The Usage-Based Study of Language Learning and Multilingualism 9781626163997, 9781626163249, 9781626163256, 2015030182, 1626163995

When humans learn languages, are they also learning how to create shared meaning? InThe Usage-based Study of Language Le

241 93 4MB

English Pages 308 [307] Year 2016

Report DMCA / Copyright

DOWNLOAD FILE

Polecaj historie

The Usage-Based Study of Language Learning and Multilingualism
 9781626163997, 9781626163249, 9781626163256, 2015030182, 1626163995

Table of contents :
Cover......Page 1
Contents......Page 8
Illustrations......Page 10
Preface......Page 16
1. Introduction: The Vibrant and Expanded Study of Usage-based Language Learning and Multilingualism......Page 18
PART I. USAGE-BASED DEVELOPMENT OF LANGUAGE ACROSS THE LIFESPAN......Page 30
2. A Multimodal Approach to the Development of Negation in Signed and Spoken Languages: Four Case Studies......Page 32
3. Why Don’t You Just Learn it from the Input? A Usage-based Corpus Study on the Acquisition of Conventionalized Indirect Speech Acts in English and German......Page 54
4. Prepositional Phrases as Manner Adverbials in the Development of Hebrew L1 Text Production......Page 72
Appendix: Statistical methods and main outcomes......Page 89
5. Negative Constructions in Nonliterate Learners’ Spoken L2 Finnish......Page 92
6. How do Multilinguals Conceptualize Interactions Among Languages Studied? Operationalizing Perceived Positive Language Interaction (PPLI)......Page 108
PART II. THE CORPUS-AIDED, USAGE-BASED STUDY OF LEARNER LANGUAGE......Page 130
7. A Friendly Conspiracy of Input, L1, and Processing Demands: That-variation in the Language of German and Spanish Learners of English......Page 132
Appendix 7–1......Page 152
8. Measuring Lexical Frequency: Comparison Groups and Subject Expression in L2 Spanish......Page 154
9. Article Omission: Toward Establishing How Referents Are Tracked in L2 English......Page 172
10. Measuring L2 Explicit Knowledge of English Verb-Particle Constructions: Frequency and Semantic Transparency at Two Proficiency Levels......Page 188
PART III. THE EXPERIMENTAL STUDY OF USAGE-BASED PROCESSING AND LEARNING......Page 204
11. Can English-Spanish Emerging Bilinguals Use Agreement Morphology to Overcome Word Order Bias?......Page 206
12. Miniature Artificial Language Learning as a Complement to Typological Data......Page 228
PART IV. MULTILINGUALISM IN THE WILD: USAGE-BASED INSIGHTS......Page 250
13. Patterns of Interaction in Doctor-Patient Communication and Their Impact on Health Outcomes......Page 252
14. Toward a Model of Multilingual Usage......Page 272
Contributors......Page 292
B......Page 298
C......Page 299
E......Page 300
I......Page 301
L......Page 302
M......Page 303
P......Page 304
S......Page 305
V......Page 306
Z......Page 307

Citation preview

The Usage-based Study of Language Learning and Multilingualism

Georgetown University Round Table on Languages and Linguistics—Selected Titles Languages in Africa: Multilingualism, Language Policy, and Education

Elizabeth C. Zsiga, One Tlale Boyer, and Ruth Kramer, Editors

Measured Language: Quantitative Studies of Acquisition, Assessment, and Variation Jeffrey Connor-Linton and Luke Wander Amoroso, Editors

Discourse 2.0: Language and New Media

Deborah Tannen and Anna Marie Trester, Editors

Arabic Language and Linguistics

Reem Bassiouney and E. Graham Katz, Editors

Language in Use: Cognitive and Discourse Perspectives on Language and Language Learning

Andrea E. Tyler, Mari Takada, Yiyoung Kim, and Diana Marinova, Editors

◆◆◆

The Usage-based Study of Language Learning and Multilingualism Lourdes Ortega, Andrea E. Tyler, Hae In Park, and Mariko Uno, Editors

Georgetown University Press Washington, DC

Georgetown University Press, Washington, DC www.press.georgetown.edu © 2016 by Georgetown University Press. All rights reserved. No part of this book may be reproduced or utilized in any form or by any means, electronic or mechanical, including photocopying and recording, or by any information storage and retrieval system, without permission in writing from the publisher. Library of Congress Cataloging-in-Publication Data The usage-based study of language learning and multilingualism / Lourdes Ortega, Andrea E. Tyler, Hae In Park, & Mariko Uno, editors. pages cm — (The Georgetown University Round Table of Languages and Linguistics series) ISBN 978-1-62616-399-7 (hardcover: alk. paper) — ISBN 978-1-62616-324-9 (pbk. : alk. paper) — ISBN 978-1-62616-325-6 (ebook) 1. Language acquisition. 2. Second language acquisition. 3. Multilingualism. 4. Language and languages—Usage. I. Ortega, Lourdes, editor. II. Tyler, Andrea, editor. III. Park, Hae In, 1982editor IV. Uno, Mariko, editor V. Series: Georgetown University round table on languages and linguistics series (2004) P118.U76 2016 401’.93—dc23 2015030182 This book is printed on acid-free paper meeting the requirements of the American National Standard for Permanence in Paper for Printed Library Materials. 17 16   9 8 7 6 5 4 3 2 First printing Printed in the United States of America Map in figure  12-3 used by permission, ©2009 SIL International, Ethnologue: Languages of the World, Sixteenth Edition, further redistribution prohibited without permission. Cover design by Pam Pease

We dedicate this book to the memory of James E. Alatis (1926–2015), in celebration of his legacy in the fields of applied linguistics and TESOL, and his vision for the Georgetown University Round Table on Languages and Linguistics (GURT).

This page intentionally left blank

‹‹ Contents

Illustrations ix Preface xv 1. Introduction: The Vibrant and Expanded Study of Usage-based Language Learning and Multilingualism ◆◆

1

Lourdes Ortega and Andrea E. Tyler

Part I. Usage-based Development Of Language Across The Lifespan 2. A Multimodal Approach to the Development of Negation in Signed and Spoken Languages: Four Case Studies ◆◆

15

Aliyah Morgenstern, Pauline Beaupoil-Hourdel, Marion Blondel, and Dominique Boutet

3. Why Don’t You Just Learn it from the Input? A Usage-based Corpus Study on the Acquisition of Conventionalized Indirect Speech Acts in English and German ◆◆

37

Ursula Kania

4. Prepositional Phrases as Manner Adverbials in the Development of Hebrew L1 Text Production ◆◆

◆◆

55

Gilad Brandes and Dorit Ravid

Appendix Statistical methods and main outcomes

72

5. Negative Constructions in Nonliterate Learners’ Spoken L2 Finnish

75

Taina Tammelin-Laine and Maisa Martin

6. How do Multilinguals Conceptualize Interactions Among Languages Studied? Operationalizing Perceived Positive Language Interaction (PPLI) ◆◆

91

Amy S. Thompson

vii

viii

Contents

Part Ii. The Corpus-Aided, Usage-based Study Of Learner Language 7. A Friendly Conspiracy of Input, L1, and Processing Demands: That-variation in the Language of German and Spanish Learners of English ◆◆

◆◆

Appendix 7–1

135

8. Measuring Lexical Frequency: Comparison Groups and Subject Expression in L2 Spanish

137

Bret Linford, Avizia Long, Megan Solon, and Kimberly L. Geeslin

9. Article Omission: Toward Establishing How Referents Are Tracked in L2 English ◆◆

155

Monika Ekiert

10. Measuring L2 Explicit Knowledge of English Verb-Particle Constructions: Frequency and Semantic Transparency at Two Proficiency Levels ◆◆

115

Stefanie Wulff

171

Helen Zhao and Fenfen Le

Part Iii. The Experimental Study Of Usage-based Processing And Learning 11. Can English-Spanish Emerging Bilinguals Use Agreement Morphology to Overcome Word Order Bias? ◆◆

12. Miniature Artificial Language Learning as a Complement to Typological Data ◆◆

189

Silvia Marijuan, Sol Lago, and Cristina Sanz

211

Maryia Fedzechkina, Elissa L. Newport, and T. Florian Jaeger

Part Iv. Multilingualism In The Wild: Usage-based Insights 13. Patterns of Interaction in Doctor-Patient Communication and Their Impact on Health Outcomes ◆◆

Diana Slade, Christian M.I.M. Matthiessen, Graham Lock, Jack Pun, and Marvin Lam

◆◆

Michel Achard and Sarah Lee

14. Toward a Model of Multilingual Usage

235

255

Contributors 275 Index 281

‹‹ Illustrations Figures Figure 2–1. Pluri-semiotic resources used by the children, according to the channels at their disposal (visual/auditory). SGS refers to gestures shared by gesturing speakers and signers, and ONOM refers to onomatopoeia. 20 Figure 2–2. Rate of actions, gestures, and speech, and number of occurrences per category in Ellie’s data 21 Figure 2–3. Rate of actions, gestures, and speech, and numbers of occurrences per category in Madeleine’s data 22 Figure 2–4. Rate of actions, gestures, and speech, and number of occurrences per category in Charlotte’s data 23 Figure 2–5. Rate of actions, gestures, and speech, and number of occurrences per category in Illana’s data 24 Figure 2–6. Charlotte 1;00 IndexWave 27 Figure 2–7. Illana 1;00 IndexWave + vocal-verbal production 27 Figure 2–8. Illana 1;00 Father-child alignment: IndexWave + verbal production or pouting 28 Figure 2–9. Illana 1;06 IndexWave 28 Figure 2–10. Illana 1;06 Father takes up the PalmUp gesture 29 Figure 2–11. Ellie 3;00 PalmUp gesture 29 Figure 2–12. Madeleine 3;00 PalmUp-Shrug (one hand) Figure 2–13. Madeleine 3;00 PalmUp-Shrug (two hands) Figure 2–14. Text grid and spectrogram for “Et lequel alors?” (‘So which one is it then?’) Figure 2–15. Illana 2;00 IndexWave Figure 2–16. Illana 2;00 PalmUp-Shrug Figure 4–1. Mean number of manner PPs per clause by age Figure 4–2. Mean number of words per manner PP by age and genre Figure 4–3. Proportions of manner PPs’ internal structures by age in narratives Figure 4–4. Proportions of manner PPs’ internal structures by age in expositories Figure 4–5. Proportions of manner PPs’ syntactic functions by age in narratives ix

30 31 32 32 33 61 62 62 62 64

x

Illustrations

Figure 4–6. Proportions of manner PPs’ syntactic functions by age in expositories 64 Figure 6–1. Visual representation of the operationalization of PPLI. 95 Figure 6–2. A visual representation of the complex nature of PPLI as a dynamic construct 101 Figure 7–1. Main effect of LengthComplement 124 Figure 7–2. Interaction between LengthComplementSubj and ComplementType 125 Figure 7–3. Interaction between LengthComplementSubj and LengthMatrixSubj 126 Figure 7–4. Interaction between LengthMatrixSubj and LengthMCVerbCC 126 Figure 7–5. Interaction between LengthMatrixSubj and LengthMCSubjMCVerb 127 Figure 7–6. Interaction between LengthMCVerbCC and LengthCIM 128 Figure 7–7. Interaction between LengthMCVerbCC and LengthMCSubjMCVerb 128 Figure 7–8. Interaction between Mode and ComplementType 129 Figure 7–9. Interaction between Mode and LengthComplementSubj 130 Figure 7–10. Interaction between L1 and Mode 130 Figure 7–11. Interaction between L1 and LengthMatrixSubj 131 Figure 7–12. Interaction between L1 and DeltaPWC 132 Figure 8–1. Overt SP production in same versus switch reference contexts for NNSs and NSs 144 Figure 8–2. Rates of overt SP by frequency (internal and external) and switch reference for NNSs 147 Figure 8–3. Rates of overt SP by frequency (internal and external) and switch reference for NSs 148 Figure 10–1. Sample of the cloze test 179 Figure 10–2. Interaction of frequency and semantic transparency in the cloze test 181 Figure 11–1. Word-by-word self-paced reading 200 Figure 11–2. Accuracy (Trial Order × Condition). The dots represent the target items; the dotted red lines represent the participants’ competency at the beginning of the experiment, and the solid red lines represent the Trial Order effects (the improvement over the course of the experiment) 202 Figure 11–3. Mean residual reaction times (ms) in each condition (OclVS sentences). Sample sentence: “Now the girl(s) is (are) looking for him (them) impatiently.” 203 Figure 11–4. Mean residual reaction times (ms) for SV sentences. Sample sentence: “Sometimes the dog(s) sleep(s) in the bed.” 205

Illustrations

xi

Figure 12–1. Example of a training item (left panel) and a comprehension test item (right panel) used in Fedzechkina, Jaeger, and Trueswell (2015). Pictures are still images of the videos accompanied by sentences in a miniature artificial language presented auditorily during the experiment (shown here in print for convenience). 214 Figure 12–2. Two paradigms used in miniature artificial language learning research. In both paradigms, participants are exposed to different tokens of two or more structural types—a crosslinguistically frequent (typical) and a crosslinguistically infrequent (atypical) structure. Paradigm I uses a between-subject design where different groups of participants are consistently exposed to tokens of either typical or atypical structural type and assesses successful learning via accuracy measures. Paradigm II employs a within-subject design where the same group of participants is exposed to tokens of typical and atypical type within the same language and successful learning is assessed through the degree of preference for tokens of either type in participants’ productions. 216 Figure 12–3. Hypothesized relationship between individual biases and language structures 223 Figure 12–4. Between-participant variability on the final day of training in Fedzechkina, Newport, and Jaeger (in press). Case-marker preferences of individual learners in the fixed (panel A) and flexible (panel B) constituent order languages. 226 Figure 12–5. Between-participant variability on the final day of training in Fedzechkina, Newport, and Jaeger (in press). Uncertainty vs. production effort trade-off. Solid shapes with dark borders represent  input languages. Solid shapes without borders represent mean output languages. Open shapes represent language outputs produced by individual learners. The error bars represent bootstrapped 95 percent confidence intervals. 226 Figure 13–1. Speech functions in Crystal’s consultation 243 Figure 13–2. Speech functions in Sam’s consultation 243 Figure 13–3. Symptoms as construed by lexis in Sam’s and Crystal’s consultations 245 Figure 13–4. Major and minor shifts in field in the HEPC stages of Crystal’s and Sam’s consultations 246 Figure 13–5. Interpersonal and experiential dimensions in the HEPC stage of Crystal’s consultation 248 Figure 13–6. Interpersonal and experiential dimensions in the HEPC stage of Sam’s consultation 250

xii

Illustrations

Tables Table 2–1. Recordings and negative productions of the four children 19 Table 2–2. Number of occurrences of PalmUp-Shrugs and IndexWaves in the four girls’ data 26 Table 3–1. Formal frequencies of the target constructions 42 Table 3–2. Results for function 44 Table 3–3. Age of acquisition (AoA) for comprehension and production 46 Table 3–4. Frequency of variation sets 47 Table 5–1. Inflection of the verb asua (to live) in the present tense indicative mode (personal ending underlined, question suffix italicized) 79 Table 5–2. Summary of participant information 80 Table 5–3. The data of the study 81 Table 5–4. Negative constructions in the data 82 Table 5–5. Some examples of the inflection of Finnish verbs 87 Table 6–1. Factor names from Thompson and Erdil-Moody (2014) 99 Table 6–2. PPLI and NPPLI results from Thompson and Erdil-Moody (2014) 99 Table 6–3. Factor names from Thompson and Khawaja (2015) 100 Table 6–4. PPLI/non-PPLI results from Thompson and Khawaja (2015) 100 Table 6–5. Factor names from Thompson and Aslan (2014) 100 Table 6–6. PPLI and NPPLI results from Thompson and Aslan (2014) 100 Table 6–7. Foreign languages studied other than English 102 Table 6–8. Language-specific interaction types 103 Table 6–9. Specific types of interactions reported by the participants 105 Table 6–10. Perceived Positive Language Interaction Questionnaire (PPLIQ) 108 Table 7–1. Data sample of the present study 122 Table 8–1. Participant characteristics 142 Table 8–2. Summary of switch reference categories 143 Table 8–3. Rates of subject forms with 1sg referents for NSs and NNSs 144 Table 8–4. The frequent verbs (i.e., those that constitute 1% or more of the corpus) according to each LF measure, in order of frequency, with the raw count and percentage of the corpus that they represent 145 Table 8–5. Rates of overt SPs for NNSs based on each measure of LF 146 Table 8–6. Rates of overt SPs for NSs based on each measure of LF 146 Table 9–1. Average frequency of determiners for three groups in the video retelling (normed per 300 words, N = 80) 162 Table 9–2. Average frequency of determiners in the bidirectional translation task (normed per 300 words, N = 15) 166

Illustrations

Table 10–1. Table 10–2. Table 10–3. Table 11–1. Table 11–2. Table 11–3. Table 11–4. Table 11–5. Table 11–6. Table 11–7. Table 11–8. Table 13–1. Table 13–2.

xiii

The target phrasal verbs (PVs) of four subcategories Comparisons of the four phrasal verb (PV) subtypes Mean accuracy of the phrasal verb (PV) types Properties of third-person direct object clitics in Spanish Sample set of experimental items Mean accuracy in each experimental condition (OclVS sentences) Logistic regression for OclVS sentences (trial order and condition as predictors) Effect sizes (std. coef.) for OclVS sentences (accuracy) Mean residual reaction times (ms) in each region and condition (OclVS sentences) Fixed-effects linear model for residual reaction time data (OclVS sentences) Effect sizes for ungrammatical sentences in Foote (2011) and the present study Moves in an exchange Tracking and challenging moves

178 180 181 195 199 200 201 202 203 204 205 241 242

This page intentionally left blank

‹‹ Preface Lourdes Ortega, Andrea E. Tyler, Hae In Park, and Mariko Uno, Editors

The present collection gathers cutting-edge research that investigates language learn-

ing and multilingualism from varied usage-based perspectives. The book comprises 13 original studies, representing a selection of the Georgetown University Round Table  on Languages and Linguistics (GURT)  2014, which we co-organized and hosted at Georgetown University on March 13–16, 2014. We have dedicated this book to the memory of James E. Alatis (1926–2015), who created GURT in 1949 and was the leader behind its success; the conference has taken place uninterruptedly every year since then, and it is still going strong to this date. The theme we chose for GURT 2014, Usage-based Approaches to Language, Language Learning, and Multilingualism, reverberated with the theme of Language in Use featured about a decade earlier at GURT 2003, which was organized by Andrea Tyler and a dedicated team of graduate students (and which led to an edited collection by the same title, published in 2005 by Georgetown University Press and co-edited by Tyler, Takada, Kim, and Marinova). As in 2003, the levels of intellectual stimulation and excitement were extremely high in 2014. But the 2014 conference experience showed us just how much of the landscape had been transformed in the intervening decade, and what tremendous growth and momentum usage-based linguistics is witnessing as of this writing. The 2014 conference established a sort of coming of age of usage-based linguistic research. It revealed tremendous and widespread interest in this research perspective into the acquisition and use of language: We received 260 proposals, and the time and expertise of 65 reviewers was critical in identifying the best 61 percent of submissions. It also demonstrated that the usage-based study of language is a truly international line of inquiry: The conference submissions spanned 34 different countries, including the traditionally English majority–speaking geographies of the United States, Australia, Canada, New Zealand, and the United Kingdom, but also many other European countries, as well as many other countries such as Brazil, China, Colombia, Hong Kong, Israel, Japan, Jordan, Kuwait, Nigeria, Pakistan, The Philippines, The Russian Federation, Saudi Arabia, Taiwan, and the United Arab Emirates. This geographic diversity is also reflected in the sum of author affiliations of the present volume and the main research contexts they investigated, which xv

xvi

Preface

include Australia, France, Germany, Hong Kong, Israel, Malaysia, Sweden, and the United States. All chapters  in the book have been refereed rigorously, and we are grateful to the following colleagues who lent their generous expertise: José Alemán Bañon (University of Reading, UK), Vera Busse (Carl von Ossietzky Universität Oldenburg, Germany), Catherine Davies (University of Alabama, USA), Annick De Houwer (Erfurt University, Germany), Jan H. Hulstijn (University of Amsterdam, The Netherlands), Terry Jansen (University of Manitoba, Canada), Kendall King (University of Minnesota, USA), Minna Kirjavainen (Osaka Gakuin University, Japan), Narges Mahpeykar (Georgetown University, USA), Carol L. Moder (Oklahoma State University, USA), Akira Murakami (University of Birmingham, UK), Marianna Ryshina-Pankova (Georgetown University, USA), Naoko Taguchi (Carnegie Mellon University, USA), Mariko Uno (Georgetown University, USA), Nina Vyatkina (The University of Kansas, USA), and Eve Zyzik (University of California Santa Cruz, USA). We also want to thank the 2014 Conference Organizing team, doctoral students Stephanie Kramer, Julie Lake, Narges Mahpeykar, Vitaly Nikolaev, Kate Riestenberg, Mariko Uno, and Jeremy Wegner. Their energy, enthusiasm, competence, and generosity were instrumental in making the conference, and thus this volume, happen. We also thank the Faculty of Languages and Linguistics at Georgetown University for their generous support. Our thanks also go to David Nicholls and Hope LeGro and their team at Georgetown University Press for their attention to detail and their support. Foremost, we are grateful to the authors of the book for offering their high-quality research to us and for their hard work in the preparation of this volume. We hope readers will enjoy and benefit from the work gathered in this collection as much as we have. And we look forward to many more years of vibrancy in the usage-based study of language learning and multilingualism.

1 ‹‹ Introduction: The Vibrant and Expanded Study of Usage-based Language Learning and Multilingualism Lourdes Ortega and Andrea E. Tyler

Georgetown University

THE FAMILY OF USAGE-BASED perspectives posits that humans deploy (or learn to deploy) language in order to create shared meaning. In this first chapter, we introduce the studies in the volume, structuring our commentaries of the chapters  around four broad themes: language development across the lifespan in family and classroom contexts, corpus studies of elicited learner language, experimental methodologies in the laboratory, and multilingualism in the wild. We finish with a brief forecast of main themes and promising areas for future research.

Introduction

Usage-based perspectives on language represent a family of approaches which share two key notions. One is that language and language learning are meaning based, grounded in general cognitive-social processes, and therefore learned rather than genetically endowed or modularly specialized. The other notion is that the central challenge for communication between speakers and listeners is to establish a convergence of attention on similar concepts within each individual interlocutor’s mind; thus, the goal of employing language is for interlocutors to make what Langacker calls mental contact (1987). As a corollary of these two central ideas, humans deploy (or learn to deploy) linguistic resources in order to create shared meaning, and all units of language—from lexical items to morphemes to sentence patterns to discourse patterns—are meaningful. Thus, no linguistic form exists which is not linked to a semantic component. Just as the form of a linguistic unit can vary in complexity, so can the semantic component vary in complexity, ranging from relatively 1

2

Introduction

straightforward lexical meaning to meaningful syntactic patterns (e.g., passive voice with focus on the patient/undergoer of an action) to discourse functions (e.g., a pattern used to convey new, focused information). This form-meaning pairing is the essence of a linguistic construction, a theoretical construct that is central in usage-based linguistics. Twenty plus years of empirical research have provided us with evidence that, like all human learning, language learning is exemplar based: The learner, whether an infant or an adult, first gains understanding of individual instances of language occurring in local, meaningful communication with others. Over a history of usage, the learner then gradually creates more abstract, interactive schematic representations, or a mental grammar. The chapters in this collection contribute usage-based insights to several questions: How do proficient speakers accomplish ‘mental contact’ or communication through the available semiotic linguistic resources they share with other members of their various discourse communities? How do young children learn to accomplish this? And how do speakers of multiple languages learn to accomplish this across languages? In this first chapter, we provide readers with a commentary on each of the studies that comprise this volume, highlighting the main contributions they make to the usage-based study of language learning and multilingualism.

Usage-based Studies of Language Development across the Lifespan

The first five studies in the collection address diverse dimensions of language development, including the linguistic and gestural expression of negation, the production and interpretation of indirect speech acts, the expansion of prepositional phrases for increasingly more complex functions in written discourse, and metalinguistic awareness of how knowledge of multiple languages interacts in the multilingual mind. The populations investigated are equally diverse, from preschool children, to school-aged adolescents, to limited-education and low-literacy adults, to college-educated adults. The target languages include: English, French, French sign language, German, and Hebrew. The methodologies featured include three longitudinal investigations of naturally occurring interactions between infants and caretakers and between students and teachers, one study involving across-age cross-sectional elicitation of writing, and one study with self-report survey data. Morgenstern, Beaupoil-Hourdel, Blondel, and Boutet (chapter  2) investigate the development of the expression of negation through symbolic gestures, particularly the PalmUp-Shrug (moving the palms up with a shrug to mean, e.g., ‘I don’t know’) and the IndexWave (waving the index finger to express, e.g., ‘That’s a no-no’). They inspected evidence recorded every six months over the first three years of life of four girls, with a triple focus on negative multimodal utterances, symbolic gestures, and actions. The languages and modalities (spoken or signed) differed: Three of the children were growing up monolingually in French, British English, or French sign language (langue des signes française, LSF). A fourth child

Introduction

3

was growing up bilingually with bimodal input in French (from the mother) and LSF (from the father). The modality comparison undertaken by Morgenstern et al. is illuminating. For example, the IndexWave sign is shared with the gestural repertoire in the adults’ child-directed input to all four girls, yet the two signing girls use it more frequently and to express a broader range of negation functions than the two non-signing girls, who do not use it up to age 3;00. Morgenstern et al. hypothesize the signing children are in contact with many more occurrences of that gesture, since it is part of the sign language system as well as the gesture system of the larger community. The PalmUp-Shrug, which also occurs with high frequency in the input to all four children, shows quite a different developmental trajectory. The monolingual LSF signing child rarely uses it, while the three hearing children all use it. Morgenstern et al. argue this is likely related to the fact that, as shown in the data, the signing parents use more specific signs that cover much of the semantic territory of this gesture for the hearing children. In sum, all four children have a common gestural inventory available in the input, yet their learning trajectories appear to vary based on the frequency of the input and communicative pressures related to audience. Furthermore, audience and communicative pressures seem to push each child’s flexibility in using the salient communication resources at her disposal. In chapter 3, Kania investigates the development of questions such as Why don’t you do X, a case of an Indirect Speech Act (ISA) which, interpreted literally, issues a request for information, but is in fact regularly understood as an order by language users, including young children. The traditional assumption is that, when it comes to ISAs, listeners first process the literal meaning and, if the literal interpretation is implausible, secondarily consider the indirect interpretation. Cognitive approaches, however, predict that when processing ISAs, language users operate with cognitive models and activate a specific part of a model in order to more directly process the so-called nonliteral meaning. This is because a shared mechanism is posited for both metaphoric and literal language comprehension (Dancygier and Sweetser 2014). Kania inspects evidence for this prediction from three child-parent language corpora available in CHILDES (MacWhinney 2000) and concludes that constructions that were predominantly used as ISAs by the caretakers were first comprehended and produced as such by the children. Although there were some exceptions, the indirect use is mastered by the three children in both comprehension and production before the direct one. As the cognitive model explanation posits, high-frequency ISAs become entrenched as constructions and are thus accessed directly. Overall, the study is important in indicating that children are indeed sensitive to the communicative function of ISAs in the input, and that this factor helps predict in which order the respective form-function mappings are learned. Drawing from construction grammar (Goldberg 2003), Brandes and Ravid (chapter 4) examine prepositional phrases that serve as manner adverbials (manner PPs) in Hebrew (be-adinut ‘in-gentleness = gently’, bli reshut ‘without permission’) and are a site for the expansion of intra-sentential complexity in the development of L1 writing. The data come from eighty written narrative and expository texts elicited cross-sectionally to represent a developmental continuum along age and schooling,

4

Introduction

from fourth graders (aged nine to ten) all the way to graduate university students (aged twenty-five to thirty). Conceptualizing manner PPs in terms of a construction—that is, a unique pairing of form (PP) and meaning (manner)—allows the authors to view them as unified wholes in discourse syntax, linking empirical patterns of increasing complexity to the changing discourse roles that manner PPs play in writing. The results showed that: (1) manner PPs grow more prevalent, and also more internally complex, with age and schooling; and (2) most manner PPs modify verbs, but some function as noun modifiers, adjective modifiers, and predicates. Unexpectedly, (3) some of the more complex manner PPs emerge earlier in expository texts than narratives; and (4) many manner PPs do not constitute optional adjuncts, but rather function as obligatory, focal elements of clauses. Regarding the third finding, Brandes and Ravid argue that expository discourse pushes writers towards more linguistic complexity early on simply because it presents higher conceptual demands; writers are pushed to articulate more general, abstract content as part of the communicative demands of this genre. In other words, the expansion of the complexity of syntactic constructions ties to expansions in the complexity of the content within discourse, an interesting example of form-meaning relations at the discourse level. As for the unexpected obligatory function of manner PPs, as in I want to live in peace with my surroundings and with myself, it was argued that such manner PPs articulate new, interesting information, a focal use that suggests speakers consistently attempt to add new information to the ongoing discourse and that languages develop a variety of ways to express focus on this new information. This is a functional or communicative usage-based constraint that cannot be explained by a simple grammatically correct or incorrect perspective on syntax. Tammelin-Laine and Martin (chapter  5) investigate a ten-month-period of development of Finnish expressions of negation in four multilingual women in a database drawn from regular L2 language classroom observations in an adult L2 literacy program. Assuming a construction grammar, exemplar-based approach rather than a rule-based approach to development, the study is distinct in tracking very low-proficiency adult learners who are not literate in their own L1s and in examining Finnish, a non-Indo-European language with a rather complex negative construction system. Similar to many other studies of the acquisition of negation (see Ortega 2014), the researchers found that the first negative utterances were simplified relative to the target norms, simply involving the presence of the Swedish negative particle ei. Indeed, the data revealed high levels of simplification generally, with only 23 percent of the utterances containing any verb form. Three of the learners gradually added some form of the bare verb stem, but at the end of data collection, the majority of their utterances still lacked verbs. Tammelin-Laine and Martin’s findings support the usage-based position that learners gradually learn exemplars of the negative construction, rather than general rules. This is consistent with what Eskildsen (2012) has shown for adult English L2 learners, who he argues “do not operate on the basis of an early general no + verb/phrase rule, contrary to what is traditionally postulated” (353). In the Finnish, as in the English study, negation development is “a much more dynamic issue […being] both patterned- and learner-specific” (Eskildsen 2012, 353).

Introduction

5

It is likely not an exaggeration to say that, if low-proficiency adult learners who are not literate in their own L1s are woefully understudied, college-educated students of languages represent the most widely studied group of L2 learners or multilinguals. Thompson (chapter 6) argues, however, that although we know a good deal about this group’s strictly linguistic development, we know very little about changes in multilinguals’ subjective metalinguistic awareness and beliefs about how their languages interact in their mind and support new learning. To address this gap, she has proposed the construct of Perceived Positive Language Interaction (PPLI), which can be studied via content analysis of the answers to the following open-ended question: “If you have studied other languages in the past, do you think that this has helped or hindered your ability to learn subsequent languages? Please provide specific examples where appropriate.” In her data, Thompson identified two types of multilinguals. Some firmly believe that knowing and learning multiple previous languages facilitates the new language learning challenges they undertake; such learner beliefs tend to be associated with stronger ideal L2 self and show less L2 English class performance anxiety. For youth in many world geographies (e.g., those studied by Thompson in Brazil, Turkey, China, Sweden, or Saudi Arabia), increasing numbers of college-bound or college-educated youth embark on the study of one or more additional languages beyond English, thus becoming budding “English plus” multilinguals. Across these varied contexts, it is important to understand how these young multilinguals make sense of their successive experiences of language learning and how knowledge of multiple languages interact in their mind not only psycholinguistically (Cook and Li Wei 2016) but also metalinguistically and emically, as Thompson proposes.

Corpus-aided, Usage-based Studies of Learner Language

The next four studies continue with the investigation of multilingual development. But they do so with cross-sectional data sampled from adult learners enrolled in contexts of formal language instruction, in a research tradition that is often called “learner language” in the field of second language acquisition (e.g., Ellis and Barkhuizen 2005; Granger, Gilquin, and Meunier 2013; Han and Tarone 2014). Corpus techniques are central in these investigations, whether employed in full or just partially applied in order to analyze elicited data more systematically or in order to design the elicitation instrument. In addition, other elicitation-based methodologies are featured in these four studies, such as retellings and cloze tests. Two of the language phenomena investigated represent subtle usage patterns that typically do not receive overt attention from language teachers, whereas the other two involve highly complex usage patterns that are typically the overt object of attention in textbooks and classrooms, although often with poor results. Languages vary in terms of the amount of variability in the production of the complementizer their speakers exhibit. Some languages, such as English, allow substantial variability (I thought [that] Jason likes candy), others like Spanish allow almost no zero que-complementizer, while others, such as German, fall midway on the continuum: Production of dass-complementizer is obligatory in adjectival complements

6

Introduction

but optional in subject and direct object complements. Wulff (chapter 7) offers a contrastive corpus-based analysis of variable that-complementizer use. Using a binary logistic regression model, she found that her intermediate-to-advanced-level English learners exhibited fine-grained exemplar-based knowledge of verb-construction associations that was similar to the knowledge shown by native speakers. In addition, nonnative speakers also showed knowledge of mode-dependent variation in that, just like the native speaker participants, they produced that more often in writing than in speech. However, L1 Spanish learners adopted a more conservative approach towards dropping the complementizer than L1 German learners and modality impacted learners’ choices, particularly among German learners more so than among their Spanish peers. Wulff’s findings are suggestive of a fundamental continuity between L1 and L2 usage patterns, but one that is modulated by L1-tuned patterns. In a nutshell, and adopting a usage-based, construction grammar perspective, Wulff argues that L2 development is a complex process founded on exemplar-based knowledge obtained from the input, modulated by preferences acquired in the L1, and heavily constrained (especially in spoken discourse) by processing demands. The complementizer is dropped under ideal processing conditions. As Wulff ’s study illustrates, experience of language, and particularly its specific instantiation of frequency, has been established as an important factor in language learning and processing (Ambridge et al. 2015). But a pending question is how researchers might reasonably investigate frequency in ways that best represent the input to which learners are exposed. In chapter  8, Linford, Long, Solon, and Geeslin address this question, focusing on a comparison of L1 and L2 Spanish users’ expression of overt versus null subject, the latter being the default. As was the case in Wulff’s study (chapter 7), this is a well-studied phenomenon in the L1 that typically does not receive explicit attention from language teachers in the L2. The researchers created a local corpus of sociolinguistic interviews in Spanish with thirteen (bilingual native-speaking) L1 users and twelve (highly advanced nonnative speaking) L2 users and studied how the two groups used, in particular, first-person singular subject expression (e.g., sé, creo, tengo, and estoy + overt or null SP yo) in the discourse condition of plus or minus switch reference. The relative token verb frequency (LF) was measured in several ways. Corpus-internal measures of LF were extracted based on three calculations: the raw count of the verb tokens in the L1 speech of the bilinguals, the raw count of the verb tokens in the L2 speech of the learners, and the L2 + L1 combined verb token frequencies. An external measure of LF was derived as well from the one-hundred-million-word L1-only Corpus del español (http://www.corpusdelespanol.org). The various analyses undertaken showed similarities and differences of results, depending on group (L1, L2) and corpus (local vs. external). Among the most striking differences is the finding that switch reference was significantly associated with more overt-yo use for the frequent verbs only when LF was measured as L2 + L1 combined verb frequencies within the local corpus. Thus, operationalizing verb frequency using the internal versus the external corpus led to different interpretations of the L2 patterns in the data (but not the L1 patterns). Linford et al. cautiously conclude that, while it is impossible with the present data to offer full explanations for these L2 findings, researchers must be advised to

Introduction

7

carefully consider how best to measure frequency and what corpus source to use for such measurement, always making those decisions for particular studies and their specific purposes. Not all languages use the same linguistic resources, such as the English article system, to convey definiteness; in non-article languages, definiteness is often inferred through discourse context or expressed through a range of contextualized factors such as word order and deictic markers. Recognizing that L2 English learners whose L1 does not contain articles have particular difficulty learning the English article system, Ekiert (chapter 9) examines how alternative linguistic resources are used to track entities and definiteness in writing by L2 English users, whose L1 Polish does not have articles. Her data revealed that the bilingual Polish-English learners significantly underused articles relative to the L1 English users. In contrast, all other non-article determiners (e.g., possessives, demonstratives, etc.) were used with greater frequency by the L2 English participants. Particularly interesting is the fact that the Polish L1 discourse contained much lower levels of non-article determiners, indicating that the patterns found in the L2/bilingual discourse was not just a result of L1 transfer. Ekiert hypothesizes that non-article determiners functioned as idiosyncratic interlanguage tools to track reference, suggesting that the L2 participants were struggling with form-meaning connections behind definiteness in the target language. Researchers have long investigated the attributes of English Phrasal Verbs (PVs) which make them difficult for L2 learners to master. Zhao and Le (chapter 10) found that previous work had considered frequency and semantic transparency (i.e., degree of semantic compositionality) but not the additional variable of L2 learner proficiency and decided to investigate the interaction of these three factors. Theirs was a sample of 89 Chinese L1 learners of English, approximately half of whom were independently deemed to be at an advanced level and half at a low-intermediate level of L2 proficiency. Using corpus tools, Zhao and Le developed an untimed discourse-level cloze test that measured semantically transparent PVs of high (give away) and low (lock out) frequency and less transparent PVs of high (make up) and low (live down) frequency. Results indicated main effects for all three variables, with no interactions. First, responses to PVs of higher frequency were more accurate than responses to lower-frequency PVs; this finding is squarely in line with a usage-based perspective. Second, responses to semantically transparent PVs were more accurate than responses to less-transparent PVs; this suggests that learners may rely on a semantic compositional strategy of considering a potential contribution from the individual lexical components to the overall meaning of the construction. Finally, general knowledge of PVs increased with higher English proficiency, except that accuracy on low frequency, low transparency PVs lagged. In sum, nontransparent, low-frequency PVs turned out to be the most problematic category even for the more advanced L2 learners, a finding that leads Zhao and Le to call for more support for instruction of such PVs. A key question, left unaddressed by Zhao and Le, is what instruction to offer to support better learning of PVs. Recently, Mahpeykar (2014) and Mahpeykar and Tyler (2015) have proposed an alternative to the typical approach of rote

8

Introduction

memorization. Using a cognitive linguistics (CL) principled polysemy analysis, these researchers conclude that nearly all meanings of PVs are compositional: Both the verbs and the particles participating in PV constructions have multiple meanings (including extended, nonliteral meanings), as well as a so-called literal meaning. For instance, the literal, high-frequency PV take out has the transparent (i.e., literal) meaning of ‘remove X from a container’ but it also has several nontransparent meanings, such as ‘destroy an enemy.’ The analysis by Mahpeykar (2014) and Mahpeykar and Tyler (2015) reveals that the meanings of the PVs can involve an extended meaning from either the verb or the particle or both. Preliminary experimental studies (Mahpeykar, Tyler, Akiyama, and Jan 2015) suggest that learners benefit from CL-informed instruction, which emphasizes the polysemous nature of English verbs and particles and the variable contribution of each to the multiple meanings of PVs. Combining these new findings on the compositionality of English PVs with Zhao and Le’s findings provides important potential insights into improved L2 pedagogy in this thorny area.

The Experimental Study of Usage-based Processing and Learning

As the chapters in the book show, a key aspect of understanding first versus second or third language learning (L1, L2, L3) involves the role of learning the appropriate categorization and labeling of events, spatial relations, social phenomena, et cetera, of the target discourse community. The L1 learner approaches this task fresh, with no preexisting categories in place. In contrast, the L2/L3 learner comes to the language learning situation with established categories, many of which are inevitably different from the L2. A major task therefore involves learning a new system for categorizing the world. The chapters thus far reviewed addressed this question in contexts involving the family, the classroom, or elicited products of language. With chapters  11 and 12, the book features two important experimental approaches to answering the same relevant questions: the paradigms of self-paced reading and artificial language learning. Both are employed to investigate universal-like biases that are thought to be at work in language: the use of differentially weighted cues to the interpretation of sentences (MacWhinney and Bates 1989) and patterns of crosslinguistic typology (Greenberg 1963). Marijuan, Lago, and Sanz (chapter  11) report on an experimental study that examined the question of whether adult English-speaking advanced learners of Spanish shift from an L1 sentence processing strategy which relies heavily on the cue of word order (N V N = Subject Verb Object) to a more appropriate L2 processing strategy which involves close attention to morphology agreement as a cue (in NcliticV N = Object Verb Subject sequences). Employing a self-paced reading paradigm, the researchers found that their budding bilinguals exhibited variable processing patterns, sometimes inappropriately relying on the L1 word order strategy but also at times successfully using clitic and morphological cues to accurately interpret subject versus object. Moreover, the most favorable condition in order to appropriately process sentences with OVS word order contained an informative (and targetlike)

Introduction

9

mismatch between the preverbal nominal (clitic) and the agreement morphology on the verb, as in Locl-sg estánvb-pl mirandovb las chicaspl (‘The girls are looking at him’). The key findings offered by Marijuan et al. are that L2 learners gradually shift their sentence processing strategies (or their attention to processing cues) and that, when the OclVS sentences convey contrastive agreement, the importance of morphology rises to a high enough level of saliency for the L1 English speakers to begin to reset their processing strategies or the L2 settings for cue validity. Importantly, shifts in processing strategies are not all or nothing. Even when learners can use morphological marking to appropriately assign the roles of subject/agent and object in contrastive OclVS conditions, they do not consistently use this information in non-contrastive situations. This finding highlights the usage-based claim that language learning is exemplar driven and hence gradual. Fedzechkina, Newport, and Jaeger (chapter  12) focus on the investigation of language typologies and argue that the relatively new research paradigm of miniature artificial language learning allows researchers to test specific hypotheses about the origin of frequently occurring, crosslinguistic patterns. Methodologically, participants (infants, children, or adults) are exposed to mini–artificial languages that are designed to focus on certain properties. The authors focus on three that have received attention in both typology and language learning: crosslinguistic preferences for consistent headedness, predictable variation, and trade-offs between competing cues to sentence meaning. From a language learning perspective, the studies show that children tend to produce a version of the artificial language that converges towards established universals, while adults tend to be more conservative, replicating the input more closely. The authors note that even small tendencies toward certain cognitive processing biases on the part of children can have a gradual, cumulative effect over several generations or iterations of learning a language. These tendencies, in turn, result in language systems that, in essence, encode certain learner biases as a basic part of the grammar. Ultimately, this research points to gradual language change whose basis is grounded in human cognitive learning and processing biases.

Multilingualism in the Wild: Usage-based Insights

Finally, the last two chapters in the collection represent exciting usage-based work undertaken in a hospital context in Hong Kong and in naturally occurring adult conversations in Malaysia. Unlike the rest of studies in the book, these contexts are found outside the family, the classroom, the elicited corpus data, or the laboratory. While very different from each other in scope and purpose, the two studies share in common a fundamental concern with human communication and with understanding it in novel theoretical ways that address both human cognition and social situatedness as two irreducible and mutually constitutive elements of human language. Slade, Matthiessen, Lock, Pun, and Lam (chapter 13) use a systemic-functional grammar (SFG) framework and present a fine-grained analysis of two naturally occurring, audio-recorded doctor-patient emergency room consultations in multilingual

10

Introduction

Hong Kong. Considering one consultation as effective and the other as less effective, based on a post-consultation patient satisfaction survey, Slade et al. argue that the doctor-centered discourse of the less effective consultation makes it more difficult for the patient to introduce additional information that may be relevant to the diagnosis, and to have patient concerns and anxieties adequately addressed. This leads directly to the unsatisfactory outcomes of the consultation, with potential negative consequences for patient safety. Conversely, in the more patient-centered discourse of the effective consultation, the patient is able to communicate to the doctor additional information relevant to his diagnosis, and to have his concerns more fully discussed. Slade et al. richly demonstrate the advantages of applying the SFG technical apparatus to take both the interpersonal and experiential aspects of discourse into account. It is this dual analytical perspective that reveals very different patterns in the two consultations, even though they take place in the same hospital context and involve patients with similar presenting conditions. What constitutes effective communication is contextual (e.g., the patterns comprising effective doctor-patient interactions are likely to diverge from effective doctor-doctor surgical communication), emergent, and dependent on a dynamic interaction between the interpersonal and informational. The volume closes with Achard and Lee (chapter 14), who offer a new perspective on code-switching and its multiple effects on the structure of language in high-contact, multilingual contexts. To our knowledge, the present study is the first published exploration of code-switching using the theoretical framework of conceptual blending (Fauconnier and Turner 2002). The data under consideration are naturally occurring recordings of everyday interactions in Kuala Lumpur, Malaysia (KL), where six distinct languages are spoken. The analysis establishes that use of the codes, or languages, spoken in KL not only indexes speaker group identity and status but also speaker stance towards idealized social models (such as modernity and city life versus traditional, rural life). Furthermore, the grammar of each code/language spoken contains blended structures that incorporate input from the neighboring social and linguistic spheres. Hence, each language bears traces of the multilingual usage events the speakers engage in on a daily basis, providing strong support for a usage-based model of language and language change. In other words, each of KL’s distinct languages is in fact a blended grammar, with each language reflecting elements of the others—even in monolingual exchanges. These blended units are shown to be pervasive at all levels of language in all of the individual languages spoken in KL. In the stratified society of KL, speaking Cantonese, English, Mandarin, or Tamil certainly carries important social, group identity relevance, yet each of these codes has structurally accommodated to the presence of the other surrounding codes. In sum, Achard and Lee demonstrate the consequences of the multilingual environment on the internal structure of the different individual languages that speakers of KL use in their daily interactions, highlighting that language emerges through use. The study further establishes the usefulness of examining naturally occurring, everyday interactions through the perspective of blending theory. Finally, it establishes the

Introduction

11

importance of a conceptual blending framework as part of a dynamic, usage-based approach to language.

Looking Forward

In a nutshell, the studies in this book illuminate several insights that the usage-based view offers to the study of language and multilingual development: 1. Meaning is central to the shape of language and guides its acquisition. 2. What is acquired is the whole of the communicative system—that is, an integrative system composed of a variety of semiotic means that include the verbal (or signed) language as well as other semiotic resources such as gestures, actions, facial expressions, gaze, and intonation. 3. Function and discourse shape language usage, whether in L1 or L2; constructions are unified wholes in discourse syntax and have psycholinguistic reality. 4. Input, in the sense of meaningful language use by the surrounding discourse community, guides the emergence of language; awareness of audience design and flexibility of salient semiotic resources are also important forces that aid language learning as emerging from usage. 5. Frequency has a deep effect on developmental trajectories; the mappings most frequently found in the input will be mastered first, based not just on sheer frequency but on cognitive models that first search for meaning in any language-and-action event. 6. Learners learn exemplars rather than general rules. 7. Language learning, whether monolingual or multilingual, is characterized by a good deal of individual variation. Based on the usage-based view of language and language learning that the chapters  in this book showcase, we would like to submit several tentative generalizations that can inspire future research in this domain. First, the application of the usage-based perspective is fruitful across all contexts of language learning, be it the family context (at a young age prior to the onset of schooling), classroom contexts (for school-aged children, youth, and adult learners), elicited usage, the laboratory, or contexts for language use in the wider society (as in hospitals and public encounters). Second, at the broadest level, the same usage-driven forces affect native and nonnative users. Thus, contrary to what is often assumed in deficit-oriented models of ultimate attainment by nonnative speakers, L2 users can derive probabilistic rules from the input without explicit instruction, just as children can. It follows that L1-attuned preferences will add to this general picture, but without creating any discontinuity, nor any fundamental difference. Finally, this book shows that while developmental and corpus linguistics methodologies are abundantly employed in usage-based research, a rich palette of methods, including experimental tools, can be profitably employed as well. We look forward to a vibrant and expanded landscape of usage-based research into language learning and multilingualism in many years to come.

12

References

Introduction

Ambridge, Ben, Evan Kidd, Caroline F. Rowland, and Anna L. Theakston. 2015. “Frequency Effects in First Language Acquisition: A Review [Target Article].” Journal of Child Language 42: 239–73. Cook, Vivian, and Li Wei. 2016. Cambridge Handbook of Linguistic Multicompetence. New York: Cambridge University Press. Dancygier, Barbara, and Eve Sweetser. 2014. Figurative Language. Cambridge, UK: Cambridge University Press. Ellis, R., and G. Barkhuizen. 2005. Analyzing Learner Language. New York: Oxford University Press. Eskildsen, Søren W. 2012. “L2 Negation Constructions at Work.” Language Learning 62: 335–72. Fauconnier, Gilles, and Mark Turner. 2002. The Way We Think: Conceptual Blending and the Mind’s Hidden Complexities. New York: Basic Books. Goldberg, Adele E. 2003. “Constructions: A New Theoretical Approach to Language.” Trends in Cognitive Sciences 7: 219–24. Granger, Sylviane, Gaëtanelle Gilquin, and Fanny Meunier. 2013. Twenty Years of Learner Corpus Research: Looking Back, Moving Ahead. Louvain, Bel.: Presses Universitaires de Louvain. Greenberg, Joseph H. 1963. “Some Universals of Grammar with Particular Reference to the Order of Meaningful Elements.” In Universals of Human Language, edited by Joseph H. Greenberg, 73–113. Cambridge, MA: MIT Press. Han, Zhaohong, and Elaine Tarone. 2014. Interlanguage 40 Years Later. Amsterdam: John Benjamins. Langacker, Ronald W. 1987. Foundations of Cognitive Grammar. Vol. 1, Theoretical Prerequisites. Stanford, CA: Stanford University Press. MacWhinney, Brian. 2000. The CHILDES Project: Tools for Analyzing Talk. 3rd ed. Mahwah, NJ: Lawrence Erlbaum. MacWhinney, B., and E. Bates. 1989. The Crosslinguistic Study of Sentence Processing. New York: Cambridge University Press. Mahpeykar, Narges. 2014. “A Principled Cognitive Linguistics Account of English Phrasal Verbs.” PhD diss., Georgetown University. Mahpeykar, Narges, and Andrea Tyler. 2015. “A Principled Cognitive Linguistics Account of English Phrasal Verbs with Up and Out.” Language and Cognition 7: 1–35. Mahpeykar, Narges, Andrea Tyler, Yuka Akiyama, and Hana Jan. 2015. “Applying Cognitive Linguistics to Issues in L2: English Phrasal Verbs.” Paper presented at the American Association of Applied Linguists, Toronto, Can., March 21–24. Ortega, Lourdes. 2014. “Trying Out Theories on Interlanguage: Description and Explanation over 40 Years of L2 Negation Research.” In Interlanguage: 40 Years Later, edited by Zhaohong Han and Elaine Tarone, 173–202. Amsterdam: John Benjamins.

◆◆◆ PART ONE

Usage-based Development of Language across the Lifespan

This page intentionally left blank

2 ‹‹ A Multimodal Approach to the Development of Negation in Signed and Spoken Languages: Four Case Studies Aliyah Morgenstern Pauline Beaupoil-Hourdel

Sorbonne Nouvelle University Marion Blondel

Centre National de la Recherche Scientifique, UMR 7023 Structures Formelles du Languages Dominique Boutet

Université D’evry-Val-d’Essonne

We address the expression of negation in four longitudinal studies including (1) Madeleine,

a hearing child in multimodal French interactions, (2) Ellie, a hearing child in multimodal English interactions, (3) Charlotte, a deaf child of deaf parents in monomodal French sign language (langue des signes française, LSF) interactions, and (4) Illana, a hearing child with one deaf, one hearing parent in bimodal bilingual (French-LSF) interactions. All the negative multimodal utterances (including French, English, or LSF, symbolic gestures, and actions) were coded and analyzed between twelve and thirty-six months for the four children. We draw the four pathways to illustrate how each child combines symbolic categories and visual/aural modalities in successive steps with respect to her own linguistic environment: Madeleine and Ellie use gestures less as they enter verbal negation but keep using the gestural cues when necessary or for emphasis. Charlotte uses more and more manual and nonmanual combinations including coverbal gestures and LSF items. Illana uses fewer French or LSF negations than her monolingual peers but she exploits a wide range of bimodal combinations. She seems always aware of the visual information she has to maintain in her output. We present quantitative and qualitative analyses of the children’s multimodal path into negation, focusing on PalmUp-Shrug and IndexWave gestures, and also underline 15

16

Morgenstern, Beaupoil-Hourdel, Blondel, and Boutet

systematic properties at the junction between coverbal gestures and signs. To do so, we show (a) their contrastive physiological patterns, and (b) their semantic and pragmatic value for the expression of negation.

Introduction

The expression of negation begins very early in infancy. The study of negation in a pragmatic context is especially propitious in the context of language acquisition. Children learn how to use negation with a variety of semiotic means as a tool to express their needs, their desires, and, ultimately, their will, which is part of establishing their own identity (Morgenstern 2006, 10). Previous research on first language acquisition has highlighted a tight relation between actions, gestures, and speech to express negation. As discussed by Spitz (1957) and Clark (1978), children’s first negative constructions seem to take over from early gestures of rejection and avoidance. The study of the expression of negation in longitudinal data of adult-child conversations is therefore a privileged locus for a multimodal approach to language acquisition, particularly in the framework of a comparison between signing and speaking children. In this exploratory study, the paths of four children’s early language development will be described with a focus on their expression of negation in different modalities and on the use of two conventional gestures: the IndexWave and the PalmUp-Shrug gestures. We use the label PalmUp-Shrug in order to include the main formal components of the gesture that involves the hands, the shoulders, and the head (sometimes even the mouth as defined in Streeck 2009). Through our qualitative and quantitative analyses of the children’s negative productions, the following questions will be investigated: To what extent do they resort to gestural means? Is the use of those gestures linked to the language(s) used by the children, signed or spoken? Does exposure to two languages in two modalities play a role in the use of gestures? We first make a brief overview of the issues at stake. We then present quantitative and qualitative analyses of the children’s multimodal paths into negation and finally discuss their uses of IndexWave and PalmUp-Shrug gestures.

A Multimodal Approach to Language Acquisition

Our work on child language development in hearing and deaf children led us to adopt a multimodal approach to the expression of negation in order to take all the dimensions of children’s communicative systems into account. Like Kendon (2004) and McNeill (2014), we do not restrict the definition of language to the use of words or signs. They are one dimension of a complex system involving other paradigms such as gestures, actions, facial expressions, gaze, and intonation. We consider language as being composed of a vast set of semiotic means on which speakers rely to construct meaning. Since negation is conventionally expressed by gestures (such as the HeadShake, the IndexWave, the PalmUp-Shrug gestures), or words and signs, children progressively acquire various negative symbolic forms in both the visual

Multimodal Development of Negation

17

and aural channels. We thus chose negation as a relevant topic to illustrate how our definition of “language” can be broadened to an integrative system composed of a variety of semiotic means.

Gesture, Speech, and Negation in Language Acquisition

Previous studies on children’s negative spoken productions have shown that no (and its equivalent in other languages) is the most consistently used word throughout the single word utterance period (Pea 1980, 170). Children begin using spoken productions for negation around eighteen months (Tomasello 2003, 228–29). However, gestures and actions precede grammatical words (Clark 1978; Spitz 1957). Guidetti (2005) showed that gestures of negation are among the first symbolic gestures children use, and that when children fully enter the spoken modality, they mostly use words. Several studies have analyzed the role of gestures prior to speech as well as during the one- and two-word utterance period and have concluded that gestures trigger language development (Iverson and Goldin-Meadow 2005). Previous studies on the gesture-word relation have highlighted that symbolic gestures tend to develop in tandem with early words (Acredolo and Goodwyn 1988), which indicates that they are a manifestation of the same cognitive development and represent an alternative means of expression. On the same line, it has been observed that cross-modal combinations (1 word + 1 gesture) facilitate the transition to the two-word stage (Capirci et al. 1996). All studies also highlight a striking individual variation in the extent to which children resort to symbolic gestures and gesture–word combinations (Acredolo and Goodwyn 1988; Capirci et al. 1996; Guidetti 2005). Some children who enter the verbal modality precociously seem to use very few symbolic gestures outside pointing at first (Morgenstern 2009). Such a variation is ascribed to the influence of different factors. On the one hand, symbolic gestures might be easier to process in comprehension and easier to produce than equivalent verbal expressions, which might explain their early emergence in a lot of children; on the other hand, they might be more or less exploited by the child according to social factors like parents’ input or response. The negative HeadShake and the gesture of raising palms up in the air for ‘all gone’ (disappearance) or combining it with a shoulder shrug to express ‘I don’t know’ are conventional gestures, because their meaning is specific to certain cultures and they are learned as such by children in the same types of situations as words or signs. Our study will enable us to make an overview of four speaking and signing children’s negative productions in their longitudinal data with a focus on two specific gestures in order to question whether their productions are linked to the children’s linguistic environment: the PalmUp-Shrug and the IndexWave. The gestural construction, which we have called the PalmUp-Shrug, is a composite posture described as a “compound enactment” (Streeck 2009, 189). It can combine palm-up flips, lifted shoulders, and a lateral HeadTilt (Kendon 2004; Streeck 2009). French speakers in

18

Morgenstern, Beaupoil-Hourdel, Blondel, and Boutet

particular might add mouth shrugs also called lip-pout. These various elements can either separately or combined express absence, uncertainty, incapacity, or helplessness (Debras 2014). The IndexWave is a gesture that has been conventionalized and is used by English and French speakers as well as signers of LSF as an expression of negation: it is an oscillating movement of the index finger held in line with the palm, with the other fingers most often folded back.

Data and Method

This study is part of a larger project on multimodal and multilevel language development in twelve monolingual and children, speaking French, English, Italian, Brazilian Portuguese, and LSF. Our team is composed of specialists of syntax, phonology, semantics, pragmatics, discourse, gesture, and sign who work together on the same data set.1 Our aim is to describe and analyze each child’s individual pathway and find regularities and individual differences in their use of the visual and aural channels as well as the various linguistic levels. In this chapter, we focus on multimodal analyses of negations in the productions of four children between the ages of 1;00 and 3;00 interacting with their parents in various linguistic environments.2 We analyze negative productions in the longitudinal data of Madeleine, a monolingual French child; Ellie, a monolingual British child (Beaupoil-Hourdel, Morgenstern, and Boutet, 2015); Charlotte, a deaf signing child with input in LSF (Limousin 2011); and Illana, a hearing bilingual bimodal child in contact with French and LSF (Tuller, Blondel, and Niederberger 2007). The four children were filmed once a month for one hour at home. Charlotte, the deaf signing child, is raised by her two deaf signing parents and was filmed by Fanny Limousin (then a PhD student), who is a native deaf signer. Illana was filmed in a bilingual, bimodal environment by a bilingual observer. Her hearing mother is bilingual (LSF-French) and her deaf father sometimes produces code-blends. LSF was more prevalent when the deaf father was present and he was a very engaged participant (he was absent during two sessions in the data analyzed for this chapter). We used the videos and the transcriptions when they were available (the data in signed language were not entirely glossed but were tagged for negations). All the parents gave permission for us to use the data in papers and show video clips of the data as well as the children’s faces in pictures. The ethics review board of our university approved the protocol. All the data are spontaneous; we added no experimental design. A specific coding system was developed, combining the use of CLAN and ELAN with the video data and the transcriptions in order to make micro and macro analyses of the functions of the different forms of negation according to context in dialogue. For this study, we restricted the data to one session every six months from the ages of one to three for the four children. This enabled us to code twenty hours of data as shown in table 2–1. We followed a two-step coding process: 1. We used our films and transcriptions in CLAN (spoken data) and ELAN (signed and bimodal bilingual data) to find all forms of negation.

Multimodal Development of Negation

19

‹‹ Table 2–1.  Recordings and negative productions of the four children Negative productions

Hours of video

Ellie

256

5

Madeleine

202

5

Charlotte

117

5

Illana

216

5

Total

791

20

2. We coded them in Excel grids in order to make micro and macro analyses of the type of modality and the functions of the different forms of negation according to context in dialogue. This chapter presents a multilevel study of the emergence and development of negation. We adopted a multimodal methodological approach by analyzing all negative communicative acts the child expressed. This involved coding all spoken negative productions in the child’s data as well as all Acts of Bodily Communication (Zlatev and Andrén 2009) that could be conveyed with actions or gestures. Our research investigates the interface between the visual modalities—actions and gestures— and the auditory modalities—vocalization and speech—and takes into account the combinations of these modalities in the construction of negative meaning. The negative functions (refusal/rejection, nonexistence/absence, denial, negative assertion, epistemic negation, and prohibition) were coded according to three types of forms: 1. Actions, such as pushing away an object. 2. Symbolic conventional gestures: HeadShake for refusal, Shrug for epistemic negation (Streeck 2009, 190), PalmUp for nonexistence (Kendon 2004, 277). We also included what we called SGS (Symbolic Gesture/Sign): gestures shared by gesturing speakers and signers, such as IndexWaves. We only coded gestures used for communicative purposes in interaction. 3. Speech or sign including negative markers in each language. For example, in French non, pas, y a plus, rien; in English no, don’t, not anymore, nothing; in LSF, signs meaning none or I don’t want; as well as lexical negation such as French arrête or English stop it. We thus make a distinction between actions and gestures, but the difference between the two modalities is not always easy to draw when it comes to young children (Willems and Hagoort 2007). We coded the behavior as an action when the movement produced by the child is a reaction to the environment rather than being conventionalized and when the movement does not seem to carry a communicative intention (Liebal and Call 2012, 119). However, it is interpreted by the parents as a negation and integrated in the ongoing dialogue. The children have a complex system at their disposal that includes symbolic and non symbolic means of expressing negation. Figure 2–1 presents the resources at

20

Morgenstern, Beaupoil-Hourdel, Blondel, and Boutet

‹‹Figure 2–1.  Pluri-semiotic resources used by the children, according to the channels at their disposal (visual/ auditory). SGS refers to gestures shared by gesturing speakers and signers, and ONOM refers to onomatopoeia. their disposal. The only child who can use all the resources is Illana, who is brought up bilingual French-LSF and is a hearing child. In this study, we focus our analyses on the use of the visual-gestural and auditory-vocal modalities and will now present the results of our coding of actions, gestures, speech, and sign (and combinations of modalities) for each child.

Analyses of the Four Children’s Multimodal Expression of Negation Ellie’s longitudinal data (monolingual English)

As shown in figure 2–2, at the beginning of the data, action seems to be sufficient for the child to express negation. Beaupoil-Hourdel, Morgenstern, and Boutet (2015) have shown that Ellie begins to use gestures and enters a symbolic mode of expression at 1;02. But as early as 1;06, she is already using an important proportion of symbolic means of expression, and predominantly with gestures (over 30 percent gesture in isolation and 35 percent combinations of speech and gesture). After 1;06, speech is the predominant modality, either in isolation (over 35 percent) or combined with gestures (20 percent) or actions (15 percent). At 2;06, there is a decrease in the use of gestures (less than 10 percent and always in combination with speech). However, she reintroduces gestures at 3;00 in combination with speech (almost 20 percent of all negative productions) and in isolation. Overall, the child uses actions in 35 percent of her negative productions and gesture in almost 30 percent. Even though 71 percent of the child’s productions involve speech, only 35 percent of them are only verbal. Of the negative productions coded for this study, 36 percent are combinations of speech and either an action or a gesture.

Multimodal Development of Negation

21

‹‹Figure 2–2.  Rate of actions, gestures, and speech, and number of occurrences per category in Ellie’s data As speech becomes predominant, Ellie’s spoken negative utterances are more and more complex. We observe that her spoken productions for negation are constantly becoming more elaborate. At 3;00, the child’s negative utterances have a mean length of utterance (MLU) just below three, which is high when you think of the number of occurrences of “no” in isolation that children and adults produce. At 1;00, she only uses the grammatical marker “no.” At 2;00, she can produce constructions such as “all gone” or “couldn’t do it.” At 3;00, she is producing more elaborate utterances like “he can’t push the baby,” “now Pepper, you mustn’t move my toys” (speaking to her cat), or “I don’t like cheese, Mummy” and uses all functions of negation (refusal, epistemic negation, negative assertions). Thus, in only three years, she has developed a good mastery of her mother tongue. Even though speech becomes predominant around 2;00, an analysis of negations restricted to speech would leave aside a great proportion of Ellie’s productions and the role of the visual-gestural modality in her pathway. The comeback of the visual modality in the role of coverbal actions and gestures at 3;00 also indicates that once Ellie has acquired the verbal means to express negations, she can still resort to actions and gestures to complement her speech.

Madeleine’s longitudinal data (monolingual French)

As can be gleaned from figure 2–3, Madeleine’s pathway is quite different from Ellie’s, as she does not use gestures during an intermediary period to enter the symbolic expression of negation. She is already producing speech at the beginning of the data, but mostly in combination with actions interpreted as negative by her addressee and the coder (pushing away toys, turning away from her mother, avoiding spoonfuls of food). At 2;00, her speech in isolation is highly predominant (over 90 percent of her productions). However, coverbal gestures start emerging and are part of 45 percent of her productions at 3;00. Madeleine uses gestures once she has a

22

Morgenstern, Beaupoil-Hourdel, Blondel, and Boutet

‹‹Figure 2–3.  Rate of actions, gestures, and speech, and numbers of occurrences per category in Madeleine’s data finer mastery of speech. As she has been extensively studied in the CoLaJE project, we know that Madeleine’s mastery of speech is quite precocious (Morgenstern and Parisse 2012) and that as of 2;03, she has acquired the French phonological system, uses quite a variety of grammatical tenses (Parisse and Morgenstern 2012), produces 3 argument clauses, prepositions, and connectives (Sekali 2012), refers to herself in the first person (Caët 2013), starts using complex sentences (Sekali 2012), and can self-repair her utterances (Morgenstern, Leroy, and Caët 2013). At 1;00, Madeleine uses the grammatical marker “non” in isolation. The phonological realization of her use of “non” is not yet complete as she pronounces them [næ]. At 2;00, she expresses various functions of negation using a variety of syntactic forms as in “télé éteinte fait rien” (‘TV shut do nothing’), “non pas les brocolis” (‘no, not broccoli’), or “pas fini mon lait” (‘not finished my milk’). Contrary to Ellie at the same age, Madeleine does not use chunks or frozen verbal expressions to convey her negations. At 3;00, Madeleine’s negations have a complex syntactic structure, such as “moi je l’avais ramassé mais maintenant je sais plus où il est” (‘I picked it up but now I don’t know where it is anymore’).

Charlotte’s longitudinal data (monolingual LSF)

Zeshan (2006) underlines that “the relation between signing and gesturing, with both manual and non-manual aspects is important . . . in negation” (29–30). Indeed, the numerous lexical and morphosyntactic forms involved in the sign languages studied so far are tightly linked to the speakers’ gesture systems in the speaking community the signers live in. The coding for Charlotte’s data (see figure 2–4 for main results) is thus slightly different since a number of the signs used in LSF to express negations are shared with the gestural repertoire of both signers and speakers in the

Multimodal Development of Negation

23

‹‹Figure 2–4.  Rate of actions, gestures, and speech, and number of occurrences per category in Charlotte’s data French community. This includes mostly IndexWaves and HeadShakes for negations. Charlotte benefits from input through a single modality, the visual modality, including actions, gestures, and signs, as well as visual input derived from the vocal modality: mouthing and lip reading. Typical LSF (in white in the graph) refers to the core LSF lexicon that hearing people would not use as gestures (the manual lexical signs NON, IL-N-Y-A-PAS we would translate as ‘no’ ‘none’, and the predicative signs incorporating negation such as NE-PAS-VOULOIR (‘don’t want’) or NE-PAS-AIMER (‘don’t like’) as opposed to gestures such as the IndexWave or the HeadShake). We can observe a larger number of shared gesture/sign productions (in black) throughout the data than of specific signs. The LSF core lexicon is only used as of 2;00 and predominantly combined with other gestures in the same production (between 20 and 30 percent of overall productions as of 2;00 combine a shared gesture and a typical LSF sign). As shown in Limousin (2011), Charlotte’s LSF productions during that period are more and more complex. At the beginning of the data, Charlotte mostly expresses rejection like the other children analyzed in our study through what we categorized as actions. She expresses refusal with HeadShakes and IndexWaves. They are gestures that have been incorporated in the sign language system as grammatical signs. All her actions or gestures/signs are produced in isolation. As she gets older, Charlotte produces those same forms but in combination with facial expressions in two or three sign utterances. At 1;06, she starts using negative predicates as well: for example, PT13 NE-PAS-VOULOIR (‘I don’t want’) and PT1 NE-PAS-SAVOIR (‘I don’t know’). Between two and three years of age, her signed productions become more sophisticated and include up to four signs together, as in Neg-index PT1 VOULOIR (‘want’) PT[>food] when she is 3;00.

Illana’s longitudinal data (bilingual bimodal LSF-French)

Illana is a hearing child growing up in a bilingual, bimodal environment. She has all semiotic means to express negation at her disposal, but gesture plays a predominant role in her productions, especially when her deaf father is present during the recordings. Illana’s negation patterns are summarized in figure 2–5.

Morgenstern, Beaupoil-Hourdel, Blondel, and Boutet

24

‹‹Figure 2–5.  Rate of actions, gestures, and speech, and number of occurrences per category in Illana’s data Interestingly enough, Illana does not produce any LSF lexical signs of negation in the selected sessions. Her input is predominantly in French during the sessions, and at 2;00 and 3;00, her deaf father is not present during the recordings, thus isolated French takes up 50 percent of her productions. However, she uses shared gestures much more than the other children and they are part of 37 percent of her productions over the whole data (combined or in isolation). When her father is present, isolated French occurs in 20 to 25 percent of her occurrences only and her use of the visual modality amounts to 73 percent of her productions. The visual forms do not decrease to the advantage of the vocal forms since Illana continues to use HeadShakes, IndexWave negations, and different symbolic gestures or facial expressions. In example 2–1 at 2;06, Illana is answering her father while addressing both her father and mother. They are playing cards with pictures of animals, and the child mixes vocal and labial French with negative symbolic gestures: ‹‹Example 2–1 FAT: CHI: FAT: CHI: FAT: CHI: CHI:

CROCODILE (‘crocodile’) Neg-index / nan trompé c’est pas … c’est un crocodile (‘No, you are wrong this is not… it’s a crocodile’) SE-TROMPER ILLANA (‘You’re wrong Illana’) palm-down gesture/ nan crocodile ! (‘Let’s give up, no, a crocodile’) BALEINE (‘Whale’) nan/Headshake baleine (‘No, this is not whale’) her father takes one card call gesture/ñan c’est à pa(pa), c’est à maman (‘Hey, it is not your turn, it is Mummy’s’)

The majority of her productions are accessible to both her parents, hearing and deaf. As she gets older, she seems to resort more and more to the combination of vocal productions with symbolic gestures. She uses all the bimodal semiotic

Multimodal Development of Negation

25

resources at her disposal to express her negations. She is therefore predominantly a bimodal child who makes use of the visual modality to adjust to her bilingual bimodal environment.

Summary of longitudinal findings

The analysis of each child’s individual pathway into negation clearly demonstrates their differences but there are some common features. All the children who have access to both the gestural-visual and auditory-vocal modalities use them both and they all start expressing their rejections and refusals with actions that are clearly interpreted by their interlocutors and integrated in the ongoing dialogue. The hearing children with no sign language input enter negation at first through actions, but then there follows a period when they either use symbolic gestures (Ellie) or speech (Madeleine). They get more or less rapidly involved in speech and seem to abandon gestures for a while, during what McNeill (2014, 159–61) calls the “Dark Age.” But gestures make a comeback with the use of coverbal gestures of negation when speech seems to be already quite elaborate. Charlotte, the deaf child, also first expresses negation with bodily actions, then uses symbolic gestures that are incorporated as signs in the LSF linguistic system and which are present from the beginning in their input. Illana, the bilingual bimodal child has created an efficient transitional system during her developmental path by combining modalities. The necessity to enter two languages at once and to speak both to deaf and hearing addressees might have an influence on the management of the visual-gestural modality, which is a stable resource to rely on in all the types of linguistic environments she experiences. The visual modality is of course crucial for Illana when she wants to address her deaf father. In order to make finer distinctions between the children’s use of conventional gestures, we will now focus on the PalmUp-Shrug and the negative IndexWave.

Focus on Two Specific Gestures: the PalmUpShrug and the Negative IndexWave

We conducted a closer analysis of the use of the four children’s IndexWave and PalmUp-Shrug gestures in the data. Table 2–2 shows the results of our coding. The number of occurrences produced in our data is not very high but we could observe differences that might indicate general trends to be tested on more signing/ non-signing deaf and hearing children. 1. The signing children, Illana and Charlotte, are the only ones to use IndexWaves (nineteen and ten occurrences, respectively) in the data.4 2. The hearing children use PalmUp-Shrugs when the deaf child does not in the sessions we coded for this study. 3. The bilingual bimodal hearing child uses more IndexWaves and PalmUp-Shrugs than all the other children together. She seems to be resorting to the visual modality more than all the other hearing children together in the data as we’ve

Morgenstern, Beaupoil-Hourdel, Blondel, and Boutet

26

‹‹ Table 2–2.  Number of occurrences of PalmUp-Shrugs and IndexWaves in the four girls’ data Ages  

1;00

1;06

2;00

2;06

3;00

Total

PUShrug

0

0

0

7

5

13

Index

0

0

0

0

0

0

PUShrug

0

9

6

1

3

19

Index

0

0

0

0

0

0

PUShrug

0

0

0

0

0

0

Index

1

1

2

6

9

19

PUShrug

0

8

10

14

2

34

Index

4

0

4

0

2

10

All gestures

Madeleine 13

Ellie 19

Charlotte 19

Illana 44

seen in part 3.4, especially when her father is present, as she maintains visual information even when she is speaking. In the next two sections, we will present qualitative analyses in order to explore the semantic and pragmatic functions in dialogue of the IndexWave and the PalmUp-Shrug gestures that might account for the quantitative differences observed between the uses of the signing versus non-signing as well as hearing versus deaf children.

Qualitative analyses of the use of the IndexWave

The IndexWave is used in many cultures as a gesture expressing negation, but with various functions ranging from refusal, negative assertions to deontic values such as prohibition or negative obligation (Calbris 1990; Jorio 2000). In LSF, the IndexWave is a gesture/sign used extremely frequently, often associated with the HeadShake and possessing the same range of negative values (Limousin 2011). In our data, the IndexWave is not used by the two non-signing girls up to 3;00. We have only found it in the adults’ data in interaction with their children when they are expressing a deontic modality meaning ‘you are not allowed to…’ or ‘don’t do that.’ We do observe in the Paris Corpus (Morgenstern and Parisse 2012) that it is occasionally used by some children after 3;06 to forbid something to their dolls, their dogs, or to warn the adults not to do something. For example, when Ellie is 3;07 (after the period analyzed in this study), she is cutting star-shaped cookies in raw dough and putting them on a tray. She tells her grandmother that they should not be cooked by saying “Not cook yet [+ IndexWave]” the grandmother reformulates the gesture and

Multimodal Development of Negation

27

words together by saying “Not to cook yet, ok” understanding Ellie’s production as carrying a deontic value. The IndexWave thus seems to be produced by our hearing speakers to express a deontic modality, which is often used when speakers have an asymmetric status in the dialogue and can exert some pragmatic ‘power’ over their interlocutors. Children thus begin using it with animals or dolls but do eventually address it to adults when they are more expert speakers and make more subtle and complex multimodal productions with subjective positioning. The two signing girls, however, make a much more frequent use of the IndexWave in a broader range of negative contexts. Our hypothesis is that they are in contact with many more occurrences of that gesture with a larger variety of functions since it is part of the sign language system and is produced with or without the HeadShake to express refusal as well as negative assertions in a variety of contexts. Charlotte is already using it in the first session of the data coded for this study, at 1;00, as shown in example 2–2. ‹‹Example 2–2 At 1;00, Charlotte’s mother is changing her diaper. After she takes off the dirty diaper, Charlotte starts wiggling her legs and arms. Her mother explains she shouldn’t move and uses both an IndexWave and a HeadShake in her explanation. Charlotte smiles then takes up the IndexWave repeatedly. She then produces a HeadShake.

‹‹Figure 2–6.  Charlotte 1;00 IndexWave As of 2;00, all Charlotte’s productions of IndexWaves are refusals and negative assertions (‘I’m not washing my hair’ for example at 2;00 when she is in the bath). Figures 2–7 and 2–8 in example 2–3 in Illana’s data illustrate the same use in a bilingual context. ‹‹Example 2–3 At 1;00, Illana is having dinner in her high chair and her mother has just refused to give her the cheese the child was pointing at. The mother offers her yogurt or a fruit, the father a glass of water. The child answers both parents by saying “nan” (a proto “non”) associated with an IndexWave. ‹‹Figure 2–7.  Illana 1;00 IndexWave + vocal-verbal production

Morgenstern, Beaupoil-Hourdel, Blondel, and Boutet

28

The father then turns to the mother, takes up his daughter’s IndexWave, and adds a pouting mouth. The more frequent use of the IndexWave by Charlotte and Illana thus seems to be linked to the fact that it carries a wider range of functions: for the two little girls surrounded by sign language, the same form first used as an isolated gesture is incorporated into the child’s sign language system and can thus be for‹‹Figure 2–8.  Illana 1;00 Father-child alignment: mally categorized as a sign. IndexWave + verbal production or pouting

Qualitative analyses of the use of the PalmUp-Shrug gesture

In adult language the PalmUp-Shrug can express absence, uncertainty, incapacity, or helplessness (Debras 2014). In our adult-child data up to 3;00, the children only express either absence, which could be glossed as ‘gone’ / ‘no more’ / ‘not here,’ or lack of knowledge, capability, or responsibility, corresponding to ‘I don’t know’ / ‘I can’t do it’ / ‘what now?’. There are very few occurrences of any version of the PalmUp-Shrug gesture in Charlotte’s whole data and we have found zero occurrences in the five sessions coded for this study. The three hearing children all use that gesture, English-speaking Ellie as well as French-speaking Madeleine and Illana. In what follows, we discuss examples that illustrate Illana’s, Ellie’s, and Madeleine’s uses. Illana, the bilingual bimodal hearing child, uses thirty-four PalmUp-Shrugs in the data. Example 2–4 shows how she can use the gesture with her deaf father. We will also revisit her uses in the next section, and comment on her multimodal uses in more depth. ‹‹Example 2–4 At 1;06, as her father has just forbidden her to drink the bath water by using a very distinct IndexWave, Illana looks at the bath, at her toys and then at her father and makes a series of PalmUp gestures that seem to mean ‘what now, what can we do now?’ as if she did not know what to play with next.

‹‹Figure 2–9.  Illana 1;06 IndexWave Her father then takes up the gesture that seems very close to the LSF sign QUOI for ‘what?’ The pictures show him with both his hands and arms open.

Multimodal Development of Negation

29

Ellie uses quite a number of PalmUp gestures with no addition of shoulder shrugs at the beginning of the data when she does not yet produce many vocalizations. Both her mother and grandmother are enchanted by her early use of this conventional gesture and reinforce it by imitating her or by formulating their own interpretation. For example the grandmother says after Ellie has done ‹‹Figure 2–10.  Illana 1;06 Father takes up the PalmUp gesture the PalmUp gesture as she is looking for her basket “We don’t know where it is do we, Ellie?” and then turns to the mother to explain “She says ‘I don’t know’.” At 2;00, Ellie starts combining her PalmUp gestures with vocal productions. All Ellie’s PalmUp gestures up to 2;00 are interpreted in context as meaning ‘Where is it?’, ‘Gone,’ or ‘Done’ and are accomplished without a marked shoulder component. As of 2;06, Ellie starts diversifying her use of the components and functions of the PalmUp-Shrug. She adds shoulder shrugs and HeadTilts and some occurrences can be interpreted as meaning ‘I don’t know.’ But as of 3;00, Ellie can combine gesture and word in a complementary fashion. ‹‹Example 2–5 E llie

(3;00)

E llie : A c t i on : E llie : Gaze: G es t ure : Mother : A c t i on :

0 [=! making a horse noise]. Plays with her horse toy. Where’s the yellow horsy? looks up. PalmUp gesture on the right hand.

[=! neighs]! Plays with a horse toy.

‹‹Figure 2–11.  Ellie 3;00 PalmUp gesture In the excerpt shown in example 2–5, Ellie is playing with her mother. She puts a horse toy into a small stable her mother built for her, then she looks up, flips her

Morgenstern, Beaupoil-Hourdel, Blondel, and Boutet

30

right hand up and asks “Where’s the yellow horsy?” At 3;00, Ellie is able to combine her gestures with speech. The gesture she produces is less visible than the ones she used before 2;00. As Ellie grows older, her gestures are less and less distal as in this example, when she produces a hand flip with her hand held close to her body and not extended in the adult’s direction. At 3;00, Ellie’s gestures are less visibly displayed in the environment and more subtle. In this sequence, Ellie produces the gesture with speech and uses the word where. She is now able to put her thoughts into words without her mother’s or her grandmother’s help. However, the combination of gesture and speech is not redundant. Her spoken utterance is a question and therefore asks for missing information; the gesture shows that the referent is not within the child’s field of vision. In this combination, the speech is not syntactically or semantically negative, yet in context, the gesture is interpreted with negative meaning and expresses absence. Madeleine does not use as many gestures as Ellie at the beginning of our data, and she only starts using the PalmUp-Shrug as a coverbal gesture. She does so as of 2;06 and at 3;00 has already become an expert at using all the semiotic resources at her disposal to express subtle nuances. ‹‹Example 2–6 M adeleine Mother :

G es t ure : Mother : Gaze: M ad :

G es t ure : Mother : G es t ure : Gaze:

(3;00) Là il nous offrait un jaune et bleu (‘We need to find a yellow and blue one’) points to a fish on the board game Il est pas loin(‘it’s very close’) looks at the board game. Je le vois c’est celui+là [=! whispers]. (‘I can see it, it’s this one’) points to a fish on the board game Non bah 0 [=! ‹‹Figure 2–12.  Madeleine 3;00 PalmUp-Shrug (one smiles] (‘No’) hand) Shrug looks at Madeleine

Multimodal Development of Negation Mother : M ad :

G es t ure : Gaze: Mother :

M ad : G es t ure :

31

Mais non! (‘No it isn’t’)

[=! upset]? (‘So which one is it then?’) PalmUp-Shrug on the board game. Bah cherche! (‘Well, look for it!’) Moi je sais pas! (‘I don’t know!’) PalmUp-Shrug, HeadTilt

‹‹Figure 2–13.  Madeleine 3;00 PalmUp-Shrug (two hands) In example 2–6, Madeleine is playing a board game with her mother. She is looking for a yellow and blue fish but she can’t find it. She gets upset and asks “Et lequel alors ?” (‘So which one is it then?’). She asks for the fish and at the same time she produces a PalmUp-Shrug gesture. The rising prosodic contour of her utterance and the gesture forms she uses together contribute to express her exasperation. As illustrated in the textgrid and spectrogram in figure 2–14, Madeleine starts her utterance with a pitch at 149 Hz on “et” and her maximum pitch reaches 459 Hz on the second syllable of “lequel.” When Madeleine pronounces “lequel,” she starts shrugging and her palm-up gesture expands on “alors.” The combination of high pitch and a proximal gesture (shoulder-lift) ending in a distal gesture (palm-up configuration), along with the child’s gaze set on the game and not on the mother, led us to code this utterance and the gesture as an expression of both exasperation and powerlessness rather than an actual question directed at her mother. The second instance of Madeleine’s gesture is combined with the spoken utterance “Moi je sais pas !” (I don’t know) and is composed of three distinct forms: a PalmUp on both hands, a Shrug, and a HeadTilt. This gesture, because it is coupled with the spoken utterance ‘I don’t know,’ has an epistemic meaning but like the previous one it appends powerlessness and exasperation to the whole meaning of the spoken utterance. Interestingly, we note that the second instance of the gesture is more emphatic than the previous one and is composed of three components.

Morgenstern, Beaupoil-Hourdel, Blondel, and Boutet

32

‹‹Figure 2–14.  Text grid and spectrogram for “Et lequel alors ?” (‘So which one is it then?’) The three speaking children thus make expert uses of the PalmUp-Shrug gesture they have replicated and learned in specific relevant contexts from their linguistic and cultural environment. Its forms and functions evolve according to the children’s age. On the other hand, in her mono-modal LSF interactions, Charlotte does not use the PalmUp-Shrug before her signing becomes more complex in order to express disengagement and lack of knowledge. This may be due to the facts that she is less exposed before 2;00 to the outside hearing community’s conventional gesture system, and that her signing parents use the sign to mean ‘Where?’ or ‘I don’t know,’ which she takes up in the middle of her second year.

The role of gestures in Illana’s multimodal productions

Illana uses fewer French or core LSF negations than her monolingual peers but she exploits a wide range of shared gesture and bimodal combinations. She is always aware of the visual information she has to maintain in her output when her father is part of the conversation. In examples 2–7 and 2–8 she uses both an IndexWave and a PalmUp-Shrug gesture in a very short period of time. ‹‹Example 2–7 At 2;00, Illana is playing with her deaf father and her bilingual hearing mother with a memory game. The mother has just pretended to associate the cards with a picture on the board and Illana indicates her disagreement by saying “Non non non” (‘no, no, no’) associated with a salient IndexWave.

‹‹Figure 2–15.  Illana 2;00 IndexWave

Multimodal Development of Negation

33

‹‹Example 2–8 A little later, Illana’s father takes advantage of the game to check that his daughter understands the signs for different animals represented on the cards. He points towards the picture of an animal with an inquisitive gaze towards her. Illana does not know the answer and uses a PalmUp-Shrug to express her ignorance. ‹‹Figure 2–16.  Illana 2;00 PalmUp-Shrug Even though Illana’s signed repertoire is not as rich as her spoken one during the length of the data, she seems to have perfectly understood how to construct a multimodal system to communicate with both her hearing mother and deaf father simultaneously by combining speech with very clear shared conventional gestures and some signs. We make the hypothesis that she derives the PalmUp-Shrug mainly from the French cultural input but uses it more often than the other hearing children for her father’s benefit. She learns the IndexWave with its wider range of functions mainly from the signed input and uses it a little less than the deaf signing child. The total of her negative gestures is quite higher than the other children in a wider variety of functions and contexts.

Conclusion

Ellie, Madeleine, Charlotte, and Illana follow four very different pathways into negation, but for all four children, gestures play an important role in their itineraries. Ellie enters conventional negation through her use of gestures. Once she masters speech, gestures are used to reinforce or complement her spoken utterances. Madeleine enters the verbal modality from the very beginning of the data, but she starts combining coverbal gestures with her speech in a complex manner sooner and gives quite subtle indications as to her affects and positioning thanks to multimodal means. Charlotte also enters conventional negation through gestures. Most of those gestures are then incorporated in her signed grammar but she continues to produce ‘co-signed’ gestures. Bilingual bimodal Illana’s use of gestures is quite special. She enters conventional negations through gestures as well, but her use of gestures does not decrease during the recordings as she maintains them as a communication mode suitable for both hearing and deaf interlocutors. The four little girls we have studied follow trajectories that are logically linked to the repertoire of resources available in the input. They all have a common gestural inventory available in the input, but the deaf-signing child does not resort to PalmUp-Shrugs and the non-signing children do not use the IndexWave gesture for communicative purpose during the period of the study. We do find those gestures more often in the data we filmed when they are older.

34

Morgenstern, Beaupoil-Hourdel, Blondel, and Boutet

This shared gestuality needs to be explored further and with older children as it is integrated very differently in the four types of input and in the linguistic systems that each child constructs in interaction. Gestures are part of the fabric of language. A more thorough investigation of their role in language development and language use could lead to a better understanding of such stuff as both spoken and signed language are made on.

Acknowledgment

This study was supported by the Education, Audiovisual, and Culture Executive Agency (EACEA), Project number 543264-LLP-1-2013-1-IT-KA2-KA2MP, Sign Language. We are grateful to Fanny Limousin, who defended one of the first PhDs in France on sign language acquisition and collected the first longitudinal dataset of a deaf signing little girl, which she allowed us to use for this chapter.  We thank our anonymous reviewer for his or her precious comments on our chapter and the editors of this volume for their constant care and help during the editing process.

Notes

1. Methodologies and Evaluation Tools, SignMET project: http://www.istc.cnr.it/project/signmet. 2, The data were collected in the framework of the CoLaJE project (Communication Langagière chez le Jeune Enfant) ANR-08-COM-O21 funded by the Agence Nationale de la Recherche: http:// colaje.scicog.fr/. 3. PT represents pointing, PT1 is a self-point. 4. Madeleine produces two IndexWaves when she repeats a French nursery rhyme with the conventional gestures associated with it on the word guère (old French for ‘not much’). Even though the use of that gesture in a song might have an impact on her later use of the same gesture, we have decided to only code communicative gestures in interaction.

References

Acredolo, Linda, and Susan Goodwyn. 1988. “Symbolic Gesturing in Normal Infants.” Child Development 59: 450–66. Beaupoil-Hourdel Pauline, Boutet Dominique & Morgenstern Aliyah. (2015). A child’s multimodal negations from 1 to 4 : The interplay between modalities. In Pierre Larrivée and Chungmin Lee, Eds. Negation and Polarity : Experimental Perspectives, Language, Cognition, and Mind, chapter 4, pp. 95–123. Springer International Publishing Switzerland. DOI:10.1007/978-3-319-17464-8_5 Caët, Stéphanie. 2013. “Référence à soi et à l’interlocuteur chez des enfants francophones et anglophones et leurs parents.” PhD diss., Université Sorbonne Nouvelle – Paris 3. Calbris, Geneviève. 1990. The Semiotics of French Gestures: Advances in Semiotics. Bloomington: Indiana University Press. Capirci, Olga, Jana M. Iverson, Elena Pizzuto, and Virginia Volterra. 1996. “Gestures and Words during the Transition to Two-word Speech.” Journal of Child Language 23: 645–73. Clark, Eve V. 1978. “From Gesture to Word, on the Natural History of Deixis in Language Acquisition.” In Human Growth and Development, edited by Jerome Seymour Bruner and Alison Garton, 85–120. Wolfson College Lectures 1976. Oxford, UK: Oxford University Press. Debras, Camille. 2014. Multimodal Stance-taking in a videotaped corpus of discussions about environmental issues in British English. Unpublished Ph.D. dissertation, English Linguistics, Sorbonne Nouvelle, Paris.

Multimodal Development of Negation

35

Guidetti, Michèle. 2005. “Yes or No? How Young French Children Combine Gestures and Words to Agree and Refuse.” Journal of Child Language 32: 911–24. Iverson, Jana M., and Susan Goldin-Meadow. 2005. “Gesture Paves the Way for Language Development.” Psychological Science, 16: 368–71. Jorio, Andrea de. 2000. Gesture in Naples and Gesture in Classical Antiquity: A Translation of La Mimica Degli Antichi Investigata Nel Gestire Napoletano, Gestural Expression of the Ancients in the Light of Neapolitan Gesturing. Translated by Adam Kendon. Bloomington: Indiana University Press. Kendon, Adam. 2004. Gesture: Visible Action as Utterance. Cambridge, UK: Cambridge University Press. Liebal, Katja, and Joseph Call. 2012. “The Origins of Non-human Primates’ Manual Gestures.” Philosophical Transactions of the Royal Society B: Biological Sciences 1585: 118–28. Limousin, Fanny. 2011. “Acquisition de la référence personnelle en LSF: Analyse longitudinale des pointages, des formes nulles et des noms signés chez une enfant sourde de parents sourds.” PhD diss., Université Paris 8. McNeill, David. 2014. “Gesture–Speech Unity: Phylogenesis, Ontogenesis, and Microgenesis.” Language, Interaction and Acquisition 2: 137–84. Morgenstern, Aliyah. 2006. Un JE en Construction. Paris: Presses de la Sorbonne Nouvelle. ———. 2009. L’enfant dans la langue. Paris: Presses de la Sorbonne Nouvelle. Morgenstern, Aliyah, and Christophe Parisse. 2012. “The Paris Corpus.” French Language Studies 22: 7–12. Morgenstern, Aliyah, Marie Leroy, and Stéphanie Caët. 2013. “Self- and Other-repairs in Child-Adult Interaction at the Intersection of Pragmatic Abilities and Language Acquisition.” Journal of Pragmatics 56 151–167. Parisse, Christophe, and Aliyah Morgenstern. 2012. “The Unfolding of the Verbal Temporal System in French Children’s Speech between 18 and 36 Months.” Journal of French Language Studies 22: 95–114. Pea, Roy D. 1980. “The Development of Negation in Early Child Language.” In The Social Foundations of Language and Thought: Essays in Honor of Jerome Seymour Bruner, ed. David R. Olson, 156–86. New York: W. W. Norton. Sekali, Martine. 2012. “The Emergence of Complex Sentences in a French Child’s Language from 0;10 to 4;01: Causal Adverbial Clauses and the Concertina Effect.” Journal of French Language Studies 22: 115–41. Spitz, René A. 1957. No and Yes: On the Genesis of Human Communication. New York: International Universities Press. Streeck, Jürgen. 2009. Gesturecraft: The Manu-facture of Meaning. Amsterdam and Philadelphia: John Benjamins. Tomasello, Michael. 2003. A Usage-based Theory of Language Acquisition. Cambridge, MA: Harvard University Press. Tuller, Laurie, Marion Blondel, and Nathalie Niederberger. 2007. “Growing Up in French and French Sign Language.” In Handbook of French, Applied Linguistics, edited by Ayoun Dalila, 334–76. Amsterdam: John Benjamins. Willems, Roel M., and Peter Hagoort. 2007. “Neural Evidence for the Interplay between Language, Gesture, and Action: A Review.” Brain and Language 101: 278–89. Zeshan, Ulrike. 2006. “Negative and Interrogative Constructions in Sign Languages: A Case Study in Sign Language Typology.” In Interrogative and Negative Constructions in Sign Languages, edited by Ulrike Zeshan, 28–68. Sign Language Typology Series 1. Nijmegen, Nld.: Ishara Press. Zlatev, Jordan, and Matts Andrén. 2009. “Stages and Transitions in Children’s Semiotic Development.” In Studies in Language and Cognition, edited by Jordan Zlatev, Matts Andrén, C. Lundmark, and M. Johansson Flack, 380–401. Newcastle, UK: Cambridge Scholars.

This page intentionally left blank

3 ‹‹ Why Don’t You Just Learn it from the Input? A Usage-based Corpus Study on the Acquisition of Conventionalized Indirect Speech Acts in English and German Ursula Kania

University of Liverpool

Within traditional approaches, the comprehension of all indirect speech acts (ISAs) involves the consideration and rejection of the ‘literal’ interpretation. More recent cognitive approaches assume that language users operate with cognitive models (e.g., for the act of ‘giving’) and that ISAs activate a specific part of a model (e.g., ‘the hearer is able and willing to perform the act’) which results in the spread of activation to other parts of the model (via metonymic links; Panther and Thornburg 2003). Within this framework, utterances conventionally used as ISAs are potentially entrenched as linguistic constructions whose ‘indirect’ functions are directly associated with the respective forms (Stefanowitsch 2003), which is reminiscent of earlier attempts to conceptualize them as idiomatic expressions. Since linguistic conventions are established during language acquisition, approaching this issue from a developmental perspective promises to be particularly enlightening. Most developmental studies to date, however, are experimental and either do not consider the degree of conventionality of the ISAs under investigation or rely on a priori assumptions about which ISAs are conventionalized (e.g., Bucciarelli, Colle, and Bara 2003). Furthermore, crosslinguistic evidence is not considered. The current study addresses these issues on the basis of three CHILDES corpora (Thomas and Adam for English, Leo for German; MacWhinney 2000). Analyses of two English constructions (Can I X?/Why don’t you X?) and their German counterparts show that constructions which are predominantly used as ISAs by the caretakers are first comprehended and produced as such by the children. Furthermore, they are sometimes paraphrased by ‘direct’ expressions, which may help the child to establish 37

38

Ursula Kania

associative links between these constructions and to expand his linguistic repertoire. Overall, the study indicates that children are indeed sensitive to the degree of conventionality of ISAs in the input and that this factor helps predict in which order the respective form-function mappings are learned.

Introduction

Usage-based, constructivist approaches assume that a speaker-hearer’s knowledge about language consists of a “structured inventory of conventional linguistic units” (Langacker 1987, 57). These units are conceptualized as constructions (i.e., form-function pairings with varying degrees of abstractness and complexity; Goldberg 1995, 2006). They are posited to be organized in a network, with associative links between formally and/or functionally related constructions. Children are assumed to make use of their general cognitive and socio-cognitive skills (e.g., intention-reading and pattern-finding) in language learning, slowly building their linguistic repertoire through interaction with mature language users (Tomasello 2003). In this process, learning is expected to proceed in a piecemeal fashion, with children only gradually abstracting away from lexically specific constructions. Furthermore, the absolute and relative frequency with which a construction is found in the input is seen as a driving factor. In the current study, this framework is applied to an analysis of the acquisition of two constructions which are commonly assumed to be conventionalized indirect speech acts (henceforth: ISAs). Consequently, the following section provides an overview of the relevant theoretical background and previous research. I then introduce the data, after which I present the results on formal coding, functional coding, the children’s age of acquisition for production as well as comprehension, and variations sets, respectively. The chapter concludes with a discussion along with suggestions for further research.

Theoretical Background and Previous Research

The notion of ‘indirect speech act’ was first introduced by Searle in the context of his speech act theory (1969, 1975), which he developed on the basis of work done by Austin (1962). One crucial assumption within this framework is that there is a prototypical mapping between the syntactic form (or: sentence mood) of an utterance and its illocutionary force (i.e., its intended meaning). For example, syntactic questions are prototypically intended to serve as sincere requests for information (e.g., What’s this?), which is why these cases are considered ‘direct’ or ‘literal’ uses. A syntactic question used to a different end (e.g., as a request for action like Can you close the window, please?) is consequently termed ‘indirect.’ This mismatch between the form of the utterance and its pragmatic function is traditionally assumed to pose a problem for the recipient. For example, within the standard pragmatic model (Searle 1975, 62–63), correct interpretation is suggested to involve a complex inferencing process, with the hearer initially considering but ultimately rejecting the ‘literal’ interpretation. While Searle states that “the problem is made more complicated [emphasis added] by the fact that some sentences seem almost to be conventionally used as indirect

Why Don’t You Just Learn it from the Input?

39

requests” (1975, 60), others propose that these cases can consequently be conceptualized as speech act idioms (Sadock 1974). However, this approach was criticized on the grounds that expressions conventionally used as ISAs can evoke both the literal as well as the nonliteral meaning while this is (presumably) not the case for idiomatic expressions (Munro 1979). Within usage-based Cognitive Linguistics, the theory of speech act metonymies has been proposed in order to account for the speaker’s motivation for and likelihood of choosing a particular expression and the hearer’s ability to arrive at the intended meaning without noticeable effort (Panther and Thornburg 1998, 2003, 2004; Thornburg and Panther 1997). In this approach, ISAs involve so-called illocutionary metonymies (i.e., they activate [peripheral] parts of a particular speech act scenario, with the activation eventually spreading to the whole scenario by means of basic conceptual relationships). For example, the sentence Can you pass the salt? evokes the notion that ‘the hearer is able and willing to perform the act’ and the activation spreads to the core of the model since there is a basic conceptual relationship between ability and action (which, in turn, is part of the scenario of ‘giving’). Crucially, this model allows for the indirect meaning to be derived ad hoc with relative ease; it can, however, “through frequency of use, become a conventionalized meaning, stored separately in the lexicon” (Panther and Thornburg 2004, 97). This possibility is closely related to Stefanowitsch’s (2003) proposal to conceptualize conventionalized ISAs as constructions (i.e., pairings of form and ‘indirect’ meaning; Goldberg 1995, 2006).1 The model is dynamic in nature, since metonymic links may be strengthened or weakened, depending on the speaker-hearer’s previous experience with particular scenarios and expressions. Unfortunately, only very few studies to date provide quantitative, corpus-based analyses of expressions which are conventionally used in the context of particular scenarios (but see Adolphs 2008 and Aijmer 1996). There is even less relevant research in the area of language acquisition, even though the conventionalization of the respective linguistic constructions is a socio-pragmatic process which takes place during early language development. In general, it can be stated that “corpus-based research into utterance function is a relatively under-explored area at present” (Adolphs 2008, 15), which may be due to the fact that not all corpora contain sufficient contextual information to allow for functional coding. Moreover, functional coding is very time-consuming, considerably limiting the amount of data that can be analyzed. Within language acquisition research, the issue of ISAs has been addressed on the basis of both corpora and experiments. Corpus-based evidence is often merely anecdotal, with researchers often expressing surprise at the fact that rather young children (i.e., below the age of three) are apparently able to interpret ISAs correctly, as is shown in this statement about Why don’t you X?: What has always amazed me about these pseudo questions is that Adam, long after he was answering genuine Why questions with ‘Because,’ and some explanation or other, did not once apply this syntactic and semantic analysis to fake Why questions, but invariably treated them as orders—which is what they, pragmatically speaking, were. (Brown 1977, 21)

40

Ursula Kania

Similar observations can be found in Dore (1974), Ervin-Tripp (1977), Garvey (1975), and Horgan (1977). Particularly noteworthy is the work by Shatz (1978, 1979), which provides an in-depth analysis of mothers’ use of questions with different discourse functions and the corresponding responses by children. Shatz (1979) focuses on indirect versus direct speech acts without, however, investigating the functions expressed by particular constructions. Furthermore, in both of her studies, the children’s own uses of the constructions were not considered, the corpora were cross-sectional and rather small, and all analyses were collapsed across subjects, thus potentially masking individual variation. There is also a considerable body of experimental research, mostly on children’s comprehension of ISAs (e.g., Bara and Bucciarelli 1998; Bernicot and Legros 1987; Bucciarelli, Colle, and Bara 2003; Carrell 1981; and for a review, see Gibbs 1994). While providing insights into children’s comprehension skills under carefully controlled conditions, comprehension strategies employed by children in rather artificial tasks may differ from the ones used in naturalistic contexts. Moreover, many studies focus on older children (i.e., above the age of three or even four; Carrell 1981) and usually there is no comparison of conventionalized versus nonconventionalized ISAs; for example, Carrell merely states that “not all interrogative forms are more difficult [to interpret as ISAs] than all declarative forms” (Carrell 1981, 341) while Bernicot and Legros (1987) explicitly focus on nonconventionalized ISAs only. To my knowledge, the only studies to date to test for the comprehension of conventionalized versus nonconventionalized ISAs are those conducted by Bara and Bucciarelli (1998) and Bucciarelli, Colle, and Bara (2003), who worked with Italian children aged 2;60–7;00. However, their classification of ISAs as either conventionalized or nonconventionalized is based on introspection only. Overall, it can be said that the acquisition and naturalistic use of constructions assumed to be conventionalized ISAs is still underresearched.

Research Questions

The current corpus-based study aims at addressing some open questions by investigating two English constructions which have been suggested to be conventionalized ISAs (Can I X?/Why don’t you X?; Adolphs 2008; Aijmer 1996) and their (assumed) German counterparts (Kann ich X?/Warum X Du nicht?; Weigand 1989). Since usage-based research has found that children only slowly abstract away from rote-learned formulas to more abstract constructional schemas (e.g., Dąbrowska 2000; Tomasello 2003), it is to be expected that these partially filled constructions provide a fruitful level for analysis. The questions to be addressed are: 1. Which discourse functions are actually expressed by these constructions in child (-directed) speech? 2. When are the different uses (‘direct’ versus ‘indirect’) comprehended and produced by the children? To what extent is the mastery of direct versus indirect uses related to the distribution of ‘direct’ versus ‘indirect’ uses in the input?

Why Don’t You Just Learn it from the Input?

41

3. How are the constructions embedded in child-caretaker interaction? In particular, do instances of the ‘indirect’ uses enter into paraphrase relations with their ‘direct’ counterparts?

The Data

For the current study, use was made of three longitudinal corpora made available through the open-source Child Language Data Exchange System (CHILDES; MacWhinney 2000). For English, the Thomas corpus (Lieven, Salomo, and Tomasello 2009; age span 2;0.12 [years;months.days]–4;11.20) and the Adam corpus (Brown 1973; age span 2;3.04–5;02.12) were chosen. For German, the Leo corpus (Behrens 2006; age span 1;11.12–4;11.05) was selected. Both the Thomas and the Leo corpora are high-density corpora (with a total of 379 and 501 recording sessions, respectively), offering the possibility to analyze the target children’s language development on a rather fine-grained level. Even though the Adam corpus is low-density (with a total of 55 recordings), it was included in the analysis since the target child uses a very high number of syntactic yes-no questions overall (over four thousand tokens; Estigarribia 2010, 73).

Results Formal coding

Using the COMBO command of the CLAN program, all utterances including either of the strings Can I X?/Kann ich X? and Why don’t you X?/Warum X Du nicht? were extracted. The results were then exported to MS Excel for further coding. Since the search also yielded irrelevant results (e.g., wh- questions like What can I do with that?, structures without a full verb, like Can I?) and some utterances contained unclear material that made a further analysis impossible, these cases were discarded manually, as were imitations, self-repetitions, and cases of Can I X?/Kann ich X? containing the negator not in its full or contracted form. Additionally, all target utterances were extracted along with a discourse frame of +/– ten utterances in order to facilitate functional coding and to allow for the identification of variation sets. Table 3–1 summarizes the results of the extraction procedure just described. As shown in table  3–1, even though, in total, the number of utterances produced by the adults and the children is roughly the same (622 and 607, respectively), the numbers are not evenly distributed over speakers and constructions. While the children make more use of Can I X?/Kann ich X? than their caretakers, the picture is reversed for Why don’t you?/Warum X Du nicht? While syntactic complexity certainly is a factor here, discourse function may also play a role. This issue will be briefly addressed later. Additionally, crosslinguistic differences can be observed: In the English corpora, the constructions under investigation are used with a higher relative frequency than in the German corpus. For example, Kann ich X? accounts for only 0.07% of Leo’s utterances (n = 112.813; Steinkrauss 2009, 73), while Can I X? accounts for 0.5% of Adam’s utterances (n = 46.498; Diessel 2004, 8). If these constructions can indeed be shown to be used (predominantly) as ISAs, this distribution would

Ursula Kania

42

‹‹ Table 3–1.  Formal frequencies of the target constructions Construction

Corpus

Speaker

Can I X?

Thomas

INP

290

Thomas

CHI

292

Adam

INP

12

Adam

CHI

228

Kann ich X? Why don’t you

Warum X Du nicht?

Number

Leo

INP

24

Leo

CHI

78

Thomas

INP

131

Thomas

CHI

2

Adam

INP

144

Adam

CHI

6

Leo

INP

21

Leo

CHI

1

Total INP

622

Total CHI

607

Note. INP = all utterances addressed to the child, collapsed across speakers; CHI = all utterances by the child.

be in line with research showing that ISAs are more common in English than in German (House and Kasper 1981), even though it must be kept in mind that this observation is based on three corpora only.

Functional coding

In comparison with formal coding, functional coding procedures are often less straightforward, more time-consuming, and potentially more prone to error, since discourse function is hard to operationalize, coding usually has to be done manually, and judgments are often subjective to some degree. Therefore, this section discusses the reasoning behind the coding choices made in the current study in some detail. Within usage-based, constructionist approaches, children are assumed to make use of their intention-reading skills in order to interpret the intended meaning of particular utterances in particular situations (Tomasello 2003, 3–4). In this context, intended meaning is often equated with discourse function, since if “someone uses a piece of language with a certain communicative intention, […] we may say that that piece of language has a certain function” (Tomasello 2003, 3). This definition of function is closely related to the one in traditional speech act theory, where it is in turn associated with the notion of illocutionary force (Austin 1962; Searle 1969). It will be applied here for two reasons: apart from focusing on the aspect that is probably most relevant to the child, this approach also allows for the classification of uses in terms of the traditional distinction between ‘direct’ versus ‘indirect.’ A direct speech act involves a prototypical mapping between the structural form

Why Don’t You Just Learn it from the Input?

43

of an utterance and its intended meaning. In the case of syntactic questions, the (assumed) prototypical meaning is a sincere request for information. Consequently, all other intended meanings (e.g., requests for action) are traditionally defined as indirect. For Can I X?/Kann ich X?, a direct use thus involves A asking whether (B believes) (s)he is capable of doing X, as exemplified in 3–1 with the child Leo (CHI) and his mother (MOT): ‹‹Example 3–1.  Leo, 3;06.242 CHI: MOT:

Kann ich schon das bauen? [Can I already build this?] Das weiss ich nicht, Leo, das musst du selber ausprobieren und [...] dann feststellen, ob du ‘s schon kannst. [I don’t know, Leo, you have to try for yourself and see whether you can already do it.]

For Why don’t you X/Warum X Du nicht?, a direct use consists of A asking B why (s)he doesn’t X: ‹‹Example 3–2.  Thomas, 4;00.09 CHI: MOT: MOT:

You remember why I don’t want milk. No. Why don’t you want milk?

As an indirect speech act, Can I X?/Kann ich X? can be used as a request for either action or permission. Since the focus of the current analysis is on ‘direct’ versus ‘indirect’ uses, both intended meanings have been subsumed under the category ‘indirect’ and no further differentiation was made. Besides, both meanings can be implied simultaneously. In these cases, the recipient can respond to both the request for permission and the request for action by producing a verbal and nonverbal reaction, as exemplified in 3–3: ‹‹Example 3–3.  Thomas, 3;06.03 CHI: MOT: MOT: MOT:

Can I have some [of] yours, can I? Okay. Yes. There you are.

Indirect uses of Why don’t you X?/Warum X Du nicht? involve a request for action, as exemplified by 3–4: ‹‹Example 3–4.  Adam, 2;07.14 MOT: CHI:

Why don’t you put that on the table? Okay.

Ursula Kania

44

It has been argued that Why don’t you X?/Warum X Du nicht? may also function as a suggestion, leaving the recipient “free to decide whether or not he/she wants to act on the suggested proposition” (Adolphs 2008, 52). However, due to the hierarchical nature of caretaker-child interactions, the child recipient is usually not free to decide. This is also mirrored in the fact that Why don’t you X? is sometimes paraphrased by direct orders, a fact which will be discussed in more detail in the Results section on variation sets below. In order to assess reliability, a second coder categorized two hundred randomly selected utterances (Cohen’s kappa = 0.93; Cohen 1960). Table 3–2 summarizes the results of the functional extraction procedures. It can be seen in table 3–2 that, in the vast majority of cases (75–100 percent), Can I X?/Kann ich X as well as Why don’t you X? are used as ISAs by both the caretakers and the children (i.e., these constructions exhibit a very consistent mapping of form and ‘indirect’ function). Even though no strong conclusions can be drawn from such a small sample, it is obvious that the reversed pattern is found for Warum X Du nicht?: It is predominantly used as a ‘direct’ speech act by the caretakers and the only time Leo himself uses this structure, it also constitutes a ‘direct’ use. This is contrary to the claim made in the literature that this construction is conventionally used as an ISA (e.g., Weigand 1989, 226), showing how important it is to analyze actual language data. Overall, it can be said that the assumption that the other three constructions are conventionalized ISAs is borne out by the data, as is the prediction that the distribution of ‘direct’ versus ‘indirect’ uses in the child’s own productions will be heavily influenced by the pattern found in the input. It is important to note, however, that the current analysis only looks at one of the many ways in which patterns ‹‹ Table 3–2.  Results for function Function Construction

Corpus

Speaker

Can I X?

Thomas

Kann ich X? Why don’t you?

Warum X Du nicht?

Indirect

Direct

Total

INP

277 (95%)

13 (4%)

290 (100%)

Thomas

CHI

285 (98%)

7 (2%)

292 (100%)

Adam

INP

10 (83%)

2 (17%)

12 (100%)

Adam

CHI

221 (97%)

7 (3%)

228 (100%)

Leo

INP

18 (75%)

6 (25%)

24 (100%)

Leo

CHI

75 (96%)

3 (4%)

78 (100%)

Thomas

INP

113 (86%)

18 (14%)

131 (100%)

Thomas

CHI

2 (100%)

0

2 (100%)

Adam

INP

134 (93%)

10 (7%)

144 (100%)

Adam

CHI

6 (100%)

0

6 (100%)

Leo

INP

4 (19%)

17 (81%)

21 (100%)

Leo

CHI

0

1 (100%)

1 (100%)

Note. INP = all utterances addressed to the child, collapsed across speakers; CHI = all utterances by the child.

Why Don’t You Just Learn it from the Input?

45

found in the input may exert an influence on the child’s language development: a semasiological perspective was adopted, showing that children are sensitive to the distribution of ‘direct’ versus ‘indirect’ uses within one particular linguistic form. Of course, it would also be interesting to tackle this issue from an onomasiological perspective, taking into account all forms used to express the same (or at least a very similar) discourse-pragmatic function (for a developmental study taking into account onomasiological/conceptual frequency, see Kania 2013). The analyses in the following section evaluate the prediction that the mapping most commonly found in the input will also be the one to be mastered first by the child in both comprehension and production.

Production/comprehension

Children’s language comprehension cannot be measured directly, so all methods used to this end are by definition indirect. In the current case, the utterances require the recipient to produce an overt (verbal and/or nonverbal) reaction, which makes it possible to judge whether the intended meaning has been understood. However, assessing the comprehension of particular structures with the use of corpus data only is difficult, since one cannot control for factors that may guide the interpretation process beside the utterance under investigation (e.g., linguistic as well as nonlinguistic context, nonverbal cues like gestures). In other words, we cannot be certain how exactly the child has arrived at the correct interpretation of the target utterance. Then again, the usage-based approach proceeds from the assumption that knowledge about language emerges from language use (Tomasello 2003, 5), and in everyday language use utterances are usually embedded in a communicative environment rich with information. Consequently, speakers are expected to make use of additional sources of information in language comprehension under nonexperimental conditions. Based on this reasoning, the age of acquisition (AoA) for comprehension of the respective structure was operationalized as the second instance in which the child produced a pragmatically appropriate (verbal and/or nonverbal) response. Besides using the second (and not the first) instance, the measure is also made more conservative by the fact that cases in which the child failed to produce an appropriate reaction were coded as ‘not comprehended correctly,’ even though the child may have understood the utterance but was just not willing to comply without making this explicit (as has already been noted elsewhere; see Ervin-Tripp 1977, 179); there are also cases which do not require the child to react since there is a shift in attention (e.g., with the child pointing at something he has just discovered). At first glance, assessing children’s age of acquisition for the production of a particular structure on the basis of corpus data is far more straightforward. However, it has to be kept in mind that even a dense sampling regime only captures about 7–15 percent of the child’s language (Lieven and Behrens 2012, 226). Therefore, having no cases of the target structure in the transcript does not necessarily mean that the child does not (yet) produce this structure. Moreover, even if a structure is not used at all, this does not automatically entail that it is not part of the child’s linguistic repertoire. Yet, even if the respective structure were potentially available for production, its nonuse indicates that it is not communicatively useful (enough) for the child

Ursula Kania

46

and/or its representation is too weak to be activated at this point. With these issues in mind, the age of acquisition for production was operationalized as the second token used by the child (note that imitations and self-repetitions had already been discarded from the original dataset). It should also be kept in mind that the primary focus of the analysis is on the comparison between the active/passive mastery of indirect versus direct uses, not on comprehension versus production. The results are summarized in table 3–3. As can be seen in table  3–3, at first glance the commonly observed tendency for comprehension to precede production can also be found here (note that there are, however, also areas where the opposite is the case and production thus precedes comprehension; Grimm et al. 2011). Within both comprehension and production, Can I X?/Kann ich X? is mastered before Why don’t you X? (with Warum X Du nicht? still not being mastered by the end of the recording period), which is probably due to the difference in structural complexity. Crucially, the indirect use is mastered in both comprehension and production before the direct one, with the exception of Why don’t you X? in the Thomas corpus and Warum X Du nicht? in the Leo corpus. In the former case, both the direct and the indirect meanings are acquired at the same time; in the latter case, the direct use is mastered first. This is, however, in line with the prediction that the mapping most frequently found in the input will be mastered first, regardless of whether this constitutes a ‘direct’ or an ‘indirect’ use in the traditional sense. What has to be kept in mind when looking at these results is that the frequency of occurrence of direct versus indirect forms is not comparable (with direct uses being far less frequent than indirect uses overall). Still, what the analysis shows is that the traditional assumption that the ‘direct’ mapping is somehow more primary and is thus used earlier than the ‘indirect’ one is not borne out by the data. In the next and final section of the Results, the discourse context, which may help the children establish associative links to related constructions, is explored by analyzing so-called variation sets.

Variation sets

The analysis of variation sets, or paraphrase relations, is one of many ways of taking into account (linguistic) context. They are defined as “sequence[s] of utterances with a constant intention but varying form” (Küntay and Slobin 2002, 7). Child-directed ‹‹ Table 3–3.  Age of acquisition (AoA) for comprehension and production Construction Can I X? Kann ich X? Why don‘t you X?

Corpus Thomas Adam Leo

Thomas Adam Warum X Du nicht? Leo

AoA comprehension Indirect Direct 2;01.03 2;08.06 3;01.26 — 2;00.16 3;11.12

AoA production Indirect Direct 2;10.07 3;07.03 3;02.21 3;03.04 2;09.13 3;06.24

2;01.25 2;03.04 2;07.22

4;11.08 4;01.15 —

2;01.24 2;08.16 2;05.20

— — —

Why Don’t You Just Learn it from the Input?

47

speech is particularly suited for the analysis of variation sets, since one adult turn often consists of several utterances, which are often redundant and contain many reformulations (Furrow, Nelson, and Benedict 1979; Hoff-Ginsberg 1990; Snow 1972). They are of interest for the current study since they may provide insights into (the psychological reality of ) different types of links between constructions (Fischer 2008). In particular, they make functional and/or formal relationships between similar constructions salient for the child, which is why they have been suggested to play a facilitative role in language development (Kania, forthcoming; Küntay and Slobin 2002). Even though a discussion of this issue is beyond the scope of this chapter, variation sets may also be used to analyze empirically the role played by speech act scenarios in actual discourse, since different paraphrases activate different parts of the respective scenario. For the current analysis, only child-directed speech was considered and a discourse window of +/– ten utterances was chosen. This was done mainly on pragmatic grounds, since a manual inspection of the whole transcripts in order to identify so-called “speech act episodes” (Adolphs 2008, 88) was not deemed feasible. The results are summarized in table 3–4. Overall, as table 3–4 shows, 11–38 percent of all target utterances formed part of variation sets, with the exception of Can I X? in the Adam corpus and Warum X Du nicht? in the Leo corpus, for which no variation sets were found. Interestingly, only ‘indirect’ uses entered into paraphrase relations, which may, however, also be due to the fact that ‘direct’ uses were much less frequent overall. What is more, all target utterances were used initiatingly (i.e., they were paraphrased in one of the following utterances but they did not ever function as paraphrases of preceding utterances). This indicates that they are indeed the unmarked option for the speaker when it comes to expressing this particular meaning (at least in these specific situations), since potential functional competitors are activated and used only later on in the same discourse episode. It has been suggested that “the interactive function of most variation sets seems to be to repeat the same content in order to maximize the chance of comprehension and/or compliance on the part of the hearer” (Küntay and Slobin 2002, 9). Therefore, it might be argued that the caretakers use ‘direct’ paraphrases because the child has failed to understand the ‘indirect’ utterance. However, it was shown in ‹‹ Table 3–4.  Frequency of variation sets Construction Can I X?

Corpus

Speaker

Variation sets

Number

Thomas

INP

31 (11%)

290

Adam

INP



12

Kann ich X?

Leo

INP

9 (38%)

24

Why don’t you

Thomas

INP

19 (14%)

131

Adam

INP

25 (17%)

144

Leo

INP



21

Warum X Du nicht?

Note. INP = all utterances addressed to the child, collapsed across speakers.

Ursula Kania

48

the preceding section that the children understand the ‘indirect’ constructions from rather early on. Furthermore, the majority of variation sets in the current dataset occur after the respective child has already been shown to have mastered the construction in both comprehension and production (e.g., after 2;10 for Can I X? in the Thomas corpus). Therefore, it is more likely that these variation sets aim at maintaining joint attention and/or at increasing the child’s willingness to comply. Even though the caretaker may not be doing this consciously, the paraphrases also offer the child the opportunity to establish/strengthen links to related constructions and/or to expand his linguistic repertoire, as can be observed in example 3–5: ‹‹Example 3–5.  Thomas, 2;08.04 MOT: MOT: MOT: MOT:

Can I just have a look in your mouth, Thomas? […] Let Mummy have a little look, sweetheart. Open your mouth. Open wide.

In this case, all paraphrases constitute more ‘direct’ instances of a request for action. The dataset also contains instances where Can I X? is used as a request for permission and is consequently paraphrased by an alternative structure expressing this function, in this case May I X?: ‹‹Example 3–6.  Thomas, 2;06.29 MOT: CHI: MOT: MOT:

Can I have it please? No. If you drop that on the floor and it spills I shall be very cross. Please may I have it.

While in example 3–6, the child articulates his unwillingness to comply, providing a potential motivation for the mother to use a paraphrase, there are also examples in which the paraphrase is offered even though the caretaker (Mechthild, or MEC, an observer who frequently visits Thomas and his mother at home) can assume that the child is willing to comply. This is illustrated in example 3–7: ‹‹Example 3–7.  Leo, 2;04.06 M EC : M EC : M EC : CHI:

Pilot an Flughafen-Tower, bitte melden sie sich, bitte melden sie sich. [Pilot to airport-tower, please respond, please respond]. Kann ich bei ihnen landen? [Can I land here?] Darf ich bei ihnen landen, in China? [May I land here, in China?] Ja. [Yes.]

For Why don’t you X?, variation sets contain a direct order, with more or less lexical overlap, as illustrated in examples 3–8 and 3–9, respectively:

Why Don’t You Just Learn it from the Input?

49

‹‹Example 3–8.  Thomas, 2;08.13 MOT: MOT:

Why don’t you go and show Dimitra this? Go and show her what you’ve got.

‹‹Example 3–9.  Adam, 3;08.14 MOT: MOT:

Why don’t you try to throw an animal in the bag? […] Pick the animals up and throw them in.

Even though no strong conclusions can be drawn from the current analysis, it should be kept in mind that the absolute and relative frequency of variation sets is potentially even higher than indicated in table 3–4, since a discourse window of only +/– 10 utterances is relatively narrow considering the nature of caretaker-child interactions. Moreover, only the caretakers’ variation sets were included, and only those cases were counted in which the paraphrases were provided by the same speaker, even though there are quite a few instances in which the caretaker provides a paraphrase and/or expansion in response to the child’s utterance, as in example 3–10: ‹‹Example 3–10.  Thomas, 3;07.02 CHI: MOT: MOT: CHI: MOT: CHI:

Mummy, can I have that because I wanted it? Well what do you say? […] Please may I have. Please can I [have] that playdoh? Please may I have it. Please can I have it?

In this example, the mother not only wants Thomas to use the politeness marker please but would also like him to use may instead of can. However, Can I X? as a request for permission is apparently already so entrenched at this point that Thomas continues to use it, expanding the utterance with please but not substituting may for can. Interestingly, this case can also be considered a variation set for the child. Of course, overall we would expect children’s variation sets to be far less frequent, since their linguistic repertoire is not adult-like in size (yet) and more mature interlocutors are usually very cooperative and responsive to children’s vocalizations. Still, there are a few interesting cases, which show that the child knows alternative constructions for expressing his communicative intention: ‹‹Example 3–11.  Adam, 3;02.21 CHI: CHI: CHI:

Can I have some sugar? […] I want some sugar. I like sugar.

50

Ursula Kania

Summary of results

In sum, the results reveal that Can I X?/Kann ich X? as well as Why don’t you X? are predominantly used by the caretakers to express discourse functions that have traditionally been classified as ‘indirect,’ while Warum X Du nicht? was used primarily as a sincere request for information. As is predicted within the usage-based approach, the patterns found in the input have a strong influence on the acquisition of these constructions by the children: In both comprehension and production, the ‘indirect’ meaning for Can I X?/Kann ich X? as well as Why don’t you X? is acquired first, while the reverse pattern can be observed for (the comprehension of ) Warum X Du nicht? A discussion of so-called variation sets provided converging evidence for the fact that Can I X?/Kann ich X?/Why don’t you X? are indeed the unmarked (i.e., highly conventionalized) option to realize the respective ‘indirect’ speech act.

Discussion and Suggestions for Further Research

Since Why don’t you X?/Can I X?/Kann ich X? exhibit a very consistent mapping of form and ‘indirect’ function in both the input and the children’s speech, they are indeed highly conventionalized ISAs, thus allowing for their conceptualization as constructions in the sense of Stefanowitsch (2003). Consequently, the pairing of form and ‘indirect’ function is the prototypical meaning for these constructions, and it is this transparent mapping that is acquired first by the children. Therefore, it is likely that the constructions only become potentially ambiguous later on, at a point where contextual information—in particular the preceding discourse context—may aide the interpretation process (e.g., Adolphs 2008). Furthermore, the constructions were found to be part of so-called variation sets, which provide further evidence for the suggestion that they are often the unmarked option for realizing the respective communicative act. Additionally, variation sets may provide insights into associative links between constructions. However, the current findings must be interpreted with caution, since they are based on three longitudinal corpora only. Therefore, it remains to be seen if and to what extent they represent linguistic conventions on a more general level. More data would also allow for a more comprehensive and systematic analysis of variation sets (e.g., with regard to their potential role in the activation of [parts of] speech act scenarios). Furthermore, just how item-specific the linguistic representations are is an open question, even though the data analyzed for the current study already reveal interesting patterns; for example, in the Thomas corpus, all ‘direct’ uses of Why don’t you X? contain either like (n = 6) or want (n = 12) as the main verb, while Can I have X? is always used as an ISA and accounts for a very large proportion of Can I X? overall in both the input (39%) and the child’s speech (45%). It has been suggested that the systematic investigation of larger lexico-grammatical patterns may help reduce the ambiguity claimed to be inherent in many cases (Sinclair 1996), which ties in with usage-based claims about the prevalence of item-specific constructions or ‘prefabricated units’ (e.g., Wray 2002).

Why Don’t You Just Learn it from the Input?

51

More data is also needed in order to address potential associative links to (formally and/or functionally) similar constructions (e.g., the corresponding ‘direct’ expressions or other ‘indirect’ expressions evoking different parts of the respective speech act scenario). This requires an onomasiological (i.e., function-based) approach, as it is necessary to take into account all linguistic forms that express a given function. Since it is not possible to identify all relevant structures automatically, this type of research requires a lot of resources and studies using this approach are consequently still sparse (but see Aijmer 1996; Kania 2013). In addition to analyzing more corpora containing child (-directed) speech, peer-to-peer interactions between older children and adolescents may also be of interest, since they would provide potential insights into when and how children’s use of the respective constructions becomes adult-like. Lastly, the focus of this study was, as is still common practice, on an analysis of the transcripts, which for the most part merely provide an orthographic representation of child (-directed) speech. As a consequence, other aspects of the interaction (prosodic characteristics, nonverbal signals like gestures and facial expressions, other visual and auditory information available to the interlocutors) could not be taken into account, even though they form an integral part of the communicative environment.

Conclusion

Overall, the results of the current study indicate that children are sensitive to the degree of conventionality of particular form-function mappings found in the input. For the case of ISAs, this means that structures that are highly conventionalized are probably learned as pairings of form and ‘indirect’ function right away, which also has implications for the representation of these constructions in adult language users. However, the analyses presented here are only a first step, with many issues awaiting further investigation. In particular, more quantitative, corpus-based research is needed in order to shed further light on the acquisition and use of constructions associated with particular speech act scenarios.

Notes

1. Note that, as a result, the constructions are not ‘indirect’ speech acts at all. Moreover, “in a corpus-based approach, there is no need for the distinction between direct and indirect speech acts [since] the starting point is the speech act expression rather than the sentence mood” (Adolphs 2008, 11). In order to acknowledge this while still allowing for the application of the traditional and hence well-established terminology, both ‘direct’ and ‘indirect’ will be used with inverted commas. 2. To ease readability, all examples from the transcripts are presented in conventional writing.

References

Adolphs, Svenja. 2008. Corpus and Context: Investigating Pragmatic Functions in Spoken Discourse. Studies in Corpus Linguistics v. 30. Amsterdam: John Benjamins.

52

Ursula Kania

Aijmer, Karin. 1996. Conversational Routines in English: Convention and Creativity. Studies in Language and Linguistics. London: Longman. Austin, John L. 1962. How to Do Things with Words: The William James Lectures Delivered at Harvard University in 1955. Cambridge, MA: Harvard University Press. Bara, Bruno G., and Monica Bucciarelli. 1998. “Language in Context: The Emergence of Pragmatic Competence.” In Cognition and Context, edited by Ana C. Quelhas and Frederico Pereira, 317–45. Lisboa: Instituto Superior de Psicologia Aplicada. Behrens, Heike. 2006. “The Input-Output Relationship in First Language Acquisition.” Language and Cognitive Processes 21: 2–24. Bernicot, Josie, and Suzanne Legros. 1987. “Direct and Indirect Directives: What do Young Children Understand?” Journal of Experimental Child Psychology 43: 346–58. doi:10.1016/0022-0965(87)90012-9. Brown, Roger. 1973. A First Language: The Early Stages. Cambridge, MA: Harvard University Press. ———. 1977. “Introduction.” In Talking to Children: Language Input and Acquisition; Papers from a Conference Sponsored by the Committee on Sociolinguistics of the Social Science Research Council (USA) [held 6–8 September 1974 … in Boston, Massachusetts], edited by Catherine E. Snow and Charles A. Ferguson, 1–27. Cambridge, UK: Cambridge University Press. Bucciarelli, Monica, Livia Colle, and Bruno G. Bara. 2003. “How Children Comprehend Speech Acts and Communicative Gestures.” Journal of Pragmatics 35: 207–41. doi:10.1016/ S0378-2166(02)00099-1. Carrell, Patricia L. 1981. “Children’s Understanding of Indirect Requests: Comparing Child and Adult Comprehension.” Journal of Child Language 8: 329–45. doi:10.1017/S0305000900003226. Cohen, Jacob. 1960. “A Coefficient of Agreement for Nominal Scales.” Educational and Psychological Measurement 20: 37–46. doi:10.1177/001316446002000104. Dąbrowska, Ewa. 2000. “From Formula to Schema: The Acquisition of English Questions.” Cognitive Linguistics 11: 83–102. Diessel, Holger. 2004. The Acquisition of Complex Sentences. Cambridge, UK: Cambridge University Press. Dore, John. 1974. “A Pragmatic Description of Early Language Development.” Journal of Psycholinguistic Research 3: 343–50. Ervin-Tripp, Susan M. 1977. “Wait for Me, Roller-skate!” In Child Discourse, edited by Susan M. Ervin-Tripp and Claudia Mitchell-Kernan, 165–88. New York: Academic Press. Estigarribia, Bruno. 2010. “Facilitation by Variation: Right-to-left Learning of English Yes/No Questions.” Cognitive Science 34: 68–93. doi:10.1111/j.1551-6709.2009.01053.x. Fischer, Kerstin. 2008. “Deriving Relations between Grammatical Constructions from Child Directed Speech.” Dritte Internationale Konferenz der Deutschen Gesellschaft für Kognitive Linguistik, Leipzig, September 26. http://www.uni-leipzig.de/~gcla08/upload/abstr13.pdf (website no longer active). Furrow, David, Katherine Nelson, and Helen Benedict. 1979. “Mothers’ Speech to Children and Syntactic Development: Some Simple Relationships.” Journal of Child Language 6: 423–42. doi:10.1017/S0305000900002464. Garvey, Catherine. 1975. “Requests and Responses in Children’s Speech.” Journal of Child Language 2: 41–63. doi:10.1017/S030500090000088X. Gibbs, Raymond W. 1994. The Poetics of Mind: Figurative Thought, Language, and Understanding. Cambridge, UK: Cambridge University Press. Goldberg, Adele E. 1995. Constructions: A Construction Grammar Approach to Argument Structure. Chicago: University of Chicago Press. ———. 2006. Constructions at Work: The Nature of Generalization in Language. New York: Oxford University Press. Grimm, Angela, Anja Müller, Cornelia Hamann, and Esther Ruigendijk, eds. 2011. Production-Comprehension Asymmetries in Child Language. Studies on Language Acquisition 43. Berlin: Mouton De Gruyter. Hoff-Ginsberg, E. 1990. “Maternal Speech and the Child’s Development of Syntax: A Further Look.” Journal of Child Language 17: 85–99. Horgan, Dianne. 1977. “How to Answer Questions When You’ve got Nothing to Say.” Journal of Child Language 5: 159–65.

Why Don’t You Just Learn it from the Input?

53

House, Juliane, and G. Kasper. 1981. “Politeness Markers in English and German.” In Conversational Routine: Explorations in Standardized Communication Situations and Prepatterned Speech, vol. 2, Janua Linguarum, edited by Florian Coulmas, 157–85. Series maior 96. The Hague: Mouton Publishers. Kania, Ursula. 2013. “‘Can I Ask You Something?’: The Influence of Functional Factors on the L1-Acquisition of Yes-No Questions in English.” Yearbook of the German Cognitive Linguistic Association 1: 109–27. ———. Forthcoming. The L1-Acquisition and Use of Yes-No Questions in English. Tübingen, Ger.: Gunter Narr. Küntay, Aylín C., and Dan I. Slobin. 2002. “Putting Interaction back into Child Language: Examples from Turkish.” Psychology of Language and Communication 6: 5–14. Langacker, Ronald W. 1987. Foundations of Cognitive Grammar. Vol. 1. Stanford: Stanford University Press. Lieven, Elena V. M., and Heike Behrens. 2012. “Dense Sampling.” In Research Methods in Child Language: A Practical Guide, edited by Erika Hoff, 226–39. Malden, MA: Wiley-Blackwell. Lieven, Elena V. M., Dorothé Salomo, and Michael Tomasello. 2009. “Two-year-old Children’s Production of Multiword Utterances: A Usage-based Analysis.” Cognitive Linguistics 20: 481–507. doi:10.1515/COGL.2009.022. MacWhinney, Brian. 2000. The CHILDES Project: Tools for Analyzing Talk. 3rd ed. Mahwah, NJ: Lawrence Erlbaum. Munro, Allen. 1979. “Indirect Speech Acts Are Not Strictly Conventional.” Linguistic Inquiry 10: 353–56. Panther, Klaus-Uwe, and Linda Thornburg. 1998. “A Cognitive Approach to Inferencing in Conversation.” Journal of Pragmatics 30: 755–69. doi:10.1016/S0378-2166(98)00028-9. ———. 2003. “Introduction: On the Nature of Conceptual Metonymy.” In Metonymy and Pragmatic Inferencing, edited by Klaus-Uwe Panther and Linda Thornburg, 1–2. Philadelphia: John Benjamins. ———. 2004. “The Role of Conceptual Metonymy in Meaning Construction.” Metaphorik.de 6: 91–116. Sadock, Jerrold M. 1974. Toward a Linguistic Theory of Speech Acts. New York: Academic Press. Searle, John R. 1969. Speech Acts: An Essay in the Philosophy of Language. Cambridge, UK: Cambridge University Press. ———.1975. “Indirect Speech Acts.” In Speech Acts, edited by Peter Cole and James L. Morgan, 59–82. New York: Academic Press. Shatz, Marilyn. 1978. “Children’s Comprehension of Their Mothers’ Question-directives.” Journal of Child Language 5: 39–46. ———.1979. “How to Do Things by Asking: Form-function Pairings in Mothers’ Questions and Their Relation to Children’s Responses.” Child Development 50: 1093–99. Sinclair, John. 1996. “The Search for the Units of Meaning.” Textus 9: 75–106. Snow, Catherine E. 1972. “Mothers’ Speech to Children Learning Language.” Child Development 43: 549–65. Stefanowitsch, . 2003. “A Construction-based Approach to Indirect Speech Acts.” In Metonymy and Pragmatic Inferencing, edited by Klaus-Uwe Panther and Linda L. Thornburg, 105–26. Amsterdam: John Benjamins. Steinkrauss, Rasmus G.-A. 2009. “Frequency and Function in WH Question Acquisition: A Usage-based Case Study of German L1 Acquisition.” PhD diss., University of Groningen. Thornburg, Linda L., and Klaus-Uwe Panther. 1997. “Speech Act Metonymies.” In Discourse and Perspective in Cognitive Linguistics, edited by Wolf-Andreas Liebert, Gisela Redeker, and Linda R. Waugh, 205–19. Amsterdam: John Benjamins. Tomasello, Michael. 2003. Constructing a Language: A Usage-based Theory of Language Acquisition. Cambridge, MA: Harvard University Press. Weigand, Edda. 1989. Sprache als Dialog: Sprechakttaxonomie und Kommunikative Grammatik. Berlin: Mouton De Gruyter. Wray, Alison. 2002. Formulaic Language and the Lexicon. Cambridge, UK: Cambridge University Press.

This page intentionally left blank

4 ‹‹ Prepositional Phrases as Manner Adverbials in the Development of Hebrew L1 Text Production Gilad Brandes and Dorit Ravid

Tel Aviv University

The present investigation centers on the development of Hebrew written discourse syntax, focusing on prepositional phrases (PPs) as a specific syntactic construction. Specifically, we examined prepositional phrases that function as manner adverbials (e.g., Be-itiyut, “in-slowness = slowly”) across development. We analyzed 160 narrative and expository texts produced by eighty writers from pre-adolescence to adulthood. Analysis focused on the effects of age and genre on the prevalence of manner PPs, the complexity of their internal structures, and their different syntactic functions. Additionally, we examined the discourse-specific uses of manner PPs in each developmental stage and in each genre. Results showed that (1) manner PPs grow more prevalent and more internally complex with age and schooling; (2) most manner PPs modify verbs, but some function as noun modifiers, adjective modifiers, and predicates; and (3) many manner PPs do not constitute optional adjuncts, but rather function as obligatory, focal elements of clauses. The use of such focal manner PPs and of their optional counterparts is discussed in terms of choices that writers make to achieve certain rhetorical goals in discourse. The study contributes to our understanding of the development of Hebrew written discourse syntax, and sheds new light on the nature of manner PPs and their uses in discourse.

Introduction

Advanced syntactic abilities are often associated with clause combining, especially with different types of subordination, in both native language users (e.g., Nippold et al. 2005; Saddler and Graham 2005) and second language learners (Ortega 55

56

Brandes and Ravid

2003). However, a plethora of evidence points to clause-internal syntax as a domain of significant attainments, especially in later L1 developmental stages (Beers and Nagy 2009; Brandes 2015; Hunt 1965), and advanced L2 proficiency levels (Byrnes, Maxim, and Norris 2010; Norris and Ortega 2009). This is clearly evidenced in the well-documented age-related increase in mean clause length in L1 text production (Berman and Ravid 2009; Scott 1988), mostly determined by numerous and elaborated clause-internal phrases (Berman, Nayditz, and Ravid 2011; Ravid and Levie 2010). Consider the following excerpt (all examples are translated from the Hebrew study texts): In my opinion, the issue of theft must be handled in all schools with the utmost severity. This clause starts out with a PP functioning as a discourse marker (in my opinion), followed by the extended subject noun phrase (NP), whose head (the issue) is modified by a PP (of theft). The PP consists of a modal auxiliary (must), an infinitive copula (be), and a passive participle (handled), modified by two PPs—a place adverbial (in all schools) and a manner adverbial (with the utmost severity)—both of which contain extended complement NPs, whose head nouns (schools; severity) are modified by a quantifier and an adjective, respectively (all; utmost). Constructing such a hierarchical structure of elaborated phrases nested within each other, to concisely convey a detailed argument regarding abstract concepts, is an ability generally attained by mature, experienced writers (Berman 2008). The current study concerns syntactic development in Hebrew L1 written text production, focusing on a specific clause-internal construction—prepositional phrases functioning as manner adverbials (henceforth: manner PPs)—exemplified by with the utmost severity above. Manner PPs are examined across the school years as participating in increased clausal complexity of the kind demonstrated above, in the service of written expression. Complexity, a central construct in research on both native language development (Ravid 2005, 2013; Berman, forthcoming), and second language learning (Bulté and Housen 2012; Ortega 2012), is conceptualized in the current context not only in structural terms (Pallotti 2015), but as deriving from a form-function approach. Our construal reflects the view of syntax (shared by most contemporary linguistic frameworks) as having both structural and semantic/functional facets (Croft 2001; Givón 1993; Goldberg 2003; Langacker 1986). As syntactic structures convey meanings in different communicative contexts, syntactic complexity by nature should be expressed in both form and content, with structurally complex syntactic construction serving to deliver complex content within discourse (Berman and Nir-Sagiv 2009). Increased syntactic complexity is expressed structurally in the amount and variety of elements in a construction, as well as in the layeredness of their hierarchical organization (Givón 2009; Ravid and Berman 2010). In a clause, this is manifested (i) in numerous clause-internal phrases nested within each other, and (ii) in the complexity of the phrases themselves, which in turn is expressed in the number and variety of modifiers and the incorporation of phrase-internal coordination and subordination (Berman 2007; Scott 1988). Importantly, complex clausal structure goes hand in hand with semantic complexity, expressed in an elaborated description of different aspects of the situation that the clause delineates. For instance, heavy NPs give more

Prepositional Phrases as Manner Adverbials

57

detailed characterizations of participant entities or concepts (e.g., the utmost severity; a pretty good friend with similar interest to my own), and adverbials elaborate on surrounding circumstances like time, place, and manner (Fillmore 1994; e.g., in all schools). The development of NPs has been investigated extensively in several studies on text production in different languages, including Hebrew (Ravid and Berman 2010; Ravid and Cahana-Amitay 2005; Ravid and Zilberbuch 2003; Ravid et al. 2002), revealing a marked increase in NP complexity as a function of age and schooling. In contrast, adverbials have not yet been thoroughly investigated in developmental text production. Manner adverbial PPs, such as handle with the utmost severity; talk in a hurtful way, contribute to clausal complexity in both structure and content, uniquely referring to the manner aspect of situations. This developmental study offers new usage-based findings regarding the employment of manner PPs in written narrative and expository discourse by native Hebrew users. The introduction proceeds with a discussion of the role of PPs as manner adverbials in Hebrew, and concludes with presenting the motivations for studying manner PPs in discourse.

PPs as Manner Adverbials

Manner adverbs are regarded as verb modifiers that describe some manner in which the event designated by the verb occurs or is performed (Hengeveld 1992; Shaer 2003). Following Halliday (1985), manner is conceived here as a broad semantic space, ranging over quality (e.g., explain nicely), means (solve through dialogue), degree (change completely), and comparison (think differently). Modern Hebrew lacks a productive morphological class of manner adverbs (like English -ly adverbs), using instead syntactic PPs as the prevalent manner adverbial device (e.g., Be-adinut, “in-gentleness = gently”; Berman and Nir-Sagiv 2011; Glinert 1989).1 Thus, Hebrew manner adverbials typically take the form of periphrastic (often multi-lexemic) phrases, while semantically functioning as units of meaning, and can hence be viewed as intermediate entities, lying between syntax and the lexicon (Ravid and Shlesinger 2000). A typological factor that enhances the intermediate status of Hebrew manner adverbials is the fact that the four basic prepositions, including the primary manner-conveying be- (“in = with”), are written as single letters attached to the following word (Ravid 2012). Given Olson’s (1994) script-as-model, this phenomenon is certain to shape literate Hebrew users’ perception of PPs. Specifically, it encourages fusion processes, whereby originally syntactic PPs have become opaque lexical items (Nir and Berman 2010). Hebrew adverbial PPs are best viewed as delineating a continuum, from completely fused words (e.g., Le’at, “to-slowly = slowly”), through intermediate cases, which include most manner PPs (e.g., Be-rakut, “in-softness = softly”), to separately represented word pairs (e.g., Bli reshut “without permission”). The construction grammar approach (Goldberg 1995, 2003) provides a useful framework for the analysis of Hebrew manner PPs along a lexicon-syntax continuum (Langacker 1990). Conceptualizing manner PPs in terms of a construction—that is, a unique pairing of form (PP) and meaning (manner)—allows us to view them as unified wholes in discourse syntax. The Hebrew manner PP construction consists of two open slots, the first for a preposition, typically be- ‘in/with’, the second for

58

Brandes and Ravid

a complement, usually an abstract noun (e.g. Be-xavana, “in-intention = intentionally”). This flexible construction creates manner adverbials by inserting different NPs into the second slot, often modified in various ways to yield complex manner expressions (e.g., Be-otsma adira, “in-power great = very powerfully”).

Manner PPs in discourse

Our motivation for studying manner adverbials in the context of written discourse across adolescence is threefold, relating to the role of manner PPs in developing clause-internal complexity, their syntactic functions, and their uses in discourse. Clausal complexity

Investigating manner PPs contributes to our understanding of the growth of clausal complexity across development. Older, more-skilled writers were expected to use more manner PPs in clauses, for packing in additional, manner-oriented information. Moreover, manner PPs themselves were expected to become longer and more complex with age, expressing more sophisticated manner meanings, as in answer with as much detail and elaboration as possible. Syntactic functions

The category of manner adverbials is typically restricted to elements functioning as verb-modifiers (e.g., Berman and Nir-Sagiv 2011; Shaer 2003). In the current analysis, we defined manner PPs as pairings of form and meaning, with no additional syntactic constraints—thus expanding the traditional category boundaries, allowing for manner PPs in syntactic contexts other than modifying verbs. This was motivated by the fact that nouns and adjectives also both designate predicative content such as events, states, processes, and relations (Hengeveld 1992; Ravid and Cahana-Amitay 2005), and can thus also receive manner modification. Moreover, PPs that are themselves predicates can present a state of affairs with a built-in sense of manner. The following excerpts are examples of manner PPs functioning respectively (i) as a predicate, (ii) a noun modifier, and (iii) an adjective modifier: (i) I was in atomic stress; (ii) a solution through legislation might help this issue; (iii) anyone who is different in skin color. We expected to find increasing syntactic flexibility, with older writers using a wider range of syntactic alternatives for expressing manner, based on the insight that developing linguistic abilities involves recruiting new forms for designating familiar functions (Berman 2009; Berman and Slobin 1994). Uses in discourse

Examining manner PPs within written texts highlights their discourse-level roles (Goldberg 2003; Slobin 1977). The question of genre is of special significance in this context, since each genre constitutes a distinct social practice with different communicative purposes, giving rise to unique cognitive requirements and the deployment of appropriate linguistic forms (Blum-Kulka, Huck-Taglicht, and Avni 2004; Longacre 1983). Narratives, which revolve around people participating in events that unfold in a temporal context, seem to be the most natural environment for manner adverbials for qualifying events and states of mind. Rather than driving the storyline

Prepositional Phrases as Manner Adverbials

59

forward, manner adverbials are expected to be incorporated into eventive units to convey evaluative (descriptive or interpretative) information (Berman 1997; Labov 2010; Ravid and Berman 2006), thus going beyond recounting events to express the writer’s attitudes, signaling the mental and emotional states of protagonists (Kang 2003; McCabe, Peterson, and Connors 2006; e.g., Paul came to class with 5 jars of praying mantes which he showed to his classmates with pride). As expository discourse deals with general, abstract situations rather than actual events, we might expect fewer manner adverbials in expository writing. An exception might be to serve what Giora (1990) calls the ‘poetic’ function—that is, as elements used to attract the reader’s attention, making the text more aesthetically appealing, as in the following example: there exists racism in abundance. Thus, in accordance with their optional status in clausal syntax, manner PPs were also predicted to play a supplementary role in discourse as ‘digressive’ elements, straying away from the basic event sequence of a narrative or the line of argumentation running through an essay. However, when actual usage is considered, some surprising outcomes might emerge. We show below that a great portion of Hebrew manner PPs are neither clausally optional nor discursively digressive.

Method Participants and materials

The texts analyzed in this study were collected in the framework of a large-scale crosslinguistic project (Berman and Verhoeven 2002).2 The subset used here consists of 160 narrative and expository written Hebrew texts produced by eighty participants in four age/schooling groups: fourth graders (aged nine to ten), seventh graders (aged twelve to thirteen), eleventh graders (aged sixteen to seventeen), and graduate university students (aged twenty-five to thirty). Participants were normally developing native Hebrew speakers from middle to high socioeconomic backgrounds. Each participant wrote a personal experience story about problems between people and an expository essay discussing this topic (for further details, see Berman and Katzenberger 2004).

Analysis

All manner PPs were identified, counted, and coded using three criteria: (1) length in words; (2) internal structure; and (3) syntactic function. The data were manually coded by each of the two authors working independently. Cases of discrepancy were resolved by discussion. To calculate manner PP length, we counted words as strings appearing between two spaces. As prepositions may or may not appear as separate words (see above), they were excluded from this calculation, so that the length of a manner PP was determined solely by the length of its complement. Regarding internal structure, each manner PP was classified into one of seven structure types, ranked on a scale of complexity. The scale was constructed along similar lines as previous scales of different linguistic domains (e.g., Ravid 2006; Ravid and Berman 2010; Ravid and Levie 2010), largely based on research by the

60

Brandes and Ravid

second author and colleagues on NP complexity (Ravid and Berman 2010; Ravid and Zilberbuch 2003). In line with our multifaceted approach to complexity, each level of the scale is motivated by structural, semantic, and functional considerations: 1. Noun: a single-word NP (e.g., Matsáti otam tsoxakot be-shéket, “found-I them laughing in-quietness = I found them laughing quietly”) 2. Noun + modifier: an NP with a single modifier such as adjective or quantifier, extending the head-noun meaning (e.g., Ha-dvarim histaymu be-riv gadol, “the-things ended in-fight big = things ended in a big fight”) 3. Nominal compound: an NP consisting of two nouns in a construct construction (Ravid and Shlesinger 1995; e.g., Menagev et ha-yad shelo be-tnu’at^gó’al, “wipe acc the-hand his in-gesture^disgust = wipe his hand with a gesture of disgust”) 4. Noun + extended modifier: an NP containing an extended AP or PP (e.g., Hem kilelu exad et ha-sheni be-klalot me’od me’od gasot, “they cursed one acc the-other in-curses very very rude = they cursed each other with very-very rude curses”) 5. Coordinated head: two nouns or adjectives coordinated in a single head, giving a double description of manner (e.g., Hu hikpid lehatsig et atsmo ke-mutslax ve-tov miméni, “he insisted to-present acc himself as-successful and-good from-me = he insisted on presenting himself as better and more successful than I am”) 6. Pronominal/demonstrative reference: an inflected pronoun or demonstrative, referencing a constituent of a preceding clause, thus promoting discourse-level cohesion (e.g., hi patra be- o.k. u-va-ze nistayma sixaténu, “she dismissed in- o.k. and-in-this ended converstion-ours = she dismissed me with an o.k. and thus ended our conversation”) 7. Subordinate construction: an NP containing a subordinate clause, so that the situation designated by the main verb is described while presenting a second situation (e.g., Hu hikpid lehityaxes elay ke-el yéled katan she-tsarix letapel bo, “he insisted treating to-me as-to child little that-needs caring in-him = he insisted on treating me like a little child that needs taking care of”) In instances where two or more structure types were incorporated within a single manner PP (e.g., a nominal compound governing a relative clause), the one ranked as more complex was taken into account. Regarding syntactic function, each manner PP was classified as verb modifier, noun modifier, adjective modifier, or predicate. Finally, a detailed qualitative analysis was conducted, aimed at revealing the different discourse-specific uses of manner PPs in each age/schooling group and in each of the two genres. Details regarding the statistical methods employed, along with the main outcomes, are found in the appendix.

Results and Discussion Prevalence of manner PPs

Overall, manner PPs were not very prevalent in our corpus consisting of 160 texts and 2,191 clauses. We identified 161 manner PPs in total, which, on average, means

Prepositional Phrases as Manner Adverbials

61

‹‹Figure 4–1.  Mean number of manner PPs per clause by age one instance per text and roughly one in every 14 clauses. Figure 4–1 presents the mean number of manner PPs per clause by age. As predicted, the prevalence of manner PPs increased with age, with older writers using these constructions more frequently (see appendix).3 On the average, grade school and junior high students produced one manner PP per roughly 22 clauses. The use of manner PPs increased to about one in every 13 clauses in high school and more than doubled by adulthood, reaching an average of one manner PP in roughly every 10 clauses. These statistics point to the role played by manner PPs in generating the increase of mean clause length throughout development—using more manner PPs being one way for skilled writers to construct longer and more elaborate clauses. Additionally, the results show that older writers have a greater tendency to relate to aspects of manner regarding the situations that they write about, manifesting the kind of richness and diversity of content characteristic of mature expression. Regarding the effect of genre on the prevalence of manner PPs, our findings were somewhat counterintuitive. Recall that the eventive nature of narratives, compared with the abstractness of expositories, led us to predict that manner PPs would be more frequent in the former. However, no significant effect of genre was found (see appendix). That is, across development, writers were as likely to use manner PPs when writing an essay as when they wrote a story. The findings reported in this section call our perception of manner PPs into question, emphasizing the need for a discourse-based investigation into the ways in which they are employed in different genres.

Internal structure of manner PPs

Another aspect of the development of using manner PPs is the construction of longer and more complex expressions by older writers (see appendix). Figure 4–2 presents the mean length of manner PPs in each age group by genre. Figures 4–3 and 4–4 present the proportions of the different types of internal structures of manner PPs in each group in narratives and expositories respectively.

62

‹‹Figure 4–2.  Mean number of words per manner PP by age and genre

‹‹Figure 4–3.  Proportions of manner PPs’ internal structures by age in narratives

‹‹Figure 4–4.  Proportions of manner PPs’ internal structures by age in expositories

Brandes and Ravid

Prepositional Phrases as Manner Adverbials

63

The developmental trends were especially clear in the narratives. Manner PPs in grade school narratives were, without exception, of the shortest and simplest one-word kind. In junior high and high school, the average manner PP length nearly doubled, reaching two words (min length: 1; max: 6), and the proportion of one-word manner PPs dropped to roughly one-half and one-third of all manner PPs respectively. Three new structure types emerged in this period, representing levels 2–4 in the manner PP complexity scale: noun + modifier, nominal compound, and noun + extended modifier (see examples earlier). Thus, junior high and high school students in many cases produced longer and more complex manner PP expressions. However, the set of syntactic means that they recruited for doing so was still relatively limited, restricted to the lower and intermediate levels of the complexity scale. The process of manner PPs becoming more elaborate continued into adulthood. The manner PPs in the adult narratives had an average length of three words (min: 1; max: 13), exhibiting the widest variety of internal structure types, including those ranked most complex: coordinated head, pronominal/demonstrative reference, and subordinate construction. Adults not only referred to aspects of manner more frequently, but did so using more complex, informatively heavy expressions. The next excerpt, from an adult woman’s narrative, represents the peak of manner PP complexity in our corpus: Some drivers drive in a physically violent manner, which they wouldn’t have allowed themselves if they weren’t sitting behind a wheel in a closed car. The data from expositories revealed similar developmental trends. Here too, manner PPs grew longer and more complex with age. The average manner PP length increased from just under two words in grade school (min: 1; max: 4) to almost three words in adulthood (min: 1; max: 9), and the proportion of simple manner PPs diminished from roughly two-thirds to less than a quarter. A qualitative analysis of the data comparing the two genres revealed that writing expositories yielded greater complexity earlier on. Manner PPs with even a level 2 complexity rank did not emerge in narratives prior to junior high, but were already found in expositories written by grade schoolers. Similarly, manner PPs with coordinated heads, found only in adult narratives, already emerged in high school expositories. These are examples of how writers are pushed to recruit more advanced linguistic resources in order to meet the higher conceptual demands of producing expository discourse, achieving the literate register characteristic of this genre (Nippold et al. 2005; Ravid 2005).

Syntactic function of manner PPs

Recall the current construal of manner PPs differed from manner adverbs or manner adverbials as they were previously conceptualized (e.g., Berman and Nir-Sagiv 2011; Ravid and Shlesinger 2000) in the breadth of their possible syntactic functions. Our target constructions were all PPs conveying a sense of manner, regardless of whether they were verb modifiers or had some other syntactic function. Unsurprisingly, verb modification was by far the most common function of manner PPs, applying in roughly 80 percent of the cases. Yet it was not the only syntactic function of manner PPs. The remaining 20 percent were PPs judged as having manner semantics, syntactically functioning as predicates, noun modifiers, and adjective

64

Brandes and Ravid

‹‹Figure 4–5.  Proportions of manner PPs’ syntactic functions by age in narratives

‹‹Figure 4–6.  Proportions of manner PPs’ syntactic functions by age in expositories modifiers. Figures 4–5 and 4–6 present the proportion of each type of manner PP in each age group in narratives and expositories respectively. Predicative manner PPs are special instances of a PP functioning as the predicate of a verbless clause (Hengeveld 1992), for example, Ani hayíti dey ba-rosh: “I was quite in-the-head = I was pretty much at the lead”; anáxnu brógez: “we in-anger = we are not speaking.” The predicative manner PP relates to the subject entity, saying how it “is.” Typically, the impression is that the subject entity is not completely passive, though there is no verb to delineate its actions. The manner PP can be viewed as an alternative means for conveying some information regarding the manner this entity operates in. Overall, seventeen predicative manner PPs were identified, making up about 10 percent of all manner PPs. They were especially prevalent in grade school narratives, where they constituted over one-half of all instances of manner PPs. In absolute terms, however, this is due to only four instances, three of which were of the form brógez, a fused juvenile expression from be-rógez, “in anger” describing a situation where two children are not on speaking terms. Noun-modifying manner PPs qualify events that are coded in nouns. Only seven cases were identified, constituting 4 percent of the manner PPs. An examination of

Prepositional Phrases as Manner Adverbials

65

these manner PPs reveals that they make quite a homogenous group—they were almost all used for pointing to the means for solving a certain problem (e.g., Ani lo xoshev she-ptira al yedey onashim ta’azor, “I not think that-solving by punishments will-help = I don’t think that solving this with punishments would help”). Noun-modifying manner PPs sporadically occurred in expositories—one instance in junior high, one in high school, and four in adult expositories. The narratives contained only a single noun-modifying manner PP, produced by a seventh grader. Note that grade schoolers did not produce any noun-modifying manner PP. Adjective-modifying manner PPs can be viewed as delineating a manner in which the quality designated by the adjective is descriptive of an entity (e.g. Mishpat ze pashut le-havana, “sentence this simple to-understanding = this sentence is simple to understand”; be’ayot beyn anashim nefotsot kmo anashim, “problems between people common as people = problems between people are as common as people”). Seven adjective-modifying manner PPs were identified, constituting 4 percent of the total. Notably, they were found solely in adult expositories. Developmentally, different trends emerged in each of the two genres (see appendix). Expositories showed the predicted pattern of manner PPs becoming more syntactically diverse with age and schooling. Fourth-grade expositories contained only the most prototypical verb-modifying manner PPs, while predicates and noun modifiers were added in junior high and high school. Finally, adult writers used all three types, with the addition of adjective modifiers. This finding testifies to growing linguistic flexibility, allowing older writers to recruit new syntactic mechanisms for the expression of manner. The opposite process was evident in narratives. Here, the proportion of the less typical non-verb-modifying manner PPs diminished from more than half in grade school to roughly a quarter in junior high to virtual absence in the two oldest age groups. That is, when writing a story, adult writers used only verb-modifying manner PPs, illustrating their need to elaborate on the manner facets of the event structure. With regard to what seems to be a greater diversity of syntactic functions in the youngest age group, recall that the large proportion of predicative manner PPs in fourth-grade narratives was mainly due to a few instances of the fused expression brógez, “in anger.”

Uses of manner PPs in discourse

We now turn to discuss the different uses of verb-modifying manner PPs within narrative and expository discourse. Underlying this discussion of the uses of verb-modifying manner PPs is the unexpected discovery that many of them did not at all function as the optional adjuncts typical of the adverbial function. Rather, these manner PPs appeared to be obligatory, in the sense that omitting them would render the clause meaningless or ungrammatical, or, alternatively, cause a dramatic change in meaning (Huddleston and Pullum 2002). Consider the following examples: 1. Ani rotsa lixyot be-shalom im ha-sviva ve-im atsmi: “I want to-live in-peace with the-surroundings and-with myself = I want to live in peace with my surroundings and with myself.”

66

Brandes and Ravid

2. Hu mitnaheg elay kmo yalda ktana: “he behaves to-me as girl little = he treats me like a little girl.” 3. Efshar ligmor et ha-inyanim ha-éle be-draxim yafot: “possible to-end acc the-matters these in-ways pretty = these matters can be resolved nicely.” Fillmore (1994) discusses similar cases, where ‘circumstantial’ adverbial elements seem to be obligatory, suggesting that such adverbials are specified in the semantic frames of the verbs they are associated with. We do not believe that such an account captures the source of obligatoriness of the various manner PPs in our sample. For instance, the manner PP be-shalom (“in peace”) in (1) above does not seem to be licensed by the verb to-live, since I want to live does not raise any difficulty. Goldberg and Ackerman (2001) suggest a different account, relating the clause’s information structure. Taking such a perspective, any clause must convey some new information not already given in the discourse, or otherwise there would be no reason to produce it (Chafe 1984; Halliday 1985). The part of a clause that conveys this new information is sometimes referred to as focus (Ward and Birner 2001). Goldberg and Ackerman (2001) suggest that the obligatory status of some adverbials is a result of their being a part of the clause’s focus domain. Thus, omitting them would not leave enough new information to justify the existence of the clause in discourse, rendering it pragmatically improbable. Such an account seems to hold well for the various relevant cases in our sample. Consider (1), repeated in its English version as (6) below, and its variations in (4) and (5). 4. I want to live. 5. #I want to live with my surroundings and with myself. 6. I want to live in peace with my surroundings and with myself. As mentioned above, (4) seems completely probable. This is because this stripped-down version of the clause advances a contrastive reading, where the desirable state of being alive is contrasted with its obvious alternative of being dead. Saying that one prefers living over dying is adequately new information to justify producing this clause. With the addition of the two complements with my surroundings and with myself in (5), the verb phrase no longer delineates the mere state of being alive, but rather a state much more specific than that, which rules out a contrastive reading. However, these complements, which defuse the clause’s original focus, do not convey enough new information to replace it; everyone, whether they want to or not, is alive (at least at the time the discourse is produced), and necessarily live both with themselves and their surroundings. This leaves a clause that does not say anything pragmatically worth saying—salvaged, however, with the addition of the manner PP in peace in (6). What the clause now means is that given the obvious state of the writer being alive, she wants her necessarily occurring interactions with herself and her surroundings to be peaceful. The manner PP conveys the focal point that the writer is making, and hence constitutes the clause’s focus. Accordingly, this type of manner PP will be henceforth referred to as focal.

Prepositional Phrases as Manner Adverbials

67

Note that focal manner PPs were not at all marginal. In fact, they constituted roughly one-half of all verb-related manner PPs in our sample. We suggest that using them represents a choice made by writers to highlight the manner aspect of a situation, not as some additional elaboration but as its main content—a choice that has unique rhetorical effects within discourse. Particularly, the use of manner PPs often promotes generality and abstraction over concrete specifics, for example, from an expository: 7. Yesh anashim paxot mekubalim ve-eleyhem mityaxasim be-tsura shona me-asher le-anashim mekubalim: “there-are people less popular and-to-them treat in-form different from-that to-people popular = there are less popular people and they are treated differently than popular people.” This writer discussed the way less popular people are treated, but without giving any details. His goal is not to describe specific behaviors that they are exposed to, but rather to make a comparative claim: these are not the same behaviors popular people are exposed to. The use of a focal manner PP conspires with the choice of a generic noun and the timeless present tense in creating a general, abstract stance (Berman 2005). It seems that such an effect of using focal manner PPs is well-suited to the expository genre, which deals with general issues, relating to their broad social contexts rather than to specific events. In contrast, using focal manner PPs in narratives is a choice somewhat contradictory to the typically specific, involved nature of the genre: 8. Oto gábi hitnaheg klapay le-xol órex^ha-bikur be-min sug shel hitnas’ut she-haya xadash li: “this Gabi behaved toward-me to-all length^the-visit in-sort kind of condescension that-was new to-me = that Gabi treated me throughout my visit with a kind of condescension that was new to me.” The writer does not describe a certain event, rather making a generalization, derived out of a series of specific behaviors occurring over a period of time. Additionally, he does not say what this friend/rival actually did, instead focusing on the way in which he perceived his actions and the feelings that they stirred in him— conveyed chiefly by the focal manner PP. Taking such a generic, almost detached stance when recounting a personal and emotionally charged experience and focusing on internal mental states rather than external actions is characteristic of the narratives of literate adults (Ravid and Chen-Djemal 2015). The rest of the manner PPs—those that did indeed function as optional adjuncts—had different uses than their focal counterparts. In expositories, most optional manner PPs resembled intensifiers, which Giora (1990) regards as digressive “attention-grabbers,” serving a poetic function: 9. Tsarix letapel ba-inyan ha-ze be-mesirut: “need to-attend in-the-matter the-this in-devotion = this matter needs to be attended to with devotion.” 10. Kvutsot ha-nilxamot be-kana’ut be-kvutsot axerot: “groups that-fight in-zealousness in-groups others = groups that fight other groups zealously.”

68

Brandes and Ravid

11. Ha-texnológya shinta et xayéynu li-vli heker: “the-technology changed life-ours to-without recognition = technology has changed our lives completely.” Notice how each verb is tailored with its own intensifying expression: you fight zealously, but attend to something with devotion, and when something really changes a lot, it becomes so different that you cannot even recognize it anymore. A rich representation of the various intensifiers and of the appropriate contexts for using each of them reflects high lexical quality (Perfetti 2007), necessarily acquired gradually with the accumulation of discursive experience. The use of intensifying manner PPs is an opportunity for writers to impress their audience with their command of valued linguistic conventions, and indeed seems to serve a poetic function. In narrative discourse, some of the optional manner PPs functioned as true digressive elements: 12. Matsáti otam tsoxakot be-shéket: “found-I them laughing in-quietness = I found them laughing quietly.” 13. …higí’a pol la-kita ve-ito 5 tsintsanot im gmaley^shlomo she-her’a be-ga’ava le-kitato: “… came Paul to-the-class and-with-him 5 jars with camels^solomon that-showed in-pride to-class-his = …Paul came to class with 5 jars of praying mantes which he showed to his classmates with pride.” 14. Ishti ha-yekara… ha-muxana be-zot ha-réga letapes im tat mikla yorek esh el koma shlishit: “wife-mine the-dear… that-prepared in-this the-moment to-climb with sub machinegun spitting fire to floor third = my dear wife… who is prepared to instantly climb with a fire-spitting machinegun to the third floor.” These manner PPs seem to signal the social–emotional significance of the recounted events as subjectively perceived by the narrator, and thus constitute evaluative devices (Labov and Waletzky 1967). Such evaluative manner PPs add some flavor to what would otherwise be a dry, matter-of-fact account of events. Used skillfully, they can create an almost cinematic effect by promoting the visualization of events, which in turn leads to a greater effect on the reader. This is probably closest to the intuitive perception of manner adverbials and certainly what we had in mind when we first started this study. A discourse-based investigation revealed that it is just one of the ways in which manner PPs are actually used.

General Discussion and Conclusion

While the centrality of clausal complexity to writing development is widely acknowledged (Beers and Nagy 2009; Berman and Ravid 2009), few studies have traced the developmental courses of specific clausal elements, with the exception of the much-investigated NP (Ravid and Berman 2010). The current study regarded the development of clausal complexity in writing narrative and expository texts in Hebrew, illuminating the hitherto unexplored contribution of manner PPs. Adopting a view of the form and function of syntactic constructions as tightly connected facets of syntactic complexity, manner PPs have been shown as Hebrew-specific constructions which advance clausal complexity in both structure

Prepositional Phrases as Manner Adverbials

69

and content, conveying information regarding the manner aspects of events and situations. The findings presented here reveal a developmental growth in manner PP prevalence, which drives the age-related lengthening of clauses, reflecting a growing tendency to relate to the manner aspect of situations in narrative and expository discourse alike. Moreover, manner PPs themselves became longer and increasingly complex with age and schooling, similarly to the previously reported increase in NP complexity. This points to a growing ability to construct more elaborate phrase structures concurrent with more elaborate descriptions of manner. The study extended the definition of manner adverbials, traditionally restricted to verb modifiers (Ravid and Shlesinger 2000; Shaer 2003), to more syntactic environments. Acknowledging the role of elements other than verbs in designating situations and events (Hengeveld 1992), we presented examples of noun-modifying, adjective-modifying, and predicative manner PPs. These analyses suggest a possible rethinking of the traditional boundaries of the category of manner adverbials. Finally, investigating syntax within text, as opposed to relying on structured experiments (Costa et al. 2014; Ravid and Saban 2008), enabled us to go “beyond the sentence” (Berman 2009) to reveal the discourse-level uses of manner PPs— yielding some unexpected outcomes. Some manner PPs functioned in the predicted way, as digressive evaluative devices in narratives, and poetic attention-grabbers in expositories. A great portion, however, unexpectedly constituted obligatory, focal elements of clauses shown to promote the generality and abstraction of expository discourse and the detached stance characteristic of mature narration—highlighting the crucial importance of usage-based research.

Acknowledgement

This study was supported by a chief scientist (education ministry) grant no. 0607015371 to Dorit Ravid.

Notes

1. Broad phonemic transcriptions of Hebrew phrases are given in italics. They are meant to make the pronunciation of these phrases accessible for readers with no Hebrew background and generally do not represent Hebrew orthography. The voiceless velar fricative, which is absent from English phonology, is transcribed as x. The glottal stop and the voiceless pharyngeal fricative, which are in many cases omitted in casual speech, are marked with an apostrophe only in word-middle position, as in ga’ava (“pride”). Word stress is marked only when non-ultimate, and therefore words transcribed with no stress mark are stressed on their final syllable. Following the transcription, and given between double quotes, are the exact literal translations, meant to present the reader with the morpho-syntactic structure of the phrase, and the looser translations into English. Grammatical elements that constitute independent words in English, but in Hebrew are orthographically attached to the following word, are separated from the words following them with a hyphen, in both the transcription and the literal translation (e.g., Be-simxa, “in-gladness = gladly”). 2. Data collection was funded by a major research grant from the Spencer Foundation, Chicago, for the study of developing literacy across different text types, ages, and languages (1997–2000), Ruth Berman, Principal Investigator.

70

Brandes and Ravid

3. We are aware that age and schooling are two different constructs that do not always correlate. In this specific population, however, it was the case that the older a participant was, the more education (s)he had received. The separate effects of age and schooling cannot be teased apart in this study. Also keep in mind that due to this study’s cross-sectional design, the developmental trends are not directly observed over time, but inferred from comparing different participant groups.

References

Beers, Scott F., and William E. Nagy. 2009. “Syntactic Complexity as a Predictor of Adolescent Writing Quality: Which Measures? Which Genre?” Reading and Writing 22: 185–200. Berman, Ruth A. 1997. “Narrative theory and narrative development: The Labovian impact.” Journal of Narrative and Life History 7:235–44. ———. 2007. “Developing Linguistic Knowledge and Use across Adolescence.” In Blackwell Handbook of Language Development, edited by Erika Hoff and Marilyn Shatz, 347–67. Oxford, UK: Blackwell. ———. 2008. “The Psycholinguistics of Text Construction.” Journal of Child Language 35: 1–37. ———. 2009. “Beyond the Sentence: Language Development in Narrative Contexts.” In Handbook of Child Language, edited by Edith Bavin, 354–75. Cambridge, UK: Cambridge University Press. ———. Forthcoming. “Development of Complex Syntax: From Early Clause-combining to Text-embedded Syntactic Packaging.” In Handbook of Communication Disorders, edited by Amalia Bar-On and Dorit Ravid. Berlin: Mouton de Gruyter. Berman, Ruth A., and Irit Katzenberger. 2004. “Form and Function in Introducing Narrative and Expository Texts: A Developmental Perspective.” Discourse Processes 38: 57–94. Berman, Ruth A., Ronit Nayditz, and Dorit Ravid. 2011. “Linguistic Diagnostics of Written Texts in Two School Age Populations.” Written Language and Literacy 14: 161–87. Berman, Ruth A., and Bracha Nir-Sagiv. 2009. “Clause Packaging in Narratives: A Crosslinguistic Developmental Study.” In Crosslinguistic Approaches to the Psychology of Language: Research in the Tradition of Dan I. Slobin, edited by Jiansheng Guo, Elena Lieven, Nancy Budwig, Susan Ervin-Tripp, Keiko Nakamura, and Şeyda Özçalişkan, 149–62. New York: Taylor & Francis. ———. 2011. “Manner Adverbials in Modern Hebrew: Evidence from Text Analysis.” Helkat Lashon 43:178–200 [in Hebrew]. Berman, Ruth A., and Dorit Ravid. 2009. “Becoming a Literate Language User: Oral and Written Text Construction across Adolescence.” In The Cambridge Handbook of Literacy, edited by David R. Olson and Nancy Torrance, 92–111. Cambridge, UK: Cambridge University Press. Berman, Ruth A., and Dan I. Slobin. 1994. Relating Events in Narrative: A Crosslinguistic Developmental Study. Hillsdale, NJ: Lawrence-Erlbaum. Berman, Ruth A., and Ludo Verhoeven. 2002. “Developing Text Production Abilities in Speech and Writing.” Written Languages and Literacy 5: 1–22. Blum-Kulka, Shoshana, Deborah Huck-Taglicht, and Hanna Avni. 2004. “The Social and Discursive Spectrum of Peer Talk.” Discourse Studies 6: 307–28. Brandes, Gilad. 2015. “Prepositional Phrases as Manner Adverbs in the Development of Hebrew Written Discourse Syntax.” MA diss., Tel Aviv University [in Hebrew]. Bulté, Bram, and Alex Housen. 2012. “Defining and Operationalizing L2 Complexity.” In Dimensions of L2 Performance and Proficiency: Complexity, Accuracy and Fluency in SLA, edited by Alex Housen, Folkert Kuiken, and Ineke Vedder, 21–46. Amsterdam: John Benjamins. Byrnes, Heidi, Hiram H. Maxim, and John M. Norris. 2011. “Realizing Advanced Foreign Language Writing Development in Collegiate Education: Curricular Design, Pedagogy, Assessment.” Modern Language Journal, special issue, 94(5). Chafe, Wallace. 1984. “How People Use Adverbial Clauses.” In Proceedings of the Tenth Annual Meeting of the Berkeley Linguistics Society, edited by Claudia Brugman and Monica Macaulay, 437–49. Berkeley, CA: Berkeley Linguistics Society. Costa, João, Naama Friedmann, Carolina Silva, and Maya Yachini. 2014. “The Boy That the Chef Cooked: Acquisition of PP Relatives in European Portuguese and Hebrew.” Lingua 150: 386–409.

Prepositional Phrases as Manner Adverbials

71

Croft, William. 2001. Radical Construction Grammar: Syntactic Theory in Typological Perspective. Oxford, UK: Oxford University Press. Fillmore, Charles J. 1994. “Under the Circumstances (Place, Time, Manner, etc.).” In Proceedings of the Twentieth Annual Meeting of the Berkeley Linguistics Society: General Session Dedicated to the Contributions of Charles J. Fillmore, edited by Susanne Gahl, Andy Dolbey, and Christopher Johnson, 158–72. Berkeley, CA: Berkeley Linguistics Society. Giora, Rachel. 1990. “On the So-called Evaluative Material in Informative Text.” Text 4: 299–319. Givón, Talmy. 1993. English Grammar: A Function-based Introduction. Vol. 1. Amsterdam: John Benjamins. ———. 2009. The Genesis of Syntactic Complexity: Diachrony, Ontogeny, Neuro Cognition, Evolution. Amsterdam: John Benjamins. Glinert, Lewis. 1989. The Grammar of Modern Hebrew. Cambridge, UK: Cambridge University Press. Goldberg, Adele E. 1995. Constructions: A Construction Grammar Approach to Argument Structure. Chicago: Chicago University Press. ———. 2003. “Constructions: A New Theoretical Approach to Language.” Trends in Cognitive Sciences 7: 219–24. Goldberg, Adele E., and Farrell Ackerman. 2001. “The Pragmatics of Obligatory Adjuncts.” Language 77: 798–814. Halliday, Michael A. K. 1985. An Introduction to Functional Grammar. London: Edward Arnold. Hengeveld, Kees. 1992. Non-verbal Predication: Theory, Typology, Diachrony. Berlin: Mouton de Gruyter. Huddleston, Rodney, and Geoffrey K. Pullum. 2002. The Cambridge Grammar of the English Language. Cambridge, UK: Cambridge University Press. Hunt, Kellogg W. 1965. Grammatical Structures Written at Three Grade Levels (Research Rep. No. 3). Urbana, IL: National Council of Teachers of English. Kang, Jenifer U. 2003. “On the Ability to Tell Good Stories in Another Language: Analysis of Korean EFL Learners’ Oral ‘Frog Story’ Narratives.” Narrative Inquiry 13: 127–49. Labov, William. 2010. “Oral Narratives of Personal Experience.” In Cambridge Encyclopedia of the Language Sciences, edited by Patrick C. Hogan, 546–48. Cambridge, UK: Cambridge University Press. Labov, William, and Joshua Waletzky. 1967. “Narrative Analysis: Oral Versions of Personal Experience.” In Sociolinguistics: The Essential Readings, edited by Christina B. Paulston, and G. Richard Tucker, 74–104. Oxford, UK: Blackwell. Langacker, Ronald. 1986. “An Introduction to Cognitive Grammar.” Cognitive Science 10: 1–40. ———. 1990. Concept, Image, Symbol. Berlin: Mouton de Gruyter. Lehmann, Christian, and Yong-Min Shin. 2005. “The Functional Domain of Concomitance.” In Typological Studies in Participation, edited by Christian Lehmann, 9–110. Berlin: Akademie (Studia Typologica, 6). Longacre, Robert, E. 1983. The Grammar of Discourse. New York: Plenum. McCabe, Allyssa, Carole Peterson, and Dianne M. Connors. 2006. “Attachment Security and Narrative Elaboration.” International Journal of Behavioral Development 30: 398–409. Nippold, Marilyn A., Linda J. Hesketh, Jill K. Duthie, and Tracy C. Mansfield. 2005. “Conversational versus Expository Discourse: A Study of Syntactic Development in Children, Adolescents, and Adults.” Journal of Speech, Language, and Hearing Research 48: 1048–64. Nir, Bracha, and Ruth A. Berman. 2010. “Parts of Speech as Constructions: The Case of Hebrew ‘Adverbs’.” Constructions and Frames 2: 242–75 Norris, John M., and Lourdes Ortega. 2009. “Towards an Organic Approach to Investigating CAF in Instructed SLA: The Case of Complexity.” Applied Linguistics 30: 555–78. Olson, David. 1994. The World on Paper. Cambridge, UK: Cambridge University Press. Ortega, Lourdes. 2003. “Syntactic Complexity Measures and Their Relationship to L2 Proficiency: A Research Synthesis of College-level L2 Writing.” Applied Linguistics 24: 492–518. ———. 2012. “Interlanguage Complexity: A Construct in Search of Theoretical Renewal.” In Linguistic Complexity: Second Language Acquisition, Indigenization, Contact, edited by Benedikt Szmrecsanyi and Bernd Kortmann, 127–55. Berlin: Mouton de Gruyter. Pallotti, Gabriele. 2015. “A Simple View of Linguistic Complexity.” Second Language Research 31:117–34. Perfetti, Charles. 2007. “Reading Ability: Lexical Quality to Comprehension.” Scientific Studies of Reading 114: 357–83.

72

Brandes and Ravid

Ravid, Dorit. 2005. “Emergence of Linguistic Complexity in Written Expository Texts: Evidence from Later Language Acquisition.” In Perspectives on Language and Language Development, edited by Dorit Ravid and Hava Bat-zeev Shyldkrot, 337–55. Dordrecht, Netherlands.: Kluwer. ———. 2006. “Semantic Development in Textual Contexts During the School Years: Noun Scale Analyses.” Journal of Child Language 33: 791–821. ———. 2012. Spelling Morphology: The Psycholinguistics of Hebrew Spelling. New York: Springer. ———. 2013. “Syntactic Complexity in Discourse Production across Different Text Types.” In Across the Line of Speech and Writing Variation, edited by Catherine Bolly and Lliesbeth Degand, 51–66. Corpora and Language in Use – Proceedings 2. Louvain-la-Neuve, France: Presses universitaires de Louvain. Ravid, Dorit, and Ruth A. Berman. 2006. “Information Density in the Development of Spoken and Written Narratives in English and Hebrew.” Discourse Processes 41: 117–149. ———. 2010. “Developing Noun Phrase Complexity at School Age: A Text-embedded Cross-linguistic Analysis.” First Language 30: 3–26. Ravid, Dorit, and Dalia Cahana-Amitay. 2005. “Verbal and Nominal Expressions in Narrating Conflict Situations in Hebrew.” Journal of Pragmatics 37: 157–83. Ravid, Dorit, and Yehudit Chen-Djemal. 2015. “Spoken and Written Narration in Hebrew: A Case Study.” Written Language & Literacy 18: 56–81. Ravid, Dorit, and Ronit Levie. 2010. “Hebrew Adjectives in Later Language Text Production.” First Language 30: 27–55. Ravid, Dorit, and Ronit Saban. 2008. “Syntactic and Meta-syntactic Skills in the School Years: A Developmental Study in Hebrew.” In Language Education in Israel: Papers in Honor of Elite Olshtain, edited by Anat Stavans and Irit Kupferberg, 75–110. Jerusalem: Magnes Press. Ravid, Dorit, and Yitzhak Shlesinger. 1995. “Factors in the Selection of Compound-type in Spoken and Written Hebrew.” Language Sciences 17: 147–79. ———. 2000. “Modern Hebrew Adverbials: between Syntactic Class and Lexical Category.” In Between Grammar and Lexicon, edited by Ellen Contini-Morava and Yishai Tobin, 333–51. Amsterdam: John Benjamins. Ravid, Dorit, Janet van Hell, Elisa Rosado, and Anita Zamora. 2002. “Subject NP Patterning in the Development of Text Production: A Crosslinguistic Study of Dutch, English, Hebrew, and Spanish.” Written Language and Literacy 5: 69–94. Ravid, Dorit, and Shoshana Zilberbuch. 2003. “The Development of Complex Nominals in Expert and Non-expert Writing: A Comparative Study.” Pragmatics and Cognition 11: 267–96. Saddler, Bruce, and Steve Graham. 2005. “The Effects of Peer-assisted Sentence-combining Instruction on the Writing Performance of More and Less Skilled Young Writers.” Journal of Educational Psychology 97: 43–54. Scott, Cheryl M. 1988. “Spoken and Written Syntax.” In Later Language Development: Ages Nine through Nineteen, edited by Marilyn A. Nippold, 49–95. Austin: Pro-ed. Shaer, Benjamin. 2003. “‘Manner’ Adverbs and the Association Theory: Some Problems and Solutions.” In Modifying Adjuncts, edited by Ewald Lang, Claudia Maienborn, and Cathrine Fabricius-Hansen, 211–260. Berlin: Mouton de Gruyter. Slobin, Dan I. 1977. “Language Change in Childhood and in History.” In Language Learning and Thought, edited by John Macnamara, 185–214. New York: Academic Press. Ward, Gregory, and Betty J. Birner. 2001. “Discourse and Information Structure.” In The Handbook of Discourse Analysis, edited by Deborah Schiffrin, Deborah Tannen, and Heidi E. Hamilton, 119–37. Oxford, UK: Blackwell.

Appendix Statistical methods and main outcomes

For manner PP prevalence, a generalized estimating equation (GEE) regression analysis was employed, with age and genre as explanatory variables. Text length was

Prepositional Phrases as Manner Adverbials

73

controlled, using the number of clauses as covariate. A significant age effect emerged (χ²(3)= 9.44, p < 0.05). A post hoc least significant difference (LSD) test revealed that there was a higher prevalence of manner PPs in adult and high school texts, compared with junior high and grade school. For manner PP length, internal structure, and syntactic function, multilevel analyses were employed, which account for a possible connection between manner PPs originating from the same text. Due to computing limitations, caused by the relatively small number of manner PPs, the narratives and expositories were analyzed separately, using age as the only explanatory variable. Regarding manner PP length, an age effect emerged in both narratives (b = 0.28, p < 0.01) and expositories (b = 0.17, p < 0.01). The mean manner PP length was significantly higher in the adult texts compared with grade school (LSD test). For internal structure, analysis regarded the proportion of the simplest one-word manner PPs (level 1 of the complexity scale), compared with all other more complex structures (levels 2–7). An age effect emerged in narratives (b = –0.88, p < 0.01), with a smaller proportion of simple structures in adult and high school texts compared with junior high and grade school. In expositories, a near-significant trend emerged (b  =  –0.42, p  =  0.066), and the proportion of simple manner PPs was significantly smaller in adults compared with grade schoolers. For syntactic function, analysis regarded the proportion of verb-modifying manner PPs, compared with all other less-typical functions. An age effect emerged in narratives (b = 1.77, p < 0.001), but not in expositories (b = 0.40, p > 0.05). In narratives, the proportion of verb-modifying manner PPs was significantly higher in adult and high school texts compared with junior high and grade school. Further details are found in Brandes 2015 (in Hebrew).

This page intentionally left blank

5 ‹‹ Negative Constructions in Nonliterate Learners’ Spoken L2 Finnish TAINA TAMMELIN-LAINE and MAISA MARTIN

University of Jyväskylä

This chapter discusses the development of Finnish expressions of negation in four initially nonliterate women with very low oral skills during their first ten-month language course. While many studies have been published describing the learning of L2 Finnish by educated adult learners, hardly any research is available on how nonliterate adults learn Finnish. Yet research-based knowledge is needed for both pedagogical and resource-related decision making. The theoretical approach to additional language learning in this study is usage-based (e.g., Bybee 2008), with construction as the unit of analysis (Eskildsen 2012) and classroom as the interactional setting. In standard Finnish, the negative construction includes the negative auxiliary verb, inflected for person and number, and an uninflected lexical verb. In the data, verbs are present in only 23 percent of all the participants’ utterances. Unexpectedly, over half of these are , mostly formulated with just the negative auxiliary verb stem, with no personal endings and no lexical verb. A potential reason for this pattern is the low use of verbs in general and the inherent complexity of the Finnish negative construction. The simple negative verb stem ei is also sufficient for getting the message across. Individual differences exist in both the number and way of using the negative construction, indicating potential developmental paths. The development of the negative construction is compared with that of literate learners both in writing and in oral tests.

Introduction

Learning to say no is an important skill. One of the first words babies learn to produce is usually no, or in Finnish ei. While this single word expresses the message, a lot more remains to be learned before one can be considered a competent speaker of a 75

76

Tammelin-Laine and Martin

language. In this chapter we chart the early steps towards this goal by four initially nonliterate women learning Finnish. Nonliterate adults are a small but growing group in Finland where almost all healthy adults have been able to read for about 250 years. Recent immigration from countries with low literacy rates has presented the educational system with a new challenge: adults who may know several languages orally but cannot read or write any of them. Nor do they know Finnish, which means that the methods normally used to teach children to read are not feasible either, as literacy cannot be based on oral skills in Finnish. An abundant body of research also shows that adults and children are different as language learners (e.g., Lightbown and Spada 2013, 93; Saville-Troike 2012, 88; Singleton 2001). The language learning of this group is now starting to attract research efforts in many countries, as evidenced by the networking within LESLLA (Low-educated Second Language and Literacy Acquisition), an international forum of researchers, practitioners, and policy makers who share an interest in the development of L2 skills by adult immigrants with little or no schooling in their country of origin. Yet very little is known so far, and even less about learning Finnish, as research on Finnish as a second language has so far focused almost exclusively on school children and educated adults. Literacy influences the learning of an additional language in many ways. It is usually acquired in school, a process that provides learners with study skills and strategies for learning, which those lacking any educational background do not automatically share. Being able to read provides many opportunities and affordances for language learning, and writing allows one to take notes and review and repeat linguistic material. Oral input all too often disappears before it has been understood and committed to memory. Sound recordings alleviate this problem but are seldom as readily available as pen and paper, even if smartphones help in this. Knowledge of any alphabetic writing system also enhances the phonological skills necessary for learning a new language (see e.g., Tarone, Bigelow, and Hansen 2009). Most adults have great difficulty with phonological discrimination and memory when they do not have the support of a writing system. Even the shapes of letters can be hard to learn without the ingrained habit of paying attention to (as such) meaningless squiggles on paper or screen (Marrapodi 2013). In her study, Tammelin-Laine (2014a) suggests that in the early stages of learning a new language, nonliterate L2 learners do not benefit from oral linguistic input in the same manner and to the same extent as learners with functional literacy skills. Learning to read, particularly in an orthographically transparent language, such as Finnish, seems to improve phonological skills, which in turn helps with learning vocabulary. In Finland, municipalities are required to draft a three-year integration plan for all adult immigrants who are not employed and who apply for unemployment benefits or other financial support. The local immigration officials and the immigrants themselves are involved in writing the plan. It lists the activities that are expected to lead to integration. For nonliterate immigrants, the first step is normally a ten-month adult literacy course consisting of approximately 1,400 lesson hours. Some participants in these courses have already lived in Finland for several years,

Negative Constructions in Nonliterate Learners’ L2 Finnish

77

usually as housewives, with scant opportunities for language learning. This study was carried out during such courses in two different locations. LESLLA learners face a formidable and time-consuming task of having to learn a new language and literacy skills at the same time. Below we will discuss a small part of this process: the learning and use of the Finnish negative constructions by four women who attended a ten-month Finnish language and literacy course. This chapter focuses on the following research questions: 1. To what extent are negative constructions used in the spoken language of the participants in the L2 Finnish classroom context? 2. What are the negative constructions like? 3. How did negative constructions develop during the ten-month course? The process of learning to say no is first discussed on the basis of research done in other languages. The Finnish negative constructions are then introduced, together with a brief description of our construction-based approach. The main part of the study consists of a presentation of the four learners, the data and methods of the study, and the findings on the development of negation in their spoken language. These are then compared with those of educated learners in the Discussion section.

Negation in Learning a New Language

The development of negation is one of the more widely researched areas of second language acquisition. Since Klima and Bellugi (1966) described the developmental sequence of negation in L1 English, L2 development has been compared with their results. Milon (1974), for example, found that a seven-year-old L2 English learner followed the L1 sequence. Comparisons of adults and children have, however, produced very different findings (Dimroth 2010, 63), with adults showing more variance. Results, or their interpretation, also vary according to the theoretical framework and language studied. Particularly within Universal Grammar–based (UG-based) studies, the order of acquisition has been assumed to be non-languagespecific, making the search for the natural order of acquisition the main motivation of these studies. The focus has been in particular on the syntactic position of the negation word in relation to the lexical verb of the sentence. In general, it has been found that in the first stage of L2 acquisition, the negation word precedes the utterance (no + X) regardless of the L1 word order (see e.g., Klein 1984 or Wode 1981). An example of an extensive study involving a language other than English is that of Hyltenstam (1977) on Swedish. His findings support the natural sequence hypothesis of language acquisition (i.e., that it is independent of the manner of learning or the L1 background). The study on German of Clahsen, Meisel, and Pienemann (1982) was part of a vast ESF project in which the acquisition of many European languages by adult immigrants was explored. German and French were subsequently compared by Meisel (1997). The findings in most of these studies support a fairly uniform developmental sequence of negation. For a brief overview, see Becker 2005, 263–67.

78

Tammelin-Laine and Martin

Interpretation of the results has also changed over time. Meisel (1997, 227) concludes that adults, “rather than using structure-dependent operations constrained by Universal Grammar (UG), rely primarily on linear sequencing strategies which apply to surface strings.” A very different theoretical approach is offered by Eskildsen (2012), who studied the development of negative constructions in spoken language produced by two adult L2 English learners. While most previous research assumes L2 acquisition to be rule-based, Eskildsen, like us, sees language learning as construction-based (see chapter  3). His findings suggest that the participants’ learning of negation proceeds from recurring exemplar-based expressions toward an increasingly target-like pattern. The development process includes non-targetlike patterns which decrease over time. However, there are also differences between the participants and the way learning proceeds. Eskildsen’s construction-based approach offers a new way of interpreting the data but has so far been little applied to any specific language. All the languages alluded to above are Indo-European and share some similarities in their negative structures (e.g., the importance of word order). Much less research is available for other types of languages. In Finnish, the key issue in learning negation is inflectional, while word order is seldom problematic. The following chapter describes the Finnish negative construction.

Negation in Finnish: A Construction Approach

Finnish verbs are inflected for person, number, time, mode, and voice, represented by suffixes (and accompanying stem changes). Table 5–1 shows the full inflectional system in the indicative mode for the present tense. The Finnish present tense negative construction consists of the auxiliary verb ei, inflected for person, and the bare stem of the lexical verb. Other tenses, modes, and the passive voice involve other forms of the lexical verb but are not discussed here as they do not occur in our data. In Finnish, there is also the undeclinable particle ei (“no”), which is used as the opposite of kyllä (“yes”). This sentence-external ei looks and sounds exactly the same as the negative auxiliary in the third person singular. (For further details, see Karlsson 2008, 28, and for the nature of Finnish negation from a typological perspective, see Miestamo 2005.) Ei particles of this type do not occur in the present data, but this feature of Finnish grammar may lead emergent learners to assume that the Finnish negation word is always ei. The negative response to both affirmative and negative yes/no questions includes conjugated ei either with or without the bare stem of the lexical verb (e.g., Asutko Jyväskylässä? En (asu). “Do you live in Jyväskylä? No [I don’t]”). In negative yes/ no questions, the question suffix is added to the conjugated ei and the word order is reversed (e.g., Etkö asu enää Jyväskylässä? “Don’t you live in Jyväskylä anymore?”). Many verbs display stem changes in inflection (e.g., nuku/n, “I sleep”; hän nukku/u, “he sleeps”). The dictionary entry is nukku/a (including the infinitive marker a). For negation, the stem is the same for all persons: en nuku (“I don’t sleep”); hän ei nuku (“he doesn’t sleep”).

Negative Constructions in Nonliterate Learners’ L2 Finnish

79

‹‹ Table 5–1.  Inflection of the verb asua (to live) in the present tense indicative mode (personal ending underlined, question suffix italicized) Person

Affirmative

Negative

Interrogative

Negative interrogative

1. sg

asun

en asu

asunko

enkö asu

2. sg

asut

et asu

asutko

etkö asu

3. sg

hän asuu

hän ei asu

asuuko hän

eikö hän asu

1. pl

asumme

emme asu

asummeko

emmekö asu

2. pl

asutte

ette asu

asutteko

ettekö asu

3. pl

he asuvat

he eivät asu

asuvatko he

eivätkö he asu

The Finnish negative construction can be described by rules that can be derived from descriptions like those above. This is not, however, the way we assume it is learned, at least not by nonliterate learners like the ones in this study. Educated learners may refer to rules or tables to find the right form for writing, but even they are more likely to learn negative construction from examples, starting from very frequent ones such as en tiedä (“I don’t know”). The overall underlying notion of language learning in this study is usage-based and cognitively oriented: acquisition takes place in encounters with an increasing number of examples of the L2. Regularities are extracted from examples by general cognitive mechanisms. The objects of study are not rules or items but constructions, loosely defined here as units of language which contain a form and a meaning, both of which can vary within some limits (e.g., Goldberg 2003). Construction Grammar (see e.g., Fillmore and Kay 1996; Goldberg 1995, 2003) has been employed to describe several structures of Finnish (see e.g., Leino, J. 2003, 2008; Leino, P., et al. 2001; Visapää 2008). As a linguistic basis for the study of second language development, the construction approach has been used previously for Finnish in Seilonen (2013) and Kajander (2013). For English it has been applied by Eskildsen (2008). In this chapter, no Construction Grammar notation is applied, but the acquisition of negative constructions is described as a process of gradually increasing control.

Participants and Their Learning Context

The participants are four women attending their first Finnish L2 language and literacy course. At the beginning of the data collection period in August 2010, their oral skills in Finnish were rather low. Amina and Asra could talk a little about their everyday life, while Husna and Rana could only say a few phrases, such as nähdään huomenna (“see you tomorrow”). As can be seen in table 5–2, none of them reported having any earlier formal schooling or acquiring reading and writing skills in any language. However, all of them knew some of the Roman alphabet. Additionally, three of them had previous experience in oral L2 acquisition. The data were collected in towns A and B, where the language and literacy training classes were provided by adult education centers (AECs). In both AECs, the total number of lessons per week was thirty-five. It was divided into contact

Tammelin-Laine and Martin

80

‹‹ Table 5–2.  Summary of participant information Adult education center Name

Country of Age* origin

Town A

Asra

24

Afghanistan Dari

Farsi

18 months none

Town B

Amina 45

Afghanistan Dari

Russian

15 months none

Town B

Husna 45

Afghanistan Dari



16 months none

Town B

Rana

Iran

Farsi

12 months none

28

Native Other Length of Earlier language languages residence* education

Sorani Kurdish

Note. *In August 2010, at start of the data collection. Mean age 35.5 years, mean length of residence 15.25 months.

teaching of approximately five lessons (forty-five minutes each) and independent work lasting two hours per day. The class size was fifteen students. In town A, the instruction at the AEC focused on functional oral language skills, with a large variety of learning activities (e.g., learning-by-doing), while literacy skills were taught along with the vocabulary for everyday life. In AEC B, the instruction was mainly reading-oriented, and vocabulary was not an explicit focus. The negative construction was briefly mentioned in the reader Aasta se alkaa (“It begins with A”), which was used in the L2 training (Laine, Uimonen, and Lahti 2006). The lessons in both classrooms mainly consisted of either teacher talk or initiation-response-feedback (IRF) cycles led by the teacher (see e.g., Tainio 2007). The participants also used Finnish occasionally with their teachers, the researcher, and the other students. Native languages were spoken frequently during the lessons. The teacher in the reading-oriented AEC B, in particular, used ungrammatical utterances such as *Tämä ei hyvä (“This no good”). Normally a copula is used in Finnish, unlike in Dari and Sorani Kurdish. Most of the meaningful interactional situations of nonliterate L2 learners in Finnish are located in the classroom, and for this reason the language used during the lessons has an essential role in their learning process (see also Elmeroth 2003; Norton Peirce 1993). In general, the explicit teaching of verbs and their use was rare in both classes.

Data and Method

The data for this study were collected in the AEC classrooms over a period of ten months, from August 2010 to May 2011. Participant observation was supported by note taking and audio recording of the lessons. At AEC A, data were collected on six days during the autumn semester and four days during the spring. At AEC B, the respective numbers of days were eight and thirteen. The observation sessions were arranged as regularly as possible. During these sessions, all the Finnish utterances produced orally by the participants were documented. The course curriculum was drawn up by the teachers. All the students, including the study participants, regarded the researcher as an assistant teacher, and natural interaction in Finnish, both with

Negative Constructions in Nonliterate Learners’ L2 Finnish

81

the participants and other students, occurred frequently. (For more details of the data collection, see Tammelin-Laine 2014a, 53–55). In the data analysis, the number of words and utterances produced by each participant were counted and utterances with one or more verbs (including ei followed by no lexical verb) were encoded according to their intended meaning and actual form. Words or sentences read aloud or words repeated after the teacher were excluded. Utterances were divided into two main categories: declarative (e.g., Mies ei hyvä, “Man not good”) and interrogative (e.g., Ei kirjoita? “No write?”), which were further divided into the subcategories of affirmative (e.g., Yksi tunti kävelee, “One hour walks”; Tämä pois kirjoittaa? “This away write?”) and negative (e.g., Aamulla ei nukku, “In the morning no sleep”; Ei kirjoita? “No write?”) utterances. The data also contain some disjunctive interrogatives (e.g., Lukee…ei kirjoita, “Read…no write”; Kirjoitta suu ja ei? “Write mouth and no?”), consisting of both an affirmative and a negative verb. Because of their more complex nature, these utterances have been classified as a separate group. (For more on interrogative utterances expressed by the participants, see Tammelin-Laine 2014b.) When no lexical verb was present, the auxiliary verb ei was distinguished from the particle ei (see the section “Negation in Finnish: A Construction Approach”) by the presence of other material in the utterance (e.g, Tämä ei hyvä, “This not good”; Koti ei, “Home not”). Although this chapter does not focus on the contexts of the utterances (e.g., IRF cycle, interaction between students), some examples of these are presented in the qualitative analysis of the utterance samples produced by the participants. The data are described quantitatively in table 5–3. Table 5–3 shows that the participants used at least one verb in only 22.7 percent of all their utterances. However, there is a clear distinction in verb use between Asra and Husna, the participants who used the most (29.9 percent) and least (11.7 percent) verbs. However, while the number of utterances including a verb produced by Husna is low, she uses proportionally a wider range of verbs than the other participants. Another noteworthy matter is that Asra and Amina were the first to crack the alphabetic code and subsequently became the most fluent readers during the observation period. Rana also succeeded in learning how to blend sounds into words, but at the end of the data collection her reading was less fluent than that of Asra or Amina (Tammelin-Laine and Martin 2015). This finding inclines us to suggest that a positive relationship exists between the development of reading skills and the use of verbs by these participants (see Tammelin-Laine 2014a, 74, 85). ‹‹ Table 5–3.  The data of the study Amina

Asra

Husna

Rana

Total

Number of words

669

512

387

635

2203

Number of utterances

264

241

179

270

954

57

73

21

66

217

21.6

29.9

11.7

24.4

22.7

15

19

12

14

31

Total of utterances with verb(s) Percentage of utterances with verb(s) Number of different verbs used

Tammelin-Laine and Martin

82

Results Quantitative findings on negation

The data include 117 negative utterances. These utterances comprise 53.9 percent of all the utterances produced by the participants. The breakdown of the participants’ negative utterances into declaratives, interrogatives, and disjunctive interrogatives with or without a lexical verb is shown in table  5–4. The negative constructions presented here are instances of learner language and mostly not formulated as they would be in standard or colloquial spoken Finnish. Table 5–4 shows that in the data, ei occurs mostly in declarative utterances without a lexical verb: Approximately 77.8 percent of all the negative constructions are declaratives containing no lexical verb. In addition, the total percentage of interrogative and disjunctive interrogative utterances with no lexical verb is 7.7. However, Amina, Asra, and Rana occasionally use lexical verbs in declaratives. There are also some examples of the use of lexical verbs in interrogative utterances with a negative or disjunctive meaning in Amina’s and Asra’s data. There are also wide individual differences: In Amina’s data, approximately 30 percent of the negative utterances include a lexical verb, while in Rana’s data, the percentage is as low as 7.3, and in Husna’s data, even lower: zero. In general, the use of a lexical verb in negative constructions among the participants is low. Possible reasons for this pattern are the rather low use of verbs in general (see table 5–3) and seeing ei as a negation particle instead of a declinable verb. The negative verb stem ei is functionally adequate for conveying the idea of negation across. Most of the utterances including ei and a lexical verb are declaratives (10.3 percent). Individual differences in the number and type of negative constructions are quite substantial: Husna’s data include just four negative utterances, considerably less than the data of the other participants, especially Asra’s (forty-five negative ‹‹ Table 5–4.  Negative constructions in the data Amina

Asra

Husna

Rana

Total

Total of utterances with verb(s)

57

73

21

66

217

Neg. declaratives with lexical verb

6

3

0

3

12

Neg. declaratives with no lexical verb

15

38

4

34

91

Neg. interrogatives with lexical verb

0

2

0

0

2

Neg. interrogatives with no lexical verb

4

0

0

2

6

Disjunct. interrogatives with lexical verb

2

1

0

0

3

Disjunct. interrogatives with no lexical verb

0

1

0

2

3

Total of neg. constructions

27

45

4

41

117

Percentage of neg. constructions with lexical verb

29.6

13.3

0

7.3

14.5

Percentage of neg. constructions

47.4

61.6

19.0

62.1

53.9

Negative Constructions in Nonliterate Learners’ L2 Finnish

83

utterances). All the negative utterances produced by Husna date from the last three months of the data collection. They are all declaratives, without a lexical verb (e.g., Tämä ei tyttö tämä poika, “This no girl this boy”). This deviates from the other participants, whose data include negative constructions from October onward, occasionally with a lexical verb.

Qualitative findings: Development of the form and use of negative constructions

In the data, the participants use the Finnish negation auxiliary without inflection (i.e., only in the stem form ei). Throughout the data, it is mostly used in utterances without a lexical verb, followed or preceded only by nouns, adjectives, or adverbs. However, in some examples it is followed by a lexical verb in its default form (third person singular) or by the grammatically correct stem of a lexical verb. Below, the development of negative constructions by each participant is presented by providing examples of their first occurrence. Amina, Asra, and Rana use negative constructions both with and without a lexical verb, while no lexical verbs occur in Husna’s negative constructions. In Amina’s, Asra’s, and Rana’s data, the first negative constructions occur from October onward. The first type of negative construction in their data is ei + no lexical verb in a declarative utterance, as illustrated in examples (1) through (3): (1) Amina: Tämä ei This no+SG3

(target: Sitä en tiedä) “That one I don’t know”

(2) Asra:

Koti ei Home no+SG3

(target: Hän ei ole kotona) “He is not at home”

(3) Rana:

Ei hyvä mies No+SG3 good man

(target: Mies ei ole hyvä) “The man is not good”

The examples show that the first negative constructions are short and simple. Their meaning is difficult to understand without knowledge of the context. Example (1) expresses the idea that the participant does not know something while examples (2) and (3) refer to someone else (third person singular) not being somewhere or not being something. One potential reason for the missing lexical verb in Example (1) is that Amina has not yet memorized the construction en tiedä (I don’t know) even though it has probably occurred frequently in the classroom. In Examples (2) and (3) Asra and Rana do not use the copula. This is in line with the participants’ low use of the copula in affirmative utterances as well (see Tammelin-Laine, 2015). In Puro’s (2002) oral data, the copula was the most frequently used verb following the auxiliary verb ei. The low use of the copula in the present study may be transferred from the native languages of the participants, Dari and Sorani Kurdish, which do not use the copula the way it is used in Finnish or in the languages spoken by the participants in Puro’s study (see Thackston 2006; mylanguages.org 2011).1 In addition, the

Tammelin-Laine and Martin

84

teacher-talk used in both AECs (especially in AEC B) may have given learners the impression that the copula is not frequently used in Finnish either. Rana’s example (3) also shows the variation typical of L2 learners, as on the same day she uses a different word order (mies ei hyvä) to express the same idea. In January, the first disjunctive interrogative utterance occurs in Asra’s data when she produces a negative construction of the type ei + a lexical verb, shown in Example (4): (4) Asra:

Kirja ei kirja? (target: Kirjoitanko vai en?) Write+SG3 no+SG3 write+SG3? “Shall I write or not?”

It is interesting that the first instance of lexical verb use in Asra’s data was found in a complex disjunctive interrogative utterance instead of, for instance, declarative or basic question. Although her construction is not target-like, it clearly expresses the idea of opposition and a question; the interrogative is marked with a rising intonation towards the end of the utterance. However, the construction contains no inflectional elements referring to the speaking subject. In the data from March, Asra uses ei with a lexical verb in both an interrogative (Example 6) and a declarative (Example 7) utterance for the first time, while Amina and Husna use ei without a lexical verb—Amina in an interrogative (Example 5) utterance and Husna in a declarative (Example 8). This is also the first occurrence of a negative construction in Husna’s data. The illustrations are as follows: (5) Amina: Ei koira? No+SG3 dog?

(target: Eikö sinulla ole koiraa?) “Don’t you have a dog?”

(6) Asra:

Ei kirjoita? No+SG3 write+NEG?

(target: Enkö kirjoita tätä?) “Shall I not write this?”

(7) Asra:

Minä ei tiedä I no+SG3 know+NEG

(target: (Minä) en tiedä) “I don’t know”

(8) Husna: Käsi ei hyvä Hand no+SG3 good

(target: Käsi ei ole terve) “The hand is not well”

In examples (6) and (7), the typical pattern of Asra’s negative constructions can be seen. In the data, when she uses a new type of negative construction for the first time, she includes a lexical verb. Later, the presence of a lexical verb depends on the target verb: The nonuse of the copula is systematic and there are no examples of its use in the data. On the other hand, Asra occasionally uses verbs like nukkua (“to sleep”) and kirjoittaa (“to write”) in negative constructions. The data show only one exception to this pattern, when in April Asra again uses the construction Minä ei tiedä in exactly the same format as in March. It is likely that ei tiedä has been picked

Negative Constructions in Nonliterate Learners’ L2 Finnish

85

up from the teacher’s speech and memorized as a construction. However, Asra has added minä to it, which shows development in her language skills. Example (6) shows that she can also use other verbs in the same construction. The use of the personal pronoun minä varies in Asra’s examples, as it also does in spoken Finnish. For Amina, the negative interrogative presented earlier in Example (5) is the second type of negative construction occurring in her spoken language. She has clearly noticed that in Finnish it is possible to ask both affirmative and negative questions, even if her construction in Example (5) is non-target-like. However, when compared to her first negative construction expressed five months earlier, this example shows clear development in her language skills. In Husna’s data, three out of four negative constructions are of exactly the same type as Example (8) with ei in second position in the utterance and without the copula. It is only in her last example from May (Ei yksi kilo maito, “No one kilo milk”) that the negative auxiliary is in the initial position. The lexical verb, kantaa (“to carry”), is missing. Based on the data, Husna seems to start using negative constructions notably later than the other participants, and their number is particularly small. In April, Amina uses a lexical verb both in a declarative utterance and in a disjunctive interrogative, as shown in Examples (9) and (10): (9) Amina: Tämä ei kirjoittaa This no+SG3 write+SG3

(target: Tätä en kirjoita) “This one I don’t write”

(10) Amina: Lukee…ei kirjoita (target: Luenko vain, en kirjoita?) Read+SG3 no+SG3 write+NEG “Shall I just read this, not write?” Based on these examples above and those earlier in this chapter, it is evident that the negative constructions used by Amina have reached a more complex structure over time even if they are not yet completely target-like. Finally, in the data from May, Rana uses several new types of negative constructions: first ei + a lexical verb in a declarative utterance, then ei without a lexical verb in a disjunctive interrogative and in a basic negative interrogative. These types are shown in Examples (11), (12), and (13): (11) Rana:

Ei nukkuu No+SG3 sleep+SG3

(target: En nuku) “I don’t sleep”

(12) Rana:

Hyvä ja ei hyvä Good and no+SG3 good

(target: Onko tämä hyvä vai ei?) “Is this good or not?”

(13) Rana:

Ei kotona opettaja? No+SG3 at home teacher?

(target: Eikö mennä kotiin, opettaja?) “Don’t we go home, teacher?”

86

Tammelin-Laine and Martin

Rana’s data shows clear development in using negative constructions. In addition to the number of new negative construction types expressed by her in May, the number of negative utterances in total increases notably during the last month of observation. Moreover, all the negative constructions that include a lexical verb occur in May, and the complexity of negative expressions increases over time, even if at this point none of them are target-like.

Discussion

The data we have presented provide us with a window on the very beginning of L2 development. This is a level clearly below A1, which is the lowest level that has been included in the previous studies on the development of negation in learner Finnish. Furthermore, L2 acquisition is complicated by the lack of reading skills and explicit grammatical knowledge that enable most learners to benefit from descriptions of negative constructions, such as the one in the earlier section on Negation in Finnish.

A construction approach

The first negative constructions consist simply of ei. It either precedes or follows a noun phrase (tämä ei / ei tämä “this not / not this”). Both word orders are used with similar frequency by Amina (8/9), Asra (14/16) and Husna (3/1), but Rana fronts the particle ei most of the time (9/22). This is not in complete agreement with previous results on the acquisition of Indo-European languages, where the first phase has been found to be placement of the negative word first in the utterance (see the section on Negation in Finnish: A Construction Approach). However, a usage-based approach would predict this result, as in Finnish the negative word is normally in the second position while the first position is also possible, when the speaker wishes to emphasize the negation. The frequent absence of a lexical verb, obligatory in target-like negative constructions in Finnish, persists throughout the data collection period both in declaratives and questions. This is common in other languages as well (Becker 2005, 305). There are also examples of this in the research on educated learners. Puro (2002) studied the oral production of Finnish negative constructions during the first semester of a university Finnish course. Although Puro found some examples of negative constructions without a lexical verb in the participants’ spoken Finnish, in most cases it was included. Honkimäki and Kulta (2006, 151), who conducted an experimental study in which university students at different levels of L2 proficiency in Finnish produced negative answers to a large set of questions, found that the lexical verb was missing in 6 percent of instances at the lower elementary level and that 36 percent of the utterances included a lexical verb in a non-target form (5 percent and 21 percent, respectively, at the higher elementary level). In the large Cefling (2009) corpus of written data, independently assessed to represent the Common European Frame of Reference for Languages (CEFR) levels A1–C2, Halttunen (2014, 52) found that adolescents at the A1 CEFR level of Finnish included a lexical verb very systematically while adults at the same level sometimes omitted it.

Negative Constructions in Nonliterate Learners’ L2 Finnish

87

The present participants did not conjugate the auxiliary ei at all during this study. This also occurred in the oral data of Honkimäki and Kulta (2006, 153), although noncongruent instances only amounted to 3 percent at the lowest level but, surprisingly, were 16 percent at the next level, falling again at the higher levels. In Halttunen’s (2014) data, noncongruence is also fairly rare, even at the A1 level, particularly among the younger Cefling participants, who attend Finnish schools and mainly acquire Finnish in interaction with peers, unlike Puro’s (2002) and Honkimäki and Kulta’s (2006) university students and the participants of this study, who have few Finnish contacts outside the classroom. The negative construction with a lexical verb in the target-like stem form does not appear in this data, but then it is also the last one mastered by educated learners. This is probably due to the irregular relationship between the affirmative and negative forms, as illustrated in table 5–5. In verbs like syödä, the stem form and the 3. sg. affirmative and negative are identical but learners also encounter the past tense stem söi (“ate”), even if they do not yet use it themselves. For verbs of the asua type, the difference is the length of the unstressed vowel, which most learners find quite hard to detect. Consonant gradation, as in nukkua, is quite common in Finnish words and adds to the difference (and saliency) between the default form (3. sg) and the stem form. Occasional errors in lexical verb conjugation persist up to the highest levels of language acquisition, particularly with verbs like asua (Halttunen 2014). In Honkimäki and Kulta (2006, 151), only 85 percent of the test utterances were completely target-like even at the advanced level. The pattern found by both Puro (2002) and Halttunen (2014), where both the auxiliary and lexical verb are conjugated in a person other than 3. sg. (*en asun), was not found in the present data. In her study on the writing of Swedish-speaking school children and teenagers learning L2 Finnish, Grönholm (2007) suggests that the process of learning negative constructions begins with using the uninflected ei and a lexical verb in the third person singular (e.g., *minä ei menee; target [minä] en mene). The second phase includes uninflected ei with a lexical verb inflected in the affirmative form (e.g., *minä ei menen), while in the third phase ei is also inflected (e.g., *minä en menen), resulting in double inflection. The final product of the learning process is the target-like negative construction including the inflected auxiliary verb and the bare stem of the lexical verb (e.g., [minä] en mene). Similar overall phases were also found by Honkimäki and Kulta (2006).

‹‹ Table 5–5.  Some examples of the inflection of Finnish verbs Person

syödä ‘to eat’

asua ‘to live’

nukkua ‘to sleep’

1. sg affirm.

minä syön

minä asun

minä nukun

1. sg negat.

minä en syö_

minä en asu_

minä en nuku_

3. sg affirm.

hän syö

hän asuu

hän nukkuu

3. sg negat.

hän ei syö

hän ei asu_

hän ei nuku_

88

Tammelin-Laine and Martin

The negative constructions of Finnish produced by learners seem to develop through several stages marked by the inflection of the auxiliary and the lexical verb. Word order is of minor importance. A bare X + ei or ei + X is the starting point on the gradual road to more target-like constructions. The participants of this study, as well as those of the other studies discussed above, also show individual differences in the number of negative constructions they produce, the emergence of the first occurrence of ei within an utterance, the time of the first occurrence of the negative construction with a lexical verb, and the greater complexity of negative constructions at the end of the observation period. The most complex negative constructions (ei + a lexical verb in a question / disjunctive question) are used only by the most fluent readers (see Tammelin-Laine 2014b). The least fluent readers also use the least number of negative constructions and verbs in general. Thus the development of negative constructions seems closely related to other aspects of the growing language skills.

Conclusion

The first research question (see the Introduction section) concerned the frequency of negation in the very early stages of learning Finnish. Somewhat surprisingly, more than half of all the utterances were negative. The need to learn how to express negation is thus obvious. The second research question asked what the negative constructions were like and was answered by Examples (1) through (13) given in the Results section presenting the qualitative findings. The development (research question 3) was described in the Discussion section. The data presented here complement the studies previously conducted on the learning of a negation system based primarily on inflection, rather than word order, by describing the very first steps. It only presents constructions for the active voice and present tense as no other grammatical categories occurred in the data. A longer observation period is needed to obtain a more complete description of the paths of oral development for comparison with the pseudo-longitudinal study of written data (Halttunen 2014) or oral test data (Honkimäki and Kulta 2016). The construction approach functions here as a broad background statement. More work and a more extensive set of data are needed to determine how learner constructions evolve from the simple negating of a noun phrase toward target-like verb phrases, with in-between steps where learners try out the limits of the new construction in various ways. Comparisons with other inflection-rich languages could also shed more light on the path learners follow. Better knowledge of the path could also help remove stumbling blocks through improved instruction.

Notes

1. Puro’s participants were nine university students with five different native languages: English, French, Dutch, Russian, and German. Of these, only Russian lacks a copula.

Negative Constructions in Nonliterate Learners’ L2 Finnish

References

89

Becker, Angelika. 2005. “The Semantic Knowledge Base for the Acquisition of Negation and the Acquisition of Finiteness.” In The Structure of Learner Varieties, edited by Henriëtte Hendricks, 263–314. Berlin: Mouton de Gruyter. Bybee, Joan. 2008. “Usage-based Grammar and Second Language Acquisition.” In Handbook of Cognitive Linguistics and Second Language Acquisition, edited by Peter Robinson and Nick C. Ellis, 216–36. New York: Routledge. Cefling. 2009. “Cefling—Linguistic Basis of the Common European Framework for L2 English and L2 Finnish.” University of Jyväskylä. Accessed February 3, 2015. www.jyu.fi/cefling. Clahsen, Harald, Jürgen M. Meisel, and Manfred Pienemann. 1982: Deutsch als Zweitsprache: Der Spracherwerb Ausländischer Arbeiter. Tübingen, Ger.: Gunter Narr. Dimroth, Christine. 2010. “The Acquisition of Negation.” In The Expression of Negation, edited by Laurence R. Horn, 39–72. Berlin: Mouton de Gruyter. Elmeroth, Elisabet. 2003. “From Refugee Camp to Solitary Confinement: Illiterate Adults Learn Swedish as a Second Language.” Scandinavian Journal of Educational Research 47: 431–49. Eskildsen, Søren W. 2008. “Constructing a Second Language Inventory – The Accumulation of Linguistic Resources in L2 English.” PhD diss., University of Southern Denmark. ———. 2012. “L2 Negation Constructions at Work.” Language Learning 62: 335–72. Fillmore, Charles J., and Paul Kay. 1996. “Construction Grammar.” Lecture notes. Center for the Study of Language and Information, Stanford. Goldberg, Adele E. 1995. Constructions. A Construction Grammar Approach to Argument Structure. Cognitive Theory of Language and Culture. Chicago: The University of Chicago Press. ———. 2003. “Constructions: A New Theoretical Approach to Language.” Trends in Cognitive Science 7: 219–24. Grönholm, Maija. 2007. PT-teoria ja negaation oppiminen suomen kielessä toisena kielenä [Processability theory and the learning of negation in L2 Finnish]. A presentation at Kasvatustieteen päivät Nov. 22, 2007, in Vaasa, Fin. Accessed January 29, 2015. http://arkisto.kasvatus.net/kasvatustieteenpaivat2007/Gronholmsuo.pdf. Halttunen, Sari. 2014. “‘Sori mä en voin tullan’. Negaatioilmaukset suomi toisena kielenä -oppijoiden teksteissä” [‘Sori mä en voin tullan’. Negative utterances in L2 Finnish learners’ texts]. MA thesis, University of Jyväskylä. Honkimäki, Annukka, and Anne Kulta. 2006. “‘Sä ei tulit mun luo’. Suomi toisena ja vieraana kielenä -negaation oppimisesta” [“Sä ei tulit mun luo.” Learning negation in Finnish as a second or foreign language]. MA thesis, University of Jyväskylä. Hyltenstam, Kenneth. 1977. “Implicational Patterns in Interlanguage Syntax Variation.” Language Learning 27: 383–410. doi:10.1111/j.1467-1770.1977.tb00129.x. Kajander, Mikko. 2013. Suomen eksistentiaalilause toisen kielen oppimisen polulla [Finnish existential sentence in the path of L2 learning]. Jyväskylä Studies in Humanities 220. Jyväskylä, Fin.: The University of Jyväskylä. Karlsson, Fred. 2008. Finnish: An Essential Grammar. 2nd ed. London: Routledge. Klein, Wolfgang. 1984. Zweitspracherwerb. Königstein/Ts, Ger.: Athenäum. Klima, Edward, and Ursula Bellugi. 1966. “Syntactic Regularities in the Speech of Children.” In Psycholinguistic Papers, edited by John Lyons and Roger J, 183–208. Edinburgh: Edinburgh University Press. Laine, Rauno, Eila Uimonen, and Ylermi Lahti. 2006. Aasta se alkaa [It begins with A]. Helsinki: Opetushallitus. Leino, Jaakko. 2003. Antaa sen muuttua. Suomen kielen permissiivirakenne ja sen kehitys [Let it change. Finnish permissive construction and its development]. Helsinki: Suomalaisen Kirjallisuuden Seura. ———, ed. 2008. Constructional Reorganization, Vol. 5. Amsterdam: John Benjamins. Leino, Pentti, Ilona Herlin, Suvi Honkanen, Lari Kotilainen, Jaakko Leino, and Maija Vilkkumaa. 2001. Roolit ja rakenteet. Antaminen, allatiivi ja suomen datiivin arvoitus [Roles and structures. Giving, allative, and the mystery of Finnish dative]. Helsinki: Suomalaisen Kirjallisuuden Seura.

90

Tammelin-Laine and Martin

Lightbown, Patsy, and Nina Spada. 2013. How Languages Are Learned. 4th ed. Oxford, UK: Oxford University Press. Marrapodi, Jean. 2013. “What Doesn’t Work for the Lowest Level Literacy Learners and Why?” In Low-Educated Second Language and Literacy Acquisition. Proceedings of the 8th Symposium, edited by Taina Tammelin-Laine, Lea Nieminen, and Maisa Martin, 46–64. Jyväskylä, Fin.: The University of Jyväskylä. Meisel, Jürgen M. 1997. “The Acquisition of the Syntax of Negation in French and German: Contrasting First and Second Language Development.” Second Language Research 13: 227–63. Miestamo, Matti. 2005. Standard Negation: The Negation of Declarative Verbal Main Clauses in a Typological Perspective. Empirical Approaches to Language Typology. Berlin: Walter de Gruyter. Milon, John P. 1974. “The Development of Negation in English by a Second Language Learner.” TESOL Quarterly 8: 137–43. Mitchell, Rosamond, Florence Myles, and Emma Marsden. 2013. Second Language Learning Theories. London: Routledge. mylanguages.org. 2011. “Learn Dari.” Accessed February 3, 2015. www.mylanguages.org. Norton Peirce, Bonny. 1993. “Language Learning, Social Identity, and Immigrant Women.” Unpublished PhD diss., University of Toronto. Puro, Tarja. 2002. “Suomi toisena kielenä -aikuisoppijan verbien kehittyminen alkeiskurssilla” [The development of verbs by adult L2 Finnish learners during a basic language course]. Licentiate thesis, University of Jyväskylä. Saville-Troike, Muriel. 2012. Introducing Second Language Acquisition. 2nd ed. Cambridge, UK: Cambridge University Press. Seilonen, Marja. 2013. Epäsuora henkilöön viittaaminen oppijansuomessa [Indirect references in Finnish learner language]. Jyväskylä Studies in Humanities 215. Jyväskylä, Fin.: The University of Jyväskylä. Singleton, David. 2001. “Age and Second Language Acquisition.” Annual Review of Applied Linguistics 21: 77–89. Tainio, Liisa, ed. 2007. Vuorovaikutusta luokkahuoneessa. Näkökulmana keskustelunanalyysi [Interaction in the classroom. Discourse analysis as a perspective]. Helsinki: Gaudeamus. Tammelin-Laine, Taina. 2014a. Aletaan alusta. Luku- ja kirjoitustaidottomat aikuiset uutta kieltä oppimassa [Let’s start from the beginning. Non-literate adults learning a new language]. Jyväskylä Studies in Humanities 240. Jyväskylä, Fin.: The University of Jyväskylä. ———. 2014b. “‘Tämä hyvä?’ Kurkistus luku- ja kirjoitustaidottomien suomenoppijoiden esittämiin kysymyksiin” [“This good?” A glimpse into the questions asked by non-literate Finnish learners]. Puhe ja kieli 34: 81–99. Tammelin-Laine, Taina. 2015. “No Verbs, No Syntax: The Development and Use of Verbs in Non-Literate Learners’ Spoken Finnish.” In Low Educated Second Language and Literacy Acquisition. Proceedings of the Ninth Symposium, edited by Maricel G. Santos, Anne Whiteside, Hilaire Fong, and Stephanie Wells. Morrisville, NC: Lulu Publishing, 249–273.Tammelin-Laine, Taina, and Maisa Martin. 2015. “The Simultaneous Development of Receptive Skills in an Orthographically Transparent Second Language.” Writing Systems Research 7: 39–57. Tarone, Elaine, Martha Bigelow, and Kit Hansen. 2009. Literacy and Second Language Oracy. Oxford, UK: Oxford University Press. Thackston, Wheeler M. 2006. Sorani Kurdish: A Reference Grammar with Selected Readings. Cambridge, MA: Harvard University Press. Tomasello, Michael. 2003. Constructing a Language. A Usage-based Theory of Language Acquisition. Cambridge, MA: Harvard University Press. Visapää, Laura. 2008. Infinitiivi ja sen infiniittisyys. Tutkimus suomen kielen itsenäisistä A-infinitiivikonstruktioista [Infinitive and its infinity. A study on Finnish independent A-infinitive constructions]. Helsinki: Suomalaisen Kirjallisuuden Seura. Wode, Henning. 1981. Learning a Second Language: 1, An Integrated View of Language Acquisition. Tübinger Beiträge für Linguistik. Tübingen, Ger.: Gunter Narr.

6 ‹‹ How do Multilinguals Conceptualize Interactions among Languages Studied? Operationalizing Perceived Positive Language Interaction (PPLI) Amy S. Thompson

University of South Florida

Perceived positive language interaction (PPLI) refers to the perception held by many (but not all) multilinguals that languages studied in the past are interrelated in a positive way that can support and expand a multilingual’s ability to learn subsequent languages. This chapter explains the origins of the construct of PPLI, detailing its operationalization and summarizing key findings from three previous studies with a sample of multilingual students in Turkey. Qualitative PPLI data from the Turkish participants are explored in order to illuminate the specific types of positive language interactions perceived by these multilinguals. The chapter ends with thoughts for future research and issues with PPLI that need to be further examined.

Introduction

The concept of PPLI refers to the emic perception held by many (but not all) multilinguals that languages studied in the past are interrelated in a positive way that can support and expand a multilingual’s ability to learn subsequent languages. I formulated the concept of PPLI and tested it for the first time in northeastern Brazil in Thompson (2009). Since then, I have investigated PPLI empirically in a number of additional contexts, including Turkey, the results of which will be highlighted in this chapter. However, the inception of PPLI goes back to a much earlier time in my experience as a multilingual and a student of multiple languages. The first time that I can specifically remember thinking about how my languages were interacting in my mind was when I was at a romance language movie series watching Philippe 91

92

Amy S. Thompson

Muyl’s Le papillon (The Butterfly). The movie was in French, a language that I speak, and I was watching it with a Spanish-speaking friend with whom I communicated almost exclusively in Spanish. As I leaned over to whisper an undoubtedly witty comment about the movie to my friend, my utterance came out in French instead of Spanish. As my friend doesn’t speak French, he looked at me and whispered, “¿Qué?” I tried again, but despite my best effort, I couldn’t speak in Spanish while listening to French. “C’est rien” I finally said to him, as I sunk back into my seat, utterly perplexed at what had just happened. This seemingly insignificant event triggered my scholarly inquiry into the conceptualization of multilingualism and PPLI from a learner emic perspective. What does it mean to be multilingual? How are various languages distinctly represented in the minds of different multilingual individuals? How might these various languages interact and help or hinder subsequent language learning? In attempting to answer these questions, my position is that perhaps the emphasis on multilingual inquiry should turn to learner perceptions, and that researchers of multilingualism should not neglect the learners themselves and how they see the language systems within themselves and the cross-language interactions therein. The concepts of language transfer, and later crosslinguistic influence, interested me, and my curiosity was particularly piqued when I re-read Kellerman’s studies from the late 1970s and 1980s, immediately followed by DeAngelis’ book Third or Additional Language Acquisition when it came out in 2007. These sources and thoughts, along with my own experiences of studying multiple languages, helped me formulate the concept of PPLI. In this chapter, I first present the origins of PPLI, detail how I operationalized the concept, and summarize key findings from three previous studies that I conducted with a sample of multilingual students in Turkey. In the second part of the chapter, I report on the qualitative PPLI data from the three aforementioned Turkish studies in a more detailed manner, taking a closer look at the specific languages involved in the participants’ perceived positive interactions across their foreign languages, as well as the specific types of interactions perceived by these multilinguals. The chapter ends with thoughts for future research and issues with PPLI that need to be further examined.

Perceived Positive Language Interaction (PPLI): The Origins

Studies that make the participants’ language backgrounds the focal point of inquiry make up a relatively small portion of SLA research, although researchers such as Cook (2009) and Ortega (2014) have argued for a closer examination of the linguistic backgrounds of research participants. Indeed, the benefits of bilingualism/ multilingualism have been investigated by several researchers to date. For example, individuals with two or more languages in their linguistic repertoire have been shown to have higher metalinguistic awareness (e.g., Bialystok 2006), more efficient learning strategies (e.g., Nayak et al. 1990; Wilson and Sperber 2006), lower language learning anxiety (e.g., Dewaele, Petrides, and Furnham 2008), and a higher tolerance for ambiguity (e.g., Dewaele and Wei 2013; Thompson and Lee 2013).

How do Multilinguals Conceptualize Language Interactions?

93

Additionally, researchers such as DeAngelis (2005a, 2005b, and 2007) have suggested that a relatively low level of proficiency is needed for the previous language learning experience to have an effect on the subsequent language learning process. A benefit that is specific to multilinguals pertains to the possibility that learning one foreign or second language makes subsequent language learning easier and/ or more successful. The relationship between multilingualism and subsequent language acquisition, for example, was explored by Gibson and Hufeisen (2003) with multilingual participants in Germany. They investigated whether knowing multiple languages would facilitate the task of translating a text in an unknown language (Swedish); a positive relationship between number of languages known and success at the Swedish text translation was found. Similarly, Kemp (2001) found that learning Basque was easier for those participants who knew more languages. These studies are related to the concept of interlingual identification (Odlin 2008), or the action of using a previously known language for subsequent language acquisition. Although some scholars have used this term exclusively to refer to negative transfer (e.g., Weinreich 1953), Odlin emphasized that the idea can be used for positive transfer as well, stating that many learners “recognize at least the possibility of making an interlingual identification, whether or not they actually choose to do so” (444). One factor thought to influence whether interlingual identification happens or not in a particular instance is the notion of perceived language distance, as first proposed by Kellerman (1979): “transfer… goes hand in hand with the learner’s perception of NL-TL ‘distance,’ or to put it another way, the typological relationship between the two languages” (38–39, underlining in original). The closer the learners perceive the languages in question to be, the more likely they are to engage in transfer. Kellerman explored this concept with a study involving L1 Dutch–L2 English participants, finding that “the less representative of the prototypical meaning a usage of a given form is, the lower its transferability” (65). In other words, Kellerman found that with his participants, the notion of centrality is crucial to perceived language distance. Kellerman’s (1995) Transfer to Nowhere Principle (a reaction to Anderson’s 1983 Transfer to Somewhere Principle) illustrates the idea that learners’ perceptions have an effect on what gets transferred from one language to the other, specifically, “there can be transfer which is not licensed by similarity to the L2, and where the way the L2 works may very largely go unheeded” (137). Although Kellerman’s work focused on L1 to L2 transfer, there has been some work which focuses on the differences between transfer when learning an L2 versus an L3. A notable example is Dewaele (1998), who studied L1 Dutch learners, some of whom have French as an L2 and some of whom have French as an L3. He found that, oftentimes, L2 knowledge is transferred into the L3, as opposed to the language learner relying on the L1 when learning the L3: “It appears that the French L2 speakers have a higher level of activation for Dutch (L1), whereas the French L3 speakers have a higher level of activation for IL English (L2). … It appears that the L1 is not necessarily always the dominant active language and that access to its lemmas could accordingly be limited” (488). In addition to emphasizing the importance of learner perceptions, PPLI also espouses the idea that multilingualism is dynamic, as illustrated in the multilingual

94

Amy S. Thompson

competence described in the Dynamic Model of Multilingualism (DMM; Herdina and Jessner 2002; Jessner 2006, 2008). In this model, language systems are seen as interdependent: “Interdependent language systems [form] part of an overall multicomponential psycholinguistic system” (Herdina and Jessner 2002, 86). Jessner (2006, 33) illustrates interactions of a multilingual’s language systems in the following formula: LS1 + LS2 + LS3 + LSn + CLIN + M = MP (LS = language system; CLIN = crosslinguistic interaction; M = multilingualism factor; MP = multilingual proficiency), and Jessner (2008, 276) created a figure of this model that illustrates the “butterfly effect” of multilingual competence. The DMM illustrates the interconnected nature of the language systems, indicating that a change in one system will affect a change in another system. Linking the DMM back to the concept of perceived language distance and PPLI, as the perception of the relationships between languages shift over time, so does the multilingual proficiency of a specific language learner. In sum, the notion of PPLI rests on two premises that are well known in second language acquisition. One is that learner judgments are important to the language learning process and “are by definition subjective” (Odlin 2008, 443) and that these learner perceptions have an effect on many aspects of subsequent language learning experiences. The other is that multilingualism is dynamic in nature (Herdina and Jessner 2002). Both premises are central in the gestation and operationalization of PPLI. A potential benefit of multilingualism is that learning one foreign or second language might make subsequent language learning easier and/or more successful. To further explore this potential benefit, we must continue to investigate whether and under what conditions multilinguals hold an emic belief in this benefit; one of the ways to investigate this possible advantage of multilingualism is via the notion of PPLI.

Operationalization of PPLI

In previous work on PPLI (e.g., Thompson 2009, 2013; Thompson and Aslan 2014; Thompson and Erdil-Moody 2014; Thompson and Khawaja 2015), the tool for exploring the PPLI framework for multilingualism has been learner responses to an open-ended question on a background questionnaire with a subsequent content analysis. Specifically, the question posed was: “If you have studied other languages in the past, do you think that this has helped or hindered your ability to learn subsequent languages? Please provide specific examples where appropriate.” Upon performing a content analysis of the answers, participants could be classified in one of two groups, PPLI or NPPLI, representing three underlying cases (PPLI multilinguals, NPPLI multilinguals, and NPPLI bilinguals). Figure 6–1 provides a pictorial representation of the PPLI construct, based on the operational coding classification that will be detailed here. For all studies, the English translations of the PPLI coding were either double-rated for the Brazilian data (Thompson 2009, 2013) or triple-rated for the Turkish data (Thompson and Aslan 2014; Thompson and Erdil-Moody 2014; Thompson and Khawaja 2015) and 100 percent agreement was reached. The examples included in this chapter come from the Turkish data.

How do Multilinguals Conceptualize Language Interactions?

95

‹‹Figure 6–1.  Visual representation of the operationalization of PPLI. Participants were placed into the PPLI group if they reported that they perceived positive language interaction between any foreign languages with which they had experience. Responses that qualified for a PPLI classification can be language- and skill-specific, or they can be quite general. Example 6–1 is illustrative of a non-skill-specific response that would qualify for a PPLI classification: ‹‹Example 6–1 Kesinlikle olumlu etkisi oldu. İngilizce biliyor olmam sayesinde Fransızca öğrenirken daha bilinçli ve daha hazırdım. It absolutely helped me positively. Thanks to knowing English I was more conscious and ready to learn French (Participant 71). Answers that indicate cross-language interaction within a specific skill also qualified for a PPLI classification, as in example 6–2: ‹‹Example 6–2 Evet (yes). Vocabulary transfer. Grammar transfer (Participant 33, comment given directly in English). Participants who perceived no positive cross-language interaction between any foreign languages in their repertoire were classified as No Positive Perceived Language Interaction (NPPLI). For the NPPLI classification, sometimes, the responses were simple, such as “Olmadı” (“No, there wasn’t,” Participant 56), indicating that no interactions between foreign languages were perceived. Oftentimes, however, there were more details of how the interactions across languages were actually hindering learning the subsequent foreign language, such as in example 6–3: ‹‹Example 6–3 Bazen İngilizce’deki gramer kurallarını ya da kelimelerin anlamlarını düşünüp, Almanca’da hata yapıyordum.

Amy S. Thompson

96

Sometimes, I would think about English grammar rules or word meanings and make mistakes in German (Participant 58). Those with experience in only one language after the L1 were automatically classified as NPPLI learners, since by definition the concept of PPLI requires that the language learners be multilingual—that is, that they have studied more than one language beyond the L1. It is of course possible to receive answers from bilingual participants that are quite insightful, such as in example 6–4: ‹‹Example 6–4 Anadilim (Türkçe) ve İngilizce’den başka bir dil bilmiyorum…İngilizce seviyem yükseldikçe ve Türkçe ile aralarındaki benzerlik ve farklılıkları gördükçe anadilimden herhangi bir bilgiyi İngilizce’de kullanmaya daha açığım. Örneğin Türkçe’deki bazı söz kalıplarının İngilizce’de de mevcut olduğunu fark ettiğimden beri olası kalıpları kontrol ederek kullanıyorum. Ayrıca Türkçe dilbilgisi bilgimin İngilizce üzerinde olumlu etkileri olduğunu düşünüyorum. Mesela sıfatın ya da zarfın ne olduğunu, ne zaman, nerede kullanıldığını Türkçe’de bildiğim için sözlükten baktığım bir kelime sıfat, zarf ise onu da cümle içinde doğru yerde kullanabiliyorum… My native language is Turkish. I don’t speak any other language than Turkish… as my English improved and I began to see the differences and similarities between Turkish and English, I was more open to use my native language knowledge. For example, ever since I noticed that some structures in Turkish exist in English as well, I try to be more careful using them. In addition, I believe that my Turkish knowledge has positive influence on my English. For instance, since I know what adjectives and adverbs are and when and where they are used in Turkish, I can use an adjective or adverb correctly in the sentence once I look it up in a dictionary… (Participant 1). However, a perceived positive interaction between the L1 and a foreign language, while important, does not meet the requirements for a PPLI classification; the way that this concept is operationalized is that the positive interaction needs to be between two second or foreign languages. This particular participant in example 6–4 was therefore classified as NPPLI, because there was only one foreign language in question. Two qualifications are in order. First, a language learner could perceive both positive and negative language interactions between foreign languages and still be classified as a PPLI learner. The crucial aspect is that at least some sort of positive interaction is perceived, and that the perceived interaction is primarily positive, such as in example 6–5: ‹‹Example 6–5 Tabii ki etkisi oldu. İngilizce sayesinde diğer dilleri daha kolay öğrendim çünkü dilbilgisi kuralları hemen hemen aynıydı. Ancak başta özellikle Almanca konusunda sorun yaşadım. Örneğin; «und» yazacağıma «and» dedim defalarca kez. Bu gibi ufak detaylar dışında olumlu etkisi oldu. Of course there was an interaction. Thanks to English, I learned other languages more easily because the grammar rules were almost the same. However,

How do Multilinguals Conceptualize Language Interactions?

97

I had trouble at first with German. For instance, I would write ‘and’ instead of ‘und.’ Except for these it affected positively (Participant 59). If a participant were to list both perceived positive and perceived negative interactions, concluding that learning multiple languages has “mostly been problematic” (or a similar idea), this participant would be classified as NPPLI. Similarly, if a participant were to list both perceived positive and perceived negative interactions, concluding that learning multiple languages has “mostly been beneficial” (or a similar idea), this participant would be classified as PPLI. As long as some sort of positive interaction is stated, along with some sort of indication that the perceived positive interactions outweighed the perceived negative interactions, the participant would be classified as PPLI. When coding for the PPLI classification, the researcher should keep the following notions in mind: 1. Verify that the participants in question are multilingual (i.e., that they have studied at least two languages beyond the L1). The information should be independently collected in a separate part of the background questionnaire. Bilingual participants (those with only one language beyond the L1) cannot be classified as PPLI. 2. Read the response and identify if a perceived positive language interaction has been stated. 3. Eliminate responses that indicate positive interactions involving the L1 and an L2 (these are interesting, but are outside the scope of the PPLI framework). 4. Remember that responses such as “I’m not sure” or “Neither positive nor negative” do not qualify for PPLI. 5. If a participant states both positive and negative interactions, the PPLI coding can be used if the participant’s overall feeling is that the interactions are positive (see example 5 above). A NPPLI coding will be used if the participant’s overall feeling is that the interactions were negative. 6. PPLI coding is by default subjective. It is always a good idea to have a second or third rater coding the answers for interrater reliability. Second, as a corollary of multilingualism itself being dynamic (Herdina and Jessner 2002), the classification of PPLI learners is dynamic, and learners at any time can be reclassified from NPPLI to PPLI be it because of a change in a learner’s perception of positive interactions between foreign languages, or because a learner began to learn a new language, thus moving from bilingual to multilingual and having more than one language beyond the L1 in their language learning repertoire. This is highly likely, for example, in the participant who produced the extensive comment in example 6–4 above; with such detailed introspection about the relationship between the L1 (Turkish) and English, it can be predicted that if this person were to learn an additional language, a positive interaction would be seen between the two languages. In sum, the construct of PPLI can be tapped with the open-ended question “If you have studied other languages in the past, do you think that this has helped or hindered your ability to learn subsequent languages? Please provide specific examples

98

Amy S. Thompson

where appropriate.” Additionally, as indicated in figure 6–1, bilinguals (defined as those with one second or foreign language learning experience) are always placed into the NPPLI group, along with the multilinguals who perceive no interaction or negative interaction between their languages beyond the L1. The PPLI group is composed of those multilinguals (those learners who have experience with more than one second or foreign language) who see positive interaction between their additional languages.

Key Findings about PPLI in the Turkish Context

The first empirical study to be published using the PPLI framework was a PPLI and language aptitude study in the context of multilingual students in northeastern Brazil (Thompson 2013). The results indicated that those learners who perceived positive interactions between foreign languages studied had higher language aptitude, as measured by the CANAL-FT (Grigorenko, Sternberg, and Ehrman 2000). In this section, I will present and discuss key PPLI results of three subsequent studies carried out in the context of Turkey on the topics of language learning motivation (Thompson and Erdil-Moody 2014), anxiety (Thompson and Khawaja 2015), and beliefs (Thompson and Aslan 2014). All three studies explored the same sample of volunteers, who answered a series of questionnaires via SurveyMonkey (www.surveymonkey.com). Although the survey contained more items, the results of the responses to the following parts of the survey will be discussed in this chapter: twenty motivation items about the ideal L2 self and ought-to L2 self from Dörnyei and Taguchi (2010), thirty-three items from the Foreign Language Classroom Anxiety Scale (FLCAS; Horwitz, Horwitz, and Cope 1986) for anxiety data, thirty-six items from a modified Beliefs about Language Learning Inventory (BALLI; Horwitz 1988) for language learning beliefs data, and an extensive background questionnaire. All of the items on the survey were presented in both Turkish and English so as to include language learners from a wide variety of English language proficiency skills, and the data were analyzed using SPSS version 22. Of the original volunteers, 166 completed at least the background and motivation questionnaires, although 7 participants were eliminated because of an L1 other than Turkish, making the total number of participants 159 adult Turkish EFL learners. Most of these participants (72.3 percent) were between eighteen and twenty-two years old, and there were 49 male and 110 female participants. The background questionnaire was presented first, followed by the motivation questionnaire, the FLCAS, and then the BALLI. As is expected in survey-based research, some participants did not finish the entire questionnaire; thus, the anxiety analysis had 156 participants and the beliefs analysis had 153 participants. Each individual questionnaire underwent an exploratory factor analysis (EFA), Cronbach’s Alpha internal reliability tests, and one-way ANOVAs to answer the research questions. For one of the research questions, all of the participants were placed into groups based on the NPPLI/PPLI distinction illustrated in the previous section.

How do Multilinguals Conceptualize Language Interactions?

99

Examining first the main motivation results reported in Thompson and Erdil-Moody (2014), tables 6–1 and 6–2 show that the NPPLI and PPLI groups differ significantly with regard to F1, the ideal L2 self (p