is pronounced in the source language French, in which, as explained in 2.4, “vowel reduction is much less evident” (Payne et al. 160 5 Case studies 2012: 204) than it is in a language such as English. The alternatives / ˈmɑːkɒtɪdʒ/ and / mɑːˈkɒtɪdʒ/ are, by contrast, regular pronunciations: The former is due to the morphological analysis mar . cot - + tage , i.e., because the string age occurs in English as a suffix (e.g., linkage ) and therefore repels stress, stress in marcottage should occur in the beginning of its remaining disyllabic string mar . cot -. At the same time, however, because the string cottage occurs in English as a separate word, an alternative morphological segmentation of marcottage is into the prefix mar and the root cottage ; stress in marcottage can therefore also be penultimate. Notice finally that of the 270 orthographically non-identical instances of the variation categories “u/ a,” “u/ 4,” and “u/ p/ a,” 140 items (~51.85 %) have no attestations in the BNC / 51 items (~18.89 %) occur in Speake & LaFlaur’s (2002) dictionary of foreign terms. The bulk of these categories is constituted by loanwords that are synchronically still felt to be foreign words (i.e., these words can thus be regarded as poorly-institutionalized in contemporary English, which is reflected in their low frequency of use). An example is the Russian loanword samovar , “[a] highly decorated tea urn used in Russia” ( OD ), which, in addition to being pronounced with regular antepenultimate stress as / ˈsaməvɑː/ ( OD ), abiding by the segmental length principle ( sa . mo - + var ), is on some occasions also stressed finally, as / ˌsaməˈvɑː/ ( OD ). Given this stress pattern, it is important to emphasize that final stress in samovar in the target language English is not necessarily the preserved final stress of самовар in the source language Russian. Since final stress is in general “a stress pattern foreign to English” (Poldauf 1984: 77), pronouncing samovar / ˌsaməˈvɑː/ may simply be a means of marking the foreignness of the word samovar , its low degree of institutionalization in Present-day English. Similarly, when a Present-day English speaker pronounces Ramadan (which occurs in the BNC 75 times) / ˌraməˈdɑːn/ (OD), instead of using the regular stress pattern / ˈramədan/ ( OD ), this is mainly because the concept signified by Ramadan —“The ninth month of the Muslim year, during which strict fasting is observed from dawn to sunset” ( OD )—is alien to contemporary British and American societies (i.e., Muslims do not constitute the majority of these societies). 5.1.3.4 Other categories Of the 432 instances of the variation categories “a/ 4,” “p/ 4,” “a/ 5,” “4/ 5,” “4/ 6,” “p/ a/ 4,” “p/ 5,” “a/ 6,” “5/ 6,” “a/ 4/ 5,” and “p/ 6,” 98 (~22.69 %) are ory and ary -words. E.g., regulatory is in British English interchangeably stressed / ˈrɛɡjᵿlət(ə)ri/ and / ˌrɛɡjᵿˈleɪt(ə)ri/ ( OED ), preserving either the primary or secondary stress of the more frequently used word regulation : / ˌreɡjəˈleɪʃən/ ( LDOCE ). (It is tempting 5.1 Stress variation in the OED 161 to say that initial stress in regulatory is also the preserved stress of regulate , but note that regulatory has 1,188 attestations in the BNC whereas regulate occurs only 641 times. The derivation of regulatory from regulation , which has 2,569 attestations in the BNC , is thus much more probable.) Similarly, applicatory is in British English interchangeably stressed / ˈaplᵻkət(ə)ri/ , / əˈplɪkət(ə)ri/ , or / ˌaplᵻˈkeɪt(ə)ri/ ( OED ). On the one hand, because applicatory does not have attestations in the BNC whereas application occurs in the corpus 9,937 times, one way of stressing the former is to preserve the primary or / and secondary stress of the latter. Thus, because application is stressed / ˌæplɪˈkeɪʃən/ ( LDOCE ), applicatory can be stressed / ˈæplᵻkət(ə)ri/ or / ˌæplᵻˈkeɪt(ə)ri/ , with the former pronunciation promoting the secondary stress of application to primary stress and de-stressing the primary-stressed syllable / ˈkeɪ/ (→ / kə/ , which is because the string atory occurs in English as a suffix (e.g., accusatory ) and should thus be stress-repellent). At the same time, however, because apart from the plicatory -word applicatory , there is also the plicatory -word explicatory , the morphological structure of applicatory may also be ap - + plicatory , with the location of stress in / əˈplɪkət(ə)ri/ being thus the root-prefix boundary location. The second most frequent type is 40 (~9.26 %) words ending in (especially) able , uble , and ible . Consider, e.g., the plicable -adjectives applicable and explicable , both of which vacillate between initial and pen-initial stress. The stress pattern / -ˈplɪ-/ is due to the analysis of the shared string plicable as the root of these adjectives, whereas especially the stress pattern / ˈæp-/ of applicable is due to the fact that the just named semantically related word application is considerably more frequent in the BNC than both applicable and applicant : 9,937 vs. 1,410 and 1,230. Accordingly, because the primary stress of the more frequently used application cannot be preserved in the less frequently used applicable and applicant , for this would mean placing stress upon the suffixes able and ant (note that in / ˌæplᵻˈkeɪt(ə)ri/ and / ˌrɛɡjᵿˈleɪt(ə)ri/ , the preservation of the primary stress of the more frequently used words application and regulation segments the less frequently used words applicatory and regulatory into the disyllabic strings appli -/ regu and the trisyllabic catory / latory , which is in accordance with the segmental length principle), applicable and applicant are pronounced / ˈæplɪkəbəl/ and / ˈæplɪkənt/ ( LDOCE ), with the secondary stress of application being promoted to primary stress. Similarly, although explicable is more frequent in the BNC than both the longer noun explication and the shorter verb explicate (131 vs. 32 / 30), it is considerably less frequent in the corpus than the semantically related noun explanation , which occurs 4,622 times. Accordingly, the back-derivation analysis can also be invoked to account for the stress pattern / ˈɛks-/ of explicable , i.e., initial stress in this adjective is the promoted secondary stress of / ˌekspləˈneɪʃən/ ( LDOCE ). 162 5 Case studies Fairly numerous are also interchangeably-stressed ive -words: 33 / 432 (=~7.64 %). E.g., occupative is / ˌɒkjᵿˈpeɪtɪv/ vs. / ˈɒkjᵿpətɪv/ in British English and only / ˌɑkjəˈpeɪdɪv/ in American English. Since occupative does not have attestations in the BNC while occupy and occupation occur in the corpus 1,029 and 2,212 times respectively, the two stress patterns of the former can be regarded as the preserved stress of the latter: / ˈɒkjəpaɪ/ and / ˌɒkjəˈpeɪʃən/ ( LDOCE ). A different case is indicative , which is / ɪnˈdɪkətɪv/ vs. / ˈɪndɪkeɪtɪv/ according to the OED (but only / ɪnˈdɪkətɪv/ according to LDOCE ; note also that for justificative , the OD gives the transcription / ˈdʒʌstᵻfᵻˌkeɪtɪv/ , where the penult / ˌkeɪ/ is marked as bearing secondary stress. Given this fact, we assume that also the penult / keɪ/ of / ˈɪndɪkeɪtɪv/ bears stress as well). Antepenultimate stress in indicative is most likely due to the prefixation analysis in - + dicative (cf. predicative , which occurs in the BNC 89 times), whereas in the case of occupative , the prefixation analysis oc - + cupative is far less intuitive because the only other English cupative -word nuncupative has only two attestations in the corpus. The initially-stressed / ˈɪndɪkeɪtɪv/ is, by contrast, the promoted secondary stress of indication , which is considerably more frequent in the BNC than indicative : 2,315 vs. 518. (Note also that of the 344 ative -words in the OD , only 31 (~9.01 %) are / -eɪtɪv/ -words, such as cultivative , renovative , motivative , etc. Because the string ative occurs in English as a suffix (e.g., talkative ) and is therefore stress-repellent, the usual (~89.83 %) phonetic realization of the orthographic string ative is / -ətɪv/ , with the penult containing a reduced vowel. E.g., in contrast to American English speakers, who pronounce innovative / ˈɪnəˌveɪtɪv/ ( LDOCE ), preserving the stresses of the more frequently used / ˌɪnəˈveɪʃən/ (LDOCE), British English speakers use the pronunciation / ˈɪnəvətɪv/ (LDOCE), in which the penultimate syllable contains a reduced vowel. 27 ative words in LDOCE are words such as innovative ( educative , imitative , meditative , penetrative , speculative , etc.), which are pronounced / -ətɪv/ by British English speakers vs. / -eɪtɪv/ by American English speakers, but do notice that some of these words are interchangeably pronounced / -ətɪv/ and / -eɪtɪv/ by American English speakers (and in the case of the above mentioned occupative , it is British English speakers who vacillate between the segmental structures / -ətɪv/ and / -eɪtɪv/ ). E.g., operative is only / ˈɒpərətɪv/ in British English, but American English speakers vacillate between / ˈɑːpərətɪv/ and / ˈɑːpəreɪtɪv/ ( LDOCE ); the same is true of decorative , which is only / ˈdekərətɪv/ in British English, but it is interchangeably / ˈdekərətɪv/ and / ˈdekəreɪtɪv/ in American English ( LDOCE ). As both the pronunciations / -ətɪv/ and / -eɪtɪv/ are in accordance with the stress assignment principles that were established in the previous chapter, we argue that the choice between the segmental structures / -ətɪv/ and / -eɪtɪv/ is entirely idio- 5.1 Stress variation in the OED 163 syncratic, i.e., there is no particular reason why the pronunciation / ˈɑːpərətɪv/ should be preferred to the pronunciation / ˈɑːpəreɪtɪv/ , or vice versa.) An interesting case is the variation / ˈænɪkdəʊtɪst/ vs. / əˈnɛkdətɪst/ of anecdotist ( OED ), which is the only ecdotist -word in English. The location of stress in / əˈnɛkdətɪst/ cannot thus be analyzed as the root-prefix boundary location. To account for this stress pattern, we can, however, invoke the vowel effect that, as pointed out in 5.1.3.1, is responsible for penultimate stress in combining forms such as metallo -. The trisyllabic string anecdo ends orthographically in o and should therefore be stressed penultimately rather than antepenultimately. Likewise, the stress patterns / kənˈtrɛərɪnɪs/ of contrariness , / hiːˈdʒɛmənɪ/ of hegemony , / mᵻˈdɪkəm(ə)nt/ of medicament , etc. ( OED ) can be attributed to the fact that the trisyllabic strings contrari -, hegemo -, and medica end orthographically in i / o / a and should therefore receive penultimate stress (and the righthand monosyllabic strings ness , ny , and ment are stress-repellent). Note also that because the base Berkeley ends orthographically in y , it can only be stressed / ˈbɜːkli/ ( LDOCE ). The derivative Berkelium , whose lefthand component Berkeli ends orthographically in i , can, by contrast, be stressed not only / ˈbɜːklɪəm/ , preserving the stress of Berkeley , but also / bəˈkiːlɪəm/ ( OED ). 5.1.4 Left-/ right-prominence vs. word stress As observed in 5.1.1, kilometer , whose segmentation into the components kilo and meter is completely transparent both formally and semantically, is in contemporary English more frequently stressed / kɪˈlɒmɪtə(r)/ than / ˈkɪləmiːtə(r)/ . The “incorrect” stress pattern is due to the fact that of the 254 items in the OD that end orthographically in meter , 181 items (~71.26 %) end in ometer (e.g., accelerometer , aerometer , barometer , chronometer , electrometer , manometer , thermometer , etc.), 14 of which end in lometer : alcoholometer , bolometer , ceilometer , cephalometer , cyclometer , intervalometer , milometer , etc. The alternative morphological segmentation of kilometer is thus into the monosyllabic prefix ki and the trisyllabic root lometer . Likewise, altimeter is, according to LDOCE , stressed / ˈæltɪˌmiːtə/ in British English and / ælˈtɪmɪtər/ in American English. Given this variation, it is very much possible that at some point in the future, the stress pattern / -ˈtɪmɪtər/ will also be exhibited by centimeter , for which, as pointed out above, LDOCE currently gives only the “correct” stress pattern / ˈsentəˌmiːtə/ . A very similar case is methylamine (← methyl + amine ), one of whose stress patterns is / mᵻˈθʌɪləmiːn/ (OED; only British English) because apart from occurring in the word methylamine , the string thylamine also occurs in the words ethylamine and naphthylamine . The stress pattern / mᵻˈθʌɪləmiːn/ is thus, again, 164 5 Case studies due to the “incorrect” prefixation analysis: me - + thylamine . At the same time, however, since both the lefthand component methyl and the righthand component amine occur in English as separate words, the “correct” segmentation methyl + amine still remains a plausible morphological analysis of methylamine , which is therefore often (and in American English exclusively) accented as a compound: / ˈmɛθɪl-/ vs. / -ˈeɪmiːn/ ( OED ), with either the stressed syllable of the lefthand component methyl being made more prominent than the stressed syllable of the righthand component amine , or vice versa. An interesting case is the variably-stressed theocrasy , which, similar to democracy , was formed in English under the influence of the Greek compound θεοκρᾱσία ← θεός , “god,” + κρᾶσις , “mingling” ( OED ). With regard to stress placement, theocrasy vacillates between the pronunciations / ˈθiːəʊˌkreɪsɪ/ and / θiːˈɒkrəsɪ/ ( OED ). In the former case, theocrasy is pronounced as a prototypical (left-prominent) compound, with the initial syllable of the lefthand disyllabic component the . o being made more prominent than the initial syllable of the righthand disyllabic component cra . sy . The alternative pronunciation / θiːˈɒkrəsɪ/ is, by contrast, due to the “incorrect” segmentation of theocrasy into the prefix the and the root ocrasy (cf. idiocrasy ). Notice, however, that of the 56 cracy -words that occur in the OD dictionary, 53 (~94.64 %) are ocracy -words, i.e., an English word that ends in cracy thus almost always has the ending ocracy : autocracy , gerontocracy , meritocracy , ochlocracy , plutocracy , technocracy , etc. By contrast, the dictionary has only three crasy -words: acrasy , idiocrasy , and idiosyncrasy , of which only one word— idiocrasy —is an ocrasyword. The “incorrect” prefixation analysis makes therefore more sense in the case of democracy , which is stressed only / dɪˈmɒkrəsi/ ( OD ), not respecting its “correct” segmentation into demo and cracy . Note also that according to the tool Morphological Analysis, democracy is not demo - + cracy but democrat + cy . Thus, because the righthand string cy is a suffix and thus repels stress, stress in democracy should occur upon the penultimate syllable of its remaining lefthand string democra -, which ends orthographically in a . (The same is true of theocra of theocrasy .) A similar formation oligarchy , for which Morphological Analysis returns the segmentation oligarch + y , is, by contrast, stressed / ˈɒlɪɡɑːki/ ( OD ), preserving the stress of its base / ˈɒlɪɡɑːk/ (OD), because the lefthand trisyllabic string oligar -, which is the string eligible to receive stress in oligarchy , ends orthographically in r (i.e., the vowel effect that applies to democra of democracy does not apply to oligar of oligarchy ). Noteworthy is also the fact that of the 413 lefthand combining forms in the OD dictionary, 290 (~70.22 %) end orthographically in o (34 end in i and 22 in a ). As pointed out in 5.1.3.1, a trisyllabic combining form that ends orthographically in o often has a penultimately-stressed alternative even when from a 5.1 Stress variation in the OED 165 morphological point of view, stress in it should be antepenultimate. Thus, for instance, analyzing the stress variation / kɑːˈsɪnədʒ(ə)n/ vs. / ˈkɑːsᵻnədʒ(ə)n/ of carcinogen ( OED ; both in British and American English), we note that the segmentation carcino - + gen is fully transparent both formally and semantically (cf. antigen ). Stronger stress in carcinogen should thus be the antepenultimate stress of the lefthand combining form carcino -, which means “cancer.” At the same time, however, because the last orthographic vowel in which carcino ends is o , carcinogen is in contemporary English usually stressed / kɑːˈsɪnədʒ(ə)n/ . Likewise, stress patterns such as / əˈkɛːrᵻsʌɪd/ of acaricide , / akˈtɪnəlʌɪt/ of actinolite , / əˈdɛnᵻfɔːm/ of adeniform , / ænˈθrəʊpəʊmɔːf/ of anthropomorph , / kəˈmɛləpɑːd/ of camelopard , / hɪˈriːsɪɑːk/ of heresiarch , / hʌɪˈpɛrənɪm/ of hyperonym , / ɒˈstrakədəːm/ of ostracoderm , etc. ( OED ) can be attributed to the fact that the lefthand components acari -, actino -, adeni -, anthropo -, camelo -, heresi -, hypero -, and ostraco end orthographically in i / o . E.g., to account for the fact that in addition to being stressed / ˈkæmɪləʊˌpɑːd/ ( OED ), camelopard is also stressed / kəˈmɛləpɑːd/ ( OED ), we cannot invoke the prefixation analysis ca - + melopard : No other English word apart from camelopard contains the righthand trisyllabic string elopard . The location of stress in / kəˈmɛləpɑːd/ is thus not the root-prefix boundary location. The vowel effect (i.e., the orthographic ending o in the lefthand component camelo -) does, by contrast, provide an explanation for this stress pattern. Consider also idempotence , for which, in addition to the right-prominent pronunciation ˌ idem ˈ potence , the OED also gives the antepenultimately-stressed transcription / ʌɪˈdɛmpət(ə)ns/ (both British and American English). Because no other English word apart from idempotence contains the trisyllabic righthand string empotence , the location of stress in / ʌɪˈdɛmpət(ə)ns/ is, again, not the root-prefix boundary location. To account for this stress pattern, we can, however, assume that for some English speakers, the morphological structure of idempotence is idempo - + tence (rather than idem + potence ), of which the righthand monosyllabic string tence is stress-repellent and the lefthand monosyllabic string idempo ends orthographically in o and receives therefore penultimate stress. Similarly, experience , which is the only erience -word in the OD , is stressed / ɪkˈspɪərɪəns/ ( OD ), because the righthand monosyllabic string ence is like a suffix and thus repels stress and the remaining lefthand trisyllabic string experi ends orthographically in i . An example involving a tetrasyllabic combining form is periodogram . In addition to being stressed / ˈpɪriədəˌɡræm/ (OED; only American English), preserving the stress of period , periodogram is also stressed / ˌpɪərɪˈɒdəɡram/ ( OED ; both in British and American English), with the tetrasyllabic lefthand combining form periodo -, which ends orthographically in o , receiving penultimate stress. 166 5 Case studies Similar examples that involve penultimate stress as an alternative to (compound-like) left-/ right-prominence are monophthong and polyptych . The morphological structure of these words is, quite obviously, mono - + phthong (cf. diphthong ) and poly - + ptych (cf. triptych ). Abidance by these segmentations results in the left-prominent pronunciations / ˈmɑnə(f)ˌθɔŋ/ and / ˈpɑlᵻpˌtɪk/ (OED), with the stressed syllables of their lefthand components mo . no -, and po . ly being made more prominent than the righthand monosyllabic components phthong and ptych . Notice, however, that according to the OD , these righthand components do not occur in English as separate words. These strings are therefore more like suffixes so that stress in the trisyllabic suffixed derivatives monophthong and polyptych should, in accordance with the segmental length principle, fall upon the segmentally longer ults of their lefthand disyllabic strings mo . noph and po . lyp -. The alternative stress patterns are therefore / məˈnɑpθɔŋ/ and / pəˈlɪptᵻk/ ( OED ). Finally, observe that many English abbreviations are interchangeably pronounced as alphabetisms and acronyms. For instance, ABTA stands in a UK -related context for “Association of British Travel Agents” and in a U. S.-related context—for, e.g., “American Brain Tumor Association.” The two possible pronunciations of this abbreviation are / ˌeɪˌbiˌtiˈeɪ/ and / ˈabtə/ or / ˈæbdə/ ( OED ). In the latter case, stress is initial because the penult has the same segmental length as the ult/ ABTA ends orthographically in a , whereas in the former case stress is final in accordance with the Nuclear Stress Rule (Chomsky & Halle 1968: 94), i.e., when individual words are concatenated into phrases, the strongest accent is placed upon the rightmost primary-stressed syllable. Thus, in the underlying phrase Association of British Travel Agents or American Brain Tumor Association the strongest accents are placed on the stressed syllables of the rightmost words agents and association . The right-prominent accentuation / ˌeɪˌbiˌtiˈeɪ/ is thus the mere preservation of the corresponding accent contour of the underlying phrases Association of British Travel Agents or American Brain Tumor Association . (Note also that in / ˌeɪˌbiˌtiˈeɪ/ , each syllable, which corresponds to a particular word in an underlying phrase, bears stress. This is a manifestation of what Torsuev (1960: 5, 6) called the semantic factor of English stress.) 5.1.5 Stresslessness The term “stresslessness” refers here to an OED transcription that does not contain the primary stress symbol (ˈ). This is mainly typical of words borrowed into English from French and languages alike, where there is no word stress. The point here is that “[t]hose English people who speak foreign languages well often use the foreign pronunciations, or approximations to them, in the 5.1 Stress variation in the OED 167 modern foreign words which are used in English” ( Jones 1917: xxxix). According to Jones (1917: xxxix), “the languages which are sufficiently well known to some English people […] are French, German, Italian, Spanish, Russian, and Hindustani,” of which only “French and Hindustani […] make no use of lexically significant word-stress” ( Jones 1917: xl). As explained in 2.4, a language such as French is different from a language such as English mainly with regard to the phenomenon of vowel reduction. Unstressed syllables in English often contain reduced vowels, whereas the nuclei of syllables in French are usually filled by full rather than by reduced vowels. E.g., sabotage is according to the OED pronounced either / ˈsæbətɑːʒ/ , with the vowel in the unstressed penult being (a qualitatively reduced) schwa, or / sabɔtaʒ/ , with each of the three syllables containing a full vowel. For an audio illustration of a stressless pronunciation, the reader might consider how department is pronounced by a British English speaker in Which seaport is the capital of the Gironde department ? at 00: 27: 23.880 --> 00: 27: 28.880 of the YouTube video ci JTN dy2HxI (09. 04. 2017); this video contains four occurrences of department in France-related geographical contexts, one of which ( of the Bouches-du-Rhone department , at 00: 27: 08.200 --> 00: 27: 10.400) is an instance of the English pronunciation / dɪˈpɑːtm(ə)nt/ ( OD ), while the other three are imitations of how the word is pronounced in French. 5.1.6 Hiatus resolution Finally, there are words in which hiatus resolution is responsible for a change in the location of stress. To be more precise, in both / ˈlaʊʃən/ and / lɑːˈəʊʃən/ of Laotian ( OED ), stress is penultimate, but while in the disyllabic pronunciation / ˈlaʊʃən/ the stressed penultimate syllable is / ˈlaʊ/ , it is / ˈəʊ/ in the trisyllabic pronunciation / lɑːˈəʊʃən/ . Similarly, in both / dʒɪˈɒɡrəfi/ and / ˈdʒɒɡrəfi/ of geography ( OED ; only British English), stress is antepenultimate, but while in the tetrasyllabic pronunciation / dʒɪˈɒɡrəfi/ the stressed antepenultimate syllable is / ˈɒ/ , in the trisyllabic alternative / ˈdʒɒɡrəfi/ the stressed antepenult is / ˈdʒɒ/ . As one can notice, in / ˈdʒɒɡrəfi/ , the hiatus / ɪˈɒ/ of / dʒɪˈɒɡrəfi/ is resolved via deleting the preceding vowel / ɪ/ , whereas / ˈlaʊʃən/ involves both vowel deletion and diphthongization as hiatus-resolution strategies, i.e., to resolve the hiatus in / lɑːˈəʊʃən/ , we first need to delete the following vowel / ˈə/ and then to diphthongize the remaining vowel sequence / ɑːˈʊ/ . Cases of hiatus resolution such as these raise the question of when a hiatus sequence in an English word is likely to be resolved vs. to be retained. (Note that in this section, we are only concerned with hiatus sequences such as / ɪ-ˈɒ/ of geography , which involve an unstressed vowel being immediately 168 5 Case studies followed by a stressed one. Other cases of hiatus resolution—e.g, / di-ˈmen(t)-shə/ vs. / di-ˈmen(t)-shē-ə/ of dementia ( MWO ), with hiatus being made up of two unstressed vowels—go beyond the scope of the present monograph and are therefore not taken into consideration.) In a fairly recent study of how hiatus sequences of rising sonority such as, e.g., i . a of Maria are dealt with in different Romance languages, it is suggested that one of the answers to this question is the location of a hiatus sequence in a word: It is known that diphthongs and hiatus sequences differ in terms of duration […]. In principle, we expect that heterosyllabic sequences will tend to be recategorised as diphthongs in those positions where they tend to be realised with shorter duration. Conversely, we hypothesise that the initiality and stress conditions that have been identified for the preservation of hiatus in Spanish may relate to the fact that in those positions sequences are realised with greater duration, reducing the likelihood of their recategorisation as diphthongs. (Chitoran & Hualde 2007: 48) In other words: the reason for historical reduction to diphthong sometimes being blocked in word-initial position is because [hiatus] sequences in this position tend to be longer than in other positions… Word-initial sequences are separated from the left word boundary by one consonant (Chitoran & Hualde 2007: 48-49) Proceeding from this, it was established by the author that in the MWO dictionary, syllabified phonetic transcriptions containing at least one hiatus sequence such as / ɪ-ˈɒ/ of geography are provided for 1,046 solidly-spelled polysyllabic English words. In the case of 973 of these words, a hiatus-containing transcription is either the sole transcription (e.g., fiasco is only / fē-ˈas-(ˌ)kō/ , with the orthographic hiatus sequence i . a being also phonetically realized as hiatus) or a hiatus-involving pronunciation is given before a hiatus-free alternative (e.g., Laotian is / lā-ˈō-shən/ vs. / ˈlau̇ -shən/ ). As for the other 73 words, a hiatus-free transcription is, by contrast, placed in the dictionary before a hiatus-containing one. E.g., in the case of extraordinary , the hiatus-free transcription / ik-ˈstrȯr-də-ˌner-ē/ is given before the hiatus-involving / ˌek-strə-ˈȯrdə-ˌner-ē/ (proceeding from which we assume that extraordinary is more frequently pronounced / ɛkˈstrɔːdɪnərɪ/ than / ɛkstrəˈɔːdɪnərɪ/ , with the hiatus / əˈɔː/ of the latter being resolved in the former via deleting the preceding vowel / ə/ ). (The fact that of the 1,046 words that contain hiatus sequences such as / ɪ-ˈɒ/ of geography , only 73 (~6.98 %) are more frequently pronounced in a hiatus-resolving manner strongly suggests that hiatus sequences such as / ɪ-ˈɒ/ of geography are as a rule retained in contemporary English.) 5.1 Stress variation in the OED 169 What is important is that of the 973 words such as fiasco and Laotian , 368 (~37.82 %) are words such as fiasco and Laotian , in which hiatus occurs initially (i.e., at the first-second syllable boundary). By contrast, in the case of the 73 words such as extraordinary , initial hiatus occurs in no more than eight words: ~10.96 %. E.g., koala is / kə-ˈwä-lə/ vs. / kō-ˈälə/ , with the former pronunciation resolving the initial hiatus o . a by means of so-called intrusive w , whereas pya is / pyä/ vs. / pē-ˈä/ , with the initial hiatus y . a being resolved in the former via replacing the preceding vowel y through the phonetically similar glide / j/ . (The OD gives for pya , “[a] monetary unit of Burma,” only the monosyllabic transcription / pjɑː/ .) Since the difference of 368 / 973 vs. eight / 73 is statistically very significant— χ 2 (1) = 21.281, p = 0.000004—we are justified in concluding that (also in the English language) initiality inhibits hiatus resolution. E.g., Laotian is more frequently / lɑːˈəʊʃən/ than / ˈlaʊʃən/ because the hiatus sequence a . o occurs in the word initially; its resolution is therefore inhibited by the initiality effect. In extraordinary , by contrast, the very same hiatus sequence occurs post-initially; the word is therefore more frequently pronounced / -ˈstrɔː-/ than / -əˈɔː-/ . Likewise, it is only British English speakers who frequently pronounce geography / ˈdʒɒɡrəfi/ , whereas in American English, the only pronunciation used is still the hiatus-involving / dʒiˈɑɡrəfi/ ( OED ). As in the case of Laotian , we can argue that since hiatus in geography is initial, its resolution should be inhibited by the very same initiality effect. Note also that many variably-stressed English words have hiatus resolution only accompanying a stress variation, without, however, being directly responsible for it. E.g., octoate is / ˈɒktəʊeɪt/ vs. / ɒkˈtəʊeɪt/ in British English and / ˈɑktəˌweɪt/ or / ˈɑkdəˌweɪt/ vs. / ˌɑkˈtoʊət/ in American English ( OED ). As one can see, when stress in octoate in American English is antepenultimate, the hiatus / ʊ-ə/ of the penultimately-stressed version / ˌɑk-ˈtoʊ-ət/ is resolved via replacing the preceding vowel / ʊ/ through the phonetically similar glide / w/ . This, however, is clearly not the cause of the stress variation in octoate , which is strongly supported by the fact that the British English transcriptions / ˈɒktəʊeɪt/ and / ɒkˈtəʊeɪt/ differ from each other only with regard to the location of the stress symbol. What accounts for the stress variation in octoate is that on the one hand, the segmental length principle requires initial stress in its lefthand disyllabic string oc . to -, whereas the segmentation oc - + toate (cf. pantoate , whose meaning is “[a] salt or ester, or the anion, of pantoic acid” ( OD ), whereas octoate likewise has a chemistry-related meaning “[a] salt or ester of an octoic acid”) results in penultimate stress, with the location of stress being the root-prefix boundary location. 170 5 Case studies Note also that laryngeal is, according to the MWO , / lə-ˈrin-jəl also -jē-əl; ˌla-rən-ˈjē-əl/ , with the hiatus e . a being resolved only when laryngeal is stressed la ˈ ryngeal . Likewise, fantasia is / fan-ˈtā-zhə, -z(h)ē-ə; ˌfan-tə-ˈzē-ə/ (MWO), with the hiatus i . a being, again, resolved only when stress in fantasia is antepenultimate. The explanation for these facts is that under both the hiatus-free trisyllabic syllabifications la . ryn . geal / fan . ta . sia and the hiatus-involving tetrasyllabic syllabifications la . ryn . ge . al / fan . ta . si . a , the analysis of the strings ryngeal and tasia as the roots of laryngeal and fantasia is in accordance with the segmental length principle (i.e., these strings are longer than the lefthand strings la -/ fan -). The monosyllabic strings geal and sia cannot, by contrast, be the roots of the trisyllabic la . ryn . geal and fan . ta . sia . To count morphologically as the root, these strings should be syllabified ge . al and si . a ; the tetrasyllabic la . ryn . ge . al / fan . ta . si . a will then be made up of the disyllabic prefixes la . ryn -/ fan . ta and the disyllabic roots ge . al / si . a (cf. laryn - + gitis of the above mentioned tetrasyllabic laryngitis , whose root is also disyllabic). Cf. Caribbean / pharyngeal and Aristotelean , of which the former are still more frequently stressed / -ˈbiːən/ and / -ˈdʒiːəl/ than / -ˈrɪbɪən/ and / -ˈrɪn(d)ʒɪəl/ , whereas the latter is more frequently stressed / -ˈtiːlɪən/ than / -ˈliːən/ . Because the 2,373 ean / ian / eal / ial -words in the OD dictionary contain on average ~4.54 orthographic vowels, with 1,089 (~45.89 %) words containing no less than five orthographic vowels, stress in such words should naturally be antepenultimate rather than penultimate, i.e., the latter can be a stress pattern in a tetrasyllabic word such as Caribbean and pharyngeal , but a hexasyllabic word such as Aristotelean should not be pronounced / ˌærɪstɒtɪˈliːən/ (preserving the stress of the Latin base Aristotelēus ( OED ), in which stress is penultimate because the vowel in the penult is long): The disyllabic string le . an is segmentally too short to count morphologically as its root. 5.2 Stress variation in YouTube Finally, using YouTube data, the chapter will attempt to answer the question of why in the case of English words with stress doublets (in the OED ), one stress pattern is as a rule more frequently used by contemporary English speakers than an alternative stress pattern. To avoid repetitions, the focus of this section will be on those words whose preferred stress patterns run counter to the tendencies that were established in Chapter 4. E.g., because trisyllables that end orthographically in a / i / o in general prefer penultimate to antepenultimate stress, the finding that paprika is more frequently stressed / pəˈpriːkə/ than / ˈpæprɪkə/ is not particularly interesting. Why is, however, enema —in spite of its orthographic ending—in Present-day English no longer stressed / ɛˈniːmə/ ? 5.2 Stress variation in YouTube 171 Since the focus of the monograph is on location-of-stress variation, cases such as / məˈˌnɪŋɡoʊ-/ of meningococcic and / ˈˌædʒədənt-/ of adjutant general , which exemplify degree-of-stress variation (i.e., when components of a morphologically complex word / combination of words occur with varying degrees of stress) will in the following be left out of consideration. Likewise, cases such as / ˈsæbətɑːʒ/ vs. / sabɔtaʒ/ of sabotage , when a pronunciation involving stress co-exists with a stressless alternative, imitating how a word is pronounced in a foreign language, are of relatively little theoretical interest and will therefore not be discussed. As mentioned in 1.2, words with stress doublets are sometimes represented by low frequency words. E.g., the spoken occurrences of the word Achillean , which is / əˈkɪlɪən/ vs. / əkɪˈliən/ in British English and / əˈkɪliən/ vs. / ˌækəˈliən/ in American English ( OED ), were attested only twice in the YouTube videos 0 WH 0 AXMRG q4 (04. 09. 2016) and OX -a BJ 3StUs (04. 09. 2016), in both of which it is pronounced by American English speakers with antepenultimate stress, i.e., / ə-ˈkɪ-li-ən/ . Cf. Euclidean , which was found to have been pronounced by 14 (predominantly American) native English speakers / juː-ˈklɪ-dɪ-ən/ ; the penultimately-stressed alternative / juː-klɪ-ˈdiː-ən/ , which is also given for Euclidean in the OED , was not attested. Needless to say, it is Euclidean rather than Achillean in the case of which we are more justified in claiming that antepenultimate stress represents the word’s default (or perhaps even sole) stress pattern in contemporary English (even though also in Achillean , stress was found to have been exclusively antepenultimate). To deal with the problem of low frequency words, Kunter (2011: 177), who studied variably-accented Noun + Noun compounds, restricts the data to only those types for which at least seven tokens are available. For within-speaker variability this means that there are seven or more tokens of the same type from a given speaker; for across-speaker variability seven or more tokens have to be available in identical contextual environments from different speakers. (Kunter 2011: 177) In a similar vein, the present monograph will regard only those results as (completely) reliable that were obtained for words pronounced by at least 10 different native English speakers. Another problem is posed by words in the case of which different stress patterns were used by near-identical numbers of (different) speakers. E.g., in the case of the word alveolus , pronunciations with stress falling upon the antepenult vs. the penult were used by 13 vs. 11 (mainly) American English speakers, i.e., the numbers of speakers stressing alveolus / -ˈvi-ə-ləs/ vs. / -ˈoʊ-ləs/ were found to be only insignificantly different from each other. Cf. nucleolus , which, according to the OED , is / ˌnjuːklɪˈəʊləs/ vs. / njuːˈklɪələs/ in British English and 172 5 Case studies only / n(j)uˈkliələs/ in American English. In disagreement with this, however, the numbers of American English speakers pronouncing nucleolus / -ˈkli-ə-ləs/ vs. / -ˈoʊ-ləs/ were found to be eight vs. 19, i.e., penultimate stress in nucleolus (which, according to the OED , should occur only in British English) thus appears to be more frequent in contemporary American English than antepenultimate stress. To distinguish between cases such as alveolus vs. nucleolus , we can count the percentage difference between the numbers of speakers using different stress patterns. The percentage difference between two numbers is the absolute value of the difference between them divided by their arithmetic mean. E.g., the percentage difference between 13 and 11 is (13-11)/ ((13+11)/ 2)=~16.67 % and that between eight and 19 is ~81.48 %. (Note, however, that the percentage difference should be counted only when the two values under comparison are above zero, i.e., when one of the values is zero (which in our case means that one of the stress patterns could not be attested in captioned YouTube videos), the result is always 200 %, irrespective of the other value, which is non-zero. E.g., if the two values are 10 and zero, the percentage difference is (10-0)/ ((10+0)/ 2)=200 %; likewise, if the two values are 200 and zero, the result is also (200-0)/ ((200+0)/ 2)=200 %.) As a threshold, we can (somewhat arbitrarily) choose the percentage difference of 25 % (cf. Balteiro 2007: 128), below which all frequency differences are to be regarded as doubtful. The case of alveolus , with the percentage difference between the numbers of speakers pronouncing it with penultimate vs. antepenultimate stress being only ~16.67 %, counts therefore as doubtful. On the basis of the finding reported above, we cannot then say that antepenultimate stress is the preferred stress pattern of alveolus in contemporary English. Finally, it is observed that the results presented below are biased towards American English because in general, the voices of American English speakers were attested far more often than the voices of speakers featuring other native English accents. Thus, the number of British English speakers pronouncing a particular word with stress doublets was as a rule less than 10 and with regard to other varieties, only the voices of Australian English speakers could fairly often be heard in a comparable number of videos. The fact that the findings of this section take mainly American English into account can be seen as a limitation of this study, but note that this section does not aim to be a lexicographic documentation of how different English varieties stress particular English words. The task of this section is, as declared above, to find out why in the case of many English words with stress doublets, one stress pattern is more frequently used by contemporary English speakers than an alternative stress pattern, and since the stress system of American English is not different from that of British English / any other native English variety, it does not make much difference which Present-day English variety serves as the object of the study. 5.2 Stress variation in YouTube 173 5.2.1 Overall results: YouTube vs. OED As already mentioned in 3.3, for 1,232 items with stress doublets in the OED , at least one spoken occurrence could be attested in captioned YouTube videos, i.e., at least one native English speaker was found to have pronounced these words. Only 644 (~52.27 %) of these items were non-identically-stressed (by either different or identical native English speakers, i.e., acrossvs. within-speaker variation), but observe that if we consider only those words that were pronounced by at least 10 different speakers, the proportion of non-identically-stressed items from the OED rises from ~52.27 % to (390 / 583=)~66.9 %. Thus we can conclude that words with stress doublets in the OED are indeed as a rule variably-stressed by contemporary English speakers. If individual variation categories are taken into consideration, the results are as follows. The variation type “p/ a” is represented by 131 genuine instances of p/ a-instability that were found to have been pronounced by at least 10 different native English speakers. (These words do not include cases such as ˈ adulthood vs. a ˈ dulthood , which can be seen as the inherited variation ˈ adult vs. a ˈ dult ; notice also that (very) similar cases of p/ a-instability, such as, e.g., ˈ affluence vs. afˈfluence and ˈ affluent vs. afˈfluent , are counted by the author only once.) 75 (~57.25 %) of these words were either exclusively or more frequently pronounced with antepenultimate stress (e.g., just like the majority of trisyllabic English words that end orthographically in y , anchovy is more frequently stressed / ˈæntʃəvɪ/ than / ænˈtʃəʊvɪ/ ), whereas the other 56 (~42.75 %) words were, by contrast, found to prefer penultimate stress. E.g., just like the majority of trisyllabic English words that end orthographically in a , angina is more frequently stressed / ænˈdʒaɪnə/ than / ˈændʒɪnə/ . Of 203 instances of u/ p-instability that were pronounced by no less than 10 different native English speakers (with similar words such as, e.g., detail and detailed counting as the same word), 149 (~73.4 %) were either exclusively or more frequently pronounced with initial / penultimate stress, i.e., the category “u/ p” is composed not only of disyllables, such as detail , in which initial stress is as a rule considerably more frequent than final stress, but also of several trisyllables such as, e.g., aikido , which, according to the OED , is in American English interchangeably stressed / aɪˈkidoʊ/ vs. / ˌaɪkiˈdoʊ/ , but of these, only the former stress pattern was heard by the author (i.e., just like the majority of trisyllabic English words that end orthographically in o , aikido prefers penultimate stress). Similar to disyllables that constitute the bulk of the variation type “u/ p,” also trisyllabic instances of the variation category “u/ a” were extremely rarely (~9.23 %) found to prefer final to antepenultimate stress (e.g., in the case of misdiagnose , the stress pattern -ˈ nose was more frequently used than mis ˈ di -); this 174 5 Case studies finding also “point[s] towards a peripheral status of final stress in contemporary English phonology” (Fournier 2007: 232). The variation categories “a/ 4,” “p/ 4,” “a/ 5,” “4/ 5,” “4/ 6,” “p/ a/ 4,” “p/ 5,” “a/ 6,” “5/ 6,” “a/ 4/ 5,” and “p/ 6” are represented by 57 items that were pronounced by at least 10 different native English speakers. (Once again, it is reiterated that similar items such as, e.g., interdisciplinary and multidisciplinary are counted only once). Of these 57 words, 26 (~45.61 %) were either exclusively or more frequently stressed closer to the left word edge. E.g., the already mentioned oxygenate was exclusively stressed / ˈɒksɪdʒəneɪt/ and participle was more frequently stressed / ˈpɑːtᵻsᵻpl/ ( OED ) than / pɑːˈtɪsᵻpl/ ( OED ). In the other 31 words (~54.39 %), stress usually fell closer to the right word edge. E.g., the already mentioned hydrogenate was more frequently stressed / haɪˈdrɒdʒəneɪt/ than / ˈhʌɪdrədʒəneɪt/ and evocative was exclusively stressed / iːˈvɒkətɪv/ ( OED ) and never / ˈɛvəʊkeɪtɪv/ ( OED ). It should be added, however, that instances of these variation categories were often heard to exhibit across-varietal stress differences. E.g., metallurgy is / ˈmɛdlˌərdʒi/ in American English, preserving the stress of the disyllabic noun metal , but British English speakers prefer the stress pattern / mᵻˈtalədʒi/ ( OED ), deriving metallurgy from the penultimately-stressed trisyllabic combining form metallo -, which ends orthographically in o , i.e., metallurgy ← metallo - + urgy , with the hiatus o . u being resolved via deleting the preceding vowel o . Ancillary , corollary , medullary , and pupillary are in American English usually stressed / ˈænsɪlərɪ/ , / ˈkɒrələrɪ/ , / ˈmɛdəlɛri/ , and / ˈpjupəlɛri/ ( OED ), but British English speakers prefer an ˈ cillary , co ˈ rollary , me ˈ dullary , and pu ˈ pillary . Given these systematic stress differences, it is clear that American English “incorrectly” analyses the ary of these words as a suffix, i.e., whereas in the case of pupillary , the segmentation into pupil and ary seems no less correct than into pupilla and ry , corollary and medullary can only be corolla + ry and medulla + ry . The British English stress patterns / kɒˈrɒlərɪ/ and / mɛˈdʌl(ə)ri/ (OED) thus preserve the stress of / kəˈrɒlə/ and / mɛˈdʌlə/ ( OD ), which end orthographically in a and are therefore stressed penultimately. The same is true of / ænˈsɪlə/ of ancilla ( OED ), which is why also ancillary is supposed to be stressed / ænˈsɪlərɪ/ ( OED ). For Americans, ancillary , corollary , and medullary are, however, ancil -/ corol -/ medul - + ary , with the latter repelling stress and the former receiving it upon their penults (i.e., e.g., even if we proceed from the syllabification co . rol -, the segmentally longer ult still does not achieve the segmental length of four segments, which is typically exhibited in English by finally-stressed disyllables. Stress in the string co . rol is therefore initial). Related to these are cases such as / ˈkɪləmiːtə(r)/ vs. / kɪˈlɒmɪtə(r)/ of kilometer , which also involve placing stress closer to the beginning vs. end of a word, 5.2 Stress variation in YouTube 175 respecting vs. non-respecting its compound-like morphological structure. Unfortunately, attestations from at least 10 different speakers were available for only a handful of such words— aristocrat , carcinogen , cuneiform , kilometer , melanocyte , oligomer , orangutan , physiatrist , prothonotary , and spectroscopy — which is not particularly surprising given that formations such as these are as a rule low frequency words. E.g., heresiarch was pronounced only four times in three captioned YouTube videos. The antepenultimately-stressed pronunciation / hɪˈriːsɪɑːk/ occurred in the video x BK 561Ry OUE (25. 01. 2015), whereas in the videos iI2 ON 4hbyT8 (25. 01. 2015) and khObey5 LMI 4 (25. 01. 2015), the initially-stressed pronunciation / ˈhɛrɪsɪˌɑːk/ ( OED ) could be heard. Similarly, for idempotence , only nine spoken occurrences could be attested in the videos 6d VN dFwqeKs (09. 02. 2015) and o3TuRs9 AN hs (09. 02. 2015), in both of which we hear the “correct” right-prominent pronunciation / ˌaɪdɛmˈpoʊtns/ ( OED ). As for the items that were pronounced by at least 10 different speakers, the results are as follows. A ˈ ristocrat was more frequent than ˈ aristocrat , car ˈ cinogen was more frequent than ˈ carcinogen , cu ˈ neiform was more frequent than ˈ cuneiform , kiˈlometer was more frequent than ˈ kilometer , melanocyte was only me ˈ lanocyte , o ˈ ligomer was more frequent than ˈ oligomer , orangutan was only o ˈ rangutan , phy ˈ siatrist was more frequent than physiˈatrist , pro ˈ thonotary was more frequent than protho ˈ notary , and spectroscopy was only spec ˈ troscopy . As suggested in 5.1.4, (preferred) stress patterns such as these are due to 1) “incorrect” prefixation analyses or / and 2) the vowel effect. E.g., physiatrist is morphologically physi - + atrist (cf. paediatrist , psychiatrist ) and should therefore be stressed / ˌfɪzɪˈatrɪst/ ( OED ; both British and American English). However, because the common string shared by physiatrist , paediatrist , and psychiatrist is not atrist but iatrist (and because the pronunciation / ˌfɪzɪˈatrɪst/ contains a hiatus sequence and is thus phonetically unfortunate), the stress pattern / fᵻˈzʌɪətrɪst/ ( OED ; both varieties) is more frequent than the stress pattern / ˌfɪzɪˈatrɪst/ and, similarly, paediatrist is, according to the OD , more frequently stressed / -ˈdʌɪətrɪst/ than / ˌpiːdɪˈatrɪst/ and psychiatrist is exclusively stressed / -ˈkʌɪətrɪst/ , with the (non-initial) hiatus i . a being resolved via diphthongization. The vowel effect is especially obvious in the case of prothonotary , which is the only onotary -word in the BNC ; the stress pattern / prə(ʊ)ˈθɒnət(ə)ri/ ( OED ) is thus hardly attributable to the prefixation analysis pro - + thonotary . Assuming, however, that in addition to segmenting prothonotary into protho and notary , which leads to the “correct” stress pattern / ˌprəʊθəˈnəʊtəri/ ( OED ), contemporary English speakers usually segment it into prothono and tary (of which the former is segmentally longer than the latter and, in addition to this, ends orthographically in o ), it immediately becomes clear why prothonotary is usually stressed / prə(ʊ)ˈθɒnət(ə)ri/ . 176 5 Case studies Now, we can have a closer look at individual stress-assignment factors accounting for stress-assignment preferences in individual variably-stressed English words. 5.2.2 Adjacent words Because “adjacent stressed syllables make speech sound jerky” (Kingdon 1949: 149), the stress pattern of one English word might affect the placement of stress in another (immediately preceding / following) English word. An example is the video koqWmL4uWsI (05. 07. 2015), in which in to serious birth de ˈ fects including cleft lip, cleft palate, and serious cardiac ˈ defects , stress in defects is final in the combination birth de ˈ fects , but it is initial in the combination cardiac ˈ defects , in which, in contrast to the birth of birth defects , the immediately preceding modifier cardiac is not finally-stressed; there being no stress clash when defect is stressed / ˈdiːfɛkt/ ( OD ) (which seems to have a purely semantic justification: To defect , which in accordance with the segmental length principle / synchronic back-derivation from the more frequently used longer nouns defection / defector is stressed / dɪˈfɛkt/ ( OD ), means “[a]bandon one’s country or cause in favor of an opposing one” ( OD ); a defect means, by contrast, “[a] shortcoming, imperfection, or lack” ( OD ). Since the connection between these senses is far from being obvious, stress differences are deliberately employed by English speakers to make a defect formally different from the semantically unrelated to defect ). In general, the spoken occurrences of the combination birth defect / birth defects were attested in captioned YouTube videos 473 times. With initial stress, the head nouns defect and defects are pronounced in these combinations only 168 times (~35.52 %), which means that, in disagreement with Hayes (1995: 370), arguing that stress movement in English is only leftward ( thir ˈ teen ~ ˈ thirteen men ), stress in the head noun defect does move rightward under the influence of the immediately preceding monosyllabic modifier birth . That is, the preferred stress pattern of the noun defect in the combinations birth defect and birth defects is / dɪˈfɛkt/ while in other environments a defect is usually stressed / ˈdiːfɛkt/ (which, as suggested above, is mainly a matter of semantics). A similar example is the adjective compact , which in addition to modifying other English words (especially the monosyllabic noun disc ), is itself often modified by the monosyllable more . The spoken occurrences of more compact were attested in captioned YouTube videos (featuring the voices of native English speakers pronouncing more compact ) 42 times. With initial stress, compact is pronounced in these attestations of more compact only 13 times (~30.95 %), whereas in the other 29 attestations (~69.05 %), stress in compact of more compact is final; an example of within-speaker variability is the video 7hv OGD eqm TQ 5.2 Stress variation in YouTube 177 (09. 04. 2016), in which in a more com ˈ pact interface , stress in compact is final, while it is initial in a more ˈ compact and intuitive interface . Once again, it can be seen that while in general the adjective compact prefers initial stress, because it often occurs as modifier of monosyllabic / initially-stressed polysyllabic nouns (e.g., compact disc ), stress in compact is more frequently final in the combination more compact , in which the immediately preceding modifier more is monosyllabic. Recall also the YouTube video -AnsohxXn QU (17. 09. 2016), in which one and the same American English speaker stresses applicable / ˈæ-/ in This name change will be more functionally ˈ applicable vs. / -ˈplɪ-/ in we’ve made the name change to make it more ap ˈ plicable . Of the 16 spoken occurrences of more applicable in captioned YouTube videos, only four (25 %) were heard to exemplify the initially-stressed pronunciation / ˈæpləkəb(ə)l/ . The usual pronunciation of applicable in more applicable is thus the pen-initially-stressed / əˈplɪkəbl/ , which is better from the point of view of rhythm. (Note also that according to Scherer & Wollmann (1972: 195), one of the factors accounting for the stress pattern / ɪˈlɛv(ə)n/ of eleven ( OD ) is the counting-context, eight , nine , ten , eleven , twelve , with the trisyllabic eleven being immediately preceded by the monosyllabic ten . Pen-initial stress in the former is thus from the point of view of rhythm better than initial stress.) Consider, however, the preferred stress patterns of the nouns incline and decrease in the combinations steep incline , per cent decrease , and slight decrease , in which, just like in the combinations birth defect(s) and more compact , final stress in the nouns incline and decrease would from the point of view of rhythm be better than initial stress, i.e., steep incline should be / ˈstiːp ɪnˈklaɪn/ rather than / ˈstiːp ˈɪnklaɪn/ , where one stressed syllable occurs immediately after another stressed syllable. Contrary to these expectations, however, in 51 spoken occurrences of steep incline , the pronunciation / ɪnˈklaɪn/ was heard by the author no more than two times (~3.92 %). Similarly, in 62 occurrences of decrease in per cent decrease , the pronunciation / dɪˈkriːs/ was likewise heard by the author only two times (~3.23 %) and in 70 occurrences of slight decrease , this pronunciation of decrease was heard by the author only six times (~8.57 %). The reason for this difference seems to be the (by and large) emphatic nature of initial stress in both incline and decrease , which due to obvious reasons is completely missing in compact and defect . That is, in the COCA corpus, steep is not only the most frequent modifier of the noun incline but also the seventh most frequent modifier of the noun decline , and per cent is not only the second most frequent modifier of the noun decrease but also the most frequent modifier of the noun increase (and, similarly, slight is not only the fourth most frequent modifier of decrease but also the fifth most frequent modifier of increase ). By 178 5 Case studies contrast, compact is the only pact -word that in COCA is modified by more , and defect is the only fect -word that in COCA is immediately preceded by the modifier birth . (As for fects , there are only two occurrences of birth effects alongside 692 occurrences of birth defects .) Thus, since steep can be frequently found adjacent to both incline and decline , which express related senses, and since per cent and slight collocate both with decrease and increase , which are semantically opposite, it is quite natural that stress in incline of steep incline and in decrease of per cent decrease and slight decrease be emphatically initial, while in the combinations more compact and birth defect(s) , where compact and defect do not contrast with other words ending in pact / fect , English speakers readily move stress from the penult to the ult, making thereby the pronunciations of the respective phrases more rhythmic. Cf. the combination tax increase , whose spoken occurrences were attested in captioned YouTube videos 178 times. The initially-stressed pronunciation / ˈɪnkriːs/ was heard by the author only 50 times (~28.09 %); the considerably more frequently (~71.91 %) used stress pattern of increase in tax increase is thus the finally-stressed / ɪnˈkriːs/ , which is better from the point of view of rhythm, i.e., / ˈtaks ɪnˈkriːs/ rather than / ˈtaks ˈɪnkriːs/ , which involves a stress clash. Because governments across the globe virtually never decrease taxes but only increase them (which is reflected in the fact that in COCA , the combination tax decrease has only 14 attestations (and, similarly, in captioned YouTube videos it occurs only nine times) whereas the combination tax increase occurs 1,808 times! ), the increase of the combination tax increase does not contrast semantically with decrease ; the emphatic potential inherent in its prefix in is thus very low. The usual stress pattern of increase in this combination is therefore the finally-stressed / ɪnˈkriːs/ . By contrast, of the five American English speakers who were found to have pronounced the combination tax decrease , only one speaker used the finally-stressed version / dɪˈkriːs/ . As just pointed out, because the phenomenon of a government decreasing taxes is extremely rare, the emphatic potential inherent in the prefix de of the noun decrease in the combination tax decrease is very high; this prefix is thus predestined to be emphasized via stress. Recall also the fact, mentioned in 1.2, that according to Halle & Vergnaud (1987: 271, f. 29), “the word-internal application of the Rhythm Rule […] is restricted to lexical compounds,” i.e., “retraction is (almost) obligatory in Marcel Proust , but almost impossible in Marcel left .” Indeed, in YouTube videos that were found to contain the spoken occurrences of Marcel , initial stress in this word was heard by the author only in combinations such as Marcel Proust , but not in genitive phrases such as Marcel’s book or syntactic clauses such as Marcel left . The attested combinations in which Marcel is stressed initially include Marcel Proust (1 HT 7 IW d9 JR c, 08. 06. 2015), Marcel Daniels (N4Yc6 ZPGD qA, 5.2 Stress variation in YouTube 179 30. 03. 2016), Marcel Kittel (9oKqcy-dnoE, 07. 06. 2015), Marcel Kiss (h57TsezhYI8, 07. 06. 2015), and Marcel Mettelsiefen (nxVeYVGrTaI, 07. 06. 2015). In each of these combinations, the immediately following heads Proust , Daniels , Kittel , Kiss , and Mettelsiefen are either monosyllables or initially-stressed polysyllables; hence in the immediately preceding Marcel , stress occurs initially. At the same time, the expected retraction was not observed in the combinations Mar ˈ cel ˈ Salathe (5r WK lN_nz5Y, 07. 06. 2015), Mar ˈ cel ˈ Neergaard (3X4LeUjOU74, 07. 06. 2015), Mar ˈ cel ˈ Irnie (57YKVeu0a1s, 07. 06. 2015), and Mar ˈ cel ˈ Dijkers (V8Pyf91ZsgM, 07. 06. 2015). Likewise, the expected retraction also did not occur in the combinations Marcel iron (pZ3wqLv14dE, 07. 06. 2015) and Marcel wave (2d9QUSMGXo8, 07. 06. 2015), which Halle & Vergnaud (1987) would most likely also regard as lexical compounds rather than as constructions that are “formed in the syntax.” As for the combination Marcel Proust , where retraction is said to be near-obligatory, it did occur in the aforementioned video 1 HT 7 IW d9 JR c (08. 06. 2015), but it did not occur in the videos Dri HMS xaz7A (07. 06. 2015) and R2P9- JPAHN 4 (07. 06. 2015). Summarizing, the Rhythm Rule applied in only five (~45.45 %) distinct types of Noun + Noun combinations in which the modifier Marcel is immediately followed by initially-stressed heads: Marcel Proust , Marcel Daniels, Marcel Kittel , Marcel Kiss , and Marcel Mettelsiefen . In the other six types (~54.55 %)— Marcel Salathe , Marcel Neergaard , Marcel Irnie , Marcel Dijkers , Marcel iron , and Marcel wave —the rule did not apply. The conclusion drawn by Halle & Vergnaud (1987: 270-271, f.29) that in the case of lexical compounds such as Marcel Proust , the retraction of stress in the modifier Marcel is “(almost) obligatory” is thus clearly wrong (even with reference to the combination Marcel Proust itself, which was pronounced with final stress in two of the three videos in which it is used by native English speakers). Similarly, the video Wi0gtM25g6o (07. 06. 2015) is the only video in which the variation between / dʒəˈkɑːd/ and / ˈdʒækəd/ of Jacquard ( OED ) is in accordance with the Rhythm Rule, i.e., an identical American English speaker uses the finally-stressed pronunciation / dʒəˈkɑːd/ in Look at the inside Jacquard (i.e., because inside is stressed / ɪnˈsʌɪd/ ( OD ), final stress in Jacquard in Look at the inside Jacquard is from the point of view of rhythm better than initial stress) whereas in the Jacquard print , stress in Jacquard is initial (i.e., because the following head is a monosyllable, initial stress in the modifier Jacquard in this combination is from the point of view of rhythm better than final stress). In the majority of the attributive uses of Jacquard (e.g., Jacquard loom , Jacquard jacket , Jacquard design , Jacquard bottle , etc.), stress in the modifier remains, however, final. Further similar examples include Bantu , bastille , and batik , which were finally-stressed in the combinations Bantu knot , bastille day , bastille fortress , bastille prisoner , 180 5 Case studies batik memories , batik cloth , batik costume , etc., in which the immediately following head noun is either a monosyllable or an initially-stressed polysyllable. Since, as already mentioned on several occasions, final stress represents a stress pattern foreign to English, items such as Marcel , Jacquard , Bantu , bastille , batik , etc., which are still perceived by English speakers as foreign words / denote things that are inherently foreign, are predominantly pronounced with final stress irrespective of the morphosyntactic environments in which they occur, i.e., it does not matter at all whether Marcel occurs in a combination such as Marcel Proust or in combinations such as Marcel’s book or Marcel left ; in each of these environments Marcel is as a rule stressed / mɑːˈsɛl/ ( OD ), violating the segmental length principle (cf. cancel , which in accordance with this principle is stressed / ˈkans(ə)l/ ), simply because the first name Marcel is a French name. Notice also that the aforementioned videos in which Marcel was found to have been pronounced with final stress all feature the voices of American English speakers, whereas of the five videos in which Marcel is pronounced with initial stress, three videos (those where Marcel occurs in the combinations Marcel Kittel , Marcel Kiss , and Marcel Mettelsiefen ) feature the voices of British English speakers. This finding is in accordance with the oft-mentioned tendency of words borrowed into English from French being more frequently finally-stressed in American than in British English. E.g., ballet is / ˈbaleɪ/ in British English vs. / bæˈleɪ/ in American English ( OED ) and, similarly, galanterie is / ɡəˈlɑːnt(ə)riː/ in British English vs. / ɡəˌlɑn(t)əˈri/ in American English (OED). Recognizing a word as Modern French, an American speaker as a rule considers it necessary for reasons of social prestige to signalize this, and it is obvious that final stress is the most suitable way of doing so. On the other hand, he meets these French words mostly in print and hence easily slips to ‘orthographic pronunciation’, as in [priˈmɪə] for premiere […] On the other hand, a British speaker who has a nodding acquaintance with French tends to adapt his neighbors’ words accentually, while avoiding gross phonetic distortion in a language with which he has some acquaintance. Both these tendencies probably arise from the marked stratification of British society, although this has been greatly reduced in recent decades, and from the relatively great value attached to being thought ‘educated’. (Poldauf 1984: 76-77) Another deviation from the Rhythm Rule is the combination hotel room , in the case of which retraction of stress in the finally-stressed modifier hotel was heard by the author only 21 times (e.g., in 1 EN vL-t1Y3o (11. 04. 2016), in every ˈ hotel room and the common areas of the ho ˈ tel the first hotel is stressed initially whereas in the second one stress is final); these constitute, however, only ~15.91 % of the 132 total spoken occurrences of hotel room in 70 captioned YouTube videos. 5.2 Stress variation in YouTube 181 Likewise, final stress in the modifiers adult , innate , overt , and pristine occurred even in combinations such as adult system , adult court , adult living , adult learning , adult friends , adult horses , adult brains , innate cruelty , innate freedom , innate goodness , innate passion , overt act , overt bribe , overt racism , overt rationing , pristine beaches , pristine state , pristine wilderness , etc., in which the immediately following head is either initially-stressed or monosyllabic. Cases such as these come as no big surprise because, as has been argued throughout this monograph, stress assignment in English is conditioned not phonetically and phonologically but morphologically, i.e., the location of stress in an English word is primarily a reflection of how its morphological structure has been analyzed by a particular English speaker. Thus, if, e.g., overt is morphologically o - + vert rather than over + t (and if the speaker does not wish to emphasize that the thing referred to is overt rather than covert), stress in overt will stay final even in a combination such as overt racism , in which initial stress in the modifier would from the point of view of rhythm be better than final stress. 5.2.3 Vowel effect In contrast to the words / combining forms aceto -, angina , ANOVA , Attila , basmati , cathedra , hibachi , immuno -, incognita / incognito , kanaka , manuka , paprika , patchouli , pavlova , Satsuma , sriracha , and urea , which end orthographically in a / i / o and are therefore exclusively / more frequently pronounced with penultimate stress (e.g., urea is / juːˈriːə/ and / ˈjʊərɪə/ according to the OED , but in 118 captioned YouTube videos in which urea was found to have been pronounced by native English speakers, only the stress pattern / juːˈriːə/ was heard by the author), the a / o -words buddleia , enema , gingiva , rodeo , trachea , and vertigo were, by contrast, more frequently stressed antepenultimately. (Note, however, that in British English, / trəˈkiːə/ is, according to LDOCE , the only stress pattern of trachea (which in American English is usually stressed / ˈtreɪkiːə/ ), and in the case of rodeo , penultimate stress was heard by the author only in names such as Rodeo Drive and California Rodeo Salinas whereas in all other environments, stress in rodeo was exclusively antepenultimate. The variation / ˈrəʊdɪəʊ/ vs. / rə(ʊ)ˈdeɪəʊ/ ( OED ; both in British and American English) thus seems to be a matter of semantics.) As for buddleia , which is more frequently / ˈbʌdliːə/ than / bʌdˈliːə/ ( OED ), the likeliest explanation is the suffixation analysis Buddle + ia . This analysis may be arrived at even if a contemporary English speaker does not remember that buddleia was “named in honor of the English botanist Adam Buddle” ( OD ). Cf. / ˈdeɪlɪə/ of dahlia , which was “named in honor of Andreas Dahl (1751-89), Swedish botanist” ( OD ) or, similarly, / ˈfjuːʃə/ of fuchsia , which was “named in 182 5 Case studies honor of Leonhard Fuchs (1501-66), German botanist.” Simply because the ia word buddleia is not the only botany-related ia -formation in English, the suffixation analysis Buddle + ia , which requires the pronunciation / ˈbʌdliːə/ , immediately suggests itself. As for enema , which in contemporary English is stressed only / ˈɛnɪmə/ , the explanation seems to be its formal similarity to the more frequently used enemy , which ends orthographically in y and is therefore stressed / ˈenəmi/ ( LDOCE ). Thus, as pointed out in 4.3.7, of the 801 trisyllables in LDOCE that end orthographically in y , 709 (~88.51 %) have antepenultimate stress. In trisyllables such as agony , apathy , enemy , energy , irony , etc., stress in English is almost exclusively initial. Finally, it is difficult to explain why antepenultimate stress is the preferred stress pattern of gingiva and vertigo (for which the OED also gives the penultimately-stressed pronunciations / -ˈdʒʌɪvə/ and / -ˈtaɪɡəʊ/ or / -ˈtiːɡəʊ/ , but these did not occur in 43 / 135 spoken occurrences of these words in captioned You- Tube videos). As for vertigo , a possible explanation might be the fact that its etymon is the Latin verb vertere , “to turn” (Dictionary.com), which is also the etymon of the English words verse and versus . Because the semantics of turning is present not only in vertigo but also in (the more frequently used, initially-stressed) versus , “turned so as to face (something), opposite, over against” (Dictionary.com), stress in vertigo should also occur initially. As for gingiva , the word is “the technical name for the gum” (Dictionary.com), i.e., the gums of the mouth. Given that in the monosyllabic gum , stress falls upon the word’s first and only syllable, also the semantically identical trisyllabic gingiva should be stressed / ˈdʒɪn-/ rather than / -ˈdʒʌɪvə/ . Recall also that, as suggested in 5.1.3.1 and 5.1.4, the very same vowel effect, which accounts for penultimate stress in trisyllables such as angina , also accounts for the stress patterns of morphologically complex words such as, e.g., aristocrat (cf. democrat ). Because its lefthand component aristo ends orthographically in o , the preferred stress pattern of aristocrat is not / ˈærɪstəʊˌkræt/ but / əˈrɪstəkræt/ (which cannot be attributed to the prefixation analysis a - + ristocrat : Aristocrat is the only istocrat -word in the BNC ). Likewise, because apart from cuneiform and oligomer , no other English word (in the BNC ) contains the trisyllabic righthand strings eiform / igomer , the stress patterns cu ˈ neiform and o ˈ ligomer are to be regarded as penultimate stress in the trisyllabic lefthand components cunei and oligo -, which end orthographically in i / o . The already mentioned umbilicus is more frequently stressed / -ˈbɪlɪkəs/ than / -ˈlaɪkəs/ . Because the BNC has only one attestation of cochilicus , it would be counterintuitive to attribute the stress pattern / -ˈbɪlɪkəs/ to the prefixation analysis um - + bilicus . Assuming, however, the morphological structure umbi- 5.2 Stress variation in YouTube 183 li - + cus (cf. Copernicus ), we do have an explanation for why umbilicus is usually stressed / -ˈbɪlɪkəs/ . (The alternative morphological analysis umbi - + licus results, by contrast, in the other, less frequently used stress pattern / -ˈlaɪkəs/ .) A similar case is equilibrate , which in contemporary English is usually stressed / iːˈkwɪlɪbreɪt/ ( OED ), not respecting the “correct” segmentation into equi and librate , which should lead to the stress pattern / ˌiːkwɪˈlɪbreɪt/ . Cf. equipotent , for which the OD gives the transcriptions / ˌiːkwɪˈpəʊt(ə)nt/ and / ˌɛkwɪˈpəʊt(ə)nt/ ; the segmentation equi - + potent is thus abided by. Given that equilibrate is more frequently / iːˈkwɪlɪbreɪt/ than / ˌiːkwɪˈlɪbreɪt/ and given that the OD dictionary gives the line break equili|brate, in which the only division point is the ult-penult boundary, we assume that for the majority of contemporary English speakers, equilibrate is not equi - + librate but equili - + brate , of which the latter is stress-repellent (because the string ate occurs in English as a suffix) and the former ends orthographically in i and should therefore receive stress upon its middle syllable. (Similarly, even if emancipate , which occurs in the BNC 18 times, were not a back-derivative from emancipation , which has 463 attestations, the former would nonetheless be stressed / ɪˈmansɪpeɪt/ ( OD ) simply because the lefthand trisyllabic string emanci ends orthographically in i and the righthand monosyllabic string pate is stress-repellent.) Similarly, for equivalent the OD gives the line break equiva|lent, which strongly suggests that morphologically, equivalent is equiva - + lent rather than equi - + valent . The stress pattern of equivalent is therefore / ɪˈkwɪv(ə)l(ə)nt/ ( OD ), with stress being penultimate in the segmentally longer lefthand trisyllabic string equiva -, which ends orthographically in a . Finally, observe that in contrast to the OD , the OED gives for equipotent the transcriptions / iːˈkwɪpətənt/ and / iːkwɪˈpəʊtənt/ , which means that the morphological structure of this adjective may also be equipo - + tent rather than only equi - + potent . In contrast to American English speakers, British English speakers usually pronounce artificer / ɑːˈtɪfᵻsə/ and mercantilist / məːˈkantᵻlɪst/ ( OED ), non-preserving the stress of the corresponding base forms artifice and mercantile . Because the righthand monosyllabic strings cer and list cannot take stress, the remaining lefthand trisyllabic strings artifi and mercanti -, which end orthographically in i , receive penultimate stress. Catholicism is in the OED both / kəˈθɒlɪsɪzəm/ and / ˈkæθəlɪsɪzəm/ , but in 256 spoken occurrences of this derived noun in captioned YouTube videos, the stress pattern / ˈkæθəlɪsɪzəm/ was not heard by the author at least once. To account for the fact that Catholicism does not preserve the stress of the more frequently used catholic (which has 3,936 attestations in the BNC while Catholicism occurs only 333 times), we can again invoke the vowel effect. Because the righthand 184 5 Case studies monosyllabic string cism is stress-repellent, the lefthand trisyllabic string Catholi -, which ends orthographically in i , receives stress upon its penult. 5.2.4 More on stress non-preservation Since stress preservation is the default outcome of English suffixation, the focus of the following will be on those suffixed derivatives whose preferred stress patterns are not the stress patterns of their respective bases, i.e., e.g., the findings that palatal and skeletal are more frequently stressed ˈ palatal and ˈ skeletal (than pa ˈ latal and ske ˈ letal ), preserving the stress of (the more frequently used) base forms ˈ palate and ˈ skeleton , are not particularly interesting because, as just pointed out, the stress of the base form is usually preserved in English in the derived form. Notice also that of the stress patterns / ˈsiːkrətɪv/ vs. / sɪˈkriːtɪv/ of secretive ( OED ), only the former could be attested by the author in 188 captioned You- Tube videos in which secretive was found to have been pronounced by native English speakers 219 times; in agreement with this, the OD gives for secretive only the initially-stressed transcription / ˈsiːkrɪtɪv/ ( OD ). This finding is also not particularly interesting because the only conclusion it justifies is that the derived adjective secretive is for a Present-day English speaker morphologically only secret + ive but not secrete + ive , i.e., in addition to the initially-stressed noun secret , / ˈsiːkrɪt/ ( OD ), there is also the semantically related finally-stressed verb secrete , / sɪˈkriːt/ ( OD ), which means “conceal, hide.” The stress of the derived adjective secretive is thus either the preserved stress of / ˈsiːkrɪt/ of a secret or the preserved stress of / sɪˈkriːt/ of to secrete . Note, however, that while secret occurs in the BNC 5,528 times, secrete has only 78 attestations in the corpus (the corresponding numbers from the COCA corpus are 38,920 vs. 294). The former has therefore more chances of counting for an English speaker as the base of the derived adjective secretive than the latter. The only stress pattern of secretive in contemporary English is therefore the initially-stressed / ˈsiːkrɪtɪv/ . Similarly, the finding that infiltrate is more frequently stressed / ˈɪnfɪltreɪt/ than / ɪnˈfɪltreɪt/ only means that back-derivation from infiltration is for a contemporary English speaker a more plausible morphological analysis of infiltrate than the prefixation analysis in - + filtrate (← filter + ate ). The reason for this might be the fact that in contemporary English, this verb is usually used to express the meaning “[i]ntroduce (someone) into an organization, place, etc. surreptitiously, in order for them to acquire secret information” ( OD ) rather than “(Of a liquid) permeate something by filtration” ( OD ). This weakens the impression of infiltrate being in - + filtrate , and because the shorter verb infiltrate occurs in the BNC and COCA only 117 and 722 times whereas the longer 5.2 Stress variation in YouTube 185 noun infiltration has 198 / 854 attestations, the usual stress pattern of the former is / ˈɪnfɪltreɪt/ ( OED ), with the secondary stress of the latter, i.e., / ˌɪnfɪlˈtreɪʃən/ ( LDOCE ), being promoted to primary stress. In addition to the above mentioned vowel effect, all deviations from stress preservation attested by the author in captioned YouTube videos can be attributed to the following three factors (which sometimes simultaneously apply to one and the same suffixed derivative): 1) emphasis, 2) segmental length principle, and 3) prefixation analysis (which makes more sense than the “correct” suffixation analysis). As for emphasis, consider, e.g., dinette , which is more frequently stressed / daɪˈnɛt/ ( OED ) than / ˈdaɪnɛt/ ( OED ). Because diminutiveness is an important aspect of the meaning of dinette , “[a] small room or part of a room used for eating meals” ( OD ; boldface mine), and since it is the suffix ette that expresses this meaning in dinette , stress in this derivative is more frequently final than initial. Similarly, as for the finally-stressed pronunciation / ˌkastəˈnɛt/ of castanet (OED), which is more frequent than / ˈkɑːstənɪt/ (OED), note that the Spanish etymon castañeta is diminutive of castaña , “chestnut” (OED). Although in the target language English, castanet is not segmentable into castan and et the way, e.g., toucanet , “[a] small tropical American toucan with mainly green plumage” (OD), can be segmented into the base toucan and the diminutive suffix et , diminutiveness nonetheless remains an aspect of the meaning of castanets , which are “[s]mall concave pieces of wood, ivory, or plastic, joined in pairs by a cord and clicked together by the fingers as a rhythmic accompaniment to Spanish dancing” ( OD ). Pronouncing castanet / ˌkastəˈnɛt/ can thus be seen as a means of emphasizing the diminutiveness of castanets. By contrast, millionaire , for which LDOCE incorrectly gives only the finally-stressed transcriptions / ˌmɪljəˈneə/ (British English) and / -ˈner/ (American English), was considerably more frequently pronounced in captioned YouTube videos with initial stress: / ˈmɪl-/ ( OED ). Apart from the French origin of the word millionaire ( OED ), there are no reasons to stress it / ˌmɪljəˈneə/ : The suffix aire does not have a strong emphatic potential and should therefore, just like other English suffixes, be stress-repellent. The same is true of hotelier , in the case of which the OED correctly places the pronunciation / hoʊˈtɛljər/ , which preserves the stress of the base hotel , before the finally-stressed pronunciation / ˌɔtɛlˈjeɪ/ ( OED ). As in the case of millionaire , there are no reasons why stress in the derivative hotelier should fall upon the suffix ier . Note finally that even employee was more frequently stressed em ˈ ployee than employ ˈ ee even though the suffix ee does have a strong emphatic potential (cf. employee and employer ). Because the derivative employee does not always occur 186 5 Case studies in contexts that involve a contrast with the derivative employer , the usual stress pattern of the former is the base stress / ɛmˈplɔɪiː/ ( OED ). As for the segmental length principle, consider the derivative trapezoid , whose base is the finally-stressed trapeze : / trəˈpiːz/ ( OD ). The preferred stress pattern of trapezoid is, however, not / trəˈpiːzɔɪd/ ( OED ) but / ˈtræpɪzɔɪd/ ( OED ), with stress in the derivative regularly occurring in the beginning of its lefthand disyllabic string tra . pe -. As for prefixation analyses, consider the line breaks arch|ival, hor|mo¦nal, may|or¦al, and neur|on¦al ( OD ), in which the only or the preferred division point, indicated by (|), is the penult-antepenult boundary. To begin with, observe that line breaks in the OD often deviate from traditional syllabification strategies, such as the Maximal Onset Principle. E.g., the above mentioned trapezoid is tr&/ pI/ zoId in the MRC vs. trap¦ez|oid in the OD ; similarly, exquisite is eks/ kwI/ zIt in the MRC / ex∙qui∙site in LDOCE vs. ex¦quis|ite in the OD : Whereas the former are phonetic divisions into syllables, the latter seems to represent a morphological segmentation of exquisite into the prefix ex -, the root quis (cf. acquis , croquis , maquis , marquis ) and the suffix ite ; likewise, the line break trap¦ez|oid seems to be in accordance with the suffixation analysis trapeze + oid . Noteworthy are also the following numbers. Excluding 1) compounds, 2) words that the tool Morphological Analysis considers to be suffixed derivatives, and 3) words such as Canadian , which can be regarded as trisyllables due to hiatus resolution in them (e.g., / kə-ˈneɪ-dɪ-ən/ > / kə-ˈneɪ-djən/ ), we are left with 750 line breaks given in the OD for penultimately-stressed trisyllables (according to the MRC database). Of these, 457 (~60.93 %) are line breaks such as prog|no¦sis and de|velop, in which the only or the preferred division point, indicated by (|), is the penult-antepenult boundary, i.e., e.g., the only possible division of develop is, according to the OD , de - + velop and in the case of prognosis , the division prog - + nosis is to be preferred to the division progno - + sis . By contrast, in the case of 2,267 line breaks given in the OD for antepenultimately-stressed trisyllables from the MRC database, the same can be said about only 628 line breaks: ~27.7 %. E.g., con|se¦quent, bal|cony. (Much more typical of antepenultimately-stressed trisyllables are, by contrast, line breaks such as bene|fit, cal¦cu|late, or con|sum|mate, where the only or the preferred division point is the ult-penult boundary or it does not matter whether, e.g., consummate is orthographically divided into con and summate or into consum and mate .) The difference of 457 / 750 vs. 628 / 2,267 is statistically hugely significant: χ 2 (1) = 270, p < 0.000001. Assuming that the line breaks given in the OD reflect how contemporary English speakers morphologically analyze English words (cf. Hammond 2006: 413, arguing that “[m]orphology plays a role in that hyphens are preferentially 5.2 Stress variation in YouTube 187 placed at […] morpheme boundaries, e.g. unable […] is better hyphenated as un able , rather than u nable ”), we argue that the trisyllables archival , hormonal , mayoral , and neuronal prefer penultimate to antepenultimate stress because for the majority of Present-day English speakers, these items count morphologically as prefixed disyllables (which is reflected in the line breaks arch|ival, hor|mo¦nal, may|or¦al, and neur|on¦al, in which the only or the preferred division point occurs at the penult-antepenult boundary). E.g., the fact that the stress pattern / meɪˈɔr(ə)l/ is in American English considerably more frequent than the stress pattern / ˈmeɪər(ə)l/ , preserving the stress of the disyllabic base / ˈmeɪər/ (LDOCE), means that for American English speakers, the morphological structure of mayoral is not the “correct” mayor + al , but may + oral . (Cf. British English, where the base mayor is monosyllabic: / mɛː/ ( OD ); the disyllabic derivative mayoral is therefore pronounced only / ˈmɛːrəl/ ( OD ), with stress in the word regularly occurring upon the only syllable constituting the monosyllabic base / mɛː/ .) Similarly, the line breaks hor|mo¦nal and neur|on¦al strongly suggest that the shared string onal is analyzed by English speakers as the root of these al -derivatives. Therefore, the stress pattern / ˈhɔːməʊnəl/ ( OED ), which preserves the stress of the base / ˈhɔːməʊn/ ( OD ), could not be attested in any of the 152 captioned YouTube videos in which hormonal was found to have been pronounced by native English speakers 250 times; the only stress pattern heard by the author was / hɔːˈməʊnəl/ ( OED ), which is also the only stress pattern given for hormonal in the OD . Likewise, neuronal is both in British and American English interchangeably stressed / ˈnjʊərənl/ and / -ˈrəʊnl/ ( OED ), but of these, only the latter was heard by the author in 102 captioned YouTube videos in which neuronal was found to have been pronounced by native English speakers 145 times; in agreement with this finding, the OD gives for neuronal only the transcription / njʊəˈrəʊn(ə)l/ , which does not preserve the stress of the base neuron : / ˈnjʊərɒn/ ( OD ). Another “incorrectly”-stressed al -derivative is archival . Its usual (and, according to the OD , sole) stress pattern in contemporary English is / ɑːˈkʌɪv(ə)l/ even though the base form archive is stressed / ˈɑːkʌɪv/ (OD). Given the line break arch|ival and given that also the ival -word adjectival is stressed / adʒɪkˈtʌɪv(ə)l/ ( OD ), the analysis of the string ival as the root of both archival and adjectival immediately suggests itself. (A good guess would be that this stress pattern has also something to do with the existence of the ival -words arrival and survival , which occur in the BNC 3,340 and 3,117 times respectively, while archival / adjectival have only 152 / 77 attestations.) Finally, we also discuss deviations from stress preservation among back-derivatives, in the case of which the stress of a shorter output form is also as a 188 5 Case studies rule the stress of its longer input form. For example, in contrast to the verse -adjectives adverse , inverse , and transverse , which are usually pronounced with initial stress, the verse -adjective diverse is usually pronounced with final stress, preserving the antepenultimate stress of the more frequently used noun diversity , i.e., not only in the BNC but also in the COCA corpus, the shorter word diverse has fewer attestations than the longer word diversity : 16,685 vs. 18,591. The verse -adjectives adverse , inverse , and transverse are, by contrast, the most frequent words in the COCA corpus that begin with the strings advers -, invers -, and transvers -. Accordingly, because each of these adjectives often occurs in the modifier position (e.g., inverse function ), initial stress in them is from the point of view of rhythm better than final stress. The verse -adjective diverse , which is a back-derivative from diversity , is, however, pronounced with final stress even in combinations such as diverse learners and diverse set , in which the immediately following head word is either initially-stressed or monosyllabic. The synchronic back-derivation from the longer word diversity thus blocks the application of the Rhythm Rule in the shorter word diverse . A similar case is Abdul , which in the COCA corpus has fewer attestations than Abdullah (to which it etymologically goes back): 1,769 vs. 2,211. In the BNC , by contrast, the shorter Abdul has more attestations than the longer Abdullah : 239 vs. 193. Thus, because 218 (~91.21 %) occurrences of Abdul in the BNC are occurrences in combinations such as Abdul Rahman , in which the proper noun Abdul modifies another proper noun, initial stress in Abdul in a combination such as Abdul Rahman is from the point of view of rhythm better than final stress: / ˌabdʊl ˈrɑːmən/ ( OD ). In the British variety, Abdul is therefore exclusively stressed / ˈabdʊl/ , whereas in American English, final stress in Abdul is considerably more frequent than initial stress even in combinations such as Ab ˈ dul ˈ Rahman . Counterexamples include 1) ally , which in COCA is less frequent than alliance (5,911 vs. 14,310), 2) nocturne , which in COCA is less frequent than nocturnal (81 vs. 1,157), and 3) recess , which in COCA is less frequent than recession (2,884 vs. 12,074). In spite of these facts, stress in the shorter words ally , nocturne , and recess was either exclusively or more frequently initial, non-preserving the stress of the semantically related, more frequently used longer words alliance , nocturnal , and recession . In the case of ally , the reluctance to pronounce it / əˈlʌɪ/ ( OD ), preserving the stress of the base / əˈlʌɪəns/ ( OD ), is due to the orthographic form ally . Since the line break given in the OD for alliance is al¦li|ance, we can assume that also the shorter ally is orthographically syllabified by English speakers al . ly (cf. Hammond 2006: 413, arguing that “double letters are better split by a hyphen than not, e.g. at test […] is much better than att est or a ttest ”), i.e., the ult has the 5.2 Stress variation in YouTube 189 same segmental length as the penult; stress in ally is therefore supposed to be initial: / ˈalʌɪ/ (OD). A similar case is inform , which occurs in the BNC 1,465 times and is thus considerably less frequent than information , which has 38,488 attestations. The stress pattern of the latter—/ ˌɪnfəˈmeɪʃən/ ( LDOCE )—is, however, not preserved in the former because in the case of inform , also the prefixation analysis is more than obvious: The segmentally longer ult form , which occurs in English as a separate word, has 33,817 attestations in the BNC . Stress in inform is therefore final. Similarly, fine has 12,724 attestations in the BNC and is thus more frequent than definition , which occurs in the corpus only 4,791 times. Accordingly, the prefixation analysis de - + fine should in the case of the disyllabic verb define make more sense than back-derivation from definition . Note also that of the 237 ation -tetrasyllables in LDOCE , 91 (~38.4 %) have a trisyllabic ate -counterpart. E.g., compensation - compensate . Given this association, we argue that if contemporary English had the verbs * accusate and * declarate , they would be stressed antepenultimately, preserving the (secondary) stress of the base nouns accusation and declaration . The existing disyllabic verbs accuse and declare are, by contrast, not perceived as back-derivatives from accusation and declaration and receive therefore final stress, abiding by the segmental length principle. As for nocturne , note that the meaning of the more frequently used nocturnal is “[d]one, occurring, or active at night” ( OD ). The less frequently used nocturne means, however, “[a] short composition of a romantic nature, typically for piano” ( OD ) or “[a] picture of a night scene” ( OD ). The meanings of the derivative nocturne thus represent fairly specialized extensions of the meaning of the base nocturnal . The stress difference / nɒkˈtəːn(ə)l/ vs. / ˈnɒktəːn/ ( OD ) is therefore deliberately employed by English speakers to make the derived noun formally different from the base adjective. Finally, recess is usually stressed / ˈriːsɛs/ (irrespective of whether it functions morphosyntactically as a noun or verb) because semantically, recess , “a small concavity,” “an enclosure that is set back or indented,” “a pause from doing something (as work),” “close at the end of a session,” etc. (WordNet), does not have much to do with recession , which is mainly associated with the meaning “the state of the economy declines” (WordNet). The pair recess - recession is thus different from the formally similar pair process - procession , in which to process , one of whose meanings is “march in a procession” (WordNet), is semantically easily analyzable as a back-derivative from procession . To process meaning “march in a procession” is thus stressed / prəˈsɛs/ ( OD ), preserving the stress of procession , but the semantically unrelated a process , “[a] series of actions or steps taken in order to achieve a particular end” ( OD ), and the converted verb to process meaning “[p]erform a series of mechanical or chemical operations on 190 5 Case studies (something) in order to change or preserve it” ( OD ) are stressed only / ˈprəʊsɛs/ ( OD ). Note also that when recess expresses the meaning “interruption of an activity,” it is semantically the opposite of process of be in the process of doing something ; initial stress in recess thus also seems to be a matter of emphasis. 5.2.5 More on disyllables One of the central claims of this monograph is that a disyllabic English word should have initial stress if its ult contains a string of letters that occurs in English as a suffix. Why is then cement , whose righthand string ment occurs in English as a suffix, still more frequently stressed / sɪˈmɛnt/ ( OED ) than / ˈsɛmənt/ ( OED ), preserving the penultimate stress of its Latin etymon caementum , while, e.g., comment and segment are no longer stressed / kəˈmɛnt/ and / sɛɡˈmɛnt/ (OED) even when functioning morphosyntactically as verbs (i.e., the OD still gives for to segment the finally-stressed transcription / sɛɡˈmɛnt/ , but the only stress pattern heard by the author (in occurrences of segment as a verb) was initial stress)? A possible answer to this question is that disyllables from the OED that were found to prefer final stress contain on average ~3.37 phonetic segments in the ult vs. ~2.03 phonetic segments in the penult; the average length difference is thus ~1.34. By contrast, in the case of disyllables that were found to prefer initial stress, the corresponding numbers are ~3.56 (ult) vs. ~2.51 (penult); the average length difference is thus only ~1.05. Accordingly, because the penult of a disyllable such as segment contains on average more phonetic segments than the penult of a disyllable such as cement (2.5 vs. 2), the contrast between the segmental length of the ult vs. the penult is more pronounced in the case of disyllables such as cement than segment . That is, proceeding from the syllabifications sI/ ment and seg/ m@nt (MRC), we note that the former word should have final stress because the ult in cement contains two more phonetic segments than the penult; due to the same reason, foment and lament , whose hyphenations in LDOCE are fo∙ment and la∙ment, are stressed / fəʊˈment/ and / ləˈment/ ( LDOCE ). The word segment should, by contrast, have initial stress because the ult in it contains only one more phonetic segment than the penult. Similarly, proceeding from the orthographic syllabification com∙ment ( LDOCE ), we note that stress in comment should be initial rather than final. (Cf. augment , which is stressed / ɔːɡˈment/ ( LDOCE ) even though the ult in this word contains only one more orthographic segment than the penult: aug∙ment ( LDOCE ). A possible explanation for this fact is that of the 292 polysyllabic words in LDOCE that contain the orthographic sequence au , only two (~0.68 %) are the words Menelaus and Nauru , in which this sequence is phonetically realized as hiatus, i.e., Men∙e∙la∙us 5.2 Stress variation in YouTube 191 and Na∙u∙ru ( LDOCE ). Given this fact, we might assume that the word augment is treated by English speakers in the same way as the words cement , foment , and lament rather than as comment and segment .) These claims are strongly supported by the fact that while / ɪˈneɪt/ (OD), which is the etymological penultimate stress of the Latin innātus ( OED ), still remains the preferred stress pattern of the English adjective innate (even in environments such as innate cruelty , innate freedom , innate goodness , innate passion , etc., in which initial stress would from the point of view of rhythm be better than final stress), the adjectives connate , pennate , and pinnate , which also etymologically go back to the penultimately-stressed Latin nātus -words connatus , pennatus , and pinnatus ( OD ), are in Present-day English stressed only / ˈkɒneɪt/ , / ˈpɛnət/ , and / ˈpɪneɪt/ ( OD ), with their final syllables repelling stress. What distinguishes innate from connate , pennate , and pinnate is that in the former, the ult . nate contains two more orthographic symbols than the penult in ., while in the latter case, the difference between the orthographic length of the ult . nate and that of the penults con ., pen ., and pin . is only one. Stress is therefore as a rule final in innate while it is initial in connate , pennate , and pinnate . Note also that to contact is in the OED both / ˈkɒntækt/ and / kənˈtækt/ , but of these, only the former stress pattern was heard by the author in 72 captioned YouTube videos containing the spoken occurrences of to contact . On the one hand, it can be argued that initial stress in to contact , “be in direct physical contact with; make contact” (WordNet), is simply the preserved initial stress of the base a contact . (Since the meaning of to contact represents a fairly general extension of the meaning of a contact , there is clearly no need to have different stress patterns for these words.) Why is, however, the base a contact in violation of the segmental length principle stressed / ˈkɒntækt/ (LDOCE)? Cf. the adjective intact , which in accordance with the segmental length principle is stressed / ɪnˈtækt/ ( LDOCE ). What distinguishes the initially stressed tact -word contact from the finally-stressed tact -word intact is that it is only the latter in which the difference between the segmental length of the ult . tact and that of the penult in . is two, which is prototypically exhibited in English by a finally-stressed disyllable. As for allele , which is / əˈliːl/ vs. / ˈaliːl/ in British English and only / əˈlil/ in American English ( OED ), note that according to the OD , English has the suffix le , which, among other things, forms “names of animals and plants: beetle .” Notice, however, that of the 3,361 words in the OD dictionary that end orthographically in le , not a single one counts, according to the tool Morphological Analysis, as a product of le -derivation. This string is thus not easily discernible as a suffix, hence Present-day English speakers more frequently stress the shorter output form allele / əˈliːl/ than / ˈaliːl/ , preserving thereby the stress of the longer input form allelomorph , i.e., / əˈliːləʊmɔːf/ ( OD ), in which the 192 5 Case studies trisyllabic combining form allelo -, which ends orthographically in o , is stressed upon its middle syllable. A slightly different case is bastile . Its more frequently used spelling variant bastille (which occurs in COCA 259 times, whereas bastile does not have attestations in the corpus) ends in the string ille , which, in contrast to the string ile of bastile (cf. percentile ), does not occur in English as a suffix (i.e., of the 102 words in the OD that end orthographically in ille , not a single one was analyzed by the tool Morphological Analysis as an ille -derivative). Bastille is therefore in accordance with the segmental length principle usually stressed / bɑːˈstiːl/ ( OED ). A fairly similar case is duress and largess , for which the OED gives the orthographic alternatives duresse and largesse and, what is particularly important, the variant largesse has more attestations in the COCA corpus than largess : 475 vs. 227. Thus, because in contrast to the string ess , the string esse does not occur in English as a suffix, duresse and largesse are in accordance with the segmental length principle usually stressed / djʊˈrɛs/ and / lɑːˈdʒɛs/ ( OED ). The case of largess vs. largesse cannot but remind of the orthographic difference between the initially-stressed moral , with the string al repelling stress, and the finally-stressed morale , which etymologically goes back to the French moral , “respelled to preserve the final stress in pronunciation” (OD). Because the string al occurs in English as a suffix and therefore repels stress, the spelling variant moral can only be associated with the initially-stressed pronunciation / ˈmɒr(ə)l/ ; the spelling variant morale can, by contrast, be associated with the finally-stressed pronunciation / məˈrɑːl/ ( OD ) because the string ale does not occur in English as a suffix (i.e., of the 108 words in the OD that end orthographically in ale , not a single one is, according to the tool Morphological Analysis, an ale -derivative). Among the factors accounting for different stress patterns in formally similar words is most likely also a semantic similarity to genuine suffixed derivatives that came into existence with the help of a particular suffix. For example, comparing the stress patterns / ɪˈvent/ and / ˈsaɪlənt/ ( LDOCE ), we note that etymologically, the English words event and silent are due to the Latin words ēˈventus and siˈlentium (Dictionary.com). Cf. the modern Italian words evento and silenzio , which are stressed / e.ˈvɛn.to/ and / si.ˈlɛn.tsio/ ( PONS ). What distinguishes the finally-stressed English word event from the initially-stressed English word silent is, however, the fact that the latter is semantically more similar to genuine ent -derivatives, such as, e.g., urgent . The ent of silent is therefore more stress-repellent than the ent of event . A similar case seems to be the stress difference between / ˈfəːmɛnt/ of ferment as a noun vs. / fəˈmɛnt/ of ferment as a verb ( OD ). Since the genuine suffix ment 5.2 Stress variation in YouTube 193 forms only nouns “expressing the means or result of an action: curtailment , excitement ” ( OD ), it is undeniably the ment of the noun ferment , meaning “[a]gitation and excitement among a group of people, typically concerning major change and leading to trouble or violence” ( OD ), that is more similar to the genuine suffix ment of the derivative excitement than the ment of to ferment . The noun ferment is therefore stressed / ˈfəːmɛnt/ , with the suffix-like string ment repelling stress, while the verb ferment is in accordance with the segmental length principle stressed / fəˈmɛnt/ . Note also that apart from the disyllabic words address , amine , chagrin , divan , Koran , mirage , offense , and pecan , there are also the monosyllabic words dress , mine , grin , van , ran , rage , fence , and can . The existence of, e.g., the monosyllabic word dress , which occurs in COCA 31,745 times, intensifies the impression of address being made up of the prefix ad and the root dress ; the usual stress pattern of address is therefore the finally-stressed / əˈdrɛs/ (which, as pointed out in 1.1, is the only stress pattern of address when it expresses the meaning “public speech”). Similarly, mine is in COCA considerably more frequent than amine (32,654 vs. 51), grin is considerably more frequent than chagrin (6,974 vs. 780), van is considerably more frequent than divan (31,846 vs. 309), ran is considerably more frequent than Koran (53,070 vs. 1,351), rage is considerably more frequent than mirage (8,813 vs. 1,288), with the phonetic realization of rage occurring as a separate word being, however, different from that of the righthand component rage of the disyllabic word mirage , fence is considerably more frequent than offence and almost as frequent as offense (13,112 vs. 289 and 13,634), and, finally, can is considerably more frequent than pecan (1,149,928 vs. 1,571). Thus, we have good reasons to regard the monosyllabic words mine , grin , van , ran , rage , fence , and can as the roots of the disyllabic words amine , chagrin , divan , Koran , mirage , offense , and pecan . Stress in the latter should therefore be final, falling upon the only syllables constituting the former. E.g., divan and pecan should be stressed / dɪˈvan/ and / pɪˈkan/ ( OD ), with stress in these disyllabic prefixed derivatives regularly falling upon the only syllables constituting their monosyllabic bases van and can . 5.2.6 Final stress in trisyllables Only six instances of the variation category “u/ a” (~9.23 %) were found to have been either exclusively or more frequently pronounced by native English speakers with final rather than with antepenultimate stress: bioengineer , castanet , Ménière , misdiagnose , namaste , and obsolete . As for bioengineer and misdiagnose , it is obvious that the stress patterns / ˌbaɪoʊɛndʒəˈnɪ(ə)r/ and / ˌmɪsdʌɪəɡˈnəʊz/ are from the point of view of rhythm 194 5 Case studies better than the stress patterns / ˌbaɪoʊˈɛndʒənɪ(ə)r/ and / ˌmɪsˈdʌɪəɡnəʊz/ ( OED ), which involve the following stressed syllables / ˈɛn/ and / ˈdʌɪ/ being separated from the preceding stressed syllables / ˌbaɪ/ and / ˌmɪs/ by only one unstressed syllable / no unstressed syllables, i.e., the prefixes bio and mis -, which modify the meanings of the bases engineer and diagnose in an important way, receive secondary stress; the stress patterns / ˌbaɪoʊˈɛn-/ and / ˌmɪsˈdʌɪ-/ are therefore rhythmically unfortunate. Cf. the base diagnose , which is only / ˈdaɪəɡnəʊz/ according to LDOCE , and similarly, in the OD , the initially-stressed pronunciation / ˈdʌɪəɡnəʊz/ is given before the finally-stressed / dʌɪəɡˈnəʊz/ . The case of castanet was already discussed above. Pronouncing castanet / ˌkastəˈnɛt/ is a means of emphasizing the diminutiveness of castanets. A somewhat similar case is namaste , “[a] respectful greeting said when giving a namaskar” ( OD ), which is / ˈnaməsteɪ/ vs. / naməˈsteɪ/ in British English and / ˌnɑməˈsteɪ/ vs. / ˈnɑməˌsteɪ/ in American English (OED). Even without knowing that the morphological structure of namaste in the source language Hindi is namas , “bowing,” + te , “to you” ( OD ), a contemporary English speaker can nevertheless arrive at the segmentation namas - + te because apart from namaste , English has also imported from Hindi the semantically-related word namaskar , which means “[a] traditional Indian greeting or gesture of respect, made by bringing the palms together before the face or chest and bowing” ( OD ). (It can be assumed that the mental lexica of those English speakers who have an entry for the Hindi loanword namaste also have an entry for the related Hindi word namaskar .) Pronouncing namaste / ˌnaməˈsteɪ/ and namaskar / ˌnʌmʌsˈkɑː/ ( OD ) thus means placing stress on the syllables that make these two semantically-related Hindi loanwords formally different from each other. As for Ménière , final stress in it can be regarded as a foreignness marker (which, of course, is also true of final stress in namaste and namaskar ), i.e., Ménière’s disease was named after the French physician Prosper Menière ( OED ), and in the case of obsolete , it is difficult to explain why in contrast to British English speakers, who stress obsolete / ˈɒbsəliːt/ (the only stress pattern in LDOCE / OD), American English speakers still prefer the finally-stressed version / ˌɑbsəˈlit/ ( OED ). As argued in 4.6, American English is with regard to stress a more archaic variety than British English, i.e., as observed in 5.1.3.3, final stress in the English word obsolete is the preserved penultimate stress of the Latin etymon word obsolētus . 5.2 Stress variation in YouTube 195 5.2.7 Hiatus resolution As observed in 5.1.6, cases of hiatus resolution among English words with stress doublets fall into two types: those only accompanying a change in the location of stress in a word, without, however, being responsible for it, and those that do indeed condition the emergence of an alternative stress pattern. Of the latter, al Qaeda , extraordinary , guano , Ophiuchus , and Seato were either exclusively or more frequently pronounced in a hiatus-resolving manner, i.e., / -ˈkʌɪ-/ instead of / -kɑːˈiː-/ ( al Qaeda ), / -ˈstrɔː-/ instead of / -strəˈɔː-/ ( extraordinary ), / ˈɡwæ-/ instead of / ɡjuːˈæ-/ ( guano ), / -ˈfjuː-/ instead of / -fɪˈu-/ ( Ophiuchus ), and / ˈsiː-/ instead of / sɪˈeɪ-/ ( Seato ). (Recall also the above mentioned fact that physiatrist is more frequently / fᵻˈzʌɪətrɪst/ than / ˌfɪzɪˈatrɪst/ .) By contrast, in the words geographer , geography , geometry , gouache , and Laotian hiatus was as a rule unresolved, but do observe that in videos featuring British English accents, geographer , geography , and geometry were more frequently pronounced / ˈdʒɒ-/ than / dʒɪˈɒ-/ , with the latter being, however, the preferred pronunciation among American English speakers. Gouache was more frequently pronounced / ɡuːˈɑːʃ/ than / ɡwaʃ/ and Laotian was more frequently / lɑːˈəʊ-ʃən/ than / ˈlaʊ-ʃən/ . These findings lend additional support to the claim that initiality inhibits hiatus-resolution in English. Thus in the words geographer , geography , geometry , gouache , and Laotian , in which hiatus is still as a rule unresolved in contemporary English, hiatus sequences are separated from the left word boundary only by a consonant. By contrast, in the words extraordinary , Ophiuchus , and physiatrist , the hiatus sequences a . o , i . u , and i . a are separated from the left word boundary both by a consonant and a vowel. 5.2.8 Within-speaker variation 185 items for which stress doublets are given in the OED were found to have been non-identically-stressed by identical native English speakers. These constitute ~28.73 % of the total number of the 644 items with stress doublets in the OED that were found to have been non-identically-stressed in captioned YouTube videos. Situations in which identical English speakers non-identically stress identical English words are thus not infrequent in comparison with cases of across-speaker variation, which involves non-identical English speakers non-identically stressing identical English words. If individual variation categories are taken into consideration, it is u/ p-instances in the case of which identical English speakers were most frequently heard to have used more than one stress pattern: Of the 233 u/ p-instances in the case of which stress doublets were attested not only in the OED but also in 196 5 Case studies captioned YouTube videos, 105 items (~45.06 %) were pronounced by identical native English speakers with more than one stress pattern. The second most frequent category is “u/ a”: 22 / 68=~32.35 %. This finding strongly suggests that within-speaker variation is as a rule triggered either rhythmically (e.g., cardiac ˈ defect vs. birth de ˈ fects ) or emphatically (e.g., the suffixes ee , or , ess sometimes attract stress because they distinguish words such as employee , lessor , princess , etc. from the formally and semantically related formations employer , lessee , and prince ). In addition to the causes of stress variation common to both withinand across-speaker variation, the former may also be due to a discovery that other native English speakers use a different stress pattern in a particular English word. Thus according to a colleague (an American English speaker who has been working in Australia since 2010), I think all of those competing stress patterns are regional variation. The only reason a speaker would switch between them would be if they’d been exposed to two dialects -so that now, for example, I sometimes use my original a'dult and sometimes the Australian (maybe also east US and British) 'adult. (personal communication via email in February 2014) The awareness of the fact that other English speakers prefer a different stress pattern in a particular English word may also lead to meta-linguistic comments, in which the existence of multiple stress patterns is either simply acknowledged or condemned in a prescriptive manner. An example of the former is the video b9tLdO6 LCXA (27. 01. 2015), in which chalcedony (also spelled calcedony ) is pronounced / kælˈsɛdənɪ/ , but it is pointed out, without expressing any evaluative judgments, that some people use the pronunciation / ˈkælsɪdənɪ/ ; similarly, in the videos Rao8 TT cviS0 (01. 02. 2015) and Z7-62 EM n_hI (01. 02. 2015), it is stated by an identical American English speaker that ANOVA can be pronounced both / ˈænəvɑ/ and / əˈnoʊvə/ . An example of the latter is the video Tk RP vqbEdjE (08. 01. 2015), in which a chemistry professor at the University of California at Berkeley says that “[he] get[s] upset with people who say / bəˈkiːlɪəm/ ” instead of / ˈbɜːklɪəm/ , i.e., because Berkelium was synthesized at Berkeley , the stress pattern of the latter, / ˈbɜːkli/ , should be preserved in the former. A similar case is the video ZJY3sXxku-M (09. 01. 2015), in which one American English speaker criticizes another American English speaker for using the antepenultimately-stressed pronunciation / kəˈrɪbiən/ in the context of the Pirates of the Caribbean movie title. According to the former, the only correct pronunciation of Caribbean is the penultimately-stressed / ˌkɛrəˈbi: ən/ . 5.2 Stress variation in YouTube 197 Noteworthy is also the video HKM zt8bPeuU (23. 12. 2014), which in part is devoted to a meta-linguistic discussion of whether the correct stress pattern of carillon is / ˈkærɪljɒn/ or / kəˈrɪljən/ , with most speakers in the video enthusiastically arguing for the former and against the latter. 5.2.9 Summary English words with stress doublets are those in the case of which more than one morphological analysis makes sense. Usually, one of the morphological analyses makes, however, more sense than the alternative analysis, which is why one stress pattern is as a rule more frequently used than the alternative stress pattern. E.g., archetypal is / ɑːˈkɛtɪpəl/ vs. / ˈɑːkɪtaɪpəl/ according to the OED , but of these stress patterns, only the latter was heard by the author in 131 captioned YouTube videos in which archetypal was pronounced by native English speakers 163 times. Indeed, apart from the typal -word archetypal , there are also the typal -words schizotypal and prototypal (and similarly, apart from the type -word archetype , there are also the numerous type -formations genotype , logotype , monotype , prototype , stereotype , etc.); archetypal counts therefore morphologically as a compound and should therefore be accented / ˈɑːkɪtaɪpəl/ , with stronger stress regularly occurring in the beginning of its lefthand disyllabic string ar . che - (with the e , in which this string ends, being merely a connecting vowel (interfix) and thus repelling stress). At the same time, however, in both the OD and the BNC , archetypal is the only word that ends in the string etypal (and in COCA , there is only one attestation of antetypal ). The analysis of this string as the root of archetypal makes therefore relatively little sense. Note also that because the lefthand trisyllabic string antety ends orthographically in y , the vowel effect that results in, e.g., the stress shift / ˈpɪrəmɪd/ → / pɪˈramɪd(ə)l/ ( OD ), i.e., the lefthand string pyrami of the al -derivative pyramidal ends orthographically in i and receives therefore penultimate stress, does not apply to the al -derivative archetypal , which is therefore pronounced with the stress of its base archetype . In some cases, however, there does not seem to be a reason why one morphological analysis should be preferred to an alternative analysis. E.g., extant was almost as frequently / ˈɛkstənt/ as it was / ɛkˈstænt/ (with the latter—as pointed out in 5.1.3.2—occurring even in environments such as extant species , extant creature , extant mammals , etc., in which initial stress would from the point of view of rhythm be better than final stress). On the one hand, because extant is semantically the opposite of extinct (cf. extinct species , extinct creature , extinct mammals , etc.), stress in extant is supposed to be final even when this would result in a stress clash. On the other hand, of the 73 ant -disyllables in 198 5 Case studies LDOCE , 64 (~87.67 %) have initial stress even though the ult in them has as a rule (53 / 73=~72.6 %) more orthographic segments than the penult. E.g., constant , distant , and instant are stressed initially even though their segmentally longer final string stant is clearly more like a root whereas the segmentally shorter initial strings con -, di -, and in are more like prefixes. The likeliest explanation for this fact is that the string ant occurs in English as a suffix (e.g., claimant ) and therefore repels stress. 6.1 English as a Germanic language 199 6 Concluding remarks 6.1 English as a Germanic language According to Wenszky (2004: 12), a serious methodological problem with many influential publications on English stress is that “the stress rules in them were developed on the basis of the analysis of some typical words, but not whole classes of words.” The present monograph differs from the publications criticized by Wenszky (2004: 12) in that all stress assignment tendencies identified in it are based on the analyses of all polysyllabic English words (contained in representative lexicographic collections such as Longman Dictionary of Contemporary English, the MRC Psycholinguistic Database, Oxford Dictionaries, etc.). Proceeding from the findings reported in the previous chapters, we can conclude that English still remains a (very! ) Germanic language as far as its stress system is concerned, i.e., since the location of stress in a polysyllabic English word is as a rule its root-prefix boundary location and since Germanic stress is “prevalently a delimitative stress, marking as it does through the position of the primary peak the beginning of a word or, after certain prefixes, of the word stem” (Poldauf 1984: 19), English stress can without a doubt be still regarded as an instance of Germanic stress; contemporary English thus turns out to be different from other modern Germanic languages, of which, in the view of van der Hulst (2010a: 442), only Icelandic and Faroese can still be referred to as Germanic languages if only the stress systems (of the modern Germanic languages) are taken into consideration: In accordance with the Germanic principle of fixed initial stress, primary stress in Icelandic and Faroese stills falls upon the first syllable (van der Hulst et al. 2010: 801, 804). In disagreement with van der Hulst (2010a: 442) and especially Domahs et al. (2014: 80), who conclude that “German, Dutch, and English must be considered quantity-sensitive languages, with the three languages showing very similar patterns overall,” the present monograph argues that from the point of view of stress, contemporary English is actually more similar to the genuine Germanic languages Icelandic and Faroese than it is to other modern Germanic languages, such as, e.g., German. Thus, as has been noticed by many authors, if only the most frequently used English words are taken into consideration, their stress patterns “may very well be interpreted exclusively in terms of a Germanic type of logic” (Fournier 2007: 236), i.e., “[t]he higher‐frequency words, i.e., the ones most often heard in real speech, are shorter and more likely to have just a single 200 6 Concluding remarks stressed syllable that is either the word‐initial syllable ( garbage , borrow , numbers ) or the only syllable ( trash , take , math )” (Cutler 2015: 110; cf. LDOCE , in which of the 1,542 high-frequency (solidly-spelled) polysyllables, 1,007 (~65.3 %) have initial stress. By contrast, in the case of mediumand lower-frequency polysyllables, the proportions of initially-stressed words are 1,158 / 2,042(=~56.71 %) and 1,043 / 2,153(=~48.44 %). For highvs. medium-frequency words, χ 2 (1) = 27, p < 0.000001; for mediumvs. lower-frequency words, χ 2 (1) = 29, p < 0.000001; for highvs. lower-frequency words, χ 2 (1) = 103, p < 0.000001. A decrease in the frequency of use (of a polysyllabic English word) thus correlates statistically with a decrease in the incidence of initial stress.) Similarly, according to Turk et al. (1995: 144), “[t]here is considerable evidence that the word-initial stress pattern is salient for English users.” In addition to this, it has often been observed that in words of non-Germanic origin (which were extensively borrowed by both German and English speakers), primary stress in English often falls upon a different syllable than it does in German (Arnold & Hansen 1975: 173-174). E.g., Aktivität in German, which, similar to activity in English, is both semantically and formally segmentable into the base aktiv and the suffix ität (cf. Stabilität ← stabil + ität ), has final stress: / aktiviˈtɛːt/ (Duden online). Similarly, whereas in English, fundament and fundamental are stressed / ˈfʌn-/ and / -ˈmɛnt(ə)l/ ( OD ), both the base noun and the derived adjective are stressed on the ult in German. Note also that both the initially-stressed English word nature and the finally-stressed German word Natur etymologically go back to the same Latin word natūra (Duden online), in which stress is penultimate because the vowel in the penultimate syllable is long; the penultimate syllable counts therefore as heavy and stressable. What accounts for these stress differences between English and German is that the morphological principle of not placing stress upon a string that occurs in the language as a suffix (even when that string, just like, e.g., the string ure of nature , is not an actual suffix! ) applies only to English but not to German. Thus, whereas final stress represents “a stress pattern foreign to English” (Poldauf 1984: 77), because final strings, in which polysyllabic English words end, often occur in the language as suffixes, stress patterns such as Aktiviˈtät , Aktiˈvist , akˈtiv , Funda ˈ ment , fundamen ˈ tal , Mu ˈ sik , Na ˈ tur , etc. are completely unproblematic in German, i.e., German speakers readily place stress in a polysyllabic word upon its ult even if it occurs in the language as a suffix. (It is thus very much possible that in German, stress assignment is indeed conditioned phonologically: “Primary stress is final if the vowel is long or there are two closing consonants” (van der Hulst et al. 2010: 803). E.g., Aktivität is stressed / aktiviˈtɛːt/ because the ult contains a long vowel and in, e.g., Aktivist and Fundament stress is final because the vowel in the nucleus is followed by no less than two coda conso- 6.1 English as a Germanic language 201 nants.) A very similar conclusion that English is a more Germanic language than German is also arrived at by Arnold & Hansen (1975: 173), who in the case of words of non-Germanic origin in English, speak of “eine starke Tendenz zur Angleichung der Betonung an die germanische Druckverteilung” (i.e., a strong tendency to adjust the stress of such words to Germanic stress). Since (in contrast to German) the location of stress in a polysyllabic English word crucially depends upon its morphological structure, a central question of the theory of English stress is how this structure is determined by English speakers. To begin with, recall that the majority of non-initially-stressed English words share the following two characteristics: 1) the righthand string that receives primary stress (e.g., hibit of inhibit ) occurs in at least one other English word (cf. prohibit ) and 2) the righthand string that receives primary stress is segmentally longer (i.e., contains more phonetic segments) than the remaining lefthand string that is unstressed / stressed secondary (e.g., the righthand string hibit of inhibit is segmentally longer than the lefthand string in -). The latter is responsible for the fact that penultimate stress is a frequent stress pattern in English only among trisyllabic words, which can often be segmented into a monosyllabic prefix and a disyllabic root. In tetraand pentasyllables, penultimate stress is, by contrast, a very rare stress pattern and among hexasyllables and longer English words, this stress pattern does not exist altogether. What intensifies the prefixation analysis is also a (segmentally longer) righthand string that occurs in English as a separate word. Thus, for instance, simply because the segmentally longer righthand string oral of mayoral occurs in English as a separate word (which is considerably more frequent in the COCA corpus than the longer word mayoral : 11,305 vs. 1,178), the morphological structure of mayoral is for the majority of contemporary American English speakers not the “correct” semantics-based mayor + al but may + oral , of which the segmentally longer righthand string oral counts morphologically as the root and hence attracts stress on to its first syllable. The more frequently used stress pattern of mayoral in contemporary American English is therefore not / ˈmeɪər(ə)l/ but / meɪˈɔr(ə)l/ . Note also that according to a recent introduction to morphology (Lieber 2010: 51), “[a]ffixation, compounding, and conversion are the most common ways of forming new words […] in English […] In addition, there are a number of less common ways in which new lexemes may be formed.” Among these minor morphological processes is back-formation, to a discussion of which Lieber (2010: Section 3.6) dedicates less than a page. The view that back-formation is a minor morphological process is also implicitly expressed in other recent introductions to (English) morphology (Tokar 2012; Schmid 2011; Hamawand 2011; Plag 2003), 202 6 Concluding remarks of which Tokar (2012: 168-170) dedicates three pages to a section on back-formation, Schmid (2011: 212)—only one, and Hamawand (2011) and Plag (2003) do not have such sections altogether; the latter does, however, provide a discussion of the theoretical nature of this morphological process (Plag 2003: 187). Noteworthy is also Mel’čuk’s (five volumes of) Cours de morphologie générale , in which the existence of subtractive word-formation (exemplified by back-derivation) is denied on logical grounds (Mel’čuk 2001: 48, 388; for a summary of different views on back-formation, see, e.g., Štekauer 1998: 4.4). According to the author, back-formation is a purely diachronic phenomenon, which means that, for example, the noun editor was at some point in the past mistakenly analyzed by English speakers as a product of suffixation of the verb edit by means of the agentive suffix or (cf. actor and act ). As soon as, however, a shorter word such as edit comes into existence via back-derivation, it immediately becomes the base of the longer word from which it was back-derived, i.e., e.g., editor is, according to Mel’čuk (2001: 388), synchronically one who edits . In disagreement with the views expressed by Mel’čuk (2001: 48, 388), the present monograph has argued, however, that in many cases, back-derivation is (also) a matter of language synchrony (rather than of (only) diachrony). For example, the previously mentioned verb exist is, according to the OD dictionary, “probably a back-formation from existence ” (whereas Dictionary.com regards it as a shortening of the Latin ex(s)istere ). Because 1) the shorter verb exist occurs in the BNC 5,356 times whereas the longer noun existence has 6,472 attestations and 2) the meaning of exist is, according to the WordNet semantic database, “have an existence,” we have good reasons to assume that the shorter exist still remains a back-derivative from the longer existence . A more obvious case is the shorter verb enthuse , which occurs in the BNC only 39 times and is thus considerably less frequent than the longer noun enthusiasm , which has 2,868 attestations. Whereas in the case of existence , the paraphrase “when somebody or something exists” is no less plausible than the paraphrase of exist as “have an existence,” the paraphrase of enthusiasm as, e.g., “the state of being enthused by somebody or something” is far less intuitive than the paraphrases “cause to feel enthusiasm”/ “utter with enthusiasm,” given for enthuse in the WordNet semantic database. The recognition of the fact that a shorter English word may even synchronically be secondary to a longer one is key to understanding why the stress of the former is sometimes not preserved in the latter. For example, while the shorter Carib has 14 attestations in the BNC , the longer Caribbean occurs in the corpus 1,126 times. For the more frequently used Caribbean , LDOCE gives the transcription / ˌkærəˈbiːən◀/ , which contains the stress shift symbol. The actual stress pattern of Caribbean is thus (given occurrences in the modifier position) 6.1 English as a Germanic language 203 often / ˈkærəbiːən/ . The less frequently used disyllabic Carib preserves this stress pattern: / ˈkarɪb/ ( OD ). Similarly, in contrast to the German word Eu ˈ ropa , the English word Europe is stressed / ˈjʊərəp/ ( LDOCE ) because the more frequently used word European (which occurs in the BNC 20,314 times while Europe has only 17,543 attestations) is, as pointed out in 4.3.5, usually stressed / ˈjʊərəpiːən/ rather than / ˌjʊərəˈpiːən/ . Note also that thus far, it has been suggested in this monograph that antepenultimate stress in an ate -trisyllable (e.g., imitate ) is the promoted secondary stress of a more frequently used ion -tetrasyllable (e.g., imitation ). Recall, however, that, as observed in 5.1.1, primary stress in an ion -tetrasyllable sometimes occurs not pre-finally but initially. That is, imitation is at least on some occasions stressed / ˈɪmɪteɪʃən/ rather than / ˌɪmɪˈteɪʃən/ . Accordingly, the stress pattern / ˈɪmɪteɪt/ (LDOCE) of the less frequently used back-derived verb imitate can be directly obtained from the stress pattern / ˈɪmɪteɪʃən/ of the more frequently used base noun imitation . The only non-Germanic aspect in the contemporary English stress system seems to be penultimate stress in words / components of morphologically complex words that end orthographically in a / o / i . E.g., Obama is / əʊˈbɑːmə/ ( OD ) because the last orthographic vowel in which Obama ends is a . Likewise, the fact that, e.g., extravagant is / ɪkˈstrævəɡənt/ strongly suggests that morphologically, extravagant is not extra + vagant but extrava - + gant , of which the segmentally longer lefthand trisyllabic string extrava ends orthographically in a and therefore receives stress upon its middle syllable. As suggested in 4.3.9, stress patterns such as these are due to the influence of Italian and Spanish, in which 1) stress is predominantly penultimate and 2) words often end orthographically in a / o / i (cf. Burzio 1994: 30, arguing in a different context that “English is significantly more like Italian than like Latin”). Note, however, that in a not insignificant number of cases, penultimate stress in an English a / o / i -word does not contradict the Germanic root-prefix boundary logic. E.g., analyzing the stress pattern / sɪndəˈrɛlə/ of Cinderella ( OD ), we note that penultimate stress in this word is in accordance with the fact that Cinderella is the only English erella -word (in the OD dictionary), but it is one of the many rella -words: chlorella , mozzarella , umbrella , etc. The morphological analysis Cinde - + rella makes therefore more sense than Cin - + derella . Likewise, the stress pattern / ˌsɪnsɪˈnati/ of Cincinnati ( OD ) is in accordance with the fact that apart from the nati -word Cincinnati , there is also the nati -word illuminati (which is stressed / ɪˌl(j)uːmɪˈnɑːti/ ( OD ), preserving the stress pattern of the more frequently used illumination : / ɪˌl(j)uːmɪˈneɪʃ(ə)n/ ( OD ), which occurs in the BNC 327 times while illuminati has only five attestations.) Cf. trattoria , which in accordance with the Italian Stress Rule is supposed to be stressed 204 6 Concluding remarks / tratoˈria/ (OED). At the same time, however, because apart from the toria -word trattoria , there are also the toria -words Astoria , Oscillatoria , Pretoria , Rotatoria , septoria , Victoria , and Vitoria , trattoria has also come to be associated with the antepenultimately-stressed pronunciation / træˈtɔːriːə/ ( OED ). Since stress assignment in Present-day English is by and large conditioned morphologically, variably-stressed English words are as a rule those whose morphological structure can be analyzed in more than one way, i.e., more than one morphological analysis makes sense in the case of one and the same English word. The five main scenarios, which have been discovered in the previous two chapters, include 1) back-derivation and suffixation (e.g., pulsatance → pulsate vs. pulse + ate ), 2) back-derivation and prefixation (e.g., contribution → contribute vs. con - + tribute ), 3) suffixation and prefixation (e.g., over + t vs. o - + vert ), 4) compounding and prefixation (e.g., methyl + amine vs. me - + thylamine ), 5) prefixation and morphological simplicity (e.g., pen - + ult vs. penult , which occurs in antepenult ). 6.2 Future work Although the present monograph could answer the research questions put forth in Chapter 1 (the causes of stress variation and the general stress assignment principles), there still remains room for further investigations of English stress. Compare, e.g., the pairs evict - eviction and contort - contortion , in both of which the shorter verb has fewer occurrences in the BNC than the corresponding longer ion -noun: 80 vs. 215 and 11 vs. 12 respectively. It is fairly obvious, though, that the case of evict vs. eviction , of which the longer noun has 135 more attestations in the BNC than the shorter verb, is different from the case of contort vs. contortion , in which the frequency difference is only one. Can we nonetheless regard contort as a back-derivative from contortion (given that the latter has at least one more attestation in the corpus than the former)? According to Balteiro (2007: 128), who discusses the frequency of use criterion in connection with pairs such as a call vs. to call (i.e., whether the former was converted from the latter or vice versa), frequency differences that are “equal or lower than 25 % of the average” should be regarded as doubtful. Thus, for instance, in the case of 7,943 solidly-spelled words from LDOCE that the tool Morphological Analysis considers to be suffixed derivatives, the median frequency difference is 814 (i.e., 3.971 (~49.99 %) more frequently used forms have no less than 815 more occurrences in the BNC than the corresponding less frequently used form. E.g., corrupt occurs in the corpus 605 times, whereas corruption has 1,422 attestations; the frequency difference is thus 817). The median 6.2 Future work 205 percentage difference is 150 %, which, proceeding from Balteiro’s (2007: 128) threshold of 25 % of the average, means that doubtful frequency differences are those when the percentage difference does not exceed 37.5 %. E.g., the case of evict vs. eviction is not doubtful because the percentage difference between 80 and 215 is ~91.53 %. The percentage difference between 11 and 12 is, however, only ~8.7 %, which makes the analysis of contort as a back-derivative from contortion doubtful. Likewise, because the percentage difference between 5,356 and 6,472 is only ~18.87 %, the above mentioned case of exist vs. existence should also be regarded as doubtful. A possible solution for doubtful cases such as contort vs. contortion and exist vs. existence is to allow bi-directionality in them. Thus, the shorter verbs contort and exist are back-derivatives from the longer nouns contortion and existence , which, in turn, are, however, suffixed derivatives from the shorter verbs contort and exist . This analysis is strongly supported by the fact that in natural languages, all meanings can be classified into what Mel’čuk (2012: 193) calls two semantic “parts of speech”: predicates and (semantic) names (of which only the former are of interest at the moment). A (semantic) predicate is a meaning that denotes a situation: an action, an activity, an event, a process, a state, a property, a relation, etc. A situation necessarily implies participants; therefore a semantic predicate prototypically has open slots for the meanings denoting such participants. (Mel’čuk 2012: 194-195) It is clear that both the verb exist and the noun existence are predicates that share the same participant: an entity that exists. The meaning of the longer noun existence is thus identical to that of the shorter verb exist , so that it does not really matter whether the former is derived from the latter or vice versa. (Similarly, as pointed out in 4.2.1, WordNet defines begin as “have a beginning.” At the same time, however, beginning is defined in WordNet as “the time at which something is supposed to begin.” The analysis of begin as a back-derivative from beginning is thus no less plausible than the suffixation analysis begin + ing for begin .) A somewhat similar case is nature vs. natural , which occur in the BNC 17,800 vs. 14,223 times respectively. The percentage difference is only ~22.34 %, which is why it is again not entirely clear whether natural is nature + al or nature is a back-derivative from natural . On the one hand, natural is defined in the Word- Net database as “in accordance with nature; relating to or concerning nature,” “existing in or produced by nature,” “existing in or in conformity with nature or the observable world,” etc. These paraphrases point to the secondariness of the longer word natural relative to the shorter word nature . At the same time, however, one of the meanings of nature is “the natural physical world including plants and animals and landscapes etc.” (WordNet). Given this meaning defini- 206 6 Concluding remarks tion and given that the percentage difference between the frequencies of nature and natural is lower than 37.5 %, we can argue that the analysis of nature as a back-derivative from natural — nature = that which is natural or characterized by naturalness—is for a contemporary English speaker as natural as the analysis of natural as nature + al . A very similar case is mature vs. maturity , which occur in the BNC 1,873 vs. 1,370 times respectively; the percentage difference is only ~31.02 %. On the one hand, maturity is, according to WordNet, “state of being mature,” which supports the suffixation analysis mature + ity . At the same time, however, the adjective mature is defined in WordNet as “characteristic of maturity” and for to mature , the database gives the definition “develop and reach maturity.” Accordingly, what distinguishes the initially-stressed disyllabic ure -word nature from the finally-stressed disyllabic ure -word mature is that the former can synchronically be seen as a back-derivative from natural , which is stressed / ˈnætʃərəl/ ( LDOCE ), whereas the latter is a back-derivative from maturity , which is stressed / məˈtʃʊərəti/ ( LDOCE ). Nature is therefore stressed / ˈneɪtʃə/ ( LDOCE ) while mature is stressed / məˈtʃʊə/ ( LDOCE ). Content with , which is defined in WordNet as “the state of being contented with your situation in life,” occurs in the BNC 393 times, but contented , which WordNet defines as “made content,” has 344 attestations. The percentage difference between these numbers is only ~13.3 %, which allows us to regard final stress in the adjective content as the preserved penultimate stress of contented , / kənˈtentɪd/ ( LDOCE ), which, however, can be regarded as the preserved final stress of the adjective content . (As for the question of why the noun content has over the course of time abandoned its original stress pattern / kənˈtɛnt/ , the explanation seems to be semantics. Content meaning “[t]hat which is contained in anything” ( OED ) does not have much to do with content meaning “satisfied, happy.”) Success is, according to WordNet, “an attainment that is successful,” but the frequency of use criterion suggests that the longer successful is secondary to the shorter success : 10,695 vs. 13,245. Given that the percentage difference between these frequency numbers from the BNC is only ~21.3 %, we argue that the shorter derivative success is stressed / səkˈses/ ( LDOCE ) because the longer base successful is stressed / səkˈsesfəl/ ( LDOCE ), but the longer derivative successful is stressed / səkˈsesfəl/ because the shorter base success is stressed / səkˈses/ . From a semantic point of view, the shorter word universe seems to be the base of the longer word universal (= that which applies to / is characteristic of the universe). The frequency of use criterion suggests, by contrast, that the longer word universal , which occurs in the BNC 2,570 times, is primary to the shorter word universe , which has only 2,506 attestations in the corpus. To be 6.2 Future work 207 more precise, because the percentage difference between the corresponding frequency numbers is only ~2.52 %, we can say that the back-derivation analysis is in the case under consideration no less plausible than the suffixation analysis. Accordingly, because the longer word universal is / ˌjuːnəˈvɜːsəl◀/ according to LDOCE (the actual stress pattern of universal is thus given occurrences in the modifier position usually / ˈjuːnəvɜːsəl/ ), also the shorter word universe is supposed to be stressed / ˈjuːnəvɜːs/ ( LDOCE ). (The shorter verse -word diverse is, by contrast, as pointed out above, a back-derivative from the longer word diversity and receives therefore final stress. To be more precise, because the shorter diverse occurs in the BNC 1,311 times whereas the longer diversity has 1,394 attestations—the percentage difference is thus only ~6.14 %—we can argue that diverse is a back-derivative from diversity , which, in turn, is, however, a suffixed derivative from diverse .) A similar trisyllabic example is anecdote , which occurs in the BNC 186 times and is thus only marginally more frequent than anecdotal , which has 164 attestations; the percentage difference between these numbers is ~12.57 %. Cf. COCA, where anecdotal occurs 1,856 times whereas anecdote has only 1,027 attestations; the percentage difference is ~57.51 %. Thus, because the shorter word anecdote can be regarded as a back-derivative from the longer word anecdotal , for which LDOCE gives the stress pattern / ˌænɪkˈdəʊtl◀/ (the actual stress pattern of anecdotal is thus often / ˈænɪkdəʊtl/ ), also the derived word anecdote is supposed to be stressed / ˈænɪkdəʊt/ ( LDOCE ), even though in its lefthand disyllabic string a . nec -, the ult is segmentally longer than the penult; stress in this string should thus be final rather than initial. Cases such as these strongly suggest that to better understand the stress patterns of contemporary English words, we need a much better understanding of contemporary English morphology. The standard additive approach, which states that longer English words are made up of shorter ones, is unacceptable. Apart from this, note that the present monograph cannot explain why, e.g., swastika is stressed only / ˈswɒstɪkə/ ( OD ) whereas the variably-stressed paprika prefers penultimate to antepenultimate stress, i.e., / pəˈpriːkə/ rather than / ˈpæprɪkə/ . Is it because by virtue of its association with Nazi Germany, the word swastika is for a contemporary English speaker a more Germanic word than paprika ? Recall also that when an ee -derivative contrasts semantically with an or one (e.g., mortgagee - mortgagor ), both the former and the latter often have emphatic final stress. E.g., just like mortgagee , also mortgagor is stressed not only / ˈmɔːɡᵻdʒə/ , preserving the stress of the base mortgage , but also / ˌmɔːɡᵻˈdʒɔː/ , with stress being emphatically placed upon the formally important suffix or . At the same time, however, when the opposite of an ee -derivative is an er -one, 208 6 Concluding remarks final stress is as a rule exhibited only by the former but not the latter. E.g., while employee is not only / -ˈplɔɪ-/ but also / -ˈiː/ , the only stress pattern of employer is the base stress / -ˈplɔɪ-/ ( OD ). Why are the ate -verbs frustrate , gyrate , locate , migrate , reincarnate , stagnate , truncate , and vibrate as a rule initially-stressed by American English speakers vs. finally-stressed by British English speakers? (For gyrate , locate , and vibrate the OD gives only the finally-stressed transcriptions / dʒʌɪˈrəɪt/ , / lə(ʊ)ˈkeɪt/ , and / vʌɪˈbreɪt/ and in the case of frustrate , migrate , stagnate , and truncate , the finally-stressed transcriptions / frʌˈstreɪt/ , / mʌɪˈɡreɪt/ , / staɡˈneɪt/ , and / trʌŋˈkeɪt/ are given before the initially-stressed alternatives / ˈfrʌstreɪt/ , / ˈmʌɪɡreɪt/ , / ˈstaɡneɪt/ , and / ˈtrʌŋkeɪt/ .) Not only in the BNC but also in the COCA corpus, the longer words frustration , gyrating , location , migration , reincarnation , stagnation , truncation , and vibration have more attestations than the corresponding shorter words frustrate , gyrate , locate , migrate , reincarnate , stagnate , truncate , and vibrate , with the corresponding percentage differences being ~177.77 %, ~95.58 %, ~122.93 %, ~129.86 %, ~171.98 %, ~145.74 %, ~44.25 %, and ~101.45 %. The mean of these values is ~123.69 % and the median is 126.39 %. Given these numbers, there is little doubt that also for an American English speaker, the ate -verbs frustrate , gyrate , locate , migrate , reincarnate , stagnate , truncate , and vibrate should count synchronically as back-derivatives from the corresponding longer ion / ing -word. Why should laboratory be / ləˈbɒrətri/ in British English vs. / ˈlæbrətɔːri/ in American English ( LDOCE )? As for the American pronunciation, it was already observed that stress in laboratory should be initial because both the string or and the string atory occur in English as suffixes (e.g., actor , accusatory ) and are therefore stress-repellent. The British pronunciation / ləˈbɒrətri/ suggests, by contrast, that for a British English speaker, laboratory is morphologically not labor + atory but labora - + tory , with the segmentally longer lefthand trisyllabic string labora -, which ends orthographically in a , receiving penultimate stress. Why is, however, the morphological structure of laboratory different for British and American English speakers? The present monograph is inclined to regard stress differences such as these as idiosyncrasies (i.e., there is no particular reason why an ate -dissyllable such as migrate should necessarily be initially-stressed by American English speakers vs. finally-stressed by British English speakers), but it is of course possible that a future study of English stress will arrive at an entirely different conclusion. Recall also the fact that at 00: 06: 04.526 --> 00: 06: 06.531 of the YouTube video S5hXrgfwK8o, a British English speaker can be heard saying It’s one of those ˈ cartoons for ˈ adults , with retraction of stress occurring in the preceding word cartoons . Is there a special reason why instead of pronouncing the following 6.2 Future work 209 adult with final stress (which the speaker does in many other environments, e.g., since becoming an a ˈ dult and moving out at 00: 04: 42.047 --> 00: 04: 45.307 of the video 9LkG43iZ4u4), the speaker retracts stress in the preceding cartoons ? Is It’s one of those ˈ cartoons for ˈ adults from the point of view of rhythm better than It’s one of those car ˈ toons for a ˈ dults ? From a purely methodological point of view, a limitation of this monograph is that only those words were searched for on YouTube in the case of which stress doublets are given in the OED . Since, however, even the most comprehensive dictionaries do not as a rule include all actually occurring stress patterns, You- Tube searches similar to those reported by the author in this monograph should in the ideal case be run with regard to all (or at least, the most frequently used) polysyllabic English words. Only then will we have a more objective picture of which English words do indeed have stress doublets and which stress pattern in the case of a particular English word with stress doublets is more frequently used in contemporary English. Noteworthy is also the fact that in the study reported in this monograph, only the stress patterns of existing English words (i.e., established English vocabulary, which can be found in dictionaries) were taken into consideration. An objection that can be raised in connection with such words is that their stress patterns are simply stored in English speakers’ mental lexica (e.g., Kettemann 1988: 329), but note that the phenomenon of (especially within-speaker) stress variation is a powerful argument against the listedness-model of English stress; additionally, even if the location of stress in a word is listed in the lexicon, it may nevertheless be in accordance with some rule (e.g., Hayes 1995: 112, arguing that “since English stress is lexically listed, the stress rules serve essentially as redundancy rules, describing the set of possible (as well as unmarked) patterns”). Note also that a recent study by Domahs et al. (2014) has already compared the stress patterns of nonsense words such as serdanuls , huzaldomp , folmadoth , fekomof , etc. and those of existing German, English, and Dutch trisyllabic words (cf. Walch 1972, Baker & Smith 1976, Baptista 1984, Guion et al. 2003, and Turcsan & Herment 2015, who also all conducted stress experiments with potential English words). The conclusion reported by Domahs et al. (2014: 84) is that “in all three languages, the lexical data and the experimental data largely show the same types of effect,” i.e., the stress patterns of nonsense words are in these three Germanic languages by and large similar to those of existing words. Additionally, it was mentioned in 2.4 that not only in existing but also in potential disyllabic English words, an increase in the number of onset consonants in the penultimate syllable correlates statistically with an increase in the incidence of initial stress, i.e., as Kelly (2004: 238) reports, a disyllabic nonsense word such as brontoon is more likely to be stressed initially than a disyllabic nonsense 210 6 Concluding remarks word such as bontoon . Thus, although some of the conclusions drawn in this monograph with regard to existing English words have already been verified by studies of the stress patterns of potential words, an additional study, which will be specifically concerned with, e.g., the segmental length principle as a stress assignment factor in nonsense words, will most certainly not be superfluous. The reason why such an investigation has not been carried out as part of the present monograph is a very practical one. Studies involving potential words obviously require (a not insignificant number of) test persons, whose participation in such studies requires remuneration on the part of the author. Since the research reported in this monograph has not been supported by any university or research institution, the author had neither financial nor even organizational resources (office) for running tests with human subjects. The only research data that was accessible to the author was thus dictionary and corpus data (i.e., YouTube); using this data, the author could only analyze the stress patterns of existing English words. To conclude: Although not each and every aspect of the topic of English stress could be covered in this monograph, it is nonetheless hoped that the monograph has succeeded in answering both 1) the more general question of why (existing) polysyllabic English words are stressed the way they are stressed and 2) the more specialized question of why some polysyllabic English words have more than one stress pattern. 7.2 Software and online tools 211 7 References 7.1 Dictionaries / databases and corpora British National Corpus . http: / / ota.ox.ac.uk/ desc/ 2554. (12 July, 2016.) Cambridge Dictionaries Online . http: / / dictionary.cambridge.org/ . (11 June, 2015.) Corpus of Contemporary American English . http: / / corpus.byu.edu/ coca/ . (21 June, 2015.) Corpus of Global Web-Based English . http: / / corpus.byu.edu/ glowbe/ . (24 November, 2016.) Dictionary.com . http: / / www.dictionary.com/ . (15 April, 2017.) Duden Online . http: / / www.duden.de/ woerterbuch. (09 July, 2017.) Longman Dictionary of Contemporary English . http: / / www.ldoceonline.com/ . (12 December, 2016.) Merriam-Webster Online . http: / / www.merriam-webster.com/ . (11 June, 2015.) MRC Psycholinguistic Database . http: / / ota.ox.ac.uk/ headers/ 1054.xml. (07 February, 2016.) Oxford Dictionaries . http: / / www.oxforddictionaries.com/ . (11 June, 2015.) Oxford English Dictionary . http: / / www.oed.com/ . (20 December, 2013.) Oxford Essential Dictionary of Foreign Terms in English . (Speake & LaFlaur 2002). http: / / tinyurl.com/ ycsrlgsb. (12 September, 2017.) PONS Online-Wörterbuch . http: / / de.pons.com/ . (12 July, 2016.) Russian Orthographic Dictionary . http: / / tinyurl.com/ q6x5jj6. (08 August, 2015.) StressTyp2 . http: / / st2.ullet.net/ ? (21 June, 2015.) TIME Magazine Corpus . http: / / corpus.byu.edu/ time/ . (12 July, 2016.) Word Spy . http: / / www.wordspy.com. (21 June, 2015.) WordNet . http: / / wordnetweb.princeton.edu/ perl/ webwn. (21 June, 2015.) 7.2 Software and online tools ffplay (Version N-68663-g3a2bed1). GNU Grep (Version 2.5.4). Microsoft Excel 2007 . Minitab (Version 15. 1. 30.0). Morphological Analysis . http: / / tinyurl.com/ ohnw7no. (18 July, 2017.) Praat (Version 5.4). TextCrawler (Version 2.5.0.0). YouTube . http: / / www.youtube.com/ . (18 July, 2017.) Youtube-dl . (Versions July 2014-July 2017). 212 7 References 7.3 Literature Adams, C. & R. R. Munro. 1978. In search of the acoustic correlates of stress: fundamental frequency, amplitude, and duration in the connected utterance of some native and non-native speakers of English. Phonetica 35(3). 125-156. Alber, B. 2006. Foot. In Brown, K. (ed.), Encyclopedia of language & linguistics , 539-546. 2nd edn. Boston: Elsevier. Altmann, H. & B. Kabak. 2015. English word stress in L2 and postcolonial varieties: systematicity and variation. In Gut, U., R. Fuchs & E.-M. Wunder (eds.), Universal or diverse paths to English phonology , 185-208. Berlin / Boston: De Gruyter Mouton. Arnold, R. & K. Hansen. 1975. Englische Phonetik . 2nd edn. Munich: Max Hueber Verlag. Aronoff, M. 1976. Word formation in generative grammar . Cambridge, MA : MIT Press. Baer, S. 2008. “Very sexy” constructions of the subject in law. Remarks on a task for the “Language Police.” In Olsen, F., A. Lorz & D. Stein (eds.), Language and law: Theory and society , 347-371. Düsseldorf: Düsseldorf University Press. Baker, R. G. & P. T. Smith. 1976. A psycholinguistic study of English stress assignment rules. Language and Speech 19(1). 9-27. Balteiro, I. 2007. The directionality of conversion in English: A dia-synchronic study . Bern: Peter Lang. Baptista, B. O. 1984. English stress rules and native speakers. Language and Speech 27(3). 217-233. Bauer, L. 2014. Grammaticality, acceptability, possible words and large corpora. Morphology . Published online 15 April 2014. Bell, M. J. & I. Plag. 2012. Informativeness is a determinant of compound stress in English. Journal of Linguistics 48(3). 485-520. Bell, M. J. & I. Plag. 2013. Informativity and analogy in English compound stress. Word Structure 6(2). 129-155. Berg, T. 1999. Stress variation in British and American English. World Englishes 18(2). 123-143. Berg, T. 2008. Emphatic stress shift in German. Zeitschrift für Sprachwissenschaft 27(2). 165-187. Bermúdez-Otero, R. & A. McMahon. 2006. English phonology and morphology. In Aarts, B. & A. McMahon (eds.), The handbook of English linguistics , 382-410. Malden, MA : Blackwell Publishing Ltd. Bryson, B. 1987. Troublesome words . 2nd edn. London: Penguin Books. Burzio, L. 1994. Principles of English stress . Cambridge: Cambridge University Press. Byun, K., W. Kim & H. Ahn. 2014. Reanalysis of the Russian ikan’e. Russian Linguistics . Published online 18 April 2014. Carpenter, A. C. 2010. A naturalness bias in learning stress. Phonology 27(3). 345-392. Carter, A. K. & C. G. Clopper. 2002. Prosodic effects on word reduction. Language and Speech 45(4). 321-353. 7.3 Literature 213 Chitoran, I. & J. I. Hualde. 2007. From hiatus to diphthong: the evolution of vowel sequences in Romance. Phonology 24(1). 37-75. Cho, T., J. M. McQueenb & E. A. Cox. 2007. Prosodically driven phonetic detail in speech processing: The case of domain-initial strengthening in English. Journal of Phonetics 35(2). 210-243. Chomsky, N. & M. Halle. 1968. The sound pattern of English . New York: Harper & Row. Chrabaszcz A, M. Winn, C. Y. Lin & W. J. Idsardi. 2014. Acoustic cues to perception of word stress by English, Mandarin, and Russian speakers. Journal of Speech Language and Hearing Research 57(4). 1468-1479. Colina, S. 2012. Syllable structure. In Hualde, J. I., A. Olarrea & E. O’Rourke (eds.), The handbook of Hispanic linguistics , 133-151. Malden, MA : Blackwell Publishing Ltd. Cruttenden, A. 2008. Gimson’s pronunciation of English . 7th edn. London: Hodder Education. Cutler, A. 2015. Lexical stress in English pronunciation. In Reed, M. & J. M. Levis (eds.), The handbook of English pronunciation , 106-124. Malden, MA : John Wiley & Sons, Inc. Davis, S. 1988. Syllable onsets as a factor in stress rules. Phonology 5(1). 1-19. de Haas, W. G. 1988. Phonological implications of skeleton and feature underspecification in Kasem. Phonology 5(2). 237-254. Dehaene-Lambertz, G. & D. Houston. 1998. Faster orientation latencies toward native language in two-month-old infants. Language and Speech 41(1), 21-43. Domahs, U., E. Klein, W. Huber & F. Domahs. 2013. Good, bad and ugly word stress - f MRI evidence for foot structure driven processing of prosodic violations. Brain & Language 125. 272-282. Domahs, U., I. Plag & R. Carroll. 2014. Word stress assignment in German, English and Dutch: quantity-sensitivity and extrametricality revisited. The Journal of Comparative Germanic Linguistics . 17. 59-96. Farnetani, E., C. T. Torsello & P. Cosi. 1988. English compound versus non-compound noun phrases in discourse: an acoustic and perceptual study. Language and Speech 31(2). 157-180. Field, J. 2003. Promoting perception: lexical segmentation in L2 listening. ELT Journal 57(4). 325-334. Foulkes, P. & G. Docherty. 2007. Phonological variation in England. In Britain, D. (ed.), Language in the British Isles , 52-74. Cambridge: Cambridge University Press. Fournier, J.-M. 2007. From a Latin syllable-driven stress system to a Romance versus Germanic morphology-driven dynamics: in honour of Lionel Guierre. Language Sciences 29(2-3). 218-236. Friederich, W. 1967. Englische Aussprachelehre. Das Verhältnis von Aussprache und Schreibung . 2nd rev. edn. Munich: Max Hueber Verlag. Fry, D. B. 1955. Duration intensity as physical correlates of linguistic stress. Journal of the Acoustical Society of America 32. 765-769. Fry, D. B. 1958. Experiments in the perception of stress. Language and Speech 1(2). 126-152. 214 7 References Fudge, E. C. 1984. English word-stress . London: George Alien & Unwin (Publishers) Ltd. Gandour, J. T. 1998. Phonetics and phonology. In Stemmer, B. & H. A. Whitaker (eds.), Handbook of neurolinguistics , 207-219. San Diego: Academic Press. Gay, T. 1978. Physiological and acoustic correlates of perceived stress. Language and Speech 21(4). 347-353. Gimson, A. C. 1970. An introduction to the pronunciation of English . 2nd edn. London: Edward Arnold. Gimson, A. C. 1972. Foreword. In Lewis, J. W. 1972. A concise pronouncing dictionary of British and American English , v. London: Oxford University Press. Guion, S. G., J. J. Clark, T. Harada & R. P. Wayland. 2003. Factors affecting stress placement for English nonwords include syllabic structure, lexical class, and stress patterns of phonologically similar words. Language and Speech 46(4). 403-427. Gussenhoven, C. 2006. Between stress and tone in Nubi word prosody. Phonology 23(2). 192-223. Gut, U. 2009. Introduction to English phonetics and phonology . Frankfurt am Main: Peter Lang. Hall, T. A. 2006. Syllable: phonology. In Brown, K. (ed.), Encyclopedia of language & linguistics , 329-333. 2nd edn. Boston: Elsevier. Halle, M. & J.-R. Vergnaud. 1987. An essay on stress . Cambridge, MA : The MIT Press. Halle, M. & S. J. Keyser. 1971. English stress: its form, its growth, and its role in verse . New York: Harper & Row, Publishers. Halverson, J. 1967. Stress, pitch, and juncture. ELT Journal XXI (3). 210-217. Hamawand, Z. 2011. Morphology in English. Word formation in cognitive grammar . London: Continuum International Publishing Group. Hammond, M. 2006. Prosodic phonology. In Aarts, B. & A. McMahon (eds.), The handbook of English linguistics , 411-432. Malden, MA : Blackwell Publishing Ltd. Hayes, B. 1995. Metrical stress theory: principles and case studies . Chicago: The University of Chicago Press. Henderson, A. 2010. A corpus-based, pilot study of lexical stress variation in American English. Research in Language 8. 99-113. Hickey, R. 2007. Southern Irish English. In Britain, D. (ed.), Language in the British Isles , 135-151. Cambridge: Cambridge University Press. Hieke, A. E. 1984. Linking as a marker of fluent speech. Language and Speech 27(4). 343-354. Hill, A. A. 1974. Word stress and the suffix ic . Journal of English Linguistics 8. 6-18. Huss, V. 1978. English word stress in the post-nuclear position. Phonetica 35(2). 86-105. Hyman, L. M. 1977. On the nature of linguistic stress. Southern California Occasional Papers in Linguistics 4. 37-82. Jensen, J. 1993. English phonology . Amsterdam / Philadelphia: John Benjamins Publishing Company. Jones, D. 1917. Explanations. In Jones, D. 1974[1967]. Everyman’s English pronouncing dictionary , xxii-xliv. 13th edn (edited by A. C. Gimson). London: J. M. Dent & Sons LTD . 7.3 Literature 215 Kager, R. 1995. Review article. Phonology 12(3). 437-464. Kelly, M. H. 2004. Word onset patterns and lexical stress in English. Journal of Memory and Language 50(3). 231-244. Kent, R. & R. Netsell. 1971. Effects of stress contrasts on certain articulatory parameters. Phonetica 24(1). 23-44. Kenyon, J. S. & T. A. Knott. 1944. Introduction. In Kenyon, J. S. & T. A. Knott. 1953[1944]. A pronouncing dictionary of American English , xv-xlix. Springfield, Mass.: G. & C. Merriam Company, Publishers. Kettemann, B. 1988. Die Phonologie morphologischer Prozesse im amerikanischen Englisch . Tübingen: Gunter Narr Verlag. Kingdon, R. 1949. The teaching of English stress. ELT Journal III (6). 146-152. Kösling, K. 2013. Prominence assignment in English triconstituent compounds . Siegen: University of Siegen dissertation. http: / / bit.ly/ 2u6su2x. (9 July, 2017.) Kösling, K., G. Kunter, H. Baayen & I. Plag. 2013. Prominence in triconstituent compounds: pitch contours and linguistic theory. Language and Speech 56(4). 529-554. Kraska-Szlenk, I. & M. Zygis. 2012. Phonetic and lexical gradience in Polish prefixed words. Cognitive Linguistics 23(2). 317-366. Kreidler, C. W. 1979. Creating new words by shortening. Journal of English Linguistics 13. 24-36. Kretzschmar, W. A. 2010. The development of Standard American English. In Kirkpatrick, A. (ed.), The Routledge handbook of World Englishes , 96-112, Abingdon: Routledge. Krivokapić, J. 2013. Rhythm and convergence between speakers of American and Indian English. Laboratory Phonology 4(1). 39-65. Kunter, G. 2011. Compound stress in English: the phonetics and phonology of prosodic prominence . Berlin: Walter de Gruyter. Langendoen, T. D. 1975. Some problems in the description of English accentuation. In Goyvaerts, D. L. & G. K. Pullum (eds.), Essays on the Sound Pattern of English , 205-218. Ghent: E. Story-Scientia. Levenshtein, V. I. 1966. Binary codes capable of correcting deletions, insertions, and reversals. Soviet Physics-Doklady 10(8). 707-710. Lieber, R. 2010. Introducing morphology . Cambridge: Cambridge University Press. Lutz, A. 2009. Word accent position and language contact in English and German. Anglia 127(2). 283-306. Marks, J. 1999. Is stress-timing real? ELT Journal 53(3). 191-199. Mcclean, M. D. & W. R. Tiffany. 1973. The acoustic parameters of stress in relation to syllable position, speech loudness and rate. Language and Speech 16(3). 283-290. Mel’čuk, I. A. 2001. Kurs obščej morfologii: morfologičeskie znaki [ Course in general morphology: morphological signs ]. Moscow / Vienna: Wiener Slawistischer Almanach. Mel’čuk, I. A. 2012. Semantics. From meaning to text . Amsterdam: John Benjamins. Mompéan, J. A. 2014. Stress shift in English: The case of the teen numbers. In Monroy-Casas, R. & I. Arboleda-Girao (eds.), Readings in English phonetics and phonology , 149-164. Valencia: University of Valencia. 216 7 References Morton, J. & W. Jassem. 1965. Acoustic correlates of stress. Language and Speech 8(3). 159-181. Mücke, D. & M. Grice. 2014. The effect of focus marking on supralaryngeal articulation - Is it mediated by accentuation? Journal of Phonetics 44 (3). 47-61. Niebuhr, O. 2007. The signaling of German rising-falling intonation categories - the interplay of synchronization, shape, and height. Phonetica 64(2-3). 174-193. Pater, J. 2000. Non-uniformity in English secondary stress: the role of ranked and lexically specific constraints. Phonology 17(2). 237-274. Payne, E., B. Post, L. Astruc, P. Prieto & M. del Mar Vanrell. 2012. Measuring child rhythm. Language and Speech 55(2). 203-229. Pearson, K. 1900. On the criterion that a given system of deviations from the probable in the case of a correlated system of variables is such that it can be reasonably supposed to have arisen from random sampling. Philosophical Magazine Series 50(5). 157-175. Peng, L. & J. Ann. 2001. Stress and duration in three varieties of English. World Englishes 20(1). 1-27. Pettersson, T. 1982. Review article. Russian Linguistics 7. 63-72. Plag, I. 2003. Word-formation in English . Cambridge: Cambridge University Press. Plag, I, G. Kunter & M. Schramm. 2011. Acoustic correlates of primary and secondary stress in North American English. Journal of Phonetics 39(3). 362-374. Plag, I., G. Kunter, S. Lappe & M. Braun. 2008. The role of semantics, argument structure, and lexicalization in compound stress assignment in English. Language 84(4). 760-794. Poldauf, I. 1984. English word stress: a theory of word-stress patterns in English . Oxford: Pergamon Press. Scarborough, R., P. Keating, S. L. Mattys, T. Cho & A. Alwan. 2009. Optical phonetics and visual perception of lexical and phrasal stress in English. Language and Speech 51(2 / 3). 135-175. Scherer, G. & A. Wollmann. 1972. Englische Phonetik und Phonologie . Berlin: Erich Schmidt Verlag. Schmid, H.-J. 2008. New words in the mind: concept-formation and entrenchment of neologisms. Anglia 126(1). 1-36. Schmid, H.-J. 2011. English morphology and word-formation . An introduction . 2nd edn. Berlin: Erich Schmidt Verlag. Simo Bobda, A. 2010. Word stress in Cameroon and Nigerian Englishes. World Englishes 29(1). 59-74. Sokolova, M. A., K. P. Gintovt, I. S. Tikhonova & R. M. Tikhonova. 1996. English phonetics. A theoretical course . Moscow: VLADOS . Sokolova, M. A., K. P. Gintovt, L. A. Kanter, N. P. Krylova, I. S. Tikhonova & G. A. Shabadash. 1997. Prakticheskaja fonetika anglijskogo jazyka [ Practical phonetics of the English language ]. Moscow: VLADOS . Stadnik-Holzer, E. 2009. Artikulatorische Phonetik. In Kempgen, S., P. Kosta, T. Berger & K. Gutschmidt (eds.), Die slavischen Sprachen. Ein internationales Handbuch zu 7.3 Literature 217 ihrer Struktur, ihrer Geschichte und ihrer Erforschung. Band 1 , 14-48. Berlin: Walter de Gruyter. Stašková, N. 2012. English back-formation in the 20th and the early 21st centuries. Linguistica Pragensia 22(2). 98-121. Štekauer, P. 1998. An onomasiological theory of English word-formation . Amsterdam / Philadelphia: John Benjamins Publishing Company. Tan, Y.-Y. 2015. ‘Native’ and ‘non-native’ perception of stress in Singapore English. World Englishes 34(3). 355-369. Tokar, A. 2012. Introduction to English morphology . Frankfurt am Main: Peter Lang. Torsuev, G. P. 1960. Voprosy aktsentologii sovremennogo anglijskogo jazyka [ Problems of accentuation in Modern English ]. Moscow: The Academy of Sciences of the USSR Press. Trevian, I. 2007. Stress-neutral endings in contemporary British English: an updated overview. Language Sciences 29(2-3). 426-450. Trevian, I. 2015. English suffixes: Stress-assignment properties, productivity, selection and combinatorial processes . Bern: Peter Lang. Turcsan, G. & S. Herment. 2015. Making sense of nonce word stress in English. In Mompean, J. A. & J. Fouz-González (eds.), Investigating English pronunciation: Trends and directions , 23-46. Basingstoke: Palgrave Macmillan. Turk, A.E & J. R. Sawusch. 1997. The domain of accentual lengthening in American English. Journal of Phonetics 25(1). 25-41. Turk, A. E., P. W. Jusczyk & L. Gerken. 1995. Do English-learning infants use syllable weight to determine stress? Language & Speech 38(2). 143-158. van der Hulst, H. 2010a. Word accent systems in the languages of Europe. In van der Hulst, H., R. Goedemans & E. van Zanten (eds.), A survey of accentual patterns in the languages of the world , 429-507. Berlin: Walter de Gruyter. van der Hulst, H. 2010b. Word accent: terms, typologies and theories. In van der Hulst, H., R. Goedemans & E. van Zanten (eds.), A survey of accentual patterns in the languages of the world , 3-53. Berlin: Walter de Gruyter. van der Hulst, H. 2012. Deconstructing stress. Lingua 122(3). 1494-1521. van der Hulst, H., R. Goedemans & E. van Zanten (eds.). 2010. A survey of accentual patterns in the languages of the world . Berlin: Walter de Gruyter. Van Rooy, B. 2002. Stress placement in Tswana English: the makings of a coherent system. World Englishes 21(1). 145-160. van Zanten, E., R. Stoel & B. Remijsen. 2010. Stress types in Austronesian languages. In van der Hulst, H., R. Goedemans & E. van Zanten (eds.), A survey of accentual patterns in the languages of the world , 87-111. Berlin: Walter de Gruyter. Ventcel’, E. S. 1969. Teorija verojatnostej [ Probability theory ]. 4th edn. Moscow: Nauka. Walch, M. L. 1972. Stress rules and performance. Language and Speech 15(3). 279-287. Wells, J. C. 1990. Longman pronouncing dictionary . Harlow: Longman. Wells, J. C. 2008. Introduction. In Wells, J. C. Longman pronunciation dictionary , xvii-xviii. 3rd edn. Harlow: Longman. 218 7 References Wenk, B. J. 1985. Speech rhythms in second language acquisition. Language and Speech 28(2). 157-175. Wenszky, N. 2004. Secondary stress in English words . Budapest: Akadémiai Kiadó. Wetzels, W. L. 1995. Mid-vowel alternations in the Brazilian Portuguese verb. Phonology 12(2). 281-304. Wilbur, R. & B. Schick. 1987. The effects of linguistic stress on ASL signs. Language and Speech 30(4). 301-323. Wilson, M. 1988. MRC Psycholinguistic Database: Machine-usable dictionary, version 2.00. Behavior Research Methods, Instruments, & Computers 20 (1). 6-10. Wiltshire, C. & R. Moon. 2003. Phonetic stress in Indian English vs. American English. World Englishes 22(3). 291-303. Yavaş, M. 2011. Applied English phonology . 2nd edn. Malden, MA : Wiley-Blackwell. 8 Appendix 219 8 Appendix The following is the more frequently heard stress patterns of the words that were pronounced in captioned YouTube videos by at least one native English speaker. Across-varietal stress differences are taken into consideration (e.g., a ˈ dult vs. ˈ adult in Br.E.), but remember that, as pointed out in 5.2, General American was attested far more often than other native English accents (i.e., Received Pronunciation, General Australian, etc.). The stress patterns given below are thus biased towards American English. E.g., ˈ abdomen means that it was mainly American English speakers who were heard by the author to have pronounced abdomen with antepenultimate stress. açaˈi vs. aˈçai in Br.E. ˈAbba Abˈbasid ˈabdomen Abˈdul ˈaberrance aˈberrancy in Br.E. ˈaberrant ˈAbib ˈAbipon ˈabsolute ˈabsolutely ˈabsoluteness absoˈlutism absoˈlutist ˈabsolutive ABTA / ˌeɪˌbiˌtiˈeɪ/ in Am.E. vs. / ˈabtə/ in Br.E. ˈAcademe aˈcaricide ˈaccent , v. ˈaccented ˈaccenting ˈaccess , v. ˈaccolade acˈcouchement acˈcusatory aˈceto - ˈacetyl Aˈchillean aˈchondrite aˈcorus ˈacrolein ˈActaeon ˈacumen Aˈdar adˈdress aˈdept aˈdeptly aˈdeptness Adiˈvasi ˈadmixture ˈAdonai ˈadulatory aˈdult vs. ˈadult in Br.E. aˈdulthood vs. ˈadult in Br.E. aˈdultlike aˈdultness ˈadumbrate ˈadversary 220 8 Appendix ˈadverse adˈversely adˈverseness ˈadvertisement vs. adˈvertisement in Br.E. ˈAdyghe aedeˈagus Aeˈneid Aeˈolus vs. ˈAeolus in Br.E. aeˈquorin ˈaestival afˈfixed ˈaffluence ˈaffluent ˈafrit aˈgape aˈgaric aiˈgrette aiˈkido ˈakinete al ˈQaeda vs. Qaˈeda in Br.E. ˈAlawite ˈalcazar vs. alˈcazar in Br.E. Alcoˈran Alenˈçon ˈaleurone ˈAleut algeˈbraist Alˈlah alˈlele vs. ˈallele in Br.E. alˈlelism ˈallied ˈalloy , v. ˈalloying ˈally , n. and v. ˈallying alˈnico ˈAltair alˈthea alˈti vs. ˈalti in Br.E. alˈveolar vs. ˈalveolar in Br.E. vs. alveˈolar in Aus.E. alˈveolus aˈlyssum ˈamateur dramatics ˈamateur ˈamide aˈmine vs. ˈamine in Br.E. aˈmino - ˈamperage ˈampere ˈampersand amylogluˈcosidase Aˈnancy ˈanchovy ˈancillary vs. anˈcillary in Br.E. anecˈdotist anˈgina ˈannex Aˈ NOVA anteˈcessor antiˈhelix Anˈtillean ˈantique vs. anˈtique in Br.E. aˈpophyllite apotheˈosis aˈpotheosize apˈpetitive vs. appeˈtitive in Br.E. ˈapplicable vs. - ˈplicable in Br.E. ˈappliqué ˈappliquéd ˈAquila ˈarchetypal Archiˈmedean arˈchival aˈristocrat Aristoˈtelean ˈarmillary ˈaroha vs. aˈroha in Br.E. ˈarrack 8 Appendix 221 ˈarsine Arthroˈpoda arˈticulatory ˈartificer vs. arˈtificer in Br.E. asap / eɪɛseɪˈpiː/ Asˈkari ˈasphalt ˈaspirant ˈassignat associˈation ˈatoll Atˈtila aˈtua ˈAugustin ˈautomatize ˈavatar Bacˈchante ˈbadinage ˈBalthazar baˈnal ˈbandore ˈBantu ˈBantustan barˈbiturate baˈsalt basˈmati vs. ˈbasmati in Br.E. ˈbastide baˈstille baˈtik Beˈguine Berˈlin berˈserk ˈbezant bicenˈtenary bioengiˈneer ˈbitumen ˈBodleian Bougainˈvillaea bouˈquet boˈyar ˈboyau briˈquette bucenˈtaur ˈbuddleia ˈbureau ˈburpee ˈByzantine vs. Byˈzantine in Br.E. C. diffiˈcile Cabˈbala cadaˈveric ˈcaftan ˈcaisson ˈcalcine caˈlicivirus ˈcanine ˈcanton ˈcapillary vs. caˈpillary in Br.E. ˈCapitoline ˈcaravan ˈcarburetted carˈcinogen ˈcardio - Caribˈbean ˈcaribou ˈcarillon caˈruncle ˈcashmere castaˈnet ˈcatamaran caˈthedra Caˈtholicism caˈtholicize ceˈlestite ceˈment cenˈtaurea cenˈtrifugal vs. - ˈfugal in Br.E. ˈceruse ˈcervical chaˈgrin chalˈcedony 222 8 Appendix ˈchallah ˈcharactery ˈcheechako ˈCherokee chimpanˈzee chiˈvalric Chloromyˈcetin choˈral ˈchrono - ˈcimarron clariˈnet vs. ˈclarinet in Br.E. ˈclarionet ˈclassificatory cliˈmacteric colˈleen ˈcomatose ˈcombat , v. ˈcomment , v. ˈcommenter comˈmunal comˈmunalism comˈmunalize comˈmunally ˈcommune communiˈqué vs. comˈmuniqué in Aus.E. comˈmutative ˈcompact , adj. ˈcompensate comˈplaisance comˈplaisant comˈposite vs. ˈcomposite in Br.E. ˈcomposter ˈconcentrate ˈconfiscate ˈCongolese ˈconjure conˈservator conˈservatory ˈconserve , n. conˈsistory ˈconstative ˈconstellate conˈstrue conˈstruing ˈconsult vs. conˈsult in Br.E., n. ˈconsummate ˈconsummately ˈcontact , v. ˈcontemplate ˈcontent , n. ˈcontour ˈcontoured conˈtrariness ˈcontrovert vs. - ˈvert in Br.E. controˈvertist conˈversant conˈversely ˈcorduroy ˈcorollary vs. coˈrollary in Br.E. coˈronal vs. ˈcoronal in Br.E. ˈcoronoid ˈcorpuscle corˈrosive ˈcostume ˈCoulomb cricoaˈrytenoid ˈcroupier crystalˈlo in crystallography and crystallographer cuˈneiform cuˈrator ˈcurvet curˈveting vs. ˈcurvetting in Br.E. ˈcushaw Cyˈclopean Cyˈperus ˈdada ˈDanakil debiˈtage 8 Appendix 223 ˈdecadence ˈdecadent ˈdecollate ˈdecorous ˈdecoy , n. ˈdecrease , n. ˈdefect , n. deˈfile , n. dehyˈdrogenate deˈmonstrable ˈdemonstrate ˈdenim ˈdepot deˈpurate ˈdesiccant ˈdesiccate ˈdesiccated ˈdesiccator ˈdetail ˈdetailed ˈdetour Deuteˈronomy ˈdictatory diffiˈcile digoxiˈgenin ˈdiktat vs. dikˈtat in Aus.E. diˈnar diˈnette diˈplodocus ˈdischarge , n. ˈdiscount vs. disˈcount in Br.E., v. ˈdiscourse disˈcrepancy disˈcrepant disˈputable disˈsoluble diˈvan diˈverse diˈversely ˈdoctrinal doˈdecyl Doneˈgal in Ir.E. vs. ˈDonegal in Am.E. ˈdrabbet Dulciˈnea duˈress ˈdurian eˈchino - E ˌ chinoderˈmata electroˈmetallurgy eloˈdea eˈlongate vs. ˈelongate in Br.E. eˈlongated emˈbryonal eˈmir ˈemirate emˈployee emˈpyreal emˈpyrean ˈenclave enˈclaved only in Aus.E. ˈendo - ˈenema ˈenqueue , n. ˈenthalpy ˈentomo epiˈtaxy ˈequerry eˈquilibrate eˈquilibrist eˈrythrose esˈcapee vs. escaˈpee in Br.E. ˈescritoire ˈescrow , v. ˈethynyl Euˈclidean euˈpatrid eˈvocative ˈexcerpt ˈexcise , n. exˈciseman 224 8 Appendix exˈcitant ˈexciton ˈexcretory vs. exˈcretory in Br.E. ˈexculpate eˈxedra eˈxergue ˈexo - ˈexorcism ˈexorcist ˈexorcize ˈexpert ˈexpertly ˈexpertness ˈexpletive vs. exˈpletive in Br.E. ˈexport vs. exˈport in Br.E., v. ˈexpurgate vs. exˈpurgate in Br.E. exˈquisite ˈextant ˈextirpate exˈtraordinary exˈtrapolate ˈfakir ˈfanfare fanˈtasia farˈfalle ˈFarsi ˈFaverolle ˈfeaturette ˈfiligree ˈfisetin ˈflagellant ˈflageolet fleuˈret ˈforbear , n. ˈforecast , v. ˈforecaster ˈforecasting ˈfragment , v. ˈfragmented vs. - ˈmented in Br.E. ˈfrustrate vs. frusˈtrate in Br.E. ˈfrustrated vs. frusˈtrated in Br.E. Fuˈlani fuˈsilli ˈgalago ˈgallant ˈgallantly ˈgallivant ˈgallivanting ˈgamete gaˈmetophyte Gaˈrifuna ˈgasoline vs. gasoˈline in Br.E. ˈgenocidal geˈographer vs. ˈge in Br.E. geˈography vs. ˈge in Br.E. geˈometrist geˈometry vs. ˈge in Br.E. Germiˈnal ˈgingiva gladiˈolus ˈglobose gluˈcosidase vs. - ˈsidase in Br.E. glyˈceric gouˈache gouˈrami ˈguano ˈGuatemalan ˈguillotine ˈguillotining ˈGulag ˈgunnera ˈguru ˈgyrate vs. gyˈrate in Br.E. ˈhabergeon Haˈdean Halˈlel ˈhalva ˈHanover harˈleian only in Br.E./ Aus.E. ˈharmattan vs. harˈmattan in Br.E. 8 Appendix 225 haˈshish vs. ˈhashish in Br.E. Hebriˈdean heˈgemony vs. ˈhe in Aus.E. heˈgira heˈpato - ˈheresiarch ˈheroin hiˈbachi hilˈlo Himaˈlayan ˈHindu hogmaˈnay ˈhoisin ˈholla ˈhomage ˈhomo homeoˈstasis Honˈduras horˈmonal hoˈtelier human ˈresources vs. reˈsources in Br.E. hyˈdrogenate hyˈgienist hyˈpericum hypoˈiodous ˈIban ˈibidem idemˈpotence ˈillustrate ˈillustrated ˈimbecile ˈimide imˈmuno immunoˈmodulatory imˈpactive ˈimport , n. imporˈtune ˈimprecatory imˈpregnate imˈpregnated vs. ˈim in Aus.E. imˈpregnator ˈimprovise ˈincitant ˈincline , n. incogˈnita incogˈnito ˈincrease , n. ˈinculcate inˈculpate inˈdecorous ˈindent , n. inˈdexible inˈdicative indisˈputable indisˈsoluble inˈexpert inˈferable ˈinfiltrate ˈinfiltrated ˈIngush ˈinlaid ˈinlay , n. inˈnate inopporˈtune ˈinspiratory ˈinstrument , v. interˈdisciplinary interˈmaxillary inˈterpellate inˈterstice inˈtuit ˈinundate ˈinvalid ˈinvalidism ˈinverse ˈinvite , n. iˈodophor iˈrate irreconˈcilable irreˈfutable 226 8 Appendix ˈIslam isoˈetes ˈisoprenoid ˈIvan jacaˈranda Jaˈcobi Jacˈquard vs. ˈJacquard in Br.E. jerˈboa jeuˈnesse ˈjudicatory kaiˈseki kaˈnaka ˈkaramu ˈKashmir Kerˈguelen kiˈbitzer kiˈbosh kiˈlometer kineˈmatograph ˈkiosk Koˈran Laˈotian ˈlardon larˈgess ˈlascar ˈlasso lavaˈtera ˈLeghorn ˈlegume ˈleprechaun ˈlessor ˈLevantine ˈlevee liˈatris liˈchenified ˈlocate ˈlycanthrope Maˈcabre macéˈdoine ˈmadame ˈmadeleine ˈMadras vs. Maˈdras in Aus.E. Mafˈfei ˈmaghrib maˈguey ˈmaguro ˈmajuscule ˈMakarov ˈMaksutov ˈmalapert Maˈlay ˈmaleate ˈmalvoisie ˈmama ˈmanatee manˈcala ˈmandarin ˈmandate , v. ˈmandated ˈmandolin manˈdorla mandraˈgora ˈmanganese ˈManueline maˈnuka ˈmaquis vs. maˈquis in Br.E. maˈrae ˈMarie Celeste maˈrie market ˈresearch market ˈresearcher ˈmarmoset marˈquis ˈMasai vs. Maˈsai in Br.E. ˈMascarene ˈmasseter masˈsif ˈmasticatory Matigˈnon matˈsuri 8 Appendix 227 maˈtutinal maxiˈmin ˈmayonnaise mayˈoral vs. ˈmayoral in Br.E. ˈmayoress meˈdicinable ˈmedullary vs. meˈdullary in Br.E. megagaˈmetophyte Meˈgillah ˈmelaena meˈlanocyte meˈlanophore meˈlanosome ˈmêlée ˈmembranous menaquiˈnone Méniˈère meˈningeal Mephistoˈphelean vs. - ˈtophelean in Br.E. ˈmercantilist vs. - ˈcantilist in Br.E. mercantiˈlistic ˈmessaline meˈtallo meˈtallophone ˈmetallurgist ˈmetallurgy vs. - ˈtallurgy in Br.E. metaˈmorphose metaˈmorphosed metaˈmorphosing metaˈmorphosis meˈthanotroph ˈmetheglin vs. - ˈtheglin in Aus.E. ˈmethylamine ˈmétier ˈmetis méˈtisse Meurˈsault micro-entrepreˈneur microˈglia ˈmigrate vs. miˈgrate in Br.E. ˈmigrator vs. miˈgrator in Br.E. ˈmigratory ˈmihrab miˈkan miˈlieu miliˈtarily ˈmillenary ˈmillionaire miˈmesis minaˈret vs. ˈminaret in Br.E. ˈminuscule ˈminyan miˈrage ˈmiscellany vs. - ˈcellany in Br.E. misdiagˈnose miˈsericord Miˈskito ˈmistral ˈmodulatory moiˈdore vs. ˈmoidore in Br.E. mokoˈpuna Molˈdovan ˈmomentarily Moˈnera moˈneran moneˈtarily ˈmonophthong moˈnotropism monˈsignor Monˈstera ˈMontrachet ˈmoray ˈmortgagor moˈshav Mosˈsad mousseˈline ˈmoustache vs. - ˈstache in Br.E. ˈmoustached vs. - ˈstached in Br.E. ˈmoustacheless 228 8 Appendix ˈmudra muˈezzin multiˈdisciplinary multiˈlamellar ˈmultimerize ˈmultiplicand multiˈplicative ˈmurderess Muscaˈdet musculoˈskeletal ˈmustelid Myceˈnaean vs. - ˈcenaean in Br.E. myˈdriasis myˈgalomorph ˈNahuatl naˈkhoda namaˈste narˈghile Naˈskapi nasopharynˈgeal ˈnatheless ˈNazarene vs. Nazaˈrene in Br.E. Ndeˈbele necesˈsarily ˈnecropsy neˈcrose ˈnegligée neˈmaticide neˈmatocyst ˈNemean vs. Neˈmean in Br.E. neuroˈglia neuroˈglial neuˈronal ˈNicene ˈNichiren ˈNikkei nincomˈpoopery ˈnisei ˈnitriding niˈtrosyl Niˈzam ˈNobel ˈnocturne non-ˈabsolute non-comˈbatant nonpaˈreil non-ˈpartisanship non-ˈrecourse ˈnouveau nucleˈolus Nyˈasa vs. ˈNyasa in South Afr.E. nymˈphalid nymˈphet ˈobit ˈoblate obsoˈlete vs. ˈobsolete in Br.E. ocˈcult vs. ˈoccult in Br.E. ocˈculted ocˈculting ocˈcultist ˈoctoate ocˈtuple ocˈtuplet ˈodalisque Odoˈnata oeˈnothera ofˈfense ofˈfenseful ofˈfenseless ofˈfensive ofˈfensively ofˈfensiveness oˈfuro ˈogive oˈlecranon vs. oleˈcranon in Br.E. Oˈligocene oˈligomer vs. ˈoligomer in Aus.E. ˈolivine oˈmertà ooˈkinete 8 Appendix 229 oˈpercle Oˈphiuchus ˈopportune vs. opporˈtune in Br.E. opporˈtunely opporˈtuneness opporˈtunism opporˈtunist ˈoptative oˈrangutan orˈgano - ˈorganoid ˈornithopod ˈornithopter Orˈphean ˈOsage overdiagˈnose oˈvert oˈvertly oˈvertness oˈxalis ˌ oxaloaˈcetic ˈoxygenate ˈoxygenated ˈoxygenator ˈoxygenize ˈozonide ˈ PABA / ˈpæbə/ paeˈdiatrist ˈPaiute ˈpakhawaj paˈlagi ˈpalanquin ˈpalatal palˈlavi Palˈmach ˈpalmitoyl Palˈmyrene ˈpalpating Panaˈgia ˈPanama Papaˈnicolaou ˈpapillary vs. paˈpillary in Br.E. paˈprika par excelˈlence ˈParaguay ˈparakeet Paˈranthropus paˈrashah paˈreu parfumeˈrie paroˈxysmal ˈParsee parˈticipatory vs. - ˈpatory in Br.E. ˈparticiple vs. - ˈticiple in Br.E. ˈpartisan ˈpartisanship ˈPascal ˈpassé paˈstis ˈpastoral paˈtagium patˈchouli paˈtrie ˈpatron patroˈnat ˈpatroness ˈPauillac paˈvane pavˈlova Pawˈnee ˈPeano Peaucelˈlier peˈcan vs. ˈpecan in Br.E. peˈdiculicide ˈpedregal ˈpeduncle ˈpeekaboo ˈpeekapoo ˈpeeress ˈPegu 230 8 Appendix peˈjorative ˈpembina ˈpenstemon vs. - ˈstemon in Br.E. perˈfume vs. ˈperfume in Br.E. perˈfumed perˈfumer periˈodogram ˈPerrier ˈpersecutory vs. - ˈcutory in Br.E. ˈpersiflage Pétaiˈnist ˈpetiolate ˈpetit vs. peˈtit in Br.E. phalanˈgeal pharmaˈcopoeial vs. - ˈpoeial in Br.E. pharynˈgeal vs. - ˈryngeal in Br.E. ˈPhascogale ˈphentermine ˈphilosophe ˈphlorizin phosphaˈtidyl phosˈphoramide phosˈphorylase photomonˈtage phylˈloxera Phyˈsalis phyˈsiatrist phyˈsiatry ˈpiccaninny ˈPiegan ˈpiezo ˈpigment , v. ˈpigmented ˈpigmenting ˈpilaf piˈlau piˈnacolone piˈneal piˈñon ˈpinyon ˈpipistrelle piroˈzhok ˈpissant ˈpistole ˈpistolet ˈpitot ˈplanigale plasmapheˈresis ˈplateau vs. plaˈteau in Aus.E. plaˈteaued plaˈteauing ˈplatina pleˈthoric vs. ˈplethoric in Br.E. pneuˈmato pneuˈmatocele pneuˈmatophore ˈpoetess vs. poeˈtess in Br.E. Pohnˈpeian pohutuˈkawa ˈpoikilo - ˈpoikilotherm ˈpolygonize poˈlymerase poˈlymerize poˈlyptych poˈmander poˈmelo Pomˈmard ˈpompadour ponˈceau ˈpongee popliˈteal popliˈteus ˈportage ˈportaging portmanˈteau vs. portˈmanteau in Br.E. Port-ˈRoyal ˈposthumous ˈposthumus poˈtage 8 Appendix 231 ˈPowhatan Prairiˈal ˈprajna praˈxinoscope ˈpreamble vs. preˈamble in Br.E. ˈprecedent , adj. preceˈrebral ˈprecip ˈprecog pre-conˈtract , v. ˈprefab ˈprefixed ˈprefixing ˈprelim ˈpremature premaˈturely ˈpremise , v. ˈpreparative vs. - ˈparative in Br.E. preˈparator ˈpresage , n. and v. preˈsager preˈsaging ˈprescript préˈvôt ˈpriestess priˈmarily ˈprincess ˈprioress priˈoritize vs. ˈprioritize in Br.E. priˈstine vs. ˈpristine in Br.E. ˈprivative progˈnathous proˈgrammable vs. ˈpro proˈgress , v. ˈprolamine ˈprolapse ˈprolate ˈprolix ˈpromenade vs. - ˈnade in Aus.E. ˈpromenading ˈprophetess proseˈcutrix ˈprospect , v. ˈprospecting vs. proˈspecting in Aus.E. ˈprospector ˈprospekt prosthaphaeˈresis ˈProtean proˈthonotary proˈvokable ˈprowess vs. proˈwess in Br.E. pseudoeˈphedrine vs. ˈephedrine in Br.E./ Aus.E. pteˈrido pteˈridophyte ˈptomaine public adˈdress ˈpuissance ˈpukeko ˈPunjab ˈpupillary vs. puˈpillary in Br.E. ˈPurim ˈpustular ˈpustule ˈpustulous ˈputtee pyraˈmidal in Br.E. vs. pyˈramidal in Am.E. pythoˈness qawˈwali quadruˈpedal quaˈdruply ˈquahog ˈquahogger ˈquahogging ˈQuakeress vs. - ˈress in Br.E. quetˈzal quinˈtuple quinˈtuplet raˈgout 232 8 Appendix ˈrakia ˈRamadan ramˈpage vs. ˈrampage in Am.E. ˈrani Raˈoult’s law ˈRashomon ratˈtan vs. ˈrattan in Br.E. ˈRaynaud ˈrazzmatazz ˈrecess , n. and v. ˈrecessing ˈreclaim , n. reˈcluse vs. ˈrecluse in Am.E. ˈrecognizable ˈrecondite reconˈtract , v. redisˈtribute vs. ˈdis in Br.E. reˈferable refiˈnancing vs. reˈfinancing in Br.E. reˈflect , n. ˈrefund , v. reˈfundable reˈfunded reˈfunder reˈfunding ˈregistrar ˈregulatory vs. - ˈlatory in Br.E. reinˈcarnate vs. - ˈnate in Br.E. ˈReinga ˈrelapse vs. reˈlapse in Br.E., v. ˈrelay , v. reloˈcator ˈremonstrate vs. reˈmonstrate in Br.E. reˈparative reproˈgrammable vs. ˈpro in Br. ˈresearching vs. - ˈsearching in Br. ˈreserpine ˈresource vs. reˈsource in Aus.E. ˈresourced vs. - ˈsourced in Aus.E. ˈresourcing vs. - ˈsourcing in Br.E. ˈrespirable ˈrespiratory vs. - ˈspiratory in Br.E. restiˈtutory reˈsurvey , v. retransˈfer , v. ˈreverb ˈReynard ˈrissole roˈbust roˈbustly roˈbustness ˈrodeo ˈromance , v. roˈmancer ˈrotatory ˈroundel rubeˈola ˈrunanga ˈSabaoth ˈsaccharify ˈsalamander ˈsaline ˈsamizdat ˈsamovar ˈsarco - Satˈsuma saˈvant ˈsavantism Saˈvoy Savoyˈard ˈschizo - ˈschottische ˈSeato vs. Seˈato in Aus.E. secaˈteurs in Br.E./ Aus.E. vs. ˈsecateurs in Am.E. seˈcondment ˈsecretive ˈsecretiveness ˈsegment , v. ˈsegmented 8 Appendix 233 ˈSenegal vs. Seneˈgal in Br.E. ˈsepoy seˈquestrate serˈrated ˈSeveran sexˈtuplet shaughˈraun vs. ˈshaugh in Br.E. shelˈlac vs. ˈshellac in Br.E. ˈShoah ˈsiderite sigˈnor siˈroc ˈsitar ˈskeletal vs. skeˈletal in Br.E. ˈSoho ˈsoleus ˈsolipsism ˈsolipsist ˈsolitaire ˈsolute sorˈry , n. specˈtroscopist specˈtroscopy ˈspinet ˈsqueegee ˈsquilgee sriˈracha ˈstagnate ˈstagnated ˈstagnating staˈlactite vs. ˈstalactite in Br.E. staˈlagmite vs. ˈstalagmite in Br.E. ˈstationarily ˈSteuben ˈstingaree ˈstylee vs. styˈlee in Br.E. subaˈdult ˈsubcontract vs. - ˈtract in Br.E., v. ˈsubluxate ˈsubmarine vs. - ˈrine in Br.E. ˈsubmarining subˈsidence ˈsubstantive vs. - ˈstantive in Br.E. ˈsubstantively vs. subˈstantively in Br.E. subˈstantiveness ˈsuccah Suˈdoku ˈsuffix , v. ˈsulphurous ˈsupine ˈsupplicatory vs. - ˈcatory in Br.E. Suˈrat surˈmise , n. ˈsurname , v. ˈsurvey , n. ˈsuspect , adj. sutˈtee vs. ˈsuttee in Br.E. ˈsynapse Synˈclavier synˈclinal syˈringe , v. ˈtallith ˈTalmud ˈtantivy teˈratogen tercenˈtenary terˈrine tetraˈvalent tiˈsane ˈtoreador ˈtotara in Br.E. vs. - ˈtara in Am.E. ˈtoucan touˈpee vs. ˈtoupee in Br.E. ˈtoward toˈwards traˈceur ˈtrachea vs. traˈchea in Br.E. ˈtracheal vs. traˈcheal in Br.E. ˈtracheid 234 8 Appendix ˈtracheo - Traˈfalgar traˈjectory transˈferable ˈtransmontane ˈtransverse ˈtrapezoid trattoˈria traˈversal traˈverse traˈversing triˈchiasis triˈchina Triˈdentine ˈtrigone ˈtruncate vs. trunˈcate in Aus.E. ˈtruncated vs. trunˈcated in Br.E. ˈTungus ˈTupi ˈturquoise Tuvaˈluan Tyˈrolean vs. Tyroˈlean in Br.E. tyˈrosinase ˈTyrrhene Tzotˈzil umˈbilicus under-ˈresourced ˈupsilon ˈUranus uˈrea ˈurethane ˈurinal vs. uˈrinal in Br.E. ˈvagary ˈvertigo ˈvesicatory ˈvibrate vs. viˈbrate in Br.E. ˈvicinal viˈzier Yad Vaˈshem ˈyippee Yom Kipˈpur ˈYugoslav 9 Index 235 9 Index This index does not include terms that are easily accessible through the Table of Contents (e.g., disyllables, which are the topic of Section 4.2, or hiatus, which is mentioned in the titles of 4.2.4, 5.1.6, and 5.2.7.) Likewise, the index does not contain the words whose stress patterns are discussed in the monograph, but it does contain the rightand lefthand strings in which these words end / with which they begin. E.g., ity of ability . In addition to suffixes and prefixes, the strings given below also include combining forms (e.g., urgy of dramaturgy or allelo of allelomorph ), ults that contain a suffix being preceded by an onset consonant (e.g., sid of Abbasid ), and bound roots / shared strings, which occur at the end of at least two different English words (e.g., ternal of external and internal or abolism of anabolism , catabolism , diabolism , and metabolism ). The authors cited are given in the exact form in which they appear in the main text. E.g., Adams & Munro, Byun et al., etc. -a 55, 181 a - 182 abbreviation 21, 79, 108, 166 -able 111, 114, 161 -abolism 109 abra- 127 acari- 165 acci- 113 aceto- 149, 181 -acity 110 acronyms 166 actino- 165 ad- 193 Adams & Munro 28 -ade 116 adelo- 149 adeni- 165 advant- 113 advers- 188 -age 153, 154, 160 -ageous 113 agglutinative suffixation 53, 77, 114 agro- 140 -aire 116, 185 akan’e 66 -al 7, 8, 55, 87, 107, 111, 115, 187, 192, 201, 205, 206 al 197 Alber 83 -ality 110 allelo- 192 allophonic principle 86 alphabetism 18 Alternating Stress Rule 122 Altmann & Kabak 22 Alyawarr 39 -alysis 109 amino- 149 -an 55, 111, 115 -ance 88, 103, 146 ancil- 174 -and 55 -ane 78 anecdo- 163 236 9 Index Anglo-Saxon 6 -anian 110 -anity 110 -ant 90, 111, 157, 161, 197, 198 ante- 102, 157 antety- 197 anthropo- 165 -any 45 ap- 80, 104, 161 apothe- 125 appli- 161 Apurin- 39 -ar 111, 159 ar- 138 Arabic 71, 133 -ard 55, 90, 153, 154 -arian 110, 111 -arily 113 aristo- 182 -arity 110 Arnold & Hansen 200, 201 Aronoff 6 Arth- 151 Arthr- 151 Arthro- 151 artifi- 183 -ary 111, 120, 121, 160, 174 aspiration 31 -at 111 -ata 149 -ate 7, 11, 87, 88, 99, 111, 119, 120, 147, 154, 183, 184, 203, 204, 208 -ation 110, 111 -ative 148 -atory 113, 121, 161, 208 -atrist 175 autostressed 16 back-derivation 73, 74, 77, 78, 79, 80, 88, 97, 102, 113, 117, 125, 146, 147, 161, 176, 184, 188, 189, 202, 204, 207 Baer 17 Baker & Smith 209 Balteiro 172, 204 Baptista 209 Bauer 99 bea- 50 Bell & Plag 57, 129, 130, 131, 132, 141, 142 bene- 113 Berg 9, 18, 20, 21, 22, 44, 49, 92, 116, 133 Berkeli- 163 Bermúdez-Otero & McMahon 78 bi- 50 -bilicus 183 bio- 106, 194 -brate 183 Bryson 105 Burzio 203 -bute 91 Byun et al. 66 ca- 165 cad- 127 -cadabra 127 camelo- 165 carcino- 165 cardinal numerals 12, 142 Carpenter 42 Carter & Clopper 34 Catholi- 184 -catory 161 centi- 138 -cer 183 chi-squared test 58 Chitoran & Hualde 89, 168 -chival 138 Cho et al. 25 Chomsky & Halle 9, 30, 84, 122, 128, 129, 166 Chrabaszcz et al. 66 Cin- 203 Cinde- 203 -cism 184 co- 148, 153, 154 codaless syllable 89 col- 154 9 Index 237 Colina 35 com- 103, 117, 118, 158 con- 91, 186, 198, 204 conditional probability 130 constitutive function 25 consum- 186 contrari- 163 counter- 56 -cracy 164 -crasy 164 -croglia 151 Cruttenden 9, 10, 11, 13, 14, 17, 34, 36, 46, 86, 89, 93, 115, 126, 143 crystallo- 149 cunei- 182 -cus 183 Cutler 127, 200 CVCCVC-disyllables 152 CVCV 152 CVCVCV-words 122 -cy 55, 111, 121 cy 164 Davis 39 de- 154, 178, 186, 189 de Haas 89 Dehaene-Lambertz & Houston 43 delimitative 199 demo- 164 democra- 164 -dent 117 deva- 106 di- 90, 121, 126, 168, 173, 198 diacritic feature 6, 9 -dicative 162 diminutive 49, 185 -diolus 151 diphthongization 89, 124, 167, 175 dis- 104 -dition 107 -dom 55 Domahs et al. 83, 122, 209 double-prominent 31, 139, 140, 141, 143 duration 5, 9, 25, 27, 28, 31, 34, 35, 36, 37, 43, 64, 65, 66, 81, 84, 168 Dutch 39, 122, 199, 209 Dyirbal 37 -ean 170 echino- 149 -ed 103, 111 -ee 16, 116, 185, 196, 207 -eiform 182 electromyographic activity 32 emanci- 183 -eme 158 emphatic 16, 17, 19, 55, 56, 93, 94, 116, 117, 156, 159, 177, 178, 185, 207 -en 118 -ence 103, 119, 120 ence 165 -ent 105, 111, 117 entre- 103 -eosis 125 -eous 108 equi- 183 equili- 183 equipo- 183 equiva- 183 -er 55, 87, 111, 207 -erian 110 -erity 110 -ery 111 -esce 6 -esque 116 -ess 17, 19, 55, 159, 192, 196 -est 90, 105 -et 55, 185 -etic 111 -ette 16, 17, 49, 185 eu- 150 ex- 40, 186 experi- 165 extrametrical 38, 70, 82 extrava- 203 238 9 Index Farnetani et al. 26, 140 Faroese 199 -fect 178 feet 83 -fense 16 -ficial 113 Field 68 fixed stress 37, 64 Foulkes & Docherty 133 Fournier 4, 127, 174, 199 French 3, 4, 42, 43, 59, 71, 101, 123, 138, 154, 159, 166, 167, 180, 185, 192, 194 frequency of use 18, 20, 50, 104, 106, 131, 160, 200, 204, 206 Friederich 11, 16, 68, 116 Fry 26, 28 Fudge 7, 16, 56 -ful 55 fundamental frequency 26, 27, 28, 31, 34 gameto- 149 Gandour 35 -gant 203 Gay 32, 34 -geal 170 -gen 165 German 38, 39, 59, 64, 122, 167, 182, 199, 200, 201, 203, 209 Germano- 149 Gimson 9, 46 -gitis 109, 170 -glia 151 glide 41, 89, 108, 169 Greek 37, 38, 150, 151, 164 Guierre 68 Guion et al. 209 Gussenhoven 37 Gut 34, 36 Hall 35 Halle & Keyser 2, 3, 68, 69, 128 Halle & Vergnaud 12, 30, 178, 179 Halverson 68 Hamawand 105, 201 Hammond 186, 188 Hayes 2, 3, 12, 13, 35, 36, 38, 39, 69, 70, 79, 176, 209 heavy 3, 9, 38, 47, 51, 54, 73, 80, 82, 83, 84, 97, 122, 125, 200 hegemo- 163 Henderson 21, 65 heresi- 165 HH-disyllables 80, 81 -hibit 5, 6, 109, 201 Hickey 99 Hieke 89 high-falling tone 34 Hill 105 him- 64 Hindi 194 Hindustani 43, 167 historical / combinatorial / syllabic length 36 Huss 28 Hyman 37 hyperarticulation 32 hypero- 165 hyphenations 48, 84, 85, 87, 107, 108, 111, 190 -ia 55, 108, 111, 112, 181, 182 -ial 108, 113, 170 iamb 83 -ian 55, 108, 110, 113, 170 -iatrist 175 -ible 111, 161 -ic 55, 87, 101, 105, 106, 107, 111, 115 -ica 124 -ical 105, 106, 111 -ice 88 Icelandic 199 -ician 110 -icity 110, 111 idempo- 165 -idity 110 -ie 55 9 Index 239 -ier 116, 185 -iferous 111 -ific 111 -ification 100 -ify 100, 111 ikan’e 66 -ile 192 -ility 110 im- 156 immuno- 149, 181 in- 5, 6, 109, 147, 156, 162, 178, 184, 198, 201 Indonesian 64 informativity 129, 131, 141, 142 -ing 55, 103, 111, 205, 208 -inian 110 -inity 110 inner-circle 133 institutionalization 160 intensity 27, 28, 31, 37, 64, 66 inter- 113 intrusive w 169 invers- 188 -ion 7, 55, 87, 88, 99, 100, 103, 106, 108, 110, 111, 121, 135, 146, 154, 203, 204, 208 Irish English 44, 88, 99, 137, 146 -is 55 -ise 19 -ish 55, 87, 117 -ism 111 -ist 111, 112 -istic 111 Italian 105, 108, 123, 124, 125, 151, 167, 192, 203 -ite 55, 100, 111, 150, 186 -ity 110, 111, 135, 206 -ival 187 -ive 18, 51, 111, 114, 148, 162, 184 -ivity 110 -ization 100 -ize 100, 101, 111 Japanese 21, 37 Jensen 83 Jones 43, 68, 89, 167 Kager 83 Kelly 40, 41, 42, 81, 209 Kent & Netsell 31 Kenyon & Knott 9, 69, 116 Kera 39 Kettemann 6, 68, 105, 209 ki- 163 kilo- 138, 163 Kingdon 7, 12, 18, 126, 176 Kösling 26, 132 Kraska-Szlenk & Żygis 12 Kreidler 78 Kretzschmar 133 Krivokapić 22 Kunter 26, 140, 171 la- 170 labora- 208 laby- 113 -lah 153 Lak-Dargwa 42 Lamu-lamu 39 Langendoen 12 laryn- 109, 170 Latin 2, 3, 6, 8, 9, 11, 16, 19, 39, 47, 51, 53, 73, 79, 97, 98, 105, 111, 114, 115, 116, 118, 121, 124, 125, 133, 148, 152, 159, 170, 182, 190, 191, 192, 194, 200, 202, 203 lax 9 -le 170, 191 left-branching 26 left-edge 37 -lent 183 -less 55 -let 55 Levenshtein 52, 53 lexicalization 131 LH 82, 83 240 9 Index -librate 183 -licus 183 Lieber 201 light 3, 9, 38, 47, 51, 53, 59, 62, 64, 70, 82, 83, 97, 111, 122 line breaks 85, 117, 186, 187 -ling 55 -list 183 loanword 160, 194 -logy 109, 110 -lometer 163 loudness 27, 35, 36, 65, 66 Lutz 16 luxu- 121 -ly 55, 111, 113, 121 ma- 119, 151 Macedonian 37 macro- 140, 151 Madimadi 39 Mam 39 Mandarin 66 manner-of-articulation 39 mar- 160 Marks 43 -mate 147, 186 -matic 107 Maximal Onset Principle 85, 186 Mbabaram 39 Mcclean & Tiffany 28 me- 164, 204 median 55, 73, 76, 88, 99, 100, 105, 106, 122, 204, 208 medica- 163 Mel’čuk 6, 202, 205 membrano- 149 menin- 109 -ment 19, 153, 163, 190, 192, 193 mental lexica 68, 69, 144, 194, 209 mercanti- 183 metallo- 149, 163, 174 mis- 194 -mite 150 mixed suffixes 7 modifier 12, 13, 91, 92, 93, 101, 115, 130, 135, 159, 176, 177, 179, 180, 181, 188, 202, 207 Mompéan 12, 13, 14, 64, 65 mono- 166 monolithicity 129 morph- 155 morphemes 6 morphemic principle 86 -morphose 155 -morphosis 155 Morton & Jassem 27 -mous 147 Mücke & Grice 25, 32 myceto - 149 myrmeco - 149 -nal 120 namas- 194 -nati 203 necr- 155 negative connotation 16 -neity 110 nemato- 149 -ness 55, 111, 163 Niebuhr 42 Normal Stress Rule 68 Norman Conquest 2 -nosis 186 nucleus 10, 28, 39, 41, 52, 70, 78, 200 -ny 163 -o 55, 181 o- 7, 90, 153, 164, 181, 204 oc- 150, 157, 162, 169 -ocracy 164 -ocrasy 164 -ography 111 -oid 55, 186 Old English 5, 80, 133 oligar- 164 oligo- 182 9 Index 241 -ology 109, 110, 111 -olus 151 -ometer 163 -onal 187 -onian 110 onset 5, 39, 40, 41, 81, 85, 87, 118, 155, 156, 209 -opoda 151 -or 16, 51, 111, 121, 154, 196, 202, 207, 208 ordinal numeral 75, 76 organo- 149 ornitho- 149 -ory 111, 133, 160, 208 oscheo- 149 oscillo- 149 -ose 88, 154, 155 -osis 88, 125, 155 -osity 110 ostraco- 165 -otomy 6 -our 154 -ous 55, 113, 115, 147 out-of-focus positions 27 over- 181 ox- 7 oxalo- 149 pa- 8 pallado- 149 paraphrase 74, 202 parasitic nature of stress 35 -pard 90 -pate 183 Pater 126 Payne et al. 43, 159 Pearson 58 pen- 3, 22, 119, 157, 204 Peng & Ann 22 pen-initial stress 161 periodo- 165 Pettersson 65 phonotactic principle 86 -phthong 166 phys- 175 physi- 175, 195 piezo- 149 pitch 25, 26, 27, 28, 29, 30, 31, 34, 35, 36, 37, 43, 64, 66, 129, 131, 132, 140, 141, 142, 143 Plag 25, 28, 29, 30, 50, 72, 130, 131, 132, 140, 141, 201 -plicable 161 -plicatory 161 -plish 117 -ployment 73 pneumato- 149 -poda 151 -podal 151 -podan 151 poikilo- 149 Poldauf 15, 25, 53, 56, 69, 71, 93, 123, 126, 127, 129, 152, 160, 180, 199, 200 poly- 166 polysyllabicity 18 positional size 130, 141, 142 post- 147 post-nuclear position 28, 30 post-pen-initial 37 -potent 183 pre- 119 -preneur 103, 104 prismato- 149 pro- 5, 6, 125, 175 -proach 80 productive 76 prog- 186 progno- 186 protho- 175 prothono- 175 pterido- 149 -ptych 166 pyrito- 149 quantity-sensitive 39, 199 quin- 150 242 9 Index -rage 193 re- 46, 80, 93, 157 reduced vowel 10, 29, 37, 133 regular expressions 50, 51 -rella 203 -rence 120 rhyme 38, 47, 70, 82 rhythmic pressure 9, 11 Rhythm Rule 12, 13, 14, 21, 178, 179, 180, 188 -rifugal 151 right-branching 26 right-edge 37 -rinthian 113 -rogenate 7 -roglia 151 -ronal 148 root-prefix boundary 6, 51, 109, 125, 151, 161, 163, 165, 169, 199, 203 -ropoda 151 Russian 41, 59, 65, 66, 70, 149, 160, 167 -ry 55, 111, 121, 174 -ryngeal 170 Sanskrit 159 satis- 114 Scarborough et al. 25, 27, 32, 33 Scherer & Wollmann 177 Schmid 74, 131, 201 Schofield 62 schwa 13, 42, 70, 129, 167 seg- 107 semantically-conditioned 76 semantics 5, 8, 91, 94, 104, 107, 131, 176, 181, 182, 201, 206 semi-vowel 41 Serbo-Croatian 39 sex- 150 -ship 55, 111 Shipibo-Conibo 39 shortening 19, 34, 78, 79, 80, 97, 105, 158, 159, 202 -sia 170 -sid 148 sidero- 149 Simo Bobda 22 -sis 155, 186 situ- 110 Sokolova et al. 25, 65 Solmecke 64 -some 55 sonority expansion 32 Sonority Sequencing Principle 35 Spanish 123, 124, 125, 151, 167, 168, 185, 203 speech-recognition 62 -st 55 Stadnik-Holzer 41 -stant 198 Štekauer 202 -ster 118 stress-accent 37 stress-attracting 116 stress-neutral 7, 14, 17, 93, 112, 114, 116 stress-repellent 88, 90, 100, 116, 119, 120, 121, 127, 146, 147, 154, 161, 163, 165, 183, 184, 185, 192, 208 stress-shifting 7, 112 stress-timed 43 sub- 15, 56, 112, 113, 118 subtractive word-formation 202 suf- 158 -summate 186 sur- 94 -sy 121 syllable-timed 43 syllable weight 39 -t 153, 181 -tage 160 Tan 22 tar- 159 -tarily 113 -tary 175 -tasia 170 -te 194 9 Index 243 -teen 12, 13, 14, 142, 143 -teenth 12 temporal and / or spatial expansion 32 -tence 165 tense 9 -tent 183 -ter 118 -ternal 107 tetra- 109, 110, 201 Thaypan 39 the- 164 Theatro- 149 theocra- 164 thir- 143 -thylamine 163, 164, 204 -toate 169 Tokar 201 tone 34, 36, 37, 39 -toria 204 -torial 115 Torsuev 18, 35, 166 tracheo- 149 -tractive 148 transvers- 188 Trevian 68, 112 tri- 15, 91 trigono- 149 trochee 83 -tuplet 150 Turcsan & Herment 133, 209 Turk et al. 200 Turk & Sawusch 25 -ty 55, 111, 121, 143, 144 -typal 197 -uble 161 -ule 55 -um 55 um- 183 umbi- 183 umbili- 183 -urbia 112, 113 -ure 111, 119, 200, 206 -urgy 109, 174 -ute 99, 146 -vagant 203 van der Hulst 3, 5, 28, 31, 34, 37, 38, 39, 41, 42, 43, 64, 124, 199, 200 Van Rooy 22 van Zanten et al. 64 -var 160 variable stress 13, 38, 44, 58, 138 -velop 186 Ventcel’ 81 -vert 153, 181, 204 volun- 113 Walch 209 -ward 55 -way 55 weak syllables 10, 11 weighted arithmetic mean 81 Wells 18, 22, 58, 133 Wenk 43 Wenszky 199 Western Aranda 39 Wetzels 89 Wilbur & Schick 33 Wilson 47 Wiltshire & Moon 22 -y 55, 87, 111, 121 y 164 Yates’ correction 79, 150 Yavaş 35, 85 χ 2 -statistics 49 Tokar Stress Variation in English Alexander Tokar Stress Variation in English This monograph is concerned with the question of why some English words have more than one stress pattern. E.g., 'overt vs. o'vert, 'pulsate vs. pul'sate, etc. It is argued that cases such as these are due to the fact that the morphological structure of one and the same English word can sometimes be analyzed in more than one way. Thus, 'overt is the stress pattern of the suffixation analysis over + -t, whereas o'vert is due to the prefixation analysis o- + -vert (cf. covert). Similarly, pulsate is simultaneously pulse + -ate (i.e., a suffixed derivative) and a back-derivative from pul'satance. „Tokar᾽ s approach in the use of both dictionary and corpus data holds promise of a scholarly breakthrough on the vital linguistic prosodic topic of English stress assignment of doublets and of stress assignment in general.“ (Irmengard Rauch, Professor of Germanic Linguistics at the University of California, Berkeley) Language in Performance LiP ISBN 978-3-8233-8180-8