The Routledge Handbook of Arabic Second Language Acquisition 9781138940550, 9781315674261

Citation preview

THE ROUTLEDGE HANDBOOK OF ARABIC SECOND LANGUAGE ACQUISITION

The Routledge Handbook of Arabic Second Language Acquisition introduces major current approaches in Arabic second language acquisition (SLA) research and offers empirical findings on crucial aspects and issues to do with the learning of Arabic as a foreign language and Arabic SLA. It brings together leading academics in the field to synthesize existing research and develops a new framework for analyzing important topics within Arabic SLA. This handbook will be suitable as a reference work for advanced undergraduate and postgraduate students and scholars actively researching in this area and is primarily relevant to sister disciplines within teacher training and Arabic applied linguistics. The themes and findings should, however, also be attractive to other areas of study, including theoretical linguistics, psycholinguistics, cognition, and cognitive psychology. Mohammad T. Alhawary is Professor of Arabic linguistics and second language acquisition at the University of Michigan, where he teaches courses on both Arabic language and Arabic theoretical and applied linguistics.

THE ROUTLEDGE HANDBOOK OF ARABIC SECOND LANGUAGE ACQUISITION

Edited by Mohammad T. Alhawary

First published 2018 by Routledge 2 Park Square, Milton Park, Abingdon, Oxon OX14 4RN and by Routledge 711 Third Avenue, New York, NY 10017 Routledge is an imprint of the Taylor & Francis Group, an informa business © 2018 selection and editorial matter, Mohammad T. Alhawary; individual chapters, the contributors The right of Mohammad T. Alhawary to be identified as the author of the editorial material, and of the authors for their individual chapters, has been asserted in accordance with sections 77 and 78 of the Copyright, Designs and Patents Act 1988. All rights reserved. No part of this book may be reprinted or reproduced or utilised in any form or by any electronic, mechanical, or other means, now known or hereafter invented, including photocopying and recording, or in any information storage or retrieval system, without permission in writing from the publishers. Trademark notice: Product or corporate names may be trademarks or registered trademarks, and are used only for identification and explanation without intent to infringe. British Library Cataloguing-in-Publication Data A catalogue record for this book is available from the British Library Library of Congress Cataloging-in-Publication Data A catalog record for this book has been requested ISBN: 978-1-138-94055-0 (hbk) ISBN: 978-1-315-67426-1 (ebk) Typeset in Times New Roman by Apex CoVantage, LLC

CONTENTS

List of contributors viii Acknowledgmentsx Introduction1 PART I

Arabic L2 phonology and phonetics

7

 1 Frequency and L1 transfer effects for the perception and production of Arabic lexical stress by L1 English and L1 Chinese learners of Arabic as an L2 Cheng-Wei Lin and Mohammad T. Alhawary

9

 2 Production of Modern Standard Arabic lexical stress cues by native speakers of American English Mashael Al-Aloula

38

 3 Native English speakers’ perception and production of Arabic consonants Asmaa Shehata

56

 4 The perception and production of Arabic consonants: a crosslinguistic study Sara Al Tubuly

70

 5 Arabic L2 phonological acquisition: an ultrasound study of emphatics and gutturals Amanda Eads, Jodi Khater, and Jeff Mielke

93

v

Contents

 6 The L2 acquisition of Modern Standard Arabic final consonant clusters by L1 Chinese speakers Mona Maamoun PART II

113

Arabic L2 vocabulary

137

 7 Looking at words: an eye-tracking investigation of L2 Arabic vocabulary learning Ayman A. Mohamed

139

 8 Keyword vs. context strategies among different levels of Arabic language learners Olla Najah Al-Shalchi

157

PART III

Arabic L2 morphosyntax

179

 9 The acquisition of resumptive pronouns: how do second language learners of Arabic do it? Dola Algady

181

10 Arabic L2 learners’ use of word order and subject-verb agreement for actor role assignment Jamil Al-Thawahrih

201

PART IV

Arabic L2 reading and corpus-aided language learning 11 Corpus linguistics and critical reading and thinking: proposals for teaching learning sequences based on journalistic corpora in Modern Standard Arabic Nadia Makouar PART V

223

225

Arabic L2 writing: discourse analysis and measuring production

249

12 Writing in Arabic: discourse analysis and pedagogical reflections Dris Soulaimani

251

vi

Contents

13 Comparing the complexity, accuracy, and fluency of written Arabic in the production of advanced learners and native speakers Michael Raish PART VI

265

Arabic L2 speaking and intercultural learning (in study abroad)

287

14 Code-switching in L2 Arabic collaborative dyadic interactions Khaled Al Masaeed

289

15 Research-based interventions for language and intercultural learning Emma Trentman

303

PART VII

Arabic heritage learners

329

16 Proficiency in standard Arabic and its predictors: the case of heritage speakers in college-level elementary Arabic classrooms Abdulkafi Albirini

331

17 Effect of age of acquisition on concept mediation in heritage Arabic bilinguals Iyad Ghanim

362

PART VIII

The Arabic L2 teacher: teacher training and self-positioning

385

18 Effect of using a collaborative video-based self-evaluation activity on helping AFL student-teachers tie theory to practice Raghda El Essawi

387

19 Arabic language teaching in the U.S.: two Arabic language users’ views on culture and self-positioning as teachers Brahim Oulbeid

402

Index422

vii

CONTRIBUTORS

Mashael Al-Aloula, George Mason University, USA Abdulkafi Albirini, Utah State University, USA Dola Algady, Al-Zahra College for Women, Muscat, Oman Mohammad T. Alhawary, University of Michigan, USA Khaled Al Masaeed, Carnegie Mellon University, USA Olla Najah Al-Shalchi, University of Texas, Austin, USA Jamil Al-Thawahrih, Defense Language Institute, Monterey, USA Sara Al Tubuly, Al Maktoum College of Higher Education, UK Amanda Eads, North Carolina State University, USA Raghda El Essawi, American University in Cairo, Egypt Iyad Ghanim, Montclair State University, USA Jodi Khater, North Carolina State University, USA Cheng-Wei Lin, University of Michigan, USA Mona Maamoun, Alexandria University, Egypt Nadia Makouar, INALCO, Paris, France Jeff Mielke, North Carolina State University, USA viii

Contributors

Ayman A. Mohamed, Michigan State University, USA Brahim Oulbeid, University of Massachusetts Amherst, USA Michael Raish, College of William and Mary, USA Asmaa Shehata, University of Ottawa, Canada Dris Soulaimani, San Diego State University, USA Emma Trentman, The University of New Mexico, USA

ix

ACKNOWLEDGMENTS

Numerous individuals and entities contributed to this Handbook in many ways. First, a debt of gratitude is owed to the anonymous participants of the studies reported on in this volume from different parts of the world, without whom the studies and the Handbook would not have happened. Second, I would like to express my gratitude to the authors for their dedication to and enthusiasm for the project. I also appreciate their patience with my multiple queries and requests for revisions. Third, I am extremely grateful to the anonymous reviewers of all the chapters. Their generous contributions of time and expertise contributed in no small measure to the quality of this volume. Fourth, my thanks are extended to the various co-sponsors at the University of Michigan who supported the conference “Investigating Arabic Second Language Acquisition: Empirical Findings and Trends” held at the University of Michigan, Ann Arbor, on September 23–25, 2016. The conference served as a significant discussion forum and allowed most contributors to the volume to present their research studies and receive feedback from the audience. Co-sponsors of the conference included: Department of Near Eastern Studies, Department of Linguistics, Center for Middle Eastern and North African Studies, International Institute, Institute for the Humanities, Rackham Graduate School, and University of Michigan Office of Research. My thanks also go to Andrea Hartill at Routledge, who believed in and supported this project with foresight and enthusiasm. Last but not least, I am forever grateful to my wife and children for putting up with the time taken away from them to complete this volume and for their tremendous support and enthusiasm.

x

INTRODUCTION

The present Handbook has been compiled in the hope of filling a critical gap in Arabic second language acquisition (SLA) research. The objectives are to offer a range of Arabic SLA studies which adopt state of the art research methods and techniques (qualitative and quantitative) in various L2 subfields, provide empirical data and findings on many language components and processes of Arabic as an L2, and generate more research interest in second language acquisition (SLA) studies – an area which remains underinvestigated. Because such research findings contribute to our understanding of how Arabic as an L2 is learned, the findings are also crucial to informing second language pedagogy and other related areas, such as testing and curriculum development. The vast majority of studies have so far been conducted on Arabic morphosyntax (e.g., Alhawary 2009 and Forthcoming; for a detailed review of such studies, see Alhawary 2009, pp. 21–48). Little has been investigated with respect to other L2 Arabic language components (such as phonology/pronunciation, vocabulary, and interlanguage pragmatics) and the different learning processes and skills (reading, writing, speaking, and listening), and only recently do we find studies on Arabic heritage learners, although the latter studies are part of a recently emerging subfield (for a detailed survey of such studies, see Alhawary 2018). It is hoped from this point on that Arabic SLA will continue to investigate as well as expand investigation coverage of the different Arabic language components and the different aspects related to the processes of Arabic second language learning so that adequate research coverage (both in depth and breadth) can be achieved. By the same token, it will also be important to explore other emerging approaches such as usage-based and emergentist or input frequency accounts (e.g., Bybee and Hopper 2001; Ellis 2002, 2012; Agren and Van de Weijer 2013; Rebuschat and Williams 2012; Römer et al. 2014), beside established rule-based approaches, prominent among which is the formal generative framework. Since the field of Arabic SLA is generally underinvestigated, it made little sense to invite contributors to write chapters exploring second language acquisition issues and approaches from a general perspective. Rather, the intention is to focus on works which are specifically conducted on Arabic SLA. This has the additional advantage of enriching the burgeoning Arabic SLA field and providing empirical evidence on various aspects of second language development. Furthermore, because of the close affinity between second language learning and 1

Introduction

second language teaching and because understanding how Arabic as an L2 is learned can crucially inform teaching it, I asked authors to provide, in the conclusion section of each chapter, specific implications of how their study findings may be applied in instructional contexts so that both the Arabic L2 learner and Arabic foreign language teacher may benefit from research results. The chapters included in this volume address issues and aspects in Arabic L2 phonology and phonetics, vocabulary learning, morphosyntax, reading, writing, speaking, and Arabic heritage learners, as well as the Arabic L2 teacher. The studies represent current trends and approaches in the field of second language acquisition as applied to Arabic on a range of areas of Arabic second language acquisition and second language development. The book is divided into eight parts comprising 19 chapters. The first part contains six chapters and deals with L2 Arabic phonology and phonetics. The first two chapters by Lin and Alhawary and Al-Aloula investigate an area in L2 Arabic phonology (suprasegmental features) which had been unexamined before: lexical stress. The first chapter, by Lin and Alhawary, investigates the perception and production of lexical stress by L1 English and L1 Chinese learners of Arabic and explores input frequency and L1 transfer effects. The second chapter, by Al-Aloula, examines L1 transfer and the production of lexical stress by L1 English learners of Arabic by focusing on the phonetic cues of duration, pitch, and intensity. The third chapter, by Shehata, explores the ability of L1 English learners of Arabic to perceive and produce 10 Arabic consonant contrasts and the relationship between learners’ perception and production. The fourth chapter, by Al Tubuly, investigates the role of L1 in the perception and production of Arabic emphatic, uvular, and pharyngeal consonants by Arabic L2 learners who are L1 speakers of English, Greek, Chinese, Turkish, and German. The fifth chapter, by Eads et al., is an ultrasound study of Arabic emphatic and guttural consonants and provides a detailed analysis and exploration of how L1 English learners of Arabic articulate them, at what level they are able to differentiate their articulation, and in what order. The sixth chapter, by Maamoun, investigates the validity of the universal sonority scale principle in acquiring the pronunciation of Arabic final consonant clusters by L1 Chinese learners of Arabic and the difficulty such learners encounter with Arabic word-final consonant clusters. The second part includes two chapters which focus on L2 Arabic vocabulary. Chapter 7, by Mohamed, is an eye-tracking study of L2 Arabic vocabulary learning by English-speaking learners of Arabic and explores whether or not the amount of attention to (novel) words over repeated encounters can predict readers’ incidental acquisition of multiple aspects of vocabulary knowledge. Chapter 8, by Al-Shalchi, compares experimentally between two vocabulary learning strategies (keyword and context) by L1 English learners of Arabic at various levels of proficiency to determine the impact of each on vocabulary learning and perceived workload during instruction. The third part focuses on the acquisition of Arabic L2 morphosyntax. Chapter 9, by Algady, presents a (generative) minimalist account of the role of economy conditions on the syntactic derivation of Arabic L2 resumptive pronouns within four types of relative clause constructions: direct object, indirect object, oblique, and subject relative clauses. Chapter 10, by AlThawahrih, examines the competition model’s prediction of L2 learners’ initial reliance on L1 cues of word order in processing and interpreting sentence structure by L1 English learners of Arabic and whether such learners would instead use an L2 gender verbal agreement cue in assigning the actor role. The fourth part relates primarily to Arabic L2 reading and the use of corpus linguistics and includes one chapter (Chapter 11), by Makouar. The chapter extends the application of interpretive 2

Introduction

semantics theory and use of corpus analysis tools to Arabic L2 reading as a framework which aims at developing the leaner’s metalinguistic awareness of word meanings and textual analysis in order to achieve deeper understanding of texts and for L2 learners to become critical readers. The approach is demonstrated in a case study with four French-speaking learners of Arabic at the intermediate proficiency level. The fifth part focuses on Arabic L2 writing from a discourse analysis perspective as well as measuring writing development. In Chapter 12, Soulaimani draws on a discourse analysis framework, which employs in particular the notions of voices and stance, in analyzing writings by advanced English-speaking learners of Arabic for use of discursive features and strategies and compares such L2 writing output production with that of L1 Arabic speakers. Similarly, in Chapter 13, Raish examines writings by advanced English-speaking learners by comparing them with those of native Arabic speakers and applying the CAF framework with its tripartite statistical measure of complexity, accuracy, and fluency. The sixth part deals with promoting Arabic L2 speaking skill and language and intercultural learning. The two chapters included also happen to address speaking and language and cultural learning in the context of study abroad. In Chapter 14, Al Masaeed follows a nuanced understanding of Myers-Scotton’s markedness model for code-switching and explores codeswitching (Arabic to English) functions during conversation dyads between English-speaking Arabic L2 learners and their Arabic native-speaking language partners outside of the classroom. In Chapter 15, Trentman reports (qualitatively) on two case studies of three Englishspeaking learners of Arabic. One case study includes one participant who was pursuing an ethnographic project for study abroad; the other study includes two participants who took part in a classroom telecollaboration project, one as a student and the other as a language partner. The two studies explore the impact of inclusion of “guided reflection” (that draws from linguistic and cultural evidence) on the participants’ language and cultural learning during telecollaboration and study abroad, respectively. The seventh part relates to Arabic heritage learners. In Chapter 16, Albirini analyzes the oral and written production of 29 Arabic heritage learners in terms of the three proficiency measures of fluency, (grammatical) accuracy, and (syntactic) complexity. Based on the participants’ functional command in standard Arabic, Albirini discusses why it is theoretically and pedagogically crucial not to approach heritage speakers on a par with foreign/second language learners of Arabic. In Chapter 17, Ghanim investigates the implications of a lexico-semantic model of bilingual lexical storage and retrieval and whether it is no longer possible for heritage speakers (as a population of bilinguals who learned one language, Arabic, in childhood but lost fluency on their way to adulthood) to activate the semantic information of (Arabic) lexical items directly from the conceptual (mental) domain, and whether instead they do so (indirectly) from semantic information associated with words in the other language (English). To test this prediction and to dissociate between age and proficiency effects, Ghanim employs a picture-naming and a translation task with 11 heritage Arabic-English bilinguals with varying degrees of Arabic proficiency. The book concludes with Part VIII, especially devoted to the Arabic second/foreign language teacher, due to the special emphasis of the present volume and stronger connection it attempts to make between second language acquisition research and second language pedagogy and bring more alignment between the two so that effective language learning and teaching can be achieved. More research on teacher preparation, teacher education, teacher cognition, and other internal and external variables is needed to examine arising issues and needs of the teacher so that they can be understood and better met. Some of these aspects are investigated in Chapters 18–19. Chapter 18, by El Essawi, is an evaluative study of 3

Introduction

student-teachers’ level of self-reflection following a collaborative activity which employed the Vialogues video discussion tool. Self-reflection in the study is defined as the ability to draw upon abstract theoretical knowledge gained in teacher-education and teacher-training programs in order to evaluate the effectiveness of student-teachers’ own pedagogical practices. The study focuses on five student-teacher participants who were native speakers of Arabic. Chapter 19, by Oulbeid, is a case study of two non-native Arabic speaker college-level teachers. It starts by exploring their beliefs about Arab culture and how such beliefs affect their classroom practices, then investigates how the two teachers negotiate and construct their identities as teachers of Arabic as a foreign language and culture professionals and how they categorize themselves as non-native speaking teachers. The Handbook will serve as a vital resource for Arabic SLA researchers and students, Arabic applied linguists, Arabic teachers and practitioners of Arabic foreign language teaching, Arabic language testers, and curriculum developers. It can be used as a course book in any courses related to Arabic applied linguistics, Arabic second language acquisition, and foreign language Arabic teaching methodology by graduate students and upper undergraduate students. The Handbook will also be useful to scholars and practitioners in second language acquisition and other neighboring disciplines, including theoretical linguistics, psycholinguistics, and cognitive psychology.

Transcription/transliteration symbols The transcription of all Arabic texts in the body of the Handbook follows a uniform system of simplified and slightly modified International Phonetic Alphabet (IPA) symbols, with some standard equivalents used widely in Arabic and Middle Eastern studies journals. For the purpose of the chapters included in the volume, Table 0.1 represents the phonetic chart of Arabic (MSA) consonants and their symbols used throughout. As for vowels, Table 0.2 lists MSA vowels and some variants of dialectal varieties as cited in the chapters. For the citation of Arabic titles and names of Arab authors in the references sections, a simplified transliteration system based on standard usage in Arabic and Middle Eastern studies journals has been adopted. The symbol ’ represents the hamza (glottal stop) and ‘ represents the ‘ayn (voiced pharyngeal fricative consonant).

4

Table 0.1 Arabic (MSA) phonetic chart of consonants and their symbols

‫م‬

f

ṭ

‫د‬

ḍ

k

‫ض ط‬

q

‫ك‬

ʔ

‫ق‬

‫ء‬

‫ن‬ θ

‫ف‬

d n

m

Nasal (Stop) Fricative

‫ ت‬

Palato- Palatal Velar Uvular Pharyngeal Glottal alveolar

Alveolar

t

‫ب‬

Dental

b

Stop

Emphatic Alveolar

Dental

Arabic Labial Plain Consonants

‫ ث‬

ð s

z

ð̣ ṣ

‫ز س ذ‬

‫ص ظ‬

š

x ‫ خ‬

‫ش‬

γ ħ

‫ ح غ‬

ʕ h

‫هـ ع‬

ž

Affricate

‫ج‬ r

Tap/Trill

‫ر‬ l ‫ل‬

Lateral

w

Glide

y

‫و‬

‫ي‬

w

‫و‬

By convention (see also Ladefoged 2001), consonants to the left of a cell are voiceless and those to the right are voiced; note also, in particular, /d/ and /ḍ/ are not complete (plain versus emphatic) contrasts, since the production of /ḍ/ involves additionally both sides of the tongue touching/pressing against the inner sides of the upper molars. In addition, depending on their own dialectal varieties, Arabic native speakers may produce the uvular fricatives as velar fricatives. Table 0.2 Arabic vowels (for broad transcription)

Arabic Symbol

Transliteration Symbol

Description

a

short front/back low

aa

long front/back low

u

short high back rounded

uu

long high back rounded

i e

short high front unrounded (MSA) short mid front unrounded (colloquial)

‫ي‬

ii

long high front unrounded

ْ‫ي‬

ay ee

diphthong (MSA) monophthong (colloquial)

‫ْو‬

aw oo

diphthong (MSA) monophthong (colloquial)

‫ا‬ ‫و‬

Introduction

References Agren, M. and Van de Weijer, J., 2013. Input frequency and the acquisition of subject-verb agreement in number in spoken and written French. Journal of French Language Studies, 23 (3), 311–333. Alhawary, Mohammad T., 2009. Arabic second language acquisition of morphosyntax. New Haven, CT: Yale University Press. Alhawary, Mohammad T., 2018. Empirical directions in the future of Arabic second language acquisition and second language pedagogy. In: Kassem M. Wahba, et al., eds. Handbook for Arabic language teaching professionals in the 21st century. New York: Routledge, 408–421. Alhawary, Mohammad T., Forthcoming. Arabic second language learning and effects of input, transfer, and typology. Washington, DC: Georgetown University Press. Bybee, J. L. and Hopper, P. J., 2001. Frequency and the emergence of linguistic structure. Amsterdam: John Benjamins. Ellis, N. C., 2002. Frequency effects in language acquisition: A review with implications for theories of implicit and explicit language acquisition. Studies in Second Language Acquisition, 24 (2), 143–188. Ellis, N. C., 2012. Frequency-based accounts of second language acquisition. In: S. Gass and A. Mackey, eds. The Routledge handbook of second language acquisition. New York: Routledge, 193–210. Ladefoged, Peter. 2001. A course in phonetics. Boston, MA: Heinle and Heinle. Rebuschat, P. and Williams, J. N. 2012. Statistical learning and language acquisition. Berlin: De Gruyter Mouton. Römer, U., et al., 2014. Second language learner knowledge of verb-argument constructions: Effects of language transfer and typology. The Modern Language Journal, 98 (4), 952–975.

6

PART I

Arabic L2 phonology and phonetics

1 FREQUENCY AND L1 TRANSFER EFFECTS FOR THE PERCEPTION AND PRODUCTION OF ARABIC LEXICAL STRESS BY L1 ENGLISH AND L1 CHINESE LEARNERS OF ARABIC AS AN L2 Cheng-Wei Lin and Mohammad T. Alhawary One of the areas in Arabic second language (L2) acquisition research which has for so long remained underexamined is L2 phonology. Even more woefully lacking are studies on Arabic suprasegmental features, including the perception and production of stress. Studies which have been conducted on L2 stress perception and production have mostly been examined in English L2 learners, and most such studies have not controlled for input frequency effects. To bridge this gap in both the second language acquisition literature in general and Arabic second language acquisition in particular, the present chapter examines the effect of frequency in the perception and production of Arabic L2 lexical stress as well as the role of the native language (L1). Usage-based accounts posit a positive role for input frequency, predicting faster reaction time and higher accuracy production rates. To examine these predictions, a lexical stress perception and production experiment was conducted in which the stimuli were manipulated to be frequency-contrastive in order to detect a biased performance preferring frequent stress patterns. The experiment was administered to four groups of first- and second-year L1 English and L1 Chinese learners of Arabic and a control group of native Arabic speakers. The results of the study experiment only partially support the frequency effect prediction. The effect of frequency is evident in learners’ responses only when the stimuli have less processing load and greater contrast in frequency. Although the study found that learners’ performance in the experiment is, to a great extent, conditioned by perception of acoustic cues and prosodic characteristics of their L1, frequency effects should not be disregarded. The study found that the effect of frequency captures aspects of learners’ performance that are not conditioned by their L1. Specifically, the participants exhibited higher fluency and accuracy in their production of stress when the stimuli had a relatively more frequent stress pattern, and vice versa.

9

Cheng-Wei Lin and Mohammad T. Alhawary

1.1 Introduction Unlike stress in English, which varies in position and meaning (e.g., ˈrecord vs. reˈcord), stress in Arabic is predictable, though it does not occur in a fixed position. Stress in Arabic is weight-sensitive, and the stressed syllable of a word is determined by an algorithm (i.e., a set of rules that compare the position and the weight of syllables of a word) which can vary from dialect to dialect. Some studies (on Arabic L1 phonology) focus on constructing different algorithms for the different varieties, such as Modern Standard Arabic (Angoujard 1990), Palestinian (Kenstowicz 1983), Damascene (Cowell 1964), and Egyptian (Mitchell 1975). The perception and production of stress has received some attention in Arabic L1 but almost no attention in Arabic as an L2. Most studies on Arabic L1 phonology dealt with the acoustic properties of stress production (e.g., Heliel 1982; de Jong and Zawaydeh 1999, 2002), with less focus on stress perception. However, research on stress in English as an L2 is abundant, though most of it has been conducted on English as an L2 (cf. studied on other L2s, such as Spanish (Face 2005, Dupoux et al. 2008; Alfano et al. 2009), Portuguese (Garcia 2016), and Italian (Alfano et al. 2007)). English L2 studies involved participants from different language backgrounds, such as Arabic (Sheikh 1987; Anani 1989; Guma 2003; Almbark et al. 2014; Bouchhioua 2008), Japanese (Archibald 1997; Ueyama 2000), Mandarin Chinese (Chen et al. 2001; Lai 2008; Wang 2008; Wang and Yoon 2008; Yu and Andruski 2011), Thai (Jangjamras 2011), and Vietnamese (Nguyen 2004). L2 stress perception and production studies have mainly investigated the role of L1 transfer, falling into three main strands of studies. The first strand directly compares the performance of L2 learners and that of L1 speakers of a specific L1, such as Chen et al. (2001), Lai (2008), Wang (2008), and Jangjamras (2011), who examine whether learners preferred the same set of acoustic correlates or cues of stress (duration, pitch, and intensity) as the L1 speakers or they attended to the relevant correlates that are active in their L1. The second strand investigates other properties of stress and whether L2 learners transferred prosodic structures, such as syllabic structure, syllable weight (Face 2000; Bullock and Lord 2003; Face 2005), stress position (Archibald 1993a, 1993b, 1997), extrametricity (Garcia 2016), and stress typicality (Yu and Andruski 2011), from their L1. The third strand pursued a wider typological approach to L2 stress perception by dealing with typological differences in stress perception and production between different learners. Among the third strands of studies, and most relevant to the present study, is a series of studies which were conducted to investigate the ability to identify stress by speakers of different L1s (e.g., Dupoux et al. 1997; Peperkamp et al. 1999; Dupoux et al. 2001; Peperkamp and Dupoux 2002; Dupoux et al. 2008; Peperkamp et al. 2010). In particular, Peperkamp et al. (1999) and Dupoux et al. (2001) used stimuli which consisted of nonsense words with stress manipulated in contrastive positions (e.g., ‘piki vs. pi’ki). The studies revealed that L1 speakers of French, a language with highly predictable stress, had more difficulty in perceiving stress than others, especially when the acoustic correlates were not kept constant or when the stimuli were produced by different speakers. Similar difficulty was faced by speakers of other L1s such as Finnish and Hungarian, and (marginally so) Polish. Based on the findings, the stress deafness model was advanced. The model posits that there is a negative relation between the predictability of stress in the speaker’s L1 and the speaker’s aptitude in perceiving stress in an L2. In other words, the more predictable a stress pattern in a speaker’s L1, the more difficult it is for the speaker to discriminate stress between a minimal pair with contrasting stress position. Subsequently, and using English novel words, Altmann (2006) examined stress perception and production in advanced learners of English belonging to seven distinct L1s. The study was 10

Frequency and L1 transfer effects

intended to test the predictions made by the stress deafness model (Peperkamp and Dupoux 2002) and the stress typology model (Altmann 2006). The findings revealed that speakers of different L1s perceived stress differently: that learners with a predictable stress in their L1 (e.g., French) had more difficulty perceiving stress than learners without word-level stress in their L1 (e.g., Chinese) or ones with unpredictable stress in their L1 (e.g., Spanish). However, the reversed pattern was found in stress production. Learners with predictable stress in their L1 performed most like native English speakers, while learners without stress or with unpredictable stress in their L1 performed differently than native English speakers. The results are interpreted in support of the stress typology model rather than the stress deafness model. An extensive review of studies on L2 stress is beyond the scope of the present study. Despite the fact that such studies have improved our understanding of the perception and production of stress, such studies are not without limitations. These include, as pointed out by Altmann (2006), potential bias in study design due to use of real words instead of nonsense words as stimuli without controlling for prior knowledge of such words, use of nonsense words that are not natural, and lack of sufficient and systematic manipulation of stress in contrastive contexts, among others. In addition, most such studies are mainly conducted on English as an L2, a language whose stress system is arguably distinctively different from that of Arabic. Therefore, it remains unclear to what extent the findings from these stress perception and production studies with English as an L2 could be applied to Arabic as an L2. Finally, little research has been conducted on L2 stress while taking frequency effects into account. A notable exception is Silveira (2011), who investigated whether the high frequency of penultimate stress position exhibited in both the participants’ L1 Brazilian Portuguese and their L2 English could facilitate or impede their production of stress in English. A stress production task was conducted using stimuli consisting of bisyllabic, trisyllabic, and quadrisyllabic real words, with varying stress positions being first, second, or the third syllable of the word. The experiment was conducted twice with a six-month interval, employing 16 participants with varying proficiency levels. The results did not reveal any difference in accuracy of stress production between the two periods, suggesting lack of any proficiency effect. However, participants were found to place stress on the penultimate position significantly more so than any other position of the word. It was also found that learners resorted to the preferred position of stress (penultimate) in their L1 when dealing with words that had stress-bearing suffixes. Silveira (2011) concluded that the preference for penultimate position comes from the frequency effects of both L1 and L2. The study had a number of methodological limitations, including use of only real words, lack of control for the frequency of stress patterns, and presence of a confounding variable where the penultimate is preferred in both correct and incorrect production and, therefore, it is not possible to tease apart L1 from L2 effects in either case.1 Notwithstanding the little research done on L2 phonology from a usage-based or frequency approach, usage-based and input frequency approaches have been increasingly explored in the field of psycholinguistics and first and second language acquisition (e.g., Ellis 2002, 2012; Ellis et a. 2014; Goodsitt et al. 1993; Gómez and Gerken 2000; Smith and Yu 2008).

1.1.1  Usage-based learning and acquisition of L2 stress Recent studies which have shown a strong link between frequency and language processing and learning have led to the emergence of the “usage-based approach” in language learning. Synonymous with frequency learning or statistical learning, the new approach assumes that learning is bottom-up driven, statistical, and frequency informed. In terms of first and second language acquisition, target structures and idiosyncratic patterns of language are claimed to emerge from 11

Cheng-Wei Lin and Mohammad T. Alhawary

usage or exposure to input. The learner tracks the frequency of the linguistic input and extracts regularities about the language from the abstractions of frequent linguistic elements. As the learner repeats this process of tracking and abstracting characteristics of the language from the frequency distribution of linguistic elements over the course of a lifetime of learning, the language emerges as one that has approximated the regularities of the input language. Thus, from a usage-based, frequency-informed perspective, the acquisition of lexical stress in Arabic proceeds from usage and repeated exposure to the input from which learners extract a variety of relevant statistical knowledge of how stress is distributed in a word in Arabic and use the extracted frequency information to help them perceive and produce stress in a way that they think is more frequent and more probable. Accordingly, the usage-based approach predicts that learners’ performance will be biased by the frequency of the input that they receive, and they will be more accurate and more fluent in perceiving and producing stress that is more frequent, and vice versa. In technical terms, and based on the algorithm for stress assignment in Arabic, the learner will need to utilize a variety of relevant statistical information (e.g., à la Romberg & Saffran 2010), i.e., information which includes the frequency of: (a) the syllabic structure or the “stress pattern” of the word, (b) the syllabic structure of the stressed syllable, (c) the position of the stressed syllable, and (d) the conditional probability of the stressed syllable. To illustrate this, consider the word mil.ʕa.qa “a spoon”; its syllabic structure is [ˈCVC.CV.CV], the syllabic structure of the stressed syllabic is [ˈCVC], the position of stress is on the antepenultimate position, and the conditional probability means having a stressed [CVC] not preceded by any syllable, which means the syllable [CVC] is in the word-initial position.

1.2  Research questions Following the usage-based account in language learning, it is then predicted that the frequency of stress-related structures could either individually or collectively inform the performance of the learners in stress perception and production. In addition to examining the effect of the native language of the learners, focus of the present study is on examining the relationship between the frequency of stress-related cues and learners’ performance. The investigation undertaken in the present chapter addresses the following questions: 1 Does the L1 of Arabic L2 learners influence the perception and production of Arabic lexical stress? 2 Does the proficiency of Arabic L2 learners influence their accuracy and fluency of Arabic lexical stress perception and production? 3 Does input frequency effect play a role in Arabic L2 learners’ perception and production of Arabic lexical stress? 4 How does frequency of stress-related cues influence Arabic L2 learners’ perception and production of lexical stress in terms of the accuracy and fluency of their responses?

1.3  Frequency distribution of stress and related cues in Arabic To capture the frequency distribution of stress-related cues (i.e., stress pattern, position, syllabic structure, and conditional probability of the stressed syllable), three sources were analyzed in the present study. First, the frequency dictionary for Arabic compiled by Buckwalter and Parkinson (2014) was used as the main source for estimating frequency distribution. The frequency dictionary was then further validated with frequency analysis of the Hans Wehr Arabic-English Dictionary (Wehr 1993) and the Al-Kitaab series (Brustad et al. 2004, 2007). 12

Frequency and L1 transfer effects

The frequency dictionary (Buckwalter and Parkinson 2014) was selected for its accessibility and relevance. It presents the most-frequently used 5,000 words based on a 30 millionword corpus of spoken and written words. Each entry includes information which is relevant to understanding a given word, such as its part of speech, English gloss, sample use in a sentence, and two types of frequency counts: “rank frequency” and “raw frequency.” Rank frequency is the relative frequency of an entry compared to other entries in the dictionary, whereas raw frequency indicates the frequency of a given entry in the entire corpus. For instance, the entry milʕaqa “a spoon” is scored 3,495 on the rank frequency and 873 on raw frequency, indicating that the word is the 3495th most frequent word compared to the other 4,999 words in the dictionary and appeared 873 times in the entire corpus. The raw frequency, in particular, is crucial to the analysis here, since it helps to estimate the frequency distribution of various cues for stress in Arabic based on the corpus size and content. Dictionary entries were converted to stress patterns (i.e., to minimal representation containing only information relevant for prosody) through the following steps: each MSA entry was first transcribed using IPA symbols, consonants and vowels were replaced with “C” and “V” (e.g., mil.ʕa.qa “spoon” → [CVC.CV.CV] and qadiim “old” → [CV.CVVC]), and then stress positions were determined, following (McCarthy and Prince 1990, p. 252) for MSA: “final syllables are stressed if superheavy CVVC or CVCC [or CVVCC]; penults are stressed if heavy CVV or CVC; otherwise the antepenult is stressed.” This process was conducted in Microsoft Excel, and the results were manually checked by the first author. The last step was merging the frequency count for entries that share the same stress pattern. For example, milʕaqa “a spoon” and silsila “a series” share the same stress pattern [ˈCVC.CV.CV]; their frequency counts, 873 and 1,894, respectively, were added up, resulting in a frequency count of 2,770 for the stress pattern [ˈCVC.CV.CV]. A total of 104 stress patterns with various frequency counts resulted from the 4,283 entries after excluding the dialect words which comprised a small portion of the 5,000 entries, nine entries which consisted of more than four syllables, and all monosyllabic words. Figure 1.1 illustrates the resulting frequency distribution of stress patterns (where each dot represents a stress pattern of a specific number of

Figure 1.1 Correlation between log frequency and syllable number of all stress patterns

13

Cheng-Wei Lin and Mohammad T. Alhawary

syllables) in logspace (i.e., log frequency), revealing two trends: (1) some stress patterns are frequent while others are not, as indicated by the distances found between similarly shaped dots; and (2) there is a negative correlation between the number of syllables and frequency of the stress patterns, suggesting that words with fewer syllables tend to be more frequent than ones with more syllables; that is, bisyllabic words are generally more frequent in the language than trisyllabic words, and trisyllabic words more frequent than quadrisyllabic words. The resulting frequency distribution of stress patterns from the frequency dictionary (Buckwalter and Parkinson 2014) was further statistically compared with that of the entries of the Hans Wehr English-Arabic Dictionary and the vocabulary lists of Al-Kitaab textbooks (Parts 1–3), after establishing the frequency distribution of stress patterns in the latter two sources, following the same steps in the frequency dictionary. The comparisons revealed a significantly positive correlation with the Hans Wehr Dictionary (Pearson’s r = .73, n = 104, p < 0.001), a moderately positive correlation with the first two volumes of Al-Kitaab (Pearson’s r = 0.50, n = 104, p < 0.001), and a strong positive correlation with the third volume of Al-Kitaab (Pearson’s r = 0.84, n = 104, p < 0.001). The positive correlations suggest that the frequency distribution obtained from the frequency dictionary is reflective of the frequency distribution found in the lexicon of the language and the language used in the language classroom. Although the resulting frequency stress distribution may not completely reflect actual language use, the frequency information extracted from the data is, to a strong degree of confidence, adequately representative of the language encountered by the learners, at least for the purpose of stimuli design for the experiment of the present study. Finally, in addition to frequency of stress patterns, frequency distributions for stress-related cues (of the frequency dictionary), including stress position, syllabic structure, and conditional probability of the stressed syllable, were extracted automatically and later checked manually. The resulting fine-grained frequency distributions (in log frequency) revealed a number of findings illustrated in Figures 1.1–1.4. Similar to the previous findings of Arabic stress patterns, some stress-related cues are more frequent than others. For example, stress position more frequently falls on the penultimate (72%) than on the final syllable (28%) in bisyllabic words (Figure 1.2). The syllabic structure CVV is more frequently stressed (45%) than the syllabic structure CVC (27%) in quadrisyllabic words (Figure 1.3). Finally, the conditional

Bisyllabic

Trisyllabic

Quadrisyllabic

72%

Frequency (%)

58% 49% 32%

27%

28% 19%

14%

0% Antepenultimate

Penultimate

Final

Stress Position

Figure 1.2 Frequency distribution of stress position grouped by syllable number

14

Frequency and L1 transfer effects Bisyllabic

Trisyllabic

Quadrisyllabic

Frequency (%)

45%

28% 29%27%

32% 27%

25% 19% 14%

19%

19%

14%

1% CVC

CVVC

CVV

CV

CVCC

Syllabic Structure

71%

48% 36% 36%

10%

ˈCV|CVV

ˈCVCC|CVV

ˈCVV|CVV

3% 1% ˈCVVC|CVV

1%

ˈCVCC|CVC

4% ˈCV|CVC

ˈCVC|CV

ˈCVVC|CV

ˈCVV|CV

ˈCVVC|**

ˈCVV|**

ˈCV|**

16%

13% 11%

ˈCVC|CVC

5%

2% ˈCVC|**

21%

ˈCVC|CVV

22%

ˈCVV|CVC

27%

ˈCVVC|CVC

32%

ˈCVVCC|CV

39%

ˈCVCC|CV

Frequency (%)

Figure 1.3 Frequency distribution of syllabic structure of stress grouped by syllable number

Syllabic Structure (Target|Condition)

Figure 1.4 Conditional probability of stress for all stress patterns

probability of ˈCVC|** (i.e., having a stressed CVC given that it is not preceded by any syllable) is higher (39%) than ˈCVV|** (27%), as shown in Figure 1.4. To examine the effect of frequency on Arabic learners’ stress perception and production, this frequency distribution information for frequent and infrequent stress-related cues was used in the present study to inform the design of frequency-contrastive stimuli and most effectively test L2 learners’ ability in perceiving and producing lexical stress.

1.4 Methods 1.4.1  Stimuli design Four stress-related cues, including frequency of the stress pattern, stress position, syllabic structure, and conditional probability of the stressed syllable, are focused on here. They are treated as variables due to their variability in frequency. To simplify the four variables, each individual 15

Cheng-Wei Lin and Mohammad T. Alhawary

variable is based on binary distinctions of either “relatively frequent” (henceforth “frequent”) or “relatively infrequent” (henceforth “infrequent”) compared to other stress patterns. Thus, a stress pattern [CV.ˈCVV.CV] can be frequent as a stress pattern, but its stress position can be relatively infrequent. In addition, the following were taken into account when constructing the stimuli: naturalness (i.e., where the stimuli are a balanced mix of real and nonsense words) and length of the stimuli, where focus of the study is on stress patterns that are trisyllabic and quadrisyllabic. Bisyllabic stress patterns are not included, since they are generally more frequent, which makes them too simple to detect any difference in performance, have less variety of stress patterns, and have higher expected value for random guesses to be correct. Additionally, to keep the stimuli at a manageable size with an adequate number of blocks and reasonable time for the participants to complete the tasks, focus here is limited to (1) frequency of the stress pattern and (2) position of the stressed syllable, while the syllabic structure of the stressed syllable as well as its conditional probability are controlled for in the stimuli design itself. Accordingly, by factoring in the frequency of the stress pattern and stress position in addition to the number of syllables and the real-nonsense word contrast, the size of each block was effectively reduced to 24 = 16 tokens. This yielded eight stress patterns that are contrastive in terms of their number of syllables, frequency of stress pattern, and frequency of position, as presented in Table 1.1. The log frequencies of the resulting stress patterns are also presented in the table; they were used as a basis to both identify real words and construct nonsense words as the stimuli of the experiment. To construct the stimuli, the approach was to swap out the root of a real word with an unattested one, since not all consonants can freely combine in an Arabic root due to presence of phonotactic constraints. One such attested constraint is that of homorganic consonants, which blocks consonants sharing the same place of articulation from co-occurring in first and second position, such as *mmd (see Greenberg 1950). Two sets of nonsense words were used: one set of which forms minimal pairs with real words included in the stimuli by only swapping out one consonant from the root (such as ‫ˈ نَ َغ َر‬na.ɣa.ra derived from ‫ˈ نَ َش َر‬na.ša.ra by replacing /š/ with /ɣ/) and another set which swaps all root consonants with ones that are consistently used for other stress patterns in the same block of the stimuli (such as ‫س‬ َ ‫ˈ َع َم‬ʕa.ma.sa, in which the root [ʕms] is swapped for the stress pattern and the vocalic template [ˈCV.CV.CV] is controlled). A full list of the stimuli used in the study is included in the Appendix of this chapter. The rationale for the choice of the two sets of nonsense words in the stimuli is to control for the frequency effect of the co-occurrence of the root consonants.

Table 1.1 Stress patterns and variables considered for the stimuli

Trisyllabic

Quadrisyllabic

Stress Pattern

Frequency Stress Pattern

Frequency Stress Position

Frequency Syllabic Structure

Condition Probability

ˈCV.CV.CV ˈCV.CV.CVC CV.ˈCVV.CV CVC.ˈCVC.CVC CVC.ˈCV.CV.CV CVC.ˈCV.CV.CVC CVC.CV.ˈCVV.CV CV.CV.ˈCVV.CVC

+ (5.9) − (3.2) + (5.8) − (4.0) + (5.0) − (2.6) + (4.7) − (3.8)

+ (6.2) + (6.2) − (5.9) − (5.9) + (5.3) + (5.3) − (5.0) − (5.0)

+ (5.9) + (5.9) + (6.1) + (6.1) + (5.0) + (5.0) + (5.5) + (5.5)

+ (3.6) + (3.6) + (3.3) + (2.5) + (2.5) + (2.5) + (3.3) + (3.3)

16

Frequency and L1 transfer effects

1.4.2 Participants The participants were two groups of L1 (American) English and L1 Mandarin Chinese learners of Arabic as an L2 belonging to first- and second-year groups. In addition, a group of native Arab speakers participated as a control group (see Table 1.2). Participants in the two English-speaking groups were recruited from a tier 1 private research university in the East Coast in the United States, and participants in the two Chinese-speaking groups were recruited from a tier 1 public research university in Taiwan. Both institutions offer Arabic as a major, and both curricula focus on MSA, where students do not seem to have exposure to dialects other than a few words that are occasionally mentioned in class.2 A monetary incentive of $30 was given to participants for their participation. The control group consists of nine ESL students and university students studying at a public university in the Midwest in the United States. Four of them were undergraduate students, four were graduate students, and one was a recent high school graduate. They were all born and raised and educated in their Arabicspeaking countries as follows: one from Syria, one from Yemen, one from Palestine, and six from Saudi Arabia.

1.4.3  Stress perception and production tasks The stress perception and production tasks consist of a warm-up reading activity (to help participants adjust to using the target language) and three main tasks to assess participants’ performance in stress perception and production.3 The three main tasks are: a stress production task, a stress identification task, and a lexical decision task. The experiments were administered through PsychoPy, with a 13-inch window laptop computer at the time of data collection. The participants used a number pad as the control panel to navigate and complete tasks in the experiment via keypresses. The numbers and the yes/no key all start from right to left, which is counter-intuitive to English speakers but is catered to Arabic orthography written from right to left. All tasks in the experiment consist of a practice phase and a testing phase. Finally, the three main tasks were administered in the following order: the stress production task was administered first, and then the stress identification and lexical decision tasks were concurrently administered together. For ease of reference, the tasks are discussed here in the following order (not in the order in which they were administered): the stress identification task, then the lexical decision task, followed by the stress production task.

Table 1.2 Participants Groups English L1 Group 1 (n = 20) Group 2 (n = 13) Mandarin Chinese L1 Group 1 (n = 21) Group 2 (n = 13) Arabic L1: Control (n = 9)

Length of Exposure

Credit Hours Enrolled in

Gender M/F

Age Mean

SD

Year 1 Year 2

5 5

8/12 6/7

20.9 22.7

4.2 3.5

Year 1 Year 2 –

7 5 –

6/15 4/9 7/2

19.1 19.7 25.9

0.9 0.8 7.1

M/F = Total Males/Total Females

17

Cheng-Wei Lin and Mohammad T. Alhawary

1.4.3.1  Stress identification task and procedure The purpose of the stress identification task was to generally investigate learners’ perception of stress and whether their perception was influenced by the different frequency cues within the stimuli. In this task, the stimuli were presented to participants both aurally and visually. For each token, participants were presented with a recording of the stimulus produced by a male native speaker of Arabic (from Saudi Arabia) and saw the orthography of the word at the same time on the computer screen. After listening to the word, participants were asked to identify the stressed syllable in a word by pressing the corresponding number on the first row of the control panel, where the number “1” stood for the first syllable, “2” stood for the second syllable, and “3” for the third syllable. The participants were asked to complete each token as soon as they could. Since participants might not know the syllabification of the stimuli solely based on Arabic orthography, the orthography of the word was syllabified and numbered, but vowel information (short and long) was eliminated from the visual stimuli, as in the word ‫ نتيجة‬na.ˈtiː.ža “result” appearing without vowels, as shown in Figure 1.5, so that the participants could only be shown that the stimulus had three syllable: with [n] being the onset of the first syllable, [t] the second syllable, and [ž] the third. This was deemed necessary in order to avoid making the orthography reveal the pronunciation of a word and making participants rely on the visual aspect of the stimuli alone when identifying the stressed syllable, since Arabic spelling (with vowels included) is transparent, unlike English. Another reason for concealing vowel information was to avoid scenarios where learners might impressionistically but falsely prefer a certain type of vowel quality or vowel length in the script of stimuli without attending to the acoustic cues provided by the (aural) stimuli.

1.4.3.2  Results of stress identification task Both the fluency and the accuracy of participants’ responses to the stimuli were analyzed. Fluency in this task and throughout is operationalized as the reaction time or amount of time the participants took to respond to the stimuli. Accuracy refers to whether the syllable selected by the participants matches the attested position for the stressed syllable in the language, and responses are accordingly marked as correct or incorrect. Both measures provide different dimensions of assessing participants’ performance in stress perception: their performance was deemed worse if it took them a longer time to figure out the stress position, even if they did so correctly. Similarly, no matter how fast their response time was, their performance was considered worse if they did not identify the position of the stressed syllable correctly. Equally important here is that only participants’ response to stimuli that were nonsense words were analyzed, to avoid bias introduced from familiarity effects. To examine the effect of L1 on reaction time, a two-way ANOVA was conducted. The analysis revealed a main effect for L1 (F(2, 73) = 4.98, p < 0.001, partial η2 = 0.11) but no proficiency or interaction effect was found. Follow-up post hoc comparisons using Tukey tests

Figure 1.5 Sample of stimuli presentation in the stress identification task

18

Frequency and L1 transfer effects

1.0

showed that although there was no significant difference between the non-native (English L1 and Chinese L1) groups, the non-native groups had significantly shorter reaction time than the native group (p < 0.001). A similar result was found for accuracy. A main effect was found for L1 (F(2, 73) = 11.39, p < 0.01, partial η2 = 0.23), while no proficiency or interaction effect was found. Post hoc comparisons using Tukey tests showed again that although there was no significant difference between the non-native groups, the non-native groups scored significantly higher on accuracy (p < 0 .001) (see Figure 1.6). In other words, whereas the non-native groups performed unexpectedly better than the native group on both reaction time and accuracy, there was neither a between-group difference (i.e., related to L1) nor a within-group difference (i.e., related to proficiency). Before reporting the frequency results, a modification was made to the data analysis due to a confound identified after data collection, which is the lack of control for vowel length of the stressed syllable in the stimuli. It was found that learners seem to have a strong preference toward long vowels, resulting in significantly better performance for stress patterns of syllables containing long vowels. To resolve this bias, comparison is restricted to stress patterns whose stressed syllables have identical vowel length, affecting only one stress pattern out of the total eight (see Table1.1; cf. Table 1.3). Therefore, to look at how frequency influence participants’ performance, the adjusted schema for data analysis, along with the log frequencies for respective stress pattern, is presented in Table 1.3, in which five groups of patterns are listed (three trisyllabic and two quadrisyllabic) comprising seven stress patterns. Each group consists of two stress patterns that contrast in their frequency, consolidated into one variable: “Relative Frequency”. For instance, in Group 1, the stress pattern [ˈCV.CV.CV] is more frequent in terms of the frequency of stress pattern as well as the position of the stressed syllable than [CVC.ˈCVC.CVC] and is therefore denoted with “+” in the relative frequency column, suggesting relatively higher frequency. To determine the effect of frequency on the reaction time of participants’ responses to stimuli groups 1–5 (Table 1.3), a one-way ANOVA with repeated measures was conducted with relative frequency (two levels: relatively frequent and relatively infrequent) as a betweengroup variable, participant as the random variable to control for individual variability, and

0.4

Accuracy 0.6

0.8

English L1 Chinese L1 Arabic Native

1st Year

2nd Year Proficiency

Native

Figure 1.6 Accuracy of responses in the stress identification task according to participants’ L1 and proficiency

19

Cheng-Wei Lin and Mohammad T. Alhawary Table 1.3 Adjusted schema for stimuli comparisons based on relative frequency Stress Pattern Trisyllabic

ˈCV.CV.CV CVC.ˈCVC.CVC ˈCV.CV.CV ˈCV.CV.CVC ˈCV.CV.CVC CVC.ˈCVC.CVC Quadrisyllabic CVC.ˈCV.CV.CV CVC.ˈCV.CV.CVC CVC.CV.ˈCVV.CV CV.CV.ˈCVV.CVC

Frequency of Frequency of Relative Vowel Length Group Stress Pattern Stress Position Frequency Number + (5.9) − (4.0) + (5.9) − (3.2) − (3.2) − (4.0) + (5.0) − (2.6) + (4.7) − (3.8)

+ (6.2) − (5.9) + (6.2) + (6.2) + (6.2) − (5.9) + (5.3) + (5.3) − (5.0) − (5.0)

+ − + − + − + − + −

Short Short Short Short Short Short Short Short Long Long

1 2 3 4 5

Plus Sign (+) = Frequent, Minus Sign (−) = Infrequent

reaction time as the dependent variable. This analysis was consistently done for each of the five groups of participants (i.e., first-year English L1, second-year English L1, first-year Chinese L1, second-year Chinese L1, and the native group) and for each stimuli group. To provide a more concise presentation of the results, only results that had rather consistent patterns of findings are presented (for additional results of the study experiment, see Lin 2018). In Table 1.4, the results are organized based on the stimuli groups (Table 1.3). For each stimuli group, the table presents the estimate value (i.e., the mean difference in reaction time between frequent and infrequent stress patterns obtained from the post hoc comparisons, the f statistics and the p value from the one-way ANOVA with repeated measures done for each group of participants). The estimate value was calculated by deducting the mean reaction time (in millisecond) to stimuli that had infrequent stress pattern from the mean of the reaction time to stimuli that had frequent stress patterns. For example, if a group of participants had an average of 2,330 milliseconds for their reaction time to stimuli with frequent stress pattern and 2,960 milliseconds for infrequent stress pattern, the estimate is 2,335 minus 2,960, which equals −625, as is shown for the Arabic L1 control group in their responses to the stimuli in Group 1. In other words, a negative estimate would suggest that the average reaction time for the frequent stress pattern is shorter than the average reaction time of the infrequent stress pattern, and vice versa. For accuracy of responses, if the difference is negative, the estimate suggests that the degree of accuracy of response for stimuli with frequent stress pattern is lower than that of infrequent stress pattern and vice versa. Based on the predictions made in the study for frequency effects, it is expected for the participants to have shorter reaction time and higher degree of accuracy of response for stimuli with frequent stress pattern than ones with infrequent stress pattern. In other words, for frequency effects to be evident, a negative estimate value is expected for reaction time analyses and a positive estimate value is expected for accuracy of response analysis. Accordingly, a consistent pattern was found from the responses to the stimuli group 1, as shown in Table 1.4. A trend of shorter reaction time for stimuli with frequent stress pattern than stimuli with infrequent stress pattern was found in stimuli group 1 for the second-year English L1 group (F(1,193) = 7.45, p < 0.01, partial η2 =0.05) and second-year Chinese L1 group (F(1,194) = 22.19, p < 0.001, partial η2 = 0.14). However, the reversed trend was found for the group 4 stimuli, where the effect of relative frequency preferring infrequent stress patterns is significant in all non-native groups. In other words, the participants seemed to respond to stimuli in group 1 with frequent stress pattern more rapidly than ones with infrequent stress 20

Frequency and L1 transfer effects Table 1.4 Summary of frequency effects on reaction time – stress identification task Participants Stimuli Group 1 Estimate F Statistics Arabic L1: −625 Control English L1: −297 1st English L1: −444 2nd Chinese L1: 608 1st Chinese L1: −1458 2nd

Stimuli Group 4 p Value

Estimate

F Statistics

p Value

F(1,134) = 1.8

p = 0.18

−228

F(1,134) = 0.25

p = 0.62

F(1,313) = 2.63

p = 0.11

409

F(1,312) = 4.45

p < 0.05*

F(1,193) = 7.45

p < 0.01**

497

F(1,194) = 9.26

p < 0.01**

F(1,314) = 0.6

p = 0.44

995

F(1,314) = 33.97 p < 0.001***

F(1,194) = 22.19 p < 0.001***  1291

F(1,194) = 38.51 p < 0.001***

pattern; whereas they responded more rapidly to stimuli with infrequent stress pattern for stimuli in group 4, violating the predictions made by frequency effects. A similar analysis was conducted for the accuracy of responses for the stress identification task. Since the accuracy of response is a binary value, in that a given response to each stimulus could either be correct (coded as 1 in the dataset) or incorrect (coded as 0 in the dataset), a repeated-measures logistic regression was conducted with relative frequency (two levels: frequent and infrequent) as a within-group variable, participant as the random variable to control for individual variability, and the accuracy of response (two levels: correct and incorrect) as the dependent variable. Similarly, the repeated-measures logistic regression was consistently done for each stimuli group across all groups of participants. Only the results that show obvious patterns of findings are presented in Table 1.5. Like the previous ANOVA analyses, Table 1.5 presents the estimate of the statistical analysis, the p value for the test, and the odds ratio as the effect size statistics. Although negative estimate value seemed to be the trend for the stimuli in stimuli groups 3–4, the main effect of relative frequency was insignificant in most groups. Only some groups were found to have significant difference in the accuracy of response and these groups were rather scattered across different groups of stimuli. A more general trend of significant preference did not seem to emerge. In other words, the results suggested that the participants were more successful in identifying the correct stressed syllable for stimuli with infrequent stress pattern than with frequent stress pattern, suggesting once again violation of the prediction made by frequency effects.

1.4.3.3  Lexical decision task and procedure The lexical decision task was held concurrently with the stress identification task (discussed above). The lexical decision tasks targeted participants’ lexical knowledge of the Arabic lexicon and whether this knowledge could be influenced by the frequency of stress-related cues of the stimuli. More importantly, the motivation for including the lexical decision task was to use it as a distractor task from the main purpose of the experiment, which was to examine learners’ performance on lexical stress in Arabic. Thus, the task was intended to help participants respond to the stimuli more naturally and without taking an unnecessarily longer time on each stimulus in order to strive for the correct answer. After participants completed a token in the stress 21

Cheng-Wei Lin and Mohammad T. Alhawary Table 1.5 Summary of frequency effects on accuracy of response – stress identification task Participants

Arabic L1: Control English L1: 1st English L1: 2nd Chinese L1: 1st Chinese L1: 2nd

Stimuli Group 3

Stimuli Group 4

Estimate

p Value

Odds Ratio

Estimate

p Value

Odds Ratio

−0.96

p < 0.01**

0.39

−0.33

p = 0.47

0.72

−0.66

p < 0.01**

0.52

−1.3

p < 0.01**

0.28

−0.72

p = 0.12

0.5

−1.04

p = 0.1

0.36

−0.56

p < 0.05*

0.58

−1.91

p < 0.001***

0.15

0.21

p = 0.58

1.23

−1.16

p = 0.17

0.32

identification task, they were immediately prompted with a screen on the computer to determine whether the stimulus to which they had just listened is a real word or not by keypresses on the control panel. After completing each response, participants were prompted with another token for the stress identification task. The same procedure was followed until all tokens were administered. Similar to the identification task, note again that only participants’ responses to nonsense words were analyzed to avoid biases introduced from participants’ prior lexical knowledge.

1.4.3.4  Results of lexical decision task To examine participants’ responses to stimuli that were nonsense words in the lexical decision task, two-way ANOVA tests were run. No main effect for L1, proficiency, or for interaction effect on participants’ reaction time was found. However, with respect to accuracy of participants’ responses, a main effect was found for L1 (F(2,73) = 26.2, p < 0.001, η2 = 0.41), but no effect for proficiency and no interaction effect were found. Post hoc comparisons using Tukey tests showed that the Arabic native group had significantly higher accuracy than the two nonnative groups (p < 0.001), but no significant difference was found between the English L1 and Chinese L1 groups (see Figure 1.7). In other words, the results suggest that participants in the native group were significantly more successful in rejecting nonsense words than participants in the non-native groups, even though they did not respond significantly faster than the nonnative participants. On the other hand, proficiency or additional exposure to language input (i.e., for lack of within-group effect) did not seem to help the non-native groups in either becoming more successful in rejecting nonsense words or being faster in doing so. To better understand participants’ performance in the lexical decision task, participants’ responses to stimuli that were real Arabic words were also examined, since accurate responses might simply be due to participants’ constantly pressing the same key to most of the tokens. As shown in Figure 1.8, the outcome of the analysis revealed that the non-native participants’ responses were considerably more accurate for real words and less so for nonsense words, suggesting that the non-native participants pressed “yes” (i.e., that they thought a given stimulus was a real word) to most of the tokens. Compared to their non-native counterparts, the native participants were consistent in their responses to stimuli that were either real or nonsense words. Due to the biased performance of the non-native groups, participants’ responses were converted into d-prime scores to allow for a better observation of whether the participants were 22

Figure 1.7 Accuracy rate grouped by participants’ L1 and proficiency in the lexical decision task

Figure 1.8 Comparison between participants’ accuracy responses to real and non-word stimuli in the lexical decision task

Figure 1.9 A comparison of d-prime scores for participants’ accuracy responses in the lexical decision task

Cheng-Wei Lin and Mohammad T. Alhawary

indeed able to distinguish real word from nonsense words. Figure 1.9 illustrates the d-prime scores for each participant group. One-way and two-way ANOVA tests were conducted with the d-prime scores as the dependent variable. A main effect for native language (f(2,74) = 82.6, p < 0.001, η2 = 0.69) and another for proficiency (f(1,64) = 16.3, p < 0.001, η2 = 0.20) were found. Post-hoc Tukey tests found the participants in the control group had significantly higher d-prime scores than the non-native participants (p < 0.001); so did the second-year (nonnative) participants than their first-year counterparts (p < 0.001). In other words, the analysis revealed that the native control group was confident in distinguishing the real from the nonsense words. In contrast, although improvement was found as the proficiency of participants increased, the non-native participants’ accurate responses may be due to chance or constantly pressing the same key rather than to their lexical knowledge about the stimuli, as evidenced in their low d-prime scores. This is understandable given their limited proficiency levels. However, the findings of the native speakers remain credible, since they demonstrated consistent performance across real and nonsense words with their high d-prime scores in both types of stimuli. To examine the effect of frequency on participants’ reaction time, a one-way ANOVA with repeated measures was conducted for each stimuli group. Although there seems to be a trend of negative estimate value, which would suggest shorter reaction time to stimuli with frequent stress pattern than stimuli with infrequent stress pattern, these differences were not significant in nearly all stimuli groups. Even when significant differences were found, the trend was not obvious, suggesting the findings were not in support of the prediction proposed for frequency effects. As for the effect of frequency on accuracy of responses, a rather interesting pattern emerged. As shown in Table 1.6, for the stimuli in stimuli group 1, the analyses found that all groups of participants, except for the first-year Chinese L1 group, had a negative estimate value, and the effect of relative frequency on the accuracy of response was significant. Although this preference for stimuli with infrequent stress pattern was also found in other stimuli groups, it was not as strong and consistent. A negative estimate value for accuracy of response indicates that the participants responded less correctly to stimuli with frequent stress pattern than stimuli with infrequent stress pattern. In other words, the participants were significantly less successful in rejecting nonsense words with a frequent stress pattern than ones with infrequent, contrary to the prediction proposed for frequency effects.

1.4.3.5  Stress production task and procedure The third main task of the experiment is the stress production task. To avoid any unintended confounding effect resulting from task sequencing, the production task was administered first,

Table 1.6 Summary of frequency effects on accuracy of response – lexical decision Participants

Arabic L1: Control English L1: 1st English L1: 2nd Chinese L1: 1st Chinese L1: 2nd

Stimuli Group 1 Estimate

p Value

Odds Ratio

−1.35 −1.08 −0.81 −0.24 −1.72

p < 0.01** p < 0.001*** p < 0.01** p = 0.36 p < 0.001***

0.26 0.35 0.45 0.8 0.19

24

Frequency and L1 transfer effects

before the other two tasks. Its aim was to examine accuracy of learners’ production of stress, how it is influenced by L1, by participants’ proficiency, and by the frequency of the stressrelated cues of the stimuli. Participants were presented with visual stimuli on the computer screen, which were in Arabic scripts with full vocalization. For each token, the participants were prompted with a beeping sound generated by the testing program, signaling them to read out the script once as soon as they could. For every token in the task, the script for each stimulus appeared twice consecutively, which required the participants to read out each stimulus twice. All of their production was recorded in PsychoPy and analyzed in Praat.

1.4.3.6  Results of stress production task As in previous tasks, the fluency (i.e., reaction time) and accuracy of the participants’ production reaction time analyzed was determined by calculating the difference between the time tag for each trial and the onset of their production (measured in milliseconds). The accuracy of stress production was determined automatically by an algorithm that extracted and compared the three correlates; i.e., F0, duration, and intensity, for each syllable of the production of the participants. The scoring evaluated and ranked each syllable of the production based on the three correlates. The syllable with the highest scores is predicted to be the stressed syllable produced by the participants. To illustrate, consider a hypothetical production for the trisyllabic word ‫[ ِم ْل َعقة‬mɪl. ʕa.qa] “a spoon” with the values of the correlates for respective syllables presented in Table 1.7. Since the word has three syllables, the highest score that can be assigned is 3. Accordingly, the first correlate, i.e. F0, is evaluated across all syllables. Since the first syllable has the highest F0, it receives the highest score, 3, and the second syllable receives 2 and the last 1. After evaluating the remaining correlates (i.e., duration and intensity), the total score received by each syllable can be calculated. Since the first syllable receives the highest score among the three syllables, it is therefore determined to be the stressed syllable of this production. The stressed position of this production is then compared with the attested position to determine whether the production was either correct (coded as 1 in the dataset) or incorrect (coded as 0 in the dataset). It is worth noting that, since the difference between the participants’ first and second production attempts (in the production task) were in most cases not significant, only analysis of the first attempt is reported on here. The first production attempt is also arguably less rehearsed and may be more reflective of participants’ actual performance As in the two previous tasks, analyses for this task were based on participants’ responses to stimuli that were nonsense words, to control for any bias that might be introduced by individual participants’ previous familiarity with any words that might have been included in the stimuli.

Table 1.7 Correlates for a hypothetical production and scoring Correlate Value

F0 (Hz) Duration (ms) Intensity (dB)

Score

1st Syllable

2nd Syllable

3rd Syllable

1st Syllable

2nd Syllable

3rd Syllable

200 90 70

185 60 60

150 100 50 Total

3 2 3 8

2 1 2 5

1 3 1 5

25

Cheng-Wei Lin and Mohammad T. Alhawary

Subsequent two-way ANOVA tests were then run. A main effect for L1 on reaction time was found (F(2,73) = 9.23, p < 0.001, η2 = 0.20), but no effect was found for proficiency or interaction effect. Post hoc comparisons using Tukey tests showed that the native control group had a significantly shorter (p < 0.001) reaction time than both non-native groups, and the Chinese L1 groups showed significantly shorter reaction time than their English L1 counterparts (p < 0.001). Similarly, a main effect of accuracy for L1 (F(2,73) = 27.43, p < 0.001, η2 = 0.42) and proficiency (F(2,73) = 4.45, p < 0.05, η2 =0.05) were found, but there was no interaction effect. Post hoc Tukey tests further showed that both the native control group and the Chinese L1 groups scored significantly higher on accuracy than their English L1 counterparts (p < 0.001), while no group difference was found between the native and both Chinese L1 groups, and no within-group effect was found in the Chinese L1 or English L1 participants (see Figure 1.10). In other words, the results suggest that participants in both non-native groups spent more time to begin the production of stress than the native speakers, but that the Chinese L1 participants spent significantly less time doing so than their English L1 counterparts. Moreover, participants in the Chinese L1 groups performed as well as the native speakers in their accurate production of stress and again significantly outperformed their English L1 counterparts. However, proficiency or exposure time to instructional input by participants in both non-native groups was not found to significantly contribute to the fluency and accuracy of stress production. To examine the frequency effects of participants’ responses on reaction time, a one-way ANOVA with repeated measures was conducted, as was done in the previous task. The results are consistent and indicative of the main effect of relative frequency, which is presented in Table 1.8. A consistent pattern seems to emerge from participants’ responses to stimuli in stimuli groups 1–2. A trend of negative estimate value was prevalent in the two stimuli groups, with significant effect for relative frequency found in several groups of participants, showing preference for stimuli with frequent stress patterns. The trend was stronger for the responses to stimuli group 2 than stimuli group 1, as all groups had significantly shorter reaction time to stimuli with frequent stress pattern than to stimuli with infrequent stress pattern. However, the responses for stimuli group 3 had a reverse pattern. Although the trend was less strong than that found in stimuli group 2, all groups of participants had a positive estimate value, where

Figure 1.10 Accuracy of responses by participants’ L1 and proficiency in the stress production task

26

Arabic L1 English 1st English 2nd Chinese 1st Chinese 2nd

Participants

−188 29 −162 11 −24

p < 0.01** p = 0.33 p < 0.001*** p = 0.76 p = 0.53

F(1,134) = 8.98 F(1,297) = 0.94 F(1,194) = 16.21 F(1,312) = 0.1 F(1,194) = 0.39 −223 −102 −188 −128 −203

F(1,134) = 13.72 F(1,297) = 8.71 F(1,193) = 20.73 F(1,311) = 12.25 F(1,194) = 29.52

F Statistics

Estimate

p Value

Estimate

F Statistics

Stimuli Group 2

Stimuli Group 1

Trisyllabic

Table 1.8 Summary of frequency effects on reaction time – stress production

p < 0.001*** p < 0.01** p < 0.001*** p < 0.001*** p < 0.001***

p Value

35 131 25 139 180

Estimate

F(1,134) = 0.23 F(1,297) = 17.04 F(1,193) = 0.33 F(1,313) = 11.1 F(1,194) = 19.52

F Statistics

Stimuli Group 3

p = 0.63 p < 0.001*** p = 0.57 p < 0.001*** p < 0.001***

p Value

Cheng-Wei Lin and Mohammad T. Alhawary Table 1.9 Summary of frequency effects on accuracy of response – stress production Participants

Trisyllabic Stimuli Group 1

Arabic L1: Control English L1: 1st English L1: 2nd Chinese L1: 1st Chinese L1: 2nd

Estimate

p Value

Odds Ratio

0.31 0.63 0.3 0.85 1.44

p = 0.48 p < 0.05* p = 0.39 p < 0.01** p < 0.001***

1.36 1.87 1.34 2.34 4.2

the first-year English L1 group and the first and second-year Chinese L1 groups had significantly longer reaction time for stimuli with frequent stress pattern than stimuli with infrequent stress pattern. In other words, although participants were faster in responding to stimuli with frequent stress pattern in stimuli groups 1–2, they spent longer time to respond to stimuli with frequent stress pattern for the stimuli in group 3, with the latter finding going against the prediction made by frequency effects in the present study. A more consistent pattern emerged from the examination of frequency effects of the participants’ responses on accuracy, where a repeated-measures logistic regression was conducted. The results of the analyses are presented in Table 1.9. A general trend of positive estimate was exhibited for trisyllabic stimuli, especially in stimuli group 1, where the main effect of relative frequency was significant in several groups of participants. In particular, for the responses in stimuli group 1, nearly all groups of participants, except for the native group and second-year English group, produced stress significantly more correctly for stimuli with frequent stress patterns than stimuli with infrequent stress patterns. In other words, the first-year English group and both Chinese groups were more successful in placing stress on the correct syllable in their production when the stress pattern of the stimuli is frequent, supporting the prediction for frequency effects proposed in this study.

1.5 Discussion The findings of the present study revealed a number of patterns in Arabic L2 stress in perception and production by the participants of the study, which include: 1 The Chinese L1 and English L1 participants performed similarly in the stress identification task. 2 The Arabic native speakers showed significantly slower and less accurate perception of stress than the Arabic L2 learners in the stress identification task. 3 The Arabic native speakers responded more accurately in the lexical decision task. 4 No difference in reaction time between the native and the non-native participants was detected in the lexical decision task. 5 The Arabic native speakers responded more quickly than the participants in the English L1 and Chinese L1 groups in the stress production task. 6 Participants in the Chinese L1 groups responded more quickly and more accurately in the stress production task than their English L1 counterparts. 28

Frequency and L1 transfer effects

7 Proficiency did not seem to significantly contribute to learners’ performance in the experiment. 8 Frequency effects are only partially evident in the participants’ responses within certain tasks. These findings can be grouped into three clusters related to: L1 transfer effects (Research Question 1), proficiency effects (Research Question 2), and frequency effects (Research Questions 3–4). The first cluster of findings, which is related to the role of L1 transfer or the extent to which learners’ L1 could influence (positively or negatively) the participants’ perception and production of stress, is exhibited in more than one of the study tasks. Based on the findings, the English L1 and Chinese L1 learners of Arabic did not seem to differ significantly in their perception of Arabic stress. This finding is consistent with findings from previous studies on L2 stress. Studies in L2 English stress perception have indicated that, without further manipulation of the correlates of stress, L1 Chinese speakers are capable of perceiving stress and can perceive stress as well as L1 English speakers can (e.g., Wang 2008, Yu and Andruski 2011). Similarly, in a previous study which investigated the acquisition of syllable weight as a cue for Spanish stress perception, advanced L1 English-speaking learners of Spanish were found to be able to utilize syllable weight to perceive stress and could perform closer to native L1 Spanish speakers (Face 2005). However, performance of the native Arabic speakers in the control group may seem to provide counter-evidence to the role of L1 transfer through their performance in the stress identification task, in which it took them a significantly longer time to respond to the stimuli, and they were significantly less successful in identifying the stressed syllable of the stimuli than the participants in the English L1 and Chinese L1 groups, regardless of their proficiency level. The difficulty that native speakers of Arabic encounter in stress perception might be explained by the findings from studies that focus on the typology of stress perception and stress deafness. In particular, Altmann (2006) examined the perception of English L2 stress by speakers of different L1s (including Arabic) and reported that learners with predictable stress in their native language (including Arabic) were found to have problems perceiving the location of stress. Similarly, Peperkamp and Dupoux (2002) found that speakers of Finnish and Hungarian, where stress is predictable and is not phonetically contrastive, experience the socalled stress deafness that is previously reported in L1 French speakers. Thus, it may be the case that Arabic speakers encounter the same difficulty identifying stress, since stress in Arabic is predictable, a conclusion which is itself related to L1 transfer. This also seems to be supported by statements from the Arabic native-speaking participants (during short follow-up interviews when asked to reflect on their performance on the experiment) that stress was something “they have never thought about”, and they expressed uncertainty about their performance in the tasks, which could be indicative of the general lack of attention to and awareness of stress, resulting in the type of performance which they exhibited. Another finding which demonstrates presence of L1 transfer can be detected in the performance of the Chinese L1 participants in the production task. The Chinese L1 participants seem to be more accurate in stressing the correct syllable and to do so more quickly than their English L1 counterparts. In addition, their production is not significantly different from that of the native speakers. The reason for this advantage of the Chinese L1 learners over their English L1 counterparts may be related to how they utilize stress correlates in their L1. In a study comparing the production of English stress by L1 English speakers and that of L1 Mandarin Chinese speakers, Lai (2008) found that L1 English speakers did not fully utilize all three correlates of

29

Cheng-Wei Lin and Mohammad T. Alhawary

stress (i.e., F0, intensity, and duration) to signify stress as the L1 Chinese speakers did. In particular, whereas the L1 English speakers did not utilize F0 when producing stress for verbs in English, the L1 Chinese speakers used all three stress cues for verbs and nouns. This lack of using F0 could be the case for the English L1 participants’ production of stress in Arabic as well, which makes their production disfavored by the algorithm that evaluates participants’ production of stress. The second cluster of findings, which is related to proficiency effects, is observed in all three tasks. Significant effect of proficiency was not found across all three tasks of the experiment. Again, intuitively, one would expect learners with a higher level of proficiency (or more exposure to input) to perform better in the target language, either with higher accuracy or better fluency. In fact, improved performance in both accuracy and fluency was detected in the participants’ responses as their proficiency level increased; however, the exhibited improvement did not reach statistical significance. Another possible explanation for the lack of significant proficiency effects on participants’ performance might be due to the proficiency level of the participants themselves. Both the first-year and second-year participants were recruited during the middle of their second semester, a level which may not be sufficient enough to develop a robust working knowledge of the language through which lexical learning becomes more robust (see Nation 2001; see Also Alhawary 2013 for a similar explanation of acquiring Arabic L2 vocabulary). To explore proficiency effects on Arabic L2 stress perception and production more effectively and reliably, future research should recruit Arabic L2 learners from higher levels. The third cluster of findings, which relates to frequency effects, is evident only partially in the experiment results. Overall, the present study did not find a global effect for frequency to influence all aspects of stress perception and production, but rather local or partial effects. With certain stress patterns, the effect of frequency could be evident, while for some other stress patterns, the results were sometimes mixed or even clearly contradicted the predictions. Whereas no frequency effect was found in the stress identification task, frequency effects seem to be at play in the stress production task, as participants were able to produce stress more rapidly and accurately when the stress pattern of the stimuli was frequent, especially when the stimuli were short, as significant differences in performance were often found within the trisyllabic stimuli groups and less so in the quadrisyllabic stimuli groups. Moreover, significant differences in performance that support the prediction made by frequency effects in the study are more often found within stimuli belonging to stimuli group 1. This may be due to the likelihood that the frequency contrast in stimuli group 1 is the largest, as the relatively frequent stress pattern in the stimuli group [ˈCV.CV.CV] is more frequent in terms of the frequency of stress pattern and stress position than its relatively infrequent counterpart [CVC. ˈCVC.CVC]. Thus, it is argued here that the degree of contrast in frequency as well as the length of the stress pattern seem to inform the extent to which frequency informs the perception and production of stress, which would undoubtedly require more empirical data from future studies to validate. Finally, although it is less relevant to the focus of the present study, the experiment results also showed that frequency seems to inform the accuracy of responses in the lexical decision task. Precisely, the relative frequency of the stress pattern seems to bias the participants’ ability to reject nonsense words, as the participants were found to be less willing to reject nonsense words with frequent stress patterns than ones with infrequent stress patterns. It seems to be the case that a nonsense word would appear to be more like a real word to the participants when it has a frequent stress pattern, which, in turn, would impede the participants’ willingness to reject the word as non-real.

30

Frequency and L1 transfer effects

1.6  Conclusion and pedagogical implications The study findings seem to only partially support frequency effect predictions, as no global effect for frequency on stress perception and production can be found. The findings indicate that the effect of frequency is evident only when the contrast in frequency is sufficiently large and the input is short in length. On the other hand, the native language of the learners is found to be a better indicator of learners’ perception and production of stress, as the characteristics of the prosodic system in the learners’ native language seem to continue to be evident in the learners’ perception and production of Arabic L2 stress, resulting in difference in performance between learners of the two different language backgrounds: English L1 and Chinese L1. Nevertheless, the effect of native language alone cannot account for the variability found in learners’ perception and production of stress. There are a number of limitations which should be taken into account in future related research. First, the dataset implemented to estimate the frequency distribution of stress-related cues involves only minimal inflectional morphology. As in many other dictionaries, various forms of a verb, noun, or adjective could not be exhaustively listed due to the length of the dictionary. Estimating frequency using only dictionary entries may not accurately reflect actual language and stress distribution, which, in turn, might have introduced biases toward our estimate of stress in this study. Second, inclusion of first- and second-year L2 learners of Arabic is not sufficient enough to test proficiency effects as well as frequency effects robustly and reliably. Future research should aim to recruit more participants per level as well as participants belonging to higher proficiency levels. Third, the binary categorization of frequency may be too minimal to adequately reflect the actual difference between one stress pattern and another. Rather than relying on raw frequency, it may be more reflective of how a stress pattern is positioned among other stress patterns to rely instead on frequency percentile or a normalized frequency index. Finally, other frequency cues, such as segmental cues for both vowels and consonants, should also be considered in designing the stimuli. As mentioned earlier, a confound in vowel length (which resulted in one fewer stress pattern out of the eight identified patterns) could have been rectified if segmental cues had been considered as the variable of the stimuli, which would have allowed more space for comparison. It will also be useful for future research examining frequency effects on Arabic L2 stress perception and production to incorporate other relevant frequency cues, in addition to the ones identified in the present study. Additionally, future studies should expand the pool of learners to include those who are speakers of L1s which have other distinctive prosodic systems, such as French and Japanese, which exhibit an invariable stress system and pitch-contour system, respectively. Denser data of naturally produced speech by L2 learners of Arabic will be useful in order to understand how stress is realized in context rather than in isolation. Notwithstanding the limitations mentioned, the present study does reveal some aspects relevant to pedagogical implications related to Arabic L2 stress perception and production. The general observation resulting from interacting with the study participants was similar to what Ryding (2013) noted, that lexical stress in Arabic is a topic that is seldom mentioned and rarely explicitly activated in the Arabic foreign language classrooms. This lack of awareness and activation of stress helps explain the overall performance of the non-native participants in the study experiment (especially the lack of any significant improvement from the first to second year of language instruction), which, in turn, calls for more efforts from the instructor and textbook writer to integrate more content, drills, and activities aiming at developing the ability to perceive and produce stress from early on. 31

Cheng-Wei Lin and Mohammad T. Alhawary

Another pedagogical implication from the study stems from the learners’ general performance on longer words. Recall, the study found that the participants generally performed better when the word was short but worse when the word was longer. From a usage-based perspective, and due to availability of words in the input, the better performance with short words is because stress in trisyllabic words is much better-entrenched in the minds of the learners than that of quadrisyllabic. This observation implies that in order for the learners to perform better in stress perception and production, they need more input containing quadrisyllabic words introduced from early on. Therefore, textbook writers and instructors should not refrain from including longer words just because the learners’ proficiency level is assumed to be too low, as long as inclusion of such words is done judiciously and such words are relevant to the content and functions being introduced. Finally, stress is one among many aspects of L2 phonology which need to receive adequate instructional attention in the foreign language classroom, especially, but not solely, at the early stages of learning. Yet giving due attention to stress alone is not a trivial matter, especially when one wishes ambitiously to teach 1–2 dialects or more side by side with MSA simultaneously, and especially with such dialects (e.g., Damascene and Cairene) whose stress systems are not the same. Recall, participants of the present study had exposure to one variety only (MSA). Teaching L2 phonology or pronunciation (as one sub-component part of the language) is not only about covering mere units of ideas but necessarily entails assigning sufficient drills and activities as well as making targeted forms sufficiently recycled in the input, which is the main premise of any rule-based learning, let alone frequency learning. As noted in Alhawary (2013), incorporating more items in the instructional input should translate into more instruction time in the classroom and more input related to such items available, which, in turn, should be subsequently sufficiently recycled. Hence the age-old practice of grading the language learning syllabus so that students get things right (without certain aspects getting compromised) and learning Arabic by starting initially with one register may prove to be not without foresight.

Notes 1 The study did not include a list or sample of the stimuli, and, therefore, it was not possible to evaluate the stimuli. 2 The Arabic curriculum used by the L1 English participants was that of the Al-Kitaab series (Brustad et al. 2004, 2007). In addition, they used the Hans Wehr dictionary as the main resource for dictionary and meaning look-up use. As for the L1 Chinese participants, they used more than a single set of textbooks, since their teachers had more freedom using and supplementing their instructional materials. Thus, the L1 Chinese participants used three textbooks: Al-Kitaab textbooks (Brustad et al. 2004, 2007), Al-‘Arabiyya bayna yadayk (Al-Fawzan et al., 2014), and Al-jadiid fi Al-lugha Al-‘Arabiyya (Guo and Jiang 2002). In addition, and similar to the L1 English participants, they used the Hans Wehr dictionary as the main resource for dictionary and meaning look-up use. Although frequency analysis and information of the frequency dictionary (Buckwalter and Parkinson 2014) were not compared with those of the two other textbooks of the L1 Chinese participants (besides that of Al-Kitaab), the fact that the Hans Wehr dictionary was assigned as the main dictionary suggests that the vocabulary and language distribution to which the L1 Chinese participants are exposed are similar to those used by the L1 English participants. 3 The reading task consisted of 48 isolated, real words. Speed reading of the participants in this task was examined. There was no statistically significant difference between the English L1 and Chinese L1 groups, but the two groups differed significantly from that of the control group, where the participants in the control group expectedly read the words at a significantly higher speed than their nonnative counterparts did. Also expectedly, an effect for proficiency was found, where the second-year non-native groups differed significantly from their first-year counterparts, indicating the former read significantly faster than the latter.

32

3 3 3 3 4 4 4 4

3 3 3 3 3 3 3 3

ˈCV.CV.CV ˈCV.CV.CVC CV.ˈCVV.CV CVC.ˈCVC.CVC CV.CV.ˈCVV.CVC CVC.ˈCV.CV.CV CVC.ˈCV.CV.CVC CVC.CV.ˈCVV.CV

ˈCV.CV.CV ˈCV.CV.CVC CV.ˈCVV.CV CVC.ˈCVC.CVC CV.CV.ˈCVV.CVC CVC.ˈCV.CV.CV CVC.ˈCV.CV.CVC CVC.CV.ˈCVV.CV

ˈʕa.ma.sa ˈʕa.ma.sat ʕi.ˈma:.sa mus.ˈtaʕ.mas mu.ta.ˈʕa:.mis ʔiʕ.ˈta.ma.sa ʔiʕ.ˈta.ma.sat ʔiʕ.ti.ˈma:.sa ˈma.θa.fa ˈma.θa.fat mi.ˈθa:.fa mus.ˈtam.θaf mu.ta.ˈma:.θif ʔim.ˈta.θa.fa ʔim.ˈta.θa.fat ʔim.ti.ˈθa:.fa

‫س‬ َ ‫َع َم‬ ‫َع َم َست‬ ‫اس‬ َ ‫ِع َم‬ ْ‫ُم ْستَع َمس‬ ‫ُمتَ َعا ِمس‬ ‫س‬ َ ‫اِعتَ َم‬ ْ ‫اِعتَ َم َس‬ ‫ت‬ ‫اِعتِ َما َسة‬

َ‫َمثَف‬ ‫َمثَفَت‬ َ‫ِمثَاف‬ ْ َ‫ُم ْستَمث‬ ‫ف‬ ‫ُمتَ َماثِف‬ َ‫اِمتَثَف‬ ْ َ‫اِمتَثَف‬ ‫ت‬ ‫اِمتِثَافَة‬ mθf mθf mθf mθf mθf mθf mθf mθf

ʕms ʕms ʕms ʕms ʕms ʕms ʕms ʕms ˈC1a.C2a.C3a ˈC1a.C2a.C3at C1i.ˈC2a:.C3a mus.ˈtaC1.C2aC3 mu.ta.ˈC1a:.C2iC3 ʔiC1.ˈta.C2a.C3a ʔiC1.ˈta.C2a.C3at ʔiC1.ti.ˈC2a:.C3a

ˈC1a.C2a.C3a ˈC1a.C2a.C3at C1i.ˈC2a:.C3a mus.ˈtaC1.C2aC3 mu.ta.ˈC1a:.C2iC3 ʔiC1.ˈta.C2a.C3a ʔiC1.ˈta.C2a.C3at ʔiC1.ti.ˈC2a:.C3a ‫فَقَ َد‬ ‫َعقَدَت‬ ‫قِيادَة‬ ‫ُم ْستَ ْهدَف‬ ‫ُمتَسائِل‬ ‫ق‬ َ َ‫اِ ْنطَل‬ ‫اِ ْشتَ َرطَت‬ ‫ضة‬ َ ‫اِ ْنتِفا‬

‫َس ِم َع‬ ‫َس َم َح‬ ‫ثَالثَة‬ ‫يَ ْستَ ْمتِع‬ ‫يَتَنا َول‬ ‫اِ ْشتَ َغ َل‬ ‫اِ ْلتَ َحقَت‬ ‫اِ ْستِطاعة‬

3 3 3 3 4 4 4 4

ˈC1a.C2a.C3a ˈC1a.C2a.C3at C1i.ˈC2a:.C3a mus.ˈtaC1.C2aC3 mu.ta.ˈC1a:.C2iC3 ʔiC1.ˈta.C2a.C3a ʔiC1.ˈta.C2a.C3at ʔiC1.ti.ˈC2a:.C3a

2 2 2 2 2 2 2 2

fqz fqz fqz fqz fqz fqz fqz fqz

‫ق‬ َ ِ‫َوث‬ ‫َج َعلَت‬ ‫نَتِي َجة‬ ‫ُم ْستَوْ َرد‬ ‫َتال‬ ٍ ‫ُمت‬ ‫اِ ْعتَ َم َد‬ ‫اِرْتَ َكزَت‬ ‫َجاب‬ َ ‫اِ ْست‬

ˈfa.qa.za ˈfa.qa.zat fi.ˈqa:.za mus.ˈtaf.qaz mu.ta.ˈfa:.qiz ʔif.ˈta.qa.za ʔif.ˈta.qa.zat ʔif.ti.ˈqa:.za

‫فَقَ َز‬ ‫فَقَزَت‬ ‫فِقَا َز‬ ‫ُم ْستَفقَ ْز‬ ‫ُمتَفَاقِز‬ ‫اِفتَقَ َز‬ ْ ‫اِفتَقَز‬ ‫َت‬ ‫اِفتِقَازَة‬

3 3 3 3 4 4 4 4

1 1 1 1 1 1 1 1

ˈCV.CV.CV ˈCV.CV.CVC CV.ˈCVV.CV CVC.ˈCVC.CVC CV.CV.ˈCVV.CVC CVC.ˈCV.CV.CV CVC.ˈCV.CV.CVC CVC.CV.ˈCVV.CV

Real Word

Nonsen Wd Pronunciation Root Template

# Block # Syl Stress Pattern

ˈfa.qa.da ˈʕa.qa.dat qa.ˈdʒa:.da mus.ˈtah.daf mu.ta.ˈsa:.ʔil ʔitˤ.ˈta.la.qa ʔiʃ.ˈta.ra.tˤat ʔin.ti.ˈfa:.dˤa

ˈsa.mi.ʕa ˈsa.ma.ħat θa.ˈla:.θa jas.ˈtam.tiʕ ja.ta.ˈna:.wal ʔiʃ.ˈta.ɣa.la ʔil.ˈta.ħa.qat ʔis.ti.ˈtˤa:.ʕa

ˈwa.θa.qa ˈdʒa.ʕa.lat na.ˈti:.dʒa mus.ˈtaw.rad mu.ta.ˈta:.lin ʔiʕ.ˈta.ma.da ʔir.ˈta.ka.zat ʔis.ta.ˈdʒa:.ba

+ (6.2) + (6.2) − (5.9) − (5.9) + (5.3) + (5.3) − (5.0) − (5.0)

+ (6.2) + (6.2) − (5.9) − (5.9) + (5.3) + (5.3) − (5.0) − (5.0)

+ (5.9) − (3.2) + (5.8) − (4.0) + (5.0) − (2.6) + (4.7) − (3.8) + (5.9) − (3.2) + (5.8) − (4.0) + (5.0) − (2.6) + (4.7) − (3.8)

+ (6.2) + (6.2) − (5.9) − (5.9) + (5.3) + (5.3) − (5.0) − (5.0)

+ (5.9) − (3.2) + (5.8) − (4.0) + (5.0) − (2.6) + (4.7) − (3.8)

+ (5.9) + (5.9) + (6.1) + (6.1) + (5.0) + (5.0) + (5.5) + (5.5)

+ (5.9) + (5.9) + (6.1) + (6.1) + (5.0) + (5.0) + (5.5) + (5.5)

+ (5.9) + (5.9) + (6.1) + (6.1) + (5.0) + (5.0) + (5.5) + (5.5)

(Continued)

+ (3.6) + (3.6) + (3.3) + (2.5) + (2.5) + (2.5) + (3.3) + (3.3)

+ (3.6) + (3.6) + (3.3) + (2.5) + (2.5) + (2.5) + (3.3) + (3.3)

+ (3.6) + (3.6) + (3.3) + (2.5) + (2.5) + (2.5) + (3.3) + (3.3)

Pronunciation Frequency of Frequency of Frequency Conditional Stress Pattern Stress Position of Syllabic Probability Structure

Nonsense words derived from unattested root and template:

Appendix: Stimuli

ˈCV.CV.CV ˈCV.CV.CVC CV.ˈCVV.CV

‫ق‬ َ َ‫َسه‬ ‫َرأَلَت‬ ‫إِعاغَة‬

3 3 3

6 6 6

ˈsa.ha.qa ˈra.ʔa.lat ʔi.ˈʕa:.ɣa

ˈba.da.ʔa ˈʕa.li.mat ħu.ˈku:.ma jas.ˈtaq.bil ja.ta.ˈna:.fas ʔix.ˈta.la.fa ʔin.ˈta.ðˤa.rat ʔis.ti.ˈdˤa:.fa

‫ق‬ َ َ‫َسب‬ ‫َسأ َلَت‬ ‫إِعادَة‬

ˈsa.ba.qa ˈsa.ʔa.lat ʔi.ˈʕa:.da

ˈna.ʃa.ra ˈtˤa.la.ʕat xi.ˈla:.la mus.ˈtax.dam mu.ta.ˈðˤa:.hir ʔik.ˈta.ʃa.fa ʔin.ˈda.fa.ʕat ʔis.ti.ˈqa:.la

‫نَ َش َر‬ ْ ‫طَلَ َع‬ ‫ت‬ ‫ِخال َل‬ ‫ُم ْست َْخدَم‬ ‫ُمتَظا ِهر‬ َ‫اِ ْكتَ َشف‬ ‫اِ ْن َدلَ َعت‬ ‫اِ ْستِقالَة‬ n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a.

Pronunciation

Template Real Word

Minimal Pair n.a. Minimal Pair n.a. Minimal Pair n.a.

Minimal Pair Minimal Pair Minimal Pair Minimal Pair Minimal Pair Minimal Pair Minimal Pair Minimal Pair

‫نَ َغ َر‬ ْ ‫طَثَ َع‬ ‫ت‬ ‫ِخال َز‬ ‫ُم ْست َْخثَم‬ ‫ُمتَظانِر‬ ‫اِ ْكتَ َش َز‬ ‫اِ ْن َزلَ َعت‬ ‫اِ ْستِقاخَة‬

3 3 3 3 4 4 4 4

5 5 5 5 5 5 5 5

ˈna.ɣa.ra ˈtˤa.θa.ʕat xi.ˈla:.za mus.ˈtax.θam mu.ta.ˈðˤa:.nir ʔik.ˈta.ʃa.za ʔin.ˈza.la.ʕat ʔis.ti.ˈqa:.xa

Nonsen Wd Pronunciation Root

ˈCV.CV.CV ˈCV.CV.CVC CV.ˈCVV.CV CVC.ˈCVC.CVC CV.CV.ˈCVV.CVC CVC.ˈCV.CV.CV CVC.ˈCV.CV.CVC CVC.CV.ˈCVV.CV

ˈC1a.C2a.C3a ˈC1a.C2a.C3at C1i.ˈC2a:.C3a mus.ˈtaC1.C2aC3 mu.ta.ˈC1a:.C2iC3 ʔiC1.ˈta.C2a.C3a ʔiC1.ˈta.C2a.C3at ʔiC1.ti.ˈC2a:.C3a

+ (6.2) + (6.2) − (5.9) − (5.9) + (5.3) + (5.3) − (5.0) − (5.0)

+ (5.9) + (5.9) + (6.1) + (6.1) + (5.0) + (5.0) + (5.5) + (5.5)

+ (3.6) + (3.6) + (3.3) + (2.5) + (2.5) + (2.5) + (3.3) + (3.3)

+ (6.2) + (6.2) − (5.9) − (5.9) + (5.3) + (5.3) − (5.0) − (5.0) + (6.2) + (6.2) − (5.9)

+ (5.9) − (3.2) + (5.8) − (4.0) + (5.0) − (2.6) + (4.7) − (3.8) + (5.9) − (3.2) + (5.8)

+ (5.9) + (5.9) + (6.1)

+ (5.9) + (5.9) + (6.1) + (6.1) + (5.0) + (5.0) + (5.5) + (5.5)

+ (3.6) + (3.6) + (3.3)

+ (3.6) + (3.6) + (3.3) + (2.5) + (2.5) + (2.5) + (3.3) + (3.3)

Frequency of Frequency of Frequency Conditional Stress Pattern Stress Position of Syllabic Probability Structure

+ (5.9) − (3.2) + (5.8) − (4.0) + (5.0) − (2.6) + (4.7) − (3.8)

Pronunciation Frequency of Frequency of Frequency Conditional Stress Pattern Stress Position of Syllabic Probability Structure

Nonsense words derived from real words:

# Block # Syl Stress Pattern

() = Log frequency

qsˤx qsˤx qsˤx qsˤx qsˤx qsˤx qsˤx qsˤx

َ‫بَدَأ‬ ‫َعلِ َمت‬ ‫ُح ُكو َمة‬ ‫يَستَقبِل‬ ‫يَتَنافَس‬ َ‫اِ ْختَلَف‬ ‫اِنتَظَ َرت‬ ‫اِ ْستِضافَة‬

ˈqa.sˤa.xa ˈqa.sˤa.xat qi.ˈsˤa:.xa mus.ˈtaq.sˤax mu.ta.ˈqa:.sˤix ʔiq.ˈta.sˤa.xa ʔiq.ˈta.sˤa.xat ʔiq.ti.ˈsˤa:.xa

‫ص َخ‬ َ َ‫ق‬ ‫صخَت‬ َ َ‫ق‬ ‫صا َخ‬ َ ِ‫ق‬ ‫ص ْخ‬ َ ‫ُم ْستَق‬ ‫اصخ‬ ِ َ‫ُمتَق‬ ‫ص َخ‬ َ َ‫اِقت‬ ْ ‫صخ‬ ‫َت‬ َ َ‫اِقت‬ ‫صاخَة‬ َ ِ‫اِقت‬

ˈCV.CV.CV ˈCV.CV.CVC CV.ˈCVV.CV CVC.ˈCVC.CVC CV.CV.ˈCVV.CVC CVC.ˈCV.CV.CV CVC.ˈCV.CV.CVC CVC.CV.ˈCVV.CV

4 4 4 4 4 4 4 4

3 3 3 3 4 4 4 4

Real Word

Nonsen Wd Pronunciation Root Template

# Block # Syl Stress Pattern

3 3 3 3 4 4 4 4

3 3 3 3 4 4 4 4

7 7 7 7 7 7 7 7

8 8 8 8 8 8 8 8

ˈCV.CV.CV ˈCV.CV.CVC CV.ˈCVV.CV CVC.ˈCVC.CVC CV.CV.ˈCVV.CVC CVC.ˈCV.CV.CV CVC.ˈCV.CV.CVC CVC.CV.ˈCVV.CV

ˈCV.CV.CV ˈCV.CV.CVC CV.ˈCVV.CV CVC.ˈCVC.CVC CV.CV.ˈCVV.CVC CVC.ˈCV.CV.CV CVC.ˈCV.CV.CVC CVC.CV.ˈCVV.CV

CVC.ˈCVC.CVC CV.CV.ˈCVV.CVC CVC.ˈCV.CV.CV CVC.ˈCV.CV.CVC CVC.CV.ˈCVV.CV

() = Log frequency

3 4 4 4 4

6 6 6 6 6

mus.ˈtal.sam mu.ta.ˈwa:.ɣiʕ ʔir.ˈta.ka.ħa ʔim.ˈta.ba.ʔat ʔis.ti.ˈdʒa:.ða ˈʕa.ʃa.fa ˈka.ta.sat ma.ˈði:.na mus.ˈtaq.zal ja.ta.ˈba:.dʒal ʔis.ˈta.xa.ʕa ʔiʕ.ˈta.ba.ðat ʔis.ti.ˈra:.θa ˈwa.θa.da ˈla.ra.dʒat hu.ˈfa:.ka jas.ˈtaʕ.kil mu.ta.ˈdˤa:.jiθ ʔiʕ.ˈta.ʃa.da ʔif.ˈta.qa.θat ʔis.ta.ˈdˤa:.sa

‫ُم ْست َْل َسم‬ ‫ُمتَوا ِغع‬ ‫اِرْتَ َك َح‬ ‫اِ ْمتَبَأ َت‬ ‫اِ ْستِجا َذة‬

َ‫َع َشف‬ ‫َكتَ َست‬ ‫َم ِذينَة‬ ‫ُم ْستَ ْقزَل‬ ‫يَتَبا َجل‬ ‫اِ ْستَ َخ َع‬ ‫اِعتَبَفَت‬ ‫اِ ْستِراثة‬

‫َوثَ َد‬ ‫لَ َر َجت‬ ‫ك‬ َ ‫هُفا‬ ‫يَستَع ِكل‬ ْ ِ‫ُمتَضاي‬ ‫ث‬ ‫اِ ْعتَ َش َد‬ ‫اِفتَقَثَت‬ ‫َضاس‬ ‫اِست‬ َ Minimal Pair Minimal Pair Minimal Pair Minimal Pair Minimal Pair Minimal Pair Minimal Pair Minimal Pair

Minimal Pair Minimal Pair Minimal Pair Minimal Pair Minimal Pair Minimal Pair Minimal Pair Minimal Pair

Minimal Pair Minimal Pair Minimal Pair Minimal Pair Minimal Pair

n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a.

n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a.

n.a. n.a. n.a. n.a. n.a.

‫َو َج َد‬ ‫َخ َر َجت‬ ‫ك‬ َ ‫هُنا‬ ‫يَستَع ِمل‬ ‫ُمتَضايِق‬ ‫اِ ْعتَقَ َد‬ ‫اِفتَقَ َرت‬ َ‫اِستَضاف‬

َ‫َعرَف‬ ‫َكتَبَت‬ ‫َم ِدينَة‬ ‫ُم ْستَ ْقبَل‬ ‫يَتَبادَل‬ ‫اِ ْستَ َم َع‬ ‫اِعتَبَ َرت‬ ‫اِ ْستِراحة‬

‫ُم ْست َْلزَم‬ ‫َواضع‬ ِ ‫ُمت‬ ‫َب‬ َ ‫اِرْتَك‬ ‫اِ ْمت ََلَت‬ ‫اِ ْستِجابَة‬

ˈwa.dʒa.da ˈxa.ra.dʒat hu.ˈna:.ka jas.ˈtaʕ.mil mu.ta.ˈdˤa:.jiq ʔiʕ.ˈta.qa.da ʔif.ˈta.qa.rat ʔis.ta.ˈdˤa:.fa

ˈʕa.ra.fa ˈka.ta.bat ma.ˈdi:.na mus.ˈtaq.bal ja.ta.ˈba:.dal ʔis.ˈta.ma.ʕa ʔiʕ.ˈta.ba.rat ʔis.ti.ˈra:.ħa

mus.ˈtal.zam mu.ta.ˈwa:.dˤiʕ ʔir.ˈta.ka.ba ʔim.ˈta.la.ʔat ʔis.ti.ˈdʒa:.ba

+ (5.9) − (3.2) + (5.8) − (4.0) + (5.0) − (2.6) + (4.7) − (3.8)

+ (5.9) − (3.2) + (5.8) − (4.0) + (5.0) − (2.6) + (4.7) − (3.8)

− (4.0) + (5.0) − (2.6) + (4.7) − (3.8)

+ (6.2) + (6.2) − (5.9) − (5.9) + (5.3) + (5.3) − (5.0) − (5.0)

+ (6.2) + (6.2) − (5.9) − (5.9) + (5.3) + (5.3) − (5.0) − (5.0)

− (5.9) + (5.3) + (5.3) − (5.0) − (5.0)

+ (5.9) + (5.9) + (6.1) + (6.1) + (5.0) + (5.0) + (5.5) + (5.5)

+ (5.9) + (5.9) + (6.1) + (6.1) + (5.0) + (5.0) + (5.5) + (5.5)

+ (6.1) + (5.0) + (5.0) + (5.5) + (5.5)

+ (3.6) + (3.6) + (3.3) + (2.5) + (2.5) + (2.5) + (3.3) + (3.3)

+ (3.6) + (3.6) + (3.3) + (2.5) + (2.5) + (2.5) + (3.3) + (3.3)

+ (2.5) + (2.5) + (2.5) + (3.3) + (3.3)

Cheng-Wei Lin and Mohammad T. Alhawary

References Alfano, I, et al., 2007. The perception of Italian and Spanish lexical stress: A  first cross-linguistic study. Proceedings of the 16th international congress of phonetic sciences. Saarbrücken, Germany. 1793–1796. Alfano, I., et al., 2009. Cross-language speech perception: Lexical stress in Spanish with Italian and Francophone subjects. In: S. Schmid et al., eds. AISV 2009. La dimensione temporale del parlato. Atti del 5oconvegno nazionale AISV – Associazione Italiana di Scienze della Voce. Università di Zurigo, 455–474. Al-Fawzān, D. A. R., et. al., 2014. Al-‘Arabiyya bayna yadayk. 2nd ed. Riyadh, Saudi Arabia: Arabic for All. Alhawary, M. T., 2013. Arabic second language acquisition research and second language teaching: What the teacher, textbook writer, and tester need to know. Al-‘Arabiyya, 46, 23–35. Almbark, R., et al., 2014. Acquiring the phonetics and phonology of English word Stress: comparing learners from different L1 backgrounds. Proceedings of the international symposium on the acquisition of second language speech, Concordia Working Papers in Applied Linguistics, 5, 19–35. Altmann, H., 2006. The perception and production of second language stress: A crosslinguistic experimental study. Thesis (PhD). University of Delaware. Anani, M., 1989. Incorrect stress placement in the case of Arab learners of English. International Review of Applied Linguistics in Language Teaching, 27 (1), 15–22. Angoujard, J.-P., 1990. Metrical structure of Arabic. De Gruyter Mouton. Archibald, J, 1993a. Language learnability and L2 phonology: The acquisition of metrical parameters. Dordrecht and Boston: Kluwer. ———, 1993b. Metrical phonology and the acquisition of L2 stress. In: F. R. Eckman, ed. Confluence: Linguistics, l2 acquisition and speech pathology. Amsterdam: John Benjamins, 37–48. ———, 1997. The acquisition of English stress by speakers of nonaccentual languages: Lexical storage versus computation of stress. Linguistics, 35 (1), 167–181. Bouchhioua, N., 2008 The role of vowel quality in cuing stress and accent in Tunisian Arabic, native English, and l2 English. Speech prosody 2008: Fourth conference on speech prosody, 539–542. Brustad, K., et al., 2004. Al-kitaab fii ta‘allum al-‘Arabiyya, part one & part two, second edition. Washington, DC: Georgetown University Press. ———, 2007. Al-kitaab fii ta‘allum al-‘Arabiyya, part three. Washington, DC: Georgetown University Press. Buckwalter, T. and Parkinson, D., 2014. A frequency dictionary of Arabic: Core vocabulary for learners. New York: Routledge. Bullock, B. and Lord, G., 2003. Analogy as a learning tool in second language acquisition: The case of Spanish stress. Amsterdam Studies in the Theory and History of Linguistic Science Series, 4, 281–298. Chen, Y., et al., 2001. A study of sentence stress production in Mandarin speakers of American English. The Journal of the Acoustical Society of America, 109 (4), 1681–1690. Cowell, M. W., 1964. A reference grammar of Syrian Arabic: Based on the dialect of Damascus. Washington, DC: Georgetown University Press. de Jong, K. and Zawaydeh, B. A., 1999. Stress, duration, and intonation in Arabic word-level prosody. Journal of Phonetics, 27 (1), 3–22. de Jong, K. and Zawaydeh, B. A., 2002. Comparing stress, lexical focus, and segmental focus: Patterns of variation in Arabic vowel duration. Journal of Phonetics, 30 (1), 53–75. Dupoux, E., et al., 1997. A destressing “deafness” in French? Journal of Memory and Language, 36 (3), 406–421. Dupoux, E., et al., 2001. A robust method to study stress “deafness”. The Journal of the Acoustical Society of America, 110 (3), 1606–1618. Dupoux, E., et al., 2008. Persistent stress ‘deafness’: The case of French learners of Spanish. Cognition, 106 (2), 682–706. Ellis, N. C., 2002. Frequency effects in language processing. Studies in Second Language Acquisition, 24 (2), 143–188. ———, 2012. Frequency-based accounts of second language acquisition. In S. Gass and A. Mackey, eds. The Routledge handbook of second language acquisition. New York: Routledge, 193–210 ———, et al., 2014. Second language verb-argument constructions are sensitive to form, function, frequency, contingency, and prototypicality. Linguistic Approaches to Bilingualism, 4 (4), 405–431.

36

Frequency and L1 transfer effects Face, T. L., 2000. The role of syllable weight in the perception of Spanish stress. In: H. Campos et al., eds. Hispanic linguistics at the turn of the millennium. Somerville, MA: Cascadilla, 1–13. Face, T. L., 2005. Syllable weight and the perception of Spanish stress placement by second language learners. Journal of Language and Learning, 3, 90–103. Garcia, G. D., 2016. Extrametricality and second language acquisition. In: G. Ó. Hansson, et al., eds. Proceedings of the Annual Meetings on Phonology, 3, 1–12. Gómez, R. L. and Gerken, L., 2000. Infant artificial language learning and language acquisition. Trends in Cognitive Sciences, 4 (5), 178–186. Goodsitt, J. V., et al., 1993. Perceptual strategies in prelingual speech segmentation. Journal of Child Language, 20 (2), 229–252. Greenberg, J. H., 1950. The patterning of root morphemes in Semitic. Word, 6 (2), 162–181. Guma, F., 2003. Perception and production of syllable structure and stress by adult Libyan Arabic speaker acquiring English in the UK. Thesis (Ph.D). Durham University. Guo, S. and Jiang, C. Y. 2002. al-Jadīd fī al-lugha al-‘Arabiyya. Beijing, China: Foreign Language Teaching and Research Press. Heliel, M., 1982. Stress timing in modern literary Arabic. Al-‘Arabiyya, 15 (1), 90–107. Jangjamras, J., 2011. Perception and production of English lexical stress by Thai speakers. Thesis (Ph.D). University of Florida Kenstowicz, M., 1983. Parametric variation and accent in the Arabic dialects. In: A. Chukerman, et al., eds. ePapers from the Regional Meeting. Chicago Linguistic Society Chicago, 19, 205–213. Lai, Y., 2008. Acoustic realization and perception of English lexical stress by Mandarin learners. Thesis (PhD). University of Kansas. Lin, C.-W., 2018. The perception and production of Arabic lexical stress by learners of Arabic: A usagebased account. Thesis (PhD). The University of Michigan. McCarthy, J. J. and Prince, A. S. 1990. Foot and word in prosodic morphology: The Arabic broken plural. Natural Language and Linguistic Theory, 8 (2), 209–283. Mitchell, T. F., 1975. Principles of Firthian linguistics. London: Longman. Nation, I. S., 2001. Learning vocabulary in another language. Cambridge: Cambridge University Press. Nguyen, T. A. T., 2004. Prosodic transfer: The tonal constraints of Vietnamese acquisition of English stress and rhythm. Thesis (PhD). The University of Queensland. Peperkamp, S. and Dupoux, E., 2002. A typological study of stress “deafness.” Laboratory phonology, 7, 203–240. Peperkamp, S., et al., 1999. Perception of stress by French, Spanish, and bilingual subjects. Proceedings of.Eurospeech, 99, 2683–2686. Peperkamp, S., et al., 2010. Perception of predictable stress: A cross-linguistic investigation. Journal of Phonetics, 38 (3), 422–430. Post Da Silveira, A., 2011. Frequency as (dis)advantage to word stress acquisition. Proceedings of the 17th international conference of phonetic sciences. ICPhS, 1634–1637. Romberg, A. R. and Saffran, J. R., 2010. Statistical learning and language acquisition. Wiley Interdisciplinary Reviews: Cognitive Science, 1 (6), 906–914. Ryding, K. C., 2013. Teaching and learning Arabic as a foreign language: A guide for teachers. Washington, DC: Georgetown University Press. Sheikh, I. a. M., 1987. The interlanguage stress phonology of Arab learners of English: influence of universals as well as dialect, distance, and task (psychotypology, proficiency). Thesis (PhD). Indiana University of Pennsylvania. Smith, L. and Yu, C., 2008. Infants rapidly learn word-referent mappings via cross-situational statistics. Cognition, 106 (3), 1558–1568. Ueyama, M., 2000. Prosodic transfer: An Acoustic study of L2 English vs. L2 Japanese. Thesis (PhD). University of California, Los Angeles. Wang, Q., 2008. L2 Stress Perception: the reliance on different acoustic cues. Proceedings of the speech prosody 2008 Conference. Compinas: Brazil, 635–638 Wang, Q. and Yoon, T.-J., 2008. The transfer of L1 acoustic cues in the perception of L2 lexical stress. Canadian Acoustics, 36 (3), 126–127. Wehr, H., 1993. Arabic-English dictionary: The Hans Wehr dictionary of modern written Arabic. 4th ed. Urbana, IL: Spoken Language Services. Yu, V. Y. and Andruski, J. E., 2011. The effect of language experience on perception of stress typicality in English nouns and verbs. The Mental Lexicon, 6 (2), 275–301.

37

2 PRODUCTION OF MODERN STANDARD ARABIC LEXICAL STRESS CUES BY NATIVE SPEAKERS OF AMERICAN ENGLISH Mashael Al-Aloula The production of stress patterns in a second language (L2) can be affected by a speaker’s native language (L1) in at least two ways: the placement of stress within a word and the phonetic cues used to realize prosodic contrasts. Even when the native language and the L2 share many prosodic characteristics, non-native speakers often exhibit differences from native speakers in prosodic patterning. The influence of the native language on production of lexical stress in a second language has been extensively investigated in recent literature on second language acquisition phonology. Many studies have examined prosodic transfer in first languages (L1), with English as the L2, with evidence for L1 prosodic transfer in the production of L2 stress cues. However, few studies have examined prosodic L1 transfer with American English (AE) as the L1 and Modern Standard Arabic (MSA) as the L2. The present study investigated the production of lexical stress in L2 MSA by native speakers of AE. Both languages, AE and MSA, use stress to signal prominence at the word level but also differ in prosodic patterning. Lexical stress plays an important role in native speakers’ perceptions and processing of speech. In this study, eight native speakers of AE (intermediate learners of MSA) and a control group of four native speakers of Arabic participated in an experiment. They produced 12 MSA near-minimal pairs in which stress patterns shifted across syllables. Measures of duration (ms), pitch (Hz), and intensity (dB) indicated that these acoustic cues were used to different degrees by native and non-native speakers to signal stress; specifically, the L2 learners used duration to a significantly greater degree than the native speakers of Arabic. These results are interpreted as an effect of prosodic transfer from the speakers’ native language, American English, on lexical stress patterns in the production of their L2 MSA.

2.1 Introduction The chapter investigates the effects of prosodic transfer across two languages: Standard American English (SAE) as a native language (L1) and Modern Standard Arabic (MSA) as a foreign language (L2). Studies on non-native English learners have shown that issues in acquiring English lexical stress vary depending on a speaker’s L1 (Archibald 1998; Ueyama 2000; Lee 38

Production of lexical stress cues

et al. 2006). Considering the prosodic differences between SAE and MSA, the chapter predicts a prosodic transfer effect from L1 (SAE) in the production of MSA (L2) lexical stress by native speakers of SAE. The cues of lexical stress in SAE are generally considered to be duration, pitch, and intensity, with a considerable contrast in vowel duration (Fry 1955; Lindblom 1963; Beckman 1986). On the other hand, the production of lexical stress by native speakers of Arabic does not exhibit the contrast in duration and pitch in stressed and unstressed vowels compared with native English speakers (Zuraiq and Sereno 2007). Thus, a prosodic transfer effect from L1 (SAE) in the production of MSA lexical stress by native speakers of SAE can be predicted. The chapter examines the production of L2 learners to detect any differences in the production of SAE leaners of MSA which may be the result of L1 (SAE) prosodic transfer (Fry 1955; Beckman 1986; House et al. 1987; Zuraiq and Sereno 2007). In addition, the chapter examines whether there are any characteristics notably different from native speakers of Najdi Arabic (NA) in the use of acoustic cues to realize lexical stress by non-native speakers.

2.1.1  L2 Prosody production Much of the research on the L2 acquisition of prosody is associated with the notions of “pronunciation errors” and “foreign accent” (Rasier and Hiligsmann 2007). However, the literature on L2 foreign accent does not provide a widely accepted definition of an L2 foreign accent (Pennington 2014). According to Munro (2008, p. 193), “the occurrence of foreign accents provides some of the clearest evidence that knowledge of a first language (L1) influences the acquisition of a second (L2).” At the suprasegmental level, the prosodic properties of L2 speech contribute to foreign accent detection by the noting of non-native rhythmic or intonation patterns by native listeners (Grover et al. 1987; Munro 1995; Shah 2003). The acoustic values of an L2 production may be based on parameters of the speaker’s L1.The shared conclusion reached by studies examining the acoustic output on adult L2 learners is that their foreign accents retain features of their native language (Bohn and Flege 1992; Ioup 2008) Previous research on the production of L2 prosody suggested that prosodic errors have more of a negative effect on listeners’ judgment of L2 speech comprehensibility than that of segmental errors. For example, Anderson-Hsieh et al. (1992, p. 536) examined 11 different language groups and found that overall prosody, as opposed to segmental or syllable structure errors, had a greater influence on pronunciation ratings for the standardized spoken language test (SPEAK). The score for prosody was most significantly associated with the overall score for pronunciation. Anderson-Hsieh et al.’s (1992) results provide evidence that suprasegmentals play a more important role than segmentals when it comes to L2 inteliigibility. The importance of prosodic features in L2 production on L2 comprehensibility is attested in the literature. Derwing et al. (1998) studied the effects of both segmental and suprasegmental instruction on learners’ comprehensibility ratings and concluded that the latter had a greater effect on performance in communicative contexts. In addition, Munro and Derwing (1999) used an intelligibility measure that asked 18 native speakers (NSs) of English to transcribe and rate on a Likert scale the degree of foreign accentedness and comprehensibility of excerpts of L2 English speech produced by 10 Mandarin NSs and two English NSs. Munro and Derwing (1999) found that comprehensibility of L2 speech by NSs is more significantly related to prosodic variables than segmental effects. Moreover, Trofimovich and Baker (2006) used acoustic analyses and listener judgments to detriment the degree of foreign accent in L2 English speech of NSs of Korean. They reported that suprasegmental features (stress timing, peak alignment, speech rate, pause frequency, and pause duration) contributed to listeners’ judgment of foreign accent at all levels of experience (Trofimovich and Baker 2006). 39

Mashael Al-Aloula

2.1.2  L1 Prosodic transfer (stress placement) The influence of L1 on the suprasegmental features, particularly stress placement, of the target language was examined in several empirical studies. The findings reported that the production of lexical stress in L2 is one area of difficulty for language learners. It is argued that adult language learners already have a prosodic system in place for their L1, so they have to discover what is and what is not consistent with their L1 system in order to make the necessary adjustments in their grammars. This task is not as straightforward as it sounds, and very often the differences between L1 and L2 manifest themselves in word stress errors in L2, thereby contributing to a detectable degree of foreign accent (Cooper et al. 2002; Lukyanchenko et al. 2010). Moreover, previous studies have shown that correct word stress may be more important for the comprehensibility of non-native speech than grammatical accuracy (Jung 2010). Studies indicate that incorrect stress patterns in an L2 are mainly due to transfer from learners’ L1 (Maczuga 2014). Archibald (1998) reported that L1 stress patterns might transfer to L2. In his study, the placement of stress by adult NSs of Polish living in an English-speaking country exhibited regularities that could be attributed to L1 transfer of stress patterns. In addition, Archibald (1998) reported a similar finding with adult NSs of Spanish who exhibited syllable structure errors of L2 English when they were asked to produce real English words. He proposed that the errors were due to the transfer of L1 Spanish stress rules (Archibald 1998). Moreover, Guion et al. (2004) reported that L2 learners (early Spanish-English and late Spanish-English bilinguals) have difficulties producing the appropriate L2 lexical stress when asked to produce bisllyabic English non-words. Both Polish and Spanish learners of English tend to apply their L1 stress assignment rules when performing a stress placement task with spoken L2 English words (Archibald 1998, Guion et al. 2004). The manipulation of the components of the L2’s prosodic cues is difficult for L2 learners (Rasier and Hiligsmann 2007). The literature on L1 prosodic transfer generally suggests that L2 or non-native speakers (NNs) transfer the acoustic cues that indicate stress from their first language. That is, NNs with different first languages were more likely to produce words with stress cues that correspond to the pattern of acoustic correlates of their L1. Since duration, pitch and intensity are used to cue stress in English (Fry 1955; Lieberman 1960; Beckman 1986), the current chapter examines the phonetic realization of these stress cues in speakers’ L2 MSA. As it was illustrated, the importance of the production of such suprasegmental features as stress and vowel quality is extensively examined in the literature, with a specific focus on English as an L2. However, the prosodic features of Modern Standard Arabic (MSA) as an L2 are not examined in the literature. Therefore, the present study aims to examine the production of a single prosodic element, which is MSA’s lexical stress by native speakers of SAE.

2.2  The phonological systems of MSA and SAE MSA and SAE phonological systems vary extensively, not only in the range of the sounds each language has but also in the relative importance of vowels, syllables, and suprasegmental features (El-Hassan 1994). Vowel quality in MSA is limited and straightforward. It has three different vowel qualities, each with a short and long variant [i] and [ii], [a] and [aa], [u] and [uu] (Odisho 2005), and two diphthongs [ay, aw]. It is worth noting that the difference in vowel length in MSA is not a difference in vowel quality but in duration (Holes 2004; Odisho 2005; Ryding 2005; Ryding 2014). 40

Production of lexical stress cues

On the other hand, vowels of SAE include the front vowels [i, ɪ, eɪ, ɛ, æ], the back vowels [u, ʊ, oʊ, ɔ, ɑ], the central vowels [ɝ, ə, ʌ], and the diphthongs [ɔɪ, aɪ, aʊ] (Reetz and Jongman 2011). The differences between the vowel systems of MSA and SAE is noticeable in the acoustic realization of stress cues.

2.2.1  Prosodic features of MSA and SAE From the point of view of prosody, both MSA and SAE are stressed-timed languages (Abercombie 1967). However, one of the most noticeable differences between the two languages is in stress and syllable length. While in SAE unstressed vowels are typically reduced to schwa [ə], in MSA the reduction is limited. This section provides more information on stress placement in MSA and SAE by highlighting the differences and similarities between the two languages.

2.2.1.1  Stress in MSA Even though stress placement in MSA is predictable, lexical stress patterns of MSA are not fixed to a certain position within a word but determined for the most part by syllable weight (Peperkamp and Dupoux 2002; Guion 2005). According to McCarthy and Prince (1990) (cited Ryding 2014, p. 36), “The Arabic system is obviously weight-sensitive: final syllables are stressed if superheavy CVVC or CVCC; penults are stressed if heavy CVV or CVC; otherwise the antepenult is stressed.” In MSA, no more than one heavy syllable may occur per word (Holes, 2004). Moreover, the majority of lexical stress placement in MSA is predictable and determined by the syllable structure (El-Hassan 1994; Holes 2004; Al-Thamery & Ibrahim 2005; Ryding 2005). More evidence of the predictability of stress placement in MSA can be drawn from the lack of a diacritic to mark the stressed syllable in MSA orthography. MSA’s orthography makes use of a diacritic called šaddah for the geminated sound, and no similar diacritic is used for stress (Al-Jarrah 2002). According to Al-Jarrah (2002), the lack of such marking may be due to the fact that lexical stress in Arabic is predictable and therefore non-phonemic. This proposition is supported by the findings of Larudee (1973). In his study, eight NSs of Cairene Arabic were asked to read a list of ten novel words. The result verified the NSs ability to follow a predictable pattern in stress placement even with novel words (Larudee 1973). According to Larudee (1973, p. 23), a NS of Arabic “has internalized a stress system for his language which he employs in novel situations.” This proposal has been extended to the area of second language acquisition to explain why Arabic learners of English fail to both produce and perceive the English stress pattern correctly. It is often stated that there are no native speakers of MSA; any native speaker of Arabic is a native speaker of his/her Arabic dialect (Badawi 1996; Kaye 2001). The degree of variation in lexical stress between MSA and other Arabic dialects depends on the differences between the syllable structure of the dialect and MSA (Watson 2002). In dialectal Arabic, the preferred syllable structure of the dialect is the main factor that affects lexical stress (Holes 2004; Watson 2002). According to Holes (2004, p. 80), “stress plays a minor role, and the dialectal rules that express it are in fact remarkably similar to each other across vast geographical areas, it is the input to them that is different.” In addition, MSA is the variety of Arabic most widely used in education, print media, official documents, correspondence, and as a liturgical language in the Middle East where it is listed as the official language (Holes 2004). More importantly, native speakers are usually exposed to MSA input from birth and well within the critical period window for first language acquisition. 41

Mashael Al-Aloula

The control group of the present study comprises NSs of the Najdi dialect. According to Ingham (1994), the Najdi syllable structure is identical to MSA syllable structure, except for the following three cases. Unlike MSA, the Najdi syllable structure allows an initial consonant cluster, as in ktibat “she wrote” (Abushihab 2010; Alghmaiz 2013; Ingham 1994). Second, in Najdi Arabic (NA) the long vowel [a:] is added before a consonant cluster within the word. Ingham (1994, p. 20) pointed out that such a difference in the Najdi syllable structure is only possible when the sequence [CC] “is composed of two radicals of the root,” as in kaatbah “having written feminine.” Finally, the addition of the sequence [CCC] is also possible in NA in restricted environments, as in the verb taktbiin “you singular feminine write” (Ingham 1994, p. 20). Furthermore, it is worth noting that none of the stimuli used in this chapter exhibit any of the three previous unique syllable structures that are unique to NA. Finally, lexical stress in NA resembles lexical stress in MSA. MSA exhibits a trochaic lexical stress “in its treatment of peripheral feet” (Watson 2002, p. 11). Therefore, the shared syllable structure and lexical stress pattern in MSA and NA motivates the use of NS of NA participants in the control group.

2.2.1.2  Stress in SAE In SAE, lexical stress is positively correlated with vowel duration where there is a noticeable reduction of vowels in the unstressed syllables (Lindblom 1963; Klatt 1976; Adams 1979; Zuraiq and Sereno 2007). Such prosodic differences create difficulties in intelligibility for learners who are trying to acquire the language (Field 2005). Both MSA and English are classified as stress-timed languages (Abercombie 1967). However, lexical stress in SAE is less predictable than MSA. This is due to the suggested possibility that that SAE is a language with lexically specified stress occurring in various locations depending on the word stress (Leyden and van Heuven 1996; Peperkamp and Doupoux 2002; Guion 2005). Moreover, SAE is considered to be a more stress-timed language than MSA in the literature, due to the greater degree of reduction in vowel quality and duration between stressed and unstressed syllables in SAE (Fry, 1958; Lindblom 1963; Klatt 1976; Van Summers 1987; de Jong and Zawaydeh 1999; De Jong and Zawaydeh 2002; Zuraiq and Sereno 2007). Such a drastic reduction and vowel neutralization is not observed in unstressed syllables in MSA, in which the vowel quality is often maintained in the unstressed syllable (El-Hassan 1994; Odisho 2005; Zuraiq and Sereno 2007).

2.2.2  Empirical studies on the production of lexical stress cues in Arabic L1 and SAE L2 Previous experimental studies on the production of lexical stress cues in Arabic and English examined the production of native speakers of Arabic of L2 English. Munro (1993) examined the productions of 10 English vowels in /bVt/ and /bVd/ contexts. These were elicited from a group of adult Arabic learners of English and NSs of English. Munro (1993) reported that new vowel sounds and those that are similar to L1 vowels were found to be different, even when produced by learners with several years of L2 experience. Acoustic analysis of the NNs pronunciation included vowel durations, F1 and F2 frequencies, and movement in F1 and F2, which revealed that their accent was the result of both temporal (formant movement) and spectral properties. Vowel length also varied significantly from native English pronunciation. In general, the vowels produced by the learners were shorter than equivalent vowels produced by native speakers and vowel length was uniformly presented before voiced and voiceless consonants (Munro 1993). These non-native features of their production could be attributed to the characteristics of the Arabic vowel system. 42

Production of lexical stress cues

Other empirical studies reported similar results. The results demonstrated the inability of L1 Arabic learners of English to reduce vowels in the unstressed English syllable to the same degree as native SAE speakers (Anani 1989; Youssef and Mazurkewich 1998; Zuraiq and Sereno 2007). Thus, NSs of Arabic transfer their L1 prosodic cue by not reducing vowels to indicate lexical stress (Anani 1989; El-Hassan 1994; Munro 1993; Youssef and Mazurkewich 1998; Zuraiq and Sereno 2007). As a result, Arabic learners of English fail to reduce the vowel to [ə], which is considered to be one of the most salient features of vowel dynamics in English (Odisho 2005). Examining some of the suprasegmental features of L2 MSA on word-level stress by L1 SAE speakers will shed some light on the type of errors produced by the learners and on whether their production of lexical stress cues is similar to that of NSs of Arabic. The review of the literature predicts that NSs of SAE will exhibit greater vowel contrast between stressed and unstressed MSA syllables more than NSs of Arabic. Moreover, both NSs and NNs are expected to produce vowels with higher pitch in stressed syllables than unstressed syllables. However, the NSs of Arabic are expected to produce greater difference in their use of intensity in stressed vowels than the non-native speakers. This predication is motivated by the findings of Alrajeh (2011), in which intensity was the most reliable cue of lexical stress, after duration, by native speakers of Arabic. Such a finding will support the L1 prosodic transfer that is predicted to take place in the production of SAE learners of MSA.

2.3 Methods This chapter examines the production of L2 MSA lexical stress cues by native speakers of SAE. The acoustic cues of stress are examined in MSA nominal paradigms in which stress is contrastive, in stressed and unstressed syllables. The study uses singular and dual MSA nouns. In MSA, dual is formed by adding the dual suffix [aan] to the singular noun stem (Holes 2004). Therefore, the study focuses on the following three cues of stress: vowel duration, pitch and intensity in the production of MSA words. The three stress cues, duration, pitch, and intensity, were chosen because they are considered to be the general cues of stress in English (Fry 1955; Morrill 2012). Moreover, they are considered to be perceptually salient acoustic cues of stress among native speakers of any language. Duration, pitch, and intensity are all measurable acoustic correlates of stress; thus, an examination and a comparison of the three cues of stress can provide a clearer picture of the production of SAE learners of MSA. The comparison of the measurement of the three stress cues will help us determine if and what type of L1 prosodic transfer of stress (duration, pitch, and intensity) takes place in the production of SAE learners of MSA.

2.3.1 Participants The participants in the experimental group were eight adult SAE learners of MSA (seven females and one male), ages 19 to 28 (with age mean of 20). The participants were recruited from an Arabic program at a university on the East Coast. The SAE learners of MSA received an average of 3.5 semesters of MSA instruction (with three credit hours per semester, which brings it to a total of 10 credit hours of Arabic instruction received). The self-reported language proficiency levels indicated that all of the participants considered themselves to be at an intermediate level of proficiency in MSA. All of the participants in the experimental group were native speakers of SAE, and none had any known speech disorders. The majority of the speakers did not have any significant knowledge of any other languages aside from their native language. 43

Mashael Al-Aloula

In order to ensure that the participants in the experimental group (L1 SAE, L2 MSA) met the selection criteria to be included in the experiment, they were asked to fill out a brief language background questionnaire (see Appendix A). The language background questionnaire provided more biographical information about the participants and their language background and proficiency level in MSA. The participants in the control group were four adult native speakers of Najdi Arabic (three males and one female), ages 24 to 33 (with age mean of 21). As was previously mentioned, MSA is considered the macrolanguage of all Arabic varieties spoken in the Arabic world (Holes 2004), and both MSA and NA share a very similar syllable structure that determines the placement of lexical stress. In addition, the participants’ high exposure and use of MSA took place before they reached the end of the critical period of language acquisition, with an average of 13 years of MSA instruction. All of the participants in the control group were graduate students who received their undergraduate degree from an Arabic-speaking university. Therefore, it is safe to claim that the participants reached a native-like level of proficiency in MSA. The participants in the control group were instructed to read the stimuli in MSA. None of the participants in the control group had any known speech disorders. Moreover, the critical items in the stimuli provide a similar stress pattern in both NA and MSA.

2.3.2 Stimuli The stimuli were six pairs of words (see Table 2.1). The chosen pairs have comparable stress patterns. The stimuli consisted of singular and dual MSA nouns; the words differ in stress placement. In MSA singular nouns, primary stress falls on the first syllable [ˈCVVˌCVC]. In the dual nouns, primary stress falls on the final syllable [ˌCVV CV ˈCVVC] and the secondary stress falls on the first syllable. The set of stimuli provides the Arabic open front unrounded vowel [a] in the three positions (stressed and unstressed), as follows: 1 Primary stress in singular nouns [ˈCVVˌCVC] [ˈṭaalib] “a student” 2 Secondary stress in dual nouns [ˌCVVCVˈCVVC] [ˌṭaaliˈbaan] “two students” 3 Primary stress in dual nouns [ˌCVVCVˈCVVC] [ˌṭaaliˈbaan] “two students” Table 2.1 The six pairs of singular and dual MSA nouns used in the production experiment Singular

Transcription

Dual

Transcription

‫طالب‬ “a student”

[ˈṭaalib]

‫طالبان‬ “two students”

[ˌṭaaliˈbaan]

‫كاتب‬ “a writer” ‫باحث‬ “a researcher” ‫عالم‬ “a scientist” ‫والد‬ “a parent”

[ˈkaatib]

‫كاتبان‬ “two writers” ‫باحثان‬ “two researchers” ‫عالمان‬ “two scientists” ‫والدان‬ “two parents”

[ˌkaatiˈbaan]

‫ضابط‬ “an officer”

[ˈḍaabiṭ]

‫ضابطان‬ “two officers”

[ˌḍaabiˈṭaan]

[ˈbaaħiθ] [ˈʕaalim] [ˈwaalid]

44

[ˌbaaħiˈθaan] [ˌʕaaliˈmaan] [ˌwaaliˈdaan]

Production of lexical stress cues

To avoid unfamiliar items in the stimuli that may affect the participants’ production, all of the words used in the carrier phrase were taken from the MSA textbook (Brustad et al. 2004) which was used in the Arabic program which the participants were recruited, (Brustad et al. 2004). In addition, the instructor of the Arabic courses in the Arabic program was consulted about the participants’ knowledge of the stimuli. Infrequent and unfamiliar words were excluded from the final list of stimuli and fillers. In order to ensure an accurate measurement of the target syllable (stressed and unstressed), the final syllable preceding the target item was unstressed and ended with the voiced alveolar plosive [d]. The target items appeared in the middle of the carrier phrase that consisted of three words: a verb, subject, and object (VSO). The target items were in the subject position of the carrier phrase. The target item was always preceded by the same verb wažad “he found” and object ʔal-kitaab “the book” in the carrier phrase. This syntactic structure was used to maintain the contrast in stress with the same vowel [a] and to elicit normal production of the carrier phrase. The 12 critical items in Table 2.1 were divided into two sets. The first set, singular, was created to measure the stressed vowel [a] in bisllyabic words which occurs in the first syllable [ˈCVVˌCVC] ˈṭaalib “a student.” The second set, dual, measures the same vowel in both the stressed and unstressed syllables in the trisyllabic words [ˌCVVCVˈCVVC] ˌṭaaliˈbaan “two students.” As mentioned earlier, in MSA, stress in disyllabic words falls on the longer syllable (penultimate) [ˈCVVC ˌCVC]. However, in trisyllabic words that contain two long syllables the last syllable (ultima) receives primary stress and the first syllable receives secondary stress (Holes 2004). The two sets are designed to contrast the production of the Arabic open front vowel [a] in the singular and dual forms of MSA words in three positions. The choice of stimuli was made due to the lack of MSA minimal pairs in which the stress of the same vowel is contrastive. In addition, this approach allows the measurement of the target vowel in the same syllable where it is stressed and unstressed. The six sets of sentences were interspersed with 12 filler sentences. The critical items and fillers were randomized in four different lists (see Appendix B for the stimuli used).

2.3.3 Procedure Participants were seated in a sound-attenuated booth in an acoustics lab at the participants’ linguistics program. The participants were given a randomized list of stimuli and filler sentences prior to the actual recording. They had a chance to read the sentences out loud before the actual recording to ensure a normal production of the critical items. The participants were not told of the specific purposes of the experiment but were asked to read the sentences twice at normal speaking rate. Any items containing reading errors were eliminated, and the participant was asked to repeat the entire sentence. The second attempt was chosen for the analysis. The second reading was chosen to ensure a natural production of the target items. Some of the participants used a slow reading rate when they first produced the target items. However, in the second reading their rate was at a normal level. Recordings were made using a Zoom H2 Handy Portable Stereo Recorder. The data was recorded in a stereo file that was converted into a mono file in Praat. The recordings were digitized (sampling rate of 22 kHz) and analyzed using Praat (Boersma and Weenink 2015). The critical items were located using both visual and auditory information.

2.3.4 Measurements Using Praat, the intervals were labeled for word duration, and the primary and secondary stressed syllables in the word for both test items (singular and dual words). The examined 45

Mashael Al-Aloula

vowels are referred to as “secondary” (secondary stress) and “primary” (primary stress) for both of the tested items (singular and dual words). Measurements of duration (ms) included the whole vowel interval in both positions of stress (primary and secondary) for both items (singular and dual words) separately. The ratio of the primary stressed vowel to the secondary stressed vowel was calculated for each item to control for variation in speaker and speaking rate. Measurements of mean pitch (Hz) and intensity (dB) of the labeled intervals were conducted automatically with a Praat script.

2.4 Results To test whether cues of stress in the examined vowel are affected by the learners’ L1 prosody, linear mixed-effects regression models were fitted to the data in R statistical programming language with the lme4 package (Bates et al. 2015). Vowel length ratio, mean pitch, and intensity were the dependent variables. Position of stress (primary vs. secondary) and group (experimental L2 Arabic vs. control L1 Arabic) were entered into the models as fixed effects with an interaction term. The participants and the test items were run as random effects with random intercepts. A separate linear mixed-effects model was constructed for each dependent variable. The statistical analyses of duration, pitch, and intensity were performed separately on “secondary” (secondary stress) and “primary” (primary stress) stressed syllables in both test items. In the presentation of the statistical analyses here, significant findings are presented first, followed by other relevant results.

2.4.1 Duration In order to control for the variability between the words and the speaking rate, the proportional length of the first vowel to the second vowel provides a “within item” control. Thus, the results will help us determine if there is an actual change in duration between the primary and secondary stress positions across words. The proportional length of the second vowel to the first vowel was calculated by dividing the duration of the vowel in which stress is primary (V2) by the duration of the vowel in which stress is secondary (V1) (Figure 2.1). For example, in the dual noun [ˌCVVCVˈCVVC] ˌṭaaliˈbaan “two students,” the duration of the last vowel (primary stress) was divided by the duration of the first (secondary stress). To investigate the effects of stress on vowel duration, a linear mixed-effects regression model (the full model) was fitted with position of stress (primary vs. secondary) and groups (Native vs. L2) as fixed effects, as well as an interaction term for stress and groups, with subjects and test items as random effects. Another linear mixed-effects regression model was constructed with only position of stress (primary vs. secondary) and the interaction term for stress as fixed effects, with subjects and test items as random effects. The two mixed-effects models were compared using a likelihood ratio test to evaluate the effect of group (Native (Arabic) vs. L2 (AE)) on duration. Group emerged as a significant predictor of duration of primary stress. The comparison of the two models reveals that group (Native vs. L2) contributed significantly to model fit (χ2 (2) = 27.729, p < .001). L2 speakers produced significantly longer duration ratios than native speakers in primary stress in dual nouns (β = −0.010, SE = 0.2052). L2 speakers produced significantly longer vowels in their production of primary stress as opposed to secondary stress. The effects of stress (primary vs. secondary) on duration ratio for both groups (Native vs. L2) are shown in Figure 2.1. To investigate the effect of item type, a third linear mixed-effects model was constructed with only groups (Native vs. L2) as fixed effects and the interaction term for stress as fixed 46

NS L2

0 1 2 3 4 5 6

Duration Ratio

Production of lexical stress cues

L2.1

NS.1

L2.2

NS.2

Groups

Figure 2.1 Dual words duration ratio V2/V1 of secondary (L2.1 and NS.1) to primary stress V2/V1 (L2.2 and NS.2) for native speakers (L1 Arabic) and L2 learners of MSA (L1 AE)

effects, with subjects and test items as random effects. Using the likelihood ratio test, the comparison between the full model and the model without position of stress as fixed effect revealed that position of stress contributed significantly to model fit (χ2 (2) = 67.44, p < .001). Primary stress in dual words has longer duration ratios than primary stress in singular words (β = 1.60, SE = 0.10). This finding might be the result of the process of adding inflectional morphemes to create the dual form in the Arabic items. The duration ratio of singular words (V2/V1) revealed that both groups make a distinction in duration between primary and secondary stress. However, the distinction in duration ratios between the vowels in singular words between the two groups was not significant (χ2 (1) = 0.1501, p = 0.6984). The duration of the primary stress in singular words between the Native and L2 speakers was not significant (β = −1.116e − 02, SE = 2.834e − 02). To investigate the effect of the interaction between position of stress (primary vs. secondary) and groups (Native vs. L2), a fourth linear mixed-effects model was created in the same manner as the full model but without the interaction term between position of stress and group. Comparison between the full model and the model without an interaction term using the likelihood ratio test reveals that the interaction between position of stress (primary vs. secondary) and groups (Native vs. L2) contribute significantly to model fit (χ2 (1) = 32.38, p < .001). In the case of primary stress in dual words, L2 speakers produced significantly longer duration ratios than native speakers (β = −0.40, SE = 0.07), as shown in Figure 2.1.

2.4.2 Pitch To investigate the effects of stress on vowel pitch, a linear mixed-effects regression model (the full model) was fitted with position of stress (primary vs. secondary) and groups (Native vs. L2) as fixed effects, as well as an interaction term for stress and groups, with subjects and test items as random effects. Another linear mixed-effects regression model was constructed with only position of stress (primary vs. secondary) and the interaction term for stress as fixed effects, with subjects and test items as random effects. The two mixed-effects models were compared using a likelihood ratio test to evaluate the effect of group (Native vs. L2) on pitch. The comparison of the two models reveals that group (Native vs. L2) did not contribute significantly to model fit (χ2 (2)= 0.88, p = 0.64). The results suggest that L2 speakers did not produce significantly different pitch values than native speakers (β = −0.080, SE = 0.086). The effects of stress (primary vs. secondary) on pitch for both groups (Native vs. L2) are shown in Figure 2.2. 47

400 200

NS L2

0

Pitch (Hz)

Mashael Al-Aloula

L2.1

NS.1

L2.2

NS.2

Groups

Figure 2.2 Mean pitch of secondary (L2.1 and NS.1) and primary stress (L2.2 and NS.2) for native speakers (L1 Arabic) and L2 learners of MSA (L1 AE)

To investigate the effect of item type, a third linear mixed-effects model was constructed with only group (Native vs. L2) as a fixed effect and the interaction term for stress as a fixed effect, with subjects and test items as random effects. Using the Likelihood Ratio Test, the comparison between the full model and the model without position of stress as fixed effect revealed that position of stress contributed significantly to model fit, with both groups (NSs and SAE) producing higher pitch values for dual words (χ2 (2) = 8.25, p =0.01). Pitch values are higher in dual words (β = 0.201, SE = 0.065). To investigate the effect of the interaction between position of stress (primary vs. secondary) and groups (Native vs. L2), a fourth linear mixed-effects model was created in the same manner as the full model but without the interaction term between position of stress and group. Comparison between the full model and the model without an interaction term using the likelihood ratio test reveals that the interaction between position of stress (primary vs. secondary) and groups (Native vs. L2) did not contribute significantly to model fit (χ2 (1) = 0.88, p = 0.34). Both groups produce similar pitch values for primary and secondary stress (β = −0.080, SE = 0.086).

2.4.3 Intensity To investigate the effects of stress on vowel intensity, a linear mixed-effects regression model (the full model) was fitted with position of stress (primary vs. secondary) and groups (Native vs. L2) as fixed effects, as well as an interaction term for stress and groups, with subjects and test items as random effects. Another linear mixed-effects regression model was constructed with only position of stress (primary vs. secondary) and the interaction term for stress as fixed effects, with subjects and test items as random effects. The two mixed-effects models were compared using a likelihood ratio test to evaluate the effect of group (Native vs. L2) on pitch. The comparison of the two models reveals that group (Native vs. L2) did not contribute to model fit (χ2 (2) = 0.88, p = 0.64). The results suggest that L2 speakers did not produce significantly different intensity values than native speakers (β = −0.122, SE = 0.138). The effects of stress (primary vs. secondary) on intensity for both groups (Native vs. L2) are shown in Figure 2.3. To investigate the effect of item type, a third linear mixed-effects model was constructed with only groups (Native vs. L2) and the interaction term for stress as fixed effects, with subjects and test items as random effects. Using the likelihood ratio test, the comparison between 48

Production of lexical stress cues

NS L2

50 60 70 80 90

Intensity (dB)

the full model and the model without position of stress as fixed effect revealed that position of stress contributed significantly to model fit (χ2 (2) = 27.72, p < .001). Primary stress in singular words was produced with greater intensity than primary stress in singular words (β = −0.029, SE = 0.013). The native speakers produced greater intensity values in singular words than SAE speakers. Figure 2.4 presents intensity (dB) of primary stress in both groups (native and L2). To investigate the effect of the interaction between position of stress (primary vs. secondary) and groups (Native vs. L2), a fourth linear mixed-effects model was created in the same manner as the full model but without the interaction term between position of stress and group. Comparison between the full model and the model without an interaction term using the likelihood ratio test reveals that the interaction between position of stress (primary vs. secondary) and groups (Native vs. L2) did contribute significantly to model fit (χ2 (2) = 27.7, p < .001). Intensity values are greater in primary stress for both items (singular and dual word); however, L2 speakers produced significantly less intensity values than native speakers (β = 0.078, SE = 0.014) as shown in Figure 2.4. This difference in the production of NS of intensity to cue lexical stress is consistent with what was reported in previous studies (Alrajeh 2011; Zuraiq and Sereno 2007). In Zuraiq and Sereno (2007), native speakers of Jordanian Arabic used intensity but not duration to cue lexical stress in their native language and not in their production of English L2.

L2.1

NS.1

L2.2

NS.2

Groups

60

70

NS L2

50

Intensity (dB)

80

Figure 2.3 Intensity of secondary (L2.1 and NS.1) and primary stress (L2.2 and NS.2) of dual words for native speakers (L1 Arabic) and L2 learners of MSA (L1 AE)

NS

L2 Groups

Figure 2.4 Intensity of primary stress in singular words for native speakers (L1 Arabic) L2 learners of MSA (L1 AE)

49

Mashael Al-Aloula

2.5 Discussion The results of the present study indicated that duration was a significant cue of lexical stress for L2 speakers. The L2 speakers’ use of vowel duration in dual MSA nouns showed significant duration differences between stressed and unstressed syllables. Such a difference in the production of duration to cue stress was not found in the production of the native Najdi speakers. Thus, the difference in the production of the L2 speakers supports the prediction of an L1 prosodic transfer in MSA lexical stress cues. SAE learners of MSA produce a greater degree of duration contrast in their production of stressed and unstressed syllables to cue stress, which is most likely motivated by the importance of duration contrast to cue lexical stress in SAE. (Lindblom 1963; Klatt 1976; Van Summers 1987; de Jong and Zawaydeh 1999; De Jong and Zawaydeh 2002). The results of this study revealed that pitch and intensity in stressed and unstressed vowels are produced by SAE learners of MSA in a way that is similar to NSs of NA to cue lexical stress. However, a significant difference in the production of intensity to cue lexical stress was reported in singular words for NSs. NSs of NA use intensity cues to a greater extent than L2 speakers in singular nouns. The results revealed a significant effect of stress on intensity in singular nouns for NSs of NA. As it was mentioned previously, such a difference can be explained by the reported significant use of intensity by NS of Arabic to cue lexical stress between stressed and unstressed syllables (Zuraiq and Sereno 2007). The lack of a statistically significant effect of mean pitch can be the result of the following factors. The critical items used in this study to contrast stress are morphologically related. In MSA (singular vs. dual) is generated by the addition of the dual suffix [an]. According to Flege and Bohn (1989), stress placement in words with inflectional suffixes is much more predictable than stress placement in monomorphemic words or those involving derivational affixes. Thus, one can argue that the L2 learners did not face much difficulty with the production of the native-like stress cues because of the less complicated nature of derivational suffixes in MSA. The lack of distinction in the L2 speakers’ production of pitch and intensity between stressed and unstressed vowels may be caused by the stronger correlation between pitch and intensity. A higher pitch tends to be accompanied by higher intensity (Kidder 2008). Therefore, this may explain their ability to produce a clear distinction in duration but not in pitch and intensity.

2.6  Conclusion and pedagogical implications The results of this study provide empirical evidence of the transfer of L1 (SAE) lexical stress cues in the production of MSA (L2) lexical stress. Out of the three cues of lexical stress examined in this study (duration, pitch, and intensity), duration emerged as an important acoustic cue of lexical stress in the L2 speakers’ production. Consistent with previous studies, duration in English is one of the important cues of lexical stress (Fry 1955; Lindblom 1963; Klatt, 1976; Zuraiq and Sereno 2007). In addition, the results showed that SAE learners of MSA resemble NS of Arabic in their use of pitch and intensity to cue lexical stress in MSA. This particular finding may lend further support to the argument that Arabic exhibits similar correlates to stress as English (de Jong and Zawaydeh 1999). The results provided partial support of the research question, which suggests that the results might still be widely applicable. However, the sample size presents a methodological limitation

50

Production of lexical stress cues

that can be addressed in future research. The current results suggest that an increase in the sample size may give greater power to detect the reported differences. Further investigation should examine correlates of stress at sentence level by SAE learners of Arabic. In addition, the production of L2 speakers with a higher proficiency level should be examined to see if their production of the stress cues would more closely resemble the production of native speakers. Since one of the most salient aspects of speech is accent, highlighting such a subtle difference between MSA and SAE can result in improvement in students’ pronunciation of MSA at the suprasegmental level. This can be helpful in Arabic foreign language learning settings, where natural target language input is scarce, as in the case of learners of MSA as a foreign language.

51

Mashael Al-Aloula

Appendix A: Language Background Questionnaire Participants ID # :_________ Age: ___________________ Gender:  Female/Male What is your native language? ___________________________________________________________________________ How many other languages do you speak? List them below. 1. _______________________________________ 2. _______________________________________ 3. _______________________________________ How many semesters/years have you been studying MSA at a university level? ___________________________________________________________________________ ___________________________________________________________________________ Have you ever studied Arabic outside of college? If yes, please indicate where. ___________________________________________________________________________ ___________________________________________________________________________ Do you practice Arabic outside of class? If yes, please indicate where and how often. ___________________________________________________________________________ ___________________________________________________________________________ How many varieties of the Arabic language are you familiar with? (e.g. Regional Dialects). Please indicate your proficiency level in each variety. Variety

Very Good/ Excellent

Fair

Poor

1. 2. 3.

Have you ever been to any Arab country? If yes, for how long? Country

Duration of stay

1.

1.

2.

2.

3.

3.

52

Production of lexical stress cues

If there are more countries and you need more space, please use this space here: ______________________________________________________________________________ ______________________________________________________________________________

Appendix B: Stimuli (Participants read the items randomized and, as shown, no diacritic/short vowels were provided.) ‫وجد طالب الكتاب‬ ‫وجد أحمد الكتاب‬ ‫وجد طالبان الكتاب‬ ‫وجد الرجل الكتاب‬ ‫وجد ضابط الكتاب‬ ‫وجد خالد الكتاب‬ ‫وجد ضابطان الكتاب‬ ‫وجد سمير الكتاب‬ ‫وجد والد الكتاب‬ ‫وجد أخي الكتاب‬ ‫وجد والدان الكتاب‬ ‫وجد محمد الكتاب‬

 1 2 3 4 5 6 7 8 9 10 11 12

‫وجد باحث الكتاب‬ ‫وجد المعلم الكتاب‬ ‫وجد باحثان الكتاب‬ ‫وجد عادل الكتاب‬ ‫وجد كاتب الكتاب‬ ‫وجد األخ الكتاب‬ ‫وجد كاتبان الكتاب‬ ‫وجد األخ الكتاب‬ ‫وجد عالم الكتاب‬ ‫وجد الوالد الكتاب‬ ‫وجد عالمان الكتاب‬ ‫وجد األستاذ الكتاب‬

13 14 15 16 17 18 19 20 21 22 23 24

References Abercombie, D., 1967. Elements of general phonetics. Edinburgh: Aldine Publishing Company. Abushihab, I., 2010. Phonological contrastive analysis of Arabic, Turkish and English. Journal of Language and Literature, 4, 16–24. Adams, C., 1979. English speech rhythm and the foreign learner. The Hague: Walter de Gruyter. Alghmaiz, B. A., 2013. Word-initial consonant cluster patterns in the Arabic Najdi dialect. Thesis (MA). Southern Illinois University at Carbondale. Al-Jarrah, R. S., 2002. An optimality-theoretic analysis of stress in the English of native Arabic speakers. Thesis (PhD). Ball State University. Alrajeh, N., 2011. Digital processing of stress in Standard Arabic. Journal of King Saud UniversityLanguages and Translation, 23 (2), 65–68. Al-Thamery, A. A. and Ibrahim, M. A., 2005. Word-Stress in Arabic: A Phonological Study from A Generative Perspective. Journal of the College of Arts University of Basrah, 1 (38), 22–31. Anani, M., 1989. Incorrect stress placement in the case of Arab learners of English. International Review of Applied Linguistics in Language Teaching, 27 (1), 15–22. Anderson-Hsieh, J., Johnson, R. and Koehler, K., 1992. The relationship between native speaker judgments of nonnative pronunciation and deviance in segmentals, prosody, and syllable structure. Language learning, 42 (4), 529–555. Archibald, J., 1998. Second language phonology (Vol. 17). Amsterdam: John Benjamins Publishing. Badawi, E.S.M. ed., 1996. Understanding Arabic: Essays in contemporary Arabic linguistics in honor of El-Said Badawi. Cairo:American University in Cairo Press. Beckman, M.E., 1986. Stress and non-stress accent. Vol. 7. Dordrecht: Walter de Gruyter. Boersma, P. and Weenink, D., 2015. Praat version 5.4. 08. Doing phonetics by computer. [Online]. Available from: http://www.fon.hum.uva.nl/praat/download_mac.html [Accessed 1 January 2015]. Bohn, O. S. and Flege, J. E., 1992. The production of new and similar vowels by adult German learners of English. Studies in Second Language Acquisition, 14 (2), 131–158. Brustad, K. et al., 2004. Al-kitāb fī taʿallum al-ʿArabīyah: A textbook for beginning Arabic, part one, 2nd ed. Washington, DC: Georgetown University Press.

53

Mashael Al-Aloula Cooper, N., Cutler, A. and Wales, R., 2002. Constraints of lexical stress on lexical access in English: Evidence from native and non-native listeners. Language and Speech, 45 (3), 207–228. De Jong, K. and Zawaydeh, B. A., 1999. Stress, duration, and intonation in Arabic word-level prosody. Journal of Phonetics, 27 (1), 3–22. De Jong, K. and Zawaydeh, B. A., 2002. Comparing stress, lexical focus, and segmental focus: Patterns of variation in Arabic vowel duration. Journal of Phonetics, 30 (1), 53–75. Derwing, T. M., Munro, M. J. and Wiebe, G., 1998. Evidence in favor of a broad framework for pronunciation instruction. Language Learning, 48 (3), 393–410. El-Hassan, S., 1994. English accentuation and vowel quality as pronounced by Arabs: A pedagogic statement. Papers and Studies in Contrastive Linguistics, 29, 205–216. Field, J., 2005. Intelligibility and the listener: The role of lexical stress. TESOL Quarterly, 39 (3), 399–423. Flege, J. E. and Bohn, O. S., 1989. An instrumental study of vowel reduction and stress placement in Spanish-accented English. Studies in Second Language Acquisition, 11 (1), 35–62. Fry, D. B., 1955. Duration and intensity as physical correlates of linguistic stress. The Journal of the Acoustical Society of America, 27 (4), 765–768. Fry, D. B., 1958. Experiments in the perception of stress. Language and Speech, 1 (2), 126–152. Grover, C., Jamieson, D. G. and Dobrovolsky, M. B., 1987. Intonation in English, French and German: Perception and production. Language and Speech, 30 (3), 277–295. Guion, S. G., 2005. Knowledge of English word stress patterns in early and late Korean-English bilinguals. Studies in Second Language Acquisition, 27 (4), 503–533. Guion, S. G., Harada, T. and Clark, J. J., 2004. Early and late Spanish – English bilinguals’ acquisition of English word stress patterns. Bilingualism: Language and Cognition, 7 (3), 207–226. Holes, C., 2004. Modern Arabic: Structures, functions, and varieties. Washington, DC: Georgetown University Press. House, D., Bruce, G., Lacerda, F. and Lindblom, B., 1987. Automatic prosodic analysis for Swedish speech recognition. In ECST, 1215–1218. Ingham, B., 1994. Najdi Arabic: Central Arabian. Washington, DC: John Benjamins Publishing. Ioup, G., 2008. Exploring the role of age in the acquisition of a second language phonology. Phonology and Second Language Acquisition, 36, 41–62. Jung, M. Y., 2010. The Intelligibility and comprehensibility of world Englishes to non-native speakers. Journal of Pan-Pacific Association of Applied Linguistics, 14 (2), 141–163. Kaye, A. S., 2001. Diglossia: The state of the art. International Journal of the Sociology of Language, 152, 117–130. Kidder, E., 2008. Tone, intonation, stress and duration in Navajo. In: Anon, ed. Coyote papers: Working papers in linguistics, linguistic theory at the University of Arizona, 16, 55–66. Klatt, D. H., 1976. Linguistic uses of segmental duration in English: Acoustic and perceptual evidence. The Journal of the Acoustical Society of America, 59 (5), 1208–1221. Larudee, F., 1973. Word stress in the spoken Arabic of Cairo. Language Sciences, 26, 31–34. Lee, B., Guion, S. G. and Harada, T., 2006. Acoustic analysis of the production of unstressed English vowels by early and late Korean and Japanese bilinguals. Studies in Second Language Acquisition, 28 (3), 487–513. Leyden, K.V. and van Heuven, V.J, 1996. Lexical stress and spoken word recognition: Dutch vs. English. Linguistics in the Netherlands,159–170. Lieberman, P., 1960. Some acoustic correlates of word stress in American English. The Journal of the Acoustical Society of America, 32 (4), 451–454. Lindblom, B., 1963. Spectrographic study of vowel reduction. The Journal of the Acoustical Society of America, 35 (11), 1773–1781. Lukyanchenko, A. et al., 2010. Opening your ears: The role of L1 in processing of nonnative prosodic contrasts. In: G. Granena et al., eds. Selected proceedings of the second language research forum. Somerville: Cascadilla Press, 50–62. McCarthy, J.J. and Prince, A.S., 1990. Foot and word in prosodic morphology: The Arabic broken plural. Natural Language & Linguistic Theory, 8 (2), 209–283. Maczuga, P. S., 2014. Production of German L2 stress by native speakers of English. Thesis (PhD). University of Calgary. Morrill, T., 2012. Acoustic correlates of stress in English adjective – noun compounds. Language and Speech, 55 (2), 167–201.

54

Production of lexical stress cues Munro, M. J., 1993. Productions of English vowels by native speakers of Arabic: Acoustic measurements and accentedness ratings. Language and Speech, 36 (1), 39–66. Munro, M. J., 1995. Nonsegmental factors in foreign accent: Ratings of filtered speech. Studies in Second Language Acquisition, 17 (1), 17–34. Munro, M. J., 2008. Foreign accent and speech intelligibility. In: J. G. H. Edwards and M. L. Zampini, eds. Phonology and second language acquisition. Amsterdam: John Benjamins, 193–218. Munro, M. J. and Derwing, T. M., 1999. Foreign accent, comprehensibility, and intelligibility in the speech of second language learners. Language Learning, 49 (1), 285–310. Odisho, E., 2005. Techniques of teaching comparative pronunciation in Arabic and English. Piscataway, NJ: Gorgias Press LLC. Pennington, M. C., 2014. Phonology in English language teaching: An international approach. New York: Routledge. Peperkamp, S. and Dupoux, E., 2002. A typological study of stress ‘deafness’. Laboratory Phonology, 7, 203–240. Rasier, L. and Hiligsmann, P., 2007. Prosodic transfer from L1 to L2. Methodological issues and description. Nouveaux Cahiers de Linguistique Française, 28, 41–66. Reetz, H. and Jongman, A., 2011. Phonetics: Transcription, production, acoustics, and perception. Oxford: Wiley. Ryding, K. C., 2005. A reference grammar of Modern Standard Arabic. Cambridge: Cambridge university press. Ryding, K. C., 2014. Arabic: A linguistic introduction. Cambridge: Cambridge University Press. Shah, A. P., 2003. Acoustic indicators of Spanish-accented English. Canadian Acoustics, 31 (3), 42–43. Trofimovich, P. and Baker, W., 2006. Learning second language suprasegmentals: Effect of L2 experience on prosody and fluency characteristics of L2 speech. Studies in second language acquisition, 28 (1), 1–30. Ueyama, M., 2000. Prosodic transfer: An acoustic study of L 2 English vs. L 2 Japanese. Thesis (PhD). UCLA. Van Summers, W., 1987. Effects of stress and final-consonant voicing on vowel production: Articulatory and acoustic analyses. The Journal of the Acoustical Society of America, 82 (3), 847–863. Watson, J. C., 2002. The phonology and morphology of Arabic. Oxford: Oxford University Press on Demand. Youssef, A. and Mazurkewich, I., 1998. The acquisition of English metrical parameters and syllable structure by adult native speakers of Egyptian Arabic (Cairene dialect). In: S. Flynn et al., eds. The generative study of second language acquisition. Mahwah: Lawrence Erlbaum Associates, 303–332. Zuraiq, W. and Sereno, J. A., 2007. English lexical stress cues in native English and non-native Arabic speakers. Ratio, 1 (2), 829–832.

55

3 NATIVE ENGLISH SPEAKERS’ PERCEPTION AND PRODUCTION OF ARABIC CONSONANTS Asmaa Shehata It has been well known that adult second language (L2) learners encounter difficulties learning L2 contrastive sound categories (Aoyama et al. 2004; Strange and Shafer 2008). Research on adult L2 learners has widely explored their acquisition of various English contrastive sounds such as English /l-r/ by Japanese learners (Bradlow et al. 1999; Hattori 2009; Sheldon and Strange 1982; Takagi 2002); however, less is known about Arabic learners’ perception and production of consonant contrastive sounds. In this follow-up investigation, the acquisition of Arabic consonant sounds by non-native adult speakers is examined. In particular, the study explores adult native English speakers’ ability to distinguish and accurately produce Arabic consonant contrasts and also examines the relationship between learners’ perception and production skills. To this end, 45 American English learners of Arabic from three separate levels of Arabic instructions participate in two chief tasks: perception and production. Three native Arabic speakers subsequently evaluate learners’ productions of the target stimuli as either correct or incorrect. The results show that novel Arabic contrastive consonant sounds are difficult for native English speakers to discriminate. However, the degree of difficulty varies, where the /t-ṭ/, /h-ħ/, and /s-ṣ/ pairs are found to be the most difficult and the /ħ-ʕ/ contrast is the easiest. Whereas advanced learners’ performance exceeds that of their counterparts in the other two groups, no difference in the performance on the two tasks is found between the beginner and intermediate learners of Arabic. Findings also provide some evidence for the perceptual assimilation model (PAM) predictions, where Best’s (1995) categories apply to certain learners’ perception patterns. Results suggest that even in the absence of study abroad experience and direct exposure to the Arabic language, English learners of Arabic are not only able to tackle the difficulties presented by the difficult Arabic consonant contrasts, but they show steady improvement as well.

3.1 Introduction Learning a second language entails, among other things, acquiring its phonology. This seems to be a feasible task for young L2 language learners who start acquiring the second language by 6 years of age and subsequently achieve native-like proficiency (Long 1990). However, it is found to be a challenging task for late L2 learners, who encounter difficulties in perceiving 56

Native English speakers’ perception and production of Arabic consonants

and/or producing non-native segmental sound contrasts, and their speech is commonly characterized by a foreign accent (Goto 1971, Polka 1991; Bradlow et al. 1999; Aoyama et al. 2004; Broersma 2010). For instance, Japanese speakers have difficulty distinguishing and producing English /r/ and /l/ (Best 1995; Ingram and Park 1998; Bradlow and Pisoni 1999; Takagi 2002), and native Dutch speakers strive to differentiate words containing /æ/ from words containing /ε/ (Broersma 2005; Cutler and Broersma 2005). In addition, Spanish learners of English find it difficult to discriminate and accurately produce the English /i/-/ɪ/ contrast and thus pronounce words like ‘ship’ as ‘sheep’, which results in comprehension misunderstandings (Flege and MacKay 2004). In the realm of studying L2 learners’ acquisition of novel sounds, various research studies have investigated L2 learners’ perception and production skills (Borden et al. 1983; Flege 1999). In this regard, a good number of empirical studies have highlighted the influential role of several factors such as age of arrival (AOA) and length of residence (LOR) in the target country. For example, Bohn and Flege (1990) examined the perception and production of English vowels /ɛ/ and /æ/ by two groups of native German speakers: experienced (i.e., lived in the United States (US) for more than five years) and inexperienced (i.e., lived in the US for less than six months). Findings showed that experienced learners were more accurate at perceiving and producing English vowel contrasts than their inexperienced counterparts. Bohn and Flege concluded that learners’ LOR could influence learners’ acquisition of novel L2 sounds. Moreover, Yamada (1995) found that native Japanese speakers with longer LOR and earlier AOA to the United States were more accurate at identifying synthetic English /l/ and /r/ contrasts than their counterparts with a short LOR and less L2 experience.

3.1.1 Background By and large, the link between perception and production of unfamiliar L2 sounds in adult acquisition has been of growing interest to researchers in the past few decades. Findings of previous research studies have shown three main patterns explaining the nature of the perception-production relationship. The first pattern shows that learners’ perception ability is independent of their production ability (Borden et al. 1983; Flege 1988; Bohn and Flege 1990; Flege 1993). For example, Flege (1993) found Chinese and Taiwanese learners of English to be more accurate at perceiving the English /t/ and /d/ contrast than producing them. Furthermore, Flege et al. (1999) indicated that advanced native Italian leaners of English were less accurate at producing ten English vowels /i, ɪ, e, ɛ, æ, o, ʌ, ɒ, ʊ, u/ than perceiving them. The second pattern stresses the link between perception and production that is confirmed by a number of studies. For instance, Akahane-Yamada et al. (1996) indicated that perceptual training for Japanese learners of English helped them to distinguish English /r/-/l/ minimal pairs and consequently improving both learners’ perception and production. The same finding was confirmed by Okuno and Hardison (2016) who found that perceptual training could result in improving the production of Japanese vowel duration by English learners of Japanese. The third pattern, however, posits that perception is dependent on production. For example, some studies exhibited that the Japanese learners’ accurate production of the English /l-r/ consonant contrast may precede their accurate perception (Goto 1971; Sheldon and Strange 1982). Additionally, Zampini (1998) found evidence that production develops before perception for the Spanish /p/ by native English speakers. Similarly, Hattori (2009) found that intensive training improved Japanese learners’ production of the English /l/ and /r/ but not their perception. 57

Asmaa Shehata

Taken together, the relationship between the perception and production of non-native contrastive sounds by adult L2 learners is still unclear where conflicting results were reported. Therefore, more research is still needed to better understand the nature of the link between the two abilities in adult L2 learners’ speech and consequently develop effective training that can help learners improve their perception and production skills. Thus, the first goal of this chapter is to explore the relationship between the perception and production of Arabic consonant phonemes by English learners of Arabic with different years of language instruction.

3.1.2  The Arabic language Arabic is a Semitic language that is officially spoken by over 422 million in 22 countries in the Middle East and North Africa, where there are two main varieties used simultaneously: the high variety (i.e., Modern Standard Arabic (MSA)) and the low variety (i.e., the colloquial spoken dialects). While MSA is used in formal settings and is taught in schools and universities, the low variety includes a number of regional colloquial spoken variants that are commonly used in daily life communication and vary from one region to another. This paper focuses on MSA consonant phonemes (see the Introduction to this volume: Table 0.1). While MSA and English share a number of similar consonant phonemes (e.g., /b/, /t/, /d/, /k/, /ʔ/, /f/, /θ/, /ð/, /s/, /z/, /š/, /h/, /m/, /n/, /ž/, /l/, /r/, /w/, /j/), MSA has nine additional consonants sounds that are not part of the English consonant inventory (e.g., /ṭ/, /ḍ/, /ṣ/, /ð̩/, /χ/, /γ/, /q/, /ħ/, /ʕ/) and represent a challenge for L2 Arabic learners (Alosh 1987; Shehata 2015a; Shehata 2017). In the US context, Arabic is classified as a less commonly taught language that has recently witnessed a significant student enrollment increase at a number of US universities and colleges (Allen 2004; Welles 2004; Furman et al. 2010). Previous studies have mainly focused on literacy skills on Arabic as a first language (L1) (e.g., Asaad and Eviatar 2014;SaieghHaddad 2003; Abu-Rabia and Siegel 2002; Roman and Pavard 1987), a few on Arabic L2 reading and vocabulary (Khaldieh 2001, 2000; Redouane 2003), and most on Arabic L2 morphosyntax (Alhawary 2009). Nevertheless, few studies have investigated the acquisition of Arabic phonological features by adult L2 learners (Hong and Sarmah 2009; Zahid 1996), and among those, very little research has examined Modern Standard Arabic contrastive consonants experimentally exploring the difficulties that native American English speakers encounter in acquiring them (Alosh 1987; Shehata 2015b). For example, Alosh (1987) examined whether native Arabic speakers and American learners of Arabic utilized different strategies in perceiving Arabic pharyngealized consonants /sˤ, dˤ/. This was examined by using synthetic sounds consisting of three plain vowels /aa, ii, uu/ adjacent to each of the target consonants. The results provided evidence that the two groups used different strategies in distinguishing pharyngealization. Native Arabic speakers were better at discriminating pharyngealized consonants than their American counterparts, who seemed to be more sensitive to the pharyngealization feature in the pharyngealized vowel /aa/. Findings also showed a significant interaction between learners’ Arabic proficiency level and their abilities to identify pharyngealization, where learners with high proficiency level were more accurate at discriminating pharyngealization on the adjacent vowel. Furthermore, Shehata (2015b) examined the influence of training in different talker variability presentation modes on the acquisition of Arabic pharyngeal-glottal consonant contrast /h-ħ/ by native speakers of English with no prior knowledge of Arabic using two discrimination AXB tasks: non-lexical and lexical. The results showed that learners who heard the target

58

Native English speakers’ perception and production of Arabic consonants

token produced by multiple speakers performed significantly more accurately than their counterparts in the single-talker training groups. This improvement was explained in terms of the richness of the input provided to learners in the multipletalker-training condition, which facilitated learners’ discrimination of the pharyngeal-glottal Arabic contrast. More recently, Lancaster and Gor (2016) investigated English speakers’ perception of Arabic pharyngeal /ħ/, /ʕ/ and uvular /χ/, /γ/ fricatives using discrimination and identification tasks and a cloze test to investigate the predictions of two theoretical models: the perceptual assimilation model (PAM) and the automatic selective perception model (ASPM). The results provided evidence that learners were able to discriminate the pharyngeal sounds more accurately than the uvular ones on one hand and the voiced phonemes than their voiceless counterparts on the other hand. The authors concluded that their findings matched the predictions of the ASPM rather than PAM’s hypotheses. The present study is a follow-up to Shehata (2015a) that explored the views of 107 adult native English speakers residing in the US regarding the difficulty of Arabic consonant phonemes in both perception and production. Learners’ beliefs about the essential factors in the acquisition of Arabic consonant sounds were also examined. Findings showed that the pharyngeal and the pharyngealized consonant phonemes were felt by learners to be too difficult to perceive, and produce. Moreover, all learners reported the fricative pharyngeal-glottal /h-ħ/ phoneme contrast to be the most challenging Arabic consonant contrast. This study attempts to build on earlier work by experimentally examining the perception and production of the same Arabic consonant contrasts used in Shehata’s study (2015a) by American English learners of Arabic. It aims at showing the effect of learners’ years of study on their discrimination and production abilities of the target contrastive consonants.

3.2  Statement of the research questions Modern Standard Arabic includes a number of consonant sounds that represent a substantial challenge for English learners (Alosh 1987; Shehata 2015a, 2015b). Prior research has mainly focused on exploring adult L2 learners’ perception and production of novel Arabic segments independently (Alwabari 2013; Al Mahmoud 2013). However, the interrelation between L2 learners’ perception and production of Arabic consonant phonemes is still unclear. Thus, the present research aims to fill this research gap by exploring the relationship between native English speakers’ accuracy in perceiving and producing Arabic consonant contrasts, addressing the following two main questions: 1 Do learners’ years of Arabic instruction influence their perception and production of Arabic consonant contrasts? 2 Is there a relationship between learners’ perception and production of Arabic consonant contrasts?

3.3 Methods 3.3.1 Participants There were two groups of subjects who participated in the current study. The first group included 45 native American English learners of Arabic between the ages of 18–30 years old (mean = 21) who were divided into three subgroups based on their length of study of the 59

Asmaa Shehata

Arabic language. In the beginner group, there were 15 native English speakers (ten females and five males) who spent at least one year studying Arabic (with a total of eight credit hours) and were in their third Arabic semester at the time of the study. In the intermediate group, there were 15 learners (nine females and six males) who studied Arabic for two years (with a total of 16 credit hours) and were in their fifth Arabic semester at the time of the study. The advanced group included 15 learners (eight females and seven males) who studied Arabic for three and a half years (with a total of 24 credit hours) and were in their seventh Arabic semester at the time of the study. All learners were recruited from a large Midwest university campus in the US, and they either received course credit or were paid for their participation. In the language background questionnaire that all learners completed before performing the study, they reported having no speech or hearing problems. None of them studied Arabic abroad nor spent extended periods of time in an Arabic-speaking country. The second group included three native Arabic speakers who listened to participants’ recordings in the production task grouped by target words and scored each token in turn as correct or incorrect. The three raters were graduate students at the same university where the data were collected; two of them were studying literature and the third one was studying education. Table 3.1 presents information about the three raters.

3.3.2  Stimuli and procedures Twenty monosyllabic CVC Arabic nonwords were used as stimuli in the perception experiment and as target words in the production task. These tokens comprised ten minimal pairs contrasting the target Arabic phonemes (i.e., /t-ṭ/, /d-ḍ/, /θ-ð/, /ð-ð ̣/, /s-ṣ/, /h-ħ/, /k-q/, /ʔ-ʕ/, /χ-ɣ/ and /ħ-ʕ/) in onset position (e.g., /da:k-ḍa:k/). These were the same nine consonant contrasts used in Shehata’s research (2015a), in addition to the pharyngeal contrast /ħ-ʕ/ that learners reported in their questionnaires to be a problematic contrast. The reason for choosing these sounds in particular was to be able to compare learners’ beliefs reported in the previous study with their actual performance in the current study. A native male Arabic speaker produced each of the nonwords four times in a carrier sentence, ʔuriidu ʔan ʔaquul kalimata ________ “I want to say the word _________,” in a sound-attenuated booth, using a Marantz PMD 660 recorder and a Samson QV microphone. After recording, the second production of each stimulus was extracted and normalized to 65dB using Praat (Boersma and Weenink 2012). A list of the stimuli is provided in Table 3.2. Besides the language background questionnaire, all participants completed two main tasks: (1) a read-aloud production task; and (2) an AXB discrimination perception task; these were completed in the order explained here.

Table 3.1 Three raters’ information

Rater 1 Rater 2 Rater 3

Age

Hometown

Length of Residence in the US (in years)

22 30 23

Alexandria, Egypt Cairo, Egypt Amman, Jordan

7 5 3.9

60

Native English speakers’ perception and production of Arabic consonants Table 3.2 Stimuli used in the perception and production tasks Consonant Contrast

Pseudowords (Auditory)

Arabic Script (Production)

/t-ṭ/ /θ-ð/ /s-ṣ/ /d-ḍ/ /h-ħ/ /ð-ð ̣/ /k-q/ /ʔ-ʕ/ /χ-ɣ/ /ħ-ʕ/

/ta:k-ṭa:k/ /θa:k-ða:k/ /sa:k-ṣa:k/ /da:k-ḍa:k/ /ha:k-ħa:k/ /ða:k-ð ̣a:k/ /ka:k-qa:k/ /ʔa:k-ʕa:k/ /χa:k-ɣa:k/ /ħa:k-ʕa:k/

‫طاك‬  – ‫تاك‬ ‫ذاك‬  – ‫ثاك‬ ‫صاك‬  – ‫ساك‬ ‫ضاك‬  – ‫داك‬ ‫حاك‬  – ‫هاك‬ ‫ظاك‬  – ‫ذاك‬ ‫قاك‬  – ‫كاك‬ ‫عاك‬  – ‫أاك‬ ‫غاك‬  – ‫خاك‬ ‫عاك‬  – ‫حاك‬

3.3.2.1  Production task The production task took place first in order not to expose the participants to auditory input involving the target stimuli. In a sound-attenuated booth, all learners were recorded using a Marantz PMD 660 recorder and a Samson QV microphone while producing the target minimal pairs that were presented in a word list using Arabic script (see Table 3.2). The recordings were later played to three native Arabic speakers who heard 45 recordings of each target word produced by different talkers, which they rated as either correct or incorrect. The inter-rater reliability score was 0.933, which referred to a very strong relationship between the three raters. The choice of the production task was guided by previous relevant research (Alosh 1987; Bradlow et al 1997; Flege 1993; Hattori 2009; Munro 1992; Okuno and Hardison 2016) that indicated the appropriateness of this task that measures learners’ language production abilities of the target language. Moreover, the use of nonsense words mades this controlled task a suitable one, because the goal was to check if participants could produce the target sounds and blend them into words rather than recognizing the words by sight.

3.3.2.2  Perception task After completing the production task, the same native English speakers participated in an AXB discrimination task in which they heard three nonwords (A, X, and B) and decided whether the second (X) was more similar to the first (A) or the third (B); e.g., /ta:k/, /ta:k/, /ṭa:k/. Participants registered their responses by pressing either the right or left shift keys (labeled First and Third) on a keyboard. Four test items (AAB, ABB, BAA, BBA) were generated for each of the 10 contrasts (listed in Table 3.2), as in the example here: AAB /ta:k/, /ta:k/, /ṭa:k/ ABB /ta:k/, /ṭa:k/, /ṭa:k/ BAA /ṭa:k/, /ta:k/, /ta:k/ BBA /ṭa:k/, /ṭa:k/, /ta:k/ This resulted in a total of 40 that were presented in random order per block, and each block was presented four times (40 × 4 blocks = 160 total). The task took place in a sound-attenuated booth over one session that lasted for 30 minutes. 61

Asmaa Shehata

3.4 Results 3.4.1  Consonant production Each participant’s correct production of each token, as rated by the three native speakers, was calculated and submitted to a two-way analysis of variance (ANOVA) to test the effect of contrast type (ten levels) and learner level (three levels: beginners, intermediate, and advanced) on learners’ production accuracy of the target consonant contrasts. The results indicated a significant effect of the contrast type, F(9, 42) = 5.127, p < .001 where /θ-ð/ and /χ-ɣ/ contrasts were the easiest and /h-ħ/ was the most difficult to produce, where learners’ accuracy rate was 100% and 66% respectively. There was also a significant effect of learner level, F(2, 42) = 13.925, p < .001 (see Figure 3.1). Crucially, there was no significant interaction between the effects of contrast type and learner level, F(2, 42) = .580, p = .914. As seen in Figure 3.1, the accuracy of learners in the three groups reached the ceiling for two different contrasts: the familiar contrast /θ-ð/ and the novel one /χ-ɣ/. However, their production accuracy of the emphatic contrasts (i.e., /t-ṭ/, /s-ṣ/, /ð-ð ̣/, /d-ḍ/) and the pharyngeal contrast /ħ-ʕ/ varied. Yet the pharyngeal-glottal contrast /ʔ-ʕ/ was found to be relatively less challenging, where learners’ accuracy rate was 73%. A Bonferroni post hoc test showed that the advanced learners were significantly more accurate at producing the target consonant contrasts than the beginner group (p < .001) and the intermediate group (p = .003). However, there was no significant difference between the beginner and intermediate groups (p = .110).

Figure 3.1 Mean proportion correct for learners’ production by group and contrast

62

Native English speakers’ perception and production of Arabic consonants

Figure 3.2 Mean proportion correct for learners’ perception by group and contrast

3.4.2  Consonant perception Each participant’s proportion of correct responses for the ten target contrasts was calculated, and the data were analyzed using a two-way ANOVA that examined the significance of the independent variable (contrast type). The results indicated a significant effect of contrast, F(9, 42) = 4.798, p < .001, and of learner level, F(2, 42) = 8.660, p < .001(see Figure 3.2). However, there was no significant interaction between the effects of contrast type and learner level, F(2, 42) = .248, p = .999, indicating that the three groups did not differ from one another in their performance on the target contrasts. A Bonferroni post hoc test showed that the advanced learners were significantly more accurate at discriminating the target Arabic consonant contrasts than the beginners (p < .001) and marginally more accurate than the intermediate learners (p = .072). However, the beginner group did not differ signficantly from the intermediate one (p = .179). Taken together, the results suggested that both beginner and intermediate Arabic learners had more difficulty producing and discriminating the target consonant contrasts than the advanced learners.

3.4.3  Comparing perception and production To examine the relationship between perception and production, participants’ perception and production scores were compared via a Spearman nonparametric correlation analysis, which showed a positive correlation between learners’ perception and production accuracy scores (r = .740, p < .0001), indicating the link between the two skills (see Figure 3.3). 63

Asmaa Shehata

Figure 3.3 Learners’ mean production and perception scores per contrast and group

3.5 Discussion This study sheds light on American English speakers’ perception and production of 10 Arabic consonant contrasts and also explores the role of learners’ years of language instruction in their acquisition of these target L2 phonemes. In doing so, this study provides new data regarding the relationship between speech perception and production. In accordance with PAM predictions, learners’ discrimination success rate for the familiar /θ-ð/ contrast that PAM classifies as two-category assimilation (TC) is not only excellent (100%) but also the highest across other contrast types. This finding is consistent with native English speakers’ high accuracy for the Zulu TC pair /ɬ-ɮ/ and Malay speakers’ high accuracy rate for the English TC contrasts (/t-d/ and /f-v/) reported in Best et al. (2001) and Pilus (2005), respectively. In general, the results provide evidence that native English speakers experience difficulty distinguishing novel Arabic contrastive consonants such as /t-ṭ/, and /d-ḍ/. It is interesting to see learners stumbling over the /h-ħ/ contrast, which includes one familiar sound /h/ contrasting to an unfamiliar one /ħ/, but performing better on the voiceless-voiced phalangeal /ħ-ʕ/ contrast that includes two unfamiliar phonemes. This could be explained in light of PAM’s prediction of poor discrimination for the single category (SC) contrast, which includes one familiar sound [h] and an unfamiliar one [ħ]. Therefore, all leaners regardless of their language level find the /h-ħ/ pair the hardest contrast. This can also explain learners’ poor success rate of the velaruvular contrast /k-q/. Along the same line, learners’ discrimnation accuracy for the other pharyngealized contrastive consonant sounds (i.e., /t-ṭ/, /d-ḍ/, /ð-ð ̣/, and /s-ṣ/) is slightly low, which is consistent with Alosh (1987), Alwabari (2013), and Shehata (2015a). With non-assimilable (NA) pairs that include distinct novel sounds that are not within the English phonetic space, however, learners find the /χ-ɣ/ and /ʕ-ħ/ contrasts to be easier, and their overall success rate is relatively good. In relation to production results, it is worth noting that learners perform better in the word production task in general and reach ceiling in their accurate production of /θ-ð/ and /χ-ɣ/ 64

Native English speakers’ perception and production of Arabic consonants

(100%) contrasts in particular. In general, learners’ accuracy production scores are above chance for the other contrasts except for /h-ħ/ and /d-ḍ/ contrasts, on which the intermediate learners performed above chance and the beginners performed at random chance. It is also found that the perception and production difficulties vary in light of the number of years of study. Advanced learners significantly display a different pattern across contrast types in comparison to the beginner and intermediate groups in both perception and production, and this answers the first research question. For instance, the production accuracy for the /ð-ð ̣/ contrast improves from 57% for the beginner learners to 63% for the intermediate group and reaches ceiling (93%) for the advanced learners. By and large, after three years of study, the advanced group is at near ceiling level of accuracy in production and performs above chance in perception. In contrast, the beginner group displays more errors in distinguishing and producing the target tokens than the intermediate group; however, the results provide no evidence that the performance of the two groups significantly differs on both the AXB discrimination task and the read-aloud production task. Comparing the findings of the present study to those of Shehata’s (2015a) shows similarities concerning learners’ difficulty of distinguishing the pharyngeal and pharyngealized consonant phonemes. The two studies also agree regarding the relative easiness of perceiving and producing both the /χ-ɣ/ and the /ʔ-ʕ/ pairs. However, there are some differences between the two studies regarding the level of difficulty of the Arabic contrasts at large. As shown in Figure 3.4, learners’ fears about difficulties reported in Shehata (2015a) were much higher than the actual performance of participants in the present study. For example, although learners in Shehata (2015a) reported a high level of difficulty for both /d-ḍ/and /ð-ð ̣/, which was over 70%, it is found to be less than 40% in this study. Even for the pharyngeal-glottal contrast that was found to be the most difficult phoneme to acquire in the two studies, its level of difficulty was higher (over 90%) in Shehata (2015a) than the current study (less than 48%). By and large, one of the positive findings of this study is the advanced learners’ improved overall accuracy in the identification and production of the target sounds, which gives a good message to beginner and intermediate learners that endurance pays off.

100

Perception

90

Production

80 70 60 50 40 30 20 10 0 /t-tς/

/d-dς/

/θ-ð/

/ð-ð

ς

/s-s /

/h-ħ/

/k-q/

/ʔ-ʕ/

/χ-ɣ/

Figure 3.4a Percentage of difficulty of Arabic consonant phoneme contrasts in Shehata’s study (2015a)

65

Asmaa Shehata

100

Perception

90

Production

80 70 60 50 40 30 20 10 0 /t-t/

/θ-ð/

/s-s/

/d-d/

/h-ħ/

/ð-ð/

/k-q/

/ʔ-ʕ/

/χ-ɣ/

/ħ-ʕ/

Figure 3.4b Percentage of difficulty of Arabic consonant phoneme contrasts in the present study

Importantly, the results reveal that native English learners’ production ability, measured by a controlled production task, is positively correlated with perception ability, measured via an AXB discrimination task, and this answers the second research question. The results are consistent with previous research (Yamada et al. 1994; Bradlow et al. 1999) that found a significant link between L2 learners’ accurate production and perception of L2 phonemes. These similar findings can be explained in light of using the same controlled production task to examine learners’ production ability (i.e., of pronunciation). Although the results here suggest the link between the two language skills, the findings should be interpreted with some caution, as learners’ production is only tested by a reading aloud production task that may not accurately represent learners’ productive pronunciation accuracy. It is left to be seen whether using a less controlled (more spontaneous) production task such as a picture description task, which ensures subjects produce the target sounds in a communicative way, can lead to different results. There are two additional limitations in this study. First, the study’s stimuli focus on positing target sounds only in the initial position; second, they are presented in isolation. Testing target contrasts in different phonetic positions (i.e., initial, middle, and final) and in rich communicative contexts can give more robust results. For future investigation, other production tasks can be considered (e.g., a picture-naming task) and other phonological environmental positions (i.e., middle and final) can be investigated. Future research may also explore the possibility of extending the present study to new groups of learners with different L1 backgrounds and employing perceptual training to better understand the relation between Arabic consonant production and perception accuracies.

3.6  Conclusion and pedagogical implications This chapter aims at exploring the learnability of Arabic consonant contrasts by examining the perception and production accuracies among native speakers of English from different Arabic instruction levels. Overall, the findings suggest that perception and production are dependent skills, where high production accuracy corresponds to high perception accuracy. Furthermore, 66

Native English speakers’ perception and production of Arabic consonants

the number of years of study is found to be a significant variable at the advanced level that positively influences learners’ production and discrimination accuracy of all contrastive consonant sounds. In all cases, low proficiency learners (beginner and intermediate groups) find Arabic contrastive consonant sounds more challenging to discriminate and produce than advanced learners. Additionally, the data support a number of PAM hypotheses regarding certain assimilation patterns (e.g., /θ-ð/, /ʕ-ħ/, and /h-ħ/). The study adds to the growing body of literature examining L2 learners’ acquisition of Arabic consonant sounds and has two crucial implications. First, the findings provide evidence that perception and production are connected, and thus teachers can help their students practice the two skills together. Second, results display that it is indeed the case that the pharyngeal-glottal contrast represents a challenge for L2 learners, who find it difficult to identify and produce even after years of language experience. Therefore, it is suggested that teachers need to develop instructional materials and practice drills that provide learners with various opportunities to practice these contrastive sounds and enable them to achieve perceptual and production improvements on all of them. This is the second study in the planned empirical research that is intended to provide data involving the Arabic language. Further studies are on the way that examine English learners’ perception and production of Arabic contrasts in different positions (initial, middle, and final) using different perception and production tasks at both controlled and spontaneous levels.

References Abu-Rabia, S. and Siegel, L., 2002. Reading, syntactic, orthographic, and working memory skills of bilingual Arabic-English-speaking Canadian children. Journal of Psycholinguistic Research, 31(6), 661–678. Akahane-Yamada, R., et al., 1996. Does Training in speech perception modify speech production? In: Anon, ed. Fourth International Conference on Spoken Language Processing. Philadelphia, PA: IEEE, 606–609. Al Mahmoud, M. S., 2013. Discrimination of Arabic contrasts by American learners. Studies in Second Language, 3 (2), 261–292. Alhawary, M. T. 2009. Arabic second language acquisition of morphosyntax. New Haven, CT: Yale University Press. Allen, R., 2004. Perspectives on Arabic teaching and learning. The Modern Language Journal, 88 (2), 275–278. Alosh, M. L., 1987. The perception and acquisition of pharyngealized fricatives by American learners of Arabic and implications for teaching Arabic phonology. Thesis (Ph.D). Ohio State University. Alwabari, S., 2013. Non-native production of Arabic pharyngeal and pharyngealized consonants. Thesis (MA). Carleton University. Aoyama, K., et al., 2004. Perceived phonetic dissimilarity and L2 speech learning: The case of Japanese /r/ and English /l/ and /r/. Journal of Phonetics, 32, 233–250. Asaad, H. and Eviatar, Z., 2014. Learning to read in Arabic: The long and winding road. Reading and Writing: An Interdisciplinary Journal, 27 (4), 649–664. Best, C., 1995. A  direct realist perspective on cross-language speech perception. In: W. Strange, ed. Speech perception and linguistic experience: Issues in cross-language research. Timonium, MD: York Press, 171–204. Best, C., et al., 2001. Discrimination of non-native consonant contrasts varying in perceptual assimilation to the listener’s native phonological system. Journal of the Acoustical Society of America, 109 (2), 775–794. Boersma, P. and Weenink, D., 2012. Praat: Doing phonetics by computer. [computer program]. Bohn, O-S. and Flege, J. E., 1990. Interlingual identification and the role of foreign language experience in L2 vowel perception. Applied Psycholinguistics, 11 (3), 303–328. Borden, G., et al., 1983. Production and perception of the /r/ – /l/ contrast in Korean adults learning English. Language Learning, 33(4): 499–526.

67

Asmaa Shehata Bradlow, A., et al., 1997. Training Japanese listeners to identify English /r/ and /l/: IV. Journal of the Acoustical Society of America, 101 (4), 2299–2310. Bradlow, A., et al., 1999. Training Japanese listeners to identify English /r/ and /l/: Long-term retention of learning in perception and production. Perception and Psychophysics, 61 (5), 977–985. Bradlow, A. and Pisoni, D. B., 1999. Recognition of spoken words by native and non-native listeners: Talker-, listener- and item-related factors. Journal of the Acoustical Society of America, 106 (4), 2074–2085. Broersma, M., 2005. Perception of familiar contrasts in unfamiliar positions. Journal of the Acoustical Society of America, 117 (6), 3890–3901. Broersma, M., 2010. Perception of final fricative voicing: Native and non-native listeners’ use of vowel duration. Journal of the Acoustical Society of America, 127 (3), 1636–44. Cutler, A. and Broersma, M., 2005. Phonetic precision in listening. In: W. J. Hardcastle, and J. M. Beck, eds. A figure of speech: A Festschrift for John Laver. Mahwah, NJ: Erlbaum, 63–91. Flege, J. E., 1988. Factors affecting the degree of perceived foreign accent in English sentences. Journal of the Acoustical Society of America, 84 (1), 370–389. Flege, J. E., 1993. Production and perception of a novel, second language phonetic contrast. Journal of the Acoustical Society of America, 93 (3), 1589–1608. Flege, J. E., 1999. Age of learning and second language speech. In: D. Birdsong, ed. Second language acquisition and the critical period hypothesis. Mahwah, NJ: Lawrence Erlbaum Associates, 101–131. Flege, J. E., et al., 1999. Native Italian speakers’ perception and production of English vowels. Journal of the Acoustical Society of America, 106 (5), 2973–2987. Flege, J. E. and MacKay, I. R., 2004. Perceiving vowels in a second language. Studies in Second Language Acquisition, 26 (1), 1–34. Furman, N., et al., 2010. Enrollments in languages other than English in United States institutions of higher education fall 2009. New York: Modern Language Association of America. www.mla.org/ pdf/2009_enrollment_survey.pdf. Goto, H., 1971. Auditory perception by normal Japanese adults of the sounds “l” and “r”. Neuropsychologia, 9 (3), 317–323. Hattori, K., 2009. Perception and production of English /r/-/l/ by adult Japanese speakers. Thesis (Ph.D). University College London. Hong, S. and Sarmah, P. 2009. Koryan perception of Arabic phonemes. Linguistic Interfaces, 2, 375–386. Ingram, J. C. L. and Park, S-G., 1998. Language, context, and speaker effects in the identification and discrimination of English of /r/ and /l/ by Japanese and Korean listeners. Journal of the Acoustical Society of America, 103 (2), 1161–74. Khaldieh, S. A., 2000. Learning strategies and writing processes of proficient vs. less-proficient learners of Arabic. Foreign Language Annals, 33 (5), 522–533. Khaldieh, S. A., 2001. The relationship between knowledge of lcraab, lexical knowledge, and reading comprehension of nonnative readers of Arabic. The Modern Language Journal, 85 (3), 416–431. Lancaster, A. and Gor, K., 2016. Abstraction of phonological representations in adult non-native speakers. Proceedings of the Linguistic Society of America Annual Meeting, 1 (24), 1–15. Long, M., 1990. Maturational constraints on language development. Studies in Second Language Acquisition, 12 (3), 251–285. Munro, M. 1992. Perception and production of English vowels by native English speakers of Arabic. Thesis (PhD). University of Alberta Okuno, T. and Hardison, D. M., 2016. Perception-production link in L2 Japanese vowel duration: Training with technology. Language Learning and Technology, 20 (2), 61–80. Pilus, Z., 2005. Perception of voicing in English-word-final obstruents by Malay speakers of English: Examining the Perceptual Assimilation Model. Malaysian Journal of ELT Research, 1 (1), 1–12. Polka, L., 1991. Cross-language speech-perception in adults: Phonemic, phonetic, and acoustic contributions. Journal of the Acoustical Society of America, 89 (6), 2961–297. Redouane, R., 2003. Learners’ variability in coining new words in L2. Al-‘Arabiyya, 36, 49–80. Roman, G. and Pavard, B., 1987. A comparative study: How we read in Arabic and French. In: J. K. O’Regan and A. Levy-Schoen, eds. Eye movements: From physiology to cognition. North-Holland, The Nether lands: Elsevier, 431–440. Saiegh-Haddad, E., 2003. Linguistic distance and initial reading acquisition: The case of Arabic diglossia. Applied Psycholinguistics, 24 (3), 431–451.

68

Native English speakers’ perception and production of Arabic consonants Shehata, A., 2015a. Problematic Arabic consonants for native English speakers: Learners’ perspectives. The International Journal of Educational Investigations, 2 (9), 24–47. Shehata, A., 2015b. Talker Variability and Second Language Word Recognition: A new training study. Journal of Linguistics and Language Teaching, 6 (2), 209–230. Shehata, A., 2017. Teaching Arabic pronunciation to non-natives: Cognition and practice. In: M. O’Brien and J. Levis, eds. Proceedings of the 8th Pronunciation in Second Language Learning and Teaching Conference, ISSN 2380–9566, Calgary, AL, August 2016. Ames, IA: Iowa State University, 121–131. Sheldon, A. and Strange, W., 1982. The acquisition of /r/ and /l/ by Japanese learners of English: Evidence that speech production can precede perception. Applied Psycholinguistics, 3 (3), 243–261. Strange, W. and Shafer, V. L., 2008. Speech perception in second language learners: The re-education of selective perception. In: J. G. Hansen Edwards and M. L. Zampini, eds. Phonology and Second Language Acquisition. Amsterdam: John Benjamins, 153–191. Takagi, N., 2002. The limits of training Japanese listeners to identify English /r/ and /l/: Eight case studies. Journal of the Acoustical Society of America, 111 (6), 2887–2896. Welles, E., 2004. Foreign language enrollments in United States institutions of higher education. ADFL Bulletin, 35 (2–3), 413–419. Yamada, R., 1995. Age and acquisition of second language speech sounds: Perception of American English /r/ and /l/ by native speakers of Japanese. In: W. Strange, ed. Speech perception and linguistic experience, issues in cross-language research. Timonium, MD: York Press, 305–320. Yamada, R. A., et al., 1994. The intelligibility of Japanese speakers’ productions of American English /r/, /l/, and /w/, as evaluated by native speakers of American English. In: Anon, ed. Proceedings of the international Conference on spoken language processing. Yokohama: Acoustical Society of Japan, 2023–2026. Zahid, M., 1996. Examination of a perceptual non-native speech contrast: Pharyngealized/non-pharyngealized discrimination by French-speaking adults. In: Anon, ed. Fourth international Conference on spoken language processing. Philadelphia, PA: IEEE, 2466–2469. Zampini, Mary L., 1998. The Relationship between the production and perception of L2 Spanish Stops. Texas Papers in Foreign Language Education, 3 (3), 85–10.

69

4 THE PERCEPTION AND PRODUCTION OF ARABIC CONSONANTS A cross-linguistic study Sara Al Tubuly This chapter examines the production and perception of Arabic emphatic, uvular, and pharyngeal consonants by learners from different linguistic backgrounds. The participants in the study are native speakers of English, Greek, Chinese, Turkish, and German who are learning Arabic as a second (L2) or a third language (L3). Fifty learners participated in this study, and two native speakers participated as a control group. The learners are classified according to their native tongue to examine whether their linguistic background affects the learning process. Alosh’s (1987) results indicated that there is an interaction between the perception of Arabic consonants and the linguistic background. Alwabari (2013), Al Mahmoud (2013), and Shehata (2015) found that emphatic and pharyngeal consonants are the most difficult consonants to produce or/and perceive. The current study tested these findings on Arabic learners with a range of L1 languages and examined what consonants are difficult to produce and perceive and what L1 consonants might replace Arabic target consonants. An experimental approach was employed to collect data. Participants were required to read minimal pairs items with emphatic, uvular and pharyngeal consonants in order to test their production and they also completed a discrimination test to evaluate their perception of the target sounds. The results demonstrated that pharyngeals, uvular stops /q/, and emphatics are the most difficult to produce and perceive. The Turkish learners performed better than German and English, while Greek and Chinese were less successful. Learners produced and perceived the sounds differently based on the phonological knowledge available in their native grammar; the findings supports Brown’s feature-based analysis (2000). Learners varied in the choice of the consonants that substitute the target consonants. A brief formal analysis of consonant substitutions using optimality theoretic approach was adopted to analyze the choices made by the learners in the erroneous production.

4.1 Introduction Second language learners usually face difficulty mastering foreign sounds. This learning process causes challenges to the learners who need to produce and perceive new foreign sounds. One of the main factors that influences the production and perception of foreign consonants by 70

The perception and production of Arabic consonants

second language (L2) or third language (L3) learners is the learners’ linguistic background (i.e. the language known by the learners). This study is a report on cross-linguistic second language acquisition in the field of segmental phonology. The area of examination is the accuracy of production and perception of emphatic, uvular, and pharyngeal consonants by different L1 learners and analysis of the performance of different groups of learners phonologically. A review to approaches of second language phonology is laid out in this section. I start from Lado (1957), who argues in the contrastive analysis hypothesis that phonological similarity between L1 and L2 will potentially make the learning experience easier; however, phonological dissimilarity can lead to perception and production errors. Nevertheless, other authors have found that there are other factors that might cause errors such as developmental or universal tendencies with markedness effects, which are prominent in L1 acquisition (Selinker 1972; Eckman 1977, 1991, 2008; Hancin-Bhatt 2008). In the ontogeny model, Major (1987, 2001, 2008) found that L1 has an effect in the initial period of acquisition and this effect decreases during subsequent developmental stages. Flege’s (1995) speech learning model provides an account for the learning of phonetic segments and predicts that similar sounds in L1 and L2 are the most difficult to perceive. Best’s (1995) perceptual assimilation model predicts that the learners perceive non-native sounds based on the similarity between the foreign and native sounds. However, these models do not account for both production and perception and do not account for cases when the learners produce patterns or sounds that are not part of L1 or L2. An optimality theoretical approach (Prince and Smolensky 2004) can account for such possibilities by modeling the interaction of competing factors influencing L2 production and perception. More importantly, Brown (2000) in her model of feature-based analysis argues that the perception of L2 contrast can be controlled by the presence of phonological features in L1. Brown found that if the L1 grammar does not have a certain feature which is needed to differentiate a foreign phonological contrast, the learner might face a difficulty learning this contrast; but if the feature (phonological knowledge) is present, even if it has no phonemic contrast or phonetic realization, the learners might be able to learn the distinction. Brown tested the prediction cross linguistically on Chinese, Japanese, and Korean learners of English. This study had a similar focus to the current one, as it examined the acquisition of phonological contrast by different L1 learners and examined the effect of phonological features on acquiring non-native contrast. The current study might provide an ideal platform to test some predictions of Brown’s theory. Cross-linguistically, several cases have been found in which learners experienced difficulties differentiating between non-native contrasts such as the well-known case of English /l/ and /r/ contrast which is a problem encountered by Japanese learners of English reported in Best and Strange (1992), Yamada (1995), and Guion et al. (2000) among others. While Iverson et al. (2011) investigated the /w/ and /v/ distinction by Hindi learners of English, Lengeris and Nicolaidis (2016) examined the acquisition of English consonants by Greek learners, where /ʃ/ and /ʒ/ were predicted to be difficult to produce or perceive and substituted by /s/ and /z/. Another case is the lack of distinction between the English /p/ and /b/ in the production and perception of Arab learners of English (Alanazi in progress). The majority of L2 studies focus on the acquisition of English consonants by learners from different linguistic backgrounds, and less attention is given to acquisition of less taught languages such as Arabic. The next section provides a review of previous research relating to the acquisition of Arabic consonants and shows the lack of research in Arabic linguistics generally, and in second language phonology particularly. 71

Sara Al Tubuly

4.1.1  Previous linguistic research on Arabic language There are few studies that examine the production or perception of Arabic consonants by learners. For instance, Kara (1976) questioned the claim that English speakers have difficulties learning the Arabic language, based on exploring the problems created by dissimilarities between Arabic and English. The methods used were interviews, questionnaires and class observations. The phonological and grammatical problems were reported to be the most difficult ones. Alosh (1987) compared Arabic learners’ perception and production of two emphatic consonants with native Arabic speakers’ perception. He also examined the learners’ proficiency level using identification and reading tasks and found that there is a strong interaction between language background and the perception of consonantal segments. The current study has a similar focus to Alosh’s work with regard to the language background. Zahid (1996) compared the performance of native speakers of Arabic and French speakers who have no experience of Arabic and found that the French performed poorly in the discrimination task and less consistently in the identification task. Similar to Alosh’s work, Alwabari (2013) examined the influence of learners’ proficiency in Arabic on their ability to produce Arabic pharyngeal and emphatic consonants. The findings showed that learners’ proficiency influenced the production of the consonants. The proficient learners performed better in the production-perception tasks. However, Al Mahmoud (2013) deployed the perceptual assimilation model (PAM) to examine learners’ perception of Arabic contrasts. The findings indicated that English learners were only able to perceive the contrast between consonants that have equivalents in their L1. Recently, Shehata (2015) examined the awareness and perception of adult native English speakers regarding the difficulty of acquiring Arabic consonants. The results showed that the pharyngeal and pharyngealized consonants are the most difficult Arabic consonant contrasts to acquire by learners. Lancaster and Gor (2016) examined the perception of pharyngeal fricatives by English-speaking learners phonologically and phonetically testing the prediction of two models, PAM and automatic selective perception (ASP) model. The findings demonstrated that the learners were more accurate when perceiving the pharyngeals compared to uvulars and when perceiving voiced phonemes compared to voiceless ones. Apart from Zahid’s work (1996), the aforementioned studies concentrated on the L2 acquisition of Arabic consonants by L1 English speakers; they found that emphatic and pharyngeal consonants are difficult to produce and/or perceive for English speakers. This finding needs to be tested with L2 Arabic learners with a range of L1 language backgrounds to provide a phonological account for the variation in accuracy among speakers of different L1s when they acquire non-native consonants and to see whether L2 learners would face difficulty with all non-native consonants which do not exist in their L1 or not. This is what the current study attempts to do. The study also sheds light on how learners with different L1s will replace L2 consonants with different L1 sounds. Brown (2000), Archibald (2005), and Goad and White (2006) argued that certain consonants should be easier for certain L1 speakers to learn. This is because they are part of their phonological inventory; they share similar or closer phonological features with those available in their linguistic system and/or have partial access to some phonological features that are not fully available in their L1 or L2 by the possibility of having some corresponding phonological structures in their L1. The current study examines how phonological features can affect the phonological acquisition by looking at the accuracy patterns of learners of different L1s. In order to determine what consonants are difficult to perceive and produce by different L1 learners and in order to examine the effect of learners’ L1 on production and perception, two experiments were applied to answer the following questions: (1) Which target consonants do 72

The perception and production of Arabic consonants

learners of Arabic perceive and produce accurately based on native-like production and perception of Modern Standard Arabic (MSA), and (2) how do learners’ L1s affect their accuracy in this respect, and what consonants replaced the target consonants in erroneous productions? Therefore, this study compares the three sound sets – emphatics, pharyngeals, and uvulars – with regard to their production and perception by learners from different linguistic backgrounds whose mother tongues are either English, German, Greek, Turkish, or Chinese. Two native speakers took exactly the same tests to check whether native speakers performed successfully as expected, and their performance was taken as a model to compare to. This work contributes to the field of second language phonology of Arabic by learners with different L1s. The chapter is organized as follows: the previous part was an introduction to the topic, including approaches to second language phonology and research questions; the second part includes a consonantal inventory of the target language (Arabic) and the L1 languages; the third section describes the methods used to conduct the study, including information about the participants, tasks, stimuli and the procedures followed; the fourth part contains the data analysis and findings; the fifth part contains the discussion, together with a brief optimality theoretical analysis. The final part is the conclusion.

4.2  Consonantal inventories of the target language and the learners’ L1s 4.2.1  Arabic consonants Some of the discussion here is based on Al-Ani (1970) and Watson (2002). Modern Standard Arabic is the language spoken in formal speech and generally taught in educational school systems as the primary form of the language; it is composed of 28 consonant phonemes. The consonantal inventory of the dialects is slightly different, and it differs from one dialect to another. Considering the consonantal inventory of standard Arabic (see the Introduction to this volume: Table 0.1), it can be noticed that Arabic is characterized by the presence of the emphatics /ṭ/, /ḍ/, /ṣ/, /ð ̣/; pharyngeals /ħ/ and /ʕ/; and the uvulars /q/, /x/, and /γ/. The consonant /v/ is only available in loan words. The /q/ sound is substituted by the sound /ɡ/ in certain Arabic dialects such as Jordanian, Iraqi, Libyan, and Yamani. The lateral /l/ is produced as dark /ƚ/ after emphatic consonants and as an exception in certain words. The alveolar /t/ and /d/ are considered dental in some dialects if they are replacing the /θ/ and /ð/ respectively. The /r/ in Arabic can be produced by a single tap or several successive taps on the same position by the tongue to produce a trill.

4.2.2  English consonants Following Ladefoged (2001) and Roach (2009), English contains 24 phonemes, as shown in Table 4.1. The glottal stop is not a phoneme in English but rather is part of the allophonic variation of /t/ intervocalically or in final position. The velar /x/ exists in Scottish English, but it is replaced by /k/ in English. Apart from the emphatics, pharyngeal consonants, uvular /x/, /γ/, and /q/, the rest of the Arabic consonants have equivalents in British English.

4.2.3  German consonants Standard German consists of 22 consonants displayed in Table 4.2; however, /ž/ occurs in foreign loan words. The /ç/ is an allophone of /x/ and does not exist in all varieties of German. 73

Sara Al Tubuly Table 4.1 English consonants English Consonants

Labial

Stop

pb

Fricative

fv

Dental

Alveolar

θð

sz

Palatal

Velar

Glottal

kɡ šž

h

ʧʤ m

n

Lateral Approximant

Retroflex

td

Affricate Nasal

Palatoalveolar

ŋ

l w

ɹ

y

Note: consonants to the left in pairs are voiceless Table 4.2 German consonants German Consonants

Labial

Alveolar

Stop

pb

td

Fricative

fv

sz

Affricate

pf

ʦ

Nasal

m

n

Lateral

Palatoalveolar

Palatal

Uvular

Glottal

x

h

kɡ š ŋ

l

Trill Approximant

Velar

ʀ w

y

Note: consonants to the left in pairs are voiceless

The sounds /ç/ and /x/ occur in complementary distribution (Kehoe and Lleó 2002). The glottal stop /ʔ/ is allophonic of vowels in some cases and can also precede vowels in some German varieties. The sound /ʀ/ is a uvular trill sound and has the allophonic /ʌ/ and the uvular approximant /ɣ/ (Hall 1993; Wiese 1996).

4.2.4  Greek consonants Standard Greek contains 18 consonants in addition to the /ʦ/, /ʣ/, and /y/ consonants (see Table 4.3) that are not fully considered as phonemes by some Greek linguists, such as Arvaniti (2007). The stops can be nasalized in formal speech as /ᵐb, /ᵐp/ and /ᵑɡ/ (Botinis et al. 2000). There are a number of allophones in Greek. The palatal /c/ and /Ɉ/ are allophonic segments of /k/ and /ɡ/ respectively. In addition, the palatal /ç/ and /ʝ/ are allophonic of /x/ and /ɣ/, respectively (Mennen and Okalidou 2006).

4.2.5  Turkish language According to Topbaş (2007), Standard Turkish has 20 phonemes, as demonstrated in Table 4.4. The /c/ and /Ɉ/ are allophones of /k/ and /ɡ/, correspondingly. The /v/ sound has the allophone /ʋ/ and the lateral /l/ has the dark emphatic /ƚ/ as an allophone. The voiceless glottal fricative /h/ can be recognized as /x/. 74

The perception and production of Arabic consonants Table 4.3 Greek consonants Greek Consonants

Labial

Dental

Stop

pb

td

Fricative

fv

θð

Alveolar

Velar

Uvular

Glottal

kɡ sz

Affricate Nasal

Palatal

xɣ

ʦʣ m

n

Lateral

l

Tap/flap

ɾ

Approximant

y

Note: consonants to the left in pairs are voiceless

Table 4.4 Turkish consonants Turkish Consonants

Labial

Alveolar

Stop

pb

td

Fricative

fv

sz

m

n

Palatoalveolar

Glottal

kɡ šž

Affricate Nasal

Velar

h

ʧʤ

Lateral

l

Flap/tap

ɾ

Note: consonants to the left in pairs are voiceless

4.2.6  Chinese consonants Mandarin Chinese has 19 phonemes, excluding /ɕ/, /ʨ ͪ/, and /ʨ/. The Chinese consonants are distinguished in terms of their aspiration. They are different from consonants in the other languages, where the difference between the pairs of sounds lies in the voice feature. The stops /p ͪ /, /t ͪ /, and /k ͪ / for instance have the equivalent un-aspirated sounds /p/, /t/, and /k/, and this also applies to the affricates, as shown in Table 4.5. There is an argument as to whether /r/ should be replaced by /ʐ /. Only /n/ and /ŋ/ can occupy the final position in the syllable. The un-aspirated consonantal stops and affricates /p/, /t/, /k/, /ʦ/, and /tʂ/ might be devoiced to /b/, /d/, /ɡ/, /ʣ/, and /dʐ/ respectively, in certain phonological environments (Duanmu 2002). After considering the consonantal inventory of the five languages, English, German, Greek, Turkish, and Chinese, we can divide the sounds into two groups: (1) those found in Arabic but not in the other five languages and (2) those which are found in the other languages but not in Arabic. The focus of this study is on emphatic, uvular, and pharyngeal consonants that have no identical or similar counterparts in the L1s (or the L2, English) of the learners. Thus, the other group of consonants which do not exist in Arabic is not of concern. Table 4.6 shows how Arabic differs from these languages with regard to the consonants under investigation. 75

Sara Al Tubuly Table 4.5 Chinese consonants Chinese Consonants

Labial

Alveolar

Stop

pͪ p

tͪ t

Fricative

f

s

ʂ

ʦͪ ʦ

tʂ ͪ tʂ

Affricate Nasal

m

Palatoalveolar

Retroflex

Palatal

Glottal

kͪ k x

n

Lateral

Velar

ŋ

l

Approximant

r

Note: consonants to the left in pairs are aspirated Table 4.6 Emphatic, velar/uvular and pharyngeal consonants Languages

Emphatics

Arabic English German Greek Turkish Chinese

/ṭ/ _ _ _ _ _

/ḍ/ _ _ _ _ _

Velar /Uvular /ṣ/ _ _ _ _ _

/ð ̣/ _ _ _ _ _

/q/ _ _ _ _ _

Pharyngeal /x/ _ /x/ /x/ _ /x/

/γ/ _ /ʀ/ /γ/ _ _

/ħ/ _ _ _ _ _

/ʕ/ _ _ _ _ _

4.2.7  Target forms Arabic is part of the Semitic language family and it is different from other language families such as the Germanic, the Hellenic, the Turkic, and the Sino-Tibetan, to which English and German, Greek, Turkish, and Chinese belong, respectively. The language contact between both Arabic and Turkish during the Ottoman Empire and Arab civilization has resulted in borrowed vocabulary. However, the phonemic system has not been influenced. For example, a loan word like /ramaḍaan/ fasting month would be pronounced as /ramazaan/ in Turkish, with a substitution of the emphatic sound /ḍ/ with /z/. Other Semitic languages, such as Hebrew and Maltese, have slightly lost some of the emphatic consonants, probably due to the influence of other languages on them. The emphatic consonants in Arabic have plain equivalents that share similar places of articulation, as shown in Table 0.1. Emphatics are argued to have two places of articulation: the primary coronal and the secondary pharyngeal articulation; the latter occurs by retracting the tongue toward the pharynx (Al-Ani 1970; Zawaydeh 1998). In fact, the emphatic, uvular, and pharyngeal consonants are characterized by tongue retraction in general; however, the tongue is lowered more in emphatics than uvular consonants. Uvulars have a more retracted tongue than emphatics (Ghazeli 1977; Al-Solami 2013), while the root of the tongue is retracted in toward the pharynx back wall in pharyngeal consonants (Al-Ani 1970). The pharynx is the primary place of articulation for pharyngeal consonants, where the narrowing is formed in that area between the root of the tongue and the pharynx wall, causing a friction sound (Laufer and Baer 1988). 76

The perception and production of Arabic consonants

The uvular sounds /x/, /γ/, and /q/ are produced when the back and root of the tongue (dorsum) is against the uvula, which gives them the feature [dorsal] and [pharyngeal] as primary articulations. There is a narrowing in /x/ and /γ/ causing the friction and giving the feature [continuant]. The sound /γ/ has the [voice] feature due to the vibration of the vocal cords, but the closure in /q/ leads to a release phase, and therefore it is [-continuant] and [-voice]. The pharyngeal consonants/ħ/ and /ʕ/ are produced by forming a narrowing between the root of the tongue and the vertical back wall of the throat, making the feature [pharyngeal] active; /ʕ/ has a [voice] feature as it is produced with a vibration of the vocal cords. The emphatics have the same articulation of /t/, /s/, /ð/, and /d/, so the [coronal] feature exists, but the articulation is accompanied by the retraction of the root of the tongue to the back of the throat, giving the feature [pharyngeal] (McCarthy 1994; Zawaydeh 1999). The acoustic properties of emphatics, pharyngeal, and uvulars are based on Al-Ani (1970). The emphatic stops are accompanied by burst duration of around 20–30 ms, usually at the frequency range of 1500–2400 Hz. The emphatic fricatives have noise duration of around 100–160 ms, with a first, second, and third formant at around 275, 1000, and 2350 Hz, respectively. The uvular stop /q/ is accompanied by burst around 3000 Hz with a duration of around 30–40 ms. The uvular fricatives have noise duration of around 100–160 ms and frequency range between 1300–1500 Hz, while pharyngeals have duration of 100–150 ms. The uvular, pharyngeal, and emphatic consonants are different phonetically (in terms of articulation and acoustics) and phonologically.

4.2.8  Phonemic contrast The words in Table 4.7 represent minimal pairs between emphatic, uvular, and pharyngeal consonants and their counterpart segments. The only difference between these words is the phonemic contrast that results in meaning changes. The target segments and their counterparts are also listed in Tables 4.9 and 4.10 to utilize the forms that will be used in our study to test the production and perception of the sounds.

4.3 Methods 4.3.1 Participants Fifty-two participants took part in the production and perception experiments. The participants were native speakers of English, German, Greek, Turkish, and Chinese aged from 17 to 49. There were 10 participants in each language group. More details can be found in Table 4.8. Table 4.7 Emphatic, uvular, and pharyngeal consonants Pair

Example

Gloss

Example

Gloss

/t/ and /ṭ/ /d/ and /ḍ/ /ð/and /ð ̣/ /s/ and /ṣ/ /h/and /ħ/ /ʔ/ and /ʕ/ /k/ and /q/ /x/ and /ɣ/

(1) /ti:n/ (2) /darb/ (3) /ðill/ (4) /sayf/ (5) /habb/ (6) /ʔamal/ (7) /kalb/ (8) /xaab/

“fig” “road” “humiliation” “sword” “blow” “hope” “dog” “failed”

 (9) /ṭiin/ (10) /ḍarb/ (11) /ð ̣ill/ (12) /ṣayf/ (13) /ħabb/ (14) /ʕamal/ (15) /qalb/ (16) /ɣaab/

“mud” “hit” “shadow” “summer” “spots” “job” “heart” “absent”

77

Sara Al Tubuly Table 4.8 Details of the participants Language

English German Greek Turkish Chinese

Age

Gender

Mean

Male

Female

Count

Count

4 6 5 3 4

6 4 5 7 6

23 23 26 31 25

These languages were chosen for comparison, because they vary in having/not having the various post-velar features active in their phonology. All five languages do not have pharyngeal features, and only the German and Greek languages have the uvular feature. They provide a good sample to test and answer the research questions. Two female Jordanian native speakers were recruited to act as a control group; they were educated in standard Arabic up to master degree level. All participants had knowledge of another foreign language, mainly English. Some of the learners took introductory courses in Arabic in their own country and continued learning Arabic in the UK while studying for other degrees such as international studies, politics, law, and religious studies. None of the participants included in the current research have studied Arabic in an Arabic-speaking country or stayed there for more than two weeks. They gave their consent and permission to use the data anonymously. They also completed a questionnaire to collect information about their previous linguistic background and any other information that might affect their performance. There were no reported hearing difficulties. The participants had studied standard Arabic for one year (with a total of 192 instruction hours for the whole year) in England or Scotland, covering the four core skills: speaking, listening, writing, and reading. They passed the introductory level with a score of no less than 55%. The teachers are native speakers of Jordanian, Syrian, and Libyan Arabic (from whom the learners received their input). They are qualified teachers and highly proficient in standard Arabic; they showed no dialectal effect on their pronunciation.

4.3.2  Experiment 1: perception Eight minimal pairs were used in an AXB task to test perception. The minimal pairs used were /t, ṭ/, /d, ḍ/, /ð, ð ̣/, /s, ṣ/, /h, ħ/, /ʔ, ʕ/, /k, q/, and /x, ɣ/. Four distractors were included occasionally, but the responses to these are not reported here. Four trials were generated for each pair (AAB, ABB, BAA, BBA) where /B/ is the target utterance with the pharyngeal/pharyngealized consonant. For example, in case of /x, ɣ/, the former is part of A and the latter is part of B. One sequence for each consonant was recorded by a Libyan male native speaker of Arabic who has finished his undergraduate study in an Arabic-speaking country. The sequences were produced correctly, and no dialectal effect was noticed during the process of recording. The recordings were edited using Gold Wave software to form the AXB sequences (12 × 8 = 96) as in Table 4.10 so the listeners can respond to the same phonetic realization of each consonant each time. The production was recorded using an internal microphone on the M-Audio Microtrack-II professional two-channel digital recorder. The device was set to 44.1 kHz and bit size 16. The learners were required to listen to each trial played on a laptop using Gold Wave 78

The perception and production of Arabic consonants Table 4.9 Perception stimuli /t/ – /ṭ/

/d/ – /ḍ/

/ð/ – /ð̣/

/s/ – /ṣ/

AAB ABB BAA BBA

/ata/, /ata/, /ɑṭɑ/ /ata/, /ɑṭɑ /, /ɑṭɑ/ /ɑṭɑ/, /ata/, /ata/ /ɑṭɑ/, /ɑṭɑ/, /ata/

/ada/, /ada/, /ɑḍɑ/ /ada/, /ɑḍɑ/, /ɑḍɑ/ /ɑḍɑ/, /ada/, /ada/ /ɑḍɑ/, /ɑḍɑ/, /ada/

/aða/, /aða/, /ɑð ̣ɑ/ /aða/,/ɑð ̣ɑ/, /ɑð ̣ɑ/ /ɑð ̣ɑ/, /aða/, /aða/ /ɑð ̣ɑ/, /ɑð ̣ɑ/, /aða/

/asa/, /asa/, /ɑṣɑ/ /asa/, /ɑṣɑ/, /ɑṣɑ/ /ɑṣɑ/, /asa/, /asa/ /ɑṣɑ/, /ɑṣɑ/, /asa/

AAB ABB BAA BBA

/h/ – /ħ/ /aha/, /aha/, /aħa/ /aha/, /aħa/, /aħa/ /aħa/, /aha/, /aha/ /aħa/, /aħa/, /aha/

/ʔ/ – /ʕ/ /aʔa/, /aʔa/, /aʕa/ / aʔa/, /aʕa/, /aʕa/ /aʕa/, /aʔa/, /aʔa/ /aʕa/, /aʕa/, /aʔa/

/k/ – /q/ /aka/, /aka/, /aqa/ /aka/, /aqa/, /aqa/ /aqa/, /aka/, /aka/ /aqa/, /aqa/, /aka/

/x/ – /ɣ/ /axa/, /axa/, /aɣa/ /axa/, /aɣa/, /aɣa/ /aɣa/, /axa/, /axa/ /aɣa/, /aɣa/, /axa/

software and headphones. Each trial is composed of three items, and the pairs were arranged randomly. The learners were required to decide whether the second item is similar to the first or the third one and record their response on an answer sheet. The responses were statistically analyzed using SPSS; the dependent variable is accuracy (i.e. correct identification of X as being a token of A or B). The experiment was self-paced and the participants were told to ask for a break if they needed it to maintain concentration. The purpose of this task is to test whether the learners can differentiate the minimal pairs.

4.3.3  Experiment 2: production The learners were asked to read minimal pairs (16 real words, 16 nonsense words composed of CVC for each sound, as /ṭaṭ/, /tat/, /qaq/ and /kak/, and four distractor words) at a normal speed and required to read each word three times in isolation, then move on to the next word (see Table 4.10). The production was recorded using an internal microphone on the M-Audio Micro-track-II professional two-channel digital recorder. The device was set to 44.1 kHz and bit size 16. The target words are first copied from the recorder to Gold Wave software; the items are prepared and processed by determining the onset and offset of the second item produced audibly and then transferred to Praat for further evaluation (Boersma and Weenink 2014). The recordings of all items were analyzed using Praat, with a sample rating of 8000 Hz on the researcher’s computer. The words were evaluated and transcribed by two linguists. The reliability of assessment of the judges was checked using an absolute agreement scale. A limit of 60% agreement is required between the judges during the evaluation of the items. The judges are the author and another linguist who are native speakers of Jordanian and Libyan Arabic; they are educated at a PhD level and qualified teachers. The items were prepared and processed by the author and sent with Praat images to the second judge, who evaluated the accuracy blindly in each word, based on native-like production of target consonants by the control group. They also identified the consonants that were produced inaccurately by the learners. For further evaluation, some items were also evaluated by peer review in a phonology research group held at the University of Essex. Pictures were not used in this study, as it might lead to learners not producing the target item and producing a synonym instead (Al Tubuly 2016). Therefore, one word at a time appeared on a computer screen, presented in Arabic script, and they were displayed in pseudo-random order; the experiment was self-paced. The target consonants occupy the position of the onset in the words, as it is a noticeable 79

Sara Al Tubuly Table 4.10 Production stimuli of the real words Arabic

Item

Gloss

Arabic

Item

Gloss

‫تَاب‬ ‫دَار‬ ‫َذل‬ ‫َسار‬ ‫هَب‬ ‫أَلم‬ ‫كَال‬ ‫خَاب‬

  (1)/taab/   (3)/daar/   (5)/ðall/   (7)/saar/   (9)/habb/ (11)/ʔalam/ (13)/kaal/ (15)/xaab/

“repented” “house” “humiliated” “walked” “blew” “pain” “weighed” “failed”

‫طَاب‬ ‫ضار‬ َ ‫ظَل‬ ‫صار‬ َ ‫َحب‬ ‫عَلم‬ ‫قَال‬ ‫غَاب‬

  (2)/ṭaab/   (4)/ḍaar/   (6)ð ̣all/   (8)/ṣaar/ (10)/ħabb/ (12)/ʕalam/ (14)/qaal/ (16)/ɣaab/

“cooked” “harmful” “stayed” “became” “pills” “flag” “said” “absent”

position (Archibald 1998). The procedures of the experiment were explained to the learners in advance. The experiments took place in a quiet place and were monitored by the author. Four items were used to practice before the experiments, to prepare the participants and make them familiar with the experimental procedure; the items were also displayed in a randomized order.

4.4  Data analysis and results The two native speakers achieved 100% in the production and perception, as expected, and their performance is considered as a native target model. The responses by the learners in the production and perception tasks were coded and entered into SPSS. The dependent variables are the accuracy of producing and perceiving the emphatic, pharyngeal, and uvular consonants. The measure of accuracy used here is presented as a value between 0 inaccurate and 1 accurate. The independent variables are the five languages as the main effects. Only significant results of Wald chi square were reported in production and perception sections. A descriptive analysis including means, standard deviations, and percentages was carried out, as shown in Table 4.11.

4.4.1  Perception results To determine which target consonants the learners of Arabic perceive more or less accurately, the overall percentage of correct scores for the 16 consonants have been rank ordered from the easiest to the most difficult and are provided in Table 4.12. It seems that learners successfully perceive the non-emphatic stop and fricative consonants /t/, /d/, /s/, /ð/, the glottal consonants /h/, /ʔ/, and the velar stop /k/ in accordance with the findings of Shehata (2015) and Al Mahmoud (2013), among others. The learners also have not demonstrated difficulty in perceiving the uvular fricatives /x/ and /γ/. The majority of pharyngeal and emphatic consonants were found to be more difficult to perceive among the learners. However, the /ṭ/ and /q/ consonants were best perceived by the learners comparing to other emphatics, uvular, and pharyngeal consonants, while /ð ̣/ and/ħ/ were less accurately perceived. Focus here is on the group of sounds that are considered difficult or showed lower accuracy rates, as shown in Table 4. 12.

4.4.1.1  Emphatic consonants In order to examine the data with respect to the second research question, bar charts were created to display the mean accuracy in perception of consonants by L1 group. Consider the 80

The perception and production of Arabic consonants Table 4.11 Descriptive statistics associated with the perception and production of consonants across all participants Consonants

/t/ /ṭ/ /d/ /ḍ/ /s/ /ṣ/ /ð/ /ð ̣/ /h/ /ħ/ /ʔ/ /ʕ/ /k/ /q/ /x/ /γ/

Perception

Production

Mean

Std. Deviation

Mean

Std. Deviation

1.00 .74 .88 .56 .88 .58 .78 .36 .72 .38 .84 .56 1.00 .64 .98 .90

.000 .443 .328 .501 .328 .499 .418 .485 .454 .490 .370 .501 .000 .485 .141 .303

1.00 .44 .96 .54 1.00 .58 .82 .50 .86 .26 1.00 .34 1.00 .26 .86 .82

.000 .501 .198 .503 .000 .499 .388 .505 .351 .443 .000 .479 .000 .443 .351 .388

Table 4.12 Overall correct percent scores in perception Perception

Percentage %

/t/ /k/ /x/ /γ/ /d/ /s/ /ʔ/ /ð/ /ṭ/ /h/ /q/ /ṣ/ /ḍ/ /ʕ/ /ħ/ /ð ̣/

100 100 98 90 88 88 84 78 74 72 64 58 56 56 38 36

performance of the learners of the five language groups with regard to the perception of the four emphatic consonants /ṭ/, /ḍ/, /ṣ/, and /ð ̣/. It can be observed in Figure 4.1 that the Turkish speakers performed better than the rest of the learners, while the Chinese speakers performed worst. German, English, and Greek speakers performed at roughly the same level, but German speakers were slightly better, followed by English speakers. Descriptively, the voiceless emphatic stop /ṭ/ was perceived best by the learners but was not found statistically significant. On the other hand, the difference in the perception of /ḍ/ among the language groups, 81

Sara Al Tubuly

Figure 4.1 Mean accuracy in perception of emphatic consonants by L1 group

especially between the Chinese speakers and the rest of the learners, is statistically significant (Wald chi square = 9.465, p = 0.050) which means that the perception of /ḍ/ has been affected by the language groups. The results show that the perception of voiceless emphatic /ṣ/ was better than the voiced /ð ̣/ and voiceless stop /ṭ/ was better than the voiced stop /ḍ/ in the five groups. The Chinese, English, and Greek speakers were equal in their perception of /ṣ/; however, the Chinese were the least successful and the Turkish were the most successful in the perception of /ṭ/, /ḍ/, and /ð ̣/, as is illustrated in Figure 4.1.

4.4.1.2  Pharyngeal and uvular consonants It seems that the perception of the uvular fricatives is largely unproblematic. As Figure 4.2 shows, there is a noticeably high performance by the learners of all language groups in their discrimination between /x/ and /γ/. In the perception of these two consonants, no one group of language speakers has performed obviously better than the others. However, it can be noticed that the learners find it difficult to perceive the voiceless pharyngeal fricative /ħ/, especially the Chinese and Greek learners. Moreover, the voiceless uvular stop /q/ was not easily discriminated from /k/ by the Chinese participants, giving a marginally significant result of Wald chi square = 8.990, p = 0.061. The results indicated that the perception of /q/ was affected by the language group.

4.4.2  Production results A test (Cronbach alpha) of the data’s internal consistency obtained from the two judges who evaluated the production test gave an overall score of 0.821. This implies that the agreement between the judges in the current study is strong (Larson-Hall 2010). Similar to perception, to 82

The perception and production of Arabic consonants

Figure 4.2 Mean accuracy in perception of pharyngeal and uvular consonants by L1 group

determine which target consonants the learners of Arabic produce accurately, the overall percentage of correct scores for the 16 consonants are rank-ordered from the easiest to the most difficult and are provided in Table 4.13. The learners successfully produced non-emphatic stop and fricative consonants /t/, /d/, /s/, /ð/, the glottal consonants /h/, /ʔ/, and the velar stop /k/. The voiceless uvular fricative /x/ was produced correctly in 86 % of the responses, while /γ/ was produced correctly in 82% of the responses. All the pharyngeal and emphatic consonants were found to be more difficult to produce among the learners. Therefore, focus here is on the group of sounds that are considered difficult or showed lower accuracy rates, as shown in Table 4.13.

4.4.2.1  Emphatic consonants In order to examine the data with respect to the second research question, bar charts were created to display the mean accuracy in production of consonants by L1 group. The results are interesting, as the production of the emphatics by Chinese learners mirror those in the perception task, with a minimal difference. Another aspect is that the success ranking of language group speakers has been maintained, in which the Turkish, followed by the German learners, performed better than the English, Greek, and Chinese learners, respectively, as shown in Figure 4.3. Statistically, there was a significant difference in the accuracy rates for the voiced emphatic stop /ḍ/ between certain language groups, giving a Wald chi square of 9.841 (p = 0.043), which proves that the accuracy of producing /ḍ/ was affected by the language groups.

4.4.2.2  Pharyngeal and uvular consonants The pharyngeals and uvular stop were not produced successfully by the learners of the language groups. The results demonstrated that the language group affected the production of the 83

Sara Al Tubuly Table 4.13 Overall correct percent scores in production Production

Percentage%

/t/ /k/ /s/ /ʔ/ /d/ /h/ /x/ /ð/ /γ/ /ṣ/ /ḍ/ /ð ̣/ /ṭ/ /ʕ/ /q/ /ħ/

100 100 100 100 96 86 86 82 82 58 54 50 44 34 26 26

Figure 4.3 Mean accuracy in production of emphatic consonants by L1 group

voiceless uvular stop /q/ (Wald chi square = 10.190, p = 0.037). Descriptively, overall the Turkish speakers showed better pronunciation of the pharyngeals than speakers of other languages. It can be noticed that English speakers were less accurate than the rest in producing uvular fricatives, as can be seen in Figure 4.4. In production, it seems that the current study provides evidence for differential substitutions of the target consonants among learners. The learners with different L1s substitute 84

The perception and production of Arabic consonants

Figure 4.4 Mean accuracy in production of pharyngeal and uvular consonants by L1 group.

different sounds for the L2 target consonant. The sound /q/ was mainly replaced by /k/, while /ħ/ was replaced by /h/ or /x/ and /ʕ/ was replaced by /a/ or /ʔ/. The emphatic /ṭ/ was produced plainly as /t/, and /ð ̣/ was substituted by /z/, /d/, or /s/, while the place of /ṣ/ was taken by /s/. The /ḍ/ was substituted by /d/, /z/, or /s/, as shown in Table 4.15. Comparing the perception and production results, it can be noticed that consonants which have relatively high accuracy scores on perception also score relatively highly in the production task, but this does not apply to /ṭ/, /ð ̣/, /ʕ/, and /q/. The Pearson correlation is strong between perception and production: .971 (p < .001); (t = 1.508, p = .152). Perception is slightly more successful than production: means 73.7 versus 70.2. Apart from the stop /q/, the uvulars are easier to produce and perceive. The pharyngeals are slightly more difficult to produce and perceive compared to emphatics. The Turkish learners were the best performers, followed by German and English learners. The Greek learners, followed by Chinese, were less successful in their performance.

4.5 Discussion 4.5.1  Main findings This study shows how and whether native speakers of various languages are different or similar in their production and perception of certain Arabic consonants. The results of the experiment show that the five groups of speakers of different L1s found it relatively difficult to perceive and produce emphatic and pharyngeal consonants. Their different L1s affected the accuracy of production and perception of the target consonants. Following Brown’s model (2000), this performance is mainly interpreted as due to the fact that the phonological feature [pharyngeal] is absent from their phonological systems; it is in fact not a common feature in most languages (McCarthy 1994). Therefore, having the [pharyngeal] as the primary phonological feature in /ħ/, /ʕ/ and the secondary one in /ṭ/, /ḍ/, /ð ̣/, and /ṣ/ results in difficulties in 85

Sara Al Tubuly Table 4.14 Substitution of target consonants /ṭ, ḍ, ṣ, ð ̣, ħ, ʕ, q, x, γ/ split by participant groups Participant Groups English German Greek Turkish Chinese

ṭ t ḍ d z s ṣ s ð̣ z

s d ħ h x ʕ ʔ a q k g x

2 4 4 7 5

0 0 0 0 6

8 6 6 3 5

4 8 4 8 3

6 2 6 0 1

0 0 0 2 1

0 0 0 0 5

5 7 3 9 5

5 3 7 1 5

4 5 5 7 4

0 0 2 3 0

6 5 3 0 0

2 3 1 5 2

8 7 0 5 0

0 0 9 0 8

3 2 3 7 2

7 8 7 3 0

0 0 0 0 8

1 2 2 7 1

9 8 8 3 9

k γ

x k

0 5 5 6 0 4 0 10 0 9 1 0 0 10 0 10 0 0 1 8 2 9 1 0 0 10 0 7 3 0

achieving a native-like pronunciation or perception of these sounds because it involves pharyngeal constrictions and pulling the tongue root toward the pharynx wall (Ladefoged 2001). Moreover, learners are probably able to acoustically perceive the distinction between /ṭ/ and /t/ more accurately than /ð ̣/ and /ð/, probably because the emphatic and plain fricatives are produced with a similar noise duration of 100–160 ms while the emphatic stops are produced with lower frequency, at around 1500–2400 Hz, than that of their counterparts. This group of sounds therefore has a phonetic and phonological complication, where the learners might simply fail to produce them due to the complexity of their articulation and fail to perceive them due to the lack of similar phonological representations in their L1. In contrast, the uvular fricatives were found to be easier and more accessible for the learners, particularly German, Greek, Chinese, and Turkish, because the [uvular] feature is part of the internalized phonological system in the grammar of Greek and German languages, and perhaps because Chinese and Turkish speakers have already been exposed to this feature through velar /x/. This exposure allows possible sound groups to enter their own phonological space without causing clashes with the native grammar (Brown 2000). Looking at individual groups, Chinese speakers, as shown in Figures 4.1–4.4, were placed at the bottom rank in the hierarchy of the learners’ performances. As mentioned earlier, in the Chinese sound system, stop and affricate consonants are distinguished in terms of aspiration but not in terms of voice. They are different from the other languages in the current study, as the other languages possess a distinction between the pairs of sounds based on the voice feature, while in Chinese, stops and affricates have equivalent unaspirated sounds. So, Chinese speakers lack both the [laryngeal] and [pharyngeal] features. In order to encode the contrast between pharyngeals and emphatics and their plain counterparts and to encode the contrast between the voiced and voiceless pharyngeals and emphatics, the [laryngeal] feature should be active, but it seems that they have either no access or limited access to both features. This means that both pharyngeal and sub-pharyngeal contrasts might be difficult to perceive and presumably to produce by Chinese speakers. However, they had no difficulties with the uvular fricatives, as they probably have a full access to [velar] and partial access to [uvular] features, but they were not able to perceive and produce /q/ accurately. One might suggest that they do not have access to the combination of [uvular/-continuant] in order to acquire /q/. The [uvular] feature is not part of the Chinese L1 grammar, but the [velar/±continuant] is. Brown (2000) argues that learners might be able to acquire foreign sounds if they have partial access to certain features. The Chinese speakers have a full access to [velar/±continuant] which probably allows a partial access to [uvular] feature. By comparing the three languages German, English, and Greek, it can be noticed that the German system is enriched with more phonological features at the back of the oral cavity. As [velar], [uvular], and [glottal] are active in German, it makes it possible for the Arabic features to be slightly accessible. Therefore, one might suggest that German speakers employed the 86

The perception and production of Arabic consonants

features available in their L1 to absorb this contrast, which consequently enabled them to perform better than speakers of the other languages. German learners would be advantaged by the fact that they have a phonemic contrast involving [uvular] in /x/ and /ʀ/ more than learners who just have [x] in their inventory. However, fewer corresponding posterior phonological features in English and Greek might have led to lower performance by their speakers. Only features [velar] and [glottal] are available in English and only [velar] and [uvular] are available in Greek. This supports Brown’s (2000) argument that the difficulty of learning a foreign sound is subject to whether the L1 has identical or closer phonological features to absorb the contrast and distinction related to the sound. In fact, the choice of the sound substitutions in the production errors is subject to phonological features available in the L1 (Brown 2000; Hancin-Bhatt 1994). For example, in the current study, the voiceless pharyngeal fricative /ħ/ was replaced by /x/ in errors made by Greek speakers but by /h/ in errors made by English speakers. This is perhaps because both language groups do not have access to the feature [pharyngeal] and the sound is substituted by what is present in the learners’ L1 grammar. The presence of [glottal] allows English learners to produce /ħ/ as /h/, while the presence of [uvular] allows Greek learners to produce /ħ/ as /x/. The native grammar of English learners does not prefer /ɡ/ over /k/ as a substitute for the Arabic /q/. It seems that /k/ is closer to /q/ because they share the phonological features of [-continuant] and [-voice], but they differ in the [velar] and [uvular] features, while /ɡ/ and /q/ share only the [-continuant] feature and differ in the rest. German and Greek learners do not substitute the Arabic /q/ with a uvular fricative. One might suggest that the native grammar of the learners preserves the manner feature over place. Turkish speakers performed relatively better than other language groups. My interpretation of this is based on different aspects. Linguistically, one might claim that Turkish speakers absorb the distinction between some of the sounds based on the phonological features accessible to the learners from some allophonic variations that occupy different environments (Brown 2000; Flege 1995). In particular, Turkish language has various allophones that are articulated in the posterior part of the mouth such as /c/, /ɟ/, /ƚ/, /ŋ/, /ɲ/, and /x/. Arabic loan words in Turkish that contain emphatic sounds such as /ṭ/ have a velarized /tˠ/, which means that a trace was kept from the borrower language. Therefore, I presume that the Turkish learners are slightly aware of the distinction between pairs of sounds due to language contact and other geographical, religious, and historical aspects that positively influenced their general performance, as they possibly had more exposure to standard Arabic through the citation of the Qur’an. The Turkish, German, and English learners usually tend to replace the consonants with their counterparts, but there were occasions when the Greek and Chinese learners substituted them with other consonants. These consonants tend to share similar phonological features available in their L1s. The sound /ð ̣/ and /ḍ/ were substituted by voiceless /s/ by Chinese learners. This can be attributed to the fact that the Chinese’s grammar lacks [laryngeal] feature; therefore, the voiceless /s/ was produced instead of their voiced counterparts. Interestingly, the emphatic /ḍ/ was substituted by /d/, /z/, or /s/. The /ḍ/ sound can be substituted in some spoken dialects by /d/ or pharyngealized /ẓ/. For example, speakers of Sudanese and Egyptian dialects might substitute /ḍ/ by /ẓ/, and speakers of Libyan dialect might substitute /ḍ/ by /d/. The word ḍaabiṭ “officer” might be pronounced as ẓaabiṭ and daabiṭ, respectively. It seems that certain consonants are easier for certain L1 speakers to learn. This is because they are part of their phonological inventory or because they share similar or closer phonological features with those available in their linguistic system. Brown (2000) and Archibald (2005) argue that having partial access to some phonological features that are not fully available in their L1 or having similar phonological structures in their L1 might enable learners to acquire sounds easily. 87

Sara Al Tubuly

It would be interesting to explore the reasoning behind the learners’ choices in their production grammar. In attempting to do so, a brief analysis using an optimality theoretic approach is adopted here to examine the choice of producing /x/ by Greek and Chinese learners instead of /h/ to replace /ħ/, the choice of /k/ to replace /q/ by other learners, and the choice of producing /s/ to replace /ḍ/ by Chinese learners.

4.5.2  Formal analysis of consonant substitutions Optimality theory (OT) is used to analyze patterns in language acquisition, both in terms of the L1 and the L2. Languages or grammars of learners are seen as resulting from the interaction of a set of universal constraints that can be violated. The difference between languages, including that between the language of L2 learners and native speakers, is represented in the variation of constraint ranking (Prince and Smolensky 2004). Second language learners need to re-rank the constraints found in their L1 to be in agreement with the constraints in the L2. In other words, they need to activate any inactive constraints in their L1 if they are fundamental to the L2 grammar. From an OT perspective, during learning process, the focus is always on the constraints re-ranking that are determined by the mapping or the mismatch between the input (target language) and the output (learners’ performance). In an applied and practical sense, learners need to be exposed to the language more, with a focus on their individual learning needs to facilitate the re-ranking of the constraints or rearrangement of their grammar. Within an OT framework, I will appeal to the features I previously mentioned but in the form of markedness constraints. The constraint *PHARYNGEAL resists the presence of sounds that are articulated in this region, whilst the *UVULAR is a constraint that resists the presence of uvular sounds (Lombardi 2001). However, regardless of the learners’ background, uvular fricatives were accepted in their grammar but not uvular stops; therefore, the *[q] constraint which prohibits uvular stop /q/ will be deployed. These markedness constraints will interact with the faithfulness constraint IDENT, which requires that the output form to be identical to the native form (input) (Wester et al. 2007; McCarthy 2008). There are two levels of performance: ranking for native speakers and ranking for the learners which happens through the learning process. The learner’s ranking can be language specific; therefore, *GLOTTAL, which prohibits the presence of glottal consonants, is an active constraint in the grammar of the Greek and Chinese learners’ L1; consequently, glottal consonants would not be substitutes to pharyngeal consonants, as illustrated in Tableau 4.5. Similarly, it is applied to the [laryngeal] feature that lends voice and can be added to the language-specific ranking of Chinese to disallow /d/ or /ḍ/ where *LARYNGEAL constraint can ban voiced consonants, as in Tableau 4.6. A violation of a higher ranked constraint will lead to a fatal exclusion for the consonant from the computation in the grammar of the language. However, when the constraint is lower ranked, this will give the segment a chance to be the optimal output. In Tableau 4.1, the grammar of the L2 allows [uvular] and [-continuant]; therefore,*[q] is lower ranked. The *PHARYNGEAL constraint is lower ranked to allow pharyngeal consonant to surface and to be optimal output, as in Tableau 4.3. However, in Tableaux 4.2 and 4.4, the*PHARYNGEAL and *[q] outrank IDENT to exclude /ħ/ and /q/. The rank of *PHARYNGEAL >> IDENT can be deployed to account for the mispronunciation of emphatics, as the feature [pharyngeal] is a primary in /ħ/ and /ʕ/ and secondary in /ṭ, ḍ, ṣ, ð ̣/. Learners with different L1s replace L2 consonants with different L1 sounds as in Tableaux 4.5 and 4.6, as mentioned earlier. This can be attributed to the fact that the features needed for /ħ/ and its counterpart /h/ do not exist in the L1 grammar of Greek and Chinese learners; they 88

The perception and production of Arabic consonants

will try to reach the accessible alternatives. *PHARYNGEAL and *GLOTTAL are higher ranked to allow /x/ to surface as optimal output. Similarly, Chinese participants tend to substitute the target consonants with voiceless consonants as shown in Table 4.15. For example, the /ḍ/ and its counterpart /d/ are voiced; therefore, *LARYNGEAL constraint is higher ranked, with *PHARYNGEAL to allow /s/ to be optimal candidate in Tableau 4.6. Tableau 4.1 Native speakers’ grammar accepts /q/. /q/

IDENT

q k

*!

*[q] *

Tableau 4.2 Learners’ grammar refuses /q/. /q/

*[q]

q

*!

IDENT

k

*

Tableau 4.3 Native speakers’ grammar accepts /ħ/. /ħ/

IDENT

*PHARYNGEAL *

ħ h

*!

Tableau 4.4 German, English, and Turkish learners’ grammar refuses /ħ/ /ħ/ ħ

*PHARYNGEAL *!

IDENT *

h Tableau 4.5 Greek and Chinese learners’ grammar refuses /h/ and /ħ/. /ħ/

*PHARYNGEAL

ħ

*!

*GLOTTAL

h

IDENT

*! *

x Tableau 4.6 Chinese learners’ grammar refuses /ḍ/ and /d/. /ḍ/

*PHARYNGEAL

*LARYNGEAL

ḍ

*!

*

d

IDENT

*! *

s

89

Sara Al Tubuly

Tableaux 4.2, 4.4–4.6 represent the starting point in the learning process. It is expected that the performance of the learners will improve by acquiring the features of the target language. The constraints will be re-ranked to achieve the native=like forms, which would be the end point of the learning process, as in Tableaux 4.1 and 4.3.

4.6  Conclusion and pedagogical implications The current study was based on the production and perception of a set of Arabic consonants by learners from different linguistic backgrounds. The emphatics and pharyngeal were difficult to produce and perceive by L2 Arabic learners, and the learners performed better in the production and perception of uvular fricatives. Turkish learners were found to be the most successful group, while Greek learners, followed by Chinese learners, were the least successful group. The learners’ L1 affected their accuracy of perceiving and producing some of the target consonants. The findings of this study provided support to the feature analysis approach. The relevance of the feature model goes beyond perception to include production in the current study. Moreover, learners differed in the choice of the consonants that substitute the target ones in the case of the production errors, based on the phonological features available in their L1. This study has examined several consonants and several languages, and it would be useful to conduct further experiments, narrowing the focus to one or two pairs only, and compare two languages, to be able to examine the findings in more depth. Moreover, more studies are needed to investigate whether the position of the consonant as word initially, medially, or finally will affect production or perception. In terms of the pedagogical implications, the current study has pointed out that uvular fricatives are not problematic to the five groups of learners; however, the pharyngeals and emphatics are. The teacher needs to be aware of the fact that learners from different L1 backgrounds will have different issues to deal with and the teacher also needs to make the learners aware of the value of employing L1 experience. For example, the teacher can draw to the attention of German learners that they can pronounce the /x/ sound and of the Turkish learners that they can differentiate between plain and emphatic consonants. Moreover, the students’ performance can be also improved by making them aware of which organs of speech are involved in the articulation.

References Al Mahmoud, M. S., 2013. Discrimination of Arabic contrasts by American learners. Studies in Second Language, 3 (2), 261–292. Al Tubuly, S., 2016. The production and perception of Arabic stress patterns by English speaking learners: A comparison with native speakers. Thesis (PhD). University of Essex. Alanazi, S., in progress. Production and perception of English stops by Saudi L2 learners of English. Thesis (PhD). University of Essex. Al-Ani, S., 1970. Arabic phonology: An acoustical and physiological investigation. The Hague: Mouton. Alosh, M., 1987. The perception and acquisition of pharyngealized fricatives by American learners of Arabic and implications for teaching Arabic phonology. Thesis (PhD). The Ohio State University. Al-Solami, M., 2013. Arabic emphatics: Phonetic and phonological remarks. Open Journal of Modern Linguistics, 4 (3), 314–318. Alwabari, S., 2013. Non-native production of Arabic pharyngeal and pharyngealized consonants. Thesis (PhD). Carleton University. Archibald, J., 1998. Second language phonology. Amsterdam/Philadelphia: John Benjamins. Archibald, J., 2005. Second language phonology as redeployment of NL phonological knowledge. Canadian Journal of Linguistics, 4 (50), 285–314.

90

The perception and production of Arabic consonants Arvaniti, A., 2007. Greek phonetics: The state of the art. Journal of Greek Linguistics, 8 (1), 97–208. Best, C., 1995. A direct realist view of cross-language speech perception. In: W. Strange, ed. Speech perception and linguistic experience: Issues in cross-language research. Maryland: York Press, 171–204. Best, C. and Strange, W., 1992. Effects of phonological and phonetic factors on cross-language perception of approximants. Journal of Phonetics, 20 (3), 305–330. Boersma, P., and Weenink, D., 2014. Praat: Doing phonetics by computer. [Online]. Available from: www.praat.org/. [Accessed 5 May 2017]. Botinis, A., et al., 2000. Acoustic structure of the Greek stop consonants. Glossologia, 11 (12), 167–199. Brown, C., 2000. The interrelation between speech perception and phonological acquisition from infant to adult. In: J. Archibald, ed. Second language acquisition and linguistic theory. Oxford: Blackwell, 4–63. Duanmu, S., 2002. The phonology of standard Chinese. Oxford: Oxford University Press. Eckman, F., 1977. Markedness and the contrastive analysis hypothesis. Language Learning, 27 (2), 315–330. Eckman, F., 1991. The structural conformity hypothesis and the acquisition of consonant clusters in the interlanguage of ESL learners. Studies in Second Language Acquisition, 13 (1), 23–41. Eckman, F., 2008. Typological markedness and L2 phonology. In: J. Edwards and M. Zampini, eds. Phonology and second language acquisition. Philadelphia: John Benjamins, 95–115. Flege, J. E., 1995. Second language speech learning: theory, findings, and problems. In: W. Strange, ed. Speech perception and linguistic experience: Issues in cross-language research. Timonium, Maryland: York Press, 233–277. Ghazeli, S., 1977. Back consonants and backing coarticulation in Arabic. Thesis (PhD). University Texas at Austin. Goad, H., and White, L., 2006. Ultimate attainment in interlanguage grammars: A prosodic approach. Second Language Research, 22 (3), 243–68. Guion, S., et al., 2000. An investigation of current models of second language speech perception: the case of Japanese adults’ perception of English consonants. Journal of the Acoustical Society of America, 107 (5), 2711–2724. Hall, T., 1993. The phonology of German /R/. Phonology, 10 (1), 83–105. Hancin-Bhatt, B., 1994. Segmental transfer: A  consequence of a dynamic system. Second Language Research, 10 (1), 241–69. Hancin-Bhatt, B., 2008. Second language phonology in optimality theory. In: J. Edwards and M. Zampini, eds. Phonology and second language acquisition. Philadelphia: John Benjamins, 117–146. Iverson, P., et al. 2008. Category and perceptual interference in second-language phoneme learning: an examination of English /w/-/v/ learning by Sinhala, German and Dutch speakers. Journal of Experimental Psychology: Human Perception and Performance, 34 (5), 1305–1316. Iverson, P., et al., 2011. Cross-language specialization in phonetic processing: English and Hindi perception of /w/-/v/ speech and nonspeech. Journal of the Acoustical Society of America, 130(5), 297–303. Kara, R., 1976. The problems encountered by English speakers in learning Arabic. Thesis (PhD). University of California. Kehoe, M. and Lleó, C., 2002. Intervocalic consonants in the acquisition of German: onsets, codas or something else? Clinical Linguistics and Phonetics, 16 (3), 169–182. Ladefoged, P., 2001. Vowels and consonants: An introduction to the sounds of languages. Oxford: Blackwell Publishers. Lado, R., 1957. Linguistics across cultures: Applied linguistics for language teachers. Ann Arbor: University of Michigan Press. Lancaster, A. and Gor, K., 2016. Abstraction of phonological representations in adult non-native speakers. Proc Ling Soc Amer, 24 (1), 1–15. Larson-Hall, J., 2010. A guide to doing statistics in second language research using SPSS. New York: Routledge. Laufer, A. and Baer, T., 1988. The emphatic and pharyngeal sounds in Hebrew and in Arabic. Language and Speech, 31(2), 181–205. Lengeris, A. and Nicolaidis, K., 2016. The identification and production of English consonants by Greek speakers. In: selected papers of the 21st international symposium on theoretical and applied linguistics 5–7 April 2013 Greece, 21, 224–238. [Online]. Available from: https://kar.kent.ac.uk/62017/1/522714997-1-SM.pdf. [Accessed 29 April 2017].

91

Sara Al Tubuly Lombardi, L., 2001. Segmental phonology in optimality theory: Constraints and representations. Cambridge: Cambridge University Press. Major, M., 1987. The Phonological similarities, markedness, and rate of L2 acquisition. Studies in Second Language Acquisition, 9 (1), 63–82. Major, R., 2001. Foreign accent: The ontogeny and phylogeny of second language phonology. London: Lawrence Erlbaum. Major. R., 2008. Transfer in second language phonology: a review. In: E. Hansen and M. Zampini, eds. Phonology and second language acquisition. Amsterdam: John Benjamins 63–94. McCarthy, J., 1994. The phonetics and phonology of Semitic pharyngeals. In: P. Keating, ed. Phonological structure and phonetic form: Papers in laboratory phonology III. Cambridge: Cambridge University Press, 191–234. McCarthy, J., 2008. Doing optimality theory: Applying theory to data. Oxford: Blackwell. Mennen, I., and Okalidou, A., 2006. Acquisition of Greek phonology: an overview. QMUC Speech Science Research Centre Working Papers, WP-11. Prince, A. and Smolensky, P., 2004. Optimality theory: constraint interaction in generative grammar report. [Online]. no. RuCCS-TR-2. ROA-537. New Brunswick, NJ. Rutgers University Centre for Cognitive Science. Available from: http://roa.rutgers.edu/files/537-0802/537-0802-PRINCE-1-0.PDF [Accessed 01 May 2017]. Roach, P., 2009. English phonetics and phonology: A practical course. Cambridge: Cambridge University Press. Selinker, L., 1972. Interlanguage. International Review of Applied Linguistics, 10 (3), 209–241. Shehata, A., 2015. Problematic Arabic consonants for native English speakers: Learners perspectives. International Journal of Educational Investigations, 2 (9), 24–47. Topbaş, S., 2007. Turkish speech acquisition. In: S. McLeod, ed. The international guide to speech acquisition. Clifton Park, NY: Thomson Delmar Learning, 566–579. Watson, J., 2002. The phonology and morphology of Arabic. Oxford: Oxford University Press. Wester, F., et al., 2007. Substitution of dental fricatives in English by Dutch L2 speakers. Language Sciences, 29 (3), 477–491. Wiese, R., 1996. The phonology of German. Oxford: Clarendon Press. Yamada, R. A., 1995. Age and acquisition of second language speech sounds: perception of American English/r/ and /l/ by native speakers of Japanese. In: W. Strange, ed. Speech perception and language experience: Issues in cross-language research. Maryland: York Press, 305–320. Zahid, M., 1996. Examination of a perceptual non-native speech contrast: pharyngealized/ non-pharyngealized discrimination by French-speaking adults. In: Proceedings, fourth international conference on spoken language, 4, 2466–2469. [Online] Available from: www.asel.udel.edu/icslp/cdrom/ vol4/977/a977.pdf. [Accessed 29 April 2017]. Zawaydeh, B., 1998. Gradient uvularisation spread in Ammani-Jordanian Arabic. In: A. Benmamoun, et al., eds. Perspectives on Arabic linguistics 11. Amsterdam: John Benjamins, 117–141. Zawaydeh, B., 1999. The phonetics and phonology of gutturals in Arabic. Thesis (PhD). Indiana University.

92

5 ARABIC L2 PHONOLOGICAL ACQUISITION An ultrasound study of emphatics and gutturals1 Amanda Eads, Jodi Khater, and Jeff Mielke This chapter investigates the articulation of Arabic emphatic and guttural consonants of nine L1 English learners of Arabic as an L2 enrolled in various levels of Arabic instruction at a university in the United States. The motivation for this study stems from two key issues: the small amount of literature discussing phonological acquisition of Arabic among adult second language learners and the need for more technical articulatory feedback to address students’ pronunciation errors (Ryding 2013). We found evidence that many of our participants struggle with the Arabic emphatic and guttural phones. For several participants these sounds were extremely challenging, and participants were not able to differentiate between their articulations of these phones. This coincides with Odisho (2005)’s claim that Arabic emphatics and gutturals are most difficult for beginning English L1 students because these are unfamiliar sounds. Arabic has four emphatics that add a secondary articulation to /s d t ð/. However, there is debate concerning whether the secondary articulation involves the retraction of the tongue body or the tongue root (Ali and Daniloff 1972; Ghazeli 1977; Al-Solami 2017). Previous studies generally agree that the secondary articulation is an effect of tongue backing. AlSolami (2013) discusses the three categories of Arabic guttural consonants – glottals, pharyngeals, and uvulars – as well as the debate concerning their articulation. His analysis coincides with Ghazeli’s (1977) evidence of independent retraction of the tongue root in the lower pharynx during pharyngeal articulation. For uvular articulation, Al Solami supports Catford’s (1977) finding that uvulars retract the tongue body but have more retracted tongue dorsum than emphatics. Therefore, this ultrasound study provides a detailed analysis concerning how L2 Arabic learners are articulating these phones, at what level learners are typically able to differentiate their articulation of the target phones, and in what order this occurs.

5.1 Introduction The motivation for this study stems from two key issues: the relatively small amount of literature discussing phonological acquisition of Arabic among adult second language learners and the need for more technical articulatory feedback to address students’ pronunciation errors. The emphatic and guttural phones stand out as particularly problematic for English-speaking learners of Arabic. The emphatics include [ṣ ḍ ṭ ð ̣], and the gutturals include the glottals [ʔ h], 93

Amanda Eads, Jodi Khater, and Jeff Mielke

pharyngeals [ʕ ħ], and uvulars [γ χ q].2 We expect the emphatics, pharyngeals, and uvulars to be particularly challenging for learners whose native languages lack these phones. Although learners may use some of the articulatory movements in their native phone production, such as backing of the tongue body and/or constriction of the pharynx, these movements are relatively inaccessible to the ears and eyes. The four Arabic emphatic phones [ṣ ḍ ṭ ð ̣] are not native to American English varieties. These sounds are new to first language English students learning Arabic as a second language. According to Odisho (2005), a text on teaching comparative pronunciation of English and Arabic, emphatic Arabic consonants add a secondary articulation to [s d t ð], which are phones found in American English varieties. This secondary tongue-backing feature is not described in detail in the teaching text, therefore leaving its readers to wonder how the secondary articulation is actually produced. While other teaching texts, such as Al-Kitaab (Brustad et al 2004), provide varying degrees of instruction for pronunciation, many learners continue to struggle with production of the Arabic emphatics. However, past linguistic studies have endeavored to provide more description on this secondary articulation. In a cinefluorographic study, Ali and Daniloff (1972) found that emphatics were articulated with simultaneous tongue depression and rearward movement of the tongue dorsum toward the posterior wall of the pharynx. A few years later in an X-ray study, Ghazeli (1977) found that the tongue body is pulled backwards into the upper oropharynx during the articulation of the emphatic [ṭ]. Additionally, Dolgopolsky (1977) claimed Arabic emphatics are pronounced as uvularized consonants, meaning the consonants are modified “by moving back the rear part of the tongue toward the uvula and the back wall of the pharynx” (Dolgopolsky 1977, p. 1). With somewhat less detail, Younes (1993, p. 119) claims the uvularized emphatics have a secondary articulation that “accompanies a primary articulation at another point in the vocal tract.” These descriptions of emphatic articulation range from vague to exact, but there remains confusion concerning which parts of the tongue are backed during articulation. Al-Solami (2013) discusses the debate concerning the backing of the tongue root and/or backing of the tongue body. He concludes that while specifics are not agreed upon, most scholars do agree that emphatic articulation involves a coronal articulation as well as a secondary articulation as a result of the retraction of the tongue body. In Al-Solami’s (2017) study of three dialects of Arabic, he claims the secondary articulation is “difficult to pinpoint” and further acknowledges Khattab et al. (2006) when he suggests that there is no single articulation strategy, but instead articulation may vary depending on dialect, phonological context, and social variables (Al Solami 2017, p. 25). However, he further elaborates that while the tongue dorsum retraction is to the upper oropharynx for emphatics, as Ghazeli (1977) found, the tongue dorsum retraction differs between the emphatic and uvular consonants in that the tongue body is depressed, and therefore he claims that Arabic emphatics should be considered velarized rather than uvularized or pharyngealized (Al Solami 2017, p. 34). Arabic guttural phones (the glottals [ʔ h], pharyngeals [ʕ ħ], and uvulars [γ χ q]) also present challenges to L1 American English learners of Arabic. This group of consonants is so named due to the place of articulation in the back of the throat. Previous studies have discussed the guttural consonants without the inclusion of the glottal consonants or, if included, the glottals have been termed laryngeals. While laryngeal is an appropriate term, this study will refer to them as glottals since they are “articulated at the larynx with a fully open glottis in [h] or fully constricted glottis in [ʔ]” (Al Solami 2017, p. 27). Several studies have reported no constriction in the pharynx during the articulation of the glottal consonants (Laufer and Condax 1979; Zawaydeh 1999; Al-Solami 2013). Additionally, Shahin (2011) found no retraction of the tongue root or epiglottis, and Al-Ani (1970) found no co-articulatory effect on vowels 94

Arabic L2 phonological acquisition

during articulation. While other studies have excluded the glottals due to lack of constriction in the pharynx, several students in our program report having issue with the glottal stop [ʔ], and therefore we felt it necessary to include the glottals in our study. As alluded to earlier, emphatic and uvular consonants are very similar in articulation. Both Ghazeli (1977) and Catford (1977) claim uvulars and emphatics involve retraction of the tongue body during articulation, with uvulars having a more retracted tongue dorsum. AlSolami (2017) also concludes that for uvulars the tongue dorsum retracts vertically rather than horizontally, as in emphatic articulation. Furthermore, Ghazeli (1977) reports that the articulation of voiceless uvular [χ] and [q] occurs when the tongue root and epiglottis are pulled backwards toward the posterior wall of the pharynx, but this does not occur for voiced uvular [γ]. There is some debate in the literature concerning which uvular consonant is more posterior (Delattre 1971; Ghazeli 1977; Al Solami 2017), though during all uvular articulation the back of the tongue is raised toward the uvula and the tongue body is not depressed. Catford (1977) describes this articulation as dorso-uvular because the back of the tongue moves toward the uvula while there is also constriction between the tongue dorsum and soft palate. The third class of guttural consonants, pharyngeal fricatives [ʕ ħ], are also articulated by retracting the tongue root toward the posterior wall of the pharynx, as in the case of the uvulars; however, with pharyngeals the posterior wall slightly moves forward (Delattre 1971; Ghazeli 1977). As with the other classes of consonants, there is debate concerning the exact place of articulation. It is generally agreed that “during the articulation of voiced pharyngeal fricative [ʕ] the tongue blade assumes a curved pyramidal shape almost like an inverted ‘V’ ” (Delattre 1971; Ghazeli 1977; Elgendi 2001; Al Solami 2017, p. 30). Additionally, Ghazeli (1977) describes a narrower constriction in the epiglottis for the voiceless pharyngeal fricative [ħ]. The pharyngeal articulation debate arises concerning retracted parts of the tongue and the epiglottis. Al-Solami (2013) describes an independent retraction of the tongue root in the lower pharynx but does not comment on the epiglottis. Laufer and Condax (1979) claim the epiglottis retracts independently from the entire tongue, while others claim the epiglottis and tongue root retraction varies (Boff-Dkhissi 1983; Laufer and Baer 1988). Given the scholarly debate concerning articulation of these emphatic and guttural phones, it stands to reason that some Arabic teaching texts would have a difficult time explaining to students how to “correctly” pronounce these difficult phones. Alif Baa: Introduction to Arabic Letters and Sounds, produced by Brustad et al. (2004), gives clear instructions concerning how to produce the emphatic and guttural Arabic phones. For example, to pronounce the uvular [χ], they write, “say k and pay attention to where the back of your tongue hits the roof of your mouth and blocks your windpipe. Instead of closing off the windpipe with the back of your tongue completely, block it part of the way and you will produce this sound” (Brustad et al. 2004, p. 33). While the instructions are explicit, the textbook is unable to provide feedback students need in order to know whether they are producing these sounds correctly. Unfortunately, Arabic language instructors often have large classes and little time to spend one on one with students. This is where visualization of tongue movement could benefit students’ further understanding of differentiation between difficult phones. Ultrasound technology provides an opportunity for students to visualize their tongue during sound production and allows for self-monitoring as well as speaker modification if used within the classroom setting. This visual feedback for students allows for immediate in-class corrections and use by the language instructor and/or use of native Arabic speaker recordings demonstrates expected articulatory movements. While we recognize ultrasound machines are expensive, and therefore not all instructors have access, we have found that it is a valuable pedagogical tool. 95

Amanda Eads, Jodi Khater, and Jeff Mielke

5.1.1  Consonant acquisition Amayreh and Dyson (1998) explored the acquisition of Arabic consonants as compared to the acquisition of English consonants. Their study focused on normally developing children in order to track the similarities and differences in the stages of acquisition for English and Arabic. Because of the dialect diversity in Jordan, where they gathered data, Amayreh and Dyson (1998) analyzed both the standard dialect and what they considered the acceptable non-standard variants. They examined all phones within Arabic, but for the purposes of this study it is noteworthy to mention that they found standard [t d k ħ] were acquired in the early period, standard [s χ γ h] were acquired during the intermediate period, and standard [ð ṣ ḍ ṭ ð ̣ q ʔ ʕ] were acquired in the later period. The acceptable non-standard sounds produced during those periods ranged, especially given children’s inexperience and lack of exposure to the standard language variety or what they call educated spoken Arabic (Dyson and Amayreh 2000). A later study (Amayreh 2003) explored the acquisition of the later period consonants compared with the children’s elementary textbooks. He suggests frequency and articulation difficulty may play a role in the acquisition of later period consonants. While several studies have examined phonological acquisition of Arabic in children (Amayreh and Dyson 1998; Dyson and Amayreh 2000; Amayreh 2003, Khattab 2011), far fewer have examined phonological acquisition among adult second language learners of Arabic (Zaba 2007; Durham et al. 2014; Showalter and Hayes-Harb 2015; Hayes-Harb and Durham 2016). Adult Arabic second language learning studies are critical for two key reasons. First, most obviously, Arabic is learned as an additional language; therefore, some phones will have been acquired prior to encountering the new language. However, Arabic has several consonants that are not native to English speakers. Secondly, as Ryding (2006) discusses, most adult second language learners of Arabic first encounter Modern Standard Arabic. Therefore the standard, more privileged, variety of Arabic is acquired prior to the more commonly spoken dialects. Students typically acquire language skills within a formal classroom setting before engaging with native speakers using various other dialects of Arabic. Due to this reversal of experience as compared to child Arabic learners, where adult students’ first experience with the language is through formal instruction, it is essential to examine which sounds prove most difficult for adult second language learners and whether the stages of acquisition for the non-native phones compares to those of the first language learners. While many of the previous studies have focused on acoustic properties and perception, this study seeks to use ultrasound technology to examine students’ articulation of the difficult phones.

5.2  Research questions In order to obtain a clear picture of the stages of acquisition and the articulation of Arabic phones that are most difficult for L1 English learners of Arabic, we developed a study that included students from all levels of the Arabic program at our university as well as a wide variety of articulatory contexts. Our research questions focus on how the emphatic and guttural Arabic phones are being articulated by our L2 Arabic students, at what levels the students typically acquire the target phones, and the order in which the target phones are acquired.

5.3 Methods 5.3.1  Participant questionnaire Before participating in the experiment, each participant completed a questionnaire detailing his or her previous experience with Arabic. Questions considered age of first encounter, length 96

Arabic L2 phonological acquisition

of time Arabic has been spoken, which dialects are spoken and with whom, time spent in class, time spent practicing each of the four skills, most difficult sounds to produce, suggested best way to learn Arabic, and if the participant had studied abroad. There were 23 questions in total, and all responses were self-reported.

5.3.2 Stimuli The stimuli for this experiment included 225 real Arabic words, including a minimum of 15 instances of each of the target phones: [s d t ð ṣ ḍ ṭ ð ̣ q χ γ ħ ʕ h ʔ] (see Appendix). Initial, medial, and final word positions were accounted for during the creation of the wordlist, with a minimum of five instances of the target phones in each position. Several words had multiple target phones, therefore providing more than the minimum 15 instances of some target phones. For example, ṣabaaħ “morning” fulfills the requirement for an initial [ṣ] and a final [ħ], although for the purposes of stimuli creation it was counted only for its initial [ṣ]. When selecting stimuli, words were chosen primarily for target phone placement and vowel pattern. An effort was made to include some words that may be familiar to students at various stages within our program; however, this was not the primary objective when compiling the word list. The wordlist and its English translation were distributed to participants before participating in the experiment so that they could familiarize themselves with the stimuli. The participants read randomized versions of the Arabic wordlist from a remote controlled screen that is part of an Arduino-based head movement tracking system (discussed next). The participants produced each word only once unless their head was misaligned during production, thus requiring them to produce the word again before proceeding through the list.

5.3.3  Ultrasound image acquisition Recordings took place in a phonology lab at the authors’ university in a sound-attenuated booth. The participants were asked to sit in a desk in the booth. For head stabilization, participants rested their foreheads on a headrest attached to the desk where they sat. Accelerometers attached to a headband gave feedback on alignment and recorded when the participants’ heads were out of alignment. A single degree of movement caused the accelerometer to signal misalignment. An Arduino-based screen attached to the headrest gave visual feedback that allowed participants to see if they were aligned. We allowed the participants to become comfortable with the head tracking system before proceeding with the experiment. An arm, also attached to the desk, held the ultrasound probe in place during the recording. It is important to note that this head stabilization and movement tracking system is only necessary for quantitative comparisons of tongue postures (as in this study), and it is not necessary for pedagogical use of ultrasound imaging. The ultrasound images were captured at 60 frames per second using a Terason t3000 with an 8MC3 microconvex array transducer and Ultraspeech software (Hueber et al. 2008). Audio was recorded using an Audio-Technica AT803 lavalier omnidirectional microphone attached to the arm holding the ultrasound transducer, recorded through a Sound Devices USBPre2 preamplifier and digitized at a sampling rate of 44.1 kHz with 16 bits per sample. Water based ultrasound gel was applied to the ultrasound transducer before being placed under the participant’s chin. Several test words were elicited to check for proper placement of the ultrasound probe and a clear tongue image. Once a clear tongue image was obtained, we synchronized the accelerometer to the participant’s current position in the headrest. This allowed for very little head movement during the recording, therefore giving us a relatively high degree of accurate and clear tongue images. After the participant was synced, the stimuli 97

Amanda Eads, Jodi Khater, and Jeff Mielke

became visible on the screen, and participants progressed through the wordlist by remote control. However, participants were unable to progress through the wordlist if their head was misaligned. They would use visual feedback to re-align their head. On average, the questionnaire and recording took approximately one hour.

5.3.4  Data preparation Once the ultrasound recording was complete, the audio, ultrasound, and stimuli files were uploaded to our server, where we segmented the recordings using the Penn Phonetics Lab Forced Aligner (Yuan and Liberman 2008), adapted for Arabic by us. We hand corrected tokens in PRAAT before running a script to obtain the medial ultrasound image of each target phone. Ultrasound images were selected from the midpoint of each target phone interval, and the tongue surfaces were traced using the Palatoglossatron program (Baker 2005). Between 400–500 tongue contours were traced per individual. These contours were compared using the R statistics software (R Core Team 2016). Smoothing Spline ANOVA comparisons (Gu 2002) were used to test for articulatory differences between different phones. This technique was first applied to tongue contours by Davidson (2006). In order to make comparisons all along the tongue’s curved surface, SS-ANOVAs were calculated using polar coordinates, following Mielke (2015).

5.3.5 Participants The total number of participants recorded for this project is 33, including heritage and nonheritage Arabic language learners as well as native Lebanese and native Egyptian Arabic speakers. However, the data and results presented in this article concentrate on a subset of nine students: four non-heritage participants (two males and two females) having completed the first year of Arabic language study at our university (of four credit hours per semester) and five non-heritage post-study abroad participants (three females and two males) who have completed six or more Arabic language courses. Tables 5.1 and 5.2 include the participants’ selfreported information concerning Arabic language use. The four first-year participants are labeled FY, and the five post-study abroad participants are labeled SA. All four first-year students had completed the Arabic 101 and 102 courses at university and reported speaking Modern Standard Arabic (MSA) and Egyptian Arabic. First-year participant FY02 reported learning Arabic in high school but did not report additional classes. The participants reported only speaking Arabic in class with the professor and classmates, and they did not report time using Arabic outside of class or homework assignments. The hours per week speaking reported in Table 5.1 includes class time, although it should be noted that recordings were done during summer break when classes were not in session. All participants reported that conversation and interaction with other speakers is the best way to learn language. Four of the study abroad participants participated in a study abroad program in Jordan. Post-study abroad participants SA01, 02, and 03 completed a summer semester, while SA05 completed a fall semester. Participant SA04 completed a summer study abroad semester in Table 5.1 Pertinent information regarding the first-year Arabic participants Speaker

FY01

FY02

FY03

FY04

Age at First Encounter Hours per Week Speaking Reported Difficult Phones

18 4 None Reported

16 3 [γ]

18 Minimal [ṣ]

19 0 [γ ʕ ð ̣]

98

Arabic L2 phonological acquisition Table 5.2 Pertinent information regarding the post-study abroad participants Speaker

SA01

SA02

SA03

SA04

SA05

Age at First Encounter

15

17

20

16

21

Speaks With Whom

In class, native speakers

In class, tutoring

None reported

Professors, Egyptian friend

Jordanian/ Lebanese friends, second language speakers

Dialects Spoken

MSA, Jordanian, Egyptian

MSA, Jordanian, Levantine

MSA

MSA, Egyptian

MSA, Levantine

Number of Classes

6+

11

Hours per Week Speaking

3

Reported Difficult Phones

[γ ħ]

1.5 [γ ħ χ q]

6

6

12

0

1

3

[γ ħ ʕ]

[ħ ʔ q]

[ṣ ḍ]

Oman. All speakers except SA03 reported engaging in the language outside of the classroom. SA03 did not report speaking partners, hours per week speaking, or a dialect other than MSA; however, she did report high amounts of reading and listening to the Qur’an. The participants’ number of completed classes ranges from 6–12, and the class totals include those completed abroad. The number of classes recorded by the participants is not an indication of a specific number of contact hours. For example, a conversation lab which meets one hour per week may be counted by a participant as a “class,” just as a three-hour-per-week course was counted as a “class.” It should be noted that the post-study abroad participants reported difficulty with more phones than the first-year participants.

5.4  Data analysis and results As noted in the participant questionnaire data, the first-year participants reported far fewer difficulties in pronunciation than their post-study abroad counterparts. This is likely due to an inability to perceive differences. Figure 5.1 shows first-year participant FY01 and Figure 5.2 shows post-study abroad participant SA03. Each curve represents one set of mid-sagittal tongue contours. The tongue tip is on the right and the tongue root is on the left. The thick lines are the mean curves for each category, and the thin dotted lines represent 95% confidence intervals around the mean. We interpret two categories as different for the parts of the curves where the confidence intervals do not overlap. For example, Figure 5.1 shows only a small significant difference between FY01’s articulation of [d] and the other coronal stops (which is probably not related to learning Arabic categories), but Figure 5.2 shows SA03 has significant tongue root retraction and lowered tongue body for the two emphatic consonants. FY01’s articulation of plain and emphatic coronal stops in Figure 5.1 shows no difference, i.e., there is no secondary articulation for emphatic [ṭ ḍ]. However, the SS-ANOVA of participant SA03 in Figure 5.2 shows the distinction between the plain and emphatic coronal stops. The tongue 99

-150 -200 -250

t d ṭ ḍ

-300

← down toward larynx up toward palate →

Amanda Eads, Jodi Khater, and Jeff Mielke

300

350

400

← back toward pharynx

450

500

front toward lips →

-150 -200 -250

t d ṭ ḍ

-300

← down toward larynx up toward palate →

Figure 5.1 SS-ANOVA of the participant FY01’s tongue during articulation of coronal stops. FY01 shows no difference in articulation between non-emphatic and emphatic coronal stops

300

350

400

← back toward pharynx

450

500

550

front toward lips →

Figure 5.2 SS-ANOVA of the participant SA03’s tongue during articulation of coronal stops. SA03 shows a depressed tongue body and retracted tongue root, as expected for emphatic articulation

body is depressed and the tongue root is retracted (to the bottom left corner of the graph), illustrating the secondary articulation for emphatic [ṭ ḍ]. None of the four first-year participants showed a difference in articulation between plain and emphatic coronal stops. Participants SA01 and SA04 showed slight tongue root retraction for emphatic [ṭ] but not for [ḍ]. SA01 also showed slight tongue depression for [ṭ]. SA02, 03, and 05 demonstrate the depressed tongue body and retracted tongue root expected in the articulation of Arabic emphatic consonants. Similarly, none of the four first-year participants demonstrated a depressed tongue body or retracted tongue root for the emphatic coronal fricatives [ṣ ð ̣]. There was no evidence of movement 100

Arabic L2 phonological acquisition

-200 -250 -300

← down toward larynx up toward palate →

toward a secondary articulation. In fact, we discovered participant FY01 produces the coronal stop [t] in place of the emphatic coronal fricative [ð ̣]. Three of the post-study abroad speakers also struggled with the articulation of the emphatic coronal fricatives. Figure 5.3 shows that participant SA01 slightly retracts the tongue root for [ṣ ð ̣] and also slightly depresses the tongue body for [ṣ]. The SS-ANOVA for SA04 is similar to that of SA01 in Figure 5.3, with movement toward secondary articulation for [ṣ] but not for [ð ̣]. Figure 5.4 shows that participant SA05 produces an emphatic [ṣ] with a retracted tongue root and depressed tongue body. Like participants SA01 and SA04, SA05 also struggles with the articulation of [ð ̣], but the secondary articulation is more pronounced than the others, as seen when comparing Figure 5.3 and Figure 5.4. The SS-ANOVA of participant SA02 in

s ð ṣ ð̣ 300

350

400

← back toward pharynx

450

500

front toward lips →

-150 -200 -250

s ð ṣ ð̣

-300

← down toward larynx up toward palate →

Figure 5.3 SS-ANOVA of participant SA01’s tongue during articulation of coronal fricatives. SA01shows slight tongue root retraction and slight tongue body depression for emphatic [ṣ] but not for [ð ̣]

250

300

350

← back toward pharynx

400

450

500

front toward lips →

Figure 5.4 SS-ANOVA of participant SA05’s tongue during articulation of coronal fricatives. SA05 shows expected articulation for emphatic [ṣ] but only slight tongue body depression for [ð ̣]

101

-150 -200 -250

s ð ṣ ð̣

-300

← down toward larynx up toward palate →

Amanda Eads, Jodi Khater, and Jeff Mielke

250

300

350

← back toward pharynx

400

450

front toward lips →

Figure 5.5 SS-ANOVA of participant SA02’s tongue during articulation of coronal fricatives. SA02 demonstrates the expected tongue body depression and the retracted tongue root for the emphatic coronal fricatives

Figure 5.5 shows the expected articulation of emphatic coronal fricatives with clearly defined tongue root retraction and tongue body depression. The SS-ANOVA for participant SA03 was similar to SA02 (Figure 5.5) revealing the expected primary and secondary articulation. The Arabic emphatic consonants proved challenging for first-year as well as some poststudy abroad participants in their third and fourth year of classes. The Arabic guttural consonants were equally challenging. As discussed in the introduction, the glottal consonants were included in this study due to students’ reports of difficulty, though it should be noted that tongue shape alone without acoustic study and consideration of vowel contexts will not fully address whether these phones are articulated as expected. We have designated these phones as distinguished in the discussion section (Section 5.5), with this caveat, but they require further study. The dorsal consonants examined in this study include the uvular consonants of [q χ γ] and the velar [k], which provides an easy reference as to whether the participants are retracting the tongue dorsum for the uvular consonants. As evidenced in Figure 5.6, participant FY03 does not retract the tongue body or dorsum. There is significant overlap in the articulation of [k] and [q]; in some cases the tongue dorsum is more retracted for non-uvular [k]. The two uvular fricatives are also not retracted. Participant FY04 has the same result. Participants SA01 and SA02 articulate the uvular consonants similar to Figure 5.7 for participant SA05, where the uvulars are articulated with the tongue rising and backing vertically toward the top left corner of the SS-ANOVA graph and the tongue dorsum is retracted well behind the dorsal [k]. Participants FY01 and FY02 both show movement toward expected articulation of the uvular consonants, although at varying degrees. In both cases, the participants raise or retract the tongue during articulation of [q χ γ], though the tongue dorsum is not vertically retracted as much as expected. For participant SA03 the uvular stop [q] is articulated as expected, and while the uvular fricatives [χ γ] have the tongue dorsum somewhat retracted, the vertical rearward movement is not as pronounced as expected. Similarly, SA04 produces an expected 102

-150 -200 -250

k q χ γ

-300

← down toward larynx up toward palate →

Arabic L2 phonological acquisition

250

300

350

← back toward pharynx

400

450

500

front toward lips →

-150 -200 -250 -300

k q χ γ

-350

← down toward larynx up toward palate →

Figure 5.6 SS-ANOVA of participant FY03’s tongue during articulation of dorsal consonants. FY03 does not demonstrate the retraction of the tongue root for the Arabic uvular phones

250

300

350

← back toward pharynx

400

450

500

front toward lips →

Figure 5.7 SS-ANOVA of participant SA05’s tongue during articulation of dorsal consonants. SA05 shows the rear movement of the tongue body toward the uvula and the retracted tongue root as expected in the articulation of Arabic uvular phones

[q] but not fully retracted [χ γ]. In the case of SA04, the voiced uvular fricative [γ] is raised but not retracted, and the voiceless uvular fricative [χ] has a retracted tongue body but is not raised as expected. The pharyngeal consonants of Arabic include a voiced fricative [ʕ] and a voiceless fricative [ħ]. Overall, more participants are moving toward the expected articulation of the pharyngeals as compared to the uvulars. Only two produce the expected articulation of the “inverted V” as 103

Amanda Eads, Jodi Khater, and Jeff Mielke

-150 -200 -250

ħ ʕ h ʔ

-300

← down toward larynx up toward palate →

described by Al-Solami (2017, p. 30) and others. However, six of the nine participants show movement toward expected pharyngeal articulation. Participant FY03 demonstrates no movement. SA05 in Figure 5.8 shows movement toward the inverted V-like shape. The tongue root is pushed forward with the voiceless fricative [ħ] in front of [ʕ], revealing a narrower constriction in the epiglottis. A variation of this occurs for participants FY01, FY02, and FY04 as well as SA01, SA04, and SA05. Participants SA02 and SA03 demonstrate the expected inverted V-shaped articulation, as seen in the SS-ANOVA for SA02 in Figure 5.9.

250

300

350

← back toward pharynx

400

450

500

front toward lips →

-150 -200 -250

ħ ʕ h ʔ

-300

← down toward larynx up toward palate →

Figure 5.8 SS-ANOVA of participant SA05’s tongue during articulation of the pharyngeal and glottal Arabic phones. SA05 shows movement toward an “inverted V” in pharyngeal articulation. (Al-Solami 2017, pg. 30)

250

300

350

← back toward pharynx

400

450

front toward lips →

Figure 5.9 SS-ANOVA of participant SA02’s tongue during articulation of the pharyngeal and glottal Arabic phones. SA02 demonstrates the “inverted V” (Al-Solami 2017, pg. 30)

104

Arabic L2 phonological acquisition

After reviewing each of the Arabic emphatic and guttural phones and considering whether the participants were able to produce the expected articulations according to the literature, it is necessary to clarify which phones participants appear to distinguish through articulation, which are clearly not distinguished, and which show some movement toward expected articulation but are not yet fully reflective of the literature. According to Amayreh (2003), it is important to note that spoken Arabic forms “allow for a great deal of variety in the production of some consonants.” Because this is the first study of its kind and Arabic ultrasound studies are quite rare, we have not compared the second language participants’ tongue contours to any particular dialect of Arabic. For interested readers, the students’ self-reported spoken dialects were mentioned in the participant data earlier. However, Dyson and Amayreh (2000) found little effect in articulation between local Jordanian varieties and the educated spoken Arabic in Amman during their study of Arabic L1 children. This remains to be examined in Arabic ultrasound literature.

5.5 Discussion According to Amayreh and Dyson (1998), the Arabic emphatic phones appear in the late stages of acquisition for L1 Arabic learners. Dyson and Amayreh (2000) further describe the emphatic consonants as the most difficult due to their complex articulation pattern. L1 Arabic children mispronounced emphatic consonants more often than any other phone(s) except for the uvular fricatives [χ γ]. This appears to be true for these Arabic second language learners as well. After two semesters involving formal instruction three days a week and an additional day of informal conversation class, the first-year students have not been able to distinguish between the emphatic and non-emphatic phones. In fact, there is no sign of articulation difference between the plain and emphatic consonants. The post-study abroad participants vary in their production. SA01 and SA04 demonstrate tongue root retraction for the voiceless emphatic consonants, although tongue body depression is not apparent. They show no apparent articulation difference between plain and emphatic voiced coronal consonants. Additionally, SA05 is able to distinguish between the voiceless emphatic and non-emphatic consonants as well as the voiced emphatic [ḍ] but struggles with distinguished articulation of voiced emphatic [ð ̣]. This leads us to hypothesize that voiceless emphatic consonants are distinguishable prior to their voiced emphatic counterparts. This would need much more attention before making any claims, but it is noteworthy for Arabic language instructors. Furthermore, participant SA05, who reported taking the highest number of classes, self-reported difficulty with the emphatic [ṣ] and [ḍ]; therefore, perception of emphatic phones is certainly higher for this participant compared to others who reported no issue but have not been able to fully distinguish between the plain and emphatic phones. Table 5.3 summarizes the speakers’ ability to distinguish plain and emphatic phones. The acquisition stages of Amayreh and Dyson (1998) dispersed the Arabic guttural consonants among all three periods. The pharyngeal voiceless fricative [ħ] is acquired in the early stage by native Arabic speakers. The uvular fricatives [χ γ] and glottal [h] are acquired in the intermediate period, while the uvular and glottal stops [q ʔ] as well as the pharyngeal voiced fricative [ʕ] are usually acquired in the later stage. This does not appear to be true for English L1 Arabic second language learners. For starters, glottal [h] and the glottal stop are native English sounds. For the purpose of this study, the glottals are among the first phones marked as distinguished, although we discussed certain limitations in the previous section. Future research will explore acoustic properties and vowel contexts. Given the data in Table 5.4 it is difficult to determine whether there is a sequence in which participants are 105

Amanda Eads, Jodi Khater, and Jeff Mielke Table 5.3 Ability to distinguish between plain and emphatic Arabic phones by individual speakers Speaker

[ṭ]

[ḍ]

[ṣ]

[ð̣]

FY01 FY02 FY03 FY04 SA01

~

SA02 SA03 SA04

   

SA05

~  

  ~

 





~

Note: A check mark indicates that the participant has articulated the phone as expected according to the literature and distinguishes between plain and emphatic consonants. A tilde indicates that the participant’s articulation of the emphatic phone differs from the non-emphatic counterpart in a way that we would expect but is not yet fully reflective of the literature

Table 5.4 Ability to distinguish between Arabic guttural phones by individual speakers Speaker

[q]

[χ]

[γ]

[ħ]

[ʕ]

[h]

[ʔ]

FY01 FY02 FY03

~ ~

~ ~

~ ~

~ ~

~ ~

  

  

~

~





    

  ~ ~ 

  ~ ~ 

~   ~ ~

~   ~ ~

    

    

FY04 SA01 SA02 SA03 SA04 SA05

Note: A check mark indicates that the participant has articulated the phone as expected according to the literature and distinguishes between the guttural phones. A line indicates that the participant’s articulation of the guttural phone has shifted in a way that we would expect but is not yet fully differentiated and reflective of the literature

able to differentiate between the articulation of uvulars and pharyngeals. The uvular stop [q], acquired by native speakers in a late stage of acquisition, is the only guttural consonant other than the glottals that is articulated according to the literature by all five post-study abroad speakers. In addition to the uvular stop, SA01, SA02, and SA05 also show distinct articulation of the other two uvular consonants [χ γ]. Dyson and Amayreh (2000, p. 94) claim that children produced as many errors on the uvular fricatives [χ γ] as they did on emphatics; however, we hypothesize that the ability to distinguish between Arabic uvular consonants occurs prior to the pharyngealized consonants for adult L1 English–L2 Arabic learners. While first-year participants FY03 and FY04 do not demonstrate a shift toward expected articulation for the uvular or pharyngealized consonants, the first-year Arabic participants 106

Arabic L2 phonological acquisition

FY01 and FY02 do demonstrate a shift toward the expected articulation of the uvular consonants. However, none of the first-year participants demonstrate a shift toward the expected emphatic consonant articulation. In contrast, all five of the post-study abroad participants demonstrate distinct articulation or a shift toward expected articulation of the uvular consonants. Meanwhile, only post-study abroad participants SA02 and SA03 are able to distinguish between all four of the emphatic and non-emphatic phones. Post-study abroad participant SA05 demonstrates distinct articulation for three out of the four empathic consonants and shows a shift toward expected articulation for the fourth. However, post-study abroad participants SA01 and SA04 only show a shift toward expected articulation for the voiceless emphatic consonants and no differentiation for the voiced emphatic consonants. Further study including more participants will be required in order to test our hypothesis concerning order of acquisition.

5.6  Conclusion and pedagogical implications These preliminary conclusions allow us to discuss articulation of Arabic emphatic and guttural phones. While debate remains in the literature concerning specifics of articulation with these two groups of sounds, there is ample discussion of general expected articulation. Our data demonstrates that the first-year students of Arabic included in this study are not yet able to fully distinguish between the expected articulations of emphatic/non-emphatic or guttural phones, with the exception of the glottals. The data also shows that although some of the poststudy abroad participants demonstrate the expected articulation of emphatic and guttural consonants, not all are able to do so, especially with regard to voiced emphatic consonants and guttural fricatives. This study suggests that the early L1 English–L2 Arabic learner is unable to produce the expected articulation of these difficult phones. This study focused primarily on participants’ ability to articulate Arabic emphatic and guttural phones as described in the literature. Therefore, we can comment on students’ individual abilities, but without adding an additional layer, including an acoustic study, we can only hypothesize regarding stages of acquisition. We hypothesize that L1 English–L2 Arabic students typically first acquire the glottals, voiceless emphatic consonants, and the uvular stop. These may be followed by the uvular and pharyngeal fricatives and the voiced emphatic consonants. This hypothesis would suggest that there are important differences between our current study involving L1 English–L2 Arabic learners and the first language acquisition model proposed by Amayreh and Dyson (1998) for L1 Arabic learners. Therefore, this calls for more studies into the basic phonological stages of Arabic second language acquisition in order to inform Arabic language instructors concerning their students’ greatest pronunciation needs. We began this project as a response to Ryding (2013)’s claim that teachers often lack technical training and resources to provide explicit feedback for the correction of pronunciation errors. Ultrasound technology allows visualization of the tongue during sound production, leading to more precise analysis of articulatory production, which allows us to describe and understand differential production. We have found that ultrasound can be used in class to provide students with immediate visual feedback of their own articulation. The pedagogical benefits of visual modeling for speaker modification and self-monitoring in relation to normative production allow for a technical hands-on approach to pronunciation error correction and feedback. 107

‫‪Amanda Eads, Jodi Khater, and Jeff Mielke‬‬

‫‪Appendix‬‬ ‫‪INITIAL POSITION‬‬ ‫‪Emphatics – Initial‬‬ ‫صوْت‬ ‫َ‬

‫صفة‬ ‫ِ‬

‫صيغة‬

‫صباح‬ ‫َ‬

‫صا ِدق‬

‫ضبْط‬ ‫َ‬

‫ضابِط‬

‫ضوْر‬ ‫َ‬

‫ض ْمني‬ ‫ِ‬

‫ضيق‬

‫طَريق‬

‫طائِرة‬

‫طول‬

‫ِطبّي‬

‫طينة‬

‫ظَرْف‬

‫ظا ِهرة‬

‫ظُ ْلمة‬

‫ِظ ّل‬

‫ِظهْري‬

‫‪Non-Emphatics – Initial‬‬ ‫َسبّاح‬

‫سا ِدس‬

‫َسوْداء‬

‫ِس ْمنَة‬

‫سي َرة‬

‫َدبّاب‬

‫داخل‬ ‫ِ‬

‫دَوْر‬

‫ِد َم ْشق‬

‫دين‬

‫ت َِركَة‬

‫تاجر‬ ‫ِ‬

‫توت‬

‫تِبْني‬

‫تين‬

‫َذرَفَ‬

‫ذا ِك َرة‬

‫ُذ َرة‬

‫ِذلَّة‬

‫ِذ ْهني‬

‫‪Uvulars – Initial‬‬ ‫قَريب‬

‫قا ِدم‬

‫قُرْب‬

‫قِطاع‬

‫قي َمة‬

‫غَريب‬

‫غالِب‬

‫ُغ َرباء‬

‫ِغذاء‬

‫َغ ْي ُر‬

‫خَريج‬

‫خا ِدم‬

‫ُخسْران‬

‫ِخالل‬

‫َخيْل‬

‫‪Glottals – Initial‬‬ ‫أَجْ َمل‬

‫آ ُكل‬

‫أُ ْخرُجْ‬

‫إِ ْمتِحان‬

‫إِيْقاف‬

‫هَ ْمزة‬

‫ها ِدف‬

‫هُنا‬

‫ِه َّمة‬

‫ِهياج‬

‫‪Pharyngeals – Initial‬‬ ‫َحرْب‬

‫حازم‬ ‫ِ‬

‫ُح ْكم‬

‫ِحساب‬

‫َحي ُ‬ ‫ْث‬

‫َع َربي‬

‫عا ِدل‬

‫ُعلَماء‬

‫ِعالج‬

‫َعيْن‬

‫‪108‬‬

‫‪Arabic L2 phonological acquisition‬‬

‫‪MEDIAL POSITION‬‬ ‫‪Emphatics – Medial‬‬ ‫صل‬ ‫بَ َ‬

‫إ ْقتِصاد‬

‫ُوصول‬

‫خاصرة‬ ‫ِ‬

‫فَصيل‬

‫ضب‬ ‫َغ َ‬

‫ِرضاء‬

‫عُضو‬

‫واضح‬ ‫ِ‬

‫خَضيب‬

‫َحطَب‬

‫َمطا ِعم‬

‫رُطوبة‬

‫راطب‬ ‫ِ‬

‫بَطيخ‬

‫َعظَ َمة‬

‫َمظا ِهر‬

‫حُظوظ‬

‫ُم ِظ ّل‬

‫نَظيف‬

‫‪Non-Emphatics – Medial‬‬ ‫َع َسل‬

‫فَساد‬

‫فَسولية‬

‫حاسب‬ ‫ِ‬

‫َرسيل‬

‫أَدَب‬

‫ِرداء‬

‫حُدود‬

‫خا ِدم‬

‫قَديم‬

‫َب‬ ‫َكت َ‬

‫تاحف‬ ‫َم ِ‬

‫ُكتُب‬

‫راتِب‬

‫قَتيل‬

‫َح َذ َر‬

‫َمذا ِهب‬

‫أُ ُذن‬

‫با ِذل‬

‫لَذيذ‬

‫‪Uvulars – Medial‬‬ ‫ف َق َط‬

‫ْإلتِقاء‬

‫حُقوق‬

‫عاقِل‬

‫َحقيبة‬

‫َرغَب‬

‫ُمغالَبَة‬

‫ضُغوط‬

‫را ِغم‬

‫صغير‬ ‫َ‬

‫َد َخنَ‬

‫ُمخالِف‬

‫صُخور‬

‫زاخر‬ ‫ِ‬

‫َرخيص‬

‫‪Glottals – Medial‬‬ ‫َسأ َ َل‬

‫فِئات‬

‫رُؤوس‬

‫قَبائِل‬

‫ضئيل‬ ‫َ‬

‫َسهَر‬

‫إ ْنتِهاك‬

‫سُهولة‬

‫قا ِهر‬

‫َمناهيج‬

‫‪Pharyngeals – Medial‬‬ ‫َر َحب‬

‫بِحار‬

‫صحُف‬ ‫ُ‬

‫صاحب‬ ‫ِ‬

‫َرحيل‬

‫َد َع َم‬

‫ميعاد‬

‫تَفاعُل‬

‫ال ِعب‬

‫َسعيد‬

‫‪109‬‬

‫‪Amanda Eads, Jodi Khater, and Jeff Mielke‬‬

‫‪FINAL POSITION‬‬ ‫‪Emphatics – Final‬‬ ‫أًرْخَص‬

‫أ ْشخاص‬

‫لُصوص‬

‫خالِص‬

‫ت َْخصيص‬

‫أ ْبيَض‬

‫أ ْغراض‬

‫قُروض‬

‫عارض‬ ‫ِ‬

‫َمريض‬

‫َربَط‬

‫ْ‬ ‫أخطاط‬

‫َمبسوط‬

‫ساقِط‬

‫ُمحيط‬

‫َحظَّ‬

‫إيقاظ‬

‫َمحْظوظ‬

‫الحظ‬ ‫ُم ِ‬

‫َحفيظ‬

‫‪Non-Emphatics – Final‬‬ ‫َعكَس‬

‫إحْساس‬

‫فُلوس‬

‫البِس‬

‫تأْسيس‬

‫يَد‬

‫أ ْفراد‬

‫قُيود‬

‫والِد‬

‫تَ ْغريد‬

‫جابَت‬

‫حُتات‬

‫بُيوت‬

‫ثابِت‬

‫َمبيت‬

‫لَ َّذ‬

‫لِذاذ‬

‫َمأخوذ‬

‫نافِذ‬

‫تَ ْنفيذ‬

‫‪Uvulars – Final‬‬ ‫طَبَق‬

‫ْإلحاق‬

‫سوق‬

‫غارق‬ ‫ِ‬

‫َرفيق‬

‫َم ْبلَغ‬

‫َمساغ‬

‫فُروغ‬

‫بالِغ‬

‫تَ ْفريغ‬

‫نُ َسخ‬

‫إ ْستِ ْنساخ‬

‫فُخوخ‬

‫راسخ‬ ‫ِ‬

‫تاريخ‬

‫‪Glottals – Final‬‬ ‫َم ْبدَأ‬

‫أجْزاء‬

‫سوء‬

‫َمبا ِدئ‬

‫ُمسيء‬

‫َشبَه‬

‫إتِّجاه‬

‫ُوجوه‬

‫ُمتَشابِه‬

‫تَرفيه‬

‫‪Pharyngeals – Final‬‬ ‫فَ َرح‬

‫إصْالح‬

‫َمطروح‬

‫واضح‬ ‫ِ‬

‫َمريح‬

‫قَطَع‬

‫إ ِّ‬ ‫طالع‬

‫فُروع‬

‫مانِع‬

‫َوسيع‬

‫‪110‬‬

Arabic L2 phonological acquisition

Notes 1 This research is a large ongoing project conducted at North Carolina State University. The recordings and tongue tracings for this particular study were obtained with the assistance of Jaxon Buck, Paige Duffy, Lyra Magloughlin, Frankie Pennington, Hannah Smith, and Eric Wilbanks. This work is supported by the North Carolina State University Department of English, NCSU’s Provost’s Professional Experience Program, and NSF Grant BCS 1451475 “Phonological Implications of Covert Articulatory Variation” to Jeff Mielke. 2 Throughout the chapter, we use the term “phone” rather than phoneme, for example, because our concern is with how the sounds are articulated, not their phonological content.

References Al-Ani, S., 1970. Arabic phonology: An acoustical and physiological investigation. The Hague: Mouton. Ali, L. H. and Daniloff, R. E., 1972. A contrastive cinefluorographic investigation of the articulation of emphatic-non-emphatic cognate consonants. Studia Linguistica [online], 26, 81–105. Available from: http://dx.doi.org/10.1111/j.1467-9582.1972.tb00589.x [Accessed June 20, 2017]. Al-Solami, M., 2013. Arabic Emphatics: Phonetic and Phonological Remarks. Open Journal of Modern Linguistics [online] 3(4), 314–18. Available from: http://dx.doi.org/10.4236/ojml.2013.34040 [Accessed June 20, 2017].Al-Solami, M., 2017. Ultrasound Study of Emphatics, Uvulars, Pharyngeals and Laryngeals in three Arabic Dialects. Journal of the Canadian Acoustical Association 45(1), 25–35. Amayreh, M., 2003. Completion of the consonant inventory of Arabic. Journal of Speech, Language and Hearing Research, 46 (3), 517–529. Amayreh, M. and Dyson A. T., 1998. The Acquisition of Arabic consonants. Journal of Speech, Language and Hearing Research, 41 (3), 642–653. Baker, A., 2005. Palatoglossatron 1.0 [Computer Software]. Available from: http://dingo.sbs.arizona. edu~apilab/pdfs/pgman.pdf [Accessed June 20, 2017]. Boff-Dkhissi, M. C., 1983. Contribution à l’étude expérimentale des consonnes d’arrière de l’arabe classique (locuteurs marocains). Travaux de l’Institut Phonétique de Strasbourg 15. Brustad, K., et al., 2004. Alif Baa with DVDs: An introduction to Arabic letters and sounds, part one. 2nd ed. Washington, DC.: Georgetown University Press. Catford, J. C., 1977. Fundamental problems in phonetics. Edinburgh: University Press. Davidson, L., 2006. Comparing tongue shapes from ultrasound imaging using smoothing spline analysis of variance. The Journal of the Acoustical Society of America, 120 (1), 407–415. Delattre, P., 1971. Pharyngeal features in the consonants of Arabic, German, Spanish, French, and American English. Phonetica [online], 23,129–155. Available from: http://dx.doi.org/10.1159/000259336 [Accessed June 20, 2017]. Dolgopolsky, A., 1977. Emphatic consonants in Semitic. Israeli Oriental Studies, 7, 1–13. Durham, K., et al., 2014. Native English speakers’ perception of Arabic Emphatic Contrasts. The Journal of the Acoustical Society of America [online] 135(4). Available from: http://dx.doi. org/10.1121/1.4877757 [Accessed June 20, 2017]. Dyson, A. T. and Amayreh, M., 2000. Phonological errors and sound changes in Arabic-speaking children. Clinical Linguistics and Phonetics, 14, 79–109. Elgendi, A., 2001. Aspects of pharyngeal coarticulation. Thesis (PhD). University of Amsterdam. Ghazeli, S., 1977. Back consonants and backing co-articulation in Arabic. Thesis (PhD). University of Texas at Austin. Gu, C., 2002. Springer series in statistics: Smoothing spline ANOVA models. New York: Springer-Verlag. Hayes-Harb, R. and Durham, K., 2016. Native English speakers’ perception of Arabic emphatic consonants and the influence of vowel context. Foreign Language Annals, 49 (3), 557–572. Hueber, T., et al., 2008. Acquisition of ultrasound, video and acoustic speech data for a silent-speech interface application. In: S. Rock, S. Fuchs and Y. Laprie eds. Proceedings of the eighth international seminar on speech production [online], 8–12 December 2008 Strasburg, France. INRIA, 365–369. Available from: http://issp2008.loria.fr/proceedings.html [Accessed June 20, 2017]. Khattab, G., 2011. Acquisition of Lebanese Arabic and Yorkshire English /l/ by bilingual and monolingual children: A comparative spectrographic study. In: Z.M. Hassan and B. Heselwood, eds. Instrumental studies in Arabic phonetics. Philadelphia, PA: John Benjamins, 325–354.

111

Amanda Eads, Jodi Khater, and Jeff Mielke Khattab, G., et al., 2006. Acoustic and auditory differences in /t/ and /t/ opposition in male and female speakers of Jordanian Arabic. In: S. Boudelaa, ed. Perspectives on Arabic linguistics xvi. Philadelphia, PA: John Benjamins, 131–160. Laufer, A. and Baer, T., 1988. The emphatic and pharyngeal sounds in Hebrew and Arabic. Language & Speech, 24, 39–61. Laufer, A. and Condax I. D., 1979. The Epiglottis as articulator. Journal of the International Phonetic Association, 9, 50–65. Mielke, J., 2015. An ultrasound study of Canadian French rhotic vowels with polar smoothing spline comparisons. Journal of the Acoustical Society of America, 137 (5), 2858–2869. Odisho, E., 2005. Techniques for teaching comparative pronunciation in Arabic and English. Piscataway, NJ: Gorgias Press. R Core Team., 2016. R: A Language and Environment for Statistical Computing. [Computer Software]. R Foundation for Statistical Computing, Vienna, Austria (www.R-project.org). [Accessed: June 20, 2017]. Ryding, K., 2006. Teaching Arabic in the United States. In: K. M. Wahba, et al., eds. Handbook for Arabic language teaching professionals in the 21st century. Mahwah, NJ: Erlbaum, 13–20. Ryding, K., 2013. Teaching and learning Arabic as a foreign language: A guide for teachers. Washington DC: Georgetown University Press. Shahin, K., 2011. A phonetic study of guttural laryngeals in Palestinian Arabic using laryngoscopic and acoustic analysis. In: Z.M. Hassan and B. Heselwood, eds. Instrumental studies in Arabic phonetics. Philadelphia, PA: John Benjamins, 129–140. Showalter, C. and Hayes-Harb, R., 2015. Native English speakers learning Arabic: The influence of novel orthographic information on second language phonological acquisition. Applied Psycholinguistics, 36 (1), 23–42. Younes, M., 1993. Emphasis spread in two Arabic dialects. In: M. Eid and C. Holes, eds. Perspectives on Arabic linguistic v. Philadelphia: John Benjamins, 119–145. Yuan, J. and Liberman, M., 2008. Speaker identification on the SCOTUS corpus. In: Proceedings of acoustics, 29 June – 4 July 2008 Paris. New York: AIP Publishing LLC, 5687–5690. Zaba, A., 2007. Native speakers and Arabic pharyngealization contrasts: perceptual and acoustic data. The Journal of the Acoustical Society of America [online]12. Available from: doi: http://dx.doi. org/10.1121/1.4781871 [Accessed: June 20, 2017]. Zawaydeh, B., 1999. The phonetics and phonology of gutturals in Arabic. Thesis(PhD). Bloomington, IN: Indiana University.

112

6 THE L2 ACQUISITION OF MODERN STANDARD ARABIC FINAL CONSONANT CLUSTERS BY L1 CHINESE SPEAKERS Mona Maamoun Chinese learners of Modern Standard Arabic as a foreign language show many pronunciation complexities that significantly hinder their communication with Egyptian Arabic native speakers. One of these complexities is the pronunciation of Arabic coda clusters which the Chinese phonological system lacks. Here, I test the role and validity of both the contrastive analysis approach and the universal sonority scale principle in acquiring the pronunciation of L2 Modern Standard Arabic coda clusters by L1 Chinese learners. This study is the starting point for conducting a series of studies (in progress) that target the Chinese population and focus on identifying, describing, explaining, and evaluating their acquisition of L2 Arabic (Modern Standard and/or Egyptian Colloquial) phonological system. Findings of such studies will help in developing teaching materials to address Arabic pronunciation errors that influence the intelligibility of learners’ speech. The participants of the present study are ten Chinese students who studied Modern Standard Arabic for two years at a university in China. The target Arabic clusters were elicited via a test of 200 CVCC mono-syllabic words. Words are sonority categorized into: 73 words of clusters that conform to sonority, 103 words of clusters that violate sonority, and 24 of sonority plateau. The data were collected through observation, recordings, and the reading technique. The statistical findings revealed that the Chinese learners apply seven systematic error pattern phonological processes (transfer and universal errors) to facilitate the pronunciation of Arabic coda clusters. The highest percentages of the error pattern phonological processes that occurred are related to vowel epenthesis, followed by vowel insertion and then cluster substitution. The most difficult type of Arabic clusters to be pronounced by the Chinese learners is the clusters that violated the sonority.

6.1 Introduction The influence of the learner’s first language plays a role in all levels of the acquisition and learning of the L2. This influence is known as “transfer.” The interference of L1 in the process of second language learning is regarded as negative transfer. At the opposite side is positive transfer, which covers the common features between L1 and L2 that facilitate the learning 113

Mona Maamoun

process (Corder 1967). Zhanming (2014) classified L1 transfer into four levels: sound transfer, word transfer, syntax transfer, and culture transfer. Zhanming (2014) argued that sound transfer is by far the most obvious type of L1 transfer and targets L2 pronunciation in a negative way. However, when L1 shares words with the L2, learning of the L2 becomes easier. L1 transfer can take place across all linguistic levels: phonetics, phonology, morphology, syntax, and pragmatics. Phonological transfer is the main scope of the current chapter, and it is defined by Lowie (2015) as the transfer of the L1 phonological system ranging from the individual sound to the prosodic characteristics of language. Moreover, Jarvis and Pavlenko (2007) define phonological transfer as the way in which a person’s knowledge of a specific language not only affects the production of a new language sound system but also its perception. L2 learners pronounce a foreign language with their L1 accent. This accented speech affects both their expressive and receptive skills as well as social interaction (Derwing and Munro 2005). Studies which focus on investigating the transfer of L1 prosody, such as Rasier and Hiligsmann (2007), are very few. The transfer of L1 phonological system into the L2 could work against acquiring accurate L2 pronunciation. It is observed that students of Arabic as a foreign language face some pronunciation problems that not only affect their speaking and listening skills, but they also impact their reading and writing skills to some extent. Examples of the writing errors that result from pronunciation errors can be taken from the writings of Chinese learners of Arabic of intermediate mid-level. They may substitute /s/ by /ṣ/ as in the word naas “people” or substitute /r/ by /l/ as in kibaar “oldies.” The sound /ð/ could be replaced by /z/ as in ʔustaaða “professor.” Metathesis could appear in their writing, such as the verb yaħtafil “to celebrate” written as yaftaħil. Cluster deletion also appears, as in the word šarm “an Egyptian city” written as šaam. Listening comprehension correlates with the level of pronunciation (Gilbert 1995). The degree, amount, and load of pronunciation errors correlate in an opposite relation with the intelligibility of speech. As pronunciation errors increase, the intelligibility level decreases. Phonological errors that lower the accuracy level of students’ pronunciation are not restricted to the speech of those who belong to the novice proficiency level, but these errors also exist in the speech of those who reach higher levels, and they may persist until the advanced level. In addition to L1 transfer, age of the learner is an important factor that cannot be neglected. Many studies such as Johnson and Newport (1989) noted that younger learners acquire an L2 at a higher and more near-native level than older learners. Adults vary in their levels of acquiring L2, which may due to either cognitive abilities, as noted by Hyltenstam and Abrahamsson, (2003) and DeKeyser (2013), or social factors and motivation. Nahla (2006) noted that students who pass the beginner level are supposed to be able to utter correct and acceptable pronunciations of Arabic sounds and should be able to discriminate between sounds, syllables, and stress. However, this is not the case. Rather, students at the intermediate and advanced levels seem to have major phonological errors that affect their pronunciation. One factor may be due to the lack of studies on L2 Arabic phonology, such as those of Asfoor (1982) and Shehata (2015). The lack of attention given to the importance of the pronunciation in the classroom may be another factor. As noted by Jarrah (2012), phonological training should be part of the learning process of a second language. Hence, introducing phonetics and phonology to L2 learners of Arabic at early stages leads to the acquisition of accurate pronunciation, and “intelligible pronunciation is a basic element of communication competence” Morley (1991). Chinese L1 speaking learners of Arabic face major pronunciation problems. Being intelligible to Arabic native speakers is a very challenging task to be achieved by Chinese 114

L2 acquisition of final consonant clusters

speakers. There are many interrelated factors that influence the acquisition of L2 pronunciation. One factor is the transfer of the Chinese L1 phonological system into L2 (pronunciation errors may result from the discrepancy between the L2 phonological system and the L1 one). Another factor is related to universals, and a third is related to the amount of exposure to L2. Based on the first factor, it can be said that pronunciation errors that are observed in the speech of Chinese learners of Arabic may result from the great difference between the Arabic and Chinese phonological systems. The Chinese phonological system lacks many phonological characteristics which are essential and distinctive in the Arabic (MSA) phonological system, such as coda clusters. The acquisition of these MSA coda clusters by Chinese native speakers is the target of the current study. As for the second factor, the acquisition of pronunciation of MSA coda clusters may be governed by a universal markedness of sonority such as the sonority scale principle (SSP), as explained further later. There are numerous sonoritybased studies that focus on describing the acquisition of L2 clusters by different L1 speakers, such as Boudaoud and Cardoso (2009) and Yavas and Someillan (2005). These studies focus on the acquisition of English initial clusters of an onset phoneme /s/ by learners whose L1 has clusters but lacks this specific type of clusters. Other types of study target learners whose L1 has no clusters at all. For example, Enochson (2014) examines the acquisition of L2 English initial /s/ clusters and other types of clusters by learners whose L1 is either Chinese Mandarin, Chinese Cantonese, or Japanese. Studies on L2 Arabic phonology are very rare, especially when the L1 is Chinese. The main objective of the current study is not only to describe and analyze the acquisition of L2 MSA coda clusters by L1 Chinese learners but also to illustrate how the different factors interact with each other in the acquisition of L2. The results of this study aim to inform the design of instructional materials for L2 Arabic learners in the pronunciation of clusters. It has been hypothesized by Clements (1990), Giegerich, (1992), and Kenstowicz (1994) that the distribution and sequencing of segments within syllables and across syllables could be described and explained with reference to a sonority sequencing principle (SSP). The SSP applies to coda clusters in which the sonority value decreases after the nucleus until it reaches a minimum value in the coda position. Hogg and McCully (1987) assigned sonority values which are appropriate for measuring the sonority distance between coda clusters, as displayed in Table 6.1 (following Hogg and McCully 1987). Based on the sonority scale, the Arabic (MSA) coda cluster shows three major categories: conformity (unmarked clusters), sonority violating (marked clusters), and plateaus, as is illustrated in more detail in the next section on the Arabic phonological system. Table 6.1 Sonority values of different types of sound classes The Sound Class

The Sonority Value

Low vowels Mid vowels High vowels Rhotics Laterals Nasals Voiced fricatives Voiceless fricatives Voiced stops Voiceless stops

10 9 8 7 6 5 4 3 2 1

115

Mona Maamoun

6.1.1  The Arabic phonological system The MSA phonological system is characterized by a simple vowel system and a more complex consonantal system. Hence, it can be predicted that students who learn Arabic as a foreign language face major problems in pronouncing the Arabic consonants. The Arabic phonemic inventory is composed of 28 consonant phonemes (see the Introduction to this volume: Table 0.1). Accordingly, the Arabic phonemic system contains some sound classes that could be considered as specific, marked, such as the emphatic phonemes, the pharyngeals, and the uvular /q/. This markedness may be the result of the full occupation of the Arabic sounds across the vocal tract as they extend from the labial to glottal The vowel system is much simpler. It consists of six vowel phonemes three long vowels / ii/, /uu/, /aa/ and their three short counterparts. Kopczynski and Meliani (1993) described the MSA vowel allophones as follows. The vowel /i/ has three variants: the emphatic allophone of /ii/, which is [ɨ:], occurs in a context with an emphatic sound such in ʕaṣ ɨ:r “juice”; the nonemphatic variant of /i:/ occurs elsewhere, as in Sali:m “safe”; and the lower allophone of the short /i/, which is [ɪ], occurs in unstressed syllables as in ṭa:lɪ b “student.” The vowel /a/ has its emphatic variant [a] in a phonological environment with emphatic sounds such in ṭabl “drum,” and the front variant [a] occurs elsewhere, as in sahl “easy.” The long /a:/ has also its emphatic variant [a:] in a phonological environment with emphatic sounds, as in ṭa:lɪb “student” and [a:] elsewhere, as in na:m “he slept.” The vowels /u (:)/ are lowered contiguous to emphatic sounds as qʊṭr “diameter” and ṣʊ:ra “picture.” Elsewhere, [u(:)] is found as in fu:l “beans” and kul “eat.” As for Arabic syllabic structure, it is noted that an MSA syllable is composed of an obligatory onset and a rime (Watson 2002). The rime consists of a nucleus (vowel) and an optional coda. The coda could be up to two consonants, as shown here: Syllable Onset

C

Rime Nucleus

Coda

V

(CC)

The Arabic phonotactic system has five types of syllables, as noted by many linguists (e.g., Al Ghamdi 2001). One short/weak syllable, which is the CV, two strong/ heavy syllables, which are the CVV and CVC, and two super strong/heavy CVVC and CVCC (see Table 6.2). Al Ani (1978) noted that there are six types, not five; the sixth is CVVCC, in words such as maarr “passed by.” Table 6.2 MSA syllabic structure Syllable Type

Example

CV CVV CVC CVVC CVCC

/fii/ “in” /maa/ “what” /min/ “from” /naar/ “fire” /baħr/ “sea”

116

L2 acquisition of final consonant clusters

These syllables do not occur freely in any position within the word; there are some restrictions on their distribution. Each syllable should contain a vowel. The onset is an obligatory unit. MSA syllables must begin with only one consonant onset, which means that there is no word or syllable that may start with a consonant cluster. In addition, Arabic syllables never begin with a vowel. Turning to the status of the coda in the Arabic syllables, we find that the coda is not an obligatory member of the Arabic syllable; it is optional. If the syllable ends without a coda, it is categorized as an open syllable, but if it ends with a coda, it is categorized as a closed one. MSA consonant clusters exist only in final position and have their own specifications. There are certain permissible types of CC in coda position. Al Tamimi and Al Shboul (2013), in their study of the MSA coda clusters phonotactics, from the perspective of the sonority sequencing principle (SSP), listed the different types of MSA CC. Their sample was 493 CVCC mono-syllabic words collected from The Hans Wehr Dictionary of Modern Written Arabic (Cowan 1994). These words are either deverbal nouns or underived nouns. Their findings could be grouped under seven categories with seven essential target classes: flap, nasal, lateral, voiced fricatives, voiceless fricatives, voiced stop, and voiceless stops. The target classes can be either the first member of a coda cluster C1 or a second member of coda cluster C2. The occurrence of these classes within other sound classes in coda clusters can be summarized in Table 6.3. In a context of C1C2, in which /r/ is the target sound class and it occurs as C1, the type of C1C2 cluster is the sonority-conforming, as in farm “to cut into small pieces.” In another context, if the same target sound class is C2, the type of C1C2 cluster is the sonority-violating, as in xamr “wine.” Finally, if the C1 is the same sound as C2, this context represents a sonority plateau, as in ħarr “hot weather.” It is observed that there are instances of MSA clusters that follow the SSP (conforming clusters), those that violate the SSP (violating clusters), and those of equal sonority values (plateaus). The sonority distance within sonority conforming clusters ranges from 1 up to 6, while the sonority distance within sonority violating clusters ranges from −6 up to −1, and finally sonority plateau clusters of the same sonority values are of zero sonority distance. It is hypothesized that the acquisition of clusters with negative sonority distance is more difficult than the acquisition of those with Table 6.3 MSA coda clusters The Preceding Cluster Member

The Target Sound Class

Nasal, voiced fricative, voiceless fricative, voiceless stop, voiced stop, glottal stop Nasal, voiced fricative, voiceless fricative, voiceless stop, voiced stop Nasal, voiced fricative, voiceless fricative, voiceless stop, voiced stop Voiced fricative, voiceless fricative, voiceless stop, voiced stop Voiceless fricative, voiceless stop, voiced stop, glottal stop Voiceless stop, voiced stop Voiceless stop, voiced stop

Flap Lateral Nasal Voiced Fricative Voiceless Fricative Voiced Stop Voiceless Stop

117

The Following Cluster Member Nasal, voiced fricative, voiceless fricative, voiceless stop, voiced stop, glottal stop Nasal, voiced fricative, voiceless fricative, voiceless stop, voiced stop Nasal, voiced fricative, voiceless fricative, voiceless stop, voiced stop Voiced fricative, voiceless fricative, voiceless stop, voiced stop Voiceless fricative, voiceless stop, voiced stop, glottal stop Voiceless stop, voiced stop Voiceless stop, voiced stop

Mona Maamoun

positive values. Many studies which focused on the L2 phonology, such as Broselow and Finer (1991) and Eckman and Iverson (1993), noted that the onset clusters with a positive sonority distance are less marked and acquired earlier than the onset clusters with a negative sonority distance value. With respect to the MSA clusters, those which violate the SSP are of negative sonority distance values; consequently, they are more marked than those of positive values which conform to the SSP. Hence, it is predictable that these types of clusters are the complex ones that will be acquired later. In the next section, a related review of the Chinese phonology at the same phonological levels is offered in order to highlight the phonological difference between the two.

6.1.2  Modern standard Chinese (Mandarin) phonology The Chinese language consists of several dialect families, most of which are mutually unintelligible. The Northern family, which is also known as the Mandarin family, is considered the largest dialect family. Over 70% of the Chinese population speak Mandarin. The consonant inventory of Standard Chinese (SC) consists of 22 phonemes, as noted by Duanmu (2006) and illustrated in Table 6.4. There are some constraints on the distribution of the Chinese phonemes such as that the phoneme /ŋ/ cannot occur initially. At the same time, the word-final coda position is restricted to either /n/ or /ŋ/. The palatal fricatives [tɕ],[tɕʰ], and [ɕ]/occur in limited environments, either with front vowels or with glides. Voicing is not distinctive between pairs of stops or affricates of the same place and manner of articulation; however, it is aspiration that is distinctive. Voicing is neutralized in syllable endings, thus [b] [d] [g] are realized as [p] [t] [k], respectively. As for Chinese syllabic structure, a Chinese syllable consists of two parts: the initial and final parts. The two parts are made of up to four sounds – CGVX, where C is a consonant, G a glide, V a vowel, X a nasal or an off-glide of a diphthong, and VX the rime. When both C and G are present, they are realized as one sound C with a secondary articulation. Thus, the SC word swei “age” is phonetically quite different from the English word swei “sway.” In addition to these constraints, there is one more characteristic of SC phonology closely related to the current research: SC does not use consonant coda clusters. Instead, it has syllables in which consonants must be followed by vowels. From this brief review of the MSA and standard Chinese phonological systems and from the contrastive point of view, we can summarize the differences between the two phonological systems across the phonemic and syllabic structure level as follows: 1 Chinese lacks emphatic phonemes, trill /r/, uvular, pharyngeal, and glottal places of articulation. Table 6.4 Standard Chinese consonants following Duanmu (2006, p. 351) Labial Plosive Fricative Nasal Affricate Liquids

p pʰ m

Labiodentals

Dental

f

t tʰ s

Palatal

Retroflex

Velar

ɕ

ʂ

k kʰ x

tɕ tɕʰ

ʈʂ ʈʂʰ r

n ts tsʰ l

118

ŋ

L2 acquisition of final consonant clusters

2 Voicing is not a distinctive feature to discriminate between sounds in Chinese. 3 Chinese coda consonants are restricted to either /n/or /ŋ/. 4 Chinese lacks the different types of Arabic syllabic structure, such as CVCC. It can be predicted from this comparison that L1 Chinese learners of Arabic may face difficulties not only in articulating final consonants but also in the articulation of final consonant clusters, which is more challenging for them.

6.2  Research questions The study aims at investigating the pronunciation of L2 Arabic coda clusters by L1 Chinese learners in CVCC mono-syllabic words through answering the following questions: 1 What modification strategies do Chinese-speaking learners of Arabic apply when they are pronouncing L2 Arabic CVCC? 2 What are the phonological processes that appear in their pronunciation of clusters? 3 Is there an effect of L1 transfer in their L2? 4 Does the universal concept of markedness have a role in the way L2 Arabic clusters are pronounced?

6.3 Methods 6.3.1 Participants The participants of the current study were ten Chinese students at intermediate low level (based on the ACTFL OPI proficiency test). They were five females and five males, ranging in age from 19 to 20 years. They studied MSA for two years at Ningxia University in China and then came to Alexandria University to study for a third year. Their first- and second-year Arabic language instruction consisted of six contact hours per week. When they joined the current study, they were in the middle of their third-year program in Alexandria, which offered 15 contact hours of Arabic instruction per week. Despite their previous exposure to Arabic, the participants still have major pronunciation errors, as they never have attended a pronunciation class in China. They noted that applying the strategy of listening and repeating is not sufficient to enhance their pronunciation skills.

6.3.2  Speech data In the current study, the author observed the participants for around seven months, taking notes on their phonological errors: phonemic, syllabic structure, and intonation errors. Moreover, the author collected around two hours of spontaneous speech from each individual participant during this period. A qualitative evaluation was built upon these evaluative techniques (class observation and open conversation); hence the most common phonological errors could be detected. Besides these observations, a personal interview was conducted with each participant about the most challenging Arabic speech sounds from his or her point of view. Unfortunately, the elicited spontaneous speech sample does not cover all types of MSA coda clusters. It is difficult to elicit the target words of the current study from picture-naming or story-telling techniques only. Therefore, a structured word list was designed to focus on the different types of MSA coda clusters. Many studies on L2 acquisition of clusters apply the reading technique 119

Mona Maamoun

to elicit a structured and controlled speech sample of a focused target (e.g.,; Fatemi 2012; Nguyen, A., and Brouha, C., 1998; Ahmad and Nazim 2014). The reading-aloud test which was used in the present study consisted of 200 mono-syllabic words of CVCC syllabic structure (see Appendix for the word stimuli used). A sample of the test is provided in Table 6.5. The reading technique was applied using specific procedures to avoid eliciting an unreliable speech sample, as follows: • Participants were asked to read the list of words from seven word sheets; each sheet contained 30 words except the first one, which contained 20 words (this sheet contained the participant’s personal information). • Words were vowelled (i.e., provided with short diacritic vowels). • Participants were instructed to read at a normal speech rate. • Participants were informed that the target words were not verbs but rather deverbal nouns or underived nouns that must end with a consonant. • Imitation or repeating techniques were not used. The words were categorized based on sonority-. They were classified into 73 words that conform to the SSP, 103 words that violate the SSP, and 24 of sonority plateau. The test was composed of four columns: the first was for the transcription of the target words, the second for the transcribed responses of the participant, the third for the specification of the type of cluster modification process, and the fourth for any notes (such as common error patterns occurring in the participants’ speech). Live phonemic transcription by two well-trained phoneticians (including the author) was undertaken. All the samples were also recorded on an HP Mini laptop using Praat software. Later, the recordings of each participant were played back and a transcription was done to be compared with the live version, acting as an intra-judge agreement strategy across the whole sample; the two transcribers listen to the recordings, each one separately, and transcribed the sample to compare between her transcription from the recordings and those which were done live. Then the two transcribers applied an inter-judge agreement strategy by comparing between their final results. Table 6.5 A sample from the MSA coda cluster test Target Words

‫فَرْ ْم‬ ‫فُرن‬

ْ‫َخ ْمر‬ ْ‫ُع ْمر‬

Transcription of Target Words

farm “mincing”

Participant’s Response Conforming Clusters /r/ + nasal

furn “oven”

xamr “wine” ʕumr “age”

‫ض ْم ْن‬ َ

ḍamn “guaranteeing”

‫َغ ْي ْم‬

γaym “clouds”

Violating Clusters Nasal + /r/

Plateaus Nasal

120

Cluster Modification Phonological Processes

Notes

L2 acquisition of final consonant clusters

6.3.3 Analysis When a participant gave a correct response (i.e., pronounced the coda cluster without applying any of the modification strategies, such as vowel insertion), a check mark was added in the third column; if not, the transcribed replaced word was added to the third column. The type of error pattern phonological processes was inserted in the fourth column. The percentage of correct responses under the three types of words in the speech of each participant was computed. In order to be able to reach the percentage of occurrence of each process, the following steps were followed: • The whole sample in each type of word and the percentage of each type of cluster were counted. • Cluster modification phonological processes in the speech of each participant were computed first. This allowed the calculation of high scoring versus low scoring phonological processes in each type of cluster. After that, statistical analysis of the data was applied. Data were fed to the computer and analyzed using IBM SPSS software package version 20.0. Quantitative data were described using mean and standard deviation. The significance of the obtained results was judged at the 5% level. The tests used were one-way ANOVA with repeated measures (each individual participant read three different types of words, and we compared between his/her production along the three types of words).

6.4  Results and discussion The participants of the current study produced some phonological processes in pronouncing the 200 CVCC target words. However, they were able to correctly pronounce some of these words. In each type of the sonority-based target words (i.e., sonority conforming, sonority violating, and sonority plateau), the results of the frequency of occurrences of the error types (number of times in which the participant produce the error as phonological processes) and percentages of occurrences in each participant are illustrated in Tables 6.6, 6.9, and 6.12. The rank of the mean percentages of occurrences of each error type are represented in Tables 6.7, 6.10, and 6.13. The number of correct pronunciation and their percentages in each participant are clarified in Tables 6.8, 6.11, and 6.14. The results show the occurrence of seven types of phonological processes which the Chinese participants produce as error types to overcome the difficulty of pronouncing the target words. These phonological processes are: 1 “vowel epenthesis,” in which a short vowel is inserted to break the cluster, as in šarħ “explanation” pronounced as šaraħ 2 “vowel insertion,” in which a short vowel is added after the second member of the cluster, while the final cluster shifts to become the first consonant of a final open syllable, as in šarħ pronounced as šaraħa 3 “cluster substitution,” which occurs when a cluster member is substituted by another consonant, as in šarħ pronounced as šalħ 4 “cluster assimilation” in which one member assimilates to another, as in žanb “offense” pronounced as žamb 5 “doubling,” in which the first cluster member is doubled, as in ʕilm “knowledge” pronounced as ʕillim 6 “cluster deletion,” in which one cluster member is deleted, as in laṭm “slap” pronounced as laṭaa 7 “addition of tanwiin” occurs where either the syllable /-un/ or /-an/ is added to the cluster, as in ʕilm pronounced as ʕilmun. 121

1 2 3 4 5 6 7 8 9 10 Total Error Means

Participants

1 2 1 6 7 15 24 24

1.37 2.74 1.37 8.22 9.59 20.55 32.88 32.88 13.7

23 41 46 51 18 44 12 14 37 37

31.51 56.16 63.01 69.86 24.66 60.27 16.44 19.18 50.68 50.68 44.2

%

Total occurrence

Total occurrence

%

Vowel Epenthesis

Cluster Substitution

Phonological Processes/Error Types

11 5 52 9 38 15 3 3 46 46

Total occurrence 15.07 6.85 71.23 12.33 52.05 20.55 4.11 4.11 63.01 63.01 31.2

%

Vowel Insertion

Table 6.6 Totals and % of errors in conforming clusters (total stimuli words = 73)

1 1 3 2 -

Total occurrence

First Cluster Doubling

1.37 1.37 4.11 2.74 2.4

% 1 1 1 2 2

Total occurrence

Cluster Assimilation

1.37 1.37 1.37 2.74 2.74 1.9

%

3 -

Total occurrence

4.11 4.1

%

Cluster Deletion

3 6 -

Total occurrence

Addition of tanwiin

2.74 8.22 5.5

%

L2 acquisition of final consonant clusters Table 6.7 Mean value of each error type across the ten participants Conforming Clusters Error Types

Mean Value of Total % of Errors Across the Ten Participants

Rank

Cluster substitution Vowel epenthesis (VIN between) Vowel insertion (VIN after) First cluster doubling Cluster assimilation Cluster deletion Addition of tanwiin Total mean % values of the seven error types in conforming cluster

13.70 44.25 31.23 2.40 1.92 4.11 5.48 67.26

5 7 6 2 1 3 4

Table 6.8 Total and % of correct in conforming clusters (stimuli words = 73) Participants

Total Correct

%

1 2 3 4 5 6 7 8 9 10 Mean of % correct responses across the 10 participants

41 24 3 17 19 19 50 54 11 10

56.16 32.9 4.1 23.3 26.0 26.0 68.5 73.9 15.1 13.7 33.9

Each one of these error types was produced to break the clusters in order to facilitate their pronunciation. Before reporting the results, we must highlight the fact that there is an overlap in the total number of occurrences of error types in some participants. For example, Participant 1 produced a total of 120 errors out of the total number of 103 sonority violating words. This is because a participant could apply more than one phonological process in each target word to facilitate the pronunciation. For example, the sonority conforming word xalʕ “extraction” is pronounced as xalaʔa, which means the occurrence of three phonological processes: vowel epenthesis, vowel insertion, and cluster substitution. Another example from the sonority violating words is the word nasl “offspring” pronounced as matil showing the occurrence of two processes: vowel epenthesis and cluster substitution. To avoid any unrepresentative results, the frequency (number) and percentage of occurrence of each error type in each participant is calculated separately from the number and percentage of the occurrence of the correct pronunciation of each participant, as illustrated in a subsequent section. 123

30.10

3

31

31

8

9

10

Total Error Means

2.91

-

7

7.77

16.6

30.10

-

33.01

8

12.62

34

4

2.91

13.59

6

13

3

16.50

5

3

14

2

17

54

54

46

42

54

39

82

78

92

51

57.5

52.43

52.43

44.66

40.78

52.43

37.86

79.61

75.73

89.32

49.51

%

Total occurrence

Total occurrence

%

Vowel Epenthesis

Cluster Substitution

Phonological Processes/Error Types

1

Participants

66

66

23

6

35

72

62

64

13

43

Total occurrence

43.7

64.08

64.08

22.33

5.83

33.98

69.90

60.19

62.14

12.62

41.75

%

Vowel Insertion

Table 6.9 Totals and % of errors in violating clusters (total stimuli words = 103)

7

1

9

Total occurrence

First Cluster Doubling

5.5

-

-

-

-

6.80

-

-

-

0.97

8.74

%

1

Total occurrence

Cluster Assimilation

0.97

-

-

-

-

0.97

-

-

-

-

-

%

Total occurrence

-

-

-

-

-

-

-

-

-

-

-

%

Cluster Deletion

1

9

4

Total occurrence

4.5

-

-

-

0.97

8.74

3.88

-

-

-

-

%

Addition of tanwiin

L2 acquisition of final consonant clusters Table 6.10 Mean values of each error type across the ten participants Violating Clusters Error Types

Mean Value of Total % of Errors Across the Ten Participants

Rank

Cluster substitution Vowel epenthesis (VIN between) Vowel insertion (VIN after) First cluster doubling Cluster assimilation Cluster deletion

16.61 57.48 43.69 5.50 0.97 -

4 6 5 3 1

Addition of tanwiin Total mean % values of the seven error types in violating cluster

4.53 83.50

2

Table 6.11 Total and % of correct in violating clusters (total stimuli words = 103) Participants

Total Correct

%

1 2 3 4 5 6 7 8 9 10 Mean of % correct responses across the 10 participants

27 5 5 17 10 11 41 47 4 3

26.2 4.9 4.9 16.5 9.7 10.7 39.8 45.6 3.9 2.9 16.50

6.4.1  Sonority conformity clusters In order to detect the most challenging MSA cluster types to the Chinese participants, a comparison between the scores of the different error types of each of the three cluster types is included. With respect to the conforming cluster type, the data displayed in Table 6.6 show the total number of times in which each type of error occurred together with the respective percentages by each participant. From Table 6.6, it is observed that when pronouncing the 73 sonority conforming stimuli words, the participants produced the seven different types of phonological process errors with variation. Thus, only two participants (Participants 7–8) did not produce the error type “cluster substitution,” and only one participant (Participant 3) deleted the cluster three times in the “cluster deletion” context. It is also observed that the Participants 7–8 produced the least number of errors in comparison with other participants. The overlap in the total number of errors can be detected in Table 6.6, as the total number of errors may exceed the total number of the target words (i.e., 73), as in Participants 3, 6, 9, and 10. 125

Total Error Means

14.9

12.50 12.50

3 3

20.83 4.17 -

9 10

5 1

5 6 7

12.50 37.50

-

3 9

3 4

4.17 -

8

1

17 17

7

2 14 4

15 3

4 19

42.5

70.83 70.83

29.17

8.33 58.33 16.67

62.50 12.50

16.67 79.17

%

Total occurrence

Total occurrence

%

Vowel Epenthesis

Cluster Substitution

Phonological Processes/Error Types

1 2

Participants

12 12

4

13 2

13

2 2

Total occurrence

31.3

50.0 50.0

16.67

54.17 8.33 -

54.17 -

8.33 8.33

%

Vowel Insertion

Table 6.12 Totals and % of errors in plateau clusters (total stimuli words = 24)

-

-

-

-

-

Total occurrence

First Cluster Doubling

-

-

-

-

-

-

%

-

-

1 -

-

-

Total occurrence

Cluster Assimilation

4.17

-

-

4.17 -

-

-

%

-

-

-

-

-

-

Total occurrence

-

-

-

-

-

-

%

Cluster Deletion

-

-

-

-

-

-

Total occurrence

Addition of tanwiin

-

-

-

-

-

-

%

L2 acquisition of final consonant clusters Table 6.13 Mean values of each error type across the ten participants Plateau Clusters Error Types

Mean Value of Total % of Errors Across the Ten Participants

Rank

Cluster substitution Vowel epenthesis (VIN between) Vowel insertion (VIN after) First cluster doubling Cluster assimilation Cluster deletion Addition of tanwiin Total mean % values of the seven error types in plateau cluster

14.88 42.50 31.25 4.17 60.42

2 4 3 1 -

Table 6.14 Total and % of correct in plateau clusters (total stimuli words = 24) Participants

Total Correct

%

1 2 3 4 5 6 7 8 9 10 Mean of % correct responses across the 10 participants

17 2 5 13 4 4 15 19 9 7

70.83 8.33 20.83 54.17 16.67 16.67 62.50 79.17 37.50 29.17 39.58

The percentages of occurrence of the three highest error types – “vowel epenthesis,” “vowel insertion,” and “cluster substitution” – across the ten participants is illustrated in Figure 6.1. The lowest whisker of the box plot indicates the lowest percentage of error occurrence, as in Participants 1 and 3 (1.37 %) for “cluster substitution,” Participant 7 (16.4%) for “vowel epenthesis,” and Participants 7–8 (4%) for “vowel insertion.” The highest whisker indicates the highest percentage of error occurrence, as in Participants 9–10 (32.8%) for “cluster substitution,” Participant 4 (69%) for “vowel epenthesis,” and Participant 3 (71%) for “vowel insertion.” Accordingly, the rank of the total average of the occurrence of each phonological process/ error types across the ten participants displayed in the last row is explained in Table 6.7. As shown in Table 6.7, the occurrence of vowel epenthesis errors ranked the highest across all ten participants, with a mean value of 44.23. Vowel insertion ranked the second highest, and cluster substitution ranked the third highest. Cluster assimilation ranked the lowest of error types. As for the participants’ correct responses in pronouncing the 73 target sonority conforming words, these are listed in Table 6.8. As previously mentioned, Participants 7–8 had the lowest number of errors, scoring the highest in terms of correct pronunciation of the target 73 words. 127

Mona Maamoun

Figure 6.1 Occurrence of highest error types by participants in conforming clusters

Participant 3 had the lowest number of correct pronunciations of words, and this is predictable, as he produced a total number of 103 (see Table 6.6), which means that he made more than one error in a single word.

6.4.2  Sonority violating clusters Table 6.9 lists the number of times in which each error type occurs and the percentage of occurrence of each error in each participant. From Table 6.9, it is observed that all participants produced vowel epenthesis and vowel insertion error types in pronouncing the 103 sonority violating target words. Only one participant (Participant 7) out of the ten participants did not produce the cluster substitution error type. None of the participants produced the cluster deletion error in pronouncing the target words. Figure 6.2 illustrates the lowest percentage of error occurrence was by Participants 2 and 8 (2.9%) in cluster substitution, Participant 5 (37.8%) in vowel epenthesis, and Participant 7 (5.8%) in vowel insertion. The highest percentage of error occurrence was that of Participant 6 (33%) in cluster substitution, Participant 2 (89%) in vowel epenthesis, and Participant 5 (69%) in vowel insertion. Thus, similar to the conforming cluster type, the results reveal that the highest percentages of phonological process/error type are those for vowel epenthesis, followed by vowel insertion and then cluster substitution, as shown in Table 6.10. As for the participants’ accurate performance in pronouncing the 103 target sonority violating words, these are provided in Table 6.11. Of note here, and similar to the correct production of conforming sonority words, Participants 7–8 score the highest percentages of correct pronunciation compared with the rest of the participants.

6.4.3  Sonority plateau clusters As for participants’ performance on the sonority plateau words, Table 6.12 displays occurrence of phonological process/error type together with percentages of occurrence for each of the ten participants. As Table 6.12 shows, the participants produced a small number of error types when pronouncing the 24 sonority plateau words (see also Figure 6.3). Participant 7 did not produce any of the error types except for vowel epenthesis, which he produced four times. 128

L2 acquisition of final consonant clusters

Figure 6.2 Occurrence of highest error types by participants in violating clusters

Figure 6.3 Occurrence of highest error types by participants in plateau clusters

Three error types, first cluster doubling, cluster deletion, and addition of tanwiin, were not exhibited in the speech of the Chinese participants while pronouncing the sonority plateau words. Similar to the conforming and violating cluster types, the results of the participants’ performance on sonority plateau cluster type words reveal the same pattern. Among the produced error types, vowel epenthesis ranked the highest, followed by vowel insertion, then cluster substitution. This is further illustrated in Table 6.13 (see also Table 6.14 for the distribution of correct responses by the participants when pronouncing the 24 target sonority plateau words).

6.4.4  Comparison of errors across the three coda clusters The results of correct production across the three types of cluster type words (see Tables 6.8, 6.11, and 6.14) indicate that the participants found the sonority plateau cluster the easiest in pronunciation, followed by the sonority conforming clusters, then the sonority violating ones – with 129

Mona Maamoun

the accuracy percentages of 39%, 33.9%, and 16.5%, respectively. Alternatively, Table 6.15 lists the percentages of errors of the three types of coda cluster words, allowing for a comparison in errors among the three sonority cluster types: conforming, violating, and plateau (see also Figure 6.4). Table 6.15 also lists the results of repeated measures ANOVA tests. It is observed that the Chinese participants produced the highest percentage of error types when pronouncing the sonority violating clusters, with a total mean % value for the seven error types being 83.5%. The second highest errors involved sonority conforming words (67.26%), and the third involved the sonority plateau clusters (60.42%). Repeated measures ANOVA tests revealed, in particular, that the frequency of occurrence of the vowel insertion error type across the three cluster types was statistically significant: F(2, 12) = 8.179, p = 0.004, η2p = 0.539. The value of the overall error occurrence in the sonority violating words was also found to be significantly high compared to its occurrence in the other two types of clusters: F(2, 12) = 12.977, p < 0.001, η2p = 0.590. These results support the hypothesis that there is a relationship between the value sonority distance and ease of pronunciation. As the MSA violating clusters are of negative values, they are hypothesized to be difficult for the Chinesespeaking learners of Arabic to pronounce. This was the case here, as these clusters exhibited the highest percentages of incorrect pronunciation. Table 6.15 Comparison between error types of coda clusters according to error % Error Types

Conforming Clusters

Violating Clusters

Plateau Clusters

F

P

Effect Size (η2)

Cluster substitution Vowel epenthesis Vowel insertion First cluster doubling Cluster assimilation Cluster deletion Addition of tanwiin Overall

13.70 44.25 31.23 2.40 1.92 4.11 5.48 67.26

16.61 57.48 43.69 5.50 0.97 4.53 83.50

14.88 42.50 31.25 4.17 60.42

0.527 2.867 8.179* 12.977*

0.604 0.083 0.004* < 0.001*

0.081 0.242 0.539 0.590

Figure 6.4 Occurrence of the seven error types across three cluster types

130

L2 acquisition of final consonant clusters

Although the identifiable seven error types occur with different percentages across the three types of words, the Chinese participants follow a systematic deviation. For example, in the three cluster type words, the highest percentages of error types involve vowel epenthesis, followed by vowel insertion, then cluster substitution. As for the remaining four error types, they occur with very low non-significant values. For example, first consonant doubling errors occur 24 times overall: seven times in conforming clusters and 17 times in violating clusters, as in, for example, raṭl “a weight measurement” pronounced as raṭṭala and šakl “shape” pronounced šakkala. Addition of tanwiin errors occur in the speech of the participants 23 times (nine times in sonority conforming clusters such as ʕilm pronounced ʕilmun and 14 times in sonority violating clusters) and do not occur in sonority plateau clusters. Cluster assimilation errors occur only nine times: seven instances in sonority conforming clusters, one in a sonority violating cluster, as in raml “sand” pronounced ramr, and one in a sonority plateau cluster. Finally, cluster deletion errors occur three times in sonority conforming clusters. The occurrence of the seven error types across the three cluster types indicates that the pronunciation of an MSA coda cluster is challenging for Chinese-speaking learners of Arabic with varying degrees due to several reasons. First, from a contrastive point of view, Chinese phonology lacks not only coda clusters but also consonant final word position (except for ŋ). This fact forces the L1 Chinese speakers to break the cluster context in Arabic by producing an error, such as vowel insertion and vowel epenthesis. Second, the Chinese phonemic inventory lacks many of the MSA phonemes, such as /r/, /θ/, / ð/, /ṭ/, /ḍ/, /ṣ/, /ð ̣/, /ž/, /q/, /x/, /γ/, /ħ/, and /ʕ/. If one of these phonemes appears in a cluster, a Chinese-speaking learner of Arabic may substitute it by its easier phoneme counterpart, producing the cluster substitution error. Third, the SSP as a universal concept contributes to the difficulty of clusters.

6.5  Conclusion and pedagogical implications Based on the data of the present study, the L1 Chinese learners of L2 Arabic scored the highest percentages of correct responses in clusters with sonority plateaus, followed by clusters that conform to the SSP. Conversely, the highest number of incorrect responses occurred in clusters that violate the SSP. Both the transfer of L1 and universal SSP hypotheses interact to influence the production of Arabic L2 acquisition of clusters by the L1 Chinese-speaking participants. The transfer of L1 was manifested in the form of applying the different types of error patterns across the three types of words to break the cluster. The occurrence of the error pattern “cluster substitution” is another co-evidence of the influence of L1 on L2. The major pronunciation problems which Chinese learners of L2 Arabic do face are those which do not exist in their L1 phonological system. As for the pedagogical implications of the present study, before discussing such implications, highlighting some points concerning the pronunciation class from the experience I personally gained from teaching L2 Arabic (MSA and Egyptian Colloquial Arabic) pronunciation is in order. First, the pronunciation class does show great effectiveness in enhancing L2 learners’ pronunciation and communication when it is based on phonetic and phonological principles, when it deals with the predicted articulation problems (each according to learners’ first language), and when it is managed by an expert phonetician and linguist. The pronunciation class differs from the other language skills classes in many aspects. For example, it is a training class rather than teaching instructions (it trains students to develop a skill until it turns into a habit). We apply different efficient approaches in our pronunciation classes, such as the kinetic approach and touch cues. The instructional materials are different than others, where we use vocal tract diagrams and models, equipment and tools (oral placement tools), 131

Mona Maamoun

pronunciation feedback, and different types of homework. A pronunciation training class is one where little learner exposure to academic concepts and terminology is presupposed. It is not necessary for L2 learners to know, for example, the names of the tongue parts (tip, blade, front, back, root, etc.). The pronunciation curriculum does not have a fixed specific order of lessons; on the contrary, lessons are arranged according to the learners’ pronunciation problems that result from their first language transfer. Training L2 Arabic learners, especially L1 Chinese-speaking learners, to pronounce coda clusters has several considerations. It is my opinion that when the L2 learners are not able to pronounce a consonant in isolation, we cannot target this consonant in a cluster context until he/she has mastered the pronunciation of this consonant in different word positions. Taking /r/ in cluster as an example, the acquisition of the pronunciation of /r/ in cluster could go through the following strategies: • For CC codas that show sonority plateaus, the target clusters should be /CVrr/, then /CVll/. • For CC codas that conform to the SSP, instruction should start with /r/+obstruent, then /l/+obstruent (in both contexts with stops then fricatives, and voiceless segments before the voiced ones). • For CC codas that violate the SSP, the sequence could be as follows: voiceless obstruent+/r/, voiced obstruent+/r/, voiced obstruent+/l/, voiceless obstruent+/l/. • During this stage we always select the other member of the cluster as a sound that is already pronounced correctly by the L2 learner. We avoid contexts where /r/, which is our target, is clustered with another challenging sound which learners have not been trained to pronounce yet. Along with the previous stages, there are three fixed elements in pronunciation training, which include: • Early training and auditory discrimination (listen and only listen) • Use of minimal pairs • Voice recording, self-evaluation, and self-correction.

132

Appendix: Word Stimuli Conforming Clusters /r/ + nasal ْ ُ‫ف‬ ‫رْن‬ ‫فَرْ ْم‬ /r/ + Voiced Fricative ْ ‫ِح‬ ‫فَرْ ْع‬ ‫رْز‬ /r/+ Voiceless Fricative ْ ‫ش‬ ْ‫شَرْح‬ ‫َرْخ‬ /r/+ Voiced Stop ‫شَرْ ْق‬ ْ‫َمرْء‬

ْ‫عَرْش‬

ْ ‫َح‬ ‫رْث‬

ْ‫ِحرْص‬

ْ‫غَرْس‬

ْ ‫ثُ ْل‬ ‫ث‬

ْ ‫َح ْل‬ ‫ف‬

ْ‫َش ْمس‬

ْ ‫ِح ْن‬ ‫ث‬

ْ ‫ُع ْن‬ ‫ف‬

ْ‫ِح ْمض‬

‫َح ْم ْد‬

ْ ‫ص ْم‬ ‫ت‬ َ ْ‫لَهْب‬

ْ‫ع َْذب‬

‫بُ ْع ْد‬

ْ‫ُرعْب‬

‫َر ْش ْق‬

‫ِمس ْْك‬

ْ ‫قَس‬ ‫ْط‬

ْ ‫ِز ْف‬ ‫ت‬

ْ‫ُع ْشب‬

ْ‫َو ْثب‬

‫َرصْ ْد‬

ْ‫َكسْب‬

‫عُرْف‬

ْ ‫ش‬ ‫َرْط‬

Laterals /l/ + nasal ‫قَ ْل ْم‬ ‫ِع ْل ْم‬ /l/ + Voiced Fricative ‫خ َْل ْع‬ /l/ + Voiceless Fricative ْ‫َم ْلح‬ ‫َس ْل ْخ‬ /l/ + Voiced Stop ْ‫ك َْلب‬

ْ‫َح ْلس‬

/l/ + Voiceless Stop ‫غ َْل ْق‬ ‫ُم ْل ْك‬

ْ ‫خ َْل‬ ‫ط‬

Nasal Nasal + Voiced Fricative ‫َم ْن ْع‬ ‫ص ْم ْغ‬ َ Nasal + Voiceless Fricative ْ‫قَ ْمح‬ ْ‫ِر ْمش‬

‫َر ْم ْز‬

Nasal + Voiced Stop ْ‫َج ْنب‬ ‫ِز ْن ْد‬ Nasal + Voiceless Stop ‫َح ْم ْق‬ ‫ُس ْم ْك‬ ْ‫بَهْج‬ ‫َم ْه ْد‬

Voiced Fricative + Voiceless Stop ‫لَ ْع ْق‬ ‫َد ْع ْك‬ Voiced Fricative + Voiceless Fricative ْ ‫ن َْز‬ ْ‫َرعْش‬ ْ‫ن َْزح‬ ‫ف‬ Voiceless Fricative + Voiceless Stop ْ ‫َر ْه‬ ْ ‫ش َْخ‬ ْ ‫ن‬ ‫ط‬ ‫َحْت‬ ‫ط‬ ‫نَ ْش ْىء‬ Voiceless Fricative + Voiced Stop ْ‫َمحْض‬ ْ‫َرحْب‬ ْ‫َشعْب‬ ْ‫ن َْخب‬ ‫َح ْش ْد‬ Voiced Stop + Voiceless Stop ْ ‫ضب‬ ‫ْط‬ ‫طَ ْب ْق‬ َ Violating Clusters Nasal + /r/ ْ‫ُع ْمر‬ ْ‫َخ ْمر‬

‫َحب ْْك‬

ْ ‫َكب‬ ‫ْت‬

(Continued)

Mona Maamoun

Voiced Fricative + /r/ ْ‫ِسعْر‬ ْ‫ِو ْزر‬ Voiceless Fricative + /r/ ْ‫ِسحْر‬ ْ‫ُذ ْخر‬

ْ‫قَصْر‬

ْ‫بِ ْشر‬

Voiced Stop + /r/ ْ‫هَجْر‬ ْ‫صدْر‬ َ

ْ‫َجبْر‬

ْ‫ِحبْر‬

ْ‫بِ ْكر‬

ْ ‫َس‬ ْ‫طر‬

ْ‫َس ْتر‬

ْ‫بَعْل‬

ْ‫َج ْزل‬

ْ‫ن َْذل‬

ْ‫َسحْل‬

ْ‫َمصْل‬

ْ‫َغسْل‬

ْ‫ِط ْفل‬

ْ ‫َر‬ ْ‫طل‬

ْ‫قَ ْتل‬

‫َد ْع ْم‬

ْ ‫ع‬ ‫َظم‬

‫خ َْز ْن‬

‫َح ْز ْم‬

ْ ‫ُغ‬ ‫صْن‬ ْ َ‫ل‬ ‫حْن‬

‫ُحس ْْن‬ ‫لَحْ ْم‬

‫َر ْس ْم‬ ‫س ُْخ ْن‬

‫ِج ْف ْن‬

‫اِب ْْن‬

‫تُب ْْن‬

Voiceless Stop + Nasal ‫َذ ْق ْن‬ ‫ُع ْق ْم‬

‫ُح ْك ْم‬

ْ ‫َر‬ ‫ط ْن‬

ْ َ‫ل‬ ‫ط ْم‬

‫َش ْت ْم‬

Voiceless Stop + Voiced Stop ْ ‫ع‬ ْ‫َطب‬ ‫َع ْق ْد‬

ْ‫َر ْكض‬

ْ‫َر ْكض‬ ‫َر ْد ْع‬

‫َد ْب ْغ‬

‫ُخب ْْز‬

‫نَب ْْذ‬

ْ‫قُدْس‬

‫طَب ْْخ‬

ْ‫نَبْش‬

ْ‫َحبْس‬

ْ‫قُدْس‬

‫طَب ْْخ‬

ْ‫نَبْش‬

ْ‫َحبْس‬

ْ َ‫ق‬ ‫ط ْع‬

‫قَ ْف ْز‬

‫نَ ْت ْع‬

ْ‫َع ْكس‬

ْ ‫غ‬ ْ‫َطس‬

ْ ‫ع‬ ْ ‫َط‬ ‫ف‬

‫ُر ْس ْغ‬

‫لَ ْس ْع‬

‫َر ْف ْع‬

Voiceless Stop + /r/ ْ‫بِ ْئر‬ ْ‫فَ ْقر‬ Nasal + /l/ ْ‫َر ْمل‬ Voiced Fricative + /l/ ْ‫َسهْل‬ ْ‫بَ ْغل‬ Voiceless Fricative + /l/ ْ‫أَهْل‬ ْ‫د َْخل‬ Voiced Stop + /l/ ْ‫َعدْل‬ ْ‫َحبْل‬ Voiceless Stop + /l/ ْ‫َع ْقل‬ ْ‫َش ْكل‬ Voiced Fricative + Nasal ‫َر ْغ ْم‬ ‫لَع ْْن‬ Voiceless Fricative + Nasal ‫فَ ْخ ْم‬ ‫َو ْش ْم‬ ‫َد ْه ْن‬ ‫َو ْه ْم‬ Voiced Stop + Nasal ‫َع ْد ْن‬ ‫هَ ْد ْم‬

ْ‫ع ُْذر‬

Voiced Stop + Voiced Fricative ‫َمضْ ْغ‬ ‫َوضْ ْع‬ ‫لَ ْد ْغ‬ ْ ‫َر ْد‬ ْ‫نَضْح‬ ْ‫قَدْح‬ ‫ف‬ Voiced Stop + Voiceless Fricative ْ ‫َر ْد‬ ْ‫نَضْح‬ ْ‫قَدْح‬ ‫ف‬ Voiceless Stop + Voiced Fricative ‫َو ْق ْع‬ ‫ِح ْق ْد‬ ‫َر ْك ْع‬ Voiceless Stop + Voiceless Fricative ْ ‫َس ْق‬ ْ‫َر ْقص‬ ْ‫نَ ْكش‬ ‫ف‬ Voiceless Fricative + Voiced Fricative ْ َ‫ل‬ ‫أَ ْخ ْذ‬ ‫حْظ‬ ‫َرصْ ْع‬ Plateaus Nasal ‫َغ ْن ْم‬ ‫َس ْم ْن‬ ‫ض ْم ْن‬ ِ Voiced Fricative ْ ‫َو ْع‬ ‫ظ‬ ‫ن َْز ْع‬

134

ْ‫ِجسْر‬

ْ‫َس ْفر‬

ْ‫َم ْثل‬

‫اِ ْث ْم‬

ْ ‫ِح ْف‬ ‫ظ‬

L2 acquisition of final consonant clusters

Voiceless Fricative ْ‫فَسْح‬ ‫فَس ْْخ‬ ْ ‫ز‬ ْ‫لَحْس‬ ‫َحْف‬ ْ‫َدهْس‬ ْ‫نَهْش‬

ْ ‫َخس‬ ‫ْف‬ ْ ‫ع‬ ‫َصْف‬ ْ‫فَحْش‬

ْ‫طَ ْفح‬ ْ ‫َس ْخ‬ ‫ف‬ ْ‫فَحْص‬

ْ‫نَ ْفش‬ ْ‫ش َْخص‬

ْ‫نَ ْفس‬ ْ‫بَ ْخس‬

ْ ‫نَ ْف‬ ‫ث‬ ْ ‫ع‬ ‫َصْف‬

References Ahmad, J., and Nazim, M.,2014. Saudi EFL learners’ views on English pronunciation. Journal of Humanities and Social Studies, 2, 355–366. Al Ani, S., 1978. Readings in Arabic linguistics: Phonetics and phonology. Bloomington, IN: Indiana University Press. Al Tamimi, Y., and Al Shboul, Y., 2013. Is the phonotactics of the Arabic complex coda sonority-based? Journal of King Saud University. Language and Translation, 25, 21–33. Al Ghamdi, M., 2001. Al-Ṣawtiyyāt Al-‘Arabiyya.Riyadh, Saudi Arabia: Maktabat Al-Tawbah. Asfoor, M. A.,1982. Difficulties English speakers encounter in Arabic phonology. Thesis (PhD). San Francisco University. Boudaoud, M., and Cardoso, W., 2009. The variable development of /s/ + consonant onset clusters in Farsi-English interlanguage. In : The 10th Generative Approaches to Second Language Acquisition Conference. Concordia University. Broselow, E. and Finer, D., 1991. Parameter setting in second language phonology and syntax. Second language research, 7, 35–59. Cardoso, W., 2008. The Development of sC onset clusters in interlanguage: Markedness vs. frequency effects. In: Roumyana Slabakova et al. eds. The generative approaches to second language acquisition. Somerville, MA: Cascadilla Press, 15–29. Clements., G., 1990. The role of the sonority cycle in core syllabification. In: J. Kingston and M. Beckman, eds. Papers in laboratory phonology. Cambridge: Cambridge University Press, 283–333. Corder, S. P.,1967. The significance of learners’ errors. International Review of Applied Linguistics, 5 (4), 161–170. Cowan, Milton J., 1994. The Hans Wehr dictionary of modern written Arabic. 4th ed. Beirut, Lebanon: Librairie du Liban. DeKeyser, R. M., 2013. Age effects in second language learning: Stepping stones toward better understanding. Language Learning, 63, 52–67. Derwing, T. M. and Munro, M. J., 2005. Second language accent and pronunciation teaching: a researchbased approach. TESOL Quarterly, 39 (3), 379–397. Duanmu, San. 2006. Chinese (Mandarin): phonology. In: Keith Brown, ed. Encyclopedia of language and linguistics. 2nd Ed. Oxford, UK: Elsevier, 351–355. Eckman, F. and Iverson, G., 1993. Sonority and markedness among onset clusters in the interlanguage of ESL learners. Second Language Research, 9 (3), 234–252. Enochson, K., 2014. L2 production of English onset sC and CC clusters. The international symposium on the acquisition of second language. Concordia working papers in applied linguistics, 5, 171–184. Fatemi, M., 2012. Difficulties of Persian learners of English in pronouncing some English consonant clusters. World Journal of English Language, 2 (4), 69. Giegerich, H., 1992. English phonology: An introduction. Cambridge: Cambridge University Press. Gilbert, J. B., 1995. Pronunciation practice as an aid to listening comprehension. In: D. J. Mendelsohn and J. Rubin, eds. A guide for the teaching of second language listening. San Diego, CA: Dominie Press, 97–112 Hogg,R. and McCully,C. 1987. Metrical phonology. Cambridge: Cambridge University Press. Hyltenstam, K. and Abrahamsson, N., 2003. Maturational constraints in second language acquisition. In: C. J. Doughty and M. H. Long, eds. The handbook of second language acquisition. Oxford, UK: Blackwell, 539–588. Jarrah, A., 2012. Phonology in language learning and teaching. Indian Journal of Applied Linguistics. 38(2), 107–120. Jarvis, S. and Pavlenko, A., 2007. Cross linguistic influence in language and cognition. New York: Routledge.

135

Mona Maamoun Johnson, S. and Newport, E., 1989. Critical period effects in second language learning: The influence of maturational state on the acquisition of English as a second language. Journal of Cognitive Psychology, 21, 60–99. Kenstowicz, M., 1994. Phonology in generative grammar. Oxford: Blackwell. Kopczynski, A. and Meliani, R., 1993.The vowels of Arabic and English. Papers and Studies in Contrastive Linguistics. Adam Mickicwicz University. Poznan. XXVII, 193–203. Lowie, W., 2015. L2 phonology. In: P. Andreson, ed. The Routledge Encyclopedia of second language acquisition. New York: Routledge, 348–353. Morley, J., 1991. The pronunciation component in teaching English to speakers of other languages. TESOL Quarterly, 25 (3), 481–520. Nahla, M., 2006. Teaching Arabic to non-native speakers at Alexandria University. In: K. M. Wahba, et al., eds. Handbook of Arabic language teaching professional in the 21st century. New York: Routledge, 67–80. Nguyen, A., and Brouha, C., 1998. The production of word-final consonants in English by L1 speakers of Vietnamese. Working papers in linguistics. vol 5. 73-94. George Mason university, Virginia. Rasier, L. and Hiligsmann, P., 2007. Prosodic transfer from L1 to L2: Theoretical and methodological issues. Nouveaux cahiers de linguistique française, 28, 41–66.Shehata, A., 2015. Problematic Arabic consonants for native English speakers: Learners' perspectives. International Journal of Educational Investigations, 9, 24–47. Watson, J. 2002. The phonology and morphology of Arabic. Oxford: Oxford University Press. Yavas, M. and Someillan, M., 2005. Patterns of acquisition of /s/ clusters. Multilingual Communication Disorders, 3, 50–55. Zhanming, W., 2014. Review of the influence of L1 in L2 acquisition. Studies in Literature and Language, 9 (2), 57–60.

136

PART II

Arabic L2 vocabulary

7 LOOKING AT WORDS An eye-tracking investigation of L2 Arabic vocabulary learning Ayman A. Mohamed

Paper-and-pencil studies have provided ample evidence about the benefit of repeated exposure on the amount and quality of vocabulary learning (Horst 2005; Schmitt 2010). On the other hand, eye-movement investigations on reading behavior documented cognitive effects of repetition on lexical processing and associated vocabulary learning with processing patterns in the light of the eye-mind link hypothesis (Rayner 1998; 2009). This chapter brings the two strands together by testing the hypothesis that the amount of attention to novel words over repeated encounters can predict readers’ incidental acquisition of multiple aspects of vocabulary knowledge. Thirty English-speaking learners of Arabic read 100 medium-length sentences with 12 novel (pseudo) words arranged in three exposure conditions: four, seven, or ten encounters, while their eye movements were recorded. Immediately after the reading session, participants took unannounced tests on form recognition, meaning recognition, and meaning recall. Eye-movement results showed that learners fixated more on initial encounters with target words and that their fixation times gradually decreased from first to last exposure. The longer they looked at novel words, the more learning gains they reported, particularly in meaning recognition and recall of these words. First fixation duration predicted form recognition, while total reading time was a significant predictor of all learning outcomes. A GLMM regression model showed that the amount of online processing was strongly associated with certain vocabulary outcomes after controlling for the number of exposures. Results from paper-andpencil vocabulary knowledge tests confirmed a significant effect of exposure condition, particularly on form recognition and meaning recall. Learning gains were the highest in form recognition and lowest in meaning recall, which points to a potential cognitive trajectory of incidental lexical development. The study provides practical implications and further directions for investigating the cognitive aspects of reading comprehension and incidental vocabulary acquisition in L2 Arabic.

7.1 Introduction In second language vocabulary research, there is a widely held view that lexical development can occur incidentally, particularly from reading (Fraser 1999; Kowen and Kim 2008; Matsuoka and Hirsh 2010; Nation 2001; Paribakht and Wesche 1997; Pellicer-Sánchez and Schmidt 2010; Pulido 2007; Rott 1999; Watanabe 1997; Webb 2008). One factor that has 139

Ayman A. Mohamed

received considerable attention in reading studies is repeated exposure, which reportedly accounts for substantial differences in learning gains on different aspects of word knowledge (Horst et al. 1998; Pigada and Schmitt 2006; Rott 1999; Waring and Takaki 2003; Webb 2007). Lexical gains have also been found to depend on the cognitive load of reading tasks (Hulstijn and Laufer 2001; Kim 2008; Keating 2008). However, these studies were all paper-and-pencil ones with learning measured post hoc, through testing. They captured the offline result of lexical learning, with no insights into the online cognitive processes of individuals while exposed to vocabulary or engaged in instructional tasks. Cognitive approaches to vocabulary learning basically relied on Schmidt’s (1990) noticing hypothesis, in which learners are expected to notice the target linguistic items while focusing on meaning in reading tasks (see Laufer 2005), and where the amount of attention directed to a specific lexical item would facilitate subsequent initial intake of that item after exposure (Godfroid et al. 2013). A major challenge, however, was how to capture these intricate cognitive processes as they occur in real time. Studies that used think-aloud protocols or interviews might have been the first to probe into the cognitive processes underlying lexical acquisition (e.g. Paribakht and Wesche 1997; Fraser 1999). With the recent advent of eye-tracking technology, psychology-based research has set the scene for more sophisticated language acquisition inquiries. Based on the ‘eye-mind link’ assumption that learners’ eye movements accurately reflect real-time processing (Rayner 2009), Godfroid et al. (2010, 2013) used eye tracking as an innovative technique to measure the amount of attention L2 readers paid to novel vocabulary in reading tasks. The authors found evidence of an association between attention and word recognition. More recently, Pellicer-Sánchez (2016) confirmed a relationship between online reading and vocabulary outcomes, advocating the case that learning can be tracked while reading. The initial hypothesis that motivates the current study is that frequent exposure to novel lexical items invites more attention to form and meaning, which can be reflected in more time for processing and more opportunities for incidental intake and retention. In the present research, I implement the eye-tracking methodology to investigate both online and offline aspects of incidental vocabulary learning from reading Arabic-like pseudo words embedded into short sentences. The main goal of the study is to track the cognitive effect of repeated exposure to vocabulary on the patterns of reading exhibited by English-speaking learners of Arabic, and whether the learners’ eye-movement reading measures reflect the development of different components of their vocabulary knowledge, including form and meaning recognition and meaning recall.

7.1.1  Review of literature Reviews of vocabulary studies usually indicate that incidental vocabulary learning is often slower than intentional learning (Horst 2005; Hulstijn 2001; Laufer 2005; Macaro 2003; Read 2004; Schmitt 2008). However, researchers and teachers tend to agree that both modes complement each other. This is because it is impossible to cover all vocabulary needs in the classroom context and that creating opportunities for incidental exposure can positively support partial lexical knowledge and aid explicit teaching. Previous studies on incidental learning from reading focused on how to promote engagement in reading tasks either by manipulating word presentation in text or administering different tasks with varying degrees of complexity (e.g., Hulstijn1992; Hulstijn et al.1996; Hulstijin and Trompeter 1998; Paribakht and Wesche 1997; Peters et al. 2009; Watanabe 1997). The results in these studies were mostly in favor of inducing more cognitive effort while learning new lexical items. Building on Craik and 140

Looking at words

Lockhart’s (1972) depth of processing hypothesis, Laufer and Hulstijn (2001) introduced the involvement load hypothesis to account for the pattern of results observed in previous literature. The hypothesis was based on an analysis of the cognitive and motivational involvement imposed by any given L2-vocabulary task. Involvement, a cognitive-motivational construct, was defined as the combined effects of need, search, and evaluation. Tasks that induce higher involvement were hypothesized to produce higher vocabulary gains. The hypothesis received empirical support from several studies (e.g., Keating 2008; Kim 2008; Mohamed 2016). In the same line, Schmitt (2008) referred to engagement with vocabulary as a key factor in vocabulary learning. Engagement, in his view, can be fostered by several other interventions beyond what the involvement hypothesis claimed, including, but not confined to, frequency of exposure, increased attention to target words, and increased time spent on the target items.

7.1.2  Exposure frequency in vocabulary learning Several studies demonstrated the effect of repeated exposure on vocabulary learning. Rott (1999) found that six exposures led to better word retention than two or four exposures in reading comprehension. Horst et al. (1998) found that words that appeared eight or more times in a long novel had a better chance of being learned by most of the participants than words that appeared less often. Waring and Takaki (2003) concluded that about eight exposures to target words in extensive reading led to a 50% chance of participants’ recognizing form and meaning in a delayed multiple-choice test three months later, while less than five repetitions was not enough to translate a word correctly after three months. Repetition was also found to play different roles in the development of other aspects of word knowledge besides word meaning. Webb (2007) investigated multiple components of word knowledge including spelling, associations, syntax, grammatical functions, and formmeaning mapping. He found that the group that encountered the target words more than ten times showed a better grasp of different aspects of word knowledge than other groups who received fewer exposures. Similarly, studies on extensive reading found that reliable learning of vocabulary started after ten exposures (Pellicer-Sánchez and Schmidt 2010; Pigada and Schmitt 2006). Schmitt (2008) thus concluded from previous research that an average of eight to ten repetitions could be appropriate for the development of receptive knowledge of vocabulary while leading only to relatively low gains in productive knowledge.

7.1.3  Exposure frequency and attention The present study is motivated by the fact that many factors that Schmitt (2008) identified as determinants of engagement (frequency of exposure, increased attention to target words, and increased time spent on the target items) are interrelated. This hypothesized interrelatedness was not directly addressed in earlier studies on incidental learning. As shown earlier, there were studies that investigated engagement as operationalized through the involvement load hypothesis and other studies that provided evidence for the positive effect of exposure per se. Because the involvement load can be seen as a manifestation of attention, a valid comparison between repetition and attention should rather attempt to disentangle their effects empirically and account for their interrelatedness at the same time. In terms of measuring attention in real-time processing, Godfroid et al. (2013) reviewed several methods that have been used in psychology studies including note taking, underlining, think-aloud protocols, and eye tracking. They concluded that, with all the precision and completeness it has to offer, it can provide a more sensitive measure of the amount and locus of 141

Ayman A. Mohamed

attention to linguistic input. Reviews of eye-tracking research have indicated that eye movements reflect an accurate representation of the cognitive processes in the reader’s mind. This assumption was coined the ‘eye-mind’ link hypothesis, which proposes a connection between overt and covert attention (Rayner 1998, 2009). While recording eye-movement patterns in natural reading, researchers were able to manipulate text features and word properties, such as frequency, predictability, familiarity, and other context variables in order to examine their effects on reading behavior.

7.1.4  Eye tracking in vocabulary studies Eye tracking is defined as the online recording of learners’ eye-movement behavior, which is described in terms of fixation times (how long readers look at something) and saccades (the movement of the eyes from one point to the next) (Godfroid, 2012). Eye-movement measures tend to be categorized as either early measures of processing (e.g., first fixation and first pass time) or late processing measures (e.g., second pass time and total time). First fixation duration captures the time of the first look at the target area (for example, a novel vocabulary word) when encountered for the first time during forward reading. First pass time, which is also termed gaze duration, combines first fixation duration along with any other fixation made on the target area at the initial visit before the eyes move forward or backward to the next target area. Second pass time captures the rereading episodes when the eyes reenter an interest area after they have left it. Total reading time is the sum of all fixations on the target area (see Winke et al., 2013). Eye-movement research on reading has pointed to several factors that could affect eyefixation times, including word frequency, context predictability, word length (Kliegl et al. 2006), part of speech, familiarity, and concreteness (Liversedge and Findlay 2000; Rayner 1998, 2009; Starr and Rayner 2001). Chaffin et al. (2001) reported that the familiarity of target words and context quality determined the amount of time readers spent on the targets. Raney and Rayner (1995) investigated the effects of word frequency on native English speakers’ rereading performance (reading the same text two times). They found that individuals had shorter reading times, made fewer fixations, and had longer saccades during the second reading. Moreover, shorter fixation durations were associated with high frequency words, suggesting independent effects of frequency and repetition on reading times. Hyönä and Niemi (1990) tracked Finnish readers’ performance on three text repetitions. Readers’ fixation times decreased consistently from first to third encounter with target sentences; their progressive fixations and number of regressions also decreased. Raney et al. (1995) found similar results regarding the effect of three repetitions of lexical items in a given text, and they also found frequency effects on the first two repetitions, but no further differences occurred after that. With regard to the association between online processing and learning, Williams and Morris (2004) examined the effect of word familiarity in reading comprehension and word recognition. They found that readers spent more processing time on novel words than familiar words and that there was a systematic relationship between online processing patterns, i.e., reading times, and retention of new word meanings. Brusnighan and Folk (2012) conducted a selfpaced, L1-reading study on incidental vocabulary learning. They found that readers spent more time processing sentences that contained novel compound words and that they were able to retain new word meanings from a single exposure. Recent studies on incidental vocabulary learning included Godfroid et al. (2013), who operationalized attention to novel pseudo words as a quantitative variable reflected in the participants’ eye-fixation times during reading. Twenty-eight advanced EFL learners read 12 142

Looking at words

paragraphs in English with target areas that consisted of known words, pseudo words, or a combination of both. Results showed that readers fixated longer on pseudo words than on known words, regardless of whether these pseudo words were combined with appositive cues. There was a strong association between the total fixation time on pseudo words and subsequent recognition of these words in a surprise posttest. Pellicer-Sánchez (2016) designed a story reading experiment with exposures to target words up to eight times. She found that repetition increased reading fluency and that reading times were associated with vocabulary gains of form and meaning. With such a study design, the author advocated the fact that eyemovement research has allowed us to confirm acquisition can track word learning during reading as well as after reading.

7.2  Research questions In the current study, I investigate the process of incidental vocabulary learning from reading and question the relative speed with which the different aspects of vocabulary knowledge (form recognition, meaning recognition, and meaning recall) accrue in this online process using eye-movement measures. One unique aspect of the study is dealing with a population less commonly represented in literature, i.e., English-speaking learners of Arabic. This may entail certain practical challenges in an eye-movement investigation, not in the least because of differences in script and the direction of reading. Rayner (1998, 2009) reviewed issues in eye tracking and maintains that different text direction entails differences in the perceptual span which is, in the case of Arabic text, asymmetric and larger to the left of fixation (Jordan et al. 2014). On the readability and accuracy of reading, Al-Wabil and George (2010) reported that L1 readers of Arabic would be more accurate with 14-points and 16-point of Simplified fixed Arabic rather than Traditional Arabic font type. In general, no previous empirical studies, to the best of my knowledge, have been conducted on Arabic foreign language reading using the eye-tracking technique. Experimenting with Arabic text using the eye-tracking method is likely to inform further research on the perceived challenges and implications for the learners’ online interaction with a non-Roman script during reading. In the light of the previous discussion, the present research is set to answer the following questions: 1 How do English-speaking learners of Arabic process novel lexical items in Arabic written input? And how do their eye movements change from the first encounter to subsequent encounters with new vocabulary? 2 Do L2 Arabic learners’ eye-fixation durations on novel words predict their learning gains in the vocabulary knowledge posttests? 3 What is the effect of repeated exposure to novel target words in L2 Arabic text on the incidental acquisition of receptive and productive knowledge of form and meaning of these words in vocabulary posttests?

7.3 Method 7.3.1 Participants Thirty English-speaking learners of Arabic participated in the study. Fourteen (nine females and five males) were enrolled in second-year Arabic, and 16 (13 females and three males) were in third-year Arabic. Arabic course offerings in their program consisted (per semester) of five credit hours for the first and second year and four credit hours for their third year. Their ages 143

Ayman A. Mohamed

ranged from 18 to 26, with a mean age of 21 years. Their level of proficiency in Arabic ranged between intermediate-high and advanced, as defined by the ACTFL oral proficiency guidelines. This information was obtained from recent OPI exam results, if available, or their teachers’ evaluations of their current proficiency levels.

7.3.2 Materials 7.3.2.1  Target words To control for prior word knowledge, I created a list of 12 pseudo words resembling real Arabic words for the study. The decision to use pseudo words rather than low frequency words was made to exclude any possibility that readers would have encountered the target items before or even had some partial knowledge of them. This approach has been followed in psycholinguistic studies, especially in the absence of a pre-test. In the present study, it was not practical to conduct a pre-test for the target words, because this would have brought learners’ attention to the goal of the study, which would contradict the assumption of incidental learning. The list consisted of six nouns, three verbs, and three adjectives. Word length ranged from four to six letters and followed the patterns and morphological rules of actual Modern Standard Arabic words. The pseudo words basically denoted concrete and familiar concepts that lead themselves to being inferred from sentence context. No cognates or pseudo homophones with English were used. To control for the effects of intrinsic word properties, the 12 target words were organized into three counterbalanced word lists. In each word list, four words appeared four times, four words appeared seven times, and the remaining four words appeared ten times in the reading task. To minimize the effect of visual factors on eye fixations, word lists were matched for the number of letters and part of speech within and across exposure conditions. Table 7.1 includes the list of pseudo words, given in transliteration, and their respective meanings in the reading context.

7.3.2.2  Reading material The target sentences were adapted mainly from the textbooks that students use in the second and third years. They consisted of 100 sentences followed by a yes/no question to check for comprehension and to direct learners to focus on meaning. The reading set included six practice sentences, 84 critical sentences and 10 fillers. The texts were randomized for all participants so that encounters with a given target word were distributed randomly across the reading set. Each sentence ranged in length between 15 and 24 words (for a sample of online reading material, see the Appendix). I created three counterbalanced versions of the reading material Table 7.1 Target pseudo words (in transliteration) for the reading task Target Word

Meaning

Target Word

Meaning

qaṣaba šinqiiṭ karduus ʔiftakasa ẓaʕana ʔistabṭaʕa

“city” “poor” “castle” “to wear” “to travel” “to buy”

saxiim buzuuq baaquur guʕrana γashiim mafsaħ

“sad” “money” “hospital” “job” “new” “market”

144

Looking at words

so that each set of four words occurred with a different frequency across conditions: either four, seven or ten exposures.

7.3.3 Procedures 7.3.3.1  The reading session Eye movements during reading were recorded using a desk-mounted EyeLink 1000, an eyetracker manufactured by SR Research (www.sr-research.com). The texts were presented in Simplified Arabic font, size 18, on a 19-inch computer monitor set up 55 cm from the participants’ eyes. Participants placed their heads on a chin and forehead rest during the experiment to minimize head movements. The experiment included two breaks, one after a block of 40 sentences and another after sentence 70. Eye calibration was performed at the beginning of the experiment and twice after the breaks. Participants advanced from one screen to another using a button on the right side of a hand-held controller and answered questions by pressing an upper left button to respond with ‘yes’ and an upper right button to respond with ‘no.’ Drift correction was performed before each sentence. Because Arabic text goes from right to left, the experiment options were adjusted to assign interest areas from right to left and to define eyemovement measures accordingly. The drift correction was also placed at the right of the screen. Participants were instructed to read for meaning and answer simple yes/no questions. The comprehension questions did not include the target word, and readers did not have to know the meaning of the word in order to provide the right response. For example in a sentence that read “Maha went to the shop and bought some meat and vegetables,” the question was “Did Maha buy fruit?” The purpose of this design is to have learners focus on meaning without bringing their attention to the component of new vocabulary in the research.

7.3.3.2 Testing The testing session started immediately after the eye-tracking experiment. Participants were not told that they would be tested on vocabulary. To obtain a multi-faceted picture of lexical knowledge development, it was important to include multiple measures of vocabulary knowledge. All the vocabulary tests consisted of the target 12 items of the study in addition to distractor items. Details of the tests are given below in the order they were conducted. Tests were specifically given in this order to avoid any testing-transfer effects from one vocabulary measure to the other: 7.3.3.2.1  FORM RECOGNITION TEST

Participants were given 12 sets of words, each containing four non-words, one of which was the target word. The task was to identify the target lexical item that occurred in the reading material among three other distractors/fillers. Students were instructed to circle only the word they recognized in each set. A participant received one point for a correct choice and zero for an incorrect response. 7.3.3.2.2  MEANING RECALL TEST

A simple Arabic-English translation test was designed in which participants were required to either provide a meaning or an explanation for the target word or just provide any clue that 145

Ayman A. Mohamed

they thought could be relevant to its meaning. Participants received one point for a correct meaning, close synonym, or explanation and zero for a blank or irrelevant response. 7.3.3.2.3  MEANING RECOGNITION TEST

To test this ability, a multiple-choice task was designed in which each target item was given along with four possible meaning options in English. The participants were required to circle only one meaning. Participants received one point for the correct choice and zero for an incorrect response.

7.3.4 Analyses Eye-movement analyses of the reading data included reports on first fixation durations, gaze durations, and total reading times of the target words on all exposures. I compared these three measures for all participants to capture the pattern of looking at a given word from the first to the last encounter. I compiled a list of fixation times by participant to each of the target words on every exposure. Second, I averaged fixation times across participants, which provided a record of the reading times spent on each word at every encounter. Then data were averaged over exposures, providing fixation times for each target encounter regardless of item. Mean eye-movement times were then compared through line graphs to detect significant rises and declines in reading times. To associate the online data with vocabulary learning measures, reading measures (first fixations, gaze durations, total times) were summed on all exposures for each target word per participant. In this way, the records would reflect accumulated episodes of attention for specific target items and relate these to either failure or success in retrieving a given word in the different vocabulary tests. Because the data consisted of multiple observations per participant, I used a generalized linear mixed model (GLMM) to fit a repeated measures logistic regression, which is appropriate because the vocabulary post-tests represent binary dependent variables. The purpose of the analysis is to test if the fixation durations and/or exposure condition predict the probability of form and meaning recognition on the post-test. The GLMM output quantifies probability in terms of an odds ratio (OR) which represents the predicted change in the dependent measure as a function of a one unit change in a given predictor. Finally, for the offline data, performance on the four vocabulary tests was compared holistically between tests and by exposure condition to explore the typical trajectory of incidental lexical knowledge acquisition.

7.4 Results 7.4.1  Online reading measures 7.4.1.1  First fixation durations (FFD) Mean first fixation times were tracked across exposures and averaged over participants and vocabulary items. Fixation durations started at an average of 403 ms (SD = 19) for first encounter and ended with an average of 375 ms (SD = 26) by the last exposure. The line graph created for first fixation durations (FFD) in Figure 7.1 shows slight increases in FFD values on initial exposures, with a small rise at exposure 6 before it declines to reach its lowest by exposure 10. To test the role of repeated encounters on the rise or decline of FFD, a GLMM was conducted to fit a logistic gamma regression with exposure as a fixed effect and FFD as a dependent scale variable. Regression output showed that exposure was not significant (OR = 1.00, 146

Looking at words

500 450 400

Mean FFD

350 300 250 200 150 100 50 0 1.00

2.00

3.00

4.00

5.00

6.00

7.00

8.00

9.00

10.00

exposure Figure 7.1 Mean first fixation durations (in milliseconds) across encounters with target words

95% CI = [0.991, 1.009], p = .956), meaning that first fixations might have decreased over encounters, but the decrease was not that significant. 7.4.1.1.2  GAZE DURATIONS (GD)

Gaze durations (GD) started at an average of 2148 ms (SD = 495) for first encounter and ended with an average of 1164 ms (SD = 489) by the last exposure. Unlike the FFD graph, the gaze duration decreasing pattern was more visible (see Figure 7.2). Visual inspection of the graph shows that GD plateaued from the fifth to the eighth exposure before declining again until the final target encounter. Regression output showed that exposure was a significant factor (OR = 0.98, 95% CI = [0.97, 0.99], p < .001), implying that each additional encounter with the target words would likely decrease the amount of gaze duration by around 2 %. 7.4.1.1.3  TOTAL READING TIME

The total reading times followed a gradually decreasing pattern, similar to gaze duration, with the longest time spent on the target word at first exposure and the shortest time spent at the last exposure (see Figure 7.3). Reading times started from an average of 3340 ms (SD = 958) and ended with an average of 1612 ms (SD = 545). Visually, a plateau in total times can be observed from the sixth to the ninth exposure before declining at the last target encounter. Regression output showed that exposure was a significant factor (OR = 0.96, 95% CI = [0.95, 0.97], p < .001), implying that each additional encounter with the target words would likely decrease the total amount of reading time by around 4 %. 147

Mean FFD

2100 2000 1900 1800 1700 1600 1500 1400 1300 1200 1100 1000 900 800 700 600 500 400 300 200 100 0

1.00

2.00

3.00

4.00

5.00

6.00

7.00

8.00

9.00

10.00

exposure Figure 7.2 Mean gaze durations (in milliseconds) on encounters with target words 3000

2500

Mean TFD

2000

1500

1000

500

0 1.00

2.00

3.00

4.00

5.00

6.00

7.00

8.00

9.00

10.00

exposure Figure 7.3 Mean total time (in milliseconds) on encounters with target words

Looking at words

7.4.2  Vocabulary learning measures To analyze the results of the paper-and-pencil tests, I looked at the chances of acquisition after a given number of exposures; i.e., what percentage of words individual readers could learn from having seen the word four, seven, or ten times. Individuals considerably varied in their word gains, particularly in meaning recall, as some could not recall any words and others were able to recall accurate meanings of almost up to 90% of the target words. An inspection of the percentage of learning gains for the different paper-and-pencil vocabulary tests showed consistently increased chances of learning as target word encounters increased (see Table 7.2). For example, in meaning recall there was a 9% chance of correct recall when the word was repeated four times, and this chance increased to 25% after the seventh exposure and reached 33% when the target items were seen ten times. On the other hand, learners could correctly recognize the form of words 74% of the time after they had encountered them 10 times in reading. Table 7.2 summarizes the percentages of word gains over exposure conditions and how they differed across different vocabulary tests. From Table 7.2, it is clear that learning gains developed at a high level in form and meaning recognition but to a smaller extent in meaning recall, which means that receptive knowledge, unsurprisingly, preceded productive knowledge development. Some large standard deviations reflect individual variation in vocabulary acquisition. The retrieval success ranged from a minimum of 0% to a maximum 50% for four exposures in all measures except form recognition, where some participants retrieved all word forms after four exposures. For seven and ten exposures, retrieval ranged between 0% and 100% in all measures.

7.4.3  Relationship between online reading and vocabulary learning The major question in regard to the eye-movement data was to see if the probability of word learning would increase as learners spent more total time processing (looking at) the words and whether the predictive power of online reading times would be upheld after factoring in exposures. To this end, I ran a GLMM logistic regression for each of the vocabulary tests as outcome variables with subjects as a random factor, target items as a repeated within-subject variable, and frequency of exposure as a fixed factor. Each reading measure was entered separately as a covariate to compare its effect against exposure frequency. In the light of model fit analysis, online reading measures showed mild positive effects to variable extents. First fixation duration significantly predicted form recognition scores, (OR = 1.45, 95% CI = [1.22, 1.48], p < .001). Gaze duration predicted meaning recognition (OR = 1.72, 95% CI = [1.53, 1.67], p < .001), and meaning recall (OR = 1.04, 95% CI = [1.01, 1.06], p = .041). The total time predicted all vocabulary outcomes: form recognition (OR = 1.42, 95% CI = [1.20, 1.77], p = .004), meaning recognition (OR = 1.58, 95% CI = [1.43, 1.66], p < .001), and meaning recall (OR = 1.24, 95% CI = [1.15, 1.28], p = .046). Table 7.2 Percentages of word gains with standard deviations by exposure condition

4 Exposures 7 Exposures 10 Exposures Average by Test

Form Recognition

Meaning Recognition

Meaning Recall

50 (27) 69 (28) 74 (26) 64

41 (36) 52 (27) 67 (26) 53

9 (13) 25 (21) 33 (23) 22

149

Ayman A. Mohamed Table 7.3 Vocabulary knowledge online and offline predictors

Total Exposure First Fixation Gaze Duration Total Time

Form Recognition OR (p)

Meaning Recognition OR (p)

Meaning Recall OR (p)

1.13 (p = .005) 1.45 (p < .001) NS 1.42 (p = .004)

NS NS 1.72 (p < .001) 1.58 (p < .001)

1.20 (p < .001) NS 1.04 (p = .041) 1.24 (p = .046)

Exposure frequency as an offline variable showed a significant effect on form recognition (OR = 1.13, 95% CI = [1.09, 1.22], p = .005) and meaning recall (OR = 1.20, 95% CI = [1.09, 1.31], p < .001), but it did not reach significance for meaning recognition (OR = 1.07, 95% CI = [0.95, 1.20], p = .249). Table 7.3 models all vocabulary outcomes as a function of online reading measures and offline exposure and lists significant odds ratios (OR) to reflect the effect size of the significance. Comparing the odds ratios of exposure frequency and online processing measures, it is shown that total reading times may, to some extent, explain vocabulary outcomes above and beyond the number of exposures.

7.4.4  Summary of results To summarize the results, online and offline measures confirmed the effect of repeated exposure on incidental learning gains for different vocabulary measures. Eye-tracking data showed that learners fixated longer on initial exposures of novel words and that gaze durations and total time gradually decreased, hitting a plateau in the middle before they declined once again to reach their lowest at the final exposure. First fixation times showed an increase from first to fifth exposure before they gradually declined until final encounters. A significant effect of repeated encounters was shown for both gaze durations and total times. A relationship between online processing patterns and learning gains was established for total reading times of all vocabulary outcomes; gaze duration predicted meaning recognition and recall, while first fixation duration only predicted form recognition of target words. Eyemovement results in general indicated that different fixation measures predicted different aspects of word knowledge. Regression outputs showed that, after controlling for the number of exposures, some online processing measures can still explain the variance in vocabulary learning above and beyond the effect of repetition. This may indicate that exposure per se does not depict the whole picture, because it does not cover for engagement and processing events. The overall picture of incidental vocabulary gains showed discrepancy across measures in that performance was highest in form recognition and lowest in meaning recall, pointing to a potential cognitive trajectory of second language lexical development.

7.5 Discussion The aim of the present study was to disentangle the effects of two factors known to facilitate vocabulary learning: attention and repeated exposures. Evidence for the association between online processing and offline vocabulary learning was provided by Godfroid and colleagues (2013) and Pellicer-Sánchez (2016) in ESL paragraph and story reading. Webb (2007) explored the effect of repetition on word knowledge. The current study builds on both directions of research by examining the joint effects of repetition and the amount of online processing and 150

Looking at words

exploring their association with learning gains in different vocabulary knowledge measures with Arabic L2 learners. The first question concerned the processing patterns of novel lexical items by L2 Arabic learners in sentence reading. The common observation about the fixation times is that learners recorded the longest fixations on first encounter and the shortest times on the last encounter of the target item in both gaze duration and total times. This result is in line with previous eyetracking studies that probed into the effect of frequency and repetition (e.g., Hyönä and Niemi 1990, Raney and Rayner 1995). Learners paid more attention to the pseudo words as low frequency items, but as they got repeated they may have gained frequency properties that mediated attention and made it easier for readers to retrieve them or reach a satisfactory level of word familiarity. At the level of individual participants, some readers showed a sharp rise in fixation times by exposures 6 and 7 and then a gradual decline until the last exposure. Perhaps this could be explained as potential ‘noticing’ of novel exemplars (Godfroid et al. 2013; Schmidt 1990); it may mark a peak of curiosity on the part of the learner that could promote successful meaning acquisition, after which the familiarity with the word causes reading times to decline. However, as stated earlier, this trend was not generalizable across the whole sample. What can be generalized, though, is that the pattern of online reading across participants clearly showed increased reading fluency of the target words along exposures. Further research would benefit from tapping into individual differences in reading behavior through self-reports or interviews that could probe into learners’ subjective experiences while processing the target words (Godfroid and Schmidtke 2013). Stimulated recall or self-reports can probably explain instances where individual readers’ fixation times peaked at a specific exposure or even point out instances where the participant was aware that he or she was paying attention to some item or that he or she has guessed a meaning or learned another aspect of a new word. Interviews can also provide accounts of learners’ awareness of the existence of novel words and their repetition and the strategies they used in interacting with the text. The second question sought to interpret learning gains in different vocabulary measures in terms of their potential relationship with the reading patterns of target words. Results from the present study largely supported the assumption tested by Godfroid et al. (2013) that more attention to target words, quantified as the summed total reading time spent on the lexical item, predicted how well learners demonstrated some knowledge of form and meaning of the item in the vocabulary posttest. The kind of associations found between fixation times and different types of vocabulary gain aligns with the claim that different eye-movement measures reflect different cognitive processes. Within the framework of the E-Z reader model (Reichle et al. 2003; Pollatsek et al. 2006), lexical processing has been posited to proceed in two stages: an early stage which the authors called “familiarity check” and a later stage referred to as the “completion of lexical access.” The fact that only first fixation durations predicted form recognition conforms to this hypothesis that early lexical processing is largely form focused. Gaze duration, as a somewhat later measure, predicted meaning recognition and recall, which may indicate that subsequent lexical processing of form-meaning mapping and encoding into memory becomes more important. The same principle would explain why total time, as a late measure, predicted productive recall as a type of higher-order lexical knowledge. The reading data in the present study provides a platform for further inquiry into how different eye-movement measures actually point to different cognitive processes. By including a test battery in future studies, eye-tracking researchers may be able to explore these associations in greater depth. The third question in the study concerned the cognitive trajectory of lexical development; i.e., the order in which the components of word knowledge emerge in the learners’ lexicon as 151

Ayman A. Mohamed

a function of repeated exposure. In line with previous literature on vocabulary acquisition (Nation 2001; Schmitt 2008, 2010), receptive knowledge of form seemed to be the first component to develop, followed by meaning recognition and finally meaning recall. These differential learning rates can be explained in terms of progression from lower-order to higher-order processes. They can also be related to the incremental nature of lexical development in that larger reading times initially contributed to building a sound knowledge of form and meaning, while there was still more need for additional reading and exposures to establish full knowledge of a given target word. In form recognition, the learner only needs to map a word form onto the memory traces of the orthographic form of the target word, while in meaning recall the learner needed to have enough chances to infer meanings correctly and subsequently decontextualize words and retain them in memory, which is evidently a more complex task. In general, the overall picture of vocabulary knowledge development in the current study shows predictable patterns in line with previous research. However, further research could provide more insight into how these knowledge measures relate to the quality of context and L2 readers’ meaning inference abilities. The fact that summed total times were a strong predictor of learning outcomes after controlling for total exposure might indicate that individual attention to target words can explain the variance in vocabulary learning above and beyond mere repeated exposures. This finding aligns with lexical processing data which showed that readers invested more time in initial encounters checking for familiarity and reanalyzing context. From a reader’s perspective, exposures were not equal in the amount of context and information they provided about target words. From this perspective, it is plausible to assume that the way readers utilized their repeated encounters with target words can predict learning outcomes beyond encounters, per se. Some methodological issues need to be discussed regarding the nature of tasks and participants in the present study. The first point concerns the use of pseudo words for the study. As learners were normally expected to know the real words for the target items (e.g., city, wear, or sad), they may have concluded that the novel words they encountered in reading were less frequent synonyms of the words they already knew, an impression that may have reduced the motivation or cognitive effort to incorporate the new lexical items. Moreover, the lab-controlled experiment condensed the number of exposures into one experimental session, which may not exactly match the typical incremental route that learners go through in incidental learning, where repeated exposures are spaced over longer periods of time. Conducting an eye-tracking experiment with Arabic script posed some technical issues in adjusting the default segmentation from right to left and assigning interest areas. After several trials, the font size 18 was found to be adequate for readers and for the software to segment word by word efficiently. The average reading speed for participants was below 100 words per minute, which is relatively much slower than what has been reported in studies on ESL reading (230–250 words per minute). This finding could be due to many factors, including text direction and the actual proficiency level and vocabulary size of Arabic learners. The overall vocabulary learning results largely mirror previous findings from ESL studies. However, a further study with more advanced learners and more extended text can point to specific characteristics of reading in Arabic as a foreign language. One specific feature in Arabic vocabulary is the root and pattern morphology, which is likely to facilitate guessing meaning from context. Because the present study used pseudo words, this concept was not applicable. It is thus recommended that a replication study should investigate learning real words from context in the light of root and pattern strategy. Following up the quantitative investigation of vocabulary learning with qualitative research involving interviews or stimulated recall would also add

152

Looking at words

more insight into the process of incidental vocabulary learning from L2 reading, and particularly Arabic as a foreign language.

7.6  Conclusion and pedagogical implications The design of the current study presented a model for further larger-scale extensive reading research in second language Arabic. The study provides additional insights in SLA vocabulary research and extends further understanding of the cognitive aspects of incidental vocabulary acquisition. As a newly integrated technology, the eye-tracking technique can answer specific questions about learners’ interaction with L2 material with considerable temporal and spatial accuracy. Implementing eye-tracking methodology in SLA is likely to open new avenues of investigation to uncover subtle cognitive processes in language acquisition in general and vocabulary development in particular. Learning more about the process of incidental learning would be beneficial for teachers and curriculum designers as they consider spacing and managing the lexical component in different informative contexts according to different proficiency levels and learning objectives. The optimal benefit from reading is more likely to be incremental in nature and would span over substantial reading material and over longer periods of time (Schmitt 2010). The results highlighted the cognitive aspects of engagement (Schmitt 2008) and involvement (Laufer and Hulstijn 2001), which have been emphasized in vocabulary acquisition research and particularly within the incidental learning framework. The online measures that captured the ongoing processing of new vocabulary demonstrated a distinction between learning opportunities that the text affords and the expected learning outcomes, as determined by textual features and readers’ involvement. Vocabulary learning from reading is not a byproduct of a single factor but is rather influenced by multiple variables with differing effect sizes. A significant variable not addressed here is context quality, which has been found to affect the acquisition of word meaning in incidental vocabulary learning from reading (Webb 2008). The fact that different eye measures predicted different vocabulary knowledge outcomes points to a possible interaction between encounters and context richness or the level of predictability of individual tokens within the text. Further replications can consider this variable for a more inclusive explanation of lexical gains from context. The results of the study are most clearly relevant to second language vocabulary learning and teaching. Maximizing exposure to vocabulary in rich contexts is a recommended strategy to ensure the best conditions for internalizing partially known words or acquiring new vocabulary. Exposure is not only confined to reading, but can also be extended to task-based learning where different input modalities (speaking, listening, reading, and writing) can integrate vocabulary learning goals in variable contexts. The present study discussed only the immediate effects of exposure in single sentences. Future studies are warranted to examine a widerscale extensive reading design that would target larger texts and graded readers spaced over multiple sessions, considering all factors of input, textual features and individual reading behavior as well as learners’ proficiency, vocabulary size, and motivation.

153

Ayman A. Mohamed

Appendix: Sample Online Reading Material Target

Sentence

Hospital 1 Money 1 Wear 1 Market 1 Money 2 Filler 1 Sad 1 Buy 1 Travel 1 Wear 2

.‫قال الطبيب إن والدي ليس مريض وممكن أن يمشي من الباقور ويذهب إلى البيت‬ .‫ذهبت أختي إلى السوق لتشتري مالبس ولكنها كانت تريد بزوق أكثر ألن المالبس غالية‬ .‫في كل يوم جدي يفتكس القميص والبنطلون ويذهب إلى المقهي ليجلس ويتكلم مع أصدقاءه‬ .‫سعيد يريد أن يذهب إلى المفسح ليشتري مالبس جديدة للشتاء القادم‬ .‫عندي بزوق كثيرة لكن ال أعرف ماذا أشتري لصديقتي في عيد الحب‬ .‫قال األستاذ إن الواجب جيد ولكن يريدني أن أكتبه على الكمبيوتر‬ .‫كانت سارة دائما سخيمة ألن ليس عندها أصدقاء ولكنها اآلن سعيدة بعد أن تعرفت على ليلى‬ .‫جون يريد أن يستبطع كاميرا جديدة والب توب في الكريسماس ولكن ليس معه فلوس كثير‬ .‫والدة صديقي تظعن دائما إلى أوربا في الصيف ألن الطقس يكون جميل‬ .‫سارة تريد أن تفتكس فستان جديد في الحفلة وستذهب مع صاحبتها إلى السينما بعد ذلك‬

References Al-Wabil, A. and George, R., 2010. An eye tracking study of Arabic typography readability. In: Proceedings of IADIS interfaces and human computer interaction. Freiburg, Germany: IADIS, 309–312. Brusnighan, S. M., and Folk, J. R., 2012. Incidental vocabulary acquisition: semantic transparency in novel compound word processing. Reading Research Quarterly, 47 (2), 72–190. Craik, F.I.M., & Lockhart, R. S., 1972. Levels of processing: A framework for memory research. Journal of Verbal Learning and Verbal Behavior, 11, 671–684. Chaffin, R., Morris, R. K., & Seely, R. E. (2001). Learning new word meanings from context: A study of eye movements. Journal of Experimental Psychology: Learning, Memory, and Cognition, 27 (1), 225–235. Fraser, C. A., 1999. Lexical processing strategy use and vocabulary learning through reading. Studies in Second Language Acquisition, 21 (2), 225–241. Godfroid, A., 2012. Eye tracking. In: P. Robinson, ed. The Routledge encyclopedia of second language acquisition. New York: Routledge, 234–236. Godfroid, A., et al., 2013. An eye for words: Gauging the role of attention in incidental L2 vocabulary acquisition by means of eye-tracking. Studies in Second Language Acquisition, 35 (3), 483–517. Godfroid, A., Housen, A. and Boers, F., 2010. A procedure for testing the Noticing Hypothesis in the context of vocabulary acquisition. In: M. Pütz and L. Sicola, eds. Inside the learner’s mind: Cognitive processing and second language acquisition. Amsterdam: John Benjamins, 169–197. Godfroid, A. and Schmidtke, J., 2013. What do eye movements tell us about awareness? A triangulation of eye-movement data, verbal reports and vocabulary learning scores. In: J. M. Bergsleithner, et al., eds. Noticing and second language acquisition: Studies in honor of Richard Schmidt. Honolulu: University of Hawai’i, National Foreign Language Resource Center, 183–205. Horst, M., 2005. Learning L2 vocabulary through extensive reading: A measurement study. The Canadian Modern Language Review, 61 (3), 355–382. Horst, M., Cobb, T. and Meara, P., 1998. Beyond a clockwork orange: acquiring second language vocabulary through reading. Reading in a Foreign Language, 11 (2), 207–223. Hulstijn, J., 1992. Retention of inferred and given word meanings: experiments in incidental vocabulary learning. In: P Arnaud and: H. Bejoint, eds. Vocabulary and applied linguistics. London: Palgrave Macmillan, 113–125. Hulstijn, J., Hollander, M. and Greidanus, T., 1996. Incidental vocabulary learning by advanced foreign language students: The influence of marginal glosses, dictionary use, and reoccurrence of unknown words. The Modern Language Journal, 80, 327–339. Hulstijn, J. and Laufer, B., 2001. Some empirical evidence for the involvement load hypothesis in vocabulary acquisition. Language Learning, 51 (3), 539–558. Hulstijn, J., 2001. Intentional and incidental second language vocabulary learning: A reappraisal of elaboration, rehearsal and automaticity. In: P. Robinson (Ed.), Cognition and second language instruction. Cambridge: Cambridge University Press, 258–286.

154

Looking at words Hulstijn, J. H. and Trompetter, P., 1998. Incidental learning of second language vocabulary in computerassisted reading and writing tasks. In: Albrechtsen, D., Hendricksen, et al., eds. Perspectives on foreign and second language pedagogy. Odense, Denmark: Odense University Press, 191–200. Hyönä, J. and Niemi, P., 1990. Eye movements in repeated movements of a text. Acta Psychologica, 73, 259–280. Jordan, T. R., et al., 2014. Reading direction and the central perceptual span: Evidence from Arabic and English. Psychonomic Bulletin & Review, 21 (2), 505–511. Keating, G. D., 2008. Task effectiveness and word learning in a second language: the involvement hypothesis on trial. Language Teaching Research, 12 (3), 365–386. Kim, Y., 2008. The role of task-induced involvement and learner proficiency in L2 vocabulary acquisition. Language Learning, 58(2), 285–325. Kliegl, R., et al., 2006. Tracking the mind during reading: The influence of past, present, and future words on fixation durations. Journal of Experimental Psychology: General, 135, 12–35. Kowen, S. and Kim, H., 2008. Beyond raw frequency: Incidental vocabulary acquisition in extensive reading. Reading in a Foreign Language, 20 (2), 191–215. Laufer, B., 2005. Focus on form in second language vocabulary learning. EUROSLA Yearbook, 5, 223–250. Laufer, B. and Hulstijn, J., 2001. Incidental vocabulary acquisition in a second language: The construct of task-induced involvement. Applied Linguistics, 22 (1), 1–26. Liversedge, S. P. and Findlay, J. M., 2000. Saccadic eye movements and cognition. Trends in Cognitive Sciences, 4, 6–14. Macaro, E., 2003. Teaching and learning a second language: A review of recent research. London: Continuum. Matsuoka, W. and Hirsh, D., 2010. Vocabulary learning through reading: does an ELT course book provide good opportunities? Reading in a Foreign Language, 22 (1), 56–70. Mohamed, A. A., 2016. Task-based incidental vocabulary learning in L2 Arabic: The role of proficiency and task performance. JNCOLCTL, 18, 121–157. Nation, I. S. P., 2001. Learning vocabulary in another language. Cambridge: Cambridge University Press. Paribakht, T. S. and Wesche, M., 1997. Reading and “incidental” L2 vocabulary acquisition: an introspective study of lexical inferencing. Studies in Second Language Acquisition, 21 (2), 195–224. Pellicer-Sánchez, A., 2016. Incidental L2 vocabulary acquisition from and while reading. Studies in Second Language Acquisition, 38 (1), 97–130. Pellicer-Sánchez, A. and Schmidt., 2010. Incidental vocabulary acquisition from an authentic novel: Do things fall apart? Reading in a Foreign Language, 22 (1), 31–55. Peters, E., et al., 2009. Learning L2 German vocabulary through reading: The Effect of three enhancement techniques compared. Language Learning, 59, 113–151. Pigada, M. and Schmitt, N., 2006. Vocabulry acquisition from extensive reading: A case study. Reading in a Foreign Language, 22 (1),1–28. Pollatsek, A., et al., 2006. Tests of the E-Z Reader model: Exploring the interface between cognition and eye-movement control. Cognitive Psychology, 52, 1–56. Pulido, D., 2007. The relationship between text comprehension and second language incidental vocabulary acquisition: A matter of topic familiarity? Language Learning, 57 (1), 155–199. Raney, G. and Rayner, K., 1995. Word frequency effects and eye movements during two readings of a text. Canadian Journal of Experimental Psychology, 49 (2),151–172. Rayner, K., 1998. Eye movements in reading and information processing: 20 Years of research. Psychological Bulletin, 124, 372–422. Rayner, K., 2009. in Eye movements and attention reading, scene perception, and visual search. The Quarterly Journal of Experimental Psychology, 62 (8), 1457–1506. Read, J., 2004. Research in teaching vocabulary. Annual Review of Applied Linguistics, 24, 146–161. Reichle, E.D., et al., 2003. The E-Z Reader model of eye movement control in reading: comparison to other models. Brain and Behavioral Sciences, 26, 445–476. Rott, S., 1999. The effect of exposure frequency on intermediate language learners’ incidental vocabulary acquisition and retention through reading. Studies in Second Language Acquisition, 21 (3), 589–619. Schmidt, R. W., 1990. The role of consciousness in second language learning. Applied Linguistics, 11 (2), 129–158. Schmitt, N., 2010. Researching vocabulary: A vocabulary research manual. Basingstoke, UK : Palgrave Macmillan.

155

Ayman A. Mohamed Schmitt, N., 2008. Review article: Instructed second language vocabulary learning. Language Teaching Research, 12 (3), 329–363. Starr, M. S. and Rayner, K., 2001. Eye movements during reading: Some current controversies. Trends in Cognitive Sciences, 5, 156–163. Waring, R. and Takaki, M., 2003. At what rate do learners learn and retain new vocabulary from a graded reader? Reading in a Foreign Langugae, 15, 130–163. Watanabe, Y.,1997. Input, intake and retention: Effects of increased processing on incidental learning of foreign language vocabulary. Studies in Second Language Acquisition,19, 287–307. Webb, S., 2007.The effects of repetition on vocabulary knowledge. Applied Linguistics,28, 46–65. Webb, S., 2008. The effects of context on incidental vocabulary learning. Reading in a Foreign Language, 20 (2), 232–245. Williams, R. S. and Morris, R. K., 2004. Eye movements, word familiarity, and vocabulary acquisition. European Journal of Cognitive Psychology, 16, 312–339. Winke, P. M., et al., 2013. Introduction to the special issue: Eye-movement recordings in second language research. Studies in Second Language Acquisition, 35 (2), 205–212.

156

8 KEYWORD VS. CONTEXT STRATEGIES AMONG DIFFERENT LEVELS OF ARABIC LANGUAGE LEARNERS Olla Najah Al-Shalchi Vocabulary of a language makes up approximately 75% of comprehension (Nagy and Scott 2000). Researchers agree that learning vocabulary is more effective when learners are given strategies to learn the vocabulary (Cheng 2011; Lee et al. 2010; Liu 2010; Teow et al. 2010), and although vocabulary learning is an important process in language learning, it may be even more challenging for learners who must first learn a new writing system (Muljani et al. 1998; Wang et al. 2003; Hamada and Koda 2008), like Arabic. Only through research can it be determined what strategies work best for different groups of learners while considering the perceived workload to ensure that it is not too much or little. The purpose of this study was to compare two strategies (keyword and context) with learners enrolled in a public institution consisting of undergraduate L1 English learners of Arabic in various Arabic levels (Arabic 1–3) to determine their impact on vocabulary learning and perceived workload during instruction by using a quantitative, experimental research design. Three research questions guided this study: (1) Does the strategy (keyword vs. context) effect vocabulary learning in learners in various levels of Arabic?, (2) Does the strategy (keyword vs. context) affect perceived workload in learners in various levels of Arabic?, and (3) Does the strategy (keyword vs. context) affect actual strategy use in learners in various levels of Arabic? The results of the study are presented and examined, with a focus on each level of Arabic (Arabic 1–3) independently, to explain that learners of various language abilities learn differently and therefore should be provided with different strategy usage. In addition to practical implications, suggestions for future research are provided.

8.1 Introduction 8.1.1  The keyword mnemonic strategy A common strategy in direct vocabulary instruction is the keyword mnemonic strategy. This strategy involves three steps, often referred to as the three R’s (relating, recoding, and retrieving). First, the keyword in the native language should sound similar to the foreign language vocabulary word. Next, there needs to be an image that depicts an interaction of the keyword and the meaning of the foreign language word. Last, when the learner is given the foreign language word, he/she needs to remember the association and the image to correctly define the 157

Olla Najah Al-Shalchi

word (Mastropieri et al. 1986). The importance of resemblance between the keyword and the foreign language word is essential (Shaughnessy 2003). Research has shown this strategy increases the vocabulary learning of foreign language learners (e.g., Atkinson and Raugh 1974; El Sawy 2002; Hall 1988; Raugh and Atkinson 1974; Sagarra and Alba 2006; van Hell and Mahn, 1997). The keyword mnemonic strategy received attention after Atkinson and Raugh (1975) documented how using the strategy helped learners acquire Russian vocabulary (Levin 1993). Russian was chosen as the foreign language in the study because it does not sound similar to English vocabulary (the L1), thus presenting an added challenge. After having three days to learn 120 Russian words, the keyword mnemonic strategy group could recall nearly threefourths of the words. The second group used any strategy they preferred and could recall approximately one-third of the words (Reed 2006). In two instances when using the keyword strategy in a German language class, learners learned the definitions and gender (i.e., masculine, feminine, or neutral) of German words (Desrochers et al. 1989; Desrochers et al. 1991). In both studies, the learners who were given instruction in using the keyword mnemonic strategy had better recall of vocabulary. In order for the participants to correctly determine the gender of the noun, they must have been able to correctly retrieve the image and definition of the word. If the participant was unable to retrieve the meaning, then identification of the gender of the noun was merely a guess. The keyword mnemonic strategy has been used in several instances and under different circumstances with positive results. While the keyword mnemonic strategy is a direct strategy that learners follow step by step, the context strategy forces the learners to use their background knowledge in the subject and build upon their expertise. The question is whether this strategy will work with learners that have established schema in the language they are learning or if the detailed guidance will interfere with learning.

8.1.2  Context strategy The context strategy is an indirect strategy used in vocabulary instruction. This strategy requires the use of the vocabulary word in multiple sentences. This strategy assumes the learner should be able to decipher the meaning of the word based on the sample sentences (Greenwood 2002). The context strategy prompts learners to look for cues such as synonyms, antonyms, prefixes, and suffixes that will allow them to decipher the correct meaning of the word (Nash and Snowling 2006). This strategy may help learners become independent learners (Decarrico 2001), because they are not being specifically told what the words mean and are decoding the meaning on their own. The context strategy has been praised because many words that learners may come across are low frequency, and teaching learners how to deal with these types of words is a better strategy than requiring learners to memorize lists of words which may rarely be used or encountered (Redouane 2010). When using this strategy, not only do learners learn the meaning of the word, but they can recall the syntax, pragmatics, and emotion associated with the meaning of the word (Gu and Johnson 1996). However, because learners may incorrectly infer the meaning of words, the strategy may not be practical for learners who are at a beginning level. A positive correlation has been demonstrated between the amount of vocabulary that a learner knows and the effective use of the strategy (Redouane 2010). Therefore, this strategy may not be appropriate to use with novice language learners. 158

Keyword vs. context strategies

8.1.3  A comparison between keyword mnemonic and context strategies The keyword mnemonic and context strategy represent two different instructional strategies used for foreign language learning. The design and use of these strategies can serve different purposes for different audiences. At present, there is little information comparing and contrasting the effects of each of these strategies on students, especially those with different levels of prior knowledge. In a study where the two strategies were compared, the researchers were interested in finding out if it the strategies would yield the same results (McDaniel et al. 1987). To test the effectiveness of these two strategies, 22 participants were randomly selected to learn 30 nonsense English-based words using the keyword mnemonic strategy, and 20 participants were randomly selected to learn the same words using the context strategy. The participants using the context strategy were given a short paragraph of three sentences in which the definition could be inferred. An immediate post-test showed that the keyword mnemonic group performed better, but after a one-week delayed post-test no significant differences were found. In a second study, participants were students who had studied English as a second language for at least two years (Rodriguez Sadoski 2000). The purpose of the study was to test four strategies (i.e., the keyword mnemonic, context, rote rehearsal, and context/keyword mnemonic) and determine their effectiveness. The participants who received the context/ keyword mnemonic strategy outperformed the other participants when tested in a delayed post-test. This suggests that using a combination of the keyword mnemonic strategy and the context strategy may lead to longer retention rates. Rodriguez and Sadoski (2000) also note that their findings and implications can provide foreign language teachers with vocabulary building tools, and although the keyword mnemonic strategy seems to have been the best of the four techniques used, the learners themselves had been learning a foreign language for years. The participants were not asked to report about the techniques used, so it is possible that learners had a system of their own that they were accustomed to using and continued to do so. Despite the possible confounding variable, the researchers are certain that using the keyword mnemonic strategy to teach foreign language vocabulary is an area that needs to be further studied and can easily become one of the techniques that teachers use in vocabulary learning (Rodriguez Sadoski, 2000). In addition, in many studies, participants are introduced to a strategy for only a very limited time before being tested, not allowing the learner enough time to have practiced using the strategy (Abd Ghani and Zulkiply 2008). The results of such studies may not show the true impact of using the strategy, since learners have not been exposed to the strategy for longer periods, which this study aims to do.

8.1.4  Workload with learners of various backgrounds Learners should acquire a variety of strategies because learning is a complex process, and not every strategy will meet the objective of the lesson (Chamot 2006). Instructional designers need to determine through research what strategies work best for different groups of learners while taking into consideration the amount of perceived workload to ensure that it is not too much or too little. Naturally, because not every strategy will help a learner achieve the objective of the lesson, it can be argued that strategies used by novice learners will differ from the strategies 159

Olla Najah Al-Shalchi

used by more advanced learners (Hsiao Oxford 2002). The extent to which someone is familiar with a particular subject will affect how quickly information can be stored and processed. The brain has an unlimited capacity to store information, but when dealing with a new topic, working memory is limited in how much information can be processed. Yet, even with this limitation, information that is stored in long-term memory helps restructure the new information to reduce the workload (Kalyuga 2007). Another way to reduce workload is by practice, and through practice, the information will come automatically (Cooper and Sweller 1987). Kalyuga (2007) states that the design of most instructional material is aimed at novice learners and does not take into account that learners of different levels of expertise will be using the same material. Because the expertise reversal effect has shown that novice learners need to use different strategies than those who are more advanced, instructional designers need to use different strategies with learners of different backgrounds. Novice learners require more detailed instruction and support to help build the new knowledge structures, while learners with background knowledge already have built the knowledge structures, or schema, and too many details may slow and hinder their learning (Kalyuga 2007). Learners with background knowledge in the domain may feel bored with the task if too much detail is given to them, and if the perceived workload is too little, this will cause poor performance.

8.2  Purpose of the research and research questions The purpose of this study was to compare keyword and context strategies with learners enrolled in various undergraduate Arabic courses at three proficiency levels (first, second, and third year) to determine their impact on vocabulary learning and perceived workload during instruction. Research is needed to determine which strategies work best for various groups of learners during vocabulary learning in the foreign language classrooms. The present study compared the effectiveness of two strategies, the keyword and context strategies, in the hope of adding to the research that has been done in this area and also to bring in a new perspective in determining whether the proficiency level of the language learner plays a role in the effectiveness of a strategy by examining the differences in perceived workload. Both the keyword mnemonic and context strategies have been extensively researched in different disciplines, and both have been shown to be effective (e.g., Atkinson and Raugh 1975; Hulstijin 1992; McDaniel et al. 1987; Pressley et al. 1980). Most of the research that has focused on these vocabulary strategies tested the effects of a strategy and vocabulary retention (Gu and Johnson 1996). However, very little research exists in which the two strategies were compared to determine if one is more effective than the other. Furthermore, no research has been done to determine if the learners’ proficiency level correlates with the most effective strategy to use, nor was there research done looking specifically at Arabic L2 learners. The present study was guided by three main research questions: 1 Does the strategy effect vocabulary learning in learners in various levels of Arabic respectively? 2 Does the strategy affect perceived workload in learners in various levels of Arabic respectively? 3 Does the strategy affect actual strategy use in learners in various levels of Arabic respectively? 160

Keyword vs. context strategies

8.3 Methods 8.3.1 Participants The study was conducted with undergraduate, non-heritage learners of Arabic at three institutions in the United States. A total of nine courses were used to recruit participants. A total of 116 participants were used, belonging to three proficiency groups: a beginner group (referred to hereafter as “Arabic 1”), an intermediate group (referred to hereafter as “Arabic 2), and an advanced group (referred to hereafter as “Arabic 3”). Table 8.1 summarizes the gender, age, and length of exposure to Arabic.

8.3.2 Treatments All participants within each class were randomly assigned to use either the keyword strategy or the context strategy and were provided with the training during five consecutive class sessions. Half of the participants received the keyword treatment and the other half received the context strategy treatment. The study treatments were part of the normal class sessions and were in a class activity that was done at the beginning of each class period for approximately 15 minutes. The instructional treatments consisted of a training packet on either the keyword or context strategy. This material included an explanation of the strategy with two examples. Nine vocabulary words were presented to the participant using the strategy, and then the participant was given the task of creating nine more of his or her own examples. The training packets were randomly distributed to the participants at the beginning of the first training session and then were collected and redistributed in the following four class periods. To ensure that participants were working on their training material, they were awarded a participation grade, and at the end of the five days, the participation grades were averaged, and only those who scored an 80% overall were used. Participants using the keyword strategy were provided with an explanation and two examples to become familiar with the strategy. The explanation and the examples were the same for each level, but the vocabulary words that each level learned differed (for samples, see Appendix A for the two examples given to the intermediate group). Each day for the first 15 minutes of class, the participants were given their packets to review the strategy explanation and examples and work on creating their own mnemonics to learn vocabulary. The participants were given vocabulary words and asked to think of a word in their native language that sounds similar to the given word. Then participants needed to draw an image in the space provided which included the meaning of the word and the word from their native language. Every time the foreign language word was said or read, the participant was Table 8.1 Participants Groups of Participants

Length of Exposure

Arabic Course Credit Hours

Gender F/M

Age Range

Age Means

Arabic 1 (n = 42) Arabic 2 (n = 36) Arabic 3 (n = 38)

Year 1 Year 2 Year 3

5 3 3

30/12 22/14 24/14

18–21 19–21 19–24

19.5 20 21.5

F/M = Total Females/Total Males

161

Olla Najah Al-Shalchi

asked to think of that image to recall the meaning. For the participants receiving the context strategy, they were provided with sample sentences in which the vocabulary word was used. Based on the context of the sentences, the participant wrote their own definition of the word in the space provided. They also had to create their own sentences using the vocabulary.

8.3.3  Measurement of dependent variables 8.3.3.1  Perceived workload After each training session, the instructor passed out the self-reporting scales to allow participants to rate their perceived workload. A modified self-reporting scale was adopted from NASATLX (Hart and Staveland 1988) to measure perceived workload (see Appendix B). The participants were asked to rate each of the five factors (i.e., mental demand, temporal demand, performance, effort, and frustration) that may contribute to workload. The scale ranged from very low (0) to very high (100) on all the factors that contribute to perceived workload. An average score from all of the factors was calculated to give an overall workload score for each day.

8.3.3.2  Arabic vocabulary learning All of the participants took a pre-test /post-test consisting of 14 questions. The tests were not counted as scores for the class. The test format was identical for all levels and consisted of a matching and a cloze section, with seven questions in each section. Because different levels of Arabic were being tested, the vocabulary used in each test differed.

8.3.3.3 Pre-test/post-test A pre-test was given to ensure that the participants did not know the vocabulary that they would be learning during the training. Each question answered correctly received 1 point, with a maximum of 14 points possible. Anyone scoring above a 35% was eliminated in this study. After completing all the training sessions, a post-test, identical to the pre-test, was given. The scoring of the post-test was the same as for the pre-test. The dependent variable was measured by the result of the post-test (for an example, see Appendix C for the post-test administered to the intermediate group).

8.3.3.4  Frequency of strategy use After each section on the post-test, participants reported how often they used the strategy they trained with during testing on that one section by indicating that they used the strategy on most of the questions, some of the questions, or none of the questions.

8.3.4  Study procedures An experimental quantitative methods research design approach was proposed for this study. In all courses, the textbooks used present vocabulary as a list with the Arabic word and the English equivalent. The vocabulary used in the study was randomly selected from one chapter from the textbook for each class. Then nine of the words from the chapter were randomly selected to be presented to the participants by using either the keyword or context strategy. Another nine vocabulary words were selected to allow the participants to create their own 162

Keyword vs. context strategies

mnemonic or context. The researcher and two subject matter experts designed the training material that was presented to the participants, and all of the training material was piloted prior to implementation. Ten additional volunteers from each of first-, second-, and third-year classes were given the training materials to use as well as to calculate the approximate time required to complete the training. These participants did not participate in the study beyond helping with calculating the time required to complete the training. Before the participants were given any information about the strategy, they had 15 minutes to complete the pre-test during their regular class session, and afterwards they were given the information sheets about the study. In the following class period, each participant was randomly assigned to either the keyword or context strategy by randomly distributing the packets to the participants, and in each packet, the instructions on using the strategy were provided. After all the packets were distributed, the instructors informed the class that their participation would be recorded and graded during the next 15 minutes while working on the training. The participants read the information in the packets silently and then examined the examples. Then the participants had time to create their own examples. At the end of the 15 minutes, the instructors passed out the modified NASATLX survey to rate the amount of perceived workload that was exerted. The packets were collected by the instructors at the end of the training session. For the next four class periods, at the beginning of each class, the instructor returned the packets to the participants. The same procedures were used as in day 1 of training, and all materials were returned to the instructor at the end of the 15-minute period. During the seventh class period the post-test was administered. The participants had 15 minutes to complete the test and answer two questions to rate how often they used the strategy to help answer the questions on the test.

8.4  Data analysis and results The data collection for this study consisted of the results of the post-tests, the ratings of perceived workload, and the self-reported usage of the strategy during post-testing. SPSS software was used for the statistical analysis. The quantitative results are presented here.

8.4.1  Keyword vs. context at each level (research question 1) The first research question asked if the strategy (keyword vs. context) would effect vocabulary learning for learners in various levels of Arabic. Three separate one-way ANOVA tests were calculated. The results of Arabic 1 learners show there was no significant effect of the independent variable, strategy, on the dependent variable at the p = .05 level F(1, 41) = .41, p = .53, ηp2 = .01. The mean score for the keyword group was 13.05 and for the context group 12.81. The results of Arabic 2 learners show there was a significant effect of the independent variable, strategy, on the dependent variable at the p = .05 level, F(1, 35) = 15.80, p = .000, ηp2 = .32. The mean score for the keyword group was 11.33 and for the context group 7.88. The results of the Arabic 3 learners also indicate there was a significant effect of the independent variable, strategy, on the dependent variable at the p = .05 level, F(1, 37) = 34.64, p = .000, ηp2 = .48. The mean score for the keyword group was 6.8 and for the context group 12.45.

8.4.2  Perceived workload at each level (research question 2) The second research question aimed to determine whether the strategy (keyword vs. context method) would affect perceived workload for learners in various levels of Arabic. To 163

Olla Najah Al-Shalchi

Perceived Workload

determine the effect of strategy on perceived workload, three separate two-way, within-subjects ANOVA or a 2 × 5 fixed factor were calculated. For Arabic 1, data were analyzed using a mixed-design ANOVA with a within-subjects factor of days and a between-subjects factor of strategy. Mauchly’s test indicated that the assumption of sphericity had been violated (χ2(9) = 59.19, p < .001); therefore, degrees of freedom were corrected using GreenhouseGeisser estimates of sphericity (ε = 0.56). Main effects of subscale, F(2.25, 90.08) = 66.51, p < .001, ηp2 = .62, and strategy, F(2.25, 90.08) = .37, p >.001, ηp2 = .72, were qualified by an interaction between days and strategy, F(2.25, 90.08) = 66.51, p < .001, ηp2 = .62. Figure 8.1 illustrates a downward linear trend for both strategies across the five days, with the keyword strategy having higher perceived workload than the context strategy group each day. For Arabic 2, data were analyzed using a mixed-design ANOVA with a within-subjects factor of days and a between-subjects factor of strategy. Mauchly’s test indicated that the assumption of sphericity had been violated (χ2(9) = 63.207, p < .001); therefore, degrees of freedom were corrected using Greenhouse-Geisser estimates of sphericity (ε = 0.480). Main effects of subscale, F(1.92, 65.29) = 101.30, p < .001, ηp2 = .75, and strategy, F(1.92, 65.29) = .65, p > .001, ηp2 = .52, were qualified by an interaction between days and strategy, F(1.92, 65.29) = 101.30, p < .001, ηp2 = .75. Figure 8.2 shows a downward linear trend for perceived workload across the five days for both the keyword and context strategies for Arabic 2 participants. For Arabic 3, data were analyzed using a mixed-design ANOVA with a within-subjects factor of days and a between-subjects factor of strategy. Mauchly’s test indicated that the assumption of sphericity had been violated (χ2(9) = 66.30, p < .001); therefore, degrees of freedom were corrected using Greenhouse-Geisser estimates of sphericity (ε = 0.48). Main effects of subscale, F(1.9, 68.39) = 131.81, p < .001, ηp2 = .79, and strategy, F(1.9, 68.39) = 2.54, p > .001, ηp2 = .09, were qualified by an interaction between days and strategy, F(1.9, 68.39) = 131.81, p < .001, ηp2 = .79. Figure 8.3 shows a downward linear trend for perceived workload across the five days for both the keyword and context strategies for Arabic 3 participants. To determine if there were any significant differences for each day, five separate one-way ANOVA tests were calculated for each level for each day. In order to control for the familywise type I error rate, a p = .01 was used. The only significant findings found were in Arabic 3 learners on days 4 and 5. For day 4, a significant effect of the independent variable, strategy (keyword vs. context), on the dependent variable (day) at the p = .01 level, F(1, 36) = 14.18, p = .001, ηp2 = .28 was found. The mean score for the keyword group was 25.05 and for the context group 37.21. For day 5, a significant effect of the independent variable, strategy 70 60 50 40 30 20 10 0

Keyword Context

1

2

3

4

5

Days

Figure 8.1 Workload for Arabic 1

164

Perceived Workload

Keyword vs. context strategies 70 60 50 40 30 20 10 0

Keyword Context

1

2

3

4

5

Days

Perceived Workload

Figure 8.2 Workload for Arabic 2 60 50 40

Keyword Context

30 20 10 0

1

2

3

4

5

Days

Figure 8.3 Workload for Arabic 3

(keyword vs. context), on the dependent variable (day) at the p = .01 level, F(1, 36) = 10.96, p = .002, ηp2 = .23 was found. The mean score for the keyword group was 16.16 and for the context group 26.32.

8.4.3  Strategy usage at each level (research question 3) The third research question aimed to determine whether the strategy would affect strategy use for learners in various levels of Arabic. Participants rated how often they used the strategy they were assigned while completing the post-test. A 5-point Likert scale was used to record level of strategy use, with 0 indicating that the strategy was used none of the time and 5 indicating the strategy was used all the time. For each level of Arabic, two separate one-way ANOVAs were calculated. The first ANOVA tested whether any differences existed in strategy usage on the cloze section of the test, and the second ANOVA tested whether a difference existed on the matching section. In the cloze section for Arabic 1, there was a significant effect of the independent variable, strategy, on the dependent variable at the p = .05 level for the three conditions, F(1, 40) = 5.78, p = .021, ηp2 = .13. The mean score for the keyword group was 2.95 and for the context group 2.19. For the matching section, no significant effect of the independent variable, strategy, on the dependent variable at the p = .05 level, F(1, 40) = .13, p = .72, ηp2 = .003 was found. The mean score for the keyword group was 2.76 and for the context group 2.62. For Arabic 2, for the cloze section, there was a significant effect of the independent variable, strategy, on the dependent variable at the p = .05 level, F(1, 34) = 6.92, p = .013, ηp2 = .17. 165

Olla Najah Al-Shalchi

The mean score for the keyword group was 2.22 and for the context group 2.5. For the matching section, there was a significant effect of the independent variable, strategy, on the dependent variable at the p = .05 level, F(1, 34) = 9.74, p = .004, ηp2 = .22. The mean score for the keyword group was 3.22 and for the context group 3.61. For Arabic 3, for the cloze section, there was no significant effect of the independent variable, strategy, on the dependent variable at the p = .05 level, F(1, 38) = .28, p = .60, ηp2 = .007. The mean score for the keyword group was 2.3 and for the context group 2.5. For the matching section, there was no significant effect of the independent variable, strategy, on the dependent variable at the p = .05 level, F(1, 38) = 1.28, p = .26, ηp2 = .03. The mean score for the keyword group was 2.7 and for the context group 2.25.

8.5 Discussion 8.5.1  Keyword and context strategies with different levels of learners The first research question examined the differences between the keyword and context strategy groups in regard to their post-test results. Researchers agree that there may not be one single effective strategy for vocabulary learning in foreign languages, and that many factors, such as the learning environment and motivation, must be taken into consideration when trying to conclude whether a strategy is effective in learning vocabulary (Sadeghi and Nobakht 2014). Therefore, it is not surprising that the results of the post-test scores provided three different outcomes. No significant differences were found in learners in Arabic 1. In Arabic 2, the keyword group outperformed the context group, and in Arabic 3 the context group performed better. McDaniel et al. (1987) found that learners who used the keyword strategy did better than the learners who used the context strategy when given an immediate post-test. The same results were repeated with the Arabic 2 learners. The keyword group was superior to the context group when given an immediate post-test. In Arabic 3, the context group was superior to the keyword group. It is important to note that Arabic words, like other Semitic languages, are derived from roots and patterns. Most Arabic words consist of three consonants that form a root inserted into different patterns, which may consist of vowels and other consonants, to give the meaning of the word. In English, for example, Ryding (2005) explains how the English [sng] consonants can be thought of being a root, and from that stems the words sing, sang, song, singing, etc. The different vowels when inserted with the root [sng] provide different meanings to the word, and so do any prefixes and/or suffixes that can be added. Arabic morphology functions in this same way, with the root providing a meaning and a pattern providing a separate meaning. Together the root and pattern give the meaning of the word. Because of this, logically, advanced Arabic learners would perform better using the context strategy. Not only are learners able to rely on the surrounding context of the word to help them decipher the meaning of the unknown word, but they are also able to rely on the root and/or pattern of an unknown word, since their vocabulary knowledge is more developed than novice learners. These results are also consistent with Van Hell and Mahn’s (1997) study, in which more experienced language learners found the keyword strategy to be less effective. The more experienced learners are aware of the phonological differences between a person’s native language and the foreign language (Van Hell and Mahn 1997). The more someone studies a language, the better his or her phonological memory will become, which may aid in learning new vocabulary. Van Hell and Mahn’s (1997) participants used either the keyword strategy or rote 166

Keyword vs. context strategies

rehearsal, and while rote rehearsal was not used in this study, it proposes that learners of higher proficiency levels may benefit more from using a strategy other than the keyword strategy. Interestingly, each strategy relies on different types of processing. The keyword strategy relies more on acoustic and visual images, which is a shallow type of processing. The context method uses deep processing by requiring the learner to use semantics to infer meaning, which in turn would be superior according to the depths of processing theory (Craik and Tulving 1975).

8.5.2  Decrease in perceived workload The second research question examined the perceived workload of the learners to determine if this would predict how successful a learner would be when using a strategy. In all levels, a downward linear effect was observed for workload reported over the training days. Perceived workload was highest on the first day of training, slowly decreased over time, and was lowest on the final day of training. Although no significant differences were found between the strategies in each level, all groups showed the same tendency in regard to perceived workload. Two recent studies examined whether any differences would exist in regard to workload and proficiency level of the learner. In Kor and Chuah’s (2014) study, a mathematical software was used, and no significant differences were found between the technology usage of the individual and workload. In the second study, researchers observed the cognitive load of students of different achievement levels when using an inquiry-based mobile learning model (Shih et al. 2010). Once again, no significant differences were found among the low, middle, and high achievement groups of learners. Together all three of these studies suggest that the ability of the learner does not play a role in determining perceived workload and that neither strategy places any type of extra burden on the learner and should not impact learning ability. The fact that workload decreased over time is likely a result of the learner becoming accustomed to the training. The more exposure a learner has with a given situation, the more likely it is that workload will be reduced over time (Meissner and Bogner 2012). When the learner is first given the training, it is a new experience and there is no background to rely on, which may lead to high perceived workload. More time to process how to use the strategy is necessary on the first day in comparison to the fifth day. However, as time goes on, the training becomes more familiar and the learner has practiced more, thus reducing workload. Perceived workload may predict how successful a learner may be in a given task. In an ideal learning environment, learners will exhibit high performance with low perceived workload (Galy et al. 2012). In the current study, although workload was consistently decreasing across all groups, not all groups performed equally on the post-test. In Arabic 1, the post-test scores were high; in Arabic 2 only the keyword group post-test scores were high, and in Arabic 3 only the context group scored high on the post-test. The workload for the context learners in Arabic 2 and the keyword learners in Arabic 3 decreased over time, which may be due to other factors. Two factors that may play a role in performance are motivation and emotion (Clark et al. 2006). When performance and workload are low, a lack in motivation may be the cause (Clark et al. 2006). Motivation is a key element in predicting success. Learners lacking motivation may stop investing in the workload. The vocabulary that participants learned in this study were words appropriate for the proficiency level. However, these words were not a part of the participants’ current study, and this may have caused some participants to lack motivation to learn the words. Others may have found the task overwhelming, and this may lead to some type of distraction because they feel that they cannot control whether they succeed or fail (Clark et al. 2006). Additionally, “more 167

Olla Najah Al-Shalchi

advanced learners may not be motivated to invest mental effort in learning tasks that were designed for novices, or to use approaches that are excessively structured” (Paas et al. 2005, p. 30). The keyword strategy is more structured and gives the learner more rules to follow in comparison with the context strategy. On the other hand, a task may be too easy for a learner who is not willing to invest in the workload and therefore does not learn (Paas et al. 2005).

8.5.3  Using the strategy The final research question tried to determine whether learners would use the strategy when being tested. The results are summarized in Table 8.2. In Arabic 2 and 3, in order for the learner to reach beyond a novice level in the language, the learners are moving away from just memorizing vocabulary and simple sentence structures to more complex structures and being able to speak about more complicated topics. For example, in Arabic 1, students are able to speak about families, hobbies, and daily activities. In Arabic 2, students begin to speak about well-known people in history, migration, and religion. In Arabic 3, students are often working with authentic texts and are becoming involved with media Arabic. Oxford (1990) explains that a feasible cause for not observing high usage memory strategies among more advanced learners is that these strategies are often used for more novice learners while they are still in the beginning stages of language learning. The more advanced learners become, the less likely they are to use the memory strategies. The results of the current study support Oxford’s claim, since there was a high usage of keyword found only in Arabic 1. Although no qualitative data was sought from this study, some participants wrote comments on their post-tests about the strategy use. One participant wrote that she is “a verbal learner, not a pictorial learner . . . [and] that the extra word confuses the definition.” Instead of remembering the definition of the word, she felt that she would only be able to remember the English word that she was linking the Arabic word to. This learner reported using the keyword strategy less than half of the time in both sections of the test. Prior research recommends making sure that a strategy is a preference for the learner and is compatible with his or her learning style (Oxford 2003). It is very possible that other learners felt the same way, and when a learning style and strategy do not match, it can hinder learning (Oxford 2003). In Arabic 3, two learners wrote that they used their own strategies, and they reported not using the strategy that they had received training on. These two learners scored 13 out of 14 and 12 out of 14 on the post-test (approximately 93% and 86%). Although they reported not using the strategies in the post-test, they are clearly very aware of knowing how they learn. Other research shows that students who are better able to explain what strategy they use and why are better language learners (Oxford 2003), and although the students were not asked to report about why they did not use a strategy, these two learners did. They also scored high on the post-test, which supports the claim that good language learners articulate how they learn (Chi et al. 1994).

Table 8.2 Strategy usage by section Level

Cloze Section

Matching Section

Arabic 1 Arabic 2 Arabic 3

keyword higher context strategy higher no significant differences

no significant differences context strategy higher no significant differences

168

Keyword vs. context strategies

8.6  Conclusion and pedagogical implications Often in the language class, no time is devoted to teaching vocabulary strategies. Learners are presented with lists and have to figure out on their own how to learn the words. If researchers can determine what strategies good learners use and help make learning be more effective, these strategies can then begin to be implemented in the classroom, whether it be an in-class or out-of-class activity. Research has shown that advanced language learners use strategies more often and are better able to master the foreign language (Zare 2012). While this study tried to determine if learners would use a strategy after receiving training, it may be more beneficial for future studies to evaluate what type of strategy learners use. A number of factors such as gender, motivation, and proficiency level can play a role in determining how strategies are used (Zare 2012). The current study did not examine these areas, which can be explored in more detail for future research. A number of recommendations to the current study can be made to improve this research in future investigations. First, even though the first 15 minutes of class time was allotted for the participants to use the strategy, not all the participants used all 15 minutes, as was intended. Instructors mentioned that some participants came to class 2–3 minutes late, and/or some participants finished the required problems and reviewing the material before 15 minutes. Having participants write down the time they began and ended the training would help determine the correct amount of time to allot for each strategy. Because different participants in each class were working on either the keyword method or the context method, not all the allotted time was necessary for some participants. The strategies require different tasks for the participants to complete, and the amount of time needed to use each strategy may differ. Therefore, it may be best to randomly assign the classes to receive one of the treatments rather than have each participant randomly assigned to a treatment. Finally, it is necessary to include more elaborate instructions for the participants using the strategy. Some of the participants did not use the strategy to the extent that was intended. Many of the participants of the keyword group did not include an illustration to go along with their sentence, and they may have only come up with a mental image. Even when prompted by the instructors to include an image, some participants felt that they could not draw an appropriate image. Participants of the context strategy sometimes used the vocabulary word in only one sentence, but in the examples, they were always provided with three separate sentences. It is unclear whether the participants did not know how to use the word in more than one sentence or if the participants thought that one sentence was sufficient for learning the word. Nevertheless, the results of this study suggest that as a learner begins to learn a language a more structured strategy, such as the keyword, is more effective than a less structured strategy, such as the context strategy. Special attention should be taken into consideration when dealing with intermediate learners, who are neither novices nor experts. The advanced learners have a tendency to perform better with the context strategy, which is supported by the expertise reversal effect. Perceived workload tends to decrease over time as a learner becomes more accustomed to using a given strategy. However, other factors may also affect the overall performance of a learner, and the call for future research in terms of motivation and emotion (Clark et al. 2006) is supported by the findings of this study. Finally, this research and other studies (Chi et al. 1994; Oxford 2003) suggest that rather than focusing on whether a strategy is used, it may be more beneficial to direct our attention to examining what strategies good language learners use and under what circumstances. 169

Olla Najah Al-Shalchi

As for pedagogical implications, the results of this study suggest that learners at a beginning level will benefit equally from receiving instructional guidance in using the keyword or context strategy, the intermediate learners will benefit the most from the keyword strategy, and the advanced learners will benefit the most from the context strategy. The expertise reversal effect may explain that the keyword strategy is best for lower proficiency levels and that the context strategy is best for more advanced learners. More advanced learners do not necessarily need extra instructional guidance because they already have schema built (Kalyuga et al. 2003). Furthermore, although in this study the Arabic 2 learners were labeled as having an intermediate level of proficiency, it may be that in reality they are true novices. There are no clear boundaries for when a novice learner becomes an expert; it is a continuous process (Kalyuga, personal communication, 21 March 2015). Because language learning is such a complex process, a great deal of instruction time is needed to allow a learner to reach an intermediate or advanced level. In language, subcategories (low, mid, and high) exist within one proficiency level. This may explain why the Arabic 2 learners seemed to benefit more from the keyword strategy. In regard to perceived workload, as was discussed earlier, no differences existed between the strategy and amount of workload, but upon further examination, not all groups had good performance scores. This may be due to other factors such as motivation and emotion, as Clark et al. (2006) has clarified that performance relies on more than the instruction and workload. By not having the vocabulary items directly linked to the current area of study, some participants may not have felt motivated to learn the new words. It is, therefore, suggested that the vocabulary be linked closely to the subject matter being learned so that the learner can clearly have a reason for learning the words. Lastly, with respect to the usage of the strategy, Arabic 1 keyword learners reported using the strategy more than the context group only in the cloze section, the Arabic 2 context learners reported using the strategy more often than the keyword group in both sections, and no differences were found among Arabic 3 learners. Consistent with other research, the more advanced learners are less likely to rely on memory strategies (Oxford 1990), which implies that language educators should begin with memory strategies at an early stage of language learning and then move into more inferential strategies as they advance in the language. Because language abilities differ and schema are being built, different strategy usage should be used with learners of various proficiency levels. Additional research should focus on what strategies successful learners use and when they use these strategies.

170

Keyword vs. context strategies

Appendix A: Training material for intermediate Arabic keyword mnemonic method One way of learning foreign language vocabulary is to use the keyword mnemonic method. To use this method, you need to think of a word in your native language that sounds similar to the word you’re trying to learn. Then create an image with the meaning of the words, and every time you hear the foreign language word think of that image. The more absurd the image is, the more likely you are to remember it. When you think of that image, it should help you remember the meaning of the word.

Let’s try an example: You are learning the names of different foods in Arabic and one of the words that you need to learn is ‫( سمك‬samak), fish. This word sounds similar to the English word smack. Now, create an image in which a fish is being used to smack someone. You may end up with something like this: Now every time you think of the word ‫سمك‬ (samak) think of the image of someone being smacked with a fish. This should help you remember the meaning of the word.

Let’s try another example. You are learning about different types of buildings, and one of the words that you need to learn is ‫موز‬ (moz), the Arabic for banana. This word sounds similar to Moses. Now, picture Moses dressed in a banana costume. Now it’s time to learn some vocabulary using this method. For the next week, you will be using this method to help you learn the vocabulary for this chapter. The vocabulary word and the word that it sounds similar to is written first. Then below that is a sentence using the meaning of the vocabulary word and the associated word. Each day you will review the words, and then you will create your own mnemonics for the remaining vocabulary.

The words that you will learn are: ‫نشأ‬

‫كلب‬

‫جرى‬

‫تم‬

‫انسان‬

171

‫ألف‬

‫حمل‬

‫القدس‬

‫وطن‬

Olla Najah Al-Shalchi

‫ – وطن‬What’s that hon? What’s that hon? It’s my homeland.

‫ – القدس‬Kudos There was a special kudos to the efforts of establishing peace in Jerusalem.

‫ – حمل‬Hamlet Hamlet carried Shakespeare on his shoulder.

‫ – ألف‬elephant The baby elephant weighs 1000 lbs!

‫ – انسان‬Insane He is one insane person!

‫ – تم‬Tim Tiny Tim completed his degree in Smallville.

172

Keyword vs. context strategies

‫ – جرى‬jar Hurry, run away from the jar!

‫ – نشأ‬Nashville He’s going to grow up to live in Nashville.

‫ – كلب‬kelp The dog got stuck in kelp.

Now in the space provided below, create your own mnemonics to help you in remembering the vocabulary. Create mnemonics for the following words. ‫ – اختيار‬choice ‫ – ما أحلى‬how beautiful . . . is ‫ – إلخ‬etc.

‫ – مثل‬like ‫ – طفل‬child ‫ – اُ ِخ َذ‬was taken

‫ – تاريخ‬date ‫ – سفارة‬embassy ‫ – طلب‬to request

Context Method We know that when learning/studying a foreign language, you are bound to come across words that you do not know the meaning of. When you read something in your native language, this is also sure to happen, yet you do not stop to look up every word in the dictionary. Many times, you subconsciously guess the meaning of the word based on the context of the sentence. Here, you’ll be doing the same thing. Let’s begin by looking at an example.

Example 1: Let’s say that the unknown word in these sentences is gather. Read the following sample sentences: 1 The students gather in the class every day. 2 During Thanksgiving, families gather for a feast. 3 The protesters will gather in front of this building at 8am and then go to city hall. 173

Olla Najah Al-Shalchi

After reading the sentences, are you able to define gather? You should be able to conclude that gather means a group of people coming together. Now, let’s try another example. Read the following sample sentences, and then see if you can define the highlighted word:

Example 2: 1 Tomorrow is Christmas; hence I will not be at work. 2 He has a paper due tomorrow; hence he will not watch television tonight. 3 I ate too much at dinner, hence I feel sick now. After reading the sentences, you should have been able to infer that hence means therefore. Now, we’ll use the same technique to learn some Arabic vocabulary. Read the following sentences and pay close attention to the highlighted word. Based on the sample sentences come up with a definition of the highlighted word. For the next week, you will be using this method to help you learn the vocabulary. Each day you’ll review the method, and then you will try to create your own examples using the remaining vocabulary. The words that you will learn are: ‫كلب‬

‫نشأ‬

‫جرى‬

.‫احيانا الفلسطينيون يشعرون أنه ليس لهم وطن‬ ‫لماذا لم يتم إصالح الحمام؟‬ !‫ال تمشي – اجري بسرعة‬ .‫نشأ محمد في مكة‬ .‫الكيلومتر ألف متر‬ ‫ماجد حمل الكمبيوتر معه ليكتب المالحظات من‬ .‫المحاضرة‬ ‫ارمسترونغ كان أول انسان ينزل على القمر‬ .‫في القدس يعيش المسلمون واليهود والمسيحيون‬ .‫يستطيع الكلب أن يعيش خارج البيت أو داخله‬

‫تم‬

‫انسان‬

‫ألف‬

‫حمل‬

‫القدس‬

‫وطن‬

.‫عشت كل حياتي هنا في وطني‬ .‫تم طبخ األكل قبل العشاء‬ .‫هوايته المفضلة الجري‬ ‫ولدت في كايفورنيا لكني نشأت في واشنطن‬ .‫دي سي‬ .‫كل شهر أدفع ألف دوالر لشقتي‬ .‫ندى حملت حقيبتها الكبيرة عندما سافرت‬

.‫ولدت في أمريكا وهذا وطني‬ .‫تم ترتيب البيت‬ .‫كل صباح أجري لنصف ساعة‬ .‫ولدت ونشأت في مدينة نيو يورك‬

.‫االنسان يحتفل بعيد ميالده مرة كل سنة‬ ‫كان يسكن في بيت لحم واآلن يسكن في‬ .‫القدس‬ .‫الكلب يحمي البيت‬

.‫االسان ال يحب أن يعيش وحيدا‬ .‫القدس عاصمة فلسطين‬

.‫في هذه الشركة الف موظف‬ .‫أحمل كتبي وأذهب إلى الصف‬

.‫الكلب أحسن صديق للرجل‬

Now using the space provided below create your own sentences with the remaining vocabulary. Pay special attention to context. ‫ – اختيار‬choice ‫ – ما أحلى‬how beautiful . . . is ‫ – إلخ‬etc.

‫ – مثل‬like ‫ – طفل‬child ‫ – اُ ِخ َذ‬was taken

174

‫ – تاريخ‬date ‫ – سفارة‬embassy ‫ – طلب‬to request

Keyword vs. context strategies

Appendix B: Modified NASA TLX

175

Olla Najah Al-Shalchi

Appendix C: Pre-test/post-test for the intermediate Arabic class Complete the following sentences using the words below. Some words will not be used. Do not use a word more than once. ‫ألف‬ ‫يصحو‬

‫ما احلى‬ ‫مثل‬

‫ولد‬ ‫يتكوّن‬

‫حمل‬ ‫تنتهي‬

‫أذكى‬ ‫ربما‬

.‫ هناك أكثر من __________ موظف يعمل في الشركة‬-1 .‫ في الصباح عادة ليس عندها وقت للفطور فـ _____________ قهوتها وتشربها بالسيارة‬-2 .‫ __________مدينتي! لن أعيش في أي مكان اخر‬-3 .‫ بيتنا ____________ من طابقين وأمام البيت حديقة‬-4 .‫ ال ___________ قبل الساعة السابعة‬-5 .‫ هؤالء الطالب ___________ طالب في صفنا‬-6 ‫ ريم صديقة مها وهي تدرس األدب ___________ والدتها‬-7 All of the time Some of the time Half of the time Less than half of None of the time the time For the first section on the test (fill in the blank), how often did you use the strategy you had training in?

Matching: 1. ‫انسان‬ 2. ‫اُ ِخ َذ‬ 3. ‫سفارة‬ 4. ‫نصيحة‬ 5. ‫فكرة‬ 6. ‫تم‬ 7. ‫أعمى‬

to complete was taken idea person blind embassy advice

________ ________ ________ ________ ________ ________ ________

All of the time Some of the time Half of the time Less than half of None of the time the time For the second section on the test (matching), how often did you use the strategy you had training in?

176

Keyword vs. context strategies

References Abd Ghani, K. and Zulkiply, N., 2008. Different memory techniques different memory needs: A case study for undergraduate science students. The International Journal of Learning, 15 (3), 203–210. Atkinson, R. C., and Raugh, M. R., 1975. An application of the mnemonic keyword method to the acquisition of a Russian vocabulary. Journal of Experimental Psychology: Human Learning and Memory, 104, 126–133. Atkinson, R., and Raugh, M. (1974) An application of the mnemonic keyword method to the acquisition of a Russian vocabulary. Available at: http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.398 .1795&rep=rep1&type=pdf (Accessed: 27 September 2009). Chamot, A. U., 2006. Issues in language learning strategy research and teaching. Electronic Journal of Foreign Language Teaching, 1 (1), 14–26. Cheng, E. K., 2011. The role of self-regulated learning in enhancing learning performance. International Journal of Research & Review, 6 (1), 1–16. Chi, M., et al., 1994. Eliciting self-explanations improves understanding. Cognitive Science, 18 (3), 439–477. Clark, R. E., et al., 2006. Motivational challenges experienced in highly complex learning environments. In: J. Elen and R. E. Clark, eds. Handling complexity in learning environments: Theory and research. Oxford: Elsevier, 27–42. Cooper, G. and Sweller, J., 1987. The effects of schema acquisition and rule automation on mathematical problem-solving transfer. Journal of Educational Psychology, 79, 347–362. Craik, F. I. M., and Tulving, E., 1975. Depth of processing and the retention of words in episodic memory. Journal of Experimental Psychology: General, 104, 268–294. DeCarrico, J. S., 2001. Vocabulary learning and teaching. In: M. Celce-Murcia, ed. Teaching English as a second or foreign language. Boston, MA: Heinle and Heinle, 285–299. Desrochers, A., et al., 1989. An application of the mnemonic keyword method to the acquisition of German nouns and their grammatical gender. Journal of Educational Psychology, 81, 25–32. Desrochers, A., et al., 1991. Instructional effects in the use of the mnemonic keyword method for learning German nouns and their grammatical gender. Applied Cognitive Psychology, 5 (1), 19–36. El Sawy, N., 2002. Languages: A fishy way to learn. Newsweek. [online] Available at: www.highbeam. com/doc/1G1-89922367.html [Accessed 27 September 2014]. Galy, E., et al., 2012. What is the relationship between mental workload factors and cognitive load types? International Journal Of Psychophysiology, 83, 269–275. Greenwood, S. C., 2002. Making words matter: Vocabulary study in the content areas. The Clearing House, 75 (5) 258–263. Gu, Y. and Johnson, R., 1996. Vocabulary learning strategies and languages learning outcomes. Language Learning, 46 (4), 643–679. Hall, J. W., 1988. On the utility of the keyword mnemonic for vocabulary learning. Journal of Educational Psychology, 80, 554–562. Hamada, M., and Koda, K., 2008. Influence of first language orthographic experience on second language decoding and word learning. Language Learning, 58 (1), 1–31. Hart, S. G. and Staveland, L. E., 1988. Development of NASA-TLX (Task Load Index): results of empirical and theoretical research. In: P. A. Hancock and N. Meshkati, eds. Human mental workload. Amsterdam: North Nolland Press, 139–183. Hsiao, T. and Oxford, R. L., 2002. Comparing theories of language learning strategies: A confirmatory factor analysis. The Modern Language Journal, 86 (3), 368–383. Hulstijin, J. H., 1992. Retention of inferred and given word meanings: Experiments in incidental vocabulary learning. In: P. J. L. Arnaud and H. Bejoint, eds., Vocabulary and applied linguistics. Basingstoke, UK: Palgrave Macmillan, 113–125. Kalyuga, S., 2007. Expertise reversal effect and its implications for learner-tailored instruction. Educational Psychology Review, 19 (4), 509–539. Kalyuga, S., et al., 2003. The expertise reversal effect. Educational Psychologist, 38 (1), 22–31. Kor, L. K. and Chuah, J. B., 2014. Cognitive workload and mathematics instructional design for nonusers of mathematical software applications. Electronic Journal of Mathematics and Technology, 8 (4), 274–285. Lee, H., et al., 2010. Improving self-regulation, learning strategy use, and achievement with metacognitive feedback. Educational Technology Research & Development, 58 (6), 629–648.

177

Olla Najah Al-Shalchi Levin, J. R., 1993. Mnemonic strategies and classroom learning: A twenty-year report card. The Elementary School Journal, 94 (2), 235–244. Liu, J., 2010. A  study on language learning strategies among the instructed EFL learners. US-China Foreign Language, 8 (3), 36–39. Mastropieri, M., et al., 1986. Direct vs. mnemonic instruction: Relative benefits for exceptional learners. Journal of Special Education, 20, 299–308. McDaniel, M. A., et al., 1987. Long-term retention of vocabulary after keyword and context learning. Journal of Educational Psychology, 79 (1), 87–89. Meissner, B., and Bogner, F. X., 2012. Science teaching based on cognitive load theory: Engaged students, but cognitive deficiencies. Studies in Educational Evaluation, 38, 127–134. Muljani, D. D., et al., 1998. The development of word recognition in a second language. Applied Psycholinguistics, 19 (1), 99–113. Nagy, W. and Scott, J., 2000. Vocabulary processes. In: M. L. Kamil, et al., eds., Handbook of reading research. Mahwah, NJ: Erlbaum, 269–284. Nash H. and Snowling M., 2006. Teaching new words to children with poor existing vocabulary knowledge: A controlled evaluation of the definition and context methods. International Journal of Language and Communication Disorders, 41 (3), 335–54. Oxford, R. L., 1990. Language learning strategies: What every teacher should know. New York: Newbury House. Oxford, R. (2003). Language learning styles and strategies: An overview. Proceedings of GALA (Generative Approaches to Language Acquisition) Conference, 1–25. Retrieved from http://web.ntpu.edu. tw/~language/workshop/read2.pdf Paas, F. A., et al., 2005. A motivational perspective on the relation between mental effort and performance: optimizing learner involvement in instruction. Educational Technology Research and Development, 53 (3), 25–34. Pressley, M., et al., 1980. The keyword method and foreign word acquisition. Journal of Experimental Psychology: Human Learning and Memory, 5, 22–29. Raugh, M. and Atkinson, R., 1974. A mnemonic method for the acquisition of a second-language vocabulary. Available at: http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.398.2086&rep=rep1&t ype=pdf [Accessed: 27 September 2009]. Redouane, R., 2010. Assessing instructional methods in L2 French vocabulary acquisition: guessingfrom-context method versus a word-list method. Annals of Spiru Haret University, Journalism Studies, 11, 73–87. Reed, S., 2006. Cognitive architectures for multimedia learning. Educational Psychologist, 41 (2), 87–98. Rodriguez, M. and Sadoski, M., 2000. Effects of rote, context, keyword, and context/keyword methods on retention of vocabulary in EFL classrooms. Language Learning, 50 (2), 385–412. Ryding, K., 2005. A reference grammar of Modern Standard Arabic. Cambridge: Cambridge University Press,. Sadeghi, K. and Nobakht, A., 2014. The effect of linguistic context on EFL vocabulary learning. GEMA: Online Journal of Language Studies, 14 (3), 65–82. Sagarra, N. and Alba, M., 2006. The key is in the keyword: L2 vocabulary learning methods with beginning learners of Spanish. Modern Language Journal, 90 (2), 228–243. Shaughnessy, M., 2003. An interview with Joel R. Levin. Educational Psychology Review, 15 (3), 297–309. Shih, J. L., et al., 2010. An inquiry-based mobile learning approach to enhancing social science learning effectiveness. Educational Technology and Society, 13 (4), 50–62. Teow, G., et al., 2010. Language learning strategies used by MFL students based on genders and achievement groups. US-China Foreign Language, 8 (1), 50–58. Van Hell, J. G. and Mahn, A., 1997. Keyword mnemonics versus rote rehearsal: learning concrete and abstract foreign words by experienced and inexperienced learners. Language Learning, 47 (3), 507–546. Wang, M., et al., 2003. Alphabetic and nonalphabetic L1 effects in English word identification: A comparison of Korean and Chinese English L2 learners. Cognition, 87 (2), 129–49. Zare, P., 2012. Language learning strategy use among EFL/ESL learners: A review of literature. International Journal of Humanities and Social Science, 2 (5), 162–169.

178

PART III

Arabic L2 morphosyntax

9 THE ACQUISITION OF RESUMPTIVE PRONOUNS How do second language learners of Arabic do it? Dola Algady This paper investigates the nature of interlanguage grammars of Arabic-English resumptive pronouns within relative clause constructions. It examines the role of economy conditions of the minimalist program (Chomsky, 1995) on syntactic derivations in accounting for the process of second language acquisition of Arabic resumptive pronouns. Based on syntactic differences and the cost of syntactic derivations of four different types of relative clauses varying by the extraction site – i.e., subject, direct object, indirect object, and oblique – I carried out a study with 16 adult English-speaking learners of Arabic. The participants completed three tasks: (1) a grammaticality judgement task, (2) a sentence combination task, and (3) a picture description task. The results of the study show that the distribution of resumptive pronouns within individual interlanguage relative clause constructions systematically varies depending on the extraction site which obeys the economy conditions of their syntactic derivation. Arabic and English differ with respect to the operations involved in the derivation of relative clauses, such that, while English uses move of an operator to generate relative clauses, in Arabic they are base-generated with a resumptive pronoun in the extraction site and are hence derived through (external) merge (Suaieh 1980, Alsayed 1998, Galal 2004). I show that the distribution of resumptive pronouns in the interlanguage grammar is constrained by the economy condition of merge-over-move and the shortest derivation requirement (Chomsky 1995), which are related to the number of the steps required in the syntactic derivation. I argue that these results suggest that a minimalist account can be implemented to specify what language features and operations are least accessible to the second language leaner. I conclude that the general principles of computational efficiency of syntactic derivations are operative on the process of second language acquisition.

9.1 Introduction The developments in syntactic theory under minimalism reconsider the relation between the language faculty and general cognitive systems whereby language acquisition is accomplished by the interaction of Chomsky’s (2005) three factors: (F1) a minimally specified UG (genetic endowment); (F2) primary linguistic data (PLD), i.e., input; and (F3) non-language facultyspecific considerations, including principles of efficient computation and principles of data 181

Dola Algady

analysis employed in acquisition. Based on this assumption, the current chapter examines the role of economy conditions of (F3) on syntactic derivations in accounting for the process of second language acquisition by investigating the nature of interlanguage (IL) grammars of Arabic-English resumptive pronouns (RPs) within relative clause (RC) constructions. RCs are complementizer phrases (CPs) embedded in a complex nominal expression (DP). Unlike English, Modern Standard Arabic (MSA) relatives contain an RP or in some cases a gap1 (Aoun et al. 2010) in the subordinate clause marking the initial position of the element that has been relativized 1(b). 1(a) [I know the girli [CP that Dani will meet __i]] English 1(b) [ʔa-ʕrifu 1.s-know

l-bintai the-girl

[CP llati daanii that

sa-yu-qaabilu-(haa)i]] MSA future-3.m.s-meet-her

“I know the girl that Dani will meet.”

Although it is more natural for RCs to be verb-initial (2), it is worth pointing out that in this study, the tasks construct contained subject-initial RCs2 (see Methods, Section 8.3). Examples of relatives’ extractions from subject (SU), direct object (DO), indirect object (IO), and oblique (OBL) positions, which are the focus of this empirical study, are provided in (2a–d), respectively. In the examples in (2), RPs are in bold; also note that in Arabic RPs are optional in the DO position 2(b) and are required in all other non-subject positions. 2(a) ṭurida expelled.3.m.s

l-waladu the-boy

llaðii that

mazzaqa tore.3.m.s

l-kitaaba the-book

“The boy that tore up the book was expelled.” (Aoun et al. 2010, p164)

2(b) ʔa-ʕrifu 1.s-know

l-mumaθθilata the-actress

llatii that

sa-yu-qaabilu-(haa) future-3.m.s-smeet-(her)

saamii Samy

“I know the actress that Sami will meet.”

2(c) žaaʔa came.3.m.s

t-tilmiiðu the student

llaðii that

ʔa-ʕṭaytu-hu 1.s-gave-him

l-kitaaba the-book

“The student to whom I gave the book came.”

2(d) qaraʔ-na read-1.p

l-kutuba the-books

llatii that

ʔaxbara-naa told.3.m.s-us

ʕan-haa about-them

kariimun Karim

“We read the books that Karim told us about.” (Aoun et al. 2010, p172)

Much of the existing typological research on the acquisition of RPs within RCs (e.g., Hyltenstam 1984, Pavesi 1986, Maghrabi 1997) reports the robust finding of IL grammars adhering to Keenan and Comrie’s (1977, 1979) accessibility hierarchy (AH) in (3). In 182

The acquisition of resumptive pronouns

particular, the observed performance is that when second language (L2) learners use RPs within RCs in higher positions (to the left) in the hierarchy, they also use them in all of the lower positions (to the right). (3) Keenan and Comrie’s accessibility hierarchy (AH): Subject > Direct Object > Indirect Object > Oblique > Genitive > Object of Comparison Hyltenstam’s (1984) data came from the acquisition of Swedish as an L2 by speakers of Spanish, Finnish, Greek, and Persian. Both Spanish and Finnish follow the same pattern found in Swedish of not supplying RPs, whereas Greek and Persian do supply them, although to somewhat different degrees. Results indicated that all participants produced RPs when producing RCs and that RPs were deleted roughly in the order predicted by the AH. These results led Hyltenstam to conclude that even though the learners have a marked category in their native language (i.e., the absence of RPs), they all use unmarked options in their IL grammar. In a replication of Hyltenstam (1984), Pavesi (1986) reported similar results. Her participants were Italian learners of L2 English. Despite the fact that none of their languages exhibit the RP phenomena in RCs, the IL grammars of these learners contained more instances of relatives with RPs than without. Moreover, Pavesi reported that the participants also adhered to the implicational pattern of the AH. Despite the fact that some of Hyltenstam’s (1984) results show a clear transfer effect, transfer does not explain why Spanish and Finnish learners used RPs in their IL. First language (L1) transfer does not explain Pavesi’s (1986) results either. In testing the predictions of the AH with respect to English-MSA ILs, Maghrabi (1997) reported conforming results. Maghrabi (1997) carried out a bi-directional study examining the acquisition of L2 RPs status of two groups of learners: L2 Arabic-speaking learners of English and L2 English-speaking learners of Arabic. The results of his study indicated that all participants used RPs in their use of L2 RCs more frequently. Maghrabi suggested that L2 learners retained RPs either because of L1 transfer and/or the incomprehensibility that learners encounter in processing L2 RCs without RPs. Similar to Hyltenstam (1984), Yuan and Zhao (2005) share the assumption that L1 influences the learner’s ability to acquire a new structure in the L2. They investigated whether Palestinian Arabic speakers who allow RPs in their L1 will acquire RPs in Chinese better than English speakers. Chinese is like Palestinian Arabic in that it does not allow wh-movement but does allow RPs. The results show that English-speaking participants outperformed Palestinian speakers in judging L2 Chinese RPs. Yuan and Zhao suggest that positive evidence from Chinese allows English-speaking learners to accumulate a superset of Chinese grammar that allows both RPs and gaps. They further suggest that overgeneralization of the use of RPs occurs in the case of the Palestinian Arabic-speaking participants because they perceive Chinese not to be typologically distinct from their L1, and this leads to learnability problems. The explanations for this AH effect come from processing considerations suggesting that humans operate with an economy strategy that leads to reduced time for processing (gap) when there is little cost (when the distance between the head noun and the gap is minimal) and an explicit strategy that leads to RPs in contexts of greater complexity (cf. Hawkins 1999, 2005). Some studies have investigated wh-movement and RPs within RCs. Shaheen (2013), for instance, investigated the acquisition of English RCs by Latakian Syrian-Arabic-speaking learners. Her results of RPs on the grammaticality judgement task revealed that learners showed more acceptance of resumption within embedded relatives than within simple ones. 183

Dola Algady

Moreover, learners rejected RPs but allowed them in lower positions of the AH (e.g., GEN). Shaheen offered two explanations. The first one is related to processing embedded RCs without RPs, given that embedded relatives are heavier to process than simple ones, so the participants relied on making a coreference explicit. The second explanation assumes that learners might not be sensitive to long distance movement (i.e., have not established the non-RP strategy yet). Hawkins and Chan (1997) and Bolotin (1995) have also investigated wh-movement and RPs. However, they presented contradictory results. Hawkins and Chan (1997) argued that Chinese-speaking learners of English are not able to acquire new features that are not instantiated in their L1. Their results show that Chinese speakers showed evidence for L1 transfer in their use of RPs with L2 English RCs. Their use of wh-pronouns along with RPs was explained by their L2 RCs derivation via base-generating wh-phrases in CP instead of moving an operator (OP) to [Spec, CP] bound by an overt RP. Bolotin’s (1995) study, on the other hand, contradicts these findings, because she argued that Arabic speakers are able to acquire wh-movement in English, even though Arabic arguably does not have wh-movement. Bolotin showed that her participants performed well when judging grammatical and ungrammatical wh-movement with and without RPs. Her results suggest that Arabic learners of English allow RPs, but they nevertheless know that gaps are possible in English RCs. According to the initial hypothesis of syntax (IHS, Platzack, 1996), unmarked features are present in a learners’ L1 (and L2/L3) initial state as the “default features.” Platzack suggests that the grammar of the human language is restricted by the economy principle because the child initially assumes the most economic form of syntax. With respect to second language acquisition, on the other hand, Platzack claims that the adult L2 learner, like the child acquiring L1, genetically prefers less costly structures (e.g., no overt syntactic movement), and both the L1 and the L2 initial stage will reflect this genetic or UG tendency. Take, for instance, syntactic movement involved in the derivation of wh-questions. In English, the wh-phrase overtly moves into [Spec, CP] because the Q feature in English is strong, whereas in languages where wh-phrases do not overtly move but rather remain in-situ, Q feature is assumed to be weak. Platzack believes that it is a reasonable claim if one considers that performing overt operations cost more than performing covert ones, or more than performing operations at all. One economy principle “Procrastinate” dictates that movement should be delayed as long as possible because movement costs. Similarly, Kim (2003) demonstrated that the Englishspeaking learners of Korean and Korean-speaking learners of English initially assume the most economical form of syntax, in which no overt syntactic movement (or no movement at all) is involved in the acquisition of wh-questions. Their results show that both groups of learners predominantly produced questions with the wh-phrase remaining in-situ. In this study I expand on Platzack’s (1996) view of markedness and claim that syntactically costly operations are more marked than their less costly counterparts. The following subsections examine how the different aspects of relativization are defined in terms of economy and how the cost of derivations can provide an account for IL grammars.

9.1.1  Merge-over-move and the acquisition of resumptive pronouns Of concern to this study is whether L2 learners derive RCs via an operation of merge or move and how such phenomena would be viewed under the economy conditions of the minimalist program (MP) (Chomsky 1995, 2000). In the matching analysis (Chomsky 1977, Sauerland 2000) adopted in this study, the head of the RC is base-generated outside of the RC.3 As mentioned earlier, RCs do not necessarily 184

The acquisition of resumptive pronouns

involve a gap, like English, but may also involve an RP. Unlike English, in MSA, RCs include an RP (overt or null, depending on the relativized position), coreferential with the modified DP. In the MSA examples 2(a–d), the resumptive element, appearing in bold, is coindexed with the modified DP. According to Sells (1984), the derivation with the pronoun does not involve movement of OP. Rather the OP is externally merged in [Spec, CP], binding the RP (cf. Shlonsky 1992, Galal 2004; see also Aoun et al. 2010 for an A’-movement account). I adopt the view that RCs which do not include an RP are derived through an OP movement 4(a). In the case where an RC involves an RP, the derivation of the RC does not involve movement 4(b). 4(a) [DP [D0 the book [CP OPi [C’ that [the girl reads ti]]]] 4(b) *[DP [D0 the book [CP OPi [C’ that [the girl reads iti]]]] Based on Galal’s (2004) minimalist analysis of the different types of RCs in MSA, the following subsection considers the syntactic derivation of the different types of RCs under investigation; i.e., SU, DO, IO, OBL as related to economy conditions of derivation. It is worth considering these syntactic derivations in much more detail, as they utilize the relevance of the minimalist framework to the analysis of the L2 data of this study.

9.1.2  Number of steps and the acquisition of resumptive pronouns Bearing in mind that the use of resumption is related to depth of embedding (e.g., Tsimpli 1997; Hawkins 1994, 1999, 2005), such that the further the gap (extraction site) from the filler (the relativized head DP), the more likely it is that learners will resort to resumption due to increase in processing load. My syntactic account from a minimalist perspective measures the probe-goal domain, i.e., the distance between the filler and the gap, in reference to the number of the derivational steps involved in the different RC types. The representations and diagrams here follow the phase impenetrability condition (which is a recast from Rizzi’s (1990) relativized minimality), according to which movement of a constituent out of a phase (as a syntactic domain) is only permitted if the constituent has first moved to the left edge of the phase (Chomsky 2000). By considering the number of steps involved in the derivation of RCs, I show that the derivation of OBL relatives is more costly than that of IO relatives, which are in turn more costly than DO relatives. Further, I show that SU RC is the most economical in terms of the number of steps of a derivation.

9.1.2.1  Direct object relative clauses The occurrence of RPs is in complementary distribution with the availability of movement (Suaieh 1980, Alsayed 1998, Galal 2004). Therefore, a definite DO relative in MSA involving a gap is derived by movement of an OP to [Spec, CP] which takes two steps; whereas in DO relatives involving RP, the OP is externally merged in [Spec, CP] and no movement is involved, as illustrated in the representation in 5(b). llaðii ṭ-ṭaalibu 5(a) qaraʔ-tu l-kitaaba read-1S the-book that.3.m.s the-student “I read the book that the student buys.” 185

ya-štarii-hu 3.m.s-buy-it

Dola Algady

5(b) DP the-book

CP C′

OP1

C TP that.3MS the-student

T′ vP

T

v′

2

VP v′

DP

v 3MS-buy

DP OP1/RP1

1

In the next subsections I present cases of relativization where resumption is syntactically required (i.e., IO and OBL relativization).

9.1.2.2  Indirect object relative clauses Following Larson’s (1988) approach to double-objects and Alsayed’s (1998) analysis of IO relatives in MSA, whereby the RP can be assigned case from the verb via the mediation of the null preposition, in 6(b), I show that in the case of resumption, no movement is involved and the OP is merged in its position [Spec, CP] in IO RCs. In cases where the derivation requires movement, then it will involve three steps, until the OP reaches the [Spec, CP] site. ṭ-ṭaaliba llaðii l-mudarrisatu tu-ʕṭii-hi l-žaaʔizata 6(a) raʔay-tu saw-1.s the-student that.3.m.s the-teacher 3.f.s-gives-him the-award “I met the student that the (female) teacher gives the award.” 6(b) CP OP1

C′

C

TP

that.3MS the-student.F

T

T′ vP v′

3

VP v′ v 3FS-gives

VP v′

DP v

2

v t1

DP the-award PP OP1/RP1

1

186

The acquisition of resumptive pronouns

In the next subsection, I present another case of obligatory resumption in MSA; namely, the case of OBL relativization.

9.1.2.3  Oblique relative clauses By adopting Galal’s (2004) analysis using a PP functional phrase (posited to enable feature checking in creating a Spec-head configuration), when an OBL RC 7(b) is derived by movement, the OP has to go through four steps to get to [Spec, CP], which is more costly when compared to a representation where the same syntactic structure is derived by freely merging the OP in [Spec, CP]. ṭ-ṭullaabu 7(a) ʔal-ʔustaaðu llaðii the-teacher that.3.m.s the-students “The teacher that the students are scared for.”

ya-xaaf-uuna 3.m.s-scare-pl

ʕalay-hi for-him

7(b) DP the-student.M

CP

OP1

C′

C TP that.3MS the-student.M T′ T

vP v′

4

VP V′ v 3M-scare-P

pP2 p′ p forj

3 2

pP1 PP P DP t1 OP1/RP1 1

The next subsection specifies an analysis (based on Galal 2004), which I adopt in this study for SU RCs in MSA based on principles of the economy of derivation and in equation to the analysis provided by Speas (1995) for the pro-drop phenomenon. 187

Dola Algady

9.1.2.4  Subject relative clauses The case of SU relatives in MSA is different from the other cases of DO, IO, and OBL. It is interesting because in SU RCs, it is obligatory to leave the subject position phonologically null.4 The analysis assumed by Galal (2004) (based on the pro-drop phenomena) is that MSA SU RCs are not generated by movement, but instead contain a base-generated null resumptive5 The following representation in 8(b) shows that deriving SU RCs via movement would cost one step, whereas their realization as a covert RP (PRO) is free of movement: 8(a) ʔaṭ-ṭaalibu llaðii ya-drusu the-student that.3.m.s 3.m.s-study “The student who is studying.” 8(b) DP CP

the-student OP1

C

that.3MS AgrP OP1/PRO Agr′ 3MS1

TP study

9.2  Research questions In light of the previous discussion, this study addresses the following questions: Question 1: Would individual IL grammars show an accessibility hierarchy (AH) effect in the distribution of resumptive pronouns within relative clause constructions? Question 2: Would adult interlanguage grammar show a preference toward deriving relative clauses through less costly structures by retaining/supplying resumptive pronouns within relative clause constructions? Question 3: Would IL data show a systematicity in the distribution of resumptive pronouns with regard to the number of steps required in the derivation of the relative clause?

9.3 Methods 9.3.1 Participants Study participants consisted of 16 L2 adult learners of MSA and one control group of five Arabic-speaking adults, all of whom were students at a Midwest university in the USA. All L2 participants were native speakers of English studying fourth-semester Arabic (five credit hours each), which insured that their Arabic was advanced enough to participate in the study. Moreover, a more reliable method of assessing Arabic proficiency (i.e., ACTFL OPI) was used in order to classify participants according to their level of MSA proficiency. 188

The acquisition of resumptive pronouns

All of the control participants were undergraduate students of business, except for C4, who held a master’s degree in Arabic from Saudi Arabia while pursuing his MA in linguistics. Most of the L2 participants had been studying MSA for almost two years, did not have experience living in an Arabic-speaking country and were not related to any Arabic native speakers. However, there were five exceptions: (P16) was related to Arabic native speakers by family ties and lived in Palestine for five years, (P9) studied Arabic in Egypt for a year; (P10) traveled to Iraq for eight months, (P7) spent two weeks in Morocco, and (P13) spent two summers in Egypt, studying Arabic. However, these participants’ accuracy rate in the grammaticality judgement task fell within the range of other participants. Detailed information about the participants is given in Table 9.1.

9.3.2  Data collection procedures The data of this study was collected through grammaticality judgements and elicited written controlled production tasks. All participants involved in the study completed three tasks: (1) a grammaticality judgements task (GJT), (2) a sentence combination task (SCT), and (3) a picture description task (PDT). The participants completed the tasks separately within two experimental sessions. They completed the GJT during the first session and the controlled production tasks during the second session, starting with the SCT and followed by the PDT. The instructions for the tasks were given in English, and English translation of any potentially difficult words was provided. Participants were given practice before the start of the actual task. Participants were also given as much time as they needed to complete the tasks, which lasted for one to two hours on average. Table 9.1 Information about the participants Participant/ID

Age

Gender

Native Language(s)

Arabic Proficiency

C1 C2 C3 C4 C5 P1 P2 P3 P4 P5 P6 P7 P8 P9 P10 P11 P12 P13 P14 P15 P16

25 22 19 30 26 20 20 22 26 36 20 25 20 23 36 20 22 55 23 19 NA

F F M M F M M M M M M F F M M M F F F F M

Saudi Arabic Saudi Arabic Saudi Arabic Saudi Arabic Saudi Arabic English English English English English/Spanish/Polish English English English/Bosnian/German English English English English/Urdu English English English English/Palestinian Arabic

Superior Superior Superior Superior Superior IntermediateM IntermediateM IntermediateM IntermediateM IntermediateM IntermediateH NoviceH IntermediateH IntermediateH NoviceH IntermediateL IntermediateH IntermediateH IntermediateM IntermediateL AdvancedL

M  =  Male, F  =  Female, NoviceH  =  Novice-high, IntermediateL  =  Intermediate-low, IntermediateM = Intermediate-mid, IntermediateH = Intermediate-high, AdvancedL =Advanced-low

189

Dola Algady

All of the tasks were composed by the researcher. Before proceeding to the description of the tasks, a number of general principles in their design are outlined here (for the full content of all tasks, see Algady 2013). The items were constructed on the basis of a number of points:  1 The test properties are comparable to those in the other tasks.  2 The vocabulary is accessible to less proficient speakers.  3 All the sentences used had the same tense (simple present).  4 Only animate/human relative heads were used, belonging to beginning-level vocabulary.6  5 All relative clauses were subject-initial, which is more accessible to the L2 learners than verb-initial.  6 All sentences within a given relative clause type were balanced for the number of words.  7 Because the length of a clause may affect relativization (Lee, 2006), the sentences being relativized were kept relatively short.  8 Humanness and definiteness of the head noun were controlled to be human and definite.  9 All relative clauses had lexical DPs as subjects rather than pronominal subjects. 10 All instances of relative clauses are right embedded. 11 All instances of relative clauses include single embedding.

9.3.2.1  Grammaticality judgement task The GJT included 40 items, ten for each of the attested RC types. Distractors were not used due to the already lengthy tasks, and the RC structures are already varied in number and gender of RPs, thus serving the effect of a distractor (Gass and Mackey 2007). Restrictive RC tokens involving manipulations in the use of resumption were included to test participants’ knowledge of the distribution of these forms. Therefore, ungrammatical sentences involving RPs violations were included. This was done in order to test whether the participants are sensitive to such violations. Participants were required to judge the grammaticality of the sentence and were given a grammatical and an ungrammatical choice. Where participants rated the sentence as ungrammatical, they were instructed to underline the part of the sentence that made the sentence ungrammatical and to provide a correction, as in (9). All sentences were presented in Arabic script. The English is provided here for additional illustration only (for a list of all sentences used in the study, see Algady 2013).

(9)   .‫أعرف النساء اللواتي األستاذ يتحدث مع‬ *ʔa-ʕrifu n-nisaaʔa llawaatii 1.s-know the-ladies that “I know the ladies that the teacher is talking to.”

l-ʔustaaðu the-teacher

ya-taħaddaθu talks

maʕa with

9.3.2.2  Sentence combination task The SCT allowed for the production of RPs within the four types of RCs. The task included 28 pairs of sentences, seven of each type of the attested RC extractions. Participants were instructed to complete the task in writing, in which they had to combine each pair of the given sentences by attaching the second sentence (e.g., 10(b)) to the first sentence (e.g., 10(a)) using relative complementizers of ʔallaðii-type (as in the answer provided here for illustration). The instructions indicated to always start with the first sentence. All sentences were presented in 190

The acquisition of resumptive pronouns

Arabic script. The English is provided for illustration here (for a list of all sentences used in the study, see Algady 2013). 10(b)

.‫المريض يعتمد على الممرض‬ ʔa l-mariiḍu ya-ʕtamidu ʕala l-mumarriḍi the-patient 3.m.s-relies on the-nurse “The patient relies on the nurse.”

10(a)

.‫قابلت الممرض‬ qaabal-tu l-mumarriḍa met-1.s the-nurse “I met the nurse.”  . . . ‫قابلت الممرض‬ Answer: qaabaltu l-mumarriḍa. . .

9.3.2.3  Picture description task In the PDT, the four types of RCs were tested with six pairs of pictures for each site (with a total of 24 pairs). Each pair of pictures consisted of one picture (in the background) showing one or more characters engaged in an activity and a descriptive full sentence, placed on top of the picture in Arabic script, stating what the characters are doing. The other picture (in the foreground) presented the same character/s (in the background picture) without showing them engaged in the activity in the background picture. The participants were instructed to read the description of the first picture, then describe in writing the character(s) in the other picture. The participants were given a question followed by a prompt, as illustrated in (11). The English equivalent is provided for illustration (for the full content of this task, see Algady 2013). (11)

man haaðaa? “who is this?” haaðaa huwa. . . “This is . . .”

‫من هذا ؟‬  . . . ‫هذا هو‬

9.3.3  Data analysis Implicational scales have been commonly used within SLA research (Hyltenstam 1984, Andersen 1978, Pienemann 2011, Spinner 2011). According to Andersen (1978), in an implicational scaling analysis method, data can be displayed with regard to variability and systematicity, and the use of a certain structure implies that a structure from a previous stage is a pre-requisite for the following but not the reverse. Most implicational analyses are based on binary judgements; either a presence or a non-presence of a given feature marked by a plus or a minus. One column is used for each category, and then the rows and columns are shifted in order to find out whether there is a scale pattern. Hyltenstam (1984) used implicational scaling analysis on his data, and from the tables he was able to check for individual variable behavior and to conclude that the learners used a regular route in their acquisition process. Table 9.2 shows an example of implicational scaling, in accordance with Hyltenstam (1984), of structures 1, 2, 3, 4, 5, 6 by participants P1, P2, P3, P4, P5, P6. An example of a “+” indicates the occurrence of the structure(s) by a specific participant, while “-” shows the absence of the structure. The scale in Table 9.2 shows a pattern where the presence of a variable to the left implies the presence of a variable to the right. No participant uses a variable to the left without also using it to the right except P4. We can find a marginal line here to show deviations from the perfect scale. Such a display allows us to examine individual performance and the group as a whole, as it reveals directionality of acquisition. 191

Dola Algady Table 9.2 An example of implicational scaling Participants

1

2

3

4

5

6

P3 P4 P6 P1 P5 P2

-

+

+ +

+ + +

+ + + + +

+ + + +

9.3.4  Data coding Only aspects relevant to the structure of RP status were considered in the analysis of the data. In this study the researcher used implicational scaling analysis to RP retention within the different types of RCs. A specific type of RC is given “+” if it meets a criterion of supplying/ retaining RPs 70% of the time7 and “-” if it is supplied less. Implicational Tables 9.2–9.8 were prepared where the participants are rank ordered. Participants were considered as allowing for optionality of RPs “+/-” if they accept both gaps and RPs 50% of the time.

9.4 Results The results of the study are represented in Tables 9.3–9.8, using implicational scales and following the AH scalability representation. The top row of all of the tables lists the RC types according to their extraction site. The left-most column in each table represents the participants of this study, and the cells in the row for each participant are the results describing the participant’s performance. The data is not necessarily ordered on the basis of the participant’s number; rather, the participants are sorted according to their fit within the AH scalability. Under each RC type (SU, DO, IO, and OBL) column, the status of RPs is indicated by “+/-”; where a “+” marks the retention/production of RPs, a “-” indicates the deletion/no production of RPs, and “+/-” signifies allowing for both gaps and RPs.

9.4.1  Results of GJT Table 9.3 shows the results obtained from the control participants whereby all participants show rejection of resumption within SU RCs, accept both the resumption and gap options equally within DO RCs, and retain resumption/reject gaps within both the IO and OBL RCs. The L2 participants’ results are presented in Table 9.4. No results for P10 are presented in the table because he did not complete the task. The “+/-” indicates acceptance of both gaps and RPs options equally by some participants. The results show an implicational pattern; when L2 participants rejected gaps and/or retained RPs within RCs on higher positions of the AH, they did the same for RCs on lower positions. Recall that, according to the AH, SU is the highest and OBL is the lowest. It also seems that none of the L2 participants rejected RPs in all types of RCs, and thus the results of this specific pattern in the use of RPs shows no L1 transfer effect (-RP in all positions). 192

The acquisition of resumptive pronouns Table 9.3 GJT results for controls Participants

SU

DO

IO

OBL

C1 C2 C3 C4 C5

-

+/+/+/+/+/-

+ + + + +

+ + + + +

Table 9.4 GJT results for L2 participants Participants

SU

DO

IO

OBL

P11 P12 P14 P5 P6 P7 P8 P15 P9 P2 P3 P4 P1 P13 P16

+/+/+/+/+/+/+/+/+/-

+/+/+/+/+/+/+/+/+ +/+ + + + +

+/+/+/+/+/+/+/+ + + +/+/+ + +

+/+/+/+ + + + + + + + + + + +

9.4.2  Results of SCT Table 9.5 shows that the L1 controls used RPs within all non-SU RCs. Despite the fact that the control participants showed such awareness of this structure on the GJT, when they were asked for RC productions, those same participants showed a preference toward using RPs with DO relatives over gaps. As for the L2 learners’ performance, Table 9.6 shows that most of the participants used RPs over gaps in DO (10 out of 15 participants). None of the participants provided RPs in the SU position, nor did they show a native-like pattern resembling the L1 setting of gap use in all relativization positions. Although Table 9.6 shows that P11 did not use RPs or RDPs in all positions, the raw results indicate that he or she has not used gaps all the way as dictated by their L1 rules. The analyses also reveals that eight of the participants (P3, P5, P6, P7, P10, P8, P9, and P14) neither used gaps nor RPs; instead, they retained the RDPs in the extraction site to a varying extent. For the most part, when participants retained RDPs within RCs on higher positions on the scale, they also, with a few exceptions (P3, P10), retained them on the lower positions. Six participants (P1, P4, P8, P9, P12, P13, P14, and P15) showed a preference toward using RPs within all non-SU relatives. P2 and P3 are exceptions, and no results are reported here for P16, because he did not complete the task. 193

Dola Algady Table 9.5 SCT results for controls Participants

SU

DO

IO

OBL

C1 C2 C3 C4 C5

-

+ + + + +

+ + + + +

+ + + + +

Table 9.6 SCT results for L2 participants Participants

SU

DO

IO

OBL

P11 P6 P10 P2 P1 P4 P8 P12 P13 P15 P9 P14 P3 P7 P5

RDP RDP RDP

+ + + + + + + + + + RDP RDP

+ + + + + + +/+ + RDP RDP

+/RDP +/RDP + + + + + + + +/RDP RDP RDP RDP

Note: highlighted entries indicate deviations from the perfect implicational scale in all tables throughout the paper; RDP = resumptive DP

9.4.3  Results of PDT Table 9.7 reports the results obtained from the L1 control participants, showing that in the PDT, similar to the SCT results, the L1 controls used RPs in accord with the AH. The five control participants used RPs within all non-SU relative clauses 100% of the time. Interestingly, though, similar to the findings reported for the SCT, no optional use of resumption was noted. Similar to the native control group, Table 9.8 indicates that 10 out of the 16 L2 participants produced RCs with RPs at the extraction site of all non-SU positions (this however, does not necessarily entail that they produced resumption 100% of the time). Only five participants produced RPs all the time within all of the non-SU positions. P2, P6, P7, P10, and P11 showed a greater tendency to use resumption within RC types on higher positions when they also used them in all lower positions. None of the L2 participants behaved in an English-like manner; i.e., none of them only allowed gaps and no resumptives in all RC types all the time. Again, none of the L2 participants or the L1 controls used overt RPs in the SU RC type. 194

The acquisition of resumptive pronouns Table 9.7 PDT results for the controls Participants

SU

DO

IO

OBL

C1 C2 C3 C4 C5

-

+ + + + +

+ + + + +

+ + + + +

Table 9.8 PDT results for L2 participants Participants

SU

DO

IO

OBL

P5 P10 P7 P2 P6 P11 P1 P3 P4 P8 P9 P12 P13 P14 P15 P16

-

+ + + + + + + + + + +

+ + + + + + + + + + + + + +

RDP + + + + + + + + + + + + + + +

RDP = resumptive DP

9.5 Discussion Similar to Hyltenstam (1984), Pavesi (1986), Maghrabi (1997), and Shaheen (2013), an AH effect in the distribution of RPs is clear from the results (Question 1). More specifically, the results showed that RPs appear in a particular position only if they also appear in all other lower positions in the hierarchy, with SU being the lowest and OBL the highest. Deviations from the implicational scale pattern were minor and only appeared in the production tasks. P2 and P5 provided RPs in DO positions but not IO positions. Moreover, both L1 control participants and L2 participants showed awareness of RP optionality in DO relatives in the GJT; however, they produced DO RCs with RPs rather than gaps in both the SCT and PDT. We can also notice the use of RDPs by L2 participants in the production tasks. Use of RDPs is often referred to as a familiar error in the formation of RCs by L2 learners (Bshara et al. 2013). In this study, RDPs are viewed on a par with RPs.8 In the following subsections, I examine these findings with respect to the predictions of the economy conditions of the MP (Chomsky 1995). As noted in Kim (2003), human beings want to achieve a maximum of effects at a minimum of effort, and this principle also applies to the grammar of the language. Thus, the prediction is that individual IL grammars, like natural 195

Dola Algady

languages (Adjémian 1976), conform to the economy conditions on syntactic derivations. Findings relevant to the study questions are discussed here with reference to individual IL grammars obeying the economy conditions of merge-over-move and the shortest derivation requirement.

9.5.1  AH effect In considering the syntactic derivation of RCs, I pointed out that under the matching/adjunction analysis (Chomsky 1977, 1995), the head is base-generated and the relative clause adjoins to the maximal projection NP/DP. Further, the OP moves to [Spec, CP] in RC structures where movement is involved. OP, however, is directly merged in [Spec, CP] in the cases which do not involve movement, where only resumption is allowed (Sells 1984, Shlonsky 1992). Given that, I assume that the order in the acquisition of RPs within RCs is rather dependent on computational considerations relevant to economy conditions on syntactic derivation. First, “move” is a more costly operation than “merge” (as dictated in the merge-over-move economy principle). Second, the different types of RCs under investigation involve different numbers of steps assuming the movement derivation, as showed earlier; an OBL relative involves four steps, an IO relative involves three steps, and a DO relative involves two steps, and finally an SU relative involves one step. If IL grammars allow resumption rather than gaps in a given relativization position, then this would indicate that the participants derived RC through OP “merge” rather than “move” (Galal 2004, Shlonsky 1992, Sells 1984, Bolotin 1995). If participants, however, allowed gaps, then this is evidence that they derived RC through OP movement (Bolotin 1995, Galal 2004, Bshara et al. 2013). Regarding the third question of the study, the results indicated a systematic pattern in the distribution of RPs with the different types of RCs, in that the more costly movement becomes, involving more steps, the more likely learners would resort to a merge of OP rather than moving it. Individual IL grammars are in accord with minimality in that their representations of the different types of RCs follow the shortest derivation requirement, as related to the cost imposed by the number of steps involved in the derivation. In other words, in accord with the shortest derivation requirement, the more costly movement becomes, involving more steps, the more likely learners would resort to minimality, preferring less costly operations and choosing to merge the OP rather than moving it.

9.5.2  RP optionality in DO The results indicated that the controls behaved differently on the GJT compared to the production tasks. The main difference in the results between the two types of tasks included more instances of accepting both gaps and resumptions (i.e., optionality) on DO RCs on the GJT. As for the L2 participants, there were more instances of allowing for optionality on the GJT as well. However, the analyses did not reveal the same results on the written production tasks. With respect to the GJT result, we can tell that both control and L2 participants were aware that there are two types of DO relatives, with and without RP in MSA, and that they (based on Galal’s (2004) assumptions) emanate from two different numerations that are sensitive to the semantics of the DP, whereby the non-presence of an RP is likely to create some degree of ambiguity. According to the economy condition of “merge-over move,” move is a more expensive operation than merge, and derivational steps – at which both operations are possible – are required to

196

The acquisition of resumptive pronouns

choose the “cheaper” merge operation as the next step (Chomsky 1995). As pointed out earlier there are two possible ways of deriving RCs (4a–b), one with RPs in the extraction site where no movement is involved and the other involves movement and a trace in the extraction site. Platzack (1996) claims that language learners genetically prefer less costly structures. Carrying on with the assumption that RCs which involve RPs do not involve movement (Sells 1984, Bolotin 1995, Hawkins and Chan 1997, Yuan and Zhao 2005), I argue that the participants of this study resorted to the less costly operation “merge” with RP in their production of DO RCs (Question 2).

9.5.3  DO and IO discrepancy What the results showed most clearly is that IL grammars showed an AH effect. Deviations from the perfect scale were minor and only appeared in the production tasks. P2 and P5 provided RPs in DO positions but not IO positions. Their results are reproduced in Table 9.9. One possible explanation for this acquisition pattern could be that those participants’ syntactic derivation of IO relatives is different from the derivation represented in 6(b). An alternative analysis, like the one in 12(b), involves two derivational steps rather than three, and thus equates the cost of IO relatives with DO relatives. 12(a) ʔaṭ-ṭaalibu llaðii l-ʔustaðatu manaħ-at-hu l-žaaʔizata that.3m.s the-teacher granted-3.f.s-him the-award the-student “The student that the (female) teacher granted the award to.”

12(b) DP the-student.M OP1

CP C′ C TP that.3MS the-teacher.F T′ T granted.3FSk DP 2

vP vP vP the-awardj

v′

v tk

VP DP ti

V′ V tk

1

197

DP tj

Dola Algady Table 9.9 L2 Participants inconsistent with the hierarchy pattern Participants

SU

DO

IO

OBL

P2 P5

-

+ +

-

+ RDP

It is possible that these L2 participants treated DO and IO equally, and thus their IL pattern is consistent with the economy conditions, such that when they allowed for “merge” within IO/ DO, they did the same with OBL but not with RCs in higher positions in the hierarchy (i.e., SU > DO = IO > OBL).

9.6  Conclusion and pedagogical implications We conclude that the general principles of computational efficiency of syntactic deviations are operative on the process of second language acquisition and that a minimalist account can be implemented to specify what language operations are least accessible to the learner. In general, the results of the study showed that the economy principle is operative in the process of L2 acquisition. IL grammars have not reflected the L1 value of using gaps all the time within RCs, nor the L2 value. Instead, there was a general tendency whereby when resumption was used in the less complex RC types, it was also used in all more complex ones (complexity here indicates the number of the steps involved in OP movement). In many aspects the results were consistent with the prediction that individual IL grammars are in accord with minimality, preferring less costly operations and choosing to merge the OP rather than moving it. One case that is worth further investigations is that of SU relatives. More conclusive evidence to whether the interlanguage subject relatives were derived through merge of a null operator and a null resumptive pronoun or move of a null operator could be achieved by including test items which contain islands within the RC. In fact, including test items that contain islands should strengthen the general claims of the current study even within object relative clause structures. Furthermore, the current study showed that an account based on the minimalist theory could provide an explanation for interlanguage grammatical patterns. We need to investigate whether the findings of this study can be generalized to account for other constructions in L2 Arabic, e.g., questions. In the application of the generalization of RC instruction, learners who are taught some types of RC can generalize them to the others which they were not taught (e.g., Eckman 1988). Particularly, when learners are trained on RC structures lower in the AH, they performed better with the structure higher up in the hierarchy. This suggests that it is the most marked aspects of a target language from which it should be possible for a learner to gain maximal generalization of their learning. The results of this study seem to also support this hypothesis. However, the direction is the opposite if we are talking about resumption use within RCs, i.e., OBL > IO > DO > SU. Thus, if one were forced to choose only one RP structure to teach, the structure must be taught through the more marked one, i.e., DO RCs.

Notes 1 Both Galal (2004) and Aoun et al. (2010) share the position that subject relatives only allow a gap in the relativized subject position. However, it is worth mentioning that there are cases where a strong pronoun can occur in a subject relativized position when the subject is in contrastive focus. See

198

The acquisition of resumptive pronouns Shaheen (2013) for data from Syrian Arabic ‫ المؤلف الذي هو كتب الكتاب‬and data from Classical Qur’anic Arabic verses such as [Qur’an 26:79] ‫والذي هو يطعمني ويسقين‬. 2 In fact, in MSA examples for RCs cited in Aoun et al. (2010), Galal (2004), and Mohammad (2000), the predominant order is verb initial. However, Alqurashi and Borsley (2012, p. 28) pointed out that “[b]oth definite and indefinite relatives are normally verb initial, but subject-initial sometimes occur.” 3 The debate is, however, very controversial on how relative clauses are syntactically represented (e.g., Chomsky, 1965, 1977, Kayne, 1994, Aoun and Li, 2003, Galal, 2004). Within generative grammar, different hypotheses have been proposed for the formal structure of relative clauses across languages. 4 See Footnote 1 for cases where a strong pronoun can occur in a subject relativized position. 5 See Suaieh (1980) for a non-movement solution in which the subject pronoun is assumed to delete, and Shlonsky (1992) for a movement analysis. 6 According to Adani (2011), animate/inanimate referents seem to play different roles in both adult and children studies. Hence, unless animacy is taken as an experimental factor, it is necessary to control for these effects. 7 It is hard to decide on a clear-cut point for whether that acquisition has occurred. However, in this study I am following the 70%–80% criteria, as it has been used in the acquisition literature. 8 Bshara et al. (2013) analyze use of RDPs as movement of the relative operator without the copy of the relative head.

References Adani, F., 2011. Rethinking the acquisition of relative clauses in Italian: Towards a grammatically based account. Child Language, 38, 141–165. Adjémian, C., 1976. On the nature of interlanguage systems. Language Learning, 26 (2), 297–320. Algady, D., 2013. The acquisition of relative clauses: How do second language learners of Arabic do it? Thesis (PhD). University of Wisconsin-Milwaukee. Alqurashi, A. and Borsley R. D., 2012. Arabic relative clauses in HPSG. Paper presented at the Proceedings of the 19th International Conference on Head-Driven Phrase Structure Grammar, Chungnam National University Daejeon. Stanford, CA: CSLI Publications. Alsayed, A., 1998. A government-binding approach to restrictive relatives, with particular reference to restrictive relatives in Standard Arabic. Thesis (PhD). University of Essex. Andersen, R. W., 1978. Implicational model for second language research. Language Learning, 28, 221–281. Aoun, J., et al., 2010. The syntax of Arabic. Cambridge: Cambridge University Press. Aoun, J. and Li, Y.A., 2003. Essays on the representational and derivational nature of grammar: The diversity of Wh-constructions. Cambridge MA: MIT press. Bolotin, N. A., 1995. The effect of age on parameter resetting: Arabic-speaking children and adults’ acquisition of English relative clause structure as assessed through interlanguages data and a universal constraint. Thesis (PhD). Harvard University. Bshara, R., et al., 2013. (Unique) errors in the acquisition of relative clauses in Palestinian Arabic and their (movement) account. In: I. Chohen, ed. Proceedings of IATL 2012, Oct 25, 26–2012, Hebrew University of Jerusalem. MITWPL #68. 25–38. Chomsky, N. 1965. Aspects of the theory of syntax. Cambridge, MA: MIT Press. Chomsky, N., 1977. On wh-movement. In: A. Akmajian, et al., eds. Formal syntax. New York: Academic Press, 71–132. Chomsky, N., 1995. The minimalist program. Cambridge MA: MIT Press. Chomsky, N., 2000. Minimalist inquiries: the framework. In: R. Martin, et al., eds. Step by step: Essays on minimalist syntax in honor of Howard Lasnik. Cambridge MA: MIT Press. 89–155. Chomsky, N., 2005. Three factors in language design. Linguistic Inquiry, 36, 1–22. Eckman, F., 1988. On the generalization of relative clause instruction in the acquisition of English as a second language. Applied Linguistics, 9, 1–20. Galal, M.M., 2004. A minimalist approach o relative clauses in Modern Standard Arabic. Thesis (PhD). University of Kansas. Gass, S. M., and Mackey, A., 2007. Data elicitation for second and foreign language research. Mahwah, N.J.: Lawrence Erlbaum Associates. Hawkins, J., 1994. A performance theory of order and constituency. Cambridge: Cambridge University Press.

199

Dola Algady Hawkins, J., 2005. Efficiency and complexity in grammars. Oxford: Oxford University Press. Hawkins, J.A., 1999. Processing complexity and filler-gap dependencies across grammar. Language, 75 (2), 244–285. Hawkins, R., and Chan, C., 1997. The partial availability of universal grammar in second language acquisition: The “failed functional features hypothesis.” Second Language Research, 13, 187–226. Hyltenstam, K., 1984. The use of typological markedness conditions as predictors in second language acquisition: the case of pronominal copies in relative clauses. In: R. Andersen, ed. Second languages: A cross-linguistic perspective Rowley, MA: Newbury House, 39–58. Kayne, R. S., 1994. The antisymmetry of syntax. Cambridge, MIT Press: Mass. Keenan, E. L. and Comrie, B., 1977. Noun phrase accessibility and universal grammar. Linguistic Inquiry, 8 (1), 63–99. Keenan, E. L and Comrie, B., 1979. Noun phrase accessibility revisited. Language, 55 (3), 649–664. Kim, J. T., 2003. L2 initial syntax: Wh-movement and the most economical syntactic derivation. In: B. Barbara Beachley et al., BUCLD Proceedings, 27 (1), 414–424. Larson, R., 1988. On the double object construction. Linguistic Inquiry, 19, 335–391. Lee, I., 2006. Development of morphosyntactic features of the complementizer phrase in L2 relative clauses of Korean learners. Thesis (PhD). Michigan State University. Maghrabi, A. A., 1997. The roles of psycholinguistic constraints and typological influence in the acquisition of pronominal copies in relativization by Arabic and English learners. Thesis (PhD). Georgetown University. Mohammad, A. M., 2000. Word order, agreement and pronominalization in standard and Palestinian Arabic. Amsterdam: John Benjamins. Pavesi, M., 1986. Markedness, discoursal modes, and relative clause formation in a formal and an informal context. Studies in Second Language Acquisition, 8, 38–55. Pienemann, M., 2011. Explaining developmental schedules. In: M. Pienemann and J. Keßler, eds. Studying processability theory; An introductory textbook. Amsterdam: John Benjamins, 50–63. Platzack, C., 1996. The initial hypothesis of syntax: a minimal perspective on language acquisition and attrition. In: H. Clahsen, ed. Generative perspectives on language acquisition: empirical findings, Theoretical considerations and crosslinguistic comparisons. Amsterdam: John Benjamins, 369–414. Rizzi, L., 1990. Relativized minimality. Cambridge MA: MIT Press. Sauerland, U., 2000. Two structures for English restrictive relative clauses. In: M. Saito et al., eds. Proceedings of the Nazan GLOW. Nagoya, Japan: Nazan University, 351–366. Sells, P., 1984. Syntax and semantics of resumptive pronouns. Thesis (PhD). University of Massachusetts. Shaheen, B., 2013. A comparative study of restrictive relative clauses in Latakian Syrian Arabic and English and the acquisition of English restrictive relative clauses by first language speakers of Latakian Syrian Arabic. Thesis (PhD). University of Essex, UK. Shlonsky, U., 1992. Resumptive pronouns as a last resort. Linguistic Inquiry, 23 (3), 443 468. Speas, M., 1995. Economy, agreement and representation of null arguments. Thesis (MA). University of Massachusetts at Amherst. Spinner, P., 2011. Second language assessment and morphosyntactic development. Studies in Second Language Acquisition, 33, 529–561. Suaieh, S., 1980. Aspects of Arabic relative clauses: A study of the structure of relative clauses in Modern Written Arabic. Thesis (PhD). University of Indiana. Tsimpli, I. M., 1997. Resumptive strategies and L2A: A minimalist account. In: E. Hughes, et al., eds., Proceedings of the 21st annual Boston University Conference on Language Development, BUCLD 21 (2), 639–655. Yuan, B. and Zhao, Y., 2005. Resumptive pronouns in English-Chinese and Arabic-Chinese interlanguages. International Review of Applied Linguistics in Language Teaching, 43, 219–237.

200

10 ARABIC L2 LEARNERS’ USE OF WORD ORDER AND SUBJECTVERB AGREEMENT FOR ACTOR ROLE ASSIGNMENT Jamil Al-Thawahrih Studies have shown that English speakers use first noun strategy in NVN order to interpret sentences as SVO and that they use second noun strategy in VNN and NNV in order to interpret sentences as VOS and OSV, respectively (MacWhinney et al. 1984). In contrast, Taman (1993) found that Arabic native speakers rely primarily on gender agreement, followed by case marking and animacy, to assign agency. This chapter investigates whether and how L2 learners of Arabic use word order and subject-verb agreement to assign an actor role in simple sentences. It assesses the role of L1 (English) in processing Arabic sentences and how L2 processing develops with increasing exposure to the language. The purpose is: (1) to determine whether L2 learners of Arabic enrolled in their first year show L1 transfer and use word order to assign a subject role and (2) if so, to what extent additional exposure to Arabic will decrease reliance on word order and increase reliance on verb agreement. Language use patterns collected from three groups of L2 learners enrolled in first-, second-, and third-year Arabic classes were examined. The frequency of choosing the first noun (1N) was calculated as a function of subject-verb variation for each group of students within the VNN and NVN word order. This study tentatively suggests that there might be an L1 transfer at the beginner stages of learning Arabic as an L2 and that learners start to use the L2 cues in processing L2 sentences with more exposure to the language. It also suggests some implications for L2 pedagogy. Whether it is because of L1 transfer or universal strategies, the beginning group did not utilize the verb agreement to assign actor role regardless of the fact that they received explicit instructions about verb agreement in Arabic.

10.1 Introduction This study investigates how second language (L2) learners of Arabic use word order and subject-verb agreement to assign an actor role in simple sentences. It aims to assess the role of a first language (L1, English) in processing Arabic sentences and how L2 processing develops with increasing exposure to the language.1 One of the theoretical models in sentence processing that addresses these two questions is the competition model (CM, henceforth).2 Building on the CM model for L1 acquisition, L2 processing studies predicted that L2 learners would initially transfer their L1 knowledge and would process L2 sentences by means of the most valid cue in 201

Jamil Al-Thawahrih

their L1. Then, as their proficiency increased they would shift to using the cues that are valid in that L2 (MacWhinney 2005). Studies in CM have most frequently made use of subject identification tasks. In subject identification task studies, participants are typically asked to determine which of two nouns in a simple sentence the actor/doer of the action is. In these studies, researchers constructed experimental sentences using two nouns and one transitive verb. They manipulated cues such as verb agreement, word order, animacy, case marking, and stress as independent variables. To illustrate, a word order could be manipulated to construct NVN, VNN, and NNV sentences; animacy could be varied as the following: both nouns are animate, only the first noun is animate, only the second noun is animate, and both nouns are inanimate.3 The manipulation of the cues inevitably yields some grammatical and some ungrammatical sentences, depending on the language. In English, the sentence the boy the girl hit is ungrammatical unless it is followed by another clause.4 In experiments, cues are sometimes presented in a competing or converging order. For instance, in the ungrammatical sentence, the horses licks the cow, word order and verb agreement compete; word order suggests that the agent of this sentence is the horses, while the verb agreement points toward the cow. However, in the sentence the horses lick the cow, word order and verb agreement converge, and both refer to the horses as the subject of the sentence. In the CM framework, word order wins in sentences where word order and verb agreement compete if participants choose the horses as the subject in spite of non-agreement with the verb. Thus, word order, according to the CM, is considered to be a stronger cue than verb agreement in English. A number of L2 studies within the CM framework confirm the transfer predictions the model anticipates. For example, McDonald (1987b) conducted one of the first CM studies on L2 processing with English-Dutch and Dutch-English L2 learners at different proficiency levels and English and Dutch native speakers (NSs) as control groups. She manipulated word order, subject-verb agreement, case marking, and animacy. She predicted that English-Dutch learners would rely on word order (English cue) at low proficiency levels to comprehend sentences and that Dutch-English learners would rely on case marking (Dutch cue) to complete a subject identification task. The results show that English speakers’ reliance on word order declines while reliance on case inflections, the cue that Dutch NSs rely on most, increases as students become more proficient in Dutch. Similarly, Dutch speakers in advanced English classes rely more on word order than those who were in beginner classes. On the other hand, some researchers who conducted studies to test the CM hypotheses did not find L1 influence when participants encountered L2 sentences. For example, Gass (1987) predicted that Italian-speaking English learners would use their strongest L1 cue, which is subject-verb agreement, to interpret English sentences. Contrary to expectations, she found that this group of learners relied on animacy, with no evidence of L1 transfer. However, this cross-linguistic study shows that Italian learners do in fact demonstrate a developmental shift toward word order as a cue to process English sentences. Based on these results, Gass (1987) suggests that semantic cues might be universal cues in L2 processing, but further studies on L2 learners have not supported such claims (Kilborn and Cooreman 1987). However, it is worth mentioning that other researchers proposed that L2 learners use universal strategies to process L2 sentences. For example, VanPatten (2007) claims that comprehension is effortful and cognitively costly at the beginning phases of learning a language. Thus, L2 learners resort to using local processing strategies such as assigning the first noun in the utterance as the subject of the sentence. However, VanPatten’s first noun principle (FNP) was tested mainly with English NSs learning different languages. As mentioned earlier, English NSs have a first noun bias, which makes it hard to determine whether NSs of English transfer their L1 knowledge or chose the first noun similar to VanPatten’s FNP predictions (Brandl 2013, p. 26). 202

Arabic L2 learners’ use of word order

To the best of the author’s knowledge, only two Arabic language studies have been conducted to test the CM framework: (1) Taman (1993), who looked at how only NSs use case marking, gender verbal agreement, and animacy to assign an actor to a sentence, and (2) Abu Radwan (2002), who examined how two groups of beginner and intermediate English-speaking learners of Arabic used verbal gender agreement, animacy, and case marking cues to assign a subject role in Arabic sentences. In addition to the two groups of learners, he used nine Arabic NSs as a control group. By manipulating these cues, Abu Radwan constructed 54 sentences using two nouns and one transitive verb.5 Participants were asked to circle the subject of the sentence. The percentage of participants identifying the first noun (1N) as the subject was calculated. While Abu Radwan’s main purpose was to study non-native speakers (NNSs), the study itself is important, as it is the second study used to test Arabic NSs in order to confirm CM prediction. The results from the NSs revealed that they relied mostly on case marking, secondarily on gender, and lastly on animacy. Both Taman and Abu Radwan’s results in regard to NSs followed different patterns. Whereas Taman found that native speakers rely on gender, Abu Radwan’s speakers rely on case marking. The conflicting findings are noteworthy but irrelevant to this current study for two reasons. First, the statistical differences between case and gender cues are small in both studies (Abu Radwan 2002, p. 206). Second, case marking is not a factor in this study. Nevertheless, both studies on NSs show that they rely on verb agreement as a strong cue to assign an agency role in simple sentences. Interestingly, Abu Radwan found that learners in their first semester of learning Arabic also relied predominantly on case marking, followed by gender and animacy, and did not transfer L1 behavior. The same pattern was observed in the intermediate group. Abu Radwan attributes the lack of transfer from L1 to the effect of explicit instruction (p. 204). Abu Radwan used a fixed word order (VNN) and manipulated the case marking on both nouns. Within this manipulation there were two conditions in which the two nouns were either accusative or nominative. In VNN Arabic sentences, only the subject is marked with the nominative marker, and only the object is marked with the accusative marker. Thus, having two nouns marked as accusative or nominative constitutes an ungrammatical sentence. However, Abu Radwan (2002) mentions that he did not use ungrammatical sentences (p. 195). Neither did he provide the list of stimuli used in his study nor an explanation of how he did not use ungrammatical sentences. Moreover, neither he nor Taman investigated the role of word order but instead employed only VNN word order. On a final note, unlike other CM studies, both studies used only 2–3 nouns and one transitive verb to conduct their stimuli, without any distractor sentences, which, in addition to the limitations mentioned earlier, restrict the generalizability and the predictive power of these studies. The primary limitations of empirical research conducted under the CM framework relate to: (1) the use of ungrammatical sentences and (2) the use of simple sentences. McLaughlin and Harrington (1989) argued that participants can be expected to find difficulties processing ungrammatical sentences and that participants might apply other problem-solving strategies instead of using the sentence-processing strategies being investigated. Thus, ungrammatical sentences might introduce unwanted elements into the analysis.6 Second, the majority of studies conducted within the CM framework used simple sentences, although some studies explored more complex structures such as relative clauses (Bates et al. 1999). Since complex structures exist naturally, studying only simple sentences limits the predictive power of the CM model (Harrington 2001). Finally, Fender (2001, p. 325) states that new studies on sentence processing have found that this is an incremental process, meaning that each incoming lexical item provides some information that updates the comprehension process; individuals never wait until they hear the whole utterance to interpret its meaning. In this sense, CM 203

Jamil Al-Thawahrih

studies fail to show how real-time processing occurs. However, it should be reiterated that processing in this study is defined as assigning an agency role in simple sentences.

10.2  Research questions The current study is motivated by: (1) the lack of Arabic studies that investigate the role of L1 in L2 processing, (2) the lack of consensus about the role of L1, and (3) methodological limitations of prior studies. This study seeks to answer the following questions: 1 To what extent do L1 English learners of Arabic observed in their first year show L1 (English) transfer by using word order to assign a subject role in Arabic sentences? 2 Does more exposure to Arabic result in declining use of word order and increased reliance on subject-verb agreement (in particular that of gender agreement) – a cue that native speakers of Arabic rely on to assign agency – and to what extent does it do so? The current study accounts for one of the limitations mentioned in previous Arabic studies and clarifies the role of L1 transfer in L2 processing: all of the sentences used in this study were grammatical. Moreover, following CM non-Arabic studies, different nouns and verbs were used in successive sentences, even though each sentence contained only “two nouns and one transitive verb.” Finally, the word order cue was manipulated to study its effect on agency assignment. Researchers have identified simple sentences as a problem for the model (Year 2003), as this practice fails to account for the complex structures found in natural languages. However, when including low and intermediate level L2 learners, the research within the CM claims to provide an insight into how those learners “utilize general cognitive and semantic/ conceptual processing strategies to make meaning of the L2/ESL language” (Fender 2001, p. 358). One goal of this study is to investigate the role of L1 transfer, which is why it includes beginning students of Arabic in their first year of study. However, including these learners also limited the study; for example, syntactic structures were limited to those structures covered in the textbook in the first year, which are quite simple. In addition, vocabulary is limited at this elementary level. It is for these reasons that the author chose to limit the tasks to simple sentences with two nouns and one verb.

10.3 Methods 10.3.1 Participants Participants in the study were: 22 Arabic L2 learners (nine males and 13 females) enrolled in a first-year Arabic course (“Beginner, B” henceforth), 18 (six males and 12 females) in a second-year Arabic course (“Intermediate, I” henceforth), and six (two males and four females) in a third-year Arabic course (“Advanced, A” henceforth) at a higher education institution in the United States. Participants were in the second semester of their corresponding year, with each semester comprising five credit hours of Arabic instruction and practice. None of the participants was a heritage Arabic speaker and none had exposure to Arabic prior to joining the Arabic program in their academic institution. The age range of the beginner group is 19–33 (with a mean of 22.4), the intermediate group is 19–56 (with a mean of 24.9), and the advanced group is 20–32 (with a mean of 24.2). At the time of the study, Arabic classes used the Al-Kitaab textbook (Brustad et al. 2011) as the primary textbook for first-year and second-year classes. Third-year Arabic used various 204

Arabic L2 learners’ use of word order

authentic resources from the media and Arabic literature instead of the textbook. Students are exposed to nominal and verbal sentences from early on but do not learn these two types of sentences explicitly until Chapter 4 (Brustad, et al. 2011, pp. 84–86). Nominal sentences begin with a noun and do not necessarily include a verb. Al-Kitaab also introduces nominal sentences that include a verb as SV(O) sentences. Verbal sentences, in contrast, start with a verb. The book introduces the alternate word order used in Arabic, VNN, interpreted as VSO. Lastly, the SV(O) and VS(O) interpretations for NVN and VNN word orders are very common in the Arabic language.

10.3.2 Stimuli In total, 48 Arabic experimental sentences with VNN and NVN word order were constructed using 48 sets of two nouns and one verb. The sets were randomly generated using Excel’s (Microsoft, 2010) random function; none of the sets were identical. An example of a set of three words is: ħabiib “lover s.m,” ya-taðakkar “3.s.m-remember,” and ʔaxuu “brother.” The sets were constructed using 12 masculine nouns and their feminine counterparts and 12 transitive verbs. Ten of the nouns were feminized using the Arabic feminine morpheme {-a}. However, two of the nouns had irregular feminine counterparts, which was indicated by using a different word. The 48 sentences (24 sentences for NVN and 24 for VNN word order conditions) were varied with respect to the subject-verb agreement as follows: only the first noun (1N) agrees with the verb, only the second noun (2N) agrees with the verb, and both nouns agree with the verb (Amb). This yielded a total of 144 experimental sentences. All sentences were semantically reversible; that is, either noun could be understood to have performed the action. Moreover, the semantic plausibility was balanced in all sentences by making sure that either noun could equally be the subject of the sentence. Previous research showed that children and non-native speakers are likely to choose mother as the subject in the sentence the baby feeds the mother because in the real world this is the only plausible interpretation (Bates et al., 1984, p. 342). Studies conducted within the CM framework usually employ anywhere from one to four sentences per condition and a large number of participants (e.g., Bates et al. 1982). However, this study was limited by the small number of participants, especially those in the third year (six participants). Thus, to increase the power of the statistical analysis, there were eight sentences per condition. Twenty-four filler sentences were constructed using 12 nouns, intransitive verbs, and prepositions. Adding the fillers aims to distract the participants from the real purpose of the study and to know if some participants were not invested in the task or if they did not comprehend the sentences. All lexical items were chosen from the firstyear textbook to help ensure that participants were familiar with them (for a list of all sentences used, see Appendix). Three lists were created from the 144 experimental sentences. In each list, one of the three subject-verb agreement conditions of each “two nouns and one transitive verb” set appeared. Selecting one condition of the subject-verb agreement conditions was done to minimize the possible effect of seeing the same lexical items more than once and to force the participants to read each sentence. Choosing one variation of the three subject-verb agreement conditions in each group yielded 48 experimental sentence blocks. The same 24 filler sentences were added to each block, which created a 72-sentence block. Table 10.1 shows a sample of the sentences used. Sentences were presented only in Arabic script for participants. Each experimental sentence consisted of two noun phrases (NP1, NP2) and a transitive verb. Each noun phrase consisted of: (1) a singular noun that always referred to a human being, (2) either (a) the Arabic definite article ʔal- “the” or (b) the possessive pronoun -ii “my.” Adding 205

Jamil Al-Thawahrih Table 10.1 Stimuli examples Word Order

Example

Subject-Verb Variation

VNN

(1) ya-ḍrib ʔal-walad ʔal-bint    3.s.m-hits the-boy the-girl.    “The boy hits the girl.” (2) ta-ḍrib ʔal-walad ʔal-bint    3.s.f-hits the-boy the-girl.    “The girl hits the boy.” (3) yu-darris ʔal-walad ʔaṭ-ṭaalib    3.s.m-teaches the-boy the-student.s.m    “The boy teaches the male student.” (4) ya-drus ʔal-walad maʕa ʔal-bint    3.s.m-studies the-boy with the-girl.    “The boy studies with the girl.” (5) ʔal-bint tu-darris ʔal-walad    the-girl 3.s.f-teaches the-boy.    “The girl teaches the boy.” (6) ʔal-walad tu-darris ʔal-bint    the-boy 3.s.f-teaches the-girl.    “The girl teaches the boy.” (7) ʔaṭ-ṭaalib-a ta-ḍrib ʔal-bint    the-student-s.f 3.s.f-hits the-girl    “The female student hits the girl.” (8) ʔal-bint ta-drus maʕa ʔal-walad    the-girl 3.s.f-studies with the-boy    “The girl studies with the boy.”

1N only agrees with the verb.

NVN

2N only agrees with the verb. Ambiguous; both nouns agree with the verb. Filler 1N only agrees with the verb. 2N only agrees with the verb. Ambiguous; both nouns agree with the verb. Filler

either (a) or (b) maintained the grammaticality of the sentences by making the noun definite. Arabic sentences do not start with an indefinite noun (Mohammad 2000). Both the definite article and the possessive pronoun were used because in some cases we need to use the definite article rather than a possessive pronoun. For example, if the possessive pronoun were to be added to the word girl, the meaning would change from “girl” to “daughter.” This raised the following issue: when the possessive pronoun was added to a feminine noun, the feminine marker was no longer the last morpheme in the noun phrase, and it became less salient than if the definite article was added. The definite article was used only with four masculine nouns and their feminine counterparts (doctor, teacher, student, and boy.) An analysis was run to make sure that participants’ performance did not differ in sentences that had a more salient feminine marker than those that had a less salient feminine marker. The results did not show any significant difference. Each filler sentence consisted of a noun phrase (NP), prepositional phrase (PP), and an intransitive verb. Nominal sentences had the following orders: (PP – NP – V), and (NP – V – PP). Verbal sentences had the following orders: (PP – V – NP), and (V – NP – PP). It should be noted that sentences that started with the PP, while grammatical, are less common if the noun in the NP is definite and are used for pragmatic reasons, such as emphasis. However, they were used in this study in order not to test for a first noun bias; the subject of these sentences was always the second noun, not the noun in the PP. In an SV cluster, verbs agree with the subject in person, gender, and number. However, in a VS structure, verbs agree with the subject in person and gender, but its number is always singular. In this study, both SV and VS structures were used; thus, to control for this issue, both 206

Arabic L2 learners’ use of word order

nouns were singular. The verb in experimental, filler, and practice sentences was in the simple present tense. Finally, the verb was always inflected for third person. The following materials were also used in this study: consent form, background questionnaire, vocabulary training delivered by DMDX (Forster and Forster 2003), and a vocabulary quiz. No specific information about the purpose of the study was provided in the consent form, but it mentioned that the Arabic program might benefit from student participation. The background questionnaire gathered information about the participants’ native language, their history studying Arabic, any visual impairments, and information about what they think the study is about. The vocabulary training session was carried out using the DMDX software. It aimed to familiarize the participants with the vocabulary used in the experiment. Each lexical item used in the experiment appeared in Arabic accompanied by its English meaning in the middle of the screen. Finally, a vocabulary test was administered in which students were asked to write the English meaning of the Arabic words after they had finished the experiment.

10.3.3 Procedure Participants were randomly assigned a computer in order to have an equal number of participants per experimental block. They completed and signed the consent form. After the participants had received instructions, they completed the experiment at their own pace. The experiment started with a vocabulary training session to familiarize the participants with the lexical items. Each lexical item remained on the screen for 10 seconds or until the participant pressed the SECOND key (Left Shift). If no action was made on the tenth second, the next lexical item appeared. After that, participants performed a practice set of six sentences to familiarize themselves with the experiment (see Appendix). Then they proceeded to the actual experiment and completed the task by selecting the correct subject of the sentence. Before the experiment started, written instructions notified the participants to push FIRST (Right shift) if they think that the subject of the sentence they read was the first noun and SECOND if the subject of a sentence which they read was the second noun. DMDX presented the sentences in random order, and each sentence appeared for a maximum of 9 seconds or until the participants made a response. A pilot study, run on a beginner NNS who did not participate in this study, showed that the 9 seconds were sufficient for the participant to read the sentence and to make a response.7 Participants were instructed to ignore the sentence if it disappeared – after 9 seconds – before they responded and move on to the following one. At the end of the experiment, the participants completed a paper-and-pencil vocabulary quiz to make sure they knew the meaning of the lexical items that appeared on the test and then filled out the background questionnaire. Each participant completed the experiment within 30 minutes.

10.4 Results The frequency of choosing 1N was calculated as a function of subject-verb variation for each group of students within the VNN and NVN word order. A score of 1 was awarded when the participant chose the first noun as the subject of the sentence, and a 0 score was given when the participant chose the second noun. Missing data were eliminated. This affected 8.37% of the results. The missing data constituted 8.52% of the beginner, 7.29% of the intermediate, and 11.11% of the advanced groups’ responses. Then the average of choosing the first noun per condition was calculated for each group. After that, the data for each of the L2 learner groups were submitted to an analysis of variance (ANOVA) with subject-verb agreement (1N, 2N, Amb) and word order (NVN, VNN) as within-subject variables. Then the data were submitted 207

Jamil Al-Thawahrih

to ANOVA with group of learners (Beginner, B; Intermediate, I; Advanced, A) as a betweensubject variable, and with subject-verb agreement (1N, 2N, Amb) and word order (NVN, VNN) as within-subject variables. The interaction between word order and subject-agreement was also examined, as it illustrates the dominance of each cue (Brandl 2013, p. 88). Alpha level is set at 0.05 unless otherwise stated.8

10.4.1  1N choice The variation among the groups across the subject-verb and word order conditions to select the first noun is illustrated in Table 10.2, which shows the average for the first noun choice for each L2 NNS group across the six conditions. As predicted, the data show that regardless of the subject-verb agreement condition, the Beginner group participants selected the 1N with an average of 0.917 in the canonical word order NVN (of the three conditions combined: NVN_1N, NVN_2N, and NVN_Amb), while they chose the 1N with an average of 0.527 in the noncanonical word order VNN (of the three conditions, combined: VNN_1N, VNN_2N, and VNN_Amb) across all the subject-verb agreement conditions. For the Intermediate and Advanced groups, participants’ performance was not dependent on the word order, but it depended on the subject-verb agreement condition. Table 10.2 illustrates that Intermediate and Advanced groups selected the first noun as agent more in 1N agrees condition in both word orders, with an average of 0.875 and 0.835, respectively. It also shows that both groups followed the same pattern in the Amb condition in both word orders, with an average of 0.835 for the Intermediate group and 0.810 for the Advanced group. The percentages of first noun choice dropped in the 2N agrees condition for NVN and VNN orders, with an average of 0.385 for the Intermediate group and 0.56 for the Advanced group. It is worth mentioning that the L2 groups’ responses were more consistent (small SD in comparison with other conditions) in the NVN word order compared with the VNN word order. However, the responses were less consistent in the 2N agrees condition for the Intermediate group, and no difference was observed for the Advanced group. To evaluate the differences in the responses presented, two-way ANOVA tests were conducted. The results are presented in Table 10.3 for the three L2 groups individually. Table 10.4 also illustrates the complete ANOVA, which is necessary to show the interactions among the three L2 groups (B, I, A). The results listed in Tables 10.3–10.4 are discussed in subsequent sections.

10.4.2  Word order condition The two-way ANOVA results revealed a significant interaction between word order and group: F(2,2005) = 28.365, p < 0.05, η2p = 0.028 (see Table 10.4). The results show a main effect for Table 10.2 Means and SDs for B, I, and A groups’ first noun choice Group Beginner, B, n = 22 Intermediate, I, n = 18 Advanced, A, n =6

Mean SD Mean SD Mean SD

NVN_1N

NVN_2N

NVN_Amb

VNN_1N

VNN_2N

VNN_Amb

0.94 0.09 0.96 0.08 0.90 0.12

0.92 0.17 0.45 0.35 0.56 0.41

0.89 0.17 0.97 0.08 0.81 0.20

0.54 0.33 0.79 0.31 0.77 0.23

0.51 0.33 0.32 0.22 0.56 0.32

0.53 0.33 0.70 0.30 0.81 0.23

208

Arabic L2 learners’ use of word order Table 10.3 ANOVA results for B, I, and A groups’ first noun choice Group

Source

df

Mean Square F

Beginner, B

Verb Agreement (AG) 2 0.09 Word Order (WO) 1 40.14 AG * WO 2 0.12 Error 960 0.16 Intermediate, I Verb Agreement (AG) 2 19.64 1 7.17 Word Order (WO) AG * WO 2 0.42 Error 795 0.15 2 2.41 Advanced, A Verb Agreement (AG) Word Order (WO) 1 0.14 AG * WO 2 0.10 Error 250 0.18

Sig. (p) Partial Eta Observed Squared (effect Powera size)

0.53 249.59 0.75   128.68 46.98 2.76   13.44 0.76 0.57  

0.59 0.00 0.47   0.00 0.00 0.06   0.00 0.38 0.57  

0.00 0.21 0.00   0.24 0.06 0.01   0.10 0.00 0.00  

0.14 1.00 0.18   1.00 1.00 0.55   1.00 0.14 0.14  

Table 10.4 Complete ANOVAs for B, I, and A groups’ first noun choice Source

df

Mean Square

F

Sig. (p)

Partial Eta Squared (effect size)

Observed Powera

Verb Agreement (AG) Word Order (WO) Group (G) AG * WO AG * G WO * G AG * WO * G Error

2 1 2 2 4 2 4 2005

10.958 16.675 .216 .069 5.075 4.534 .306 .160

68.554 104.320 1.352 .432 31.748 28.365 1.914  

.000 .000 .259 .649 .000 .000 .105

.064 .049 .001 .000 .060 .028 .004

1.000 1.000 .293 .121 1.000 1.000 .582  

 

 

word order in the Beginner (F(1, 960) = 249.59, p < 0.05, η2p = 1.00) and Intermediate (F(1,795) = 46.98, p < 0.05, η2p = 1.00) groups but not in the Advanced group (see Table 10.4). In the Beginner group, participants chose the first noun 91.7% of the time in the NVN word order and only 52.7% of the time in the VNN word order. The word order accounted for 21% of the variance in the Beginner group data. The Intermediate group selected the first noun 79.33% of the time in the NVN word order and 60.33% of the time in the VNN word order. Unlike the Beginner group, word order accounted for only 6% of the variance in the data. Finally, the Advanced group selected the first noun 75.67% of the time in the NVN word order and 71.33% of the time in the VNN word order. Word order did not account for any variance in the Advanced group’s responses.

10.4.3  Subject-verb gender agreement condition As expected, the interaction between group and gender verbal agreement was significant: F(4,2005) = 31.748, p < 0.05, η2p = 0.060 (see Table 10.4). The ANOVA results showed a significant effect for the verbal agreement variable in Intermediate (F(2,795) = 128.68, p < 0.05, 209

Jamil Al-Thawahrih

η2p = 1.00) and Advanced (F(2,250) = 13.44, p < 0.05, η2p = 1.00) groups but not in the Beginner group. The verb agreement effect was not significant in the Beginner group; it did not account for any variance in the group’s responses. As Table 10.2 shows, participants in the beginner group assigned agency to the first noun with almost similar percentages across all subject-verb agreement conditions, with 74% of the time when the 1N agrees, 71.5% when the 2N agrees, and 71% for the Amb condition.9 On the other hand, the verb agreement variable accounted for 24% of variance in the Intermediate group’s responses. Although relatively small, the effect of the verb agreement on the noun choice was higher for this group than the effect of word order. The agency decision of the Intermediate group was clearly influenced by the subject-verb agreement condition. In the 1N agrees condition, participants selected the first noun 87.5% of the time, in comparison with 38.5% of the time in the 2N agrees condition. However, they chose the first noun 83.5% of the time in the Amb condition. The results for the Advanced group largely followed the Intermediate group’s pattern; however, there were some differences. The verb agreement in the Advanced group only accounted for 10% of the variance of this group’s data. Participants chose the first noun 83.5% and 81.00% of the time for 1N agrees and the Amb conditions, respectively. Nevertheless, they chose the first noun 56.00% of the time in the 2N agrees condition. Based on the results, and as predicted, the two cues used in this experiment, subject-verb agreement and word order, may be ranked as follows for each learner group: Beginner: Word order > subject-verb agreement. Intermediate: Subject-verb agreement > word order. Advanced: Subject-verb agreement > word order. In the next sections, the results of the interactions between word order and subject-verb agreement are presented for each group.

10.4.4  Beginner group The interaction between word order and verb agreement was not significant, F(2,960) = 0.75, p < 0.47, η2p = 0. Figure 10.1 demonstrates that within the same word order, the participants were consistent in selecting the first noun as agent. Participants chose the first noun as the agent 91.7% of the time in the NVN condition, regardless of agreement, while they selected the 1N with an average of 52.7% in the VNN condition, again disregarding verb agreement. The results from the VNN shows a slight tendency to choose first noun more than the second noun.

10.4.5  Intermediate group The interaction among word order and subject-verb agreement was not significant for the Intermediate group, where F(2,795)= 2.79, p < 0.06; however, it approached significance. This is illustrated in Figure 10.2, where participants chose the first noun in the NVN when the verb agreed with the first noun or the ambiguous conditions more often when the verb agreed with the second noun. The same pattern applies to the VNN order. However, their performance in the VNN order was less native-like than their performance in the NVN word order, except for the 2N agrees condition.

210

Arabic L2 learners’ use of word order

Figure 10.1 Word order and subject-verb agreement interaction for the Beginner group 1N choice

Figure 10.2 Word order and subject-verb agreement for the Intermediate group 1N choice

10.4.6  Advanced group There was no significant interaction among word order and verb agreement for this group of participants: F(2,250) = 0.57, p < 0.57. However, results shown in Figure 10.3 resemble the Intermediate group results. The lack of significance is likely due to the small number of participants (n = 6) in this group; the low observed power (OP) of this interaction (OP = 0.14) indicates that the sample size was inadequate (see Table 10.3).

10.4.7  Interactions of word order, subject-verb agreement, and group There was no significant interaction between word order, verb agreement, and group: F(4,2005) = 1.914, p < 0.105. Nevertheless, the result approached significance. For the NVN word order condition, there were no significant differences between the three groups of learners in the 1N agrees condition. However, for the 2N agrees condition, the Beginner group chose the first noun significantly more often than either the Intermediate (F(1,43) = 27.014, p  = 0.00, ηp2 = 0.386) or the Advanced (F(1,43) = 7.545, p = 0.009, ηp2 = 0.149) groups, but there was no significant difference between the Intermediate and Advanced groups’ responses. For the Amb subject-verb agreement condition, the Beginner group’s performance did not differ from that of the other groups. Interestingly, there was a significant difference between the performances of the Intermediate and the Advanced groups (F(1,43) = 5.064, p = 0.03, 211

Jamil Al-Thawahrih

Figure 10.3 Word order and subject-verb agreement for the Advanced group 1N choice

ηp2 = 0.105). For the VNN word order, only two significant differences existed in the performance of the three groups in all of the subject-verb agreement conditions: (1) the Intermediate group chose the first noun significantly more often than the Beginner group in the 1N agrees condition (F(1,43) = 6.579, p = 0.014, ηp2 = 0.133), and (2) the Intermediate group selected the first noun significantly less often than the Beginner group in the 2N agrees condition (F(1,43) = 4.120, p < 0.049, ηp2 = 0.087). In these two cases, the Intermediate group’s performance was, as expected, more native-like than the Beginner group.

10.5 Discussion 10.5.1  Main findings The Beginner group showed high reliance on word order to assign agency. Within the same word order, participants’ performance of first noun choice was not affected by the subject-verb agreement condition. This group’s reliance on word order was clearly manifested in the NVN_2N condition. In that condition, the Beginner group selected the first noun 92% of the time (Table 10.2). The native-like choice would be choosing the second noun as the subject of the sentence, which would lead to a low first noun choice percentage. Choosing the first noun for this condition with such a relatively small variation (SD = 0.17) clearly demonstrates that the participants in this group did not pay attention to the verb agreement cue in these sentences. The Beginner group’s data illustrate a contrast in participants’ behavior between the NVN order and the VNN order. The results from the NVN sentences align with research conducted on English native speakers, which shows that they interpret the NVN sentences as SVO sentences (McDonald 1987a). This suggests that the Beginner group participants used their L1 to process the sentences. However, the pattern observed in the VNN sentences neither confirmed nor refuted L1 processing. According to CM studies, English native speakers apply a second noun strategy to the noncanonical orders. This means that the Beginner group was expected to interpret the VNN sentences as VOS sentences, which would translate to a low first noun choice in all the VNN sentences regardless of the subject-verb agreement condition.10 However, participants in the Beginner group performed at chance level across all verb agreement, with a small tendency to choose the first noun as an agent. One possible explanation for the contrasts in the participants’ behavior on the word order conditions relates to the fact that English is rigidly a NVN language with SVO interpretation, except in passive sentences. On the other hand, Arabic has free word order, with NVN and VNN word orders more common than the NNV. 212

Arabic L2 learners’ use of word order

The difference between the two languages is not only in the different word order, but also in the interpretations allowed within these word orders. In the NVN word order, Arabic allows SVO, which aligns with the most common interpretation in English. Although it is rare, Arabic, unlike English, also allows OVS sentences. The interpretations of VNN word order highlight the differences between the two languages; in Arabic the most common interpretation is the VSO sentences, but the rare VOS interpretation is also allowed, while in English VNN is always interpreted as VOS (MacWhinney et al. 1984). Thus, the similarity between interpretations of the NVN word order across the two languages might have caused the L2 learners to not pay attention to verb agreement and to assume that the NVN Arabic sentences are always SVO sentences. MacWhinney (2001) refers to this sort of transfer as positive transfer (p. 80). Recall that the textbook used in first and second years explicitly introduces the NVN sentences as only SVO sentences for the sake of simplicity. On the other hand, the difference in interpreting VNN sentences between the two languages, as well as the explicit instruction received from the textbook and their instructors about the order of Arabic verbal sentences, might have helped the L2 learners to suppress their L1 interpretation (VOS) to some degree. This assumption aligns with MacWhinney’s (1997) claim that [beginner] L2 learners start learning their L2 by transferring everything that could be transferred from their L1 (p. 119). This is to say that the participants in this group might have started with VOS interpretation based on second noun bias found in their (English) L1; however, they moved to the VSO interpretation because of the explicit grammar instruction they received. MacWhinney (2005) claims that when learners detect errors caused by L1 transfer, they might be able to suppress this incorrect transfer (p. 57). High variability in the VNN order in comparison with the NVN was observed, which might suggest that some learners detected that the direct transfer to the VNN yields misunderstanding and that they needed to adjust their hypotheses about the VNN order in Arabic. Now, the question raised here is: “Do these findings support the presence of L1 transfer?” As discussed earlier, the results from the NVN order support L1 transfer, while the VNN findings did not align with prior research findings. Some researchers proposed that L2 learners resort to universal strategies in interpreting L2 sentences. For example, a first noun principle (FNP) proposed by VanPatten (2007) states that “learners tend to process the first noun or pronoun they encounter in a sentence as the subject” (p. 122). The data from the NVN align with VanPatten’s FNP. Moreover, the FNP explains the results from the VNN sentences. As with the L1 transfer explanation, we can argue that the participants start with utilizing the VSO order, and after spending almost two semesters learning the target language, their interlanguage reorganizes to integrate the verb agreement cue. That being said, it is not clear why they would shift from using the FNP in interpreting VNN sentences and start to choose the second noun as an agent if the textbook explicitly interprets the VNN word order as VSO. Thus, it is more plausible to assume that they started with their L1 assumption that VNN word order is a VOS sentence; then later, with the explicit instruction they received, they shifted toward a VSO interpretation. Whether or not L1 transfer was involved cannot be conclusively determined from the current data. The data did demonstrate, however, that the Beginner group did not use the verb agreement cue but rather word order to interpret Arabic sentences. The findings from this group show the limitations of a cross-sectional study. A longitudinal study that examines the same Arabic learners’ performance at different stages, including earlier in the first year, is needed. On the other hand, the Intermediate group relied on verb agreement rather than the word order to assign agency. However, there was no significant interaction between subject-verb agreement and word order conditions, but it approached significance: F(2,795) = 2.76, p < 213

Jamil Al-Thawahrih

0.06, η2p = 0.01. Moreover, the observed power (0.55) suggests that with a bigger sample size, the results might have reached significance. Examining the six different conditions, we find that in each word order the participants picked the first noun more often in the 1N agrees and the Amb subject-verb agreement conditions in comparison with the 2N agrees subject-verb agreement condition. That being said, their performance on the NVN word order was more consistent (small SD) and native-like than their behavior on the VNN order, except for the 2N agrees condition (Table 10.2). This is an interesting finding that we return to after examining the Advanced group’s behavior. For the Advanced group, no significant differences were detected for the different interactions. The small sample size might have been the biggest factor for not reaching statistical significance. However, verb agreement accounted for 10% of the variance in the data collected and it was, like the Intermediate group but unlike the Beginner group, significant: F(2,250) = 13.44, p < 0.05, η2p = 0.10. Word order was not significant and did not account for any variance in the Advanced group’s responses. Similar to the Beginner group, the interaction between the word order and subject-verb agreement conditions was not significant: F (2,250) = 0.57, p < 0.57. Nevertheless, the pattern of interaction between the word order and verb agreement of the Advanced group is similar to the Intermediate group, although not significant (Table 10.3). The participants in the Intermediate and Advanced groups selected the first noun, regardless of the word order, more in the 1N agrees and the Amb subject-verb agreement conditions in comparison to the 2N agrees subject-verb agreement condition (Table 10.3). The lack of significance in the interactions between word order and subject-verb agreement is likely due not to the participants in the Advanced group performing the same across the subject-verb agreement condition, but, as stated before, to the small sample size. As mentioned earlier, the results showed high variability in the Intermediate group’s behavior in the VNN sentences in comparison with the NVN sentences, except for the 2N agrees condition. This high variability in the participants’ performance in the VNN order was similar to the beginner group behavior, which, as explained before, tentatively suggests an L1 effect. The findings from these two aforementioned groups suggest that L2 Arabic learners might more readily process NVN order with its variation in Arabic than VNN order. Interestingly, the Intermediate group’s performance on the NVN_2N condition as compared to their performance on the VNN_1N condition showed the opposite pattern; the participant performance was more native-like on the VNN_2N variation, with a smaller SD in comparison with the NVN_2N variation. This behavior might also suggest an L1 transfer effect. In other words, native English speakers apply a first noun strategy in the NVN word order, which leads them to interpret the sentence as a SVO sentence. The SVO interpretation is the opposite interpretation for the NVN_2N sentences in Arabic. However, English speakers employ a second noun strategy for a noncanonical word order, such as VNN order; this leads them to prefer the VOS interpretation over the VSO, which agrees with the Arabic interpretation of the VNN_2N sentences. On the one hand, since the NVN_2N native-like interpretation goes against their L1 first noun strategy, they might have faced some difficulties processing this variation in a native-like way (by utilizing the verb agreement cue). On the other hand, the VNN_2N nativelike behavior mirrors their second noun strategy in noncanonical word order, and thus what seems like acquiring native-like strategies might only be the effect of L1 transfer. It should be reiterated that due to the lack of significant difference between the NVN_2N and the VNN_2N conditions, there is a need for a longitudinal study in order to shed light on learners’ behavior at different periods of time, which will help us understand these results. Finally, the Advanced group did not show any significant difference (due to group size) between the participants’ performances in the two word order conditions. Nevertheless, the difference in the mean of 214

Arabic L2 learners’ use of word order

first noun choice between NVN order and VNN order was only seen in the 1N subject-verb agreement condition (Table 10.2). This pattern is similar to the one found in the intermediate group, and as suggested earlier might be due to an L1 effect. Interestingly, the Intermediate and even the Beginner groups outperformed the Advanced group in the NVN_1N and the NVN_Amb conditions. Moreover, the Intermediate group outperformed the Advanced group in the NVN_2N and VNN_2N conditions (Table 10.2). The Beginner group performed better because they applied either the FNP or because they were heavily transferring their L1 processing strategies and not because they attended to the subject-verb agreement cue, as discussed earlier. However, the Intermediate group did utilize the verb agreement cue, and the question remains as to why the Intermediate group behaved in a more native-like way than the Advanced group. There are several explanations as to why this might be the case. However, before presenting these explanations, we should emphasize that there were no significant differences between the Intermediate group and the Advanced group except in the NVN_Amb condition. The Intermediate group selected the first noun more frequently than the Advanced group in the NVN_ Amb condition, but this could be merely because of the small sample size of the Advanced group. The difference that does exist could be due to the fact that the Intermediate group had explicit instruction about verb agreement from the textbook and their instructors, while the Advanced group did not receive any explicit instructions about the same grammatical structure. Alternatively, participants in the Advanced group used authentic materials from different sources, which possibly introduced them to more variations of subject-verb agreement forms such as feminine plural and dual agreement, to which participants in the Intermediate group were not exposed. This could have caused the participants from the Advanced group to reorganize their knowledge of subject-verb agreement forms. Finally, the participants’ proficiency levels might not be aligned with the year of enrollment. Participants were part of another study which administered a proficiency test for Arabic students. This proficiency test shows that Intermediate group outperformed the Advanced group on the listening and reading test. However, both groups performed similarly in the speaking test. While we cannot rely on these results because not all participants who took part in this study were part of the other study, the results of the reading and listening test are representative of the participants’ interpretative skills, and the results show that the Intermediate [sub]group outperformed the Advanced group. The differences in instruction between the two groups might be the reason behind these interesting findings. It is clear that further research with a larger group of participants, as well as a proficiency exam to control for the level of the participants, is needed in order to reach sound conclusions.

10.5.2  Limitations and suggestions Several limitations have already been discussed. The sample size, especially in the advanced group (n = 6), was small, and this caused the statistical analysis to lack the power to yield a significant difference. Also, as we discussed, the results of the VNN order from the Beginner group raised the need to conduct a longitudinal study in order to answer the first research questions regarding L1 transfer. Moreover, there is a need to examine the participants’ proficiency level using a standardized test. Finally, the task itself does not resemble a real-life situation; the sentences used in this study are simple sentences limited to the gender verbal agreement cue (with all verbs being inflected for third person singular). Furthermore, research on learners with different L1s might be beneficial to tease apart universal processing accounts from that of CM. Spanish is a rich morphological language and, 215

Jamil Al-Thawahrih

based on CM, we would expect L1 Spanish speakers to behave differently from English L1 learners processing Arabic. If correct, this would provide support for L1 transfer accounts. However, if Spanish L1 and English L1 behave similarly, this would align with the universal processing accounts such as VanPatten (2007). Future research may also use more complex structures and investigate more cues related to verbal agreement; this would increase the generalizability of the study. Lastly, subsequent research may also adapt different techniques used in psycholinguistic studies, such as selfpaced reading or listening, eye tracking, and so forth. These techniques provide real-time measurements that the agent identification task adopted in this study cannot provide.

10.6  Conclusion and pedagogical implications The hypothesis for the first research question, which predicted that the Beginner group would not utilize the verb agreement cue and would rely on L1 processing cues to assign agency, was partially confirmed: the Beginner group’s results revealed that participants did not use the verb agreement cue and used word order to complete the task. Nevertheless, participants’ performance on the VNN word order neither confirmed nor refuted L1 processing, which was observed in the NVN word order condition. In fact, other alternatives, for example, universal FNP, can explain the data from both word order conditions. Thus, a firm answer to this question cannot be provided based on the results from this study. There is a need for a longitudinal study, and until more research is conducted, the lack of evidence for the second noun strategy does not necessarily suggest that there is [no] L1 transfer. Similarly, the hypothesis for the second research question, which predicted that L2 learners’ processing mechanism would tend to shift toward native-like processing as they progress from beginning levels of learning to more advanced levels, was partially supported: there is tentative evidence that the performance of the three groups moved toward a native-like behavior as their exposure to the language increased. However, there were some cases in which the Intermediate group’s responses were more native-like than the Advanced group’s responses. To conclude, this study tentatively suggests that there might be an L1 transfer at the beginning stages of learning Arabic, and learners shift to use the L2 cues in processing L2 sentences with more exposure to the language. It also suggests some implications for L2 pedagogy. Whether it is because of L1 transfer or universal strategies, the Beginner group did not utilize subject-verb agreement to assign agency even though they had received explicit instructions about subject-verb agreement in Arabic. The results suggest that L2 learners, especially at the early stages of acquisition, do not attend to grammatical forms. As a result, pedagogical interventions that aim at activating grammatical forms such as those related to verbal agreement more effectively is needed.

216

Arabic L2 learners’ use of word order

Appendix: Stimuli Sentences 1 NVN 1.a NVN (first noun agrees condition: 1N masculine, 2N feminine) English

Arabic

My lover hits my friend. The student loves the teacher. My friend marries my neighbor. My father wants the girl. My uncle hears the doctor. The boy thanks the girl. My uncle discusses my mother. The boy teaches my sister. My husband asks the doctor. My friend knows my sister. My lover helps my neighbor. My friend remembers my sister.

.‫حبيبي يضرب صاحبتي‬ .‫الطالب يحب األستاذة‬ .‫صديقي يتزوج جارتي‬ .‫والدي يريد البنت‬ .‫خالي يسمع الدكتورة‬ .‫الولد يشكر البنت‬ .‫خالي يحاور والدتي‬ .‫الولد يدرّس أختي‬ .‫زوجي يسأل الدكتورة‬ .‫صديقي يعرف أختي‬ .‫حبيبي يساعد جارتي‬ .‫صاحبي يتذكر أختي‬

1.b NVN (first noun agrees condition: 1N feminine, 2N masculine) English

Arabic

The doctor marries my brother. My friend hears my friend. My mother thanks my neighbor. My aunt hits my brother. My friend loves my neighbor. My mother helps my brother. My lover teaches the student. My wife discusses the boy. The girl knows my lover. My aunt wants the teacher. The girl asks my brother. The girl remembers the doctor.

.‫الدكتورة تتزوج أخي‬ .‫صديقتي تسمع صاحبي‬ .‫والدتي تشكر جاري‬ .‫خالتي تضرب أخي‬ .‫صاحبتي تحب جاري‬ .‫والدتي تساعد أخي‬ .‫حبيبتي تدرّس الطالب‬ .‫زوجتي تحاور الولد‬ .‫البنت تعرف حبيبي‬ .‫خالتي تريد األستاذ‬ .‫البنت تسأل أخي‬ .‫البنت تتذكر الدكتور‬

2 VNN 2.a VNN (first noun agrees condition: 1N masculine, 2N feminine) English

Arabic

My lover remembers my mother. (f) The student helps the doctor. My husband knows the girl. The boy asks my friend (f). My uncle teaches my friend (f). The boy discusses my neighbor. My uncle thanks my sister.

.‫يتذكر حبيبي والدتي‬ .‫يساعد الطالب الدكتورة‬ .‫يعرف زوجي البنت‬ .‫يسأل الولد صاحبتي‬ .‫يدرّس خالي صديقتي‬ .‫يحاور الولد جارتي‬ .‫يشكر خالي أختي‬

217

Jamil Al-Thawahrih

English

Arabic

The boy hears my neighbor. My friend wants my sister. My friend marries the teacher. My lover loves my sister. My father hits the doctor.

.‫يسمع الولد جارتي‬ .‫يريد صديقي أختي‬ .‫يتزوج صاحبي األستاذة‬ .‫يحب حبيبي أختي‬ .‫يضرب والدي الدكتورة‬

2.b VNN (first noun agrees condition: 1N feminine, 2N masculine) English

Arabic

The girl remembers my brother. My aunt marries the man. The girl loves my father. My friend teaches the boy. My neighbor knows my friend. The teacher hits my friend. My wife wants the doctor. My lover hears my father. The student asks the teachers. My friend thanks the doctor. My friend discusses the doctor. My aunt helps my friend.

.‫تتذكر البنت أخي‬ .‫تتزوج خالتي الرجل‬ .‫تحب البنت والدي‬ .‫تدرّس صاحبتي الولد‬ .‫تعرف جارتي صاحبي‬ .‫تضرب األستاذة صديقي‬ .‫تريد زوجتي الدكتور‬ .‫تسمع حبيبتي والدي‬ .‫تسأل الطالبة األستاذ‬ .‫تشكر صديقتي الدكتور‬ .‫تحاور صديقتي الدكتور‬ .‫تساعد خالتي صاحبي‬

Filler sentences 1 NP-V- PP English

Arabic

The boy studies with the girl. The teacher lives with my neighbor (f). The boy dances with my neighbor. The student (f) goes out with the boy. My lover (f) drinks with my brother. My mother runs with my sister.

.‫الولد يدرس مع البنت‬ .‫األستاذ يسكن مع جارتي‬ .‫الولد يرقص مع جاري‬ .‫الطالبة تخرج مع الولد‬ .‫حبيبتي تشرب مع أخي‬ .‫والدتي تركض مع أختي‬

2 PP- NP-V English

Arabic

The student goes to the teacher (f). My brother runs with my friend. My lover eats at the university. The doctor (f) writes to my father. My neighbor (f) lives with the boy. The girl studies with my friend (f).

.‫إلى األستاذة الطالب يذهب‬ .‫مع صديقي أخي يركض‬ .‫في الجامعة حبيبي يأكل‬ .‫إلى والدي الدكتورة تكتب‬ .‫مع الولد جارتي تسكن‬ .‫مع صديقتي البنت تدرس‬

218

Arabic L2 learners’ use of word order

3 V-NP-PP English

Arabic

The boy studies with my friend (f). The doctor lives in Utah. The teacher runs with my friend (f). The teacher (f) goes out of the class. The student (f) drinks with the girl. My mother runs with my aunt.

.‫يدرس الولد مع صديقتي‬ .‫يسكن الدكتور في يوتا‬ .‫يركض األستاذ مع صاحبتي‬ .‫تخرج األستاذة من الصف‬ .‫تشرب الطالبة مع البنت‬ .‫تركض والدتي مع خالتي‬

4 PP-V-NP English

Arabic

The boy goes to the university. The teacher runs to the class. My father eats with my neighbor. The girl writes with my friend. The doctor (f) lives with my brother. My friend (f) studies with the teacher (f).

.‫إلى الجامعة يذهب الطالب‬ .‫إلى الصف يركض األستاذ‬ .‫مع جاري يأكل والدي‬ .‫مع صديقي تكتب البنت‬ .‫مع أخي تسكن الدكتورة‬ .‫مع األستاذة تدرس صديقتي‬

Practice sentences English

Arabic

The girl drinks the milk. My mother eats lunch. My friend (f) drinks coffee. My father studies at home. The boy eats in the class. My friend (f) goes to the university.

NVN ‫البنت تشرب الحليب‬ NVN ‫الغداء تأكل والدتي‬ VNN ‫تشرب القهوة صاحبتي‬ NV PP ‫والدي يدرس في البيت‬ PP VN ‫في الصف يأكل الولد‬ PP VN ‫إلى الجامعة تذهب صديقتي‬

Notes 1 The author is aware that the term “processing” recently has been used to describe parsing the stimuli as it comes in real time (Harrington 2001, Fender 2001). As a result, current pyscholinguistic research uses methods that can provide insights about what happens in real-time such as self-paced reading/ listening, etc. (Brandl 2013). In this chapter, however, we use the term to describe the final outcome in assigning a syntactic structure; namely, the actor role in a sentence. 2 An extensive review of CM is beyond the scope of this study; however, for a comprehensive and extensive view, see Bates and MacWhinney (1981, 1982, 1989) and MacWhinney (1987, 1997, 2001). 3 For simplicity, the following abbreviations are used: subject (S), object (O), noun (N), and verb (V). 4 The use of ungrammatical sentences has been considered problematic (Mclaughlin and Harrington 1989), a matter discussed later. 5 Abu Radwan (2002) used a 3*4*3 design which yields 36 sentences. However, the total number of stimuli used was 54 sentences. This suggests that he used unequal number of sentences per condition. Moreover, he mentioned that he only used the same two nouns and verb creating all the stimuli (p. 195). Thus, it is not clear how he constructed 54 sentences.

219

Jamil Al-Thawahrih 6 See MacWhinney et al. (1985) for the justification of using ungrammatical sentences. 7 In the reviewed literature, researchers allocated between 3 to 10 seconds for each experimental sentence (Li, et al. 1993). The researchers did not provided a rationale for the exact time provided. However, it was necessary to ensure that the allocated time is sufficient for participants to read the sentence, and for that reason, a pilot study was conducted before running the experiment on the participants. 8 For the sake of simplicity, the following abbreviations are used: (1) to refer to the word order conditions in this study, the following is used: NVN and VNN; (2) to refer to the subject-verb agreement conditions, the following is used: 1N agrees, 2N agrees, and Amb; and (3), for a specific condition yielded from the combination of the two variables (word order and subject-verb agreement), the following is used: NVN_1N, NVN_2N, NVN_Amb, VNN_1N, VNN_2N, and VNN_Amb. 9 While the paired sample t-tests did not reveal any significance difference between the subject-verb agreement conditions in the beginner group, the observed power was low (OP = 0.14; see Table 2). The lack of significance is likely due to the small sample size. 10 An example of VNN sentences that the NSs of English interpreted as VOS is: licking the pens the cat (Bates and MacWhinney 1981, p. 201). It should be noted that Bates and MacWhinney (1981) used ungrammatical sentences in their stimuli.

References Abu Radwan, A., 2002. Sentence processing strategies: An application of the competition model to Arabic. In: D. B. Parkinson and E. Benmamoun, eds. Perspectives on Arabic linguistics, xiii-xiv. Amsterdam: John Benjamins, 185–209. Bates, E., et al., 1982. Functional constraints on sentence processing: A cross-linguistic study. Cognition, 11, 245–299. Bates, E., et al., 1984. A cross linguistic study of the development of sentence interpretation strategies. Child Development, 55, 341–354. Bates, E., et al., 1999. Processing complex sentences: A cross linguistic study. Language and Cognitive Processes, 14, 69–123. Bates, E. and MacWhinney, B., 1981. Second language acquisition from a functionalist perspective: Pragmatic, semantic, and perceptual strategies. In: H. Winitz, ed. Annals of the New York academy of sciences: Native language and foreign language acquisition. New York: The New York Academy of Sciences, 190–214. Bates, E. and MacWhinney, B., 1982. Functionalist approaches to grammar. In: E. Wanner and L. Gleitman, eds. Language acquisition: The state of the art. Cambridge: Cambridge University Press, 173–217. Bates, E. and MacWhinney, B., 1989. Functionalism and the competition model. In: B. MacWhinney and E. Bates, eds. The crosslinguistic study of sentence processing. Cambridge: Cambridge University Press, 3–73. Brandl, A., 2013. Processing strategies by beginning L2 learners of English and Spanish: A crosslinguistic study. Thesis (PhD). Florida State University. Brustad K., et al., 2011. Al-kitaab fii ta‘allum al-‘Arabiyya, part 1 and part 2. 3rd ed. Washington, DC: Georgetown University Press. Fender, M., 2001. A review of L1 and L2/ESL word integration skills and the nature of L2/ESL word integration development involved in lower-level text processing. Language Learning, 51, 319–396. Forster, K. I. and Forster, J. C., 2003. Dmdx: A windows display program with millisecond accuracy. Behavior Research Methods, Instruments, and Computers, 35 (1), 116–124. Gass, S., 1987. The resolution of conflicts among competing systems: A  bidirectional perspective. Applied Psycholinguistics, 8, 329–350. Harrington, M., 2001. Sentence processing. In: P. Robinson, ed. Cognition and second language instruction. Cambridge: Cambridge University Press, 91–124. Kilborn, K. and Cooreman, A., 1987. Sentence interpretation strategies in adult Dutch-English bilinguals. Applied Psycholinguistics, 8, 415–431.

220

Arabic L2 learners’ use of word order Li, P., et al., 1993. Processing a language without inflections: A reaction time study of sentence interpretation in Chinese. Journal of Memory and Language, 32, 169–192. MacWhinney, B., 1987. The competition model. In: B. MacWhinney, ed. Mechanisms of language acquisition. New Jersey: Erlbaum, 249–308 MacWhinney, B., 1997. Second language acquisition and the competition model. In: A. M. B.de Groot and J. F. Kroll, eds. Tutorials in bilingualism: Psycholinguistic perspectives. Mahwah, NJ: Erlbaum, 113–144. MacWhinney, B., 2001. The competition model: the input, the context, and the brain. In: P. Robinson, ed. Cognition and second language instruction. Cambridge: Cambridge University Press, 69–90. MacWhinney, B., 2005. A unified model of language acquisition. In: A. M. B. de Groot and J. F. Kroll, eds. Handbook of bilingualism. Oxford: Oxford University Press, 49–67. MacWhinney, et al., 1984. Cue validity and sentence interpretation in English, German, and Italian. Journal of Verbal Learning and Verbal Behavior, 23, 127–150. MacWhinney, et al., 1985. The development of sentence interpretation in Hungarian. Cognitive Psychology, 17, 178–209. McDonald, J., 1987a. Assigning linguistic roles: The influence of conflicting cues. Journal of Memory and Language, 26, 100–117. McDonald, J. L., 1987b. Sentence interpretation in bilingual speakers of English and Dutch. Applied Psycholinguistics, 8 (4), 379–413. McLaughlin, B., and Harrington, M., 1989. Second language acquisition. Annual Review of Applied Linguistics, 19, 122–134. Mohammad, M., 2000. Word order, agreement, and pronominalization in standard and Palestinian Arabic. Amsterdam: John Benjamins. Taman, H., 1993. The utilization of syntactic, semantic, and pragmatic cues in the assignment of subject role in Arabic. Applied Psycholinguistics, 14, 299–317. VanPatten, B., 2007. Input processing in adult SLA. In: B. VanPatten and J. Williams, eds. Theories in second language acquisition. Mahwah, NJ: Lawrence Erlbaum, 115–135. Year, J., 2003. Sentence processing within the Competition Model. TESOL and Applied Linguistics 3, 1.

221

PART IV

Arabic L2 reading and corpusaided language learning

11 CORPUS LINGUISTICS AND CRITICAL READING AND THINKING Proposals for teaching learning sequences based on journalistic corpora in Modern Standard Arabic Nadia Makouar The aim of this study is to show the possibility to integrate corpus analysis tools into language learning activities along with a critical thinking approach. Many studies in second language acquisition have introduced the use of text analysis tools into the classrooms. Known as datadriven learning, this approach allows L2 learners to read and interpret texts critically and acquire language skills more easily. The approach has been adopted in many languages, but it has yet to be applied in Arabic. To this end, a media corpus was used consisting of news articles that covered the Bahrain and the Egyptian revolutions in 2011, from Al-Jazeera, Al-Sharq Al-Awsaṭ, Al-Wasaṭ, and Al-Masry Al-Yawm. For a theoretical framework, the theory of interpretive semantics has been adopted along with textometry methods (Lexico 3) for interpreting and teaching texts. Through combining theoretical principles with textometry tools, it was possible to characterize the ideological and political orientation of the different enunciators of Arabic media texts. After a careful analysis of the corpus, an exploratory case study was conducted with four French-speaking learners of Arabic. In the course of the study, the participants received the treatment of the new approach over a seven-week sequence, and their performance was qualitatively analyzed. The findings clearly point to the potential of the approach and of using corpus analysis tools and metalinguistic awareness in enhancing participants’ reading and critical reading/thinking skills. The study demonstrated that it is possible to develop a hermeneutic reading process and improve interpretative and knowledge-construction process skills and strategies, as well as build more interdisciplinary bridges in the curricula.

11.1 Introduction Considering the increasing flow of contradictory information in the media, it seems important that language learners gain access to authentic content in class and make critical observation of language data. For example, if we look at media and editorial policies regarding the recent revolutions in Egypt and Bahrain in 2011, we notice that events were not narrated in the same 225

Nadia Makouar

way. In previous research (Makouar 2014), it was shown that a differential approach of the articles could highlight the ideological orientation of the media. The goal of this case study is thus to connect, in particular, reading with a metalinguistic awareness in order to progressively integrate a critical thinking process in Arabic language learning and teaching.1 Apart from the importance of invigorating Arabic language teaching in an academic context, scientific literature has shown that it is crucial to integrate authentic materials, particularly corpora. This is indeed the opinion of Bohas (2010) or that of Imbert (2010),2 for whom it is regrettable to witness the proliferation of Arabic methods being limited to grammar and use, making it a fixed language. Imbert (2010) argues that it is essential to take into account contemporary corpora and accept the evolution of Arabic language in all its variety to teach it as it is spoken and written today. By combining qualitative interpretative semantics theory with textometric methods, this study shows how text analysis of some Arab and pan-Arab media can detect ideological and political implications during the revolution in Egypt and Bahrain. In this chapter I attempt to present an account of media literacy, text analysis, and language learning and reading from an interpretative perspective. I first discuss the background and theoretical framework. Then I present the methodology of analysis by giving some examples of the analysis and the experiment. Finally, I discuss the findings of the experiment and future research.

11.1.1 Background Many studies in second language acquisition (SLA) have introduced the use of text analysis tools into classrooms. Known as data-driven learning (DDL henceforth), this approach allows L2 learners to observe texts differently and acquire language skills more easily (Johns 1991). The approach has been developed over the past three decades (since Johns 1990, 1991) as a new approach and as a research paradigm in applied linguistics. According to this approach, which takes data as a starting point for learning, learners are led to discover and identify the language, as well as the different word uses, through contact with the data, generally gathered in corpus form (Chambers 2010). Tim Johns (1990) introduced the DDL approach to develop the idea that learners could describe and learn language by themselves. Based on this research, others use the concept of “learner as a researcher” (Schaeffer-Lacroix 2014) and “reflective learner” (Chachkine et al. 2013). Practitioners such as Chambers (2005, 2007), Kerr (2013) and Landure (2013) developed the approach in French L2, Johns (1991) in English L2, and Schaeffer-Lacroix (2014) and Vyatkina (2016) in German L2 learning. The objective of the present (case) study is to examine the combination of DDL and interpretative semantic theory as an effective approach in Arabic L2 learning and reading. My hypothesis is that a discourse analysis theory such as interpretive semantics could enable students to notice, analyze, interpret, and discuss the meaningful and purposeful linguistic and rhetorical variation present in the texts and enhance their understanding of what they read. The aim is to raise students’ awareness of media analysis and the representations that the media convey. Wolton (2009, p. 72) argues that “information is always the condition of the critical mind,” and thus “whatever else the nature of the information, the role of the receiver is necessary.” Accordingly, this research intends to combine and include principles of DDL and interpretative semantics theory in Arabic language learning and reading. The tools used to observe and analyze the texts have become increasingly efficient and accessible. There are many tools which can compute frequency statistics, collocates and key word in context (KWIC) concordances such as Lexico 3 (Université Paris 3), TXM (ENS de Lyon), GraphColl (University of Lancaster) etc. 226

Corpus linguistics and critical reading and thinking

Corpus linguistics used for DDL includes quantitative methods to carry out linguistic analysis. It allows linguists to uncover discursive patterns through frequency calculations and statistics. This kind of information structuring gives a non-linear visualization of the text components. From a non-discourse analysis, it allows one to deepen qualitative research in the language, by exploring and analyzing contexts through concordances. As previously mentioned, DDL has already been established as an L2 learning approach in many languages.

11.1.2  Theoretical framework 11.1.2.1  Interpretive semantics Interpretative semantics theory, as founded by Rastier (2005), is defined by Hébert (2006, p. 117) as “a ‘second generation’ synthesis of European structural semantics, developed after the work of Bréal and Saussure, then Hjelmslev, Greimas, Coseriu, Pottier.” Interpretative semantics is a structural approach to language. It focuses on constraints on language and its use in the text considered for establishing different levels of readings. It emphasizes the notion of difference in the context of the signified (or content of a sign). The theory states that no text is only written “in a language”: it is written in a genre within a discourse, obviously taking into account language constraints. The interpretative process theorized in the interpretative semantics theory matches critical pedagogical practices as it allows for progress to take place in the interpretative activity of texts. The critical approach in pedagogy is an imperative that can find its resources in analytical tools as an easier way to perform critical thinking. In our experiment, we use corpus semantics (a combination of interpretative semantics and corpus linguistics) as a theoretical framework in order to analyze texts. We assume that introducing textometry tools into Arabic L2 learning can help students become more critical readers. In Boydston’s compilation of Dewey’s works (1978, p. 188), the latter defined critical thought as: a) “a state of perplexity, hesitation, doubt, and b) as an act of search or investigation directed toward bringing to light further facts which serve to corroborate or to nullify the suggested belief.” By applying the same principles, research on interpretative semantics theory application in the classroom has yielded interesting results in literary texts. Although they did not focus their research in second language learning, Sanfourche (1997) and Mezaille (2003) have shown that the interpretative approach (with or without corpus analysis tools) allowed stimulating interaction and reflection in the L1 classroom. Sanfourche (1997) conducted his research on learning to read with young participants who were fifth-graders, without incorporating corpus analysis tools. By taking a short French novel written by Le Clézio as teaching material, Sanfourche wanted to deconstruct the meaningmaking processes and draw attention to the textual genre in those processes. In his study, he used interpretative semantic concepts (referential impressions, generic and specific themes, etc.) and highlighted the interpretative path that participates in the construction of text meaning. To undertake that, Sanfourche observed the assumptions and reactions of students during the linear reading of the text. This experiment took the form of a series of research studies. His study (1997) on reception is part of a very stimulating dynamic process with students, in which “the main role of the teacher is to encourage this interaction in order to transform individual strategies into shared reading skills.” According to Sanfourche, several interrelated dimensions are involved in the learning and pedagogical process: students’ interactions with (i) the text, (ii) reading and the various interpretations implied by the text, and (iii) their own interpretive textual methods. It enables students (i) to become aware of the facts of textual coherence 227

Nadia Makouar

and cohesion (Halliday and Hasan 1976), (ii) their influence on reading, and (iii) students’ relation to the text. In his study, Sanfourche assigned the teacher three objectives: 1 highlighting interpretative processes (by confronting and explicating them) and valuing students’ knowledge of the world 2 placing the reader in a situation of construction of meaning in order to show that the meaning of a text is not a given, but that it is constructed during reading 3 showing the significance of textual genre in the process of reading. He defined cohesion as a set of linguistic phenomena that link the sentences of a text, “as the formal organization of the text that ensures its semantic continuity” (Sarfati and Paveau 2003, p. 188–189). He considered coherence as an extralinguistic level related to the organization of representations and the “encyclopedic knowledge of the participants, who can then judge the conformity of textual data to the prelinguistic data which constitute their beliefs and their knowledge about the world” (Sarfati and Paveau, 2003, p. 189). Based on these considerations, Sanfourche started his study experiment by reading a little paragraph, sentence by sentence. This method enabled him to create an atmosphere of open discussion among participants. For example, upon reading the first sentence of a paragraph (“Today, August 15, 1963, the young woman named Liana is alone, sitting on the bench covered with dark green moleskin, at the back of the great hall”), he observed that the lack of knowledge of the place mentioned raised a real debate. The use of the definite article surprised and disturbed the participants’ reading. As they read “the great hall,” they reacted: “The narrator acts as if we knew where we are” or “but we have never heard about this place.” Once the second sentence was read, some assumptions made previously were called into question by some participants. By progressively and sequentially working on the text extracts, students discussed and drew upon their prior knowledge by suggesting answers about “grey areas” remaining in the text in order to give “cohesion to their reading,” since it is the semantic relationship between words that generates meaning and raise other questions and hypotheses in the mind of the reader. These were then “challenged” and continued to be challenged “by modifications, adjustments of their first representation.” The participants’ references for interpretation were found either “bookish” or “drawn from everyday experience.” Sanfourche used the notions of coherence and textual cohesion to show their functioning and interdependence in the participants’ reading strategies.3 He named the convergence of these strategies as “competence of integration,” which “operates at different levels: (i) at the sentence, inter-sentential, and textual levels by integrating the semantic data with textual coherence, (ii) at the extratextual level by reciprocal integration of world knowledge and semantics strategies of the text, (iii) at the meta-textual level by integrating strategies built before the new strategies induced, or even required, by other texts.” Sanfourche did not incorporate quantitative measures for evaluating the performance of the participants of his study and limited the study findings to qualitative observations. From the same perspective, Mezaille (2003) conducted his research with high school students (tenth-grade students). Unlike Sanfourche, Mezaille (2003) introduced a corpus analysis tool, Hyperbase (Brunet 2010), for the meaning interpretation process that consists of analyzing the occurrences of a word in literary books. Mezaille focused his research on the thematic analysis of literary texts (Proust, Zola, etc.). He proposed a “thematic construction of words within a Balzacian corpus.” The work given to participants in his study consisted of selecting a key word “joy” in the novel Le Lys dans la Vallée and transitioned from observing to understanding its use in context. Forty occurrences (through concordance lines) were submitted to the participants, 228

Corpus linguistics and critical reading and thinking

who had to examine them 48 hours before an oral examination. The goal was to make them “establish the maximum number of relations between the extracts provided in order to understand their potential unity.” The observation of immediate co-occurrences of “joy” allowed the participants to associate with it notions such as expressed, interiorized, and positive. The two research studies mentioned which adopted interpretative semantics concepts show that it can be applied in various learning contexts, with various audiences and teaching materials. In Sanfourche’s study, students were able to interact and continue to reflect until they understood and identified the places and elements described in the text. Mezaille shows that students succeeded in recognizing semantic variations and thematic relations formed by the occurrence of “joy” in the different extracts of Balzac’s novel. In interpretive semantics, the seme is the smallest unit of meaning. Two types of semes are identified: inherent semes and afferent contextual semes. Rastier (2015, p. 494) defines an “inherent seme” as an attribute having a typical value: inherited by default from type to token, unless contradicted by the context. Each of the semes in a type is an attribute with a typical value. For instance, in “crow,” the attribute (or semantic axis) “color” has “black” as its typical value. “Black” is therefore said to be an inherent seme for “crow.” But a contextual determination could very well prevent this inheritance from taking place and impose an atypical “color” value (e.g., “I see a white crow.”). No inherent seme, therefore, appears in every context. On the other, an “afferent contextual seme” is one which is activated by the linguistic context. The meaning of a word can be “perceived” with interpretative operations such as activation, inhibition, or propagation (Rastier 2015, p. 495). To explain the notions of activation, inhibition, and propagation, consider sentence (1) from our journalistic corpus Al-Jazeera. (1) ‫ في‬،‫شهدت األسابيع األخيرة محاوالت تودد من قبل إدارة الرئيس األميركي باراك أوباما لشبكة الجزيرة الفضائية‬ .‫محاولة لكسر تاريخ من الجمود شاب العالقة بين الجهتين‬ “In recent weeks, US President Obama’s administration and Al-Jazeera have attempted a rapprochement in order to break the historic stalemate that characterizes the relationship between the two sides.” In example (1), context activates the seme “political institution” in “Al-Jazeera.” For the related seme “media,” objectivity is inhibited by the context, as in the sentence, Al-Jazeera makes arrangements with the US administration to “normalize” their relations. Propagation occurs when the seme “state” on “Al-Jazeera” is propagated or distributed, hence the seme becomes a contextual afferent, because in that context especially, “Al-Jazeera” is portrayed as a one-sided and interventionist media among readers and viewers.4 Two other interpretive semantics concepts, architextuality and semantic classes, are also relevant to the present study. Architextuality means that any text placed in a corpus receives semantic determinations and could potentially modify the meaning of each of the texts that comprise it (Rastier 2001b). A semantic class is a semantic paradigm which contains words that share certain semantic properties relative to the context in which they appear. For example, the words “president,” “campaign,” and “election” share the semantic feature: “politics.” No studies using interpretative semantic theory with corpus analysis tools have been explored in second language acquisition. My proposal aims to fill this gap by examining these aspects in Arabic L2 learning. 229

Nadia Makouar

11.1.2.2  Textometry tool Lexico 3 and study corpus For the purpose of the present study, we use the textometry tool Lexico 3, which brings out regularities and disparities between texts and investigates “language facts” and compares them in each corpora.5 Through specificity calculations (hypergeometric test), co-occurrences, and concordances, the tool could confirm or invalidate assumptions that justify corpus building. The statistical results can help in the identification and exploration of language ideologies within corpora. The approach therefore consists of a contrastive analysis of corpus. Thanks to this process, the sense of a unit (noted seme) comes up in contrast with other units. The approach is twintracked: first, we analyze texts with descriptive concepts of corpus semantics, then we apply these methods to Arabic language learning through interpretation didactics by allowing learners to pursue a textual study with authentic data. Corpus semantics can identify regularities and differences in a corpus; a process that can be longer and more difficult with a linear reading. Statistical calculations integrated in Lexico 3 software (Salem et al. 2003) ​​contain many features, such as concordances and statistical calculations, the most used in our study. The software can also help target exploration by the identification of language elements. For example, textual topography is one of the software features, as illustrated in Figure 11.1. We can visualize the global distribution of the phrase ‫درع‬ ‫“ الجزيرة‬the shield of the Arabian peninsula” throughout the corpus. Here, each small square

Figure 11.1 Topography of “‫”درع الجزيرة‬

230

Corpus linguistics and critical reading and thinking

corresponds to a sentence ending with a period; and when it is gray (one occurrence of the phrase) or dark (more than one occurrence of the phrase), it means that the sentence contains a phrase. The separations indicate different corpora: (ajz-bah: Al-Jazeera, aws-bah: Al-Sharq al-Awsaṭ, and Wasaṭ: Al-Wasaṭ). The computer software computed the frequency distribution of the term or linguistic unit within the corpora. The descriptive power of such a representation seems obvious. It facilitates the comprehension and interpretation of corpora by restoring their progression. The software allows by a simple click on the square-sentence of the topography to access the context in which the word or the phrase occurs. We can see that the selected phrase is dispersed in the corpus of Al-Jazeera, whereas it is more concentrated in a part of the corpus of Al-Sharq al-Awsaṭ and almost absent in Al-Wasaṭ. Also, the analysis goes beyond the notion of presence or absence of the word. The darker the square, the more the term is present in the sentence. We choose to study the journalistic discourses during the revolutions in Egypt and Bahrain in 2011 in order to highlight, through media analysis, the ideological and political orientations that influence media around the conflict between the Gulf countries and its regional impact. By so doing, we could describe the meaning of words in context (from a contrastive analysis perspective) as used by four online Arabic news media: Al-Jazeera, Al-Sharq Al-Awsaṭ, Al-Wasaṭ, and Al-Masry Al-Yawm). The focus of the present study is on the impact (direct or indirect) of Arabic news media on events and actors by analyzing their designations in the texts. In sum, incorporating Lexico 3 in the study allowed us, among other things, to perform three main steps: identify repeated segments, identify under- or over-represented phrases in different corpora, and identify co-occurrences and create a textual topography. As far as we are aware, there has been no semantic corpus application in SLA studies, except for the use of concordances and other tools in language learning contexts with various languages (Boulton and Tyne 2014, p. 299–309).

11.1.3 Corpora For the purpose of our analysis, we have compiled two corpora: one about the “Egyptian revolution” and the second about the “Bahraini revolution.” We collected them from four online media sources and selected articles with the keywords ‫“ احتجاج‬protest” and ‫“ مظاهر‬demonstration” directly online, using search engines. The two keywords (“protest” and “demonstration”) were the only keywords chosen in order to be as neutral as possible and as least influenced by our assumptions (see Table 11.1, which displays information about the corpora of the present study). For our analysis and before conducting the study with participants, we made a main assumption. There are differences between journalistic discourses due to the editorial policy

Table 11.1 Corpus information Arabic Online Media Media Corpus

Al-Jazeera

Al-Sharq al-Awsaṭ

National Media

Bahraini Revolution Egyptian Revolution

54 articles 37 articles

59 articles 44 articles

Al-Wasaṭ 34 articles Al-Masry al-Yawm 71 articles

231

Nadia Makouar

partially linked to those holding them (from a financial point of view), and the main questions were the following: • What could characterize the differences in the journalistic discourses used in the corpora of this study? • Do the different media follow the same narrative of the same event at the same time? To answer these questions, we considered the specific terms in the corpus. We compared each of the selected corpora (Al-Jazeera, Al-Sharq Al-Awsaṭ, Al-Wasaṭ, and Al-Masry Al-Yawm) depending on the country (Revolution in Egypt or Bahraini Revolution, according to the studied countries). By using Lexico 3, we could identify in each corpus the repeated segments that characterize it. Then, the observation was carried out in the concordance (in context) in order to visualize their co-occurrences for interpreting the meaning of the selected words (see Figure 11.2, illustrating one of the concordances). By using this methodology and the concordances, we could notice, for example, that Al-Jazeera was not portrayed in the same way within the subcorpora, as illustrated in examples (2)–(4). (2) In al-Sharq al-Awsaṭ: .‫ بل وله عالقات دولية‬،‫”الجزيرة” تحولت إلى حزب اسمه “حزب الجزيرة” له “مريدوه” وأنصاره‬ “Al-Jazeera became a party whose name is ‘Al-Jazeera party’ which has its ‘partisans’ and its allies and even international relations.” In this context “Al-Jazeera” is: “political party” (3) In al-Sharq al-Awsaṭ: .‫كانت ”الجزيرة“ تضرب األرقام ضعفين أو خمسة أضعاف بالنسبة ألعداد المتظاهرين أو القتلى‬ “Al-Jazeera multiplied by two to five times the number of protestors and victims.” In this context “Al-Jazeera” is: “troublemaker,” “liar”

Figure 11.2 Concordances of ‫الجزيرة‬

232

Corpus linguistics and critical reading and thinking

(4) But in Al-Masry Al-Yawm: ‫ و الجزيرة تنقله مباشرة‬..”‫اإلعالم المرئي في مصر يتجاهل “يوم الغضب‬ “The media in Egypt ignores the ‘Day of Rage’ . . . while Al-Jazeera relays it live.” In this context “Al-Jazeera” is: “fair,” “accurate,” “solidarity.” In the case of the Arab revolutions, it is important to know the ideological orientations that led the media to be in favor or against the protest movements. The approach is thus intra- and intertextual and also allows for a contrastive analysis of the corpora. This method seeks to compare the contents of the texts of different genres within the same discourse, allowing for the emergence of meaning units in a given corpus. Regarding non-linear reading of a corpus, and according to Boulton and Tyne (2014, p. 73), the concordance appears as the tool that is mostly used and widespread in the classroom. It is simple to handle and provides targeted information about a corpus. In addition, TogniniBonelli (2001, p. 41) argues that various types of readings (vertical, fragmented, and focused on repeated segments) must be associated with a new methodology: when it comes to teaching with corpus data, there is a new methodology to teach: going back to the parameters differentiating the reading of a text from that of a corpus; [. . .] we can say that students have to be taught to read vertically, to read fragmentarily, and to focus on repeated events. This relationship between new reading practices and critical thinking matches the philosophical paradigm of knowledge. In the context of language learning, two paradigms are generally in opposition: knowledge and competence:6 [T]he appropriation of knowledge always involves an adaptation that is realized in a course of action. The exercises which allow this appropriation do not apply the theory, but contain it; in another words, “in the practical state.” [. . .] [K]nowledge is then nothing but learning, within a social practice. (Rastier 2001a, p. 161). In that context, we favor the knowledge paradigm that leads to rethinking the relationship between the learner and the interpretative practice of authentic texts. This seems to be more crucial so far as the corpus that we submit to learners is a journalistic corpus and the effects of their reading involved in the development of representations of “immediate history.” Indeed, according to Boulton (2009 cited in Johns 1988): “text . . . and the learner’s engagement with text should play a central role in the learning process. In that engagement, a key concept is that of authenticity, viewed from three points of view – authenticity of script, authenticity of purpose, and authenticity of activity.” Interpretation of meaning uses the principle of contextualization. The fact of bringing together texts generates new senses that we have not noticed before. The meaning implies a maximum contextualization both by language (the whole text) and by the situation of communication. The meaning can be represented as an interpretative path (Rastier 2001b). The attribution of meaning must result from an interpretative path by taking into account the cotext, the context, and intertext notions. 233

Nadia Makouar

11.2  Research questions This chapter uses a case study approach to explore reading and media literacy and textual analysis from an interpretive perspective while incorporating corpus analysis tools. The research questions are: 1 Is the incorporation of corpus analysis tools and a metalinguistic awareness effective in improving reading and enhancing critical reading/thinking skills? 2 What are the main challenges faced by participants from such an approach?

11.3 Methods 11.3.1 Participants Participants of the present study were four French-speaking learners of Arabic learning Arabic as an L2 in their own home institution in France. The four participants (two males and two females) were all undergraduate students and were placed by their academic institution at the intermediate proficiency level, based on language instruction of 2–4 years. Three participants had two years of Arabic formal exposure (Sophie, Taymour, and Khadija) and one had three years (Hicham). The first year is equivalent to 7.5 contact hours per week, the second year 9.5 contact hours per week, and the third year 11.5 contact hours per week. Their age range was 20–22 with the age mean of 21 years. Three of the participants were heritage Arabic speakers and one was a non-heritage L1 French speaker.

11.3.2 Procedure The procedure consisted of seven sessions or sequences where participants received exposure to the approach of incorporating corpus analysis tools and metalinguistic awareness in reading. The duration of each sequence was 1 hour and 30 minutes. The purpose of sequences 1–2 was a gradual introduction to the theoretical concepts: afferent contextual and inherent semes. For these sequences the input was all the corpus. We proposed to work on two types of documents: the concordances and texts. Before the study began, the participants were asked some questions to test their knowledge of journalistic genres and discourses. This was followed by a sequence of seven sessions held over seven weeks with a one-week interval. The topics and assignments covered during the seven sessions/sequences are listed in Table 11.2. Through written and comprehension exercises, we attempted to articulate the statistical tool possibilities with the learning sequences. To understand the dynamics of the texts that they read, participants were provided with the necessary theoretical and practical tools. Sequences 1 and 5–6 are discussed in details here. The observation of the impact of the treatment in the various sequences is demonstrated in the various responses of the participants to answers posed to them following a given sequence.

11.3.1.1  Sequence 1: introduction to inherent and afferent semes In sequence 1, we wanted to show the semantic incidence of co-occurrences. One of the tasks was to observe the different neighboring words of ‫“ ثورة‬revolution” and understand if it had the same meaning throughout the text (see Appendix). Below are sample responses from three 234

Corpus linguistics and critical reading and thinking Table 11.2 Schedule of learning sessions/sequences Sequence

Input

Interpretative Semantics Notions

Output

1

Texts and concordances

Inherent and afferent semes

2

Texts and concordances

Inherent and afferent semes, semantic classes

3

Texts and concordances

Inherent and afferent semes, semantic classes

4

Videos in relation to texts in the “Egypt corpus”

Afferent and contextual semes

5

Large contexts of occurrences (paragraphs) from opinion and informative article “Egypt corpus and statistics tables”

Contextually and socially normed semes; dialogics

Writing (in French) answers about the meaning of words in contexts Writing answers about the meaning of words in contexts; translation into Arabic; reformulation in Arabic Writing (in Arabic) an opinion text: in the “ reader’s corner” Answering analysis questions; translation into French Answering analysis questions in Arabic and French

6

7

“Egypt corpus” and statistics tables; French article about Qatar and Saudi Arabia

Semantic classes; intertextuality; architextuality Semantic classes; intertextuality; architextuality

Answering analysis questions in Arabic and French; formulating hypotheses Answering analysis questions in French

participants, answering the question (in French): Do you think that the word ‫ ثورة‬has the same meaning throughout the text? Sophie:

We see that according to its co-occurrences, the meaning of the word changes. It seems that in the early occurrences there is the idea of a popular revolution for social or political change, while in the last text the word refers rather to an abrupt change in technology. This creates a parallel or even gives the idea of complementarity between the uprising and the emergence of new communication technologies.

At this stage, Sophie encounters the overall context of the word (whole text) but also the concordances. She can observe the meaning units being superimposed on each occurrence. She could thus make a rapid and targeted observation of the context, analyze the associations, and make assumptions about the meaning of the word. Taymour: The meaning depends mainly on the context and the co-occurrences that surround it. The title of the text, where the term is quoted, evokes both the idea of “Arab Revolutions” made by the internet but also of the progress of the means of communication that have allowed this. Twice the term clearly refers to the 235

Nadia Makouar

Khadija:

revolution as a popular movement. At other times it is rather a mockery or to insist that it is a “revolution of technical progress” that has allowed these revolutions. [Taymour goes so far as to refer to the generations’ conflicts as] the “revolution of generations.” The general meaning of the term remains more or less the same; it is the idea of ​​a change or upheaval, here lived by the author as negative. No, throughout the text, this word takes different meanings which may be literal or figurative. In the article, when the word is followed by a country name, it is often in the literal meaning: “revolution” in the sense of a group of persons who protest against a force that oppresses him.

Thus, participants noticed the contextual afferences, taking up the linguistic vocabulary (meanings, co-occurrences), sometimes inaccurately, but their analyses are interesting. Sophie wrote that it is about “creating a parallel, or even giving the idea of complementarity” between the two meanings of “revolution.” For Taymour, “the general meaning of the term is more or less the same; it is the idea of a change or upheaval here lived by the author as negative.” Thus, their metalinguistic awareness helps them in determining meaning.

11.3.1.2  Sequence 5: introduction to the concept of intertextuality and textual genre and its impact on the enunciative phenomena For this sequence we wanted to test a theoretical concept of interpretative semantics, which is intertextuality and textual genre, which considers the following: “No text is only written ‘in a language’; it is written in a genre within a discourse, obviously taking into account the constraints of a language” (Rastier 2005, p. 33). We attempted to test these “constraints” of language by the intertextuality principle as a relation between two texts that leads to a relation of semantic interdependence between two texts. Participants were presented with two sub-corpora of two different genres (opinion articles and informative articles). For this sequence, we asked students to observe the formulation of the topic “internet” (see Table 11.3 for samples of short extracted texts). In this sequence, the participants were asked to compare the use of verbs between informative and opinion articles. The aim was to show how a contrastive analysis of genres could make morphosyntactic patterns more easily perceptible and could highlight the link between enunciation and genres. The occurrences and their contexts were extracted with the “topography” function in Lexico 3. Participants noticed differences in verbs and tenses. Sophie did not identify the tenses but noted that there were differences about the topic description. The sequence progression allowed the participants to point out these features which began to emerge from the second question (posed to the participants): Compare between the extracts of informative articles and opinion articles; is the topic of the internet perceived in the same way? (See Appendix.) Participants’ responses showed how word meanings and morphosyntactic features were highlighted by the phenomenon of intertextuality (comparison between texts and between different genres). Below are sample responses from three participants. Sophie:

The verbs of the first section are rather neutral and informative; they are “general truth,” whereas those in the second section are strongly affirmed; they are also more descriptive. We can conclude that the statements in the first section are objective, whereas the statements in the second section are more argumentative. 236

Corpus linguistics and critical reading and thinking Table 11.3 Short extracts of informative and opinion articles from “Egypt Corpus” Extracts from informative articles

Extracts from opinion articles ،‫ مثل ثورة تونس‬،‫ثورة “ويكيليكس” التي ما زالت تعتمل‬ ‫ هي أيضا غصن فارع من‬،‫ويعد ربانها (أسانج) بالمزيد‬ ‫أغصان شجرة اإلنترنت‬ ‫فقد استفادت “القاعدة” بشكل كبير من عالم اإلنترنت‬

‫الحكومة تقطع خدمات الهاتف الجوال واإلنترنت عن‬ ‫سكان القاهرة‬ ‫ انقطع موقعا “ تويتر” و “فيس بوك” على‬، ‫من ناحية أخرى‬ ‫ واتسمت خدمة اإلنترنت في‬، ‫مدار اليوم أكثر من مرة‬ ‫ كما توقفت خدمة “بالك بيري” في عدد‬، ‫مصر ببطء شديد‬ ‫من المحافظات مساء أمس‬ ‫وترددت أنباء على موقع “تويتر” قبل ٕاغالقه عن منع‬ ‫االتصاالت بشكل كامل اليوم (الجمعة) في مصر سواء عبر‬ ‫الهواتف األرضية أو المحمولة واإلنترنت على مدار اليوم‬ ‫في محاولة حكومية لتقليل التواصل بين المتظاهرين‬ Co-occurrences of internet : protesters, government, cut services, Blackberry, Facebook, twitter, phone(s) Verbs: to cut, to stop, to decrease Afferent semes activated in context: “threat to the authorities”

(‫بالذات بعدما اندمج اإلعالم التقليدي (تلفزيون وصحف‬ ‫باإلعالم الحديث (موبايل و ٕانترنت) وأصبح الكل يخاطب‬ ‫الكل ويحرص على احتكار‬ ‫ٕاخراج المشهد‬ Co-occurrences of internet : revolution, world (of the internet), old communication means, modern, mobile, everyone, “WikiLeaks” revolution, using old communication, modern, mobile, everyone Verbs: to become, to benefit Afferent semes activated in context: “Modern,” “spreading of ideas”

Taymour: In the informative article extracts, specific events are described, at defined times. So we noticed the quite exclusive use of the past tense maaḍii, whereas in opinion extracts, phrases are written with a freer style, mixing the past maaḍii and the present muḍaariʕ. There is thus a style that could be described as more journalistic in the “informative” articles. Hicham: In the extracts of informative articles, verbs are often in the past tenses and precede statements or citations, whereas in the opinion articles extracts, we have more variations between the past tense and the present tense because the second mode allows the author to make statements, report habits, and assert general truths. Thus, based on the participants’ responses, we can observe through learners’ observation how intertextuality allows a critical look at the textual content and the constraints of media writing practices. The discourse and genre as linguistic notions have an impact on the learner’s reading practice.

11.3.1.3  Sequence 6: architextuality and semantic classes In this sequence, the goal for the participants was to work on architextuality and semantic class concepts. This part of the sequence requested the participants to distinguish word meanings and highlight their understanding in Arabic. They were asked to formulate hypotheses on specific words in each corpus (i.e., specific occurrences). The input consisted of the statistical specificities of event-related words in the “Egypt corpus” (see Table 11.4). As can be seen from Table 11.4, there is a strong opposition in the scores between the two pan-Arab news media: Al-Jazeera and Al-Sharq Al-Awsaṭ. The Egyptian media, Al-Masry 237

Nadia Makouar Table 11.4 Statistical specificities of the event-related words within the “Egypt corpus” Expressions ‫احتجاج‬ protests ‫االحتجاجات‬ The protests ‫ثورة‬ revolution ‫الثورة‬ The revolution ‫المسيرة‬ The walk ‫االنتفاضة المصرية‬ The Egyptian Intifada ‫المواجهات‬ The clashes ‫الثورة المصرية‬ The Egyptian revolution ‫الثورة الشعبية‬ The popular revolution ‫مظاهرات‬ Demonstrations ‫ ثورة‬25 ‫يناير‬ The January 25 Revolution

Al-Jazeera

Al-Sharq Al-Awsaṭ

Al-Masry al-Yawm

9

-9

-3

7

-6

7

-8

8

4 3

5

-5

5

-5

-3

3 3

-3

3

-4

-3

3 -4

6

Al-Yawm, uses less often the words “(The) protest (s)” or “revolution,” which are more significant in the Qatari media, Al-Jazeera. The only positive specificity of the Saudi media, Al-Sharq Al-Awsaṭ, is “demonstrations.” Thus, the semantic unit of “fight” seems more present in the articles of Al-Jazeera. Indeed, the words “intifada,” “clashes,” and “revolutions” are more specific in Al-Jazeera corpora and show that the media is trying to highlight the uprising of the Egyptian people by diversifying the terms and adding qualifiers like “popular” and “Egyptian.” The analysis of the specificities is based on the use of the hypergeometric model (or Fisher’s test, see Lafon 1980). Positive specificities are the numbers that go far beyond what the model predicted, and negative specificities the numbers that are clearly lower than what the same model could hope for. These diagnoses are linked to a score of specificity which makes it possible to measure the differences observed in relation to what the model predicted. The higher the diagnosis, the greater the difference. For example, in the corpus, “The walk” is over-represented in Al-Jazeera corpora, whereas it is under-represented in Al-Masry al-Yawm and even more so in Al-Sharq Al-Awsaṭ. A negative specificity could mean that the word is absent from the corpus, but not necessarily. In this stage of analysis, we can assume this is a low frequency word or an absent word. To make sure, we have to check in context (with the concordances). Following explanation of the principle of the statistical specificities, the participants were presented with three tables and were asked to compare the more specific words in each media 238

Corpus linguistics and critical reading and thinking

source, semantically classify such words, and make assumptions based on their own observations (see Appendix). Below are samples responses from three of the participants. Hicham:

Sophie: Khadija:



1) According to the editorial policy, the words that are chosen to deal with the same subject are not the same. For example, while Al-Jazeera mostly used the word “revolution,” Al-Sharq Al-Awsaṭ used “riots.” 2) We can assume that Al-Jazeera supports the Egypt revolution; that is why the articles support them and call into question the authorities’ reactions toward protestors, whereas Al-Sharq Al-Awsaṭ obviously is not supporting the Egyptian movement, in which they saw the incivility more than a legitimate revolution. 1) I think we do not find the same results from the two newspapers, because their analytical angles are different and the type of language used and the frequency of a few words are different. 2) By observing these statistical word tables, I think that Al-Jazeera is written with a tone of protest while Al-Sharq al-Awsaṭ is more focused on events and uses a less diverse vocabulary in the discourse. 1) The results are different from one newspaper to another because the terms used by a newspaper depends on its geographical location in relation to the event mentioned, the diplomatic relationship between the country in which the newspaper is established and also where the facts described have taken place. 2) We can formulate the following hypothesis: the media studies do not portray the Egyptian revolution in the same way.

Accordingly, we can tell here that the participants succeeded in distinguishing the meaning of words and their specific uses according to the newspapers. This allowed them to elaborate a reflection beyond the surface language and, thus, to formulate hypotheses about the circumstances of appearance of words influenced by many situations (political, diplomatic, ideological, etc.). They established the link between language and journalistic practices on a specific topic and a deeper level. This corpus analysis enriched their knowledge about the conditions of enunciation and made them aware of possible formulations existing within a social practice.

11.4.  Discussion of findings Since the present exploratory study adopts the case study approach, qualitative observations of the participants’ performance was primarily relied on in gauging their performance as a result of the treatment in the different sequences. Quantitative measures to gauge learning gains from the treatment during the seven sequences, such as employing a pre- and post-test statistical design, were not adopted here, since it would not be reliable to attribute learning gains on the basis of such measures. The participants were also enrolled in other classes learning Arabic, and their number is small to warrant meaningful statistical results. Based on participants’ interaction with the approach and their specific responses, samples of which were extracted and discussed earlier, the study has shown that the treatment during the seven sequences shows the potential of incorporating corpus analysis tools and metalinguistic awareness in reading and textual understanding. Thus, the answer to the first research question about the effectiveness of incorporating corpus analysis tools and a metalinguistic awareness effective in improving reading and enhancing critical reading/thinking skills seems 239

Nadia Makouar

to be a positive one. Participants’ responses do indicate that the approach was beneficial in many respects to their reading understanding of texts and word meanings. As for the second research question, about the specific challenges faced by participants in the implementation of the approach, a debriefing survey was administered to the participants (see Appendix 2). The results of the survey showed the participants found four aspects difficult. First, one of the four participants found the pace in each sequence too fast and indicated she needed more time for each sequence. Second, two participants stated that they have been limited by the absence of short vowels in the corpus. To address such a challenge, it would be advisable to work with vocalized texts when implementing the approach. Third, all participants seem to prefer to gradually understand the specific aspects of the texts before they proceed to the big picture about the text. This is probably due to the non-advanced proficiency level of the participants, where it is widely attested in the L2 reading literature that unlike advanced and proficient/native readers, L2 learners at lower levels prefer a bottom-up reading approach to reading texts over a top-down approach. Finally, a challenge was identified in relation to use of video media introduced in Sequence 4, where all the participants except one (the one with three years of Arabic background) found the sequence difficult. This could be due to the listening skills of the participants not being on a par with that of reading.

11.5  Conclusion and pedagogical implications The aim of the study was to show how the combination of qualitative and quantitative tools could improve reading and critical thinking/reading skills. We note that our first experiment with the seven sequences discussed earlier is encouraging despite the particularity of the tool originally used by linguists. In order to conduct a quantitative analysis, future research should include more participants and for a longer period. Therefore, we are considering an application in an institutional setting where learning proposals will be conducted using action research methodology as an iterative within a longitudinal approach. This will certainly help to better adapt the method, to enable learners to use the textometric tool and evaluate more precisely the acquisition of the lexicon, its reinvestment into production, and to identify the different stages of interpretation process of the learner. This project would correspond to more integration in an interdisciplinary context. The adaptation of the analysis and results of the political corpus will find a place in this type of process. The content of the study also tends to evolve. It is about re-examining the relation between text and video media support (already implemented in sequence 4 of the experiment) to make more operational and more intuitive the interpretative movement between text and extra-textual information related to the corpus approach. We also intend to develop sequences from parallel corpora, defined by Teubert (2004, p. 188) as “the repositories of source language units of meaning and their target language equivalents.” Sequences from this type of corpora, by linking the units in the source text with units of the target texts with tools, could, in addition to developing L2 reading, enhance L2 writing development and complete courses in Arabic translation.

240

Corpus linguistics and critical reading and thinking

Appendix: Sample Sequences 1, 5–6 Sequence 1/Resources: text from Al-Sharq al-Awsaṭ (‫)حقا إنها «ثورة» االتصاالت‬ I. Preliminary questions (Collaboratively) • What genres or types of text can we find in a newspaper? • Among the kinds of text you mention, note the elements that differentiate them. II. Study of the text (individually) A. First approach to the text (You can read the first page only for the entire sequence) 1. When was the text written? What does the title of the text evoke for you? 2. What is the text about? 3. Does it seem to be objective or subjective? What are the main enunciation marks on this text? With the raised marks, can you tell what is the genre of this text? B. Understanding questions/Resource: Concordances of “‫( ”ثورة‬all the occurrences of this word on the text from Al-Sharq al-Awsaṭ (‫)حقا إنها «ثورة» االتصاالت‬ First, without referring to the text, tell what does the word “‫ ”ثورة‬mean to you? 1. In what type of text did you read it? 2. Now look at the word “‫ ”ثورة‬in the concordances and read its neighboring words (co-occurrences): • What do these co-occurrences mean to you? • Does “‫ ”ثورة‬always have the same neighboring words? • According to you, does “‫ ”ثورة‬have the same meaning throughout the text? • Why ? Sequence 5/Resources: Extracts from Opinion and Information sections in the online newspaper al-Sharq al-Awsaṭ. In this sequence, we will use extracts from information articles (section: ‫ )األخبار‬and others from opinion articles (section: ‫ )الرأي‬of the Saudi newspaper al-Sharq al-Awsaṭ.

Part 1: Extracts from ‫“ األخبار‬information” For the questions below, several answers are possible according to the extracts presented. Please answer each time by taking into account all the extracts. Create as many rows as you found answers. ‫إلى ما يؤ ّدي انقطاع‬ ‫خدمة اإلنترنت ؟‬

‫من الذين يريدون‬ ‫قطعه ؟ لماذا ؟‬

‫من هم الذين يستعملون لماذا يستعملونه ؟ كيف يسمى اإلنترنت ؟ ما هي مشكلة وسيلة‬ ‫االتصال هذه ؟‬ ‫اإلنترنت ؟‬

Part 2: Extracts from ‫الرأي‬ « opinion » ‫ كيف يوصف اإلنترنت في هذه المقتطفات ؟‬1 . ‫ هل قضية اإلنترنت تعالج بنفس الطريقة ؟‬.‫ قارن بالمقتطفات لألخبار‬2. ‫ كيف يتكلّم الكتاب في صفحتي األخبار والرأي ؟‬،‫ بصفة عامة‬3 . 241

Nadia Makouar

What differences do you see in the use of verbs and tenses between the “information” extracts and the “opinion” extracts? (Answer in French)

Sequence 6 / resources: entire corpus and tables of specificities The corpus on which we are going to work today consists of a set of texts from three different newspapers. The media texts were selected by keywords and deal with the revolutions in Egypt in 2011. Two of these newspapers are pan-Arabic: al-Jazeera and al-Sharq al-Awsaṭ. The third text is an independent Egyptian national newspaper Al-Masry Al-Yawm. The corpus is large (equivalent to 200 pages in a Word document). It is however possible to consult the words and to know each instance in which journal they are present. You will have three tables available. Each table provides information about the media and their contents. Here is an example of a table showing the specific words or expressions in Al-Jazeera corpora.

Figure 11.3 

242

Corpus linguistics and critical reading and thinking

The first column indicates the specific words ranked in decreasing order of specificity. The other two columns show respectively the total frequency of the expression in the whole corpus and the frequency of the expression in the selected part. Here the selected part is the sub-corpora Al-Jazeera. Specificity scores (last column) are obtained by using a probabilistic test which allows to observe which expression are over or underused in a sub-corpora relative to the entire corpus. According to your knowledge about the newspapers and the countries concerned (Saudi Arabia, Qatar, and Egypt), carefully observe the specificities of each sub-corpora; • Highlight expressions that seem interesting to you to remember or to see later on the concordances; • Then compare the tables, two by two. • If necessary, do not hesitate to look at the words in context (write down which ones you have consulted) 1 What observations can you make after comparing the tables? 2 Are there expressions that you can bring together under the same semantic class? Which ones? 3 Why do we not find similar results from one table to another (from one journal to another)? 4 What hypothesis can you make about these differences between newspapers?

Appendix 2: Final Survey  1 Did you find the theme of revolutions stimulating?  Yes, very much  Yes, rather  Mildly  No, not at all Why ?  2 Did you find the corpus stimulating?  Yes, very much  Yes, rather  Mildly  Not at all Why?  3 What difficulties did you encounter during the written production?  None  Lack of vocabulary  Lack of ideas  Incomprehension of some words in the corpus  Other:  4 Which (other) resources did you use for your production?  The corpus  Arabic dictionary  Arabic-French dictionary (print format)  Arabic-French dictionary (online) 243

Nadia Makouar

 Discussions with the other participants  Other : Which (other) resources did you use for your understanding?  The corpus  Arabic dictionary  Arabic-French dictionary (paper)  Arabic-French dictionary (online)  Discussions with the other participants  Other : Which difficulties did you encounter during the understanding process?  None  Incomprehension of some words in the corpus  Other : What is the advantage of studying the corpus with the textometric tool?  None  It is time saving for finding a specific word or expression  It is a good way for comparing texts with the tables of specificities  This is useful to have all the dictionary of the corpus  Other: How did you perceive the integration of videos for the understanding of the corpus in sequence 4?  That was not useful for the understanding of the corpus  This is good illustration to comprehend the corpus  This is a new approach in my Arabic learning experience  Other: In this comparative approach of texts, have you more easily perceived the specific dimensions of Arab culture (especially on practices in families, regarding, for example, the means of communication, the question of authority, etc.)?  Yes, the approach allowed me to have an easy access and perception of certain facts related to the Arab culture  No, the approach didn’t allow me to have an easy access and perception of certain facts related to the Arab culture  Other: Why ? What was the advantage of working with the other participants?  None  Sharing of ideas  Help for vocabulary (written or oral)  Help for understanding the sequences  Other : In the comparative approach of the articles “opinion” and “ information” (sequence 5), did you perceive the differences of enunciation?  Yes, easily  Yes, quite easily  Mildly  Difficult  Very difficult Why ?  

 5

 6

 7

 8

 9

10

11

244

Corpus linguistics and critical reading and thinking

12 Sort the sequences according to the level of appreciation 1 for the one you liked the most, 7 for the one you liked least Explain your first choice 13 Sort the sequences according to the level of understanding 1, for the one you understood the most, 7 for the one you understood least Explain your first choice 14 What sequence did you feel the less comfortable with? Why? 15 Which features did you find useful?  None  Concordances  Topography  Specificities  Dictionary  Other 16 On which other textual genre(s) would you like to work with this tool? 17 How would you benefit from the tool and the comparative method? (What kind of study would you do?) 18 What suggestions would you make to improve the project? 19 Are you satisfied with the project?  Yes, very much  Yes, rather  Mildly  No, no at all For which reasons? 20 Do you think you have progressed in written production?  Yes  No 21 Do you think you have progressed in understanding?  Yes  No 22 Do you have any other comments?

Notes 1 Although the study has a subcomponent writing part, focus here is on the reading process. 2 “The question that many teachers raise can be formulated thus: should we still teach some unusual rules of grammars and ignore other very common uses because they are not described?” (Imbert 2010, p. 58). 3 Sanfourche (1997) explains: “We discover quite simply and paradoxically, thanks to the segmentation in sentences, that (i) inter-phrastic relations are also assured by lexical coherence, (ii) there is no understanding without textual cohesion, and no cohesion without coherence, and (iii) readers’ strategies have taken into account these two essential dimensions of the textuality.” 4 For other examples, see Hébert (2006) and Rastier (2015, p. 31). 5 Developed by the SYLED-CLA2T team at the Sorbonne-nouvelle University, Paris 3. 6 Upon comparing the two approaches, Hirtt (2009, p. 23) states: “In constructivist pedagogy, the most important thing is not that the student succeeds, but he uses his work (and his possible errors) to progress in the discovery and mastery of knowledge. In the competency-based approach, progress in knowledge is not an objective in itself. Only the result matters.”

245

Nadia Makouar

References Bohas, G., 2010. Norme et violation de la norme. l’accord du verbe avec son sujet en arabe standard, In : N. Abi-Rached, ed. Normes et marginalités à l’épreuve. Alsace, Franc: Presses Universitaires de Strasbourg, 179–188. Boulton, A., 2009. Data-driven learning: Reasonable fears and rational reassurance. Indian Journal of Applied Linguistics, 35 (1), 81–106. Boulton, A. and Tyne, H., 2014. Des documents authentiques aux corpus: Démarches pour l’Apprentissage des Langues. Paris, France: Didier. Boydston, J. A., 1978. The middle works of John Dewey, 1899–1924: Journal articles, book reviews, miscellany in the 1910–1911 period, and how we think. 6. Carbondale, IL: Southern Illinois University Press. Brunet, E., 2010. HYPERBASE Manuel de référence. Available from : ftp://ancilla.unice.fr/manuel.pdf [Accessed 06 January 2018]. Chachkine, E., F., et al., 2013. Pour un apprenant réfléchissant. Linguistik Online 60 (3/13), 23–42. Available from: www.linguistik-online.de/60_13/chachkineDemaiziereSchaeffer-Lacroix.html [Accessed 20 September 2017]. Chambers, A., 2005. Integrating corpus consultation in language studies. Language Learning and Technology, 9 (2), 111–125. Chambers, A., 2007. Popularising corpus consultation by language learners and teachers. In: E. Hidalgo, et al., eds. 6th International conference on teaching and language corpora (TaLC 6), 4–7 July 2004 University of Granada, Spain. Amsterdam: Brill, 3–16. Chambers, A., 2010. What is data-driven learning? In: A. O’Keeffe and M. McCarthy, eds. The Routledge handbook of corpus linguistics. London, UK: Routledge, 345–358. Halliday, M., and Hasan, R., 1976. Cohesion in English, London: Longman. Hébert, L., 2006. Tools for text and image analysis: An introduction to applied semiotics. Paris: ADAGP. Hirtt, N., 2009. L’approche par compétences: une mystification pédagogique. L’école Démocratique, 39, 1–34. Imbert, F., 2010. Enseigner la grammaire arabe à l’université: réforme et devoir de réalisme linguistique, In: V. Aguilar, et al., eds. Primero Congreso Internacional sobre Enseñanza del Árabe como Lengua Extranjera (ARABELE2009), 25–26 September 2009 Murcia, Spain: Ediciones de la Universidad de Murcia, 47–62. Johns, T. 1988. Whence and whither classroom concordancing? In: P. Bongaerts, P. de Haan, S. Lobbe & H. Wekker, eds. Computer applications in language learning. Dordrecht: Foris, 9–27. Johns, T., 1990. From printout to handout: Grammar and vocabulary teaching in the context of datadriven learning. CALL, 10, 14–34. Johns, T., 1991. Should you be persuaded: Two samples of data-driven learning materials. English Language Research Journal, 4, 1–16. Kerr, B., 2013. Grammatical description and classroom application: Theory and practice in data-driven learning. Bulletin Suisse de Linguistique Appliquée, 97, 17–39. Lafon, P., 1980. Sur la variabilité de la fréquence des formes dans un corpus. Mots, 1 (1), 127–165. Landure, C., 2013. Corpus et dictionnaires: Prise de conscience linguistique chez des apprentis LANSAD. Bulletin Suisse de Linguistique Appliquée, 97, 59–76. Makouar, N., 2014. Les “mots” des révolutions: étude contrastive d’un corpus de journaux arabes, In: M. Oualdi, et al., eds. Les ondes de choc des révolutions arabes. Beyrouth, Liban: Presses de l’Ifpo, 181–197. Mézaille, T., 2003. Thématiques littéraires: enseignement des textes numériques. Revue Texto [online]. Available from: www.revue-texto.net/Reperes/Themes/Mezaille_T/Mezaille_Thematiques.html [Accessed 20 September 2017]. Rastier, F., 2001a. Sémiotique et sciences de la culture. Linx-Revue des Linguistes de L’université Paris X Nanterre, 44, 149–168. Rastier, F. 2001b. Éléments de théorie des genres. Revue Texto [online]. Available from: www.revuetexto.net/Inedits/Rastier/Rastier_Elements.html [Accessed 20 September 2017]. Rastier, F., 2005. Enjeux épistémologiques de la linguistique de corpus. In: G. Williams, ed. La linguistique de corpus. Rennes: Presses Universitaires de Rennes, 31–45. Rastier, F., 2015. Interpretative Semantics. In: N. Riemer, ed. Routledge Handbook of semantics. London: Routledge, 491–506.

246

Corpus linguistics and critical reading and thinking Salem, A., et al., 2003. Lexico3: Outils de statistique textuelle. Manuel d’utilisation. Lexi&co [online]. Available from: http://lexi-co.com/ressources/manuel-3.41.pdf [Accessed 20 September 2017]. Sanfourche, J.-P., 1997. Stratégies de lecteur et compétences de lecture: la contribution de la sémantique textuelle à la formation des lecteurs. Revue Texto [online]. Available from: www.revue-texto.net/Reperes/Themes/Sanfourche_Strategies.html [Accessed 20 September 2017]. Sarfati, G.-É. and Paveau, M.-A., 2003. Les grandes théories de la linguistique. Paris, France: Armand Colin. Schaeffer-Lacroix, E., 2014. Impact de discussions métalinguistiques sur l’apprentissage de la production écrite en allemand, langue étrangère. Linx-Revue des Linguistes de L’université Paris Ouest Nanterre La Défense, 72, 193–211. Teubert, W., 2004. Units of meaning, parallel corpora, and their implications for language teaching. Language and Computers, 52 (1), 171–189. Tognini-Bonelli, E., 2001. Corpus linguistics at work. Amsterdam: John Benjamins. Vyatkina, N., 2016. Data-driven learning for beginners: the case of German verb-preposition collocations. ReCALL, 28 (02), 207–226. Wolton, D., 2009. Informer n’est pas communiquer. Paris, France: CNRS éditions.

247

PART V

Arabic L2 writing Discourse analysis and measuring production

12 WRITING IN ARABIC Discourse analysis and pedagogical reflections Dris Soulaimani

Second language learners face challenges in their writing even at the advanced level (Silva 1993). Research in second language acquisition has investigated these challenges at the linguistic structural level (Erris 1994, Lu 2011, Myles 2005, McNamara et al. 2009). Less focus, however, has been placed on the discursive properties of the target language, especially the less commonly taught languages such as Arabic. This study investigates the ways in which second language learners of Arabic incorporate these discursive features in their writing. The study is grounded in theories of discourse analysis, which take into account not only structure but also extra-linguistic components and the social environment within which language operates (Labov 1972; Goodwin 1979, 1984). Drawing on this framework, the study analyzes sample Arabic texts written by English-speaking students of Arabic at the advanced level and compares these texts to related samples written by Arabic native speakers. Analysis of the data presupposes going beyond structure through concentration on the discursive qualities of texts and the different social processes, such as voice, generally defined as the author’s identity in texts (Bakhtin 1981, Matsuda 2001, Agha 2005) and stance, which refers to expressions of judgment or commitment (Hyland 2005, Lancaster 2016). Results show a disparity in discursive strategies used by native speakers and second language learners. Native speakers’ writings involve dialogic interactions featuring stylistic characteristics of Arabic, such as emphatics. These dialogic processes allow for interacting with potential readers and projecting authors’ views and identities through explicit linguistic units or relevant discursive resources. Second language texts also show comparable dialogic properties, but they lack authentic semiotic processes of Arabic, since students mostly resort to stylistic features of English. Based on these results, this chapter provides recommendations on how to improve the quality of Arabic second language writing.

12.1 Introduction Studies in second language acquisition have investigated structural aspects of second language (L2) writing extensively. Researchers have focused less attention, however, on the sociolinguistic features of L2 writing, especially for the less commonly taught languages such as Arabic. Written texts typically involve dialogic processes that allow for interacting with potential readers and projecting personal views through explicit linguistic units or other discursive 251

Dris Soulaimani

resources such as stance and voice. Stance broadly refers to expressions of judgments or commitments (Hyland 2005, Lancaster 2016), while voice denotes the author’s identity in texts (Bakhtin 1981, Matsuda 2001, Agha 2005). Drawing on theories of discourse analysis (Labov 1972; Goodwin 1979, 1984), this study examines comparable data sets written by Arabic L1 and L2 learners at the advanced level and investigates the discursive features of stance and voice in these texts. Scholars deploy theories of discourse analysis as viable frameworks in multiple disciplines, including anthropology, sociology, and second language acquisition. Within these trends, the sociolinguistic nature of language has been emphasized mainly through analysis of oral language. Written texts, whether L1 or L2, have received less focus, although they provide a platform for analyzing properties of language in use. This study examines the occurrence of stance markers and voice in Arabic L1 and L2 essays and illustrates the ways in which L2 texts display discursive qualities of English writing, such as self-mentions (the frequent use of firstperson pronouns: I and my). Discussion of these properties advances our knowledge of stance and voice in Arabic and directs us toward areas of improvement for Arabic language teaching, particularly in writing. Beyond structure, L2 learners of Arabic require explicit instruction on stance and voice and benefit from being aware of the social dimensions of these properties (see van Compernolle and Williams 2013). Identification of these features allows students to understand the social identities they project in writing through markers such as pronouns (ʔanaa “I”), hedges (rubbamaa “perhaps”), and boosters (li-lγaaya “extremely”).

12.2  Research questions The present study focuses on the analysis of the discursive qualities and strategies discussed above. The study addresses the following research questions: 1 How are the features of stance and voice reflected in Arabic texts? 2 How do Arabic L1 and L2 essays differ with regard to these features, and for what reasons? 3 To what extent do L2 learners of Arabic correctly incorporate the discursive properties of Arabic in their writing? 4 What are the implications of the emphasis on voice in English writing for learners of Arabic?

12.3 Methods 12.3.1 Stance Stance (or footing, in Goffman’s term), generally refers to an act of evaluation or assessment performed by participants during an interaction (Wu 2004, Du Bois 2007, Du Bois and Kärkkäinen 2012). According to Du Bois (2007), “[s]tance is a public act by a social actor, achieved dialogically through overt communicative means” (Du Bois 2007, p. 169). This dialogic view of stance and of language as a whole can be traced back to Vološinov ([1929] 1973) who tried to offer an alternative to the linguistic structural approach, considering language an “adaptable sign” (1973, p. 68) and an interactive domain. For Vološinov, the importance of a linguistic sign stems not from its stability and self-containedness, but from its being flexible and constantly changing. 252

Writing in Arabic

More recently, this interactive quality of language has been demonstratively supported within conversational analysis (see for example Jefferson 1978, Sacks 1992, Schegloff 1997) and discourse analysis (see for example Goodwin and Heritage 1990, Goodwin and Goodwin 2000). Within these frameworks the focus is on language in interaction and on the ways in which participants collaborate to construct meaning through verbal and non-verbal activities. This participation framework (Goodwin 1990) argues for an interactive view of language and calls for a linguistic theory that encompasses not only grammatical structure but also other extra-linguistic components of language such as embodied stance displays (Goodwin 2006, Du Bois 2007, Streeck 2009, Soulaimani 2017, among others). Such approaches to spoken language can be extended to analyze the sociolinguistic features of written texts. Although written language lacks some semiotic resources such as sounds and gestures, it nevertheless offers metalinguistic information that proves consequential for text quality and for determining its overall meaning. Manifestation of stance in Arabic writing, as will be discussed later, illustrates the ways in which language users display social identities and build alignments toward a topic, an idea, or a person. In this regard, Hyland (2005) discussed metadiscursive features, including interactional properties of the author’s personal involvement in written texts. These properties are made visible through different markers such as transitions (but, thus), hedges (perhaps, might), boosters (in fact), attitude markers (unfortunately, surprisingly), and self-mentions (I, my). These markers carry clear consequences for the overall quality of texts but also index various social identities of writers (Clark and Ivanič 1997, Ivanič and Camps 2001). Similarly, Lancaster (2016) discussed the organization of stance in undergraduate student essays and argued that stance-taking is a valued quality in English academic discourse. The author noted that stance is overtly present in highly graded essays through instances of contrastiveness (problem-posing), critical distance (avoiding judgments), and positive alignment (closeness to the subject matter). Overall, the author analyzed three stance properties exemplified in the following sentence from a student essay: “While that may be true, the circumstances pertaining to this case show clearly that it makes sense to allow the horizontal price-fixing.” (Lancaster 2016, p. 19, bold in the original) In this example, Lancaster argues that three stance categories work together. The first property is the attitudinal stance toward the topic under discussion, represented by “it makes sense,” which shows the author’s agreement with the subject of analysis. The second stance property is the author’s epistemic stance toward the status of knowledge, which in this case expressed by “show clearly,” thereby indicating that the writer is “committed” to the argument (p. 19). The third stance is the interactional stance, which is represented by the phrase “While that may be true,” through which the author attempts to engage and convince the reader. In the Arabic context, few studies (in English) have dealt with these discursive underpinnings. Fakhri (1994) analyzed English essays written by Arabic native speakers and noted how these essays manifest features of Arabic writing such as the excessive use of coordination (wa “and”). The author argued that this characterization is due to a lack of familiarity with English techniques of writing rather than being a question of language transfer. He also recommended more research on contrastive discourse analysis to avoid any cultural generalizations. Hatim (1997) also criticized the over-simplification in contrastive discourse involving Arabic. He cautioned against approaches that link language writing styles to cultures (for example Kaplan’s 1966 claim of the zigzagging style of Arabic compared to “the logical” and linear 253

Dris Soulaimani

writing in English). Hatim’s study, which takes translation theory as a framework, proposes instead investigating the logic specific to each way of writing; for instance, Arabic’s argumentation style is based on discussing a thesis to be supported whereas English involves a thesis to be opposed. Hatim’s contribution to the theory of contrastive rhetoric is important, and his findings on discourse are equally relevant to issues of L2 writing. For instance, the different writing preferences in Arabic and English are clearly manifest in students’ essays, as will be discussed later. A more relevant work is El-Seidi’s (2000) study, which discusses metadiscourses in English and Arabic writing and analyzes Arabic discursive expressions such as emphatics (e.g., ʔinna “indeed”), hedges (e.g. rubbamaa “perhaps”), and attitudinal markers (e.g. mina l-mudhiš “surprisingly”). The author argues for a metadiscourse curriculum in second language classrooms but focuses less on the social functions of the analyzed discursive expressions. Comparable properties are discussed here, with more emphasis on the sociolinguistic dimensions of stance and voice. Based on Hyland’s (2005) categorization in English, I discuss the occurrence of comparable as well as language-specific stance markers in Arabic writing. Comparable stance markers in Arabic include expressions such as bi-l-ʔiḍafa ʔilaa “in addition to” and laakin “but,” which function as tools of transition that establish semantic connections between clauses. Other expressions, such as yumkin “maybe,” rubbama “perhaps,” and mina l-mumkin ʔanna “it is possible that,” constitute hedges that open the floor for other views and possibilities in a text and consequently restrict the scope of discursive space. On the other hand boosters such as fi l-ħaqiiqa “in fact,” laa šakka ʔanna “there is no doubt that,” and mina l-waaḍiħ ʔanna “it is clear that” strengthen the author’s commitment to a particular argument and guide the reader in an authoritative way. This authoritativeness is normally enhanced in English through the use of first-person pronouns (see, for example, Hyland 2010, Lancaster 2016), as in ʔanaa ʔaraa “I see,” fii raʔyii “in my opinion,” denoting a clear reference to the author. A writer also might express judgments, which are revealed through attitude markers such as lil-ʔasaf “unfortunately” and wa l-ʕajiib ʔanna “it is surprising that.” Arabic also allows for different ways of expressing stance, using language-specific markers such as ʔinna (with nominal sentences) and laqad (with past form verbal sentences), both of which translate roughly as “indeed,” in addition to other Arabic stylistic features which will be explored in L1 writing in this chapter.

12.3.2 Voice Drawing on the widely influential work of Bakhtin ([1929]1984, [1935]1981), discourse can be defined as a dialogic process that applies to both written and spoken language, showing the effects of outside forces or others’ voices and judgments (see also Milroy 1980 and Agha 2005). To quote Bakhtin: It [discourse] is entangled, shot through with shard thoughts, points of view, alien value judgments and accents. The word, directed toward its object, enters a dialogically agitated and tension-filled environment of alien words, value judgments and accents, weaves in and out of complex interrelationships, merges with some, recoils from others, intersects with yet a third group: and all this may crucially shape discourse, may leave a trace in all its semantic layers, may complicate its expression and influence its entire stylistic profile. (Bakhtin 1981, p. 276) 254

Writing in Arabic

Within discourse, we find the salient feature of voice, which delineates “typifyable speaking personae” (Agha 2005, p. 39), referring to the ways in which words identify peoples’ perspectives. These are social voices, or social speech types in Bakhtin’s (1981, 1984) term (Agha 2005, p. 39), which can normally be detected in everyday interaction. Agha (2005) contributed a subcategory of voice called enregisterment, with comparable social functions and interactive qualities, normally characterized by being “indexical images of speaker-actor” (Agha 2005, p. 40). From a second language acquisition (SLA) perspective, voice, which in many ways overlaps with stance, conveys a social identity through self-representation in texts (Matsuda 2001). Whether in L1 or L2, voice also displays a writer’s opinions, authoritativeness (Ivanič and Camps 2001), and world views (Wertsch 1991). Analysis of students’ essays in English shows that voice can be explicit or hidden depending on the identities they project (Scollon et al. 1997) and the cultural environment within which writing takes place. Explicit voice is conveyed through authoritative expressions such as I think, and pronouns like he, she, or they, all of which index certain identity or alignment with a particular position. Hidden or subtle voice (also deferential) can be achieved through deliberate use of passive voice (Ivanič and Camps 2001). In all cases, voice remains an identity marker (see Halliday’s 1985 discussion of reported speech), and awareness of voice can help L2 learners control this identity in writing. Within theories of L2 writing, it has been suggested that certain cultures, such as Japanese, are less likely to express voice (Ramanathan and Atkinson 1999). Matsuda (2001), however, notes the existence of voice in Japanese discourse, although its manifestation might be realized in a different way. Matsuda (2001) summarizes properties of voice with one of the most cited definitions, “Voice is the amalgamative effect of the use of discursive and non-discursive features that language users choose, deliberately or otherwise, from socially available yet everchanging repertoires” (Matsuda 2001, p. 40). Accordingly, voice concerns the text’s feel or “reader’s impression” (Matsuda and Tardy 2007, p. 239), including its discursive and nondiscursive properties. Examples of discursive properties include stance markers (self-mentions, boosters, hedges, etc.), word selections, and structure, while non-discursive entities include punctuation, text margins, and fonts (Matsuda and Tardy 2007). In spite of its careful wording and illustrations, Matsuda’s definition has been criticized for being too general, especially by those who consider the notion of voice exaggerated. For example, Stapleton (2002) suggests that voice properties are merely grammatical structures and elements of vocabulary acquisition (2002, p. 184) rather than discursive features with social meaning. The current study, however, demonstrates that voice in L2 writing has both structural realization and social connotations. Through investigation of this discursive feature in Arabic, the study aims at contributing to the theory of SLA by presenting a cross-linguistic check of discursive features of voice and stance and their functions in writing.

12.3.3  About the data The texts analyzed in this study are drawn from essays of students taking a content course on Arabic language and culture at a major university in the United States. The course was taught entirely in Standard Arabic and covered topics such as linguistic diversity in the Middle East and North Africa, Arab nationalist movements, and the status of women in the Arab world. The students included both L1 and L2 learners of Arabic. The L2 Arabic learners were 16 Americans (six males and ten females) who had studied at least three years of Arabic (the first two years of which correspond to 20 credit hours, five credit hours for each course/semester, and the third year corresponds to eight credit hours, four credit hours for each course/semester). 255

Dris Soulaimani

The Arabic native-speaking participants were 14 college students (nine males and five females) in the same academic institution; they came from different Arab countries in the Gulf. Over the 15-week course, students were assigned a mid-term paper of 1000 words and final papers of 2000 words. The age range of the participants was 20–30. The data analyzed here draw from essays of both L1 and L2 learners, including mid-term and final assignments, from a course on Arabic culture that was offered in multiple semesters between the years 2012 and 2015. A total of 30 essays by the 30 participants were collected and analyzed. Extracted samples discussed below are from nine essays: five essays by Englishspeaking participants (two male and three female) and four by native-speaking participants (two males and two females). The selected samples are intended to serve as illustrative examples of the use of stance and voice in Arabic texts.

12.4  Analysis and results This section discusses the organization of stance and voice in Arabic L1 and L2 writing. First, I examine the occurrence of stance and voice in L2 writing and subsequently analyze the manifestation of these features in L1 writing.

12.4.1  Stance and voice in L2 writing 12.4.1.1 Self-mentions .‫( أنا أظن أن هذه الحركة الموسيقية الجديدة مهمة جدا‬1) ʔanaa ʔað̣unn ʔanna haaihi l-ħaraka l-muusiiqiyya l-žadiida muhimma židdan.1 “I think that this new musical movement is very important.” .‫( أنا أعتقد بال شك أن هناك قوة خاصة في كل فن‬2) ʔanaa ʔaʕtaqid bilaa šakk ʔanna hunaaka quwwa xaaṣṣa fi kulli fann. “I believe without a doubt [strongly] that there is a particular strength in all art.” .‫( أنا متأكد بال شك أن هذه الحركة الموسيقية ستؤثر على الحالة السياسية واالجتماعية المستقبلية‬3) ʔanaa mutaʔakkid bilaa šakk ʔanna haaðihi l-ħaraka l-muusiiqiyya satuʔaθθir ʕalaa l-ħaala s-siyaasiyya wa l-ʔižtimaaʕiyya l-mustaqbaliyya. “I’m sure with no doubt that this musical movement will affect the political and social situation in the future.”

In these examples, self-mentions are displayed through the use of first-person pronouns coupled with cognitive verbs, including ʔað̣unn “I think,” ʔaʕtaqid “I believe,” and an adjective mutaʔakkid2 “being sure.” As demonstrated in Ivanič’s (1998) analysis of English essays, selfmentions carry clear consequences for the overall impression of the writer’s “discoursal self.” Personal pronouns constitute a clear tool for self-mentions, normally used as a discourse marker to solidify the author’s thesis and show powerful commitment to a proposed statement. This commitment is shown more clearly in example (3), with the use of similar expressions “sure” and “with no doubt,” to increase the level of commitment to the proposition. Further, because Arabic is a pro-drop language (it does not require a pronoun to accompany a verb), first-person conjugation is embedded in imperfect (present) tense form through the verbal morpheme “ʔa” in ʔaʕtaqid “I believe.” This property provides another layer of emphasis through the additional verbal self-mention. While Arabic allows the use of personal pronouns to overload emphasis, excessive use of such pronouns weakens the quality of the text. Consider other examples from other L2 essays.3 256

Writing in Arabic

‫أعتقد في بعض الثقافات أحيانا‬  .‫ أعتقد أن المرأة يجب أن ترتدي الحجاب ألنها حريتهم‬.‫( أعتقد أن الجميع لديهم وجهة نظرهم الخاصة‬4) .‫الناس يعتقدون أنك لست مسلمة ألنك ال ترتدي الحجاب‬ ʔaʕtaqid ʔanna l-žamiiʕ ladayhim wižhat nað̣arihim l-xaaṣṣa. ʔaʕtaqid ʔanna l-marʔa yažib ʔan tartadii l-ħižaab liʔannahaa ħurriyatuhum. ʔaʕtaqid fii baʕḍ θ-θaqaafaat ʔaħyaanan ʔan-naas yaʕtaqiduun ʔannaki lasti muslima li-ʔannaki laa tartadii l-ħižaab. “I believe everyone has their own point of view. I believe that women should wear the hijab because that is their freedom. I believe in some cultures people might think you are not a Muslim because you do not wear the hijab.” ‫ أنا سوف أتكلم عن المشاكل‬.‫( في هذه الكتابة أنا سوف أتكلم عن اللغة األمازيغية والثقافة والهوية في القرن واحد وعشرون في المغرب‬5) .‫والخالفات وآرائي في هذا الموضوع‬ fii haaðihi l-kitaaba ʔanaa sawfa ʔatakallam ʕan l-luγa l-ʔamaaziiγiyya wa θ-θaqaafa wa l-huwiyya fi l-qarn wahid wa ʕišriin fi l-maγrib. ʔanaa sawfa ʔatakallam ʕan l-mašaakil wa l-xilaafaat wa ʔaaraaʔii fii haaðaa l-mawḍuuʕ. “In this writing I will talk about the Tamazight language, culture and identity in the twenty-first century in Morocco. I will talk about the problems and conflicts, and my views on this subject.”

In excerpt (4) self-mentions are carried through the verb ʔaʕtaqid “I believe,” which is repeated at the onset of each sentence. Likewise, in excerpt (5) this discursive feature is indicated through ʔanaa “I,” ʔatakallam “I talk,” and ʔaaraaʔii “my opinions.” Both examples are structurally correct, but they lack transition markers that indicate inter-sentence connection. More importantly, the examples manifest characteristics of English writing, with an emphasis on a strong voice, usually conveyed through self-mentions. Nevertheless, the L2 examples still exhibit dialogic qualities and interaction with possible readers. The first sentence in (4) establishes an epistemic stance toward an agreed-upon truth that people are entitled to their own opinions. Borrowing from Lancaster (2016), such constructions act as disclaim markers that explain the writer’s views against predicted opinions of the readers. This sentence sets the stage for the coming utterance in which the writer displays a second stance of commitment to the idea that women should wear headscarves because this is their freedom. The second stance display follows understandably the first one and projects the writer’s identity, or voice, as someone who values others’ opinions, and aligns with women who choose to wear a headscarf. Similarly, example (5) shows the ways in which voice is constructed through referential pronouns as in ʔanaa “I,” ʔatakallam “I talk,” and ʔaaraaʔii “my opinions.” Earlier I discussed how voice constitutes self-representation in texts, which include not only structure but also multiple semiotic resources such as word preferences, the use of tenses, pronouns, and stance markers. Following Ivanič and Camps (2001), Matsuda (2001), and Zhao (2013, 2016), these examples show a presence aspect of writing which indicates the “writer’s explicit self-reference and interactions with the intended readers in the text” (Zhao 2016, p. 3). For instance, the sentence “I believe everyone has their own point of view” represents a response to previous voices or judgments and illustrates the dialogic aspect of voice (Bakhtin 1981, Farmer 1995, Prior 2001) which renders a written text, whether L2 or L1, a field of interaction (Bakhtin 1981). From an Arabic writing perspective, however, the presence aspect seems exaggerated, and therefore negatively affects the quality of L2 writing. This overemphasis on explicit voice, self-representation and authorial presence gives these L2 examples a non-native feel. Therefore, in addition to language structure and correct transitions, L2 learners need to understand and correctly incorporate sociolinguistic tools in their writing through language-specific discourse markers such as synonyms, antonyms, and different forms of emphasis, as will be demonstrated in the next section. 257

Dris Soulaimani

‫ ولكني أود أن أركز بصفة خاصة على واحدة من ميزات‬.‫( من المستحيل أن أناقش خصائص العالم الحديث بدون مناقشة العولمة‬6) .‫العولمة المثيرة وهذه هي حركة الراب العربي‬ mina l-mustaħiil ʔan ʔunaaqiš xaṣaaʔiṣ l-ʕaalam l-ħadiiθ biduun munaaqašat l-ʕawlama. wa laakinnii ʔawadd ʔan ʔurakkiz biṣifa xaaṣṣa ʕalaa waaħida min miizaat l-ʕawlama l-muθiira wa haaðihi hiya ħarakat rraap l-ʕarabiyy. “It is not possible to discuss the characteristics of the modern world without discussing globalization. But I would like to focus in particular on one of the interesting features of globalization, and this is the Arab rap movement.”

In this L2 example, the writer uses the stance marker mina l-mustaħiil ʔan . . . biduun “it is not possible to . . . without” to note that “globalization” and “the characteristics of the new world” are intertwined topics. Conceivably, we could discuss one topic without mentioning the other, but the writer here is emphasizing the connectedness between the two topics in order to achieve an epistemic stance and increase the level of commitment to the proposition. The goal here is to present a strong introduction to the subject matter by conjoining “the modern world” and “globalization.” This introduction also reveals an interactional side of the text, which opens a dialogue to convince the reader that the connection is very evident. Further, the introductory utterance lays the ground for the subsequent sentence, in which the writer reveals the main topic of the essay. The second sentence starts with the contrastive device (wa)laakin “but,” which in English offers multiple functions, including denial of expectation (Lakoff 1971), correction (Anscombre and Ducrot 1977), and a disclaim (Martin and White 2005, Lancaster 2016). The use of laakin “but” in Arabic, however, is restricted to counter expectation (see for example Alhuqbani 2013), since other markers such as baynamaa “whereas” could be supplied (Hussein 2008). Hence, through the use of laakin “but” as a disclaim marker, the second sentence could follow the first one in English. Based on the Arabic style, however, this marker is not properly utilized, as it violates the denial of expectation criterion. The second sentence is actually confirming the first statement (the inseparability of the two topics) through discussion of the two topics (rap music in the modern world on the one hand, and globalization on the other). A possible substitute marker for laakin here could be liðaalik “for that reason,” as a statement of result. Further, the writer uses the English evaluative marker l-muθiira “the interesting” as an attitudinal stance to express appreciation. Hence, categories of stance in this example, as well as in other L2 writings, are fulfilled almost following the English argumentative style without taking into account the Arabic discursive features that could be used instead. This situation necessitates making L2 learners aware of these discourse features and alternative ways to improve the quality of their writing.

12.4.2  Stance and voice in L1 writing 12.4.2.1 Emphatics Excerpt (7), which is taken from an L1 essay, starts with a dialogic expression, almost a monologue. The dialogic dimension of the sentence is indicated through the object pronoun ka “you,” asking rhetorically if the reader would ever fully comprehend the meaning of globalization. This is a powerful stance display, through which the writer emphasizes the importance of the projected topic, guiding the reader toward the subsequent sentence, which elaborates on globalization. 258

Writing in Arabic

.‫( العولمة وما أدراك ما العولمة‬7) .‫العولمة مصطلح أو فكرة مشتركة بين المجتمع لها عدة نظريات مختلفة‬

ʔal-ʕawlama wa maa ʔadraaka maa l-ʕawlama. ʔal-ʕawlama muṣṭalaħ ʔaw fikra muštaraka bayna l-mužtamaʕ lahaa ʕiddat nað̣ariyyaat muxtalifa. “Globalization. And what will explain for you [the meaning of] globalization? Globalization is a term or a common societal theme, which is subject to different theories.”

In the subsequent sentence, the writer provides an approximate definition of l-ʕawlama “globalization,” telling the reader that the term is based on multiple theories. This multiplicity is conveyed through two interchangeable adjectives: ʕiddat “several” and muxtalifa “different.” Through the use of similar terms, the writer aims at putting more emphasis on the proposition, a typical style of Arabic writing, which values the use of synonyms as well as antonyms. More couplets are included in the following examples: .‫( منهم من يؤيد العولمة ويتقبلها ومنهم من ينفيها ويرفض تقبل الفكرة‬8) minhum man yuʔayyid l-ʕawlama wa yataqabbaluhaa wa minhum man yanfiihaa wa yarfuḍ taqabbula l-fikra. “Some people are in favor of globalization and welcome it, while others deny it and refuse to accept the idea.” .‫( ال نستطيع الهروب منها ويجب أن نتقبلها شئنا أم أبينا‬9) laa nastaṭiiʕu l-huruuba minhaa wa yažib ʔan nataqabbalahaa šiʔnaa ʔam ʔabaynaa. “We cannot escape from it and we must accept it whether we like it or not.”

The synonyms and antonyms (yuʔayyid “support;” yataqabbal “accept;” yanfii “deny;” yarfuḍ “refuse”) are meant to highlight the controversial status of globalization. When we approach these couplets, we might think that the additional words are redundant, but this is misjudging the discursive qualities of Arabic texts. The use of multiple words here reflects the various opinions that the writer wants us to imagine. It can also be an exaggeration of the way globalization should be seen as an inevitable phenomenon. This idea is well expressed through the verb yažib “must” and overly emphasized through the antonyms šiʔnaa ʔam ʔabaynaa, literally meaning “we wanted or we refused.” This exaggeration is in line with the powerful opening statement, which projected the intricacy of the topic under discussion. In this regard, Al-Masri (2007) notes that such stylistic qualities (synonyms and antonyms) enrich the Arabic text and that they become lost in English translation. In these utterances, these properties serve not only as aesthetic features but also as stance markers (boosters), which emphasize a certain topic and express a strong commitment to a proposition. More examples of synonyms taken from L1 essays are provided here. .‫( كانوا يتمسكون بهذه القيم حتى تعم وتنتشر‬10) kaanuu yatamassakuun bihaaðihi l-qiyam ħattaa taʕumma wa tantašir. “They held on to this value until it spread.” .‫( باإلضافة إلى تقديم أيادي العون والمساعدة للدول التي تواجه كوارث طبيعية‬11) bilʔiḍaafa ʔilaa taqdiim ʔayaadii l-ʕawn wa l-musaaʕada li-d-duwal l-latii tuwaažih kawariθ ṭ-ṭabiiʕiyya. “In addition to lending a helping hand and assistance to countries facing natural disasters.” .‫( حاولنا قدر المستطاع أن نبقى في حدود الموضوع لكن الكرم يجلب ويجذب‬12) ħaawalnaa qadra lmustaaṭaaʕ ʔan nabqaa fii ħuduudi l-mawḍuuʕ laakinna l-karam yažlib wa yažðib. “We tried as much as possible to stay on target, but generosity is very attractive.”

259

Dris Soulaimani

In reference to generosity, which Arabs cherish, the L1 writer here uses a number of synonyms (taʕumma wa tantašir “spread,” l-ʕawn wa l-musaaʕada “help and assistance,” and yažlib wa yažðib “attract”). As mentioned before, these synonyms are common stylistic devices in Arabic writing, and they might act as tools of persuasion (Johnstone 1991‫)‏‬. In these examples, they serve as discursive features, or boosters, that allow for heightening the degree of commitment to a certain proposition. For instance, the L1 writer in example (10) uses synonyms as a stance marker to emphasize the act of spreading generosity values as an outcome of staying steadfast to certain values. Likewise, the use l-ʕawn wa l-musaaʕada “help and assistance” in sentence (11) indexes the great amount of help given even at dire moments. Also, the verbs yažlib wa yažðib in (12) stress the fact that the topic under discussion is very attractive, which made it difficult for the writer to stay on track. ‫ إنما تأتيهم بعفوية خالصة وبشكل‬،ً ‫ فهم ال يصطنعونها اصطناعا‬،‫( والكرم يعتبر ميزة من مزاياهم الحميدة التي حباهم هللا بها‬13) .‫تلقائي‬ wa l-karam yuʕtabar miiza min mazaayaahum l-ħamiida l-latii ħabaahum l-laahu bihaa, fahum laa yaṣṭaniʕuunahaa ṣṭinaaʕan, ʔinnamaa taʔtiihim biʕafwiyya xaaliṣa wa bišakl tilqaaʔiyy. “Generosity is considered one of the praiseworthy distinguishing features that God gave them, as it is not something they made artificially; it just comes to them in a purely spontaneous manner.”

In addition to synonyms (biʕafwiyya xaaliṣa wa bišakl tilqaaʔiyy “pure spontaneity and in an improvised way”), this example shows another way of expressing emphatics. The L1 writer uses mafʕuul muṭlaq “the absolute object” (also called cognate accusative, following Johnstone (1991)), which refers typically to a combination of cognate verbs and nouns to express assertion or an adverbial quality. This is another language-specific instrument that serves as a discourse marker for emphasis. In sentence (13), the writer supplies an additional emphatic marker, ʔinnamaa, which in its correct use presupposes a negative antecedent (laa yaṣṭaniʕuunahaa “they don’t fake it”). The combination la yaṣṭaniʕuunahaa ṣṭinaaʕan “they don’t fake in an artificial way” would not make sense if translated literally into English, which perhaps explains why L2 writers avoid such structures and resort mainly to paralleled discourse markers with which they are familiar in their mother tongues. It is also noteworthy how self-mentions are significantly reduced in the current examples if compared to their L2 counterparts. The L1 writers utilize instead the collective pronoun “we” as in ħaawalnaa “we tried” and avail themselves with Arabic discursive tools to achieve assertion, persuasion, and enrich the quality of their texts. A closer look at example (13) shows the voice of the writer, who draws from various Arabic semiotic resources to overly emphasize that generosity is a genuine character of the Arab people. Other examples from the same essay on generosity give clearer illustration of voice in L1 writing. .‫( الكرم صفة من صفات العرب الطيبة والتي يعتزون ويفتخرون بها‬14) ʔal-karam ṣifa min ṣifaat l-ʕarab ṭ-ṭayyiba wa l-latii yaʕtazzuun wa yaftaxiruun bihaa. “Generosity is one among [other] good characteristics of Arabs, one in which they take pride.” ‫ ويطعم‬،‫ ويساعد الفقراء والمساكين‬،‫ وكان يحرر األسرى‬،‫ ويكرم من غير انتظار الرد أو المقابل‬،‫( فكان حاتم يعطي من غير تردد‬15) .‫ ويزوج الشباب‬،‫األطفال واألسر الفقيرة‬ fa kaana ħaatim yuʕṭii min γayr taraddud, wa yukrim min γayr intiðaari r-radd ʔaw l-muqaabil, wa kaana yuħarrir l-ʔasraa, wa yusaaʕid l-fuqaraaʔ wa l-masaakiin, wa yuṭʕim l-ʔaṭfaal wa l-ʔusar l-faqiira, wa yuzawwiž ššbaab. “So Hatim used to give with no hesitation, and be hospitable without waiting for a returned favor, and he used to free the captives and help the poor and needy, and feed the children and poor families, and help young people get married.”

260

Writing in Arabic

The writer in (14) describes generosity as one of the characteristics attributed to Arabs. This description is controlled by the stance marker “good” used as an “evaluative lexis” (Ivanič and Camps 2001, p. 13), which projects the writer’s identity as someone who values such a quality. These lexical choices index the writer’s voice in the text and constitute what Ivanič and Camps call ideational positionality, meaning “the writer’s positioning in relation to the topic they are writing about” (2001, p. 18). Such markers give the reader a window onto the writer’s preferences, beliefs, or values. In a related study, Matsuda and Tardy (2007) analyzed blind-manuscript reviewing, suggesting that some reviewers were able to construct an image of their writers, including gender, experience, or political orientation. In Bakhtinian (1981) terms, these are social voices that are appropriated in the environment and are made visible through writing (see also Agha 2005). In example (14), the author’s alignment with the topic is indicated not only through lexical items but also through structure. Had the writer chosen a slightly different wording (for example, mina ṣ-ṣifaat ṭ-ṭayyiba li-l-ʕarab “one of the good characteristics of Arabs”), the meaning would have changed and so would the stance display. The new structure would mean that Arabs might have a mix of good and bad characteristics, and that generosity belongs to a category of the good ones. This structure would convey a weaker stance display than communicated in the original text. Whether deliberate or not (Matsuda 2001), this structural choice of voice is in line with the author’s argument throughout, praising emphatically Arab generosity. The second example (15) discusses the same topic of generosity through illustration of an Arab figure from the sixth century called Hatim Al-Ta’i. Throughout the Arab world, Hatim has become an icon for endless generosity. The description of Hatim’s extraordinary character is conveyed through a series of verbs in imperfect (present) tense preceded by the verb kaana “was” to indicate a repeated action in the past. This is not a mere reporting of someone’s characteristics in a certain period in history but rather an identification with these characteristics and an attempt to persuade the reader toward a congruent stance display. In other words, the voice of the writer and his identity are transmitted through the ways in which he describes the main character as someone giving, feeding, and helping, etc. Thus, the writer reconstructs an existing discourse (generosity) and an already established character (Hatim Al-Ta’i) to support his main argument. The result is a recontextualized discourse and an image of a person belonging to a certain group characterized with certain values portrayed to the reader in the way the writer wants him to be seen.

12.5  Conclusion and pedagogical implications Although written texts might lack some of the semiotic aspects of oral language such as prosody and embodiment, they still display linguistic and non-linguistic properties through stance and voice. These features are necessary not only for determining the stylistic qualities of the text, but also for understanding the ways in which writers show social identities and display varying degrees of commitment toward an idea or a topic. Through analysis of a representative set of examples from L1 and L2 essays, this study explored multiple discursive features in these texts, including self-mentions, synonyms, emphatics, and voice. Analysis of the L2 examples reveals how L2 learners use self-mentions extensively as discursive markers to display a strong and explicit voice. Construction of both stance and voice in these texts is performed in a way that is more similar to English than Arabic writing. Additionally, L2 learners tend to omit discursive features that are unparalleled in their first language, such as mafʕuul muṭlaq “the absolute object.” Nevertheless, L2 texts manifest dialogic features and interaction with potential readers, which confirms the interactive quality of language, whether spoken or written. 261

Dris Soulaimani

On the other hand, L1 examples feature fewer self-mentions and more vocabulary couplets such as synonyms and antonyms as a way of increasing commitment to ideas under discussion. Such markers constitute language-specific discursive devices utilized not only for aesthetic reasons but also for enhancing commitment to a proposition or even exaggerating the emphasis. Hence, L1 writers avail themselves of these Arabic discursive features to write coherent texts and transmit social identities, including an appropriate voice. Researchers and practitioners in the field need to deconstruct these notions of stance and voice in original Arabic texts to determine what constitutes good quality writing. L2 learners need to learn explicitly criteria for good writing in Arabic and should benefit from awareness of the extra-linguistic dimension they project in their writing. Extensive reading of authentic texts will constitute an important resource for acquiring the discursive properties of Arabic. This will assist learners in constructing original arguments within the framework of Arabic writing, including not only correct structure but also appropriate discourse.

Notes 1 For the sake of simplification, I am omitting most case endings in transliteration. 2 This is called ism faaʕil (active participle), literally translated as the “noun of the doer,” which can function as a noun or adjective in Arabic. 3 Grammatical inaccuracies are retained and quoted as produced by the participants.

References Agha, A., 2005. Voice, footing, enregisterment. Journal of Linguistic Anthropology, 15 (1), 38–59. Al-Ani, S. H., 1972. Features of interference in the teaching of Arabic composition. An-Nashra, 5, 3–13. Alhuqbani, M. N., 2013. The English but and its equivalent in standard Arabic: Universality vs. locality. Theory and Practice in Language Studies, 3 (12), 2157–2168. Al-Masri, H., 2007. Linguistic losses in the translation of Arabic literary texts. In: D. Parkinson, ed. Perspectives on Arabic linguistics: Papers from the annual symposium on Arabic linguistics XXI. Provo, Utah. Anscombre, J. C. and Ducrot, O., 1977. Deux mais en français? Lingua, 43, 23–40. Bakhtin, M. M., [1929] 1984. Problems of Dostoevsky’s poetics. C. Emerson, ed. and trans. Minneapolis: University of Minneapolis Press. Bakhtin, M. M., [1935] 1981. Discourse in the novel. In: M. Holquist, ed., Emerson C. and Holquist M., trans. The dialogic imagination: Four essays. Austin: University of Texas Press, 259–422. Clark, R. and Ivanič R., 1997. The politics of writing. New York: Psychology Press. Du Bois, J. and Kärkkäinen, E., 2012. Taking a stance on emotion: Affect, sequence, and intersubjectivity in dialogic interaction. Text and Talk, 32 (4), 433–451. Du Bois, J. W., 2007. The stance triangle. In: R. Englebretson, ed. Stancetaking in discourse: Subjectivity, evaluation, interaction. Amsterdam: John Benjamins, 139–182. El-Seidi, M., 2000. Metadiscourse in English and Arabic Argumentative Writing. In: Z. Ibrāhīm, et. al., eds. Diversity in language: Contrastive studies in Arabic and English theoretical and applied linguistics. Cairo: American University in Cairo Press, 111–126. Erris, D.R., 1994. Lexical and syntactic features of ESL writing by students at different levels of L2 proficiency. TESOL Quarterly, 28, 414–420. Fakhri, A., 1994. Text organization and transfer: the case of Arab ESL learners. IRAL: International review of applied linguistics in language teaching, 32 (1). Farmer, F., 1995. Voice reprised: three etudes for a dialogic understanding. Rhetoric Review, 13 (2), 301–320. Goodwin, C., 1979. The interactive construction of a sentence in natural conversation. In: G. Psathas, ed. Everyday language: Studies in ethnomethodology. New York: Irvington, 97–121. Goodwin, C., 1984. Notes on story structure and the organization of participation. In: M. Atkinson and Heritage, J., eds. Structures of social action: Studies in conversation analysis. Cambridge: Cambridge University, 225–246.

262

Writing in Arabic Goodwin, C. and Goodwin M. H., 2000. Emotion within situated activity. In: A. Duranti, ed. Linguistic anthropology: A reader. Malden, MA: Blackwell, 239–57. Goodwin, C. and Heritage. J., 1990. Conversation analysis. Annual Review of Anthropology, 19, 283–307. Goodwin, M. H., 1990. He-Said-She-Said: Talk as social organization among Black children. Bloomington, IN: Indiana University. Goodwin, M. H., 2006. The hidden life of girls: Games of stance, status, and exclusion. Oxford: Blackwell. Halliday, M. A. K., 1985. An introduction to functional grammar. London: Edward Arnold. Hatim, B., 1997. Communication across cultures: Translation theory and contrastive text linguistics. Exeter, UK: University of Exeter Press. Hussein, M., 2008. The discourse marker ‘but’ in English and Standard Arabic: One procedure and different implementations. Newcastle Working Papers in Linguistics, 15, 1–16. Hyland, K., 2005. Metadiscourse. London, UK: Continuum. Hyland, K., (2010). Community and individuality: Performing identity in applied linguistics. Written Communication. 27 (2), 159–188. Ivanič, R., 1998. Writing and identity: The discoursal construction of identity in academic writing. Amsterdam: John Benjamins. Ivanič, R. and Camps, D., 2001. I am how I sound: Voice as self-representation in L2 writing. Journal of Second Language Writing, 10 (1), 3–33. Jefferson, G., 1978. Sequential aspects of storytelling in conversation. In: J. Schenkein, ed. Studies in the organization of conversational interaction. New York: Academic, 219–48. Johnstone, B., 1991. Repetition in Arabic discourse: Paradigms, syntagms and the ecology of language. Amsterdam: John Benjamins. Kaplan, R. B., 1966. Cultural thought patterns in inter-cultural education. Language Learning 16, 1–20. Labov, W., 1972. The transformation of experience in narrative syntax. In: W. Labov, ed. Language in the inner city: Studies in the Black English vernacular. Philadelphia: University of Pennsylvania, 354–396. Lakoff, R., 1971. If’s, and’s and but’s-about conjunction. In: C. J. Fillmore and Langendoen, D.T., eds. Studies in Linguistic Semantics. New York: Rinehart and Winston, 115–42. Lancaster, Z., 2016. Expressing stance in undergraduate writing: Discipline-specific and general qualities. Journal of English for Academic Purposes, 23, 16–30. Lu, X., 2011. A corpus-based evaluation of syntactic complexity measures as indices of college-level ESL writers’ language development. TESOl Quarterly, 45 (1), 36–62. Martin, J. R. and White, P. R. R., 2005. The language of evaluation: Appraisal in English. New York: Palgrave Macmillan. Matsuda, P. K., 2001. Voice in Japanese written discourse: Implications for second language writing. Journal of Second Language Writing, 10 (1–2), 35–53. Matsuda, P. K. and Tardy, C. M., 2007. Voice in academic writing: The rhetorical construction of author identity in blind manuscript review. English for Specific Purposes, 26, 235–249. McNamara, D. S., et al., 2009. Linguistic features of writing quality. Written Communication, 27 (1), 57–86. Milroy, L., 1980. Language and social networks. Oxford: Blackwell. Myles, F., 2005. The emergence of morpho-syntactic structure in French L2. In: J. M. Dewaele, ed. Focus on French as a foreign language: Multidisciplinary approaches. Clevedon: Multidisciplinary Matters, 164–190. Prior, P., 2001. Voice in text, mind, and society: Sociohistoric accounts of discourse acquisition and use. Journal of Second Language Writing, 10 (1–2), 55–81. Ramanathan, V. and Atkinson, D., 1999. Individualism, academic writing, and ESL writers. Journal of Second Language Writing, 8 (1), 45–75. Sacks, H., 1992. Lectures on conversation (vol. 2). Oxford: Blackwell. Schegloff, E. A., 1997. ‘Narrative analysis’ thirty years after. Journal of Narrative and Life History, 7, 1–4. Scollon, R., et al., 1997. Voice, appropriation and discourse representation in a student writing task. Linguistics and Education, 9 (3), 227–250. Silva, T., (1993). Toward an understanding of the distinct nature of L2 writing: The ESL research and its implications. TESOL Quarterly, 27 (4), 657–677. Soulaimani, D., (2017). Embodiment in Moroccan Arabic storytelling: Language, stance, and discourse analysis. Text & Talk. 37 (3), 335–357.

263

Dris Soulaimani Stapleton, P., 2002. Critiquing voice as a viable pedagogical tool in L2 writing: Returning the spotlight to ideas. Journal of Second Language Writing, 11 (3), 177–190. Streeck, J., 2009. Gesturecraft: The manufacture of meaning. Amsterdam: John Benjamins. van Compernolle, R. A. and Williams, L., 2013. The effect of instruction on language learners’ sociolinguistic awareness: An empirical study with pedagogical implications. System, 41 (2), 298–306. Vološinov, V. N., [1929] 1973. Marxism and the philosophy of language. Matejka, L. and Titunik, I., trans. Cambridge, MA: Harvard University. Wertsch, J. V., 1991. Voices of the mind: A sociological approach to mediated action. Cambridge: Harvard University Press. Wu, R., 2004. Stance in talk: A  conversation analysis of Mandarin final particles. Amsterdam: John Benjamins. Zhao, C. G., 2013. Measuring authorial voice strength in L2 argumentative writing: The development and validation of an analytic rubric. Language Testing, 30 (2), 201–230. Zhao, C. G., 2016. Voice in timed L2 argumentative essay writing. Assessing writing, In Press.

264

13 COMPARING THE COMPLEXITY, ACCURACY, AND FLUENCY OF WRITTEN ARABIC IN THE PRODUCTION OF ADVANCED LEARNERS AND NATIVE SPEAKERS Michael Raish This chapter compares the written performance of Arabic native speakers (NSs) and upperlevel/advanced learners of Arabic via a presentation of descriptive statistics of a number of direct measures of written complexity, accuracy, and fluency (CAF). The results presented here are exploratory in nature and represent a subset of data collected as part of a larger study investigating the measurement of written Arabic in the writing samples of Arabic learners and NSs. The specific measures selected were applied to the writing of both groups, which in turn was elicited via written prompts targeting three distinct genres of communicative writing. The measures selected target written fluency (words per minute), accuracy (ratio of error-free T-units), and complexity (average length of T-unit). In adapting these measures to Arabic, this study likewise touches on issues specific to measuring written Arabic CAF, such as the thorny issue of what to consider an “error” in writing. This study reveals that upper-level learners (determined via self-reported years of study) indeed resemble NSs in many ways, although NSs are shown to be more complex, accurate, and fluent writers. However, this study importantly sheds light on the high degree of variability in Arabic writing according to the selected measures – even within the writing of NSs. Furthermore, written genre is shown here to influence structural aspects of learner- and NS-produced writing, with persuasive writing being notably more complex, but less accurate, than writing elicited by descriptive or narrative prompts. Finally, several of the CAF measures here show promise for application in further studies of written Arabic, and future efforts will ideally expand on these results, for example by using additional measures to shed more light on the three CAF constructs and including a larger participant pool.

13.1 Introduction Issues of Arabic literacy and literacy development are of critical importance to Arabic learners, educators, and millions of native speakers. As new Arabic programs are inaugurated, and enrollment in existing academic and summer Arabic programs continues to expand (Al-Batal, 265

Michael Raish

2007), Arabic instructors and program administrators (not to mention learners) find themselves in need of useful information regarding the role of writing in Arabic instruction. The current study considers writing from a communicative perspective; that is, rather than consisting of loosely connected sentences composed by learners to demonstrate or refine knowledge of a grammatical or lexical item, communicative writing is considered very broadly to refer to tasks that involve the exchange of information, ideas, and opinions through the medium of written language (Shih 1999, pg. 21). What should communicative writing in Arabic “look like,” and how can it best be measured in an educational context? To date, however, Arabic remains under-researched in comparison with many other major languages, especially in terms of literacy development and literacy-related issues (Mahfoudhi et al. 2011, p. 1012). The question of what makes a foreign/second language user a “good writer,” as well as how to best measure or evaluate written production, is both theoretical and practical in nature. This measurement, and the constructs that it is claimed to target (e.g., proficiency, fluency, facility, etc.) is an issue that has been a source of serious debate among researchers in fields as diverse as linguistics, psychology, and neurology. In addition to representing a key avenue through which learners’ interlanguage is developed (Williams, 2012), communicative writing can be expected to be the main modality through which modern language learners interact with target cultures located overseas. Recent years have seen increased focus in the field of applied linguistics on the measurement of complexity, accuracy, and fluency (CAF) in second/foreign language (L2) learner linguistic production. The CAF framework is typically employed in relation to Skehan’s (1998) trade-off hypothesis, in which learners are assumed to possess finite attentional resources available for devotion to either linguistic complexity, accuracy, or fluency in their L2 production. According to this hypothesis, language users engaging in communicative tasks are forced to allocate differing amounts of their mental resources when attending to the complexity, accuracy, and fluency of their production. Furthermore, an increase in resources devoted to any of these three areas is understood to result in a corresponding decrease in measures pertaining to the remaining areas. In general, researchers posit a balance between form (complexity and accuracy) and fluency, with increased attention to one leading to decreased effectiveness in the other (Skehan 2009). In order to quantify these interacting constructs, researchers frequently rely on direct CAF measures that take the form of “ratios, frequencies, or formulas,” (Norris and Ortega 2009, p. 1). The CAF triad is particularly influential in the fields of applied linguistics and second language acquisition, in which researchers are often concerned with quantifiable and empirically valid measurements of learners’ interlanguage. Purpura (2013) argues that multiple grammatical structures accessed by CAF analyses seem to be acquired by L2 learners in a predictable developmental order and that “the acquisition of single structures follows a fixed developmental sequence” (p. 120). Thus far, the bulk of research employing the CAF framework has investigated oral rather than written production (Byrnes and Manchon 2014), in spite of the fact that the modality of writing is understood to play a facilitative role in the development of general L2 proficiency. Furthermore, the relationship between the complexity, accuracy, and fluency of oral production and written production is likewise somewhat unclear. Although the majority of studies employing CAF measures to date have investigated oral production, a few studies have established that each modality can elicit differential performance from learners. Recent studies have found, for example, that L2 learners may produce more grammatically complex language in the written mode than in the oral (Hakansson and Norrby 2007), that L2 learners may make more syntactical errors when writing than when speaking (Ågren et al. 2012), and that lexical complexity can be much higher in the written mode. That is, L2 learners use 266

Comparing the complexity of written Arabic

significantly more diverse vocabulary when writing about a topic as opposed to speaking about it (Ågren et al. 2012). To date, a limited number of studies have empirically investigated aspects of either L1 or L2 Arabic writing. Shakir and Obeidat (1991) present an early attempt to quantify aspects of written L2 Arabic production, finding a positive effect for a term of university Arabic study. Al-Semari (1993) found that Saudi L1 Arabic writers used similar revision strategies when composing texts in both languages. Similarly, Noor (2007) found that Saudi participants’ ability to successfully complete a Cloze test in Arabic partially predicted certain elements of their written English compositions. More generally, it is known that the literacy practices of Arabic NSs are often highly diverse. Hallajow (2016) describes the various codes and languages (including colloquial Arabic and English) used by Syrian university students online, and Bianchi (2013) similarly explores the social and communicative role of alternate script choices by L1 Arabic writers in an online environment. If the “educated native speaker” (Wahba 2006) is taken as a model for L2 Arabic acquisition, the varied communicative and digital literacy competencies of Arabic NSs should be addressed. These and similar studies argue strongly for a conceptions of L1 and L2 Arabic literacy that are capable of incorporating this diversity.

13.1.1  Syntactic complexity Although the CAF framework has been elaborated and applied to the analysis of a number of widely studied languages in both the oral and written modes, the measures it encompasses have yet to be explored in detail within the context of communicative writing in Arabic. In general, complexity in L2 production refers to the “size, elaborateness, richness and diversity of the learner’s linguistic L2 system” (Housen & Kuiken 2009, p. 464, emphasis in the original). On the syntactic level, complexity might refer to the use of sophisticated forms, the amount of subordination or other clausal measures, or mean length of a unit of production (sentence or T-unit), while lexical complexity is often operationalized in the form of ratios such as lexical diversity, sophistication, and uniqueness. It is generally assumed that learners’ linguistic production becomes more complex as they become more proficient in a language (i.e., as their interlanguage develops). At its heart, complex L2 production is that which is “elaborate and varied” (Ellis 2003, p. 340), and the production of complex language is achieved through “expanding or restructuring second language knowledge” (Wolfe-Quintero et al. 1998, p. 4). Syntactic complexity can be operationalized in a number of different ways, although these measures often take the form of ratios based on a unit of measurement; e.g., clauses, sentences, or T-units. Al-Jabr (2006) identifies three sources of syntactic complexity in Arabic and English texts, namely: a) language-specific syntactical features (e.g., whereas English tends to be characterized by a high degree of subordination, Arabic prefers coordination), b) text genre, and c) “the particular author’s (idiosyncratic) lexico-grammatical preferences” (p. 206). In the case of written Arabic, we might add d) the author’s underlying level of Arabic proficiency (in the case of l2 learners). In general, syntactic complexity is understood to increase concurrently with learners’ L2 proficiency. This has been convincingly shown to occur with clausal measures such as clauses per T-unit, dependent clauses per T-unit, and dependent clauses per total number of clauses (Kuiken AND Vedder 2012, P. 145). In the written mode, a preponderance of studies employed the T-unit as a unit of measurement. Originating with Hunt’s (1965) identification of the “minimal terminable unit,” (p. 37), the T-unit is commonly operationalized as “one main clause with all subordinate clauses attached to it” (p. 20). Researchers have some flexibility in determining the specific 267

Michael Raish

operationalization of T-unit used in individual studies and Hunt, himself, proposed several slightly divergent definitions of the T-unit (see Foster et al. 2000). Often, the T-unit represents the smallest structure that could be considered a grammatical sentence. This unit “contains one independent clause plus any number of other types of clauses, including adverbial, adjective and nominal clauses” (Wolfe-Quintero et al. 1998, p. 85). T-units are useful because they allow researchers to distinguish between simple and compound sentences, thus providing a more fine-grained basis of comparison between writing samples assumed to represent varying levels of proficiency (Schneider and Connor 1990, p. 415). Researchers may choose to view production as more syntactically complex if it contains a greater number of clauses per T-unit (Kuiken et al. 2010) or longer T-units in general (Gaies 1980). Once a stretch of discourse has been broken into units of analysis such as the T-unit, complexity can be calculated on the basis of ratios incorporating that unit. In the domain of L2 writing, mean length of T-unit, or Words per T-unit (W/T), is the single-most-used measure of complexity (Norris and Ortega 2009, p. 566; Ortega 2003; Wolfe-Quintero et al. 1998). Vann (1979) provides examples of simple T-units in English texts, contrasting an adult’s utterance, “the man who I saw yesterday runs fast,” as opposed to a child’s imitation, “I saw the man and he runs fast” (p. 6). The former utterance is composed of a single T-unit while the latter contains two, even though both could be considered intact sentences on the basis of punctuation. When glossed in Arabic, the adult’s ʔar-ražul-u l-laðii raʔaytu-hu ʔams-i yažrii sariiʕ-an consists of a main clause and a definite relative clause, the latter of which cannot stand on its own as an intact sentence (ʔal-laðii raʔaytu-hu ʔams-i is a fragment), thus this utterance consists of a single T-unit. The child’s raʔaytu r-ražul-a wa-yažrii sariiʕ-an can be broken into two self-contained, intact Arabic sentences (raʔaytu r-ražul-a and wa-yažrii sariiʕ-an). It is expected that Arabic T-units will tend to be shorter (in terms of words per T-unit) than T-units in English or other European languages due to the concatenative nature of Arabic, in which an entire predication can be expressed within a single lexical item (i.e., ‫ يشاهدونها‬yušaahiduuna-haa “they watch her”) (see Ryding 2014). The sentence (i.e., a series of main and subordinate clauses necessarily separated by a period) is an attractive alternative to the clause and T-unit in the analysis of writing primarily because it is organic, being provided by the author via punctuation. However, the sentence is problematic in the analysis of written Arabic data due to the large amount of variability in sentence type (complex, compound, simple, etc.) and length. Arabic sentences in particular are known to be longer than their English equivalents due to both “the absence of strict punctuation rules” as well as conventions of written Arabic discourse that often result in the coordination of what in English would most likely be several sentences (Farghaly and Shaalan 2009, p. 9). Thus, a “sentence” may span an entire paragraph and contain multiple T-units/clauses/ etc., joined by cohesive devices used to create coordination and subordination. Often, written data composed by learners is less than clean-cut, and may contain any number of fragments, salutations, “missing” pronouns, etc. These instances require individual researchers to make choices that emerge from their data. For example, an Arabic learner in the current study composed the following sentence in response to a description task: ‫بنايات كثيرة‬ ‫ وكل البنايات قريب من غرفتي‬binaayaat kaθiira, wa-kull l-binaayaat qariib min γurfat-ii “Many buildings, and all the buildings are close to my room.” In this selection, we find a fragment, “many buildings,” followed by an intact T-unit containing a gender agreement error. What should be done with the fragment “many buildings”? Ishikawa (1995) does not include fragments or exclamations as T-units while other researchers do. The researcher may choose to: a) count the fragment as an independent T-unit (possibly inferring that the author had intended to express “there are many buildings”); b) ignore the fragment, as it is not able to stand on its own 268

Comparing the complexity of written Arabic

as minimal sentence; or c) include the fragment with the following T-unit contained within the same sentence. Compositions would be considered less complex if analyzed according to the first option (more and shorter T-units) and more complex according to the third (fewer and longer T-units). In the current study, fragments were counted as part of adjacent T-units, rather than as independent T-units, in order to reduce the amount of guesswork incumbent upon the researcher. A number of studies have employed T-unit analysis for various purposes in Arabic. Albirini (2013) describes using T-units in the analysis of oral data produced by Egyptian and Palestinian heritage speakers of Arabic, calculating error-free T-units for accuracy and a T-unit complexity ratio for syntactic complexity. Tweissi (1990) examined the “foreigner talk” (FT) produced by L1 Arabic speakers in recorded phone conversations with both learners and other NSs, finding that Arabic FT was shown to be characterized by longer T-units than production in interaction with other NSs. Noor (2007) reports asking a population of adult Saudi L2 English students to complete a cloze test in Arabic and English as well as prepare compositions in both languages. However, the Arabic compositions were translated into English prior to analysis, and T-unit analysis was apparently only conducted on English texts. The author found that “higher competence in Arabic” (operationalized as participants’ score on the cloze test) “has a significant positive effect on the overall performance of the students’ writings in English” (p. 421). In an earlier longitudinal study of Arabic L2 writing development, Shakir and Obeidat (1991) present an attempt to quantify “maturity” in the writing of L2 Arabic learners. Utilizing a pre-/post-test design anchored to a term of study, the authors collected written compositions written by 18 adult university-level learners of Arabic produced during a standardized proficiency test. The researchers found that, after the period of study, participating Arabic learners tended to produce significantly more words per text (i.e., they were writing more fluently), although the number of sentences barely increased (i.e., their writing was more syntactically complex).

13.1.2  Lexical complexity In addition to syntactic complexity, L1 and L2 written production is known to be characterized by variable rates of lexical complexity. As vocabulary/lexical knowledge is central to the act of writing, it is expected that learners who have access to differing amounts of that knowledge should tend to produce writing of varying complexity. Additionally, lexical measures are known to be extremely important in the context of holistic ratings of linguistic performance. Within the context of Arabic, it is known that individual speakers can display differing amounts of lexical sophistication depending on the context of interaction. Tweissi (1990), for example, established that the speech of L1 Arabic speakers is significantly less lexically complex when produced in interaction with non-native Arabic speakers. Lexical complexity is often operationalized in the form of ratios representing lexical sophistication, uniqueness, or richness. Lexical sophistication often entails determining what percentage of the words in a given passage do not appear on a list of the most frequent written words in the target language. Laufer and Nation (1995) counted as “sophisticated” those items in participants’ writing samples that did not appear on a list of the 2000 most frequent words in the target language; however, the exact cut-off point for “sophisticated” Arabic lexical items when examining learners’ production remains unclear. A second widely used measure of lexical complexity, lexical richness, or lexical diversity is typically measured via a type/token ratio, in which the number of types of lexical items in a passage is compared with the total number of words (tokens) in that passage (Purpura 2013). A basic type/token ratio has been 269

Michael Raish

shown to often “artificially” inflate the richness of very short texts; however, due to the fact that shorter passages have relatively few tokens in general, they tend to repeat types less frequently than longer passages. Therefore, a number of studies employ Guiraud’s index of lexical richness in which an algebraic adjustment for text length is made by dividing the number of types by the square root of the number of tokens (e.g., Kuiken et al. 2010). Guiraud’s index has been shown to successfully compensate for disparate text lengths, an important feature in studies eliciting production from learners at disparate levels of L2 acquisition. However, the question of which words should be considered as representing different “types” is one that researchers have a great degree of freedom in determining. In English, for example, the phrase “they see her” yušaahiduuna-haa consists of three tokens and three types. However if the phrases “they see her,” “we see her,” “I see her,” etc., were repeated in a stretch of discourse, most type/token ratios would not reflect a substantial increase in lexical complexity with the addition of each phrase. However, the Arabic equivalents nušaahidu-haa, ʔušaahidu-haa, etc. would each be recognized as both distinct tokens and distinct types, greatly increasing a given stretch of discourse’s lexical complexity. Therefore, the written production presented in the current study was subjected to a process of semi-lemmatization, in which object pronouns, affixed articles (e.g., the definite article ʔal-), pronouns, “and” conjunction, prefixes, and suffixes associated with verb conjugations were stripped. This was done prior to calculating lexical diversity via Guiraud’s index.

13.1.3 Accuracy In their most basic form, empirical measures of accuracy aim to elucidate to what extent a sample of discourse is “native-like.” Accuracy may thus refer to the number (and sometimes type) of errors that are produced by a language user in a given stretch of discourse within a limited amount of time. Although often undefined (see Polio 1997), errors can be understood as deviations from the linguistic systems (phonological, morphological, syntactic, lexical, etc.) that an L1 language user might be expected to produce. Accuracy is likewise understood to be a key area in which even high-achieving L2 learners differ significantly from native speakers. Accuracy is often operationalized in the CAF framework in terms of “errors” per unit of time or linguistic production; e.g., errors per sentence, errors per 100 words, or errors per T-unit (Kuiken et al. 2010); or it can entail the classification of errors according to type or severity (Abe and Tono 2005). There is a wide-ranging strand of inquiry into the nature of accuracy in written L2 production (Polio 1997). However, the published studies investigating the written production of Arabic learners, either within the CAF framework or otherwise, remain limited. Early studies of L2 Arabic production such as those by Al-Ani (1972), Hana (1972), and Rammuny (1976) are situated within the error analysis frame and examine topics such as L1 interference and error classification. Although these error analysis studies contributed to knowledge about error production and the quantification of the same among Arabic L2 learners, they did not expand to touch on other aspects of Arabic acquisition such as stages of interlanguage development, developmental sequences, etc. More recently, Mansouri’s (1995) study of the relationship between discourse cues and learner grammatical production was based in part on learner production in the form of written essays. Importantly, advanced learners of Arabic were found to produce far fewer written linguistic errors than lower-level learners, confirming the fact that accuracy is a key measure through which advanced and beginning Arabic learners diverge. The present study employs the measurement of error-free T-units per T-unit, introduced by Scott and Tucker (1974) in their study of L2 English produced by L1 Arabic speakers. This 270

Comparing the complexity of written Arabic

measure is now widely used in CAF analyses of written L2 production. Wolfe-Quintero et al. (1998) claim in their meta-analysis of studies of L2 writing that the EFT/T measure is the most effective measure of written accuracy. They note that studies employing learners at different proficiency levels completing timed compositions have collectively found error-free T-units to be the most significant unit of analysis for accuracy. In the context of L1 Arabic, it is known that even NS writers tend to predictably commit “mistakes” while writing, especially on the orthographic/spelling level. This is in keeping with previous studies that have established predictable rates of spelling errors among Arabic NSs. Importantly, a number of NS participants in this study produced lexical items that could be considered to belong to the register of “colloquial” Arabic. The act of writing colloquial Arabic, whether in response to an experimental prompt, in the context of a newspaper article, or in any other supposedly “formal” context is inherently a controversial act (see Fahmy 2011 for a discussion of this issue). However, absent a consensus regarding this ideologically fraught issue, it is not necessarily the individual researcher’s responsibility to actively reinforce or subvert the lines that allegedly separate the various registers of Arabic writing. Therefore, given that NSs and learners used colloquial lexical items frequently in this study, such items were not coded as errors if they were appropriately inflected. For example, in response to a descriptive prompt in this study, a NS participant1 wrote: ‫ فانا شايفته جميل جدا‬، ‫ بالنسبة لحرم جامعتي‬bi-n-nisba li-ħaram žaamiʕat-ii, fa-ʔanaa šaayfat – hu žamiil židd-an, w-anti taʕrifiin-a ʔan-nii ʔuħibb-u l-ʔašžaar-a wa-l-ṭabiiʔat-a “regarding my university, I see it as very pretty, and you know that I love trees and nature.” This two-T-unit utterance contains a colloquial item šayfat – hu “I see it as” in the main clause of the first T-unit, in addition to omitting the word-initial hamza in ʔanti. Additionally, this participant used what can be described as MSA elsewhere, including the verb inflected according to the standards of MSA for the second person feminine singular taʕrifiin-a “you (f.s.) know.” Several Arabic learners used the colloquial item hoon “here,” for example, instead of the MSA hunaa. Additional artifacts of regional variation in Arabic that were present in NS production that were not counted as errors were variation in the placement, or presence, of the letter hamza. For example, an Egyptian participant produced Egyptian spelling variants such as ‫مسئول‬ masʔuul “responsible” instead of the “standard” 2.‫ مسؤول‬NSs also exhibited significant variation in the presence or absence of the hamza, with one NS writing for example ‫ يودي‬instead of ‫ ; يؤدي‬because of the sheer frequency of “missing” hamzas, no such instance was counted as an error, regardless of whether or not any other participant in this study produced a given hamzarelated spelling variant. The substitution of a haaʔ for a taaʔ marbuuṭa was likewise not counted as an error; e.g., in the NS-produced chunk ‫ فيها مباني كثيره‬fii-haa mabaanii kaθiira “there are a lot of buildings in it.” Instances of NS production that were coded as an error involved departures from the morpho-syntactic norms of MSA rather than the use of colloquial lexical items or common spelling variants. These errors included spelling errors, gender/number disagreement, and departures from the MSA case system that would normally be evident in un-vowelled texts, such as failure to represent the manṣuub case on an indefinite, masculine, singular noun or failure to correctly inflect a defective verb in the jussive mood. For example, a second NS produced the T-unit ‫ لكن يوم زوجته لم ينتهي بعد‬laakinna yawm zawžat-i-hi lam yantahii baʕd “but his wife’s day had still not ended,” in which the problematic ‫ لم ينتهي‬lam yantahii was provided instead of the correct ‫ لم ينته‬lam yantahi. Another NS participant included a superfluous preposition, describing something as fii bi-l-qurb-i min al-manzil “in in proximity to the home” (for a classification of errors used in the study, see Appendix A). Ultimately, individual researchers working with written Arabic are required to make a series of decisions in the process of defining errors that, although they may be somewhat 271

Michael Raish

arbitrary, make sense within the context of their data. In the current study, written prompts did not explicitly seek to elicit a particular register; prompts were presented to NSs in MSA, and participants were free to respond as they saw fit (for the prompts used in this study, see Appendix B). In this case, the decision was made to “accept” instances of regional variation such as colloquial lexical items and to “reject” other deviations from MSA such as superfluous prepositions, letter substitutions, etc. This taxonomy allowed for the classification of each T-unit as error-free or containing errors.

13.1.4 Fluency Fluency refers to the ability of language users to rapidly produce language (Skehan 1998). This construct can touch on how quickly (or slowly) individuals speak and write, and it likewise interacts with the constructs of accuracy and fluency according to Skehan’s (1998) tradeoff hypothesis: the more that L2 language users attend to speaking accurately, for example, the less attentional resources they have available to speak quickly, and vice versa. Fluency is a highly variable construct, and it intuitively “makes sense” that different individual language users will produce language at different rates. It is known, for example, that even L1 language users tend to vary in their rates of oral fluency (Derwing et al. 2009). In the context of L2 acquisition, even advanced L2 speakers have been shown to produce language more slowly than L1 speakers (Towell et al. 1996), making the construct of fluency potentially predictive of differences between L2 learners and NSs. The current study uses the common measure of words per minute, and initial results reveal that even NSs write Arabic at different rates, and that different task genres elicit written data produced at different rates of fluency.

13.1.5  Genre effect In addition to language users’ overall proficiency in an L2, task type has likewise been shown to have an important effect on CAF measurements. In general, it is known that written L2 proficiency is to a certain extent task-bound (Khaldieh , 2000); thus, distinct tasks targeting different genres of discourse have the potential to elicit data that contains differing structures and discourse strategies (Carrell & Connor, 1991). While certain types of linguistic production, such as description and narration, are extremely common activities for learners even at lower levels of proficiency, other types such as argumentation or persuasion are more associated with upper levels of proficiency. Familiarity with the topic of the task also influences the performance of description – learners are more able to write effectively about familiar topics than unfamiliar ones (Huang, 2009). This corresponds with Skehan’s (2009) contention that experimental tasks based on familiar information tend to advantage accuracy and fluency over complexity. Certain task types can be expected to be “easier” for writers to produce in, while others will likely be “harder.” Narratives, for example, represent a basic linguistic function that even NSs acquire at an early age and which is practiced throughout childhood (Kellogg, 2001: 177). Learners are likewise typically quite familiar with the narrative function, leading to this type of task being used frequently in both teaching and research. Argumentation or persuasion, on the other hand, is considered to be among the most difficult linguistic functions a writer can be expected to perform in an L2. Furthermore, argumentative/persuasive skills are argued to develop according to a predictable sequence, and the linguistic function of argumentation is understood to be language specific (Golder & Coirier, 1996). This is due in large part to the fact that persuasion/argumentation requires the use of linguistic structures as well as 272

Comparing the complexity of written Arabic

genre-specific rhetorical devices (Ferris, 1994). L2 learners’ production of persuasive writing is often characterized by stylistic diversions from NS norms (Kellogg, 2001). For writers of Arabic, a persuasive/argumentative task may prove to be particularly effective at revealing differences between L2 and NS writers with regard to their lexical and syntactic performance, as well as “on the frequency of cohesive markers that a particular . . . writing task can be expected to elicit” (Kormos, 2011, 149). Even NSs tend to display this variation when responding to persuasive/argumentative prompts, as the functions required to appropriately compose a response are often not fully developed at all in NSs except for in the most advanced writers (Kellogg, 2001). Given the influence of task design, two learners at comparable proficiency levels may receive quite divergent scores on a given CAF measure due to a number of factors outside their control. Even NSs will likely show some degree of variability in writing, especially in the context of a persuasive/argumentative task. Therefore, researchers aiming to capture a range of expression and to control for the influence of learner-external factors in participant writing production may choose to collect multiple samples of each participant’s written production in response to prompts within multiple task types.

13.2  Research questions This study is motivated by the following questions: 1 What are the main differences in written complexity, accuracy, and fluency between NS writers and upper-level Arabic L2 writers in terms of written CAF? 2 What is the effect of written genre on the written CAF scores of both groups?

13.3 Methods Three written prompts targeting three distinct genres of written communication were administered to NSs and Arabic L2 learners, and their responses were analyzed according to the CAF measures described earlier (see Appendix B). The results described in this chapter form part of a larger, forthcoming study examining the nature of Arabic writing when produced by learners at various levels of proficiency. Participants were allowed to compose their responses to the written prompts described here in the modality most comfortable to them, i.e., by hand writing or typing them. The implications of this choice are described later.

13.3.1 Participants The current chapter compares the writing of 14 Arabic NSs with that of a group of upper-level Arabic learners. Within the pool of NS participants, 11 were speakers of Saudi Arabic, two of Egyptian, and one of Lebanese. Participant ages were not recorded; however, current educational status is used as a proxy for age: all NSs were adult students at the time of data collection (ten BA students, three MA and one PhD). With one exception, all NS participants report having attended secondary school in Arabic, rather than a foreign language, indicating that this pool of NSs can be expected to be familiar with the type of formal writing that occurs as part of a typical Arabic language educational trajectory. However, in a reflection of the incredibly diverse array of educational systems and formats offered in the Arab world, only two of the 14 NS participants reported that the language of instruction in their university education was Arabic, and one participant reported attending university in a mixture of Arabic and English. The remaining 11 participants (eight Saudi speakers, two Egyptian speakers, one Lebanese 273

Michael Raish

speaker) report English as the language of instruction in their university-level education. When asked to rate themselves against their peers as an Arabic writer on a 1–5 Likert scale, the average of responses from this group of participants was 3.57/5 (SD = 0.76). The Arabic L2 learner group consisted of 14 participants (seven male, seven female) drawn from undergraduate and graduate courses at a university in the United States. Participants were classified as being upper-level learners if they reported completing at least four years of Arabic study at the time of data collection, which occurred close to the end of the spring semester. Information regarding participants’ study and knowledge of Arabic prior to data collection was not collected; that is, participants were included in the study solely on the basis of their selfreported number of years of Arabic study. Although age data was not recorded, all learner participants were adult students at the time of data collection – one was a first-year undergraduate student, one a second-year, one a third-year, and two were fourth-year undergraduate students. The remainder (n = 9) were graduate (MA and PhD) students. Eight of the learners reported having participated in a study abroad experience in which they studied Arabic. None of the L2 Arabic learner participants reported being heritage learners or speakers of Arabic. In keeping with general trends regarding Arabic learners in the United States, three of the upperlevel participants had studied one language in addition to Arabic, and ten had studied two or more languages in addition to Arabic.

13.3.2  Instruments and procedure Participants in this study were administered three written prompts targeting the respective genres of written L2 expression: description, narration, and persuasion/argumentation. These prompts were administered in all participants’ L1 in random order and were chosen in order to capture as representative a sample as possible of participants’ written capabilities across genres (see Appendix B for the presentation of written prompts used in this study). Manchon (2014) holds that written tasks should be intended to prompt the activation of various forms of L2 knowledge for the purpose of “communicative engagements in the personal, social, professional or academic sphere” (29). Therefore, written prompts framed as pieces of personal communication (letters, emails, etc.) may be more effective at stimulating this activation. Alanen et al. (2012) concur, stating that writing tasks used for assessment purposes should be designed to have “some measure of authenticity” in terms of the types of production and processes needed to compose them as well as to elicit types of production familiar to participants (p. 188). In the descriptive task, participants were asked to describe their university in as much detail as possible within the context of a “message/letter” to a friend. In the narration task, participants were asked to “tell the story” depicted in a picture series taken from a popular Arabic textbook. Narratives likewise represent one of the most widely taught written genres of linguistic production in foreign language instruction, especially at the beginning level (Kormos 2011, p. 148). In the final task, participants were asked to write a letter to the editor of their school newspaper, supporting or opposing a ban of smoking on their university’s campus. In order to ensure that the written production described here was collected in an ecologically sound manner, participants were allowed to compose their texts in the modality most comfortable to them. To this end, all NS participants except for one typed their responses, while all Arabic learners except for one opted to write by hand. L2 Arabic learner and Arabic NS participants were administered three written prompts in their L1 (English and Arabic, respectively) in which they were asked to respond in written Arabic in as much detail as possible. The administration of the individual prompts was randomized, and responses to each were limited to 20 minutes, although participants were allowed to cease writing before 20 274

Comparing the complexity of written Arabic

minutes if they finished their response. L2 Arabic learners completed their responses by hand in the presence of the researcher, while NS participants were given access to an electronic version of the experimental materials which they completed and returned via email. The electronic form included fields in which NS participants were asked to record their start and stop time for each prompt, and no NS participants reported taking longer than 20 minutes for any individual prompt. Handwritten learner responses were subsequently transcribed in Microsoft word, allowing for the calculation of total words, total sentences, and number of error-free T-units per response (see Appendix A for a description of the various linguistic phenomena identified as an “error” or as “acceptable” for the purposes of coding for accuracy). The remainder of calculations were conducted in Microsoft Excel. Due to the small number of participants, coding and analysis were limited to descriptive statistics for the purpose of comparison between groups. In order to arrive at a workable unit of lexical diversity, the transcribed responses of NSs and Arabic learners were further lemmatized from their original format, and these responses were analyzed via the Lexico 5 corpus analysis software. The lemmatization process entailed the removing of definite article prefixes, the conjunction wa- “and,” and a number of other affixed inflections that would complicate automatic recognition of forms. For example, in the absence of some form of lemmatization, the Lexico program would recognize ʔal-kitaab, kitaab, wa-lkitaab, wa-kitaab, kitaab-an, etc., as separate types, and in a sense “artificially” inflating the lexical diversity present in all participants’ production. Therefore, most affixed inflections were removed from nouns and verbs via the find/replace function in Microsoft Word, and instances of the examples listed were reduced to kitaab.

13.4 Results The initial results of this study reveal that NSs of Arabic are indeed more complex, accurate, and fluent writers than their high-achieving Arabic L2 learner counterparts. However, individual differences within groups indicate that both groupings of participants include writers with diverse CAF-related abilities. For example, the passage in Table 13.1 presents the first 101 words taken from a representative example of a NS response to the persuasion prompt, in which participants were asked to support or oppose a campus-wide ban on smoking in the form of a letter to the editor of their student newspaper: This passage can be compared to the 101-word sample presented in Table 13.2 in which an Arabic L2 learner responds to the English version of the same prompt: These two samples, as well as the groups they are drawn from, diverge according to all the selected measures of written Arabic CAF. However, there is often substantial overlap between Table 13.1 NS persuasion response ‫ حيث إن المجتمع‬،‫ وأنا أرى أن فى ذلك مجهود عظيم يستحق الثناء‬.‫تقوم الجامعة فى الوقت الحاضر بتنظيم حملة للقضاء على التدخين‬ ‫ ولكني أرى في الوقت نفسه أن ثمار تلك الحملة لن تتأتى كاملة اال‬.‫يرى في التدخين آفة مجتمعية يجب التخلص منها بشتى الوسائل‬ ‫ حيث أن فرض القوانين في الجامعة والعمل على تطبيقها بشكل فعال لن يجدي اذا لم تتماشى معه‬،‫باشتراك الدولة والمجتمع ككل‬ ‫ فالمادة‬،‫ وأعتقد أنه من بين أكثر الوسائل تأثيرا في هذا الشأن هي وسائل االعالم‬.‫قوانين الدولة والمجتمع ككل خارج الحرم الجامعي‬  . . . ‫ على ما يبدو‬،‫االعالمية يتأثر بها الشباب خاصة والذين يشكلون النسبة األكبر من المدخنين‬  Words/Min

Guiraud’s Index

Words/T-Unit

Error-Free T-Units/ T-Unit

8.70

9.90

15.82

0.91

275

Michael Raish Table 13.2 L2 learner persuasion response ‫ منذ أسبوع سمعت عن قرارة الجامعة لتتهى‬.‫ أتمنى أن هذه الرسالة يجدك في خير وحصة وعافية‬.‫ السالم عليكم‬.‫لمقرئ هذه الرسالة‬ ‫ منع التدخين عادة الشخص يبدأ التدخين ألن شيئا‬.‫ أسعدت بذالك الخبر وأشكركم على قرارتكم الحسنة‬.‫التدخين في منطقة الجامعت‬ ‫ بل يخطر صحته‬.‫ بينما يمكنه أن يسأل لمساعدة مشاكله أو يعالج نفسه يبدأ التدخين وال يحسن أي شيء في حياته‬.‫يحزنه أو يزعجه‬ ‫ فبال من قضاء وقته في التدخين يجب الشخص أن يصبح ما يزعجه في حياته ويسأل أهله‬.‫وصحة الناس الذين يستمعون التدخين‬ .‫ ولذالك قرارة الجامعة لتمنع التدخين سيضطر الطلبة ليبحثوا عما يساعدهم ويحسن صحتهم‬.‫وأصدقاءه ليساعده في نهاية تدخين‬ Words/Min

Guiraud’s Index

Words/T-Unit

Error-Free T-Units/ T-Unit

9.18

7.56

8.42

0.58

the selected pools of NS and Arabic learner writers. Although the group means indicate that NSs write more quickly than high-achieving learners, for example, the NS writing presented in Table 13.1 was composed at a slightly slower rate than the learner writing presented in Table 13.2. Descriptive statistics representing participant scores are presented in Table 13.3. Across all three prompts, NSs on average produced 12.45 words per minute, while learners produced 7.01 words per minute. Across all three prompts, NSs on average produced 12.45 words per minute while the L2 learners produced 7.01 words per minute. Thus, it can be said that this population of NSs are more fluent writers than their Arabic L2 learning counterparts. Interestingly, the differences in standard deviation show that there is more variability within the group of NS participants than within the Arabic L2 learners in terms of written words/minute. Within this more fluent written production, the T-units produced by NSs are more syntactically and lexically complex than those produced by the L2 learners. The average length of T-units produced by NSs in this study is 10.10 words per T-unit, while the average length of an L2 Arabic T-unit in this study is 7.37 words. NS production is also more significantly more lexically complex than that of learners, with an average Guiraud’s index score of 12.01 as compared to that of the average learner score of 9.71. It should be noted that the unit of lexical diversity reported here represents the total number of types produced by a given participant, divided by the square root of the total number of tokens produced, over the course of responses to three prompts, rather than an average of a participant’s three individual lexical diversity scores. Finally, it can be observed from these raw scores across the three genres of written production analyzed in this study that NSs of Arabic are significantly more accurate than their highachieving Arabic learner counterparts. While 93% of NS T-units were classified as “error-free” according to the coding schema used in this study, only 62% of learner T-units met that definition, which is in line with previous studies that have established that the error-free T-units/Tunit measurement can be particularly helpful for distinguishing between high-achieving learners and NSs. A further observation facilitated by these scores is that the NS participants in this study tend to exhibit a high degree of diversity in their scores on the individual measures examined, excluding the selected accuracy measure. The standard deviation of NS scores tends to be larger than that of learner scores for every measure except for EFT/T, indicating that there is a wider range between average individual NSs than between the average L2 learner participant classified as high-achieving on the basis of years studied. When L2 learner and NS scores for the various measures of complexity, accuracy, and fluency are broken down according to written genre, however, a number of key differences between and within the two groups of participants emerge. Of the three prompts, the persuasive prompt elicited the least fluent production among both groups, with L2 learners composing in this prompt at 6.24 words/minute and NSs at 12.68 words/minute (see Table 13.4 for 276

Comparing the complexity of written Arabic Table 13.3 Descriptive statistics

Fluency (words/ minute) Lexical Complexity (Guiraud’s index) Syntactic Complexity (words/T-unit) Accuracy (error-free T-units/T-unit)

Mean Std. Deviation Mean Std. Deviation Mean Std. Deviation Mean Std. Deviation

Arabic Learners

Native Speakers

Statistic

Std. Error

Statistic

Std. Error

7.01 2.60 9.71 1.66 7.37 1.41 .62 .18

.70

12.45 3.52 12.01 1.59 10.10 2.15 .93 .08

.94

.44 .38 .05

.42 .58 .02

Table 13.4 Persuasion-specific measures Construct

Fluency Lexical Complexity Syntactic Complexity Accuracy

Measure

Words/Min Guiraud’s Words/T EfT/T

NSs

Learners

Mean

Std. Deviation

Mean

Std. Deviation

12.68 7.99 13.96 .87

2.32 1.09 3.39 .16

6.24 5.89 8.84 .52

2.71 1.37 2.01 .28

persuasion scores). Again, the NSs wrote much more quickly than learners across all prompts, and the descriptive prompt elicited the most fluent production among NSs (13.12 words/minute; see Table 13.5 for description scores). However, the fluency rates of NSs were much more tightly grouped in the persuasion prompt (standard deviation = 2.32) vs. in their responses to the description prompt (standard deviation = 7.43 words/minute). The L2 learners wrote more slowly than the NSs across all three prompts, with the persuasion prompt likewise eliciting the least fluent production, although all three prompts were very similar. NSs achieved higher scores according to Guiraud’s index in each of the individual prompts than the Arabic learner group, although the persuasion prompt interestingly failed to elicit the most lexically diverse production in either the NS or the Arabic learner groups. Among Arabic learners, the persuasion prompt actually elicited the least lexically diverse writing, while the description elicited the most lexically diverse writing among both groups. Breaking down the selected CAF scores by task type reveals the fact that the three genres of written production elicit divergent levels of syntactic complexity, as operationalized in the form of mean length of T-unit. First, it is clear that NSs produced more complex (i.e., longer) T-units than learners across the three prompts and that their average T-unit length is less tightly grouped than that of learners, with larger standard deviations for each prompt than the Arabic learner group. NS participants in this study produced average T-units of similar lengths in response to the description and narration prompts (see Table 13.6 for narration scores), while the average T-units produced in response to the persuasion prompt are much longer (13.96 words/T), nearly four words on average longer than the production elicited by the description task (9.17 words/T). The learner group in general seems to produce T-units in patterns similar 277

Michael Raish Table 13.5 Description-specific measures Construct

Fluency Lexical Complexity Syntactic Complexity Accuracy

Measure

Words/Min Guiraud’s Words/T EfT/T

NSs

Learners

Mean

Std. Deviation

Mean

Std. Deviation

13.12 8.21 9.17 .95

7.43 1.40 2.85 .058

7.24 6.77 7.61 .58

3.02 1.60 1.92 .22

Table 13.6 Narration-specific measures Construct

Fluency Lexical Complexity Syntactic Complexity Accuracy

Measure

Words/Min Guiraud’s Words/T EFT/T

NSs

Learners

Mean

Std. Deviation

Mean

Std. Deviation

12.79 7.26 8.46 .93

4.96 1.38 2.32 .10

7.24 6.77 7.61 .58

3.02 1.60 1.92 .22

to those of NSs; however, learner T-units are on average shorter than those of NSs. Learners likewise produced the most complex T-units in response to the persuasion prompt (8.84 words/T). The most striking aspect of learner and NS scores on this measure is that the persuasion prompt appears to elicit production which is measurably and significantly more complex (according to the selected measure) than either the description and narration prompts, both of which resemble the other. NSs proved to be highly accurate writers according to the determination of what constitutes an error, described earlier. Generally, inclusion of colloquial lexical items, hamza omission and idiosyncratic hamza placement, as well as other deviations from general norms of MSA spelling were not counted as errors. Gender and case disagreement, superfluous prepositions, spelling errors, etc., were counted as errors. In both the description and narration prompts, the NS-produced T-units that were over 90% accurate on average; however, the persuasion prompt elicited T-units that were only 87% accurate. While still highly accurate, the production elicited by the persuasion prompt is less accurate than that elicited by the other two prompts. In order to investigate the significance of the differences in accuracy scores between persuasion and the other two investigated genres in the writing of NSs, a series of paired-samples t-tests were conducted. The difference in accuracy scores between the persuasion (M = .87, SD = .164) and narration (M = .93, SD = .10) prompts in the NS group was not revealed to be statistically significant; t(13) = 1.45, p = .17. However, the difference between accuracy in the description (M = .96, SD = .06) and the persuasion (M = .87, SD = .164) prompts was determined to be statistically significant among NSs; t(13) = −2.38, p = .033. The same pairedsamples t-test did not indicate statistical significance when comparing between persuasion and other accuracy measures in the L2 learner group, however. Among the L2 learners, there was is no statistical significance in the comparison between persuasion (M = .52, SD = .28) and description (M = .59, SD = .22); t(13) = −.88, p = .39; although the difference between accuracy in persuasion and accuracy in narration (M = .72, SD = .20) was determined to be statistically significant t(13) = 2.55, p = .024. 278

Comparing the complexity of written Arabic

13.5  Discussion and future directions The results presented in the previous section reveal that, although significant differences exist between NSs and high-achieving Arabic learners according to all measures selected for this study, the Arabic L2 learner group appears to be exhibiting many of the same trends evident in the production of NSs. These trends become particularly clear when production is broken into the three distinct genres targeted in this study. Of particular note is that the persuasion prompt appears to be associated with diminished accuracy and increased syntactic complexity in the production of both NSs and Arabic learners. Among L2 learners, Skehan (2009) finds that narrative tasks tend to elicit more complex but less fluent and accurate production, while tasks “based on concrete or familiar information advantage accuracy and fluency” (p. 511). Although NSs proved to be more lexically diverse writers than the Arabic L2 learners who participated in this study, the three prompts administered to participants do not appear to elicit highly divergent rates of lexical diversity. Lexical knowledge (e.g., the diversity of lexical forms employed in writing) has been shown to be a key area of divergence between the writing of even advanced learners and NSs. Graduate-level English learners, for example, have been shown to lack derivative knowledge (i.e., the ability to derive different forms) of academiclevel vocabulary items (Schmitt and Zimmerman 2002). Even among L1 writers, it is clear that academic discourse is not the first natural language of any writer (Tang 2012). Instead, academic writing is a specialized competence which both native and non-native speakers must acquire (Hyland 2016), typically through education. The type/token ratio employed in the current study could easily be combined with a measure of lexical sophistication based on frequency lists, for example; lexical measures have been shown to be highly predictive of holistic ratings of L2 writing (e.g., Crossley et al. 2011). As noted earlier, the relationship between scores on individual CAF measures and holistic scores on instruments such as an OPI is underinvestigated, especially in the context of a topic such as written Arabic. Future studies will hopefully attempt to collect written data concurrently with an independent measure or measure of latent Arabic proficiency, such as an OPI or a “short-cut” measure such as a cloze or C-test. The written passages that form the data described in this chapter could also be submitted for holistic ratings by Arabic instructors or another group of NS readers in order to investigate the relationship between individual CAF measures and the “quality” of the Arabic writing from which they emerge. Additionally, the fact that only 87% of NS T-units produced in the persuasion task were classified as “error-free” according to the classification taxonomy used in this study indicates that the structures and forms commonly used for persuasion in written Arabic may not be fully accessible to all literate NSs, even if they are currently university students. Admittedly, the difference in scores between persuasion and narration among NSs was not determined to be statistically significant via a paired-samples t-test, although the difference between persuasion and description was. The small sample size (14 individuals in the L2 learner and NS groups) also limits the strength of conclusions that can be made on the basis of these scores, although the results presented above do indicate that, within this group of participants, NSs seem to write differently in response to prompts of different genres. Future studies in this vein would greatly benefit from an increased sample size, especially given the fact that the NS group in this study proved to exhibit a great deal of within-group variation according to the selected measures of CAF. The decreased NS accuracy in response to the persuasion prompt can be compared to the use on the part of some participants of colloquial lexical items in response to prompts formulated in MSA. Clearly the issue of literacy (academic and otherwise) is particularly thorny 279

Michael Raish

with regard to users of Arabic, and the issue of formality/informality in Arabic written discourse is highly contentious and represents a potentially rich avenue of research for future studies. As Fahmy (2011) argues, any act of writing in colloquial Arabic is inherently a controversial, political act. Many Arabic NSs likely produce a large amount of writing composed in informal registers of Arabic (that may correspond with spoken dialects) in the form of personal emails, text messages, and writing composed to be shared via social media platforms (Hallajow 2016). More information is needed, however, regarding the literacy-related capacities, beliefs, and practices of educated NSs if Arabic L2 learners are to use their writing as an achievement goal. It is certainly possible that differences in the modality of composition could lead to differences on the various CAF measured examined (i.e., that participants may type faster than they handwrite); however, it must also be acknowledged that digital literacy is far from a main emphasis of instruction in many Arabic instructions. Typing, to the extent that it forms a part of Arabic curricula, is typically not required of learners in the institution in which this study was conducted. On the other hand, the vast majority of NS participants report engaging in digital literary practices such as writing personal emails (14/14), writing text messages (13/14), and writing on social media in Arabic (14/14) – activities that typically necessitate typing – supporting the decision to allow participants to compose Arabic in the modality most comfortable to them. Future studies may investigate what effect, if any, the decision to type or handwrite affects the quantity and quality of written Arabic on the part of both learners and NSs; however, in the absence of systematic emphasis on digital literacy in Arabic curricula, it seems likely that at least lower-level Arabic learners will be more comfortable handwriting.

13.6  Conclusion and pedagogical implications This study is exploratory in nature, and only presents the descriptive statistics associated with the written production of a group of NSs and high-achieving Arabic L2 learners. However, given the dearth of published studies investigating the thorny topic of writing in Arabic from an empirical perspective, this study presents useful information regarding effective methods for quantifying certain elements of Arabic compositions. Although NSs are more complex, accurate, and fluent writers than the group of high-achieving Arabic L2 learners who participated in this study, this group of learners superficially appears to be “moving” toward NS norms. They produce more syntactically complex language in response to a persuasive written task, for example, and this task likewise elicits the least accurate writing from both groups. Future studies in this vein will hopefully include a larger participant pool and expand these efforts to additional measures of written production, as well as examine the relationship between written Arabic CAF and independent measures of Arabic proficiency, such as an Arabic OPI. The results of this study indicate that instructors, researchers, and other Arabic language professionals may wish to specifically include a consideration of CAF when designing and administering tasks to learners, given that specific task types appear to elicit variable CAF scores among both L2 learner and NS writers. For example, this knowledge may be useful to curriculum designers who are concerned with the sequencing of pedagogic tasks.

280

Comparing the complexity of written Arabic

Appendix A: Error Classification This study required the identification of written phenomena that occurred in the Arabic data as an “error” or as “accepted” production in order to calculate the selected measurement of written accuracy. Rather than accept all NS production as “error-free,” certain error types were identified within the writing of NSs (see Raish 2017 for a more detailed exploration of these phenomena). The identification of the phenomena presented here allowed for the classification of T-units as either error-free or as containing an error. In certain cases, follow-up research was conducted to confirm that problematic instances of NS production elicited in the current study were indeed “errors” or deviations from the norms of written MSA. The following phenomena were counted as errors in the writing of Arabic NSs and Arabic L2 learners: 1 Missing, alternative, or superfluous letters. For example, a NS writer described a university as‫ في بالقرب من المنزل‬fii bi-l-qurb-i min ʔal-manzil “in in proximity to the home.” 2 Case and verbal mood, and related inflection errors. For example, a NS writer wrote that certain laws ‫ لم تتماشى‬lam tatamaašaa “are not in harmony.” According to the rules of MSA verbal mood inflection, form VI verbs such as the verb ‫ تماشى‬tamaashaa “to conform to, be in harmony with” that are composed of a triliteral root of which the final root letter is realized as an ʔalif maqṣuura should be realized with a word-final fatħa in the jussive mood (‫لم‬ ‫تتماش‬ lam tatamaaša). َ The following phenomena were not coded as errors in the writing of Arabic NSs and Arabic L2 learners: 1 Inclusion of colloquial lexical items. An Arabic NS produced the following two-T-unit passage while describing their university: ‫ وانت تعرفين أني أحب‬،‫ فانا شايفته جميل جدا‬،‫بالنسبة لحرم جامعتي‬ ‫ األشجار والطبيعة‬bi-n-nisba li-ħaram jaamiʕat-ii, fa-ʔanaa shaayfat-hu jamiil jidd-an, w-anti taʕrifiin-a ʔan-nii ʔuħibb-u l-ʔashjaar w-ṭ-ṭabiiʕa “as far as my university, I see it as very pretty, and you know that I love trees and nature.” Given that a colloquial item co-occurred with MSA markers (e.g., the MSA second person feminine singular taʕrifiin-a “you know”) in this NS-produced writing, the same phenomena was not coded as an error when produced by learners. This occurred for example in the learner-produced T-unit ‫لذلك أي موضوع‬ ‫ في عمل وفرصة هون‬،‫ الذي تريد أن تدرسه‬li-dhaalika ʔayy mawḍuuʕ ʔal-ladhii turiid ʔan tadrusa-hu, fii ʕamal wa-furṣa hoon “therefore, any subject you want to study, there’s work and opportunity here.” 2 Common spelling variants. The most frequent types of spelling variation consisted of variability in writing the letter hamza, as well as substituting the word-final letter ‫ – ـه‬ah for the feminine marker ‫ ـة‬a(t). Additionally, certain spelling practices commonly used in Egypt were not counted as mistakes, even if they might be viewed as such by NSs from a different region of the Arab world. The Egyptian spelling style that manifested most frequently in this data is the (mostly) Egyptian practice of writing a word-final ‫ ـي‬ii without the accompanying dots (i.e., as ‫)ـى‬.

281

Michael Raish

Appendix B: Written Prompts Arabic learner participants were presented instructions in English, while Arabic NSs were presented instructions in Arabic.

Instructions: ‫ قُم باإلجابة على السؤال بشكل تفصيلي وأكتب كل ما تتمكن‬.‫ دقيقة لإلجابة على الموضوع التالي‬20 ‫أمامك‬ . ‫من كتاباته‬ You will be given 20 minutes to respond to the following prompt. Please try to answer as thoroughly as possible and write as much as you can.

Description prompt: ‫صديقك من المرحلة الثانوية ينوي االلتحاق بإحدى الجامعات إلكمال دراساته العليا ويريد بعض التفاصيل عن‬ . ‫ أكتب له بريدا االكترونيا تصف فيه شكل حرم جامعتك‬.‫ درست فيه‬/‫الحرم الجامعي الذي تدرس‬ A friend from high school wants to enroll in a university to continue their higher education, and they would like some information about the campus where you study/ studied. Write them a message in which you describe your campus in as much detail as possible.

Narration prompt: .‫ احكي ما يفعله هذا الشخص خالل اليوم‬.‫يصور هذا الكاريكاتير يوما عاديا في حياة هذا الشخص‬ The following cartoon depicts a normal day in the protagonist’s life. Tell the story of what this person does during their day.

Persuasion prompt: ‫ وقد طُلب منك التعبير عن رأيك في هذا الموضوع‬،‫تدرس جامعتك مسألة منع التدخين في الحرم الجامعي‬ ‫ سود رسالة لرئيس تحرير صحيفة الجامعة تعبر فيها عن رغبتك في‬.‫بوصفك عضوا فعاال في اتحاد الطلبة‬ . ‫ اسرد في هذه الرسالة مبرراتك للقبول أو الرفض‬،‫دعم أو عدم دعم هذه الفكرة‬ Your university is studying the issue of banning smoking on campus, and you have been asked to express your opinion on this matter as an active member of the student union. Draft a letter to the editor of the university newspaper in which you express your desire to support or oppose this idea, and detail your justification for doing so.

Notes 1 This participant self-identified as a NS of Arabic and reported having completed his or her high school and university education in Arabic in a Middle Eastern country. It is possible that the wording of the prompt (i.e., “[w]rite a message to a friend . . .”) was influential in eliciting written colloquial Arabic in this case. The elicitation of various forms of written Arabic promises to be a rich topic for future study. 2 A search on the arabiCorpus website conducted on 11/2/2016 confirms that the spelling variant ‫ مسئول‬occurs almost exclusively in Egyptian subcorpora of the “All Newspapers” corpus, providing cursory evidence for the fact that it is indeed an “acceptable” spelling variant in Egyptian print media.

282

Comparing the complexity of written Arabic

References Abe, M. and Tono, Y., 2005. Variations in L2 spoken and written English: Investigating patterns of grammatical errors across proficiency levels. In: Proceedings from the corpus linguistics conference series 1. Birmingham University. Available online at www.birmingham.ac.uk/Documents/college-artslaw/ corpus/conference-archives/2005-journal/LanguageLearningandError/variationsinL2.doc Ågren, M., et al., 2012. The growth of complexity and accuracy in L2 French. In: A. Housen A, et al., eds. Dimensions of L2 performance and proficiency. Complexity, accuracy and fluency. Amsterdam: John Benjamins, 95–119. Alanen, R., et al., 2012. Designing and assessing L2 writing tasks across CEFR proficiency levels. In: I. Bartning, et al., eds. Learning to write effectively: Current trends in European Research (Eurosla Monographs, 1). Rome: EuroSLA, 21–56. Al-Ani, S. H., 1972. Features of interference in the teaching of Arabic composition. An-Nashra, 5, 3–13. Al-Batal, M., 2007. Arabic and national language educational policy. The Modern Language Journal, 91 (2), 268–271. Albirini, A., 2013. Toward understanding the variability in the language proficiencies of Arabic heritage speakers. International Journal of Bilingualism, 18 (6), 730–765. Al-Jabr, A. F., 2006. Effect of syntactic complexity on translating from/into English/Arabic. Babel, 52 (3), 203–221. Al-Semari, O., 1993. Saudi students’ revising strategies in Arabic and English essays. Dissertation Abstracts International, 55 (2), 265A.m Bianchi, R. M., 2013. Arab English: The case of 3arabizi/Arabish on Mahjoob. com. Voices in Asia Journal, 1 (1), 82–96. Byrnes, H. and Manchón, R. M., 2014. Task-based language learning: Insights from and for L2 Writing (Vol. 7). Amsterdam: John Benjamins Publishing. Carrell, L. and Connor, U., 1991. Reading and writing descriptive and persuasive texts. The Modern Language Journal, 75 (3), 314–324. Crossley, S. A., et al., 2011. Predicting lexical proficiency in language learner texts using computational indices. Language Testing, 28 (4), 561–580. Derwing, T. M., et al., 2009. The relationship between L1 fluency and L2 fluency development. Studies in Second Language Acquisition, 31 (04), 533–557. Ellis, R., 2003. Task-based language learning and teaching. Oxford: Oxford University Press. Fahmy, Z., 2011. Ordinary Egyptians: Creating the modern nation through popular culture. Stanford: Stanford University Press. Farghaly, A. and Shaalan, K., 2009. Arabic natural language processing: Challenges and solutions. ACM Transactions on Asian Language Information Processing (TALIP), 8 (4), 1–22. Ferris, D. R., 1994. Lexical and syntactic features of ESL writing by students at different levels of L2 proficiency. TESOL Quarterly, 28 (2), 414–420. Foster, P., et al., 2000. Measuring spoken language: A unit for all reasons. Applied Linguistics, 21 (3), 354–375. Gaies, S. J., 1980. T-unit analysis in second language research: Applications, problems and limitations. TESOL Quarterly, 14 (1), 53–60. Golder, C. and Coirier, P., 1996. The production and recognition of typological argumentative text markers. Argumentation, 10 (2), 271–282. Håkansson, G. and Norrby, C., 2007. Processability Theory applied to written and oral Swedish. In: F. Mansouri, ed. Second language acquisition research: Theory-construction and testing. Cambridge: Cambridge Scholars Press, 81–94. Hallajow, N., 2016. The interplay of technology and context in Syrian university students’ electronic literacy practices. Computers in Human Behavior, 55, 178–189. Hana, S. A., 1972. Quantitative measurement of errors and remedial instruction in reading Arabic. Allisan Al-arabi, 9 (1), I – VI. Housen, A. and Kuiken, F., 2009. Complexity, accuracy, and fluency in second language acquisition. Applied linguistics, 30 (4), 461–473. Huang, Q., 2009. Background knowledge and reading teaching. Asian Social Science, 5 (5), 138. Hunt, K. W., 1965. Grammatical structures written at three grade levels. NCTE Research Report No. 3. Hyland, K., 2016. Academic publishing and the myth of linguistic injustice. Journal of Second Language Writing, 31, 58–69.

283

Michael Raish Ishikawa, S., 1995. Objective measurement of low-proficiency EFL narrative writing. Journal of Second Language Writing, 4 (1), 51–69. Kellogg, R. T., 2001. Competition for working memory among writing processes. The American Journal of Psychology, 114 (2), 175. Khaldieh, S. A., 2000. Learning strategies and writing processes of proficient vs. less-proficient1 learners of Arabic. Foreign Language Annals, 33 (5), 522–533. Kormos, J., 2011. Task complexity and linguistic and discourse features of narrative writing performance. Journal of Second Language Writing, 20 (2), 148–161. Kuiken, F., et al., 2010. Communicative adequacy and linguistic complexity in L2 writing. In: I. Bartning, et al., eds., Communicative proficiency and linguistic development: Intersections between SLA and language testing research (Eurosla Monographs, 1). Rome: EuroSLA, 81–100. Kuiken, F. and Vedder, I., 2012. Syntactic complexity, lexical variation and accuracy as a function of task complexity and proficiency level in L2 writing and speaking. In: A. Housen, et al., eds. Dimensions of L2 performance and proficiency: Complexity, accuracy and fluency in SLA. Amsterdam: John Benjamins, 143–70. Laufer, B. and Nation, P., 1995. Vocabulary size and use: Lexical richness in L2 written production. Applied linguistics, 16 (3), 307–322. Mahfoudhi, A., et al., 2011. Introduction to the special issue on literacy in Arabic. Reading and Writing, 24 (9), 1011–1018. Manchón, R. M., 2014. The internal dimension of tasks: The interaction between task factors and learner factors in bringing about learning through writing. In: H. Byrnes and R. M. Manchón, eds. Task-based language learning: Insights to and from writing. Amsterdam: John Benjamins, 27–52. Mansouri, F., 1995. The acquisition of subject-verb agreement in Arabic as a second language. Australian Review of Applied Linguistics, 18 (2), 65–84. Noor, H. H., 2007. Competency in first language: Does it affect the quality of second language writing. Dirasat: Human and Social Sciences, 34 (2), 412–424. Norris, J. M. and Ortega, L., 2009. Towards an organic approach to investigating CAF in instructed SLA: The case of complexity. Applied Linguistics, 30 (4) 555–578. Ortega, L., 2003. Syntactic complexity measures and their relationship to L2 proficiency: A  research synthesis of college-level L2 writing. Applied linguistics, 24 (4), 492–518. Polio, C. G., 1997. Measures of linguistic accuracy in second language writing research. Language learning, 47 (1), 101–143. Purpura, J. E., 2013. Assessing grammar. In: A. J. Kunnan, ed. The Companion to Language Assessment. Oxford: Wiley, 100–124. Raish, M., 2017. The measurement of the complexity, accuracy, and fluency of written Arabic. Doctoral dissertation, Georgetown University, Washington, DC. Rammuny, R. M., 1976. Statistical study of errors made by American students in written Arabic. Al‘Arabiyya, 75–94. Ryding, K. C., 2014. Arabic: A linguistic introduction. Cambridge: Cambridge University Press. Schmitt, N. and Zimmerman, C. B., 2002. Derivative word forms: What do learners know?. Tesol Quarterly, 145–171. Schneider, M. and Connor, U., 1990. Analyzing topical structure in ESL essays. Studies in Second Language Acquisition, 12 (04), 411–427 Scott, M. S. and Tucker, G. R., 1974. Error analysis and English-language strategies of Arab students. Language learning, 24 (1), 69–97. Shakir, A. and Obeidat, H., 1991. Maturity in AFL student-written texts: A case study. Al-‘Arabiyya, (24), 65–81. Shih, M., 1999. More than practicing language: Communicative reading and writing for Asian settings. TESOL Journal, 8 (4), 20–25. Skehan, P., 1998. A cognitive approach to language learning. Oxford: Oxford University Press. Skehan, P., 2009. Modelling second language performance: Integrating complexity, accuracy, fluency, and lexis. Applied Linguistics, 30 (4), 510–532. Tang, R., 2012. Academic writing in a second or foreign language: Issues and challenges facing ESL/ EFL academic writers in higher education contexts. London: Continuum. Towell, R., et al., 1996. The development of fluency in advanced learners of French. Applied Linguistics, 17 (1), 84–119.

284

Comparing the complexity of written Arabic Tweissi, A., 1990. Foreigner talk in Arabic: Evidence for the universality of language simplification. In: M. Eid and E. Mccarthy, eds. Perspectives on Arabic linguistics II. Amsterdam: John Benjamins, 296–326. Vann, R. J., 1979. Oral and written language relationships: The question of universals. Available from: https://eric.ed.gov/?id=ED191051 (ERIC Document Reproduction Service No. ED191051). [Accessed 4 October 2017] Wahba, K. M., 2006. Arabic language use and the educated language user. In: K. M. Wahba, et al., eds. Handbook for Arabic Language Teaching Professionals in the 21st Century. Mahwah, NJ: Lawrence Erlbaum Associates, 139–157. Williams, J., 2012. The potential role (s) of writing in second language development. Journal of Second Language Writing, 21 (4), 321–331. Wolfe-Quintero, et al., 1998. Second language development in writing: Measures of fluency, accuracy, and complexity. Honolulu: University of Hawaii Press.

285

PART VI

Arabic L2 speaking and intercultural learning (in study abroad)

14 CODE-SWITCHING IN L2 ARABIC COLLABORATIVE DYADIC INTERACTIONS Khaled Al Masaeed

Drawing on approximately seven hours of audio-recorded data and insights from the markedness model of code-switching (Myers-Scotton 1993, 1998, 2006), the current study explored the use of code-switching (CS) in an Arabic as a second language study abroad program. The conversations from which the data were drawn took place between second language (L2) learners of Arabic and their Arabic native speaker language partners in dyads. Results of this study showed that, overall, participants used CS between English and Arabic in this particular L2 learning environment to negotiate meaning and identities in order to keep the conversation going, show solidarity, and signal the interlocutors’ acknowledgment of their mutually bilingual identities and roles in these interactions. Results also showed that CS tends to occur in the following contexts: (1) when the learners lacked a specific linguistic skill needed to maximize their speaking practice and maintain understanding between them and their speaking partners to bridge these linguistic gaps, and ask for assistance when needed; (2) when the speaking partners provided learners with chunks of talk to foster their lexical learning; and (3) when both learners and their speaking partners showed solidarity and maintained intersubjectivity as mutually bilingual speakers, normally at the periphery of speaking sessions. Based on these results, the current chapter contributes to the existing research on CS in second language learning by sociolinguistically exploring the judicious use of CS and its role in enhancing L2 learning, especially in small-group speaking practice activities. Finally, to argue for more flexibility in allowing the utilization of sensible CS in second/foreign language environments, the study concludes with a discussion about pedagogical implications for the use of CS in target language speaking practice contexts such as the ones in this study and suggests directions for future research for a better understanding of this linguistic practice.

14.1 Introduction The last few years have witnessed a rapid expansion of research exploring L2 Arabic in study abroad contexts, which has yielded valuable insights in code-switching (e.g., Al Masaeed 2014, 2016); the acquisition of phonological variants (Raish 2015); learning regional dialects and developing intercultural communicative competence (Shiri 2013, 2015a, 2015b); and issues regarding access to and contact with native speakers of Arabic outside the classroom (Trentman 2013a, 2013b). However, the field is still under-researched, especially given the 289

Khaled Al Masaeed

need for quality L2 Arabic instruction as enrollment increases in the United States in this critical target language. One area that needs attention in L2 Arabic is the use of the first language (L1) in foreign/second language contexts. This practice has usually been referred to as CS, and work in second language acquisition (SLA) has produced a significant number of studies concerning the role of the L1 in L2 learning. CS is rather commonly used by learners and their instructors (Levine 2011), and a great deal of recent L2 research has advocated the use of CS in L2 contexts for its positive role in maximizing learning (e.g., Al Masaeed 2014, 2016; Forman, 2016; Turnbull and Dailey-O’Cain 2009; von Compernolle 2015; Üstünel and Seedhouse 2005). However, research on CS in L2 learning has overwhelmingly focused on this practice in classroom settings rather than in other L2 contexts, like one-on-one speaking practice sessions between native and non-native speakers (Al Masaeed 2016, Thompson and Harrison 2014). Therefore, the present study explored the motivations and functions of CS in speaking sessions between learners of L2 Arabic and their native speaker language partners in dyads in a study abroad context.

14.1.1 Background L2 research has advocated the use of CS in L2 contexts for its positive role in maximizing learning in the L2 classroom. For example, Butzkamm (1998) demonstrated that teachers and students utilize CS not only to compensate for lexical gaps but also as a conversational lubricant for educational purposes (i.e., maximizing learners’ ability to participate in and contribute to class discussions). This point was also confirmed by Antón and DiCamilla’s (1999) seminal study in which they showed that the use of CS in pair work in the classroom goes beyond making up for lexical needs to mediate interaction and aids in scaffolding to achieve learning tasks among L2 learners in their pair work. Other studies in L2 contexts also showed that CS can serve social functions in the classroom. Canagarajah (1995) concluded that learners used their L2 for official interactions that had to do with the textbook and the lesson in question, while the L1 was used for all other personal and unofficial interactions. Similar results were reported by Hancock (1997), who pointed out that CS serves social functions in the classroom and can be of two types: (1) on-record CS, which takes place when learners are working on class activities; and (2) off-record CS, which is used for personal (unofficial) interactions. Hancock goes further to posit that the unmarked code for on-record language use is L2, whereas the unmarked language for off-record activities is learners’ L1. All these studies agree that CS can serve personal and official functions in the L2 classroom. This indicates that learners’ use of CS in L2 contexts may go beyond simple compensation for lack of lexical competence. However, these and other previous studies of CS in L2 contexts have examined this linguistic practice primarily in L2 classrooms teaching languages other than Arabic. Therefore, Al Masaeed’s (2014, 2016) studies are among the first to have examined this practice in L2 settings outside the classroom (conversations in the L2 with native speaker language partners in dyads in study abroad contexts). Al Masaeed (2014) used the conversational analysis approach to examine the functions of Arabic-English CS in conversations that took place between students and speaking partners in one-on-one speaking sessions in an Arabic as a second language study abroad program. His analysis yielded the following functions of CS: quotation or reported speech, clarification, inclusion or exclusion of certain speakers in the conversation, language play and joking, and language negotiation. Al Masaeed (2016) investigated CS between learners and their speaking partners from the vantage point of sociocultural theory and concluded that “judicious use of students’ L1 for certain purposes does indeed work as one of several interactional resources that contribute to enhancing the development 290

Code-switching in dyadic interactions

of the L2 by operating as a mediating tool to cultivate communication and L2 learning” (p. 716). As can be noted from these previous studies, while there is robust research on CS in L2 learning in the classroom, CS in SLA contexts outside the classroom has not been substantially explored and needs to be investigated further to gain a better understanding of how it is employed in these contexts. Therefore, the current study aims at exploring how CS is employed between L2 Arabic learners and their interlocutors, who are Arabic native speakers, in dyads outside the classroom. To explore how CS was used in this environment and what functions it served, the study drew on Myers-Scotton’s markedness model (MM) of CS (see, e.g., Myers-Scotton 1993, 1998, 2006). Simply put, the MM integrates micro-level and macro-level perspectives in the study of CS to explain its motivations and functions. Myers-Scotton (1998) argues that CS is indexical of the norms of the surrounding society and that “individuals exploit the relationships that become established in a community between a linguistic variety and who uses the variety, and where and how it is used” (p. 18). CS is based on the calculations of the costs and rewards of one choice over another. In other words, speakers as individuals make choices from their linguistic repertoire to achieve certain goals that are important to them. They, for example, code-switch to negotiate different identities. In fact, Myers-Scotton has been influenced by Grice’s (1975) cooperative principle, in which he describes the conventions that speakers normally adhere to in order to have a successful conversation. According to Grice, conversations adhere to the cooperative principle, which encompasses the maxims of quantity, quality, relation, and manner. The maxim of quantity states that speakers’ contributions should be as informative as required but not more; the maxim of quality states that speakers should say only what they believe to be true; the maxim of relation states that speakers’ contributions should be relevant to the discussion; and the maxim of manner states that speakers should avoid ambiguity or obscurity. Hence, Myers-Scotton (1998, p. 26) proposed five related maxims to explain the code choice that speakers make in CS. The first is The Unmarked Choice Maxim: Make your code choice the unmarked index of the unmarked rights and obligations set in talk exchanges when you wish to establish or affirm that rights and obligations set. In other words, CS is unmarked if it is predicted by the addressee, because that particular choice reveals the norms and behavioral rules that are prevalent in the speech community. The second maxim is The Marked Choice Maxim: Make a marked code choice which is not the unmarked index of the unmarked rights and obligations set in an interaction when you wish to establish a new rights and obligations set as unmarked for the current exchange. This is when speakers use CS, where it is not predicted, in order to negotiate their relative status or position. The third maxim is The Exploratory Choice Maxim: When an unmarked choice is not clear, use switching between speech varieties to make alternate exploratory choices as (alternate) candidates for the unmarked choice and thereby as an index of a rights and obligations set which you favor. 291

Khaled Al Masaeed

Here, CS occurs when the speaker is not sure what the unmarked code is in a particular context or with a particular interlocutor. The fourth maxim is the Deference Maxim: Switch to a code which expresses deference to others when special respect is called for by the circumstances. CS here is used to denote a change of footing/roles. The last maxim is the Virtuosity Maxim: Switch to whatever code is necessary in order to carry on the conversation/accommodate the participation of all speakers present. This maxim is about accommodating interlocutors for the sake of better communication (for a detailed discussion and examples of MM maxims, see Myers-Scotton 2006). These maxims offer a more comprehensive framework for analyzing CS than simply assigning a list of discourse functions to CS, as in Gumperz (1982) and Auer (2007). The choice to engage in CS is made based on speakers’ linguistic backgrounds, their experiences in a societal community, and their own goals that they would like to achieve in their interactions. Therefore, the MM is a powerful tool with substantial potential to account for CS. In addition, the MM, as Fägersten (2012) posits, “is also particularly applicable to analyses of second and foreign language classroom discourses, where the marked and unmarked codes may clearly correspond to interaction type, sociocultural values, or discursive goals” (p. 86). However, as I have demonstrated so far, research tends to examine CS in L2 contexts in classrooms and, therefore, research has been scarce when it comes to examining CS in study abroad contexts in small groups/pairs outside the classroom. Thompson and Harrison (2014), for example, point out that “studies of language choice and CS behaviors might be investigated during small-group activities and in teaching-learning contexts in which students are required to learn some of the course content independently, outside of class” (p. 334). Furthermore, the study of Arabic in study abroad contexts is still underexplored (Shiri 2013, Trentman 2013a, Al Masaeed 2016).

14.2  Research questions The current study utilized the MM to explore the motivations and functions of CS in Arabic speaking practice conversations that took place in dyads. Specifically, the study focused on CS from Arabic to English by speakers whose L1 is Arabic or English and expands on Al Masaeed (2014, 2016) by addressing the following primary questions: 1 What are the motivations and functions of Arabic-English CS that characterize the speaking practice sessions in the current study? 2 What are the pedagogical insights that can be gleaned from the data?

14.3 Methods 14.3.1  Context, data, and participants The program from which the data for the present study is drawn is an eight-week, federally funded summer study abroad program in Fes, Morocco, that offers American university students the opportunity to study Arabic as a critical language. The data for this study was collected from the 2011 program that took place from the beginning of June to the beginning of August. 292

Code-switching in dyadic interactions

L2 Arabic students who participated in the program came from different majors and fields of study and from different universities in the US. For this study there was a total of 26 participants: 17 students (three males and 14 females) and nine native Arabic-speaking partners (three males and six females). The students who participated in the study were enrolled in graduate (two students) and undergraduate (15 students) programs; all of them were native speakers of English, whose Arabic proficiency level on the American Council on the Teaching of Foreign Languages (ACTFL) scale ranged from Intermediate-Low to Intermediate-High. Prior to their departure from the US, and thus before arriving at the program site in Morocco, all students took an oral proficiency interview (OPI) by phone, and had to have taken at least two courses of Arabic. Other than graduate versus undergraduate standing and proficiency ratings, age information related to the participants was not available to the researcher. All of the native speaker language partners were undergraduate and graduate Moroccan students majoring in English studies or linguistics and were native speakers of Arabic and L2 speakers of English. Although it was possible that the speaking partners also knew other languages such as French, Spanish, or Tamazight, these languages were not the focus of this study. All speaking partners were hired to work for the program after participating in a pedagogical training workshop that took place prior the start of the program. The data consisted of 17 audiorecorded speaking sessions, about 25 minutes each, resulting in a total of approximately 7 hours of recorded data for the analysis. The recorded Arabic-English data was transcribed according to the transcription conventions and abbreviations provided in the Appendix.

14.3.2  Speaking sessions Speaking partners and students had four mandatory one-on-one speaking sessions per week (Monday-Thursday). In each session, a student met with a speaking partner for 20–30 minutes. Speaking partners were trained to conduct the conversations according to ACTFL OPI interview guidelines. The sessions addressed various topics that had already been studied and discussed in class, which depended on the materials covered within the program and what was happening in the community, the country, the Arab world, and the world in general. Some of the topics were, for example, immigration between Morocco and Europe, immigration and health care in the US, movies that students watched in classes, shopping and travelling around Morocco, and so forth. Both students and speaking partners were urged to speak Arabic as much as possible. On the use of English, they were instructed as follows: let English be your last resort; and if you must say something in English, say it and continue in Arabic.

14.4 Results As previously stated, the MM distinguishes between marked and unmarked CS. Marked CS is a switch to the marked code in order to negotiate different identities between speakers. Such a switch, according to the MM, generally occurs to increase or decrease social distance between speakers. Therefore, marked CS can be positive or negative depending on the situation and relationship between the participants. According to Myers-Scotton, CS is based on the calculation of the costs and rewards of one choice over another. Speakers as individuals make choices from their linguistic repertoire to achieve certain goals that are important to them, and thus they code-switch to negotiate different identities. Analysis of the data revealed a total of 91 instances of CS: 69 of these switches were initiated by the learners, and 22 were initiated by the speaking partners. In the current study, each one of these instances of CS is considered a marked code, because Arabic is the expected code 293

Khaled Al Masaeed

of communication in this language program, and both students and speaking partners are urged to maximize the use of Arabic to enhance learners’ speaking production. However, it is worth pointing out that an important assumption of the current study is that communicative competence (Hymes 1971, Canal and Swain 1980, Bachman 1990) includes not only linguistic competence but also the ability to use language appropriately according to context. Therefore, the use of CS in this study is looked at as one way of expressing communicative competence, through which both learners and their speaking partners take part in a dialogic process to negotiate meaning and enhance L2 learning. Next, I present five examples of CS to show the motivations and functions of instances of CS found in the data. Example (1) is a conversation in which a student (Kelley, a pseudonym, just as for all names in this chapter) and her speaking partner (Layla) are acting out a hypothetical situation where the student is calling the speaking partner to invite her to watch a movie in the cinema. They have just agreed on the time and the place, and Kelley wants to end the (hypothetical) phone call. All instances of CS from Arabic to English in the examples discussed in the present study are placed in quotation marks in the original transcript and in the italic English translation provided, when translation is needed.

EXAMPLE (1) 1 Kelley: 2 Layla: 3 Kelley: 4 Layla:

kayfa naquul (.) “see you later”? how do we say (.) “see you later”? ʔaraaka laaħiqan see you later ((in Arabic)) ʔaraaka laaħiqan see you later ((repeating in Arabic)) naʕam (.) maʕa s-salaamah yes (.) good bye

The conversation is taking place in Arabic, the unmarked code. But Kelley wants to say “see you later,” but she does not know how to say it, and therefore she switches to English to ask Layla about how to say it in Arabic. The speaking partner here is performing a “scaffolding” role where the student is feeling safe enough to rely on her for support in carrying on the roleplay situation. This kind of CS in which the learner asks his/her instructor, classmate, or a speaking partner about how to say something in the target language is quite common in L2 learning contexts, as is the response of the interlocutor; this is thus indexical and denotes a role change in which the student (Line 1) is acknowledging her role as a learner and calling on the speaking partner to position herself as the experienced interlocutor who can provide assistance. The speaking partner accepts this role and accommodates the student (Line 2) by providing the phrase in Arabic (the unmarked code). Hence, CS in this interaction is marked. In Example (2), the speaking partner (Ahmad) is asking the student (Sally) a question about health insurance in the US.

EXAMPLE (2) 1 Ahmad: ehm (.) ʕindi suʔaal ʕan ʔat-taʔmiin ʔaṣ-ṣiħii fii ʔal-wilaayaat 2 ʔal-muttaħidah (.) ʔat-taʔmiin ʔaṣ-ṣiħii,  ehm (.) I have a question on health insurance in the United States (.) health insurance, 294

Code-switching in dyadic interactions

3 Sally: ʔat-taʔmiin ʔaṣ-ṣiħii ? health insurance? fii ʔal-wilaayaat ʔal-muttaħidah (.) Obama? 4 Ahmad: in the United States (.) Obama? maaðaa yaʕnii ʔat-taʔmiin ʔaṣ-ṣiħii? 5 Sally:  what is health insurance ((saying the Arabic equivalent))? “health insurance” 6 Ahmad: 7 Sally: “oh (.) okay” 8 Ahmad: fii ʔamriikaa (.) maa hiya baʕḍ ʔal-mašaakil ʔal-latii wa limaaðaa baʕḍ ʔan-naas laa yuriiduun ʔat-taʔmiin ʔaṣ-ṣiħii ?  in America (.) what are some of the problems that ((he changes the  question)) why don’t some people want health insurance? It is clear from the transcript that the speaking partner used ʔat-taʔmiin ʔaṣ-ṣiħii “health insurance,” but the student did not know what it meant and, therefore, she asks the speaking partner to provide the meaning of that term. It is worth pointing out here that the student asked in Arabic, and that, after an unsuccessful attempt at providing the meaning through hinting (via mentioning the US and Obama) in Arabic in Line 4, the speaking partner is the one who, in Line 6, starts the switch to English in order to respond to the student’s question. It is necessary for the speaking partner to switch to English at this point to save time and frustration; all the student needs is to know what ʔat-taʔmiin ʔaṣ-ṣiħii means so they can stay focused and continue the conversation. This kind of CS is marked because the speaking partner is positioning himself as the expert who can provide assistance to the interlocutor, who is also acknowledging her position as a learner who needs some linguistic help at this point of the conversation. In other words, the switch by the speaking partner recognizes the limited knowledge of the learner and both parties’ desire to minimize frustration as well as the need to keep the conversation going. Marked CS was also found at the periphery of speaking sessions: that is when participants more naturally express their mutual multilingual identities by showing solidarity, offering compliments, making jokes, and establishing a comfortable environment for communicative interactions. Examples (3) and (4) demonstrate this kind of CS. In Example (3), the speaking session is about to conclude, and the next student (Gabe) is waiting at the door to begin a new session. Once the current student leaves the office, the speaking partner (Nour) asks Gabe, who has been standing at the door, about the time.

EXAMPLE (3) 1 Nour: kam ʔas-saaʕah? what time is it? ehm (.) “three thirty” 2 Gabe: ehm (.) “three thirty” “wow ((clapping and laughing)) on time” 3 Nour: “wow ((clapping and laughing)) on time” 4 Gabe: bi al (.) bi ʔal-ʕarabiyyah “three thirty” (.) bi ʔal-ʔingliiziyyah in (.) in Arabic “three thirty” (.) in English [bi ʔal-ʔingliiziyyah 5 Nour: [in English] 6 Gabe: θalaaθah wa niṣf 295

Khaled Al Masaeed

three thirty ((he says it in Arabic)) 7 Nour: “wow” (laugh) (2.0) ʔahlan wa sahlan “wow” (laugh) (2.0) welcome 8 Gabe: “two” ʔahlans “to you” “two” welcomes “to you” 9 Nour: shukran (laugh) (2.0) ʔahlayn (.) “three” ʔahlans “to you”  thank you (laugh) (2.0) welcome (.) “three” welcomes “to you” Gabe answers in English and then turns this switch into a joke. CS here is marked because both the student and the speaking partner are using English and a play on words for the sake of joking and showing solidarity. Such a switch is not expected in this conversation, but both speakers managed to turn it into a successful introduction to their speaking practice. Obviously, both accept the switch and seem to like it. CS in this exchange is of a potential learning benefit because it contributes to a safe and enjoyable atmosphere for language learning. A close look at the conversation reveals some linguistic creativity via the play with words, which thus enhances language gain. It is possible that the student’s answer in English at the beginning of the exchange (Line 2) is not planned, since Arabic is the unmarked code. However, Gabe creatively and successfully turns his use of English into a linguistic game, which the speaking partner likes and joins. This kind of CS is marked, because both interlocutors are signaling their ability to create with language and that they can be more informal by ignoring sticking to the unmarked code (i.e., Arabic). Example (4) is another instance of CS as the marked code that occurs at the periphery of the speaking session; it is drawn from the end of a speaking session between a speaking partner (Saleem) and a student (Mona). The speaking partner asks Mona if she is happy with her progress in speaking. The example begins with Mona’s response, in which she initiates the switch to English.

Example (4) 01 Mona: . . . uhm (.) uhm saʔadrus ʔal-mufradaat wa ʔal-ʕibaaraat (.) haaðaa muškilatii (.) laa ʔastaʕmil ʔal-ʕibaaraat (.) haaðaa yaʕnii (.) haaðaa 02 . . . um (.) um I’ll study the vocabulary and the expressions (.) This is my problem (.) I do not use the expressions (.) this means (.) this 03 Saleem: [haaðaa sayusaaʕid kathiiran [This will help a lot 04 Mona: ʔiðaa kaan ʔafʕal haaða (.) uhm (.) “like I’ll see more in a better” (). if I do this (.) uhm (.) “like I’ll see more in a better” (). 05 Saleem: “but in general (.) like (.) I mean (.) I mean this is the first time I speak with you as a speaking partner (.) but it was very interesting 06 07 (.) like the way how you speak Arabic and also like (.) how you express yourself in Arabic (.) you know (.) it really” 08 “[really?” 09 Mona: 10 Saleem: “yeah” 11 Mona: šukran Thank you 12 Saleem: maaši muškila No problem 296

Code-switching in dyadic interactions

Mona started her switch from Arabic to English to elaborate on what she thinks she needs to pay more attention to in order to improve her speaking skill in Arabic. At that point, Saleem deploys CS as a strategy to show solidarity through praise and encouragement, and to signal his bilingual identity. Then Mona switches back to Arabic to stay on task (i.e., keeping the interaction in the unmarked code) in Line 11, and so does Saleem in Line 12. One might argue that both interlocutors are aware of the fact that the session is winding down and that English may now be permitted and Mona’s desire to elaborate is most easily accomplished in the language in which she is most comfortable. However, this use of English is marked due to the context of the program and the nature of these sessions (i.e., the goal is to practice Arabic), and its marked status is also conveyed through the interlocutors’ switch back to Arabic in the last two lines. Example (5) is a clear instance of CS as the marked code. A student (Amanda) is going over with her speaking partner (Nabeel) what she wrote about a movie she watched in class recently.

EXAMPLE (5) 1 Nabeel: ʔuriidu ʔan ʔasmaʕa maaðaa katabti I’d like to listen to what you wrote yata (2.0) yataʕallamaan 2 Amanda: yata (2.0) they both learn [ah hah 3 Nabeel: 4 Amanda: “[I think it’s missing something” ah hah 5 Nabeel: 6 Amanda: laa (.) “I think it’s missing something (.) it’s messed up (.) hold on” No (.) “I think it’s missing something (.) it’s messed up (.) hold on” ehm (.) ʔanti faqaṭ ʔiqraʔii maa katabtii wa ʔanaa sawfa ʔara 7 Nabeel: ehm (.) you just read what you wrote and I’ll see “there is something wrong with it (.) look!” 8 Amanda: 9 Nabeel: aah (.) min hunaa yabdaʔ ? hunaa ʔal-bidaayah ? aah (.) it starts from here (.) this is the beginning? 10 Amanda: “this is weird (.) looks like it’s backward or something (.) see?” 11 Nabeel: ah hah (.) ʔanti faqaṭ ʔiqraʔii wa maa huwa xaṭaʔ sawfa nuṣaħħiḥuhu (.) faqaṭ kaan hunaak muškil fii ʔaṭ-ṭibaaʕah (.) ʔanti 12 13 ṭabaʕtii ʔal-waraqah wa kaan hunaak muškil fii ʔal-“software”  ah hah (.) you just read and we’ll fix what’s wrong (.) it was just a problem with the way it was printed (.) you printed out the paper and there was a problem with the “software” 14 Amanda: naʕam (.) “I’ll just read then” yes (.) “I’ll just read then” 15 Nabeel: “okay” “even if it’s screwed up?” 16 Amanda: 17 Nabeel: “okay” “because of the printer and my computer” 18 Amanda: 19 Nabeel: “okay” (laugh) ((and starts reading)) 20 Amanda: The student seems to be struggling with reading what she just printed out. This is clear in Line 2 where she pauses for about 2 seconds. Then she switches to English and resists switching 297

Khaled Al Masaeed

back to Arabic. It is possible that she is trying to figure out what might have happened to the piece she wrote. She keeps negotiating in English to make sure that the speaking partner knows it is the printer and the software that caused the written passage to look strange. Her switch to English can be explained based on her linguistic background; she wants to explain herself clearly to avoid the blame. The speaking partner, on the other hand, continues to speak Arabic, perhaps to remind or encourage the student to stay true to the session’s goal – Arabic speaking practice. The student agrees to start reading the Arabic text after she made sure that the speaking partner recognized her point. So, CS is marked in this excerpt in the sense that it is not expected (especially by the speaking partner), and this is why the speaking partner tries to keep speaking Arabic and avoid switching to English. In addition, the speaking partner is negotiating his role as a speaking partner in Arabic, not in English; he is sticking to what is expected of him from the program’s point of view. Although this instance of CS is marked, the student’s switch from Arabic to English also plays a key role in negotiating a point of understanding between the two speakers. CS in this example demonstrates why a certain code is (not) chosen over another; that is, speakers always “seek to optimize their outcome” (Myers-Scotton 1998, p. 30). CS is an important communicational tool, and not just a rejection of the rule to speak only Arabic, that speakers/learners utilize to solve learning problems like the one in this specific interaction.

14.5 Discussion Before discussing the results of the data from the vantage point of the MM (i.e., the first research question), it is worth pointing out that the analysis shows the following: (1) speaking sessions are student-dominated, as they are designed for students to practice their speaking skills and, therefore, instances of CS started by students outnumber those initiated by speaking partners (69 vs. 22); and (2) in spite of the fact that CS occurred in each speaking session at least once, the overall frequency of CS was relatively low in this context (91 total), and most of the switches were short. Overall, then, both students and speaking partners do their best to stay focused on the task to speak Arabic as much as possible during these speaking practice sessions. The application of the MM indicates that the main characteristic of CS in the context of these speaking sessions is marked. Given that this framework takes societal norms and speaker’s goals into consideration, all instances of CS in the data are found to fit under the marked choice due to the nature of the language program and the expectations of these speaking sessions that urge participants to maximize the use of Arabic (the unmarked code) to enhance learners’ skill in speaking Arabic. Moreover, participants were found (as demonstrated by the examples in the results section) to engage in CS to negotiate meaning and identities in order to keep the conversation going, show solidarity, and signal the interlocutors’ acknowledgement of their mutually bilingual identities and roles in these interactions. To this end, the data showed that a switch to English tends to occur in contexts such as those discussed here. The first is learner-initiated: When the Arabic L2 learners, who are L1 speakers of English, do not have the linguistic ability to keep the conversation going in Arabic, they utilize more CS than speaking partners. Marked CS in the current data serves different functions for the learners and their speaking partners. The students use CS to (1) request the meaning of vocabulary and expressions, as in Example 1; (2) create with language for joking and showing a bilingual identity, as in Example 3; (3) finish a thought, as in Example 4; or (4) bridge a communication gap, as in Example 5. The second context constitutes CS that is speaking-partnerinitiated: such switches by speaking partners acknowledge the limited knowledge of the 298

Code-switching in dyadic interactions

language learner, both parties’ desire to minimize frustration, and the need to keep the conversation going. For example, in cases involving Arabic lexical items unknown to the student, and which the speaking partner feels that the student would not be able to follow or that it would take too much time and the simpler solution is to quickly do the switch and then go back to Arabic, as in Example 2. In switching to English, speaking partners are able to accommodate the student while offering encouragement that will be understood. Another point that can be taken from the data is that although the instances of CS tend to be short, CS that takes place at the periphery of the speaking session tends to be a little bit longer than those instances that occur during the speaking session, as can be concluded from Examples (3) and (4), which has the consequence of creating solidarity between each other as mutual bilingual speakers. In general, the utilization of CS in these examples of CS goes hand in hand with what Myers-Scotton (1998) posits: that “[c]hoices are marked when they can be seen as negotiations to invoke a rights and obligations set other than the unmarked one for the context” (p. 6, original italics). This applies to all the student-initiated requests for Arabic lexical items or for translations that become opportunities for speaking partner-scaffolded adjacency pairs that move the conversation along, provide the student with the information requested, and, thus, construct the students as active participants and autonomous learners. It also applied to CS initiated by speaking partners who take on the role of language experts to scaffold their learners’ interactions in L2 Arabic. To address the second research question, significant pedagogical insights can be gained from the data and the previous discussion. Judicious use of CS can be seen as a tool to enhance language learning via interaction and the collaborative work between students and their speaking partners, as in the context of the current study. From a sociocultural perspective, CS functions as a scaffolding tool (Moore 2013), where learners and their speaking partners can use CS to smooth speaking and expressing ideas. The data showed that speaking partners provided students with limited chunks of talk to help them achieve this goal. CS also works as a pragmatic device that reflects learners’ communicative competence, which they can utilize to maximize learning and help them become true bilingual speakers. Language teaching and learning are becoming not only more communicative but also more interactive all the time, and the advancement of technology has increased learners’ access to instant communication with different interlocutors from different languages, cultures, and countries. This process in L2 contexts should be expected, and it requires some flexibility in terms of CS. This means that directors of L2 programs should take into consideration the reality of language use in contexts like the one in the current study, where interlocutors of the kind examined here demonstrate significant advantages of CS that bilinguals have at their disposal. In order for this to take place in a beneficial manner, speaking partners should have at least some training on when and how to utilize and allow CS in language learning contexts and how to communicate this to their students. This can be done through showing examples from the ample research that demonstrates how CS can enhance L2 learning by being employed as a resource for richer conversations at all levels among L2 learners. Learners at this stage are not expected to use L2 exclusively in all domains of their lives. There are some topics that may be expressed more naturally in the learner’s first language. And, therefore, small amounts of CS could potentially help learners’ progress toward becoming better bilinguals and increase their confidence and communicative competence. Finally, the thorny problem that remains challenging in this field is that despite the great deal of L2 research that shows ample evidence for the benefits of CS in learning in general, and in L2 in particular, advocating for CS is still a difficult task because it goes against what most policy makers believe. Makoni and Mashiri (2007), for example, argue that we should 299

Khaled Al Masaeed

develop language policies that encourage language fluidity to better understand the social realities of their speakers instead of trying to strengthen policies that attempt to hermetically seal languages off from each other. But as Creese and Blackledge (2010) point out: “Research shows that code-switching is rarely institutionally endorsed or pedagogically underpinned” (p. 105). Therefore, applied linguists (especially in the field of L2 Arabic) are urged to publish more on this topic in an attempt to keep raising the awareness of the benefits of CS for all of those involved in L2 learning and teaching contexts. This age of globalization calls upon us for more flexibility when it comes to CS in order to keep up with the linguistic reality of our modern world and authentic language use. When forming policies and curricula, educators should consider research that shows, as Macaro (2009) argues, that students and their teachers still use their L1 in L2 and bilingual environments even when policies dictate otherwise.

14.6  Conclusion and pedagogical implications The present sociolinguistic study investigated the linguistic practice of CS by Arabic L2 learners and their Arabic native speaking partners during speaking practice sessions in a study abroad context. In order to reach this goal, the MM was applied under the subfield of sociolinguistics to the Arabic-English data from this study. The data showed that CS is marked and has different potential benefits and functions in this particular context of the program. The study also demonstrated benefits of allowing the utilization of judicious CS in low-proficiency bilingual programs to mirror the linguistic reality of our modern world, with its increasingly globalized teaching focus. In this study the markedness model was applied to argue for using CS sensibly and judiciously to help learners progress toward becoming better bilinguals, boost trust among learners and their speaking partners, encourage them to signal their mutually bilingual identities, and enhance participation and smooth exchange of ideas. Future research might take the following into account: perceptions of both learners and teachers about CS, what goes on in conversations to see if participants’ use of CS does (not) mirror their perceptions, and following participants throughout the semester/period of the program to observe their use of CS. In addition, some previous studies posited that CS decreases as L2 proficiency increases, and future research should, therefore, look at the use of CS among L2 learners with different proficiency levels or even with longer periods of study abroad to gain more insights into this bilingual practice. Moreover, the present study urges applied linguists to publish more work on this topic to contribute to destigmatizing CS in SLA and bilingual programs.

Appendix Transcription Conventions and Abbreviations (.) micro-pause [ simultaneous talk [sometimes called overlapping] length of silence in seconds (2.0) (()) extra-textual information provided by the transcriber as well as transcriber’s comments ? question rising intonation , exclamation mark ! (laugh) laugh () Unintelligible 300

Code-switching in dyadic interactions

References Al Masaeed, K., 2014. Conversational Arabic-English code-switching. Academic Exchange Quarterly, 18 (2), 47–56. Al Masaeed, K., 2016. Judicious use of L1 in L2 Arabic speaking practice sessions. Foreign Language Annals, 49 (4), 716–728. Antón, M. and Dicamilla, F. J., 1999. Socio-cognitive functions of L1 collaborative interaction in the L2 classroom. The Modern Language Journal, 83 (2), 233–247. Auer, P., 2007. The pragmatics of code-switching: a sequential approach. In: Li Wei, ed. The bilingual reader. 2nd ed. New York: Routledge, 123–138. Bachman, L. F., 1990. Fundamental considerations in language testing. Oxford: Oxford University Press. Butzkamm, W., 1998. Code-switching in a bilingual history lesson: the mother tongue as a conversational lubricant. International Journal of Bilingual Education and Bilingualism, 1 (2), 81–99. Canagarajah, A. S., 1995. Functions of codeswitching in ESL classrooms: socialising bilingualism in Jaffna. Journal of Multilingual and Multicultural Development, 16 (3), 173–195. Canale, M. and Swain, M., 1980. Theoretical bases of communicative approaches to second language teaching and testing. Applied Linguistics 1 (1), 1–47. Creese, A. and Blackledge, A., 2010. Translanguaging in the bilingual classroom: a pedagogy for learning. Modern Language Journal, 94, 103–115. Fägersten, K. B., 2012. Teacher discourse and code choice in a Swedish EFL classroom. In: B. Yoon and H. K. Kim, eds. Teachers’ roles in second language learning: Classroom applications of sociocultural theory. Charlotte, NC: Information Age Publishing, 81–98. Forman, R., 2016. First and second language use in Asian EFL. Bristol, UK: Multilingual Matters. Grice, H.P., 1975. Logic and conversation. In: P. Cole and J. Morgan, eds. Syntax and semantics 3: Speech acts. New York: Academic Press, 41–58. Gumperz, J. J., 1982. Discourse strategies. Cambridge: Cambridge University Press. Hancock, M., 1997. Behind classroom code switching: Layering and language choice in L2 learner interaction. TESOL Quarterly, 31 (2), 217–235. Hymes, D., 1971. On communicative competence. Philadelphia: University of Pensylvania Press. Levine, G. S., 2011. Code choice in the language classroom. Bristol, UK: Multilingual Matters. Macaro, E., 2009. Teacher use of codeswitching in the second language classroom: exploring ‘optimal’ use. In: M. Turnbull and J. Dailey-O’Cain, eds. First language use in second and foreign language learning. Clevedon, UK: Multilingual Matters, 35–49. Makoni, S. and Mashiri, P., 2007. Critical historiography: does language planning in Africa need a construct of language as part of its theoretical apparatus? In: S. Makoni and A. Pennycook, eds. Disinventing and reconstituting languages. Clevedon, UK: Multilingual Matters, 62–89. Moore, P. J., 2013. An emergent perspective on the use of the first language in the English-as-a-foreign language classroom. Modern Language Journal, 97, 239 – 253. Myers-Scotton, C., 1993. Social motivations for codeswitching. New York: Oxford University Press. Myers-Scotton, C., 1998. A theoretical introduction to the markedness model. In: C. Myers-Scotton, ed. Codes and consequences: Choosing linguistic varieties. New York: Oxford University Press, 18–38. Myers-Scotton, C., 2006. Multiple voices: An introduction to bilingualism. Malden, MA: Blackwell. Raish, M., 2015. The acquisition of an Egyptian phonological variant by US students in Cairo. Foreign Language Annals, 48 (2), 267–283. Shiri, S., 2013. Learners’ attitudes toward regional dialects and destination preferences in study abroad. Foreign Language Annals, 46, 565–587. ———, 2015a. The homestay in intensive language study abroad: Social networks, language socialization, and developing intercultural competence. Foreign Language Annals, 48 (1), 5–25. ———, 2015b. Intercultural communicative competence development during and after language study abroad: Insights from Arabic. Foreign Language Annals, 48 (4), 541–569. Thompson, G. L. and Harrison, K., 2014. Language use in the foreign language classroom. Foreign Language Annals, 47 (2), 321–337. Trentman, E., 2013a. Arabic and English during study abroad in Cairo, Egypt: Issues of access and use. The Modern Language Journal, 97, 457–473. ———, 2013b. Imagined communities and language learning during study abroad: Arabic learners in Egypt. Foreign Language Annals, 46 (4), 545–564.

301

Khaled Al Masaeed Turnbull, M. and Dailey-O’Cain, J., 2009. First language use in second and foreign language learning. Clevedon, UK: Multilingual Matters. Üstünel, E. and Seedhouse, P., 2005. Why that, in that language, right now? codeswitching and pedagogical focus. International Journal of Applied Linguistics, 15, 302–325. von Compernolle, R. A., 2015. Interaction and second language development: A Vygotskian perspective. Amsterdam: John Benjamins.

302

15 RESEARCH-BASED INTERVENTIONS FOR LANGUAGE AND INTERCULTURAL LEARNING Emma Trentman Study abroad and telecollaboration are two methods valued for the naturalistic language and intercultural learning experiences they can provide (Coleman 2015, O’Dowd 2011). However, research on both study abroad and telecollaboration demonstrates that these activities may not necessarily lead to increased language and intercultural competencies. Research on study abroad shows that there is considerable individual variation in language and intercultural learning during study abroad (Kinginger 2009, Vande Berg and Paige 2012), and research on telecollaboration contains evidence of both successful and failed partnerships (O’Dowd and Ritter 2006, O’Dowd 2011). To improve these outcomes, researchers have emphasized the need for program interventions that move beyond “immersing” students to including guided reflection based on linguistic and cultural evidence (Vande Berg et al. 2012a, O’Dowd 2011). This chapter uses qualitative methods to describe two cases involving research-based interventions for language and intercultural learning. The first case is a student pursuing an ethnographic project for study abroad. The second case is a student and language partner participating in a classroom telecollaboration project. Both projects were designed to provide guided reflection using linguistic and cultural evidence. The findings demonstrate that there were both challenges to implementing these projects as well as benefits in terms of language and intercultural learning. The ethnographic project provided opportunities for interaction, formed a framework for interpreting hospitality experiences in everyday life, and helped with identity negotiation. The telecollaboration project allowed for learning lexical items and cultural practices and led to the discussion of topics beyond the classroom content and at a higher proficiency level. However, both projects could be further developed to better explore cultural rich points. The chapter concludes with pedagogical recommendations for these types of projects.

15.1 Introduction Study abroad and telecollaboration are two methods of extending language and intercultural learning beyond the classroom. Study abroad is valued for the linguistic and cultural “immersion” assumed to occur (Coleman 2015). Telecollaboration exchanges use the internet to pair language learners with fluent speakers to develop their language and intercultural learning

303

Emma Trentman

(O’Dowd 2011). In both cases, there is an emphasis on “authentic” experiences that cannot be gained in the classroom. However, research on study abroad and telecollaboration demonstrates that the benefits of these activities may not be as clear as popularly imagined. Research on study abroad shows that there is considerable variation in language and intercultural learning during study abroad, and that while some students make considerable gains, others do not (Kinginger 2009, Vande Berg and Paige 2012). Similarly, while research on telecollaborative projects contains instances of successful partnerships, there are also many instances of failed partnerships and communication (O’Dowd and Ritter 2006, O’Dowd 2011). As a result, researchers have emphasized the need for program interventions to ensure language and intercultural learning during both study abroad and telecollaboration, and especially the need for guided reflection based on linguistic and cultural evidence (Vande Berg et al. 2012a, O’Dowd 2011). This chapter uses a case study approach to describe two projects using research-based interventions, one during study abroad and the other during telecollaboration. Both projects were designed to provide guided reflection using linguistic and cultural evidence. The results demonstrate that despite numerous challenges in implementing the projects, there were clear benefits in terms of language and intercultural learning as well as potential for further reflection on cultural rich points.

15.1.1  Language and intercultural learning during study abroad Research on study abroad in various settings demonstrates that students do not always integrate into host country social networks and use the target language (Coleman 2013, 2015; Kinginger 2009). While much research on study abroad focuses on U.S. students traveling to European countries, research on U.S. learners of Arabic abroad reveals similar findings. Kuntz and Belnap (2001) found that students studying abroad in Morocco and Yemen reported little use of Arabic outside of class, and that using Arabic outside of class could in fact result in peer criticism for “showing off.” Dewey et al. (2013) found that students’ social networks varied in size and in the languages used (English or Arabic). Trentman (2013a, 2013b) describes how some Arabic learners in Egypt struggled to gain access to local social networks, socializing instead with other international students. Bown et al. (2015) report that students studying in Jordan mostly engaged in intermediate level “pleasantries” and struggled with proficiency gaps and unsympathetic interlocutors who were uninterested or switched to English. Student explanations for their challenges often involve the difficulty of “making friends” with locals (Allen 2010, Bataller 2010, Mendelson 2004), locals desiring to practice English with them (Kinginger 2008, Mitchell et al. 2015, Trentman, 2013a), and unpleasant incidents related to their gender, national, and racial identities abroad (Kinginger 2013), including those in which they are unprepared to engage with local discourses on these topics (Diao and Trentman 2016, Kinginger 2008). For Arabic learners, proficiency challenges may be compounded by a lack of preparation to engage in the local dialects used for everyday social activities (Kuntz and Belnap 2001, Trentman 2013a, Trentman and Diao 2017). Arabic learners also report challenges related to negotiating foreigner, heritage, racial/ethnic, and gendered identities abroad, although it is worth noting that students are also able to negotiate these same identities in ways that positively impact their language and intercultural learning (Bown et al. 2015; Diao and Trentman 2016; Kuntz and Belnap 2001; Trentman 2012, 2015a, 2015b, 2018; Trentman and Diao 2017). In short, despite the popular perception of study abroad as an experience that automatically leads to language and intercultural learning, students may face difficulties achieving these goals abroad. 304

Research-based interventions for language and intercultural learning

15.1.2  Language and intercultural learning during telecollaboration Telecollaboration projects, also referred to as “online intercultural exchange,” use the power of modern communications technology to pair learners with other learners or experts in diverse geographic locations (O’Dowd, 2010, p. 2). The media used in telecollaboration projects range from asynchronous communication such as email, video recordings, or forum communications to synchronous communication such as chat programs or video calls, including combinations of the two. As in study abroad, these more naturalistic forms of communication are expected to result in enhanced language and intercultural learning. However, research on telecollaboration shows a similar dissonance between popular expectations and research findings, where O’Dowd (2011, p. 356) emphasizes that there is a need “to avoid unfounded expectations that the activity will have magical transformative effects on students’ linguistic skills and intercultural awareness.” There is considerable individual variation, with some students developing friendships with their partners while others break off communication or develop negative or superficial evaluations of the other culture (Thorne 2003, Belz 2003, Ware 2005, O’Dowd 2003). Reviewing a variety of studies, O’Dowd and Ritter (2006) identify challenges at four different levels of telecollaborative projects: the learner, the classroom, the socioinstitutional, and the interaction, noting (p. 625) that failures are frequently due to “a combination of interconnected factors from different areas.” Learners vary in their language and intercultural proficiency as well as in their motivations and expectations for the experience (O’Dowd 2003, O’Dowd and Ritter 2006, Ware 2005). At the level of the classroom, the teacher relationship and design of the project impact its success (Belz and Müller-Hartmann 2003, O’Dowd and Ritter 2006). At the socioinstitutional level, challenges arise from differential access to technology (Belz 2003, Thorne 2003), syncing academic schedules and workload expectations (Belz 2002, Belz and MüllerHartmann 2003, Ware 2005), and mismatched language proficiencies resulting from the high prestige afforded to English as a global language (Belz and Müller-Hartmann 2003, Ware 2005, Schenker 2012, Helm et al. 2012). In the interactions themselves, cross-cultural variation in communication styles can lead to learners misinterpreting culturally specific patterns of interaction (Ware 2005; Belz 2003; O’Dowd 2003, 2010). Thus, in telecollaboration experiences, as in study abroad, researchers have emphasized the need for interventions rather than assuming that geographic or cyber proximity will automatically result in language and intercultural learning.

15.1.3  Interventions in study abroad Many study abroad programs do recognize the difficulties students face gaining access to host country social networks and intervene to promote students’ access to locals. These interventions include practices such as living with local hosts, language pledges, language partners, and direct enrollment in local university classes. However, there is evidence that simply providing easier access to local networks is not enough to guarantee positive intercultural interactions and integration (Vande Berg et al. 2012a). For example, research on the homestay experience has found that there is considerable variation in the extent to which learners are incorporated into the family’s social network and engage in interactions beyond basic dialogues (Allen 2010, Jackson, 2008). Describing a program in Tunisia where the host family became a key social network, Shiri (2015) notes the considerable effort made to place students appropriately, provide training for host families, and integrate the homestay with classroom instruction. Vande Berg et al. (2012a) conclude that in addition to providing access to local social networks, program interventions must include developing students’ self-reflexive 305

Emma Trentman

capacity to understand how meaning is created from experiences and how to apply this knowledge to subsequent intercultural encounters. There are a number of approaches that can help develop students’ self-reflexive capacity for reflecting on meaning making, including those described in the latter chapters of Vande Berg et al. (2012b). The approach described in this chapter is ethnographic projects for study abroad.

15.1.4  Ethnographic projects for study abroad Ethnographic projects for study abroad are designed to promote reflection on the creation of cultural knowledge while encouraging learners to investigate host country culture using the target language. Two major projects described in research on study abroad are the Ealing Ethnography project, which focused on U.K. learners abroad for several months in Germany or Spain (Roberts et al. 2001, Roberts 2003) and the Special English Stream for learners from Hong Kong studying in England for five weeks (Jackson 2006, 2008, 2010). In both models, students completed a preparatory class prior to studying abroad, collected data for a topic of their choice while abroad, and wrote up their findings upon their return. As Roberts (2003, p. 123) explains, the purpose of ethnographic projects for study abroad is not to produce professional ethnographies: “They [the students] are not professional ethnographers nor are they attempting a polished account. Rather they are using the ethnographic process to develop a reflexive intercultural understanding which may be transferable to many different settings in the future.” Students develop this reflexive intercultural understanding by sharpening their skills of reflection and observation in their own culture as well as abroad. The need to understand meaning making from a local perspective can push students to engage with locals in ways they might not otherwise and cause them to be less reliant upon stereotypes (Jackson 2006, Roberts et al. 2001, Roberts 2003).

15.1.5  Interventions in telecollaboration As in study abroad, research on telecollaboration emphasizes the need for guided reflection to allow students to develop their linguistic and intercultural skills (Belz 2002, Godwin-Jones 2013, O’Dowd 2011). One advantage of telecollaborative exchanges is that records of chat and email exchanges, and even audio recordings, can serve as tools for linguistic and intercultural analysis. Ware (2005) suggests having students analyze successful and unsuccessful episodes of exchanges prior to beginning their own. Belz and Vyatkina (2008) report positive results when students analyzed linguistic features that influence communication styles in their correspondence during their telecollaboration project. O’Dowd (2011, p. 355) emphasizes the importance of guided reflection in the classroom to explore cultural “rich points.” The term “rich points” was originally coined by Agar (1991, p. 176) to describe points of contact between languages that are difficult for learners to engage with, because “the problematic bit of language is puttied thickly into far-reaching networks of association and many situations of use.” He notes that rich points are frequently points of disagreement for native speakers as well. While investigations of language and intercultural learning in study abroad and telecollaboration are generally distinct (but see Kinginger and Belz 2005, O’Dowd 2010), there are clear parallels between these activities in terms of popular expectations and research findings. The common perception of these activities is that L2 learners will naturally and automatically develop their linguistic and intercultural skills when placed in geographic or cyber proximity to members of the target language community. Yet research demonstrates that there are many instances in which language and intercultural learning during these activities is limited. To 306

Research-based interventions for language and intercultural learning

maximize language and intercultural development for all students, researchers emphasize the need for program interventions that not only assist learners in socializing with target language speakers but also provide opportunities for guided reflection that draws from linguistic and cultural evidence.

15.2  Research questions This chapter uses a case study approach to examine research-based interventions for language and intercultural learning during study abroad and telecollaboration. The first case focuses on a student pursuing an ethnographic project for study abroad. The second case examines the work of a student and language partner participating in a classroom telecollaboration project. The research questions are: 1 What were the challenges faced in each project? 2 Were there benefits in terms of language and intercultural learning? 3 What pedagogical recommendations exist for future projects?

15.3 Methods 15.3.1 Participants The participant in the first case study was Juan, a third-year university student studying on a U.S. government scholarship for two semesters at a language center in the Arab Gulf. Juan was a highly motivated learner, with Languages as one of his majors. He had studied abroad previously in the Arab world on other government scholarships (in two intensive summer programs) and in South America. At the time of the study, Juan was 20 years old and had completed two three-credit hour courses at the intermediate level. His father was Hispanic and his mother was from an Asian country that sends many workers to the Arab Gulf. He was selected for the study due to his participation in the ethnographic project for language and intercultural learning. Because this project was arranged as a two-semester independent study for his home university, he was the only student at the language center participating in it. The participants in the second case study were Zane and Nermine. Zane was a second-year student (19 years old) in his fourth semester of Arabic (each of the four semesters equivalent to three credit hours), with Languages as one of his majors. He was of mixed Hispanic and Caucasian descent. He had not studied or traveled abroad before. Nermine, his native-speaker language partner, was a graduate student in Eastern languages at an Egyptian university and was in her early 20s. She had experience working with study abroad students as a language partner at her university. The telecollaboration project was conducted as part of the fourth-semester Arabic class, with seven students and four partners. Zane and Nermine were selected for this study because they were the only pair that met both of the following conditions: (1) they both consented to the analysis of their work, and (2) they were able to record all five of their Skype conversations.

15.3.2  Components of the ethnographic project Juan completed readings from Murchison (2010), a textbook geared toward undergraduates learning ethnographic methods, weekly assignments based on these readings (such as participant observations or interviews), and biweekly diary assignments. At the end of his time abroad, he submitted his analysis as a written paper. Further details of these assignments are 307

Emma Trentman

found in the Appendix. The data analyzed for this study consists of Juan’s weekly assignments, final paper, and diary entries, as well as email and chat correspondence between Juan and the teacher/researcher (me). I also interviewed Juan in late March about the ethnographic project and his study abroad experience generally.

15.3.3  Components of the telecollaboration project The telecollaboration project consisted of a ten-week exchange between my fourth-semester Arabic class and language partners at an Egyptian university, sponsored by an internal teaching grant. There were five topics, based on student interest: introduction and family, daily routine and food, religious life, politics in Egypt, and school and university experience. At the beginning and end of the project, there was a group Skype conversation. For each topic, the participants had a two-week period to write a blog in Arabic, read and comment on each other’s blogs, and chat on Skype. Students were given class time to read their partner’s blogs, work on their own, and prepare questions for their Skype conversations. They also completed a reflection portfolio where they analyzed their interactions. The project was worth 20% of students’ final grade for the class. Further details of the project are found in the appendix. The data analyzed for this study consists of Zane and Nermine’s blogs, their Skype recordings, and Zane’s reflection portfolio.

15.3.4 Analysis The data collected from the two projects were analyzed qualitatively using the MAXQDA program. The first step was thematic coding, where I coded the data for themes related to the research questions, such as difficulty understanding or evidence of language acquisition. I used a recursive process to recode previously analyzed data as the need for new codes became apparent. The second step was to use a conceptual map to further analyze the codes and their relationship to one another. The third step was to develop narrative descriptions of the case studies based on the codes and conceptual map. These narratives include key quotes from the coded data. Qualitative research emphasizes the need for triangulation between multiple perspectives and media. In this study, triangulation occurred through the variety of media and coursework components in both projects, which included my own and the participants’ perspectives. In terms of analyzing language gain, the emphasis in this study was on looking for examples that clearly related to the research-based interventions themselves. Pre- and post-measures of proficiency were not used because it would be impossible to attribute changes on these measures to the research-based interventions alone, since the participants were also engaged in Arabic classes and extracurricular activities (Juan) likely to impact their proficiency. Case study research emphasizes providing rich descriptions of particular cases, focusing on developmental factors and context (Flyvberg 2011). The goal is not to generate predictive theory or universal claims but rather context-dependent knowledge that has the potential for transferability to other contexts (Duff 2008, Flyvberg 2011). Duff (2008, p. 51) explains that: Transferability, which is sometimes also referred to as comparability, assigns the responsibility to readers to determine whether there is a congruence, fit, or connection between one study context, in all its richness, and their own context, rather than having the original researchers make that assumption for them. This emphasis on transferability and rich descriptions also means that it is not a problem to compare two cases that are very different in terms of the settings, participants, and projects, 308

Research-based interventions for language and intercultural learning

such as the ones in this study. Furthermore, the rich description of the cases necessarily includes the process as well as the outcomes of the two projects, as describing the challenges involved in classroom research projects is crucial to the development and implementation of future research (Schachter and Gass, 1996). Thus, this analysis provides rich descriptions of both the process and outcomes of these two cases that will allow readers to judge the potential transferability of these research-based interventions to their own diverse contexts.

15.4  Discussion of findings In this section, I discuss the challenges of implementing the projects, the outcomes in terms of language and intercultural learning, and the missed potential for further reflection on cultural rich points.

15.4.1  Challenges implementing the projects In both projects, the challenges related to time management, guidance, and Arabic language proficiency.

15.4.1.1  Time management Time management was especially challenging in the ethnographic project. In addition to studying Arabic full time at the language center, Juan was also taking courses online at his home university and spent part of the year volunteering, interning, or taking classes at a local university. He also traveled throughout the country and region, spent time with his local friends, applied to scholarships, and suffered illness. He explained in a chat that “The most challenging aspects of my experience is the fact that I am having trouble balancing everything in [country].” Indeed, it seems unlikely there were enough hours in the day for him to fully engage with all his responsibilities. As a result, he quickly fell behind on his work for the ethnographic project, despite email reminders of impending and missed deadlines. In the fall semester, he submitted the first few assignments on time but then fell behind starting in October. In the spring semester, a similar pattern ensued, where he began to fall behind in February. He submitted his final paper after the deadline for the spring semester, temporarily receiving an incomplete. Time management was less of a challenge in the telecollaboration project, potentially because it formed a large part of the regular Arabic class. However, Zane’s heavy course load, the tenhour time difference, and his family obligations made it challenging for him to find time to schedule and complete the Skype conversations with his partner. He arrived late to the initial group chat, and toward the end of the semester missed assignments or turned them in late. As an instructor, incorporating the telecollaboration project into the regular Arabic classroom was also challenging for me, as the class only met three hours per week, and this made it difficult to both address challenges in the telecollaboration and cover the regular fourth-semester material. For example, an entire class session could be devoted just to reading the partners’ blogs to understand the basic meaning, leaving little time for reflection and discussion of cultural meanings.

15.4.1.2 Guidance Providing adequate guidance to the participants as they completed their projects was also challenging. In the ethnographic project, guidance needed to be provided via email, chat, and Google Hangouts. Although I had previously engaged in ethnographic research as a researcher, 309

Emma Trentman

I had not advised a student on ethnographic research before, and Juan had not completed an ethnographic project before. The initial guidance I provided was through a textbook designed to introduce ethnography to undergraduates (Murchison, 2010) and weekly assignments based on chapters in the textbook. However, many of the assignments required multiple rounds of discussion, which delayed subsequent assignments. For example, it took several rounds of communication to establish a sufficiently narrow topic for the project. Although Juan expressed an interest in hospitality from the start, it was difficult for him to narrow this topic to more specific research questions. Finally, he settled on examining hospitality practices within his circle of local friends. Juan also struggled to find academic sources on his topic and write a literature review, and we worked through several revisions of this assignment. He scheduled his initial participant observations for 15–20 minutes, an amount of time in line with his busy schedule but insufficient for an ethnographic project. He also was unsure of how much detail and what to describe in his field notes. Finally, although I had developed several assignments to guide Juan through analyzing his data, he missed all of these assignments. As a result, his final paper, entitled “The Journey from a Guest to a Host” read more like a personal success narrative than an ethnographic study. In the telecollaboration project, the participants’ inexperience with blogs (perhaps because they were not an authentic medium for peer to peer communication) caused them to interpret these assignments differently than I had planned. The original goal of the blogs was to provide personal experiences that students could compare and discuss. However, while Zane generally wrote about his personal experiences on the blog, Nermine tended to write more generally about “the Egyptian people.” For example, on the daily routine topic, Zane opened his blog as follows: ‫ في االثنين واالربعاء والجمعة عندي‬٬‫ فصول‬۹ ‫ هذا الفصل الدراسي عندي‬٬ ]‫انا عندي أيام مجنونين [مجنونة‬ ‫ ثم اروح‬٬۱۲ ‫ والروسية في الساعة‬۱۱ ‫ في الصباح واللغة العربية في الساعة‬۱۰ ‫فصل الكيمياء في الساعة‬ ‫ بعض ذالك [ذلك] اروح بيتي وأحيانا اروح‬٬۳ ‫ عندي فصل الرياضيات في الساعة‬٬۱ ‫المكتبة في الساعة‬  . . . 1‫الرياضة‬ [“I have crazy days, this semester I have 9 classes, on Monday, Wednesday, and Friday I have chemistry at 10:00 in the morning and Arabic at 11:00 and Russian at 12:00, then I go to the library to study at 1:00, I have math class at 3:00, after that I go to my house and sometimes exercise . . .”] This contrasts with the beginning of Nermine’s post: ‫فى حياتنا اليومية بعيد عن المناسبات بتكون العيلة‬ . . . ‫اهال بيكم معايا فى جولة لشعب مصر االصيل والكريم‬ ]‫متجمعة مع بعضها خاصة يوم الجمعة وبيعتبر يوم مقدس بتتجمع فية [فيه] العيلة كلها على الفطار النة [النه‬  . . .  ‫يوم اجازة للمسلم والمسيحى والطلبة اجازة من الدراسة‬ [“Welcome on a tour of the authentic and generous Egyptian people . . . in our daily life, outside of celebrations, the family gathers together, especially Friday which is considered a holy day where the whole family gathers for breakfast because it’s a day off for the Muslim and Christian and a day off from school for students . . .”] In addition to the posts, participants also reinterpreted the task of commenting on the blogs. I had viewed the blog comments as a way of continuing the conversation, but with one exception that will be discussed later, Zane and Nermine either did not comment on each other’s 310

Research-based interventions for language and intercultural learning

blogs or left very short comments. In one case, Nermine simply listed the linguistic errors Zane had made. However, they did converse about their blogs in their Skype conversations.

15.4.1.3  Arabic proficiency The learners’ Arabic proficiency was the third major challenge in these projects. One benefit of ethnographic projects for study abroad is the need to use the local language in informal conversations and interviews and then write field notes or transcribe these conversations. Although Juan held informal conversations and conducted a formal interview in Arabic, he submitted his initial summaries and transcriptions in English. He wrote in his journal that he found it difficult to recall conversations in Arabic, even if this was the original language: I can recall parts of conversations in Arabic, but I still have the problem of recalling it all in Arabic. So I remember the topic or theme of the conversation, but I can only remember it in English. There are definitely Arabic undertones, but English is still the main language that controls my mind. Proficiency was an even greater challenge in the telecollaboration, as Zane was only in his fourth semester of Arabic, in a system where the classes only met three hours per week. As a result, Zane and Nermine’s Skype exchanges were marked by frequent requests for clarification, and this was the dominant theme in Zane’s reflection portfolio, where he reported that he struggled to remember and pronounce words such as “Christian” and “pray” that he had looked up beforehand, and also to continue the conversation when he ran out of pre-prepared questions. He felt that he was “not thinking in Arabic.” Nermine spoke English, and she would frequently translate words for Zane. However, both Nermine and Zane had difficulty understanding non-native speakers of their native languages, a problem that could be amplified by technical difficulties such as a weak Skype connection. For example, in several conversations Nermine did not understand Zane’s pronunciation of “Ramadan.” Similarly, in the following exchange Zane struggled to understand Nermine’s pronunciation of the word “culture”2: culture ‫ اه‬culture ‫ثقافة‬ ‫ يخلي اه مصر اه كويسة ثقافيا في الثقافة‬:N [“he makes, uh, Egypt, uh, good culturally, in culture, culture, culture, uh, culture”] uh, uh :Z culture ‫ يعني ثقافة‬culture culture :N [“culture, culture, culture means culture”] ‫ أنا مش‬calasher :Z [“calasher? I don’t”] culture :N ‫ أنا مش فاهم آسف‬:Z [“I don’t understand, sorry”] ‫ يعني ثقافة‬culture ‫ باإلنجليزي إنجليزي‬:N [“In English, English, culture means culture”] ‫ صح‬oh okay :Z [“oh, okay, right”]

311

Emma Trentman

Given the challenges of time management, guidance, and Arabic language proficiency in implementing the ethnographic and telecollaboration projects, a real question is whether these projects were successful in terms of language and intercultural learning. Yet it is crucial to remember that the goal of these projects was not the production of a professional ethnography or unbroken Skype conversation but increased language and intercultural learning, and there is evidence that the projects were successful in achieving this goal.

15.4.2  Benefits for language and intercultural learning 15.4.2.1  Ethnographic. project In the ethnographic project, the benefits for language and intercultural learning were the opportunities it provided for interaction, the way it served as a background frame for Juan’s hospitality experiences in his everyday life, and how it helped him reflect upon his identity negotiations abroad. Although Juan was a generally outgoing person, he still felt that pursuing the ethnographic project caused him to engage with people in Arabic that he otherwise might not have talked to. He explained in his interview: ‫ يقابلوني وأن يجلسوا‬،‫وفيه فرصة لآلخرين يقابلوا‬ . . . ‫بسبب المشروع كنت [كان] الزم أجلس في المقهى أكثر‬ ‫معي في الطاولة فعندما هم يقابلوني ويسألوني ممكن أجلس معك وقلت لهم نعم بالعربي وهم سألوا باإلنجليزي‬ ،‫ وبدأ يعني أشياء آخر‬،‫فهذا بدأ المناقشة يعني لماذا أنت موجود في (البلد) ولماذا أنت تتكلم اللغة العربية ف‬  . . . [ ‫أخرى عن العالقات بين العالم العربي والغرب ويعني كيف هم يفهموا الثقافة األمريكي [األمريكية‬، ‫أخرى‬ [“Because of the project, it was necessary for me to sit in the café more . . . and there’s an opportunity for others to meet, to meet me and to sit with me at the table, so when they meet me and ask me can I sit with you, and I said yes to them in Arabic and they asked in English, that started the conversation like why are you here in [the country] and why do you speak Arabic, so, and it started like other, other, other things on the relations between the Arab world and the West and like how they understand American culture”] In addition to specific interactional opportunities, the ethnographic project also formed a background framework that helped Juan pay attention to, ask questions about, and reflect upon hospitality events as they occurred in his everyday life. He explained in a chat that “this project is going hand in hand with my social life here.” He also felt that his initial difficulties in engaging with the project were due to a lack of local friends, as his close friendships did not solidify until toward the end of the first semester. Whether experiencing hospitality practices with his friends or in invitations from strangers, having the ethnographic project in mind led Juan to pay close attention to hospitality traditions as well as ask for specific details and reflect upon his experience. He noted in an email: ‫بصراحة أظن ان عندي افكار دقيق [دقيقة] و أوضح عن الضيافة (في البلد) ألني دائما أبحث عن فرص‬ ‫لتواصل [للتواصل] مع االشخاص و أالحظ االختالفات في الضيافة عندما أتصرف معهم‬ [“Truly, I think I have precise and clearer ideas about hospitality [in the country] because I am always looking for opportunities to interact with people and notice the differences in hospitality when I interact with them”] For example, during a trip to his language partner’s family home, he reported asking him about the hospitality traditions he experienced in this visit, including the specific types of food offered on different occasions and regional differences in the presentation of fruit, dates, and coffee. 312

Research-based interventions for language and intercultural learning

On another occasion, Juan sought clarification concerning the role of fruit: “Why are fruits one of the items that a host presents to a guest?” I asked casually as I aimed for a slice of an apple in the fruit bowl. [The person] who was seated to my right took the opportunity to explain it to me the sacredness of fruits.” (Final Paper) Juan’s careful observations of hospitality practices led him to develop his participation in these practices himself, including hospitality expectations among friends. Throughout the year, he began to develop a better sense of how to engage in hospitality practices with his friends through “simple purchases of tea and food” (Final Paper). He reported receiving compliments from his local friends on his understanding of hospitality practices, such as a friend who said: You know  .  .  . I do not understand how you do it, but it’s strange because I  feel that you have become Arab. The way you act around other people is different from when I met you. The way you talk and conceive ideas has also changed. Also, I do not understand how you comprehend the idea of Karam Al-Arab. To me you have become [local]. (Final Paper) Finally, engaging in the ethnographic project helped Juan analyze and understand some of the identity-related incidents he experienced related to his identities as a non-Arab Arabic speaker and an American of Asian and Hispanic heritage. The relationship between his ethnic and national identities, his Arabic-speaking abilities, and hospitality was of interest to Juan from the beginning, and his original research questions asked about the influence of racial and national background and Arabic proficiency on hospitality practices. Juan initially found the reception of his ethnic and national identities frustrating, as locals typically assumed he could not be American due to his skin color: When I meet people and I tell them that I am American, they look at me and tell me that I could not possibly be American because I do not fit the stereotypical image. At first this used to bother me a lot, because people would assume that just because of my physical features and background that I am less of an American than someone with white skin. (Chat) However, he came to realize that local perceptions of the U.S. were based on predominantly white film casts and media, and locals generally did not have access to alternative perceptions of the U.S. Thus, he could use curiosity about the apparent contradiction between his skin color and nationality, and the hospitality invitations that ensued, as opportunities to provide locals with alternate perspectives on the U.S., including discussions of racism and the unique culture of his home state. He explained in his interview:  . . . ‫عندما هم متفاجئين أو مستغربين عندما أنا أقول أنا من أمريكا عشان بشرتي سمراء فدائما هم يقولون‬ ‫ بس يعني هذه الحقيقة في‬،‫ أنت ال تقدر أن تكون أمريكي عشان أنت شكل [شكلك] مختلف ف‬،‫أنت الزم‬ ‫أمريكا نحن مش بلد مع األشخاص [أشخاص] بيضاء [بيض] نحن مكونين من األشخاص [أشخاص] مثلي‬ [ ‫هل تعرف القصة عن امريكا القصة حقيقي [الحقيقية‬ ‫من حوال [حول] العالم فالزم أقول لهم‬ 313

Emma Trentman

[“When they are surprised or astonished when I say I’m from America because my skin is brown, they always say you must, you can’t be American because your appearance is different, so, but like this is the reality in America, we’re not a country of white people, we are made of people like me from all over the world, so I have to say to them do you know the story about America, the real story”] Juan’s identity as a non-Arab, and especially American, Arabic speaker also remained salient throughout his time abroad. For example, he described meeting two men on a trip: I spoke to them in Arabic and in the middle of our conversation they asked me where I was from. . . . As soon as I told them [I was from the U.S.], they gasped, and I had to keep myself from laughing, because they could not believe that an American would ever speak Arabic. As usual the conversation led to why I was studying Arabic and so I explained. (Journal) As with the interest in his ethnic and national background, curiosity about Juan’s identity as an American Arabic speaker also opened up conversational opportunities, including hospitality invitations. While he observed that hospitality was also offered to non-Arabic speaking foreigners, being able to speak Arabic allowed him to understand the conversations around him, ask questions about hospitality practices, and engage in higher level conversations. Thus, the ethnographic project helped develop Juan’s language skills as he engaged in discussions on the United States and its relationship with the Arab world, hospitality traditions, and other topics that arose in the interactions he engaged in due to the project. In terms of his intercultural learning, reflecting upon the hospitality traditions he encountered for the project helped him develop a greater understanding of these practices and begin to participate in them himself. He was also able to reflect upon his racialized experiences abroad and move from a perspective of frustration to one that saw these incidents as opportunities for mutual learning.

15.4.2.2  Telecollaboration project In the telecollaboration project, there were also clear benefits in terms of language and intercultural learning. Zane learned various lexical items and about cultural practices such as family gatherings on Fridays. Zane and Nermine were able to easily discuss basic topics such as the weather and their families. As the project progressed, they also developed a system that allowed them to discuss more complicated topics. First, they would prepare questions for each other based on their blogs, which also served as a reference for Zane when he forgot a word or concept originally used in the blogs. They also began using the chat function in Skype to type words to each other. Zane could then use these words as a reference later in the conversation, and the transcript helped Nermine understand when Zane mispronounced a word. These strategies let them discuss topics that were not covered in the classroom and that were above Zane’s proficiency level. For example, in their discussion of Egyptian politics, Nermine explained that she did not like a particular political group because they “killed people.” Zane did not understand the word “kill,” which led to the following exchange (italics indicate English words): 314

Research-based interventions for language and intercultural learning

‫ بيقتل؟‬-‫ اي يعني اه بي‬:Z [“What does uh, ki-kill” mean?”] killed killed ‫ بيقتل بيقتل‬:N [“kill, kill, killed, killed”] oh okay :Z ‫ اه‬:N [Ah] ‫ ازي نكتب هذا [هذه] الكلمة؟‬:Z [“how do we write that word?”] ‫ اه اكتبها؟‬:N [“Uh, should I write it?”] ‫ أيوة من فضلك‬:Z [“Yes, please”] kill ‫يعني‬ . . . ‫تل‬-‫ يق‬-‫ ب‬:N (typing work in chat) [“k-ill, it means kill”] ‫ بيقتل‬:Z [“kill”] ‫ بيقتل‬:N [“kill”]

Nermine and Zane went on to discuss the words ‫“ الشرطة‬police,” ‫“ العنف‬violence,” and ‫أمن‬ “safety/security” using a similar strategy, and they remained written in the Skype chat. This allowed Zane to reuse these words or variants later in their conversation, albeit with difficulty, to describe protests against police violence happening in Albuquerque: police ‫ وفي الحقيقة اه والية [مدينة] ألباكركي ام فيها مشكلة اه اه مع ال‬:Z [“Actually, uh, in the state [city] of Albuquerque, um, there is a problem uh, uh with the police”] ‫؟‬police ‫ مع ال‬:N [“with the police?”] ‫( شرطة‬looks at chat transcript) ‫ اه‬:Z [“Uh (looks at chat transcript) police”] ‫ شرطة‬:N [“police”] (starts looking at chat transcript again) ‫ شرطة مع الشرطة ام هو اه هم ام دقيقة واحدة من فضلك‬:Z [“police, with the police, um, he, um, they, um, one minute please”]  . . . (short discussion of gender forms for please)

315

Emma Trentman

‫؟‬kill again ‫ الشرطة في اه في الباكركي ام ام هم ازي اه‬:Z [“The police in uh, in Albuquerque, um, um, they, how, uh, kill again?”] ‫ أنا كتبتها‬:N [“I wrote it”] ‫اه‬ ‫ بيقبلوا‬okay (looking at chat transcript) ‫ أنا مش عارف أنا ماعرفش الكلمة اه دقيقة واحدة‬I lost the word ‫ هم‬:Z [“They I lost the word I don’t know, I don’t know the word, uh, one minute (looking at chat transcript) okay they kill, uh”] ‫ بيقتلوا‬:N [“they kill”] often ‫ بيقتلوا اه الناس ام‬:Z [“they kill uh, the people, um, often”] ‫ الشرطة بيقتلوا الناس ليه؟‬:N [“why do the police kill people?”] ‫( أمن‬looks at chat transcript) ‫ ام‬safe again ‫ اه هم يقولوا أن الزم اه ام اه يقتلوا الناس ألن هو مش اه‬:Z [“ah, they say that they have to uh, um, uh, kill people because it’s not uh, safe again, um, safety”] ‫ مش آمنين‬:N [“they’re not safe”] ‫ أمن‬:Z [“safety”] ‫ اه مش آمن‬:N [“oh, it’s not safe”] ‫ مش آمن‬:Z [“it’s not safe”] ‫ ممم‬:N [“mmm”] ‫اه‬ ]‫ ولكن اه كثير اه المرة كثير مرة [مرات كثير‬:Z [“but uh, a lot, uh, time, time many”] ‫ مرات كثير‬:N [“many times”] ‫ مرات كثير هذا مش اه صحيح مش صحيح‬:Z [“many times that’s not uh, right, not right”] ‫ مش صح‬:N [“not right”] ‫ مش صح أيوة‬:Z [“not right, yes”] ‫ وام الناس ام في الباكركي مش مبسوطين اه مع الشرطة دالوقتي‬:Z [“and um the people um in Albuquerque are not happy um with the police now”]

Thus, by developing strategies to understand each other during the telecollaboration project, Nermine and Zane were able to discuss complicated and culturally meaningful topics. 316

Research-based interventions for language and intercultural learning

15.4.3  Cultural rich points with missed potential While there were clear instances of language and intercultural learning resulting from these projects, there were also unexplored cultural rich points in both projects, particularly in understanding how these rich points were embedded in larger cultural discourses.

15.4.3.1  Ethnographic project In the ethnographic project, these unexplored rich points included a perceived dichotomy between the West and the Arab world and multilingual versus monolingual practices during study abroad. Juan self-identified as “America’s diplomat to [country],” and many of his conversations with locals focused on explaining “the truth about America” within the framework of a misunderstanding between the West and the Arab world. He explained to one interlocutor that this misunderstanding: is one of the many reasons why I wanted to learn Arabic. I wanted to be a part of the change and part of the solution to improve relations between the Arab and Western World. (Participant Observation notes) Locals also drew upon this framework. For example, Juan reported how a local friend introduced Juan and his U.K. friend, “mentioning that we were from the West, but know how to speak Arabic” (Journal). Indeed, this Arab world–West dichotomy seemed to be a shared interpretive framework for both Juan and his interlocutors in describing their interactions. Yet the creation and maintenance of this cultural framework was never explored by Juan or locals he encountered, despite their engagement in multilingual and multicultural practices countering this dichotomy, such as speaking both Arabic and English, or adjusting their hospitality practices to accommodate multiple cultural traditions. Monolingual vs. multilingual practices during study abroad formed another unexplored rich point. As a dedicated Arabic learner, Juan was frustrated that other study abroad students seemed more interested in partying with each other than using Arabic with locals: “some of them did not take the study or the language as serious as I thought they should. But that is the mistake which people choose to make, inshallah [hopefully] they will learn later on in life what really matters, parties or studying and becoming a good Arabic Speaker” (Journal). As a result, he wrote that he hoped incoming students in the second semester would take “full advantage of their time here in [the country] by studying hard, BUT more importantly using their Arabic with actual [people from the country]” (Journal). These convictions reflect larger discourses of successful study abroad students who “immerse” themselves in target language networks abroad. Yet in many ways, Juan’s success in engaging in Arabic was actually dependent upon his multilingual identity rather than a monolingual Arabic practice. For example, the first time he met one of his local friends, their conversation focused on language learning, and Juan wrote in his journal that their complementary language skills were what made their friendship desirable. This friend then introduced Juan to his cousin, because like Juan, the cousin spoke Spanish, English, and Arabic. This cousin became Juan’s closest local friend, and they conversed in a mix of all three languages. Juan reflected that when he could use all of his languages, he felt his personality was complete: ‫كاملة ويعني أقدر أن أشرح له مشعورتي‬ . . . ‫عندما أنا مع (صديقه) أو نطلع مع بعض إنه شخصيتي‬ ‫إنجليزي‬ . . . ‫ فيه شبح‬،‫ مثال ال أعرف كيف‬،‫ مثال‬،‫[مشاعري] بهذه اللغة أو بهذه اللغة وعندي اختيار فيعني‬ 317

Emma Trentman

‫أسباني فعندما الثالثة موجودين [موجودة] هذا يتكون أنا لكن في شخص حقيقي فدائما‬ . . . ‫شبح عربي وشبح‬ ]‫عندما أتكلم واحد [واحدة] من اللغات أشعر إن [بأن] يعني فيه شخص ثاني بس مش الشخص كامل [الكامل‬ ‫ ناقص في حياتي فعندما هم موجودين [هي موجودة] مع بعض وعندي‬،‫ منقوص‬،‫ شو اسمه‬،‫فيه شيء يعني‬ ‫أعبر [عن] نفسي بهذه الثالثة لغات [اللغات الثالثة] أشعر إن [بأن] أنا شخص كامل‬ . . . ‫فرصة ل‬ [“.  . . when I’m with [his friend] or we go out together, my personality is . . . complete, and like I can explain my feelings to him in this language or this language and I have a choice, so like for example, for example, I don’t know how, there’s an English . . . ghost, an Arabic ghost, and a . . . Spanish ghost, so when the three are present this constitutes me, but in a real person, so always when I speak one of the languages, I feel that like there’s another person but not the complete person, there’s something like, what is it called . . . lost, missing in my life, so when they are present together, and I have the opportunity to . . . express myself in these three languages, I feel that I am a complete person”] Although Juan clearly articulates the importance of his multilingual identity and the opportunities this provided to him, the resulting contradiction with his beliefs in monolingual “immersion” practices remained unexplored.

15.4.3.2  Telecollaboration project In the telecollaboration project, two notable rich points with missed potential were religious and political divisions in Egypt. These are also areas in which there is disagreement among Egyptians, as in Agar’s (1991) definition of rich points. In her blog and in Skype conversations, Nermine, who was Muslim, placed considerable emphasis on the unity of Christians and Muslims in Egypt, showing a picture on her blog of an Egyptian imam and priest standing together and telling Zane on Skype that she had Christian friends and her father had gone to a church for a funeral. When she described Muslim religious traditions, such as Ramadan buffets, she emphasized that these were enjoyed by Muslims and Christians. Describing Ramadan on Skype, she explained: ‫ اه بنبقى كلنا فرحانين مع بعض وتحس إن‬،‫وبنبقى فرحانين كلنا سوا مسلمين أو مسيحيين أو أي ديانة أخرى‬ ‫ األمة كلها تقريبا في مصر‬،‫فيه ألفة بين ال‬ [“we are all happy, all of us together, Muslim or Christian, or any other religion, uh, we are all happy together, and you feel there’s a familiarity between the, almost the whole nation in Egypt”] While this frequent emphasis on religious unity in Nermine’s blogs and speech demonstrates her unhappiness with the religious violence that was simultaneously occurring in Egypt, Zane appeared to remain largely unaware of the existence of larger cultural discourses surrounding religious life in Egypt. Another rich point that was not fully explored in their discussions surrounded the exit of President Morsi from power in June 2013. In his blog on Egyptian politics, Zane described this event as a ‫“ انقالب‬coup.” This prompted Nermine to offer her only non-linguistic comment of the entire telecollaboration, as she wrote a new blog entry “correcting” Zane’s sentence to describe the event as ‫“ إرادة شعبية وثورة شعب‬the will of the people and a people’s revolution” rather than a ‫“ انقالب‬coup.” She then went over this point with Zane in their Skype conversation:

318

Research-based interventions for language and intercultural learning

‫ انقالب قلتها ليه‬:N [“coup, why did you say it?”] ‫االنقالب أو‬ ‫ ليه ليه كلمت عن‬:Z [“why, why did I talk about the coup or”] ‫ انقالب انت كتبتها في المدونة بتاعتك كلمة انقالب‬:N [“coup, you wrote it in your blog, the word coup”] (looks at his blog) ‫ كلمة انقالب اه فين؟‬:Z [“the word coup, uh, where?”] (typing) ‫ انقالب‬:N [“coup”] ‫ انقالب اه‬:Z [“coup, uh”] ‫ أنا كتبتها كتبتها‬:N [“I wrote it, I wrote it”] ‫[تعرفي] ايه عن االنقالب االنقالب‬ ‫ إنت عايزة اه تعرف‬Okay :Z ٍ [“okay, you want, uh, to know what about the coup, the coup”] ‫في المدونة بتاعتك ليه معناها ايه؟‬ ‫ إنت إنت كتبت كتبها‬:N [“you, you wrote, why did you write it on your blog, what is the meaning of it?”] ‫ ام أنا اه كتبت كتبت ام هذا كلمة [هذه الكلمة] ألن هنا في أمريكا ام اه يقولوا أن أن ام كان اه أن‬:Z [“um, I, uh, wrote, wrote um, this word because here in America um, uh, they say that, that um, there was um, that”] ‫ انقالب‬:N [“coup”] ‫في مصر‬ ‫ انقالب‬:Z [“coup in Egypt”] ‫ مم صح‬:N [“mmm, right”] ‫ أيوة‬:Z [“yes”] ‫ لكن الكلمة غلط مش صح‬:N [“but the word is wrong, it’s not right”] ‫ مش صح‬:Z [“not right”] ‫ مم اه الصح‬:N [“mm, uh, right”] ‫اه أحسن؟‬ ‫ أي الكلمة‬:Z [“what is the uh, better word?”] (types) ‫ ال كثير غلط‬:N [“no, very wrong”]

319

Emma Trentman ‫اقرأها‬ ‫ كده‬:N [“this, read it”] ‫ ثورة شعبية‬okay ‫ اه‬:Z [“uh, okay, a popular revolution”] ‫ثورة‬ ‫ ثورة‬:N [“revolution, revolution”] ‫ ثورة‬:Z [“revolution”] ‫ مم‬:N [“mmm”] ‫ شعبية‬:Z [“popular”] ‫ صح شعبية‬:N [“yes, popular”] ‫ أحسن من‬okay so :Z [“okay so better than”] ‫ انقالب‬:N [“coup”] ‫ االنقالب آسف صح‬:Z [“the coup, sorry, right”] ‫ صح‬:N [“right”] ‫ ولكن هم يعني نفس الحاجة أو ال؟‬:Z [“but do they mean the same thing or not?”] ‫ ال اه هم قصدهم انقالب يعني ام اه الجيش الجيش المصري هو ام مسك ام هو اللي بيحكم ام مصر‬:N [“no, uh, they, they mean, a coup means um, uh, the army, the Egyptian army, it, um, is in charge um, it rules um, Egypt”] mm-hmm :Z ‫ لكن دا مش صح‬:N [“but that’s not right”] okay :Z ‫ في رئيس عندنا رئيس هو اللي ماسك مصر هو الرئيس مش من الجيش‬:N [“we have a president, a president who is in charge of Egypt, he’s the president, not from the army”] okay :Z

In this discussion, Zane’s unfamiliarity with Egyptian discourses debating the nature of President Morsi’s exit from power caused him to struggle to understand whether Nermine’s correction of his use of the word “coup” was because it was linguistically incorrect or because it conflicted with her political stance. Thus, analyzing these two cases of research-based interventions reveals that while there were numerous challenges in implementing the projects, they did lead to language and 320

Research-based interventions for language and intercultural learning

intercultural learning, although there was potential for the further exploration of cultural rich points. In the next section, I discuss pedagogical recommendations for programs or instructors interested in implementing these interventions.

15.5  Conclusion and pedagogical implications This chapter has used a case study approach to analyze research-based interventions for language and intercultural learning in study abroad and telecollaboration. The first case describes the experience of an Arabic L2 learner completing an ethnographic project for study abroad while studying in an Arab Gulf country. The second case focuses on a telecollaboration project between a learner in a fourth-semester Arabic class and his Egyptian partner. Time management, guidance, and Arabic proficiency created challenges to implementing these projects. At the same time, there were clear benefits in terms of language and intercultural learning. The ethnographic project provided opportunities for Juan to interact with locals, served as a framework for reflecting upon and developing his everyday experiences of hospitality, and helped him analyze identity-related incidents he encountered abroad. The telecollaboration project assisted Zane with learning basic vocabulary and cultural information and discussing topics that were above his proficiency level and not covered in the classroom. The projects also demonstrate the missed potential for exploring certain cultural rich points. In the ethnographic project, these include the dichotomy of “the Arab world and the West” and monolingual ideologies of study abroad. In the telecollaboration project, these include religious and political divisions in Egypt. The pedagogical recommendations that follow from this study emphasize the need for programmatic support so that participants can invest sufficient time in the project and develop a shared metalanguage for reflecting upon cultural rich points. Although the results of this case study analysis are not generalizable to other contexts, they may be transferable. By providing a rich description of the two cases, I hope that readers will be able to interpret these findings for their own contexts and develop research-based interventions that promote language and intercultural learning during study abroad and telecollaboration. Future research on the process and outcomes of implementing these types of research-based interventions in diverse contexts will contribute further to our ability to develop researchbased interventions for promoting language and intercultural learning during study abroad and telecollaboration. The pedagogical recommendations of this study focus on recognizing the significant time investment required for these projects and the need to develop a shared metalanguage for discussing cultural rich points. Both projects necessitated a considerable time investment from the student and the instructor, and this investment must be factored into the design of the project and made clear to students and instructors from the outset. These projects need to be incorporated into regular classwork, and where there are multiple sections of a class, this will only be possible through strong programmatic support. Furthermore, there needs to be regularly designated class time to prepare for the telecollaborative or ethnographic activities pursued outside of class and also to reflect upon and analyze these activities after their completion. In addition to taking class time to fully explore cultural rich points, students and instructors also need a shared language to reflect upon these rich points, and specifically upon their connections to larger cultural discourses. This shared language could start with defining cultural rich points and using key concepts from Agar (1991), such as the “puttied in” associations with larger discourses and native speaker disagreement. For example, taking class time to provide multiple Egyptian perspectives on religious and political issues and explain how rich points are enmeshed in these larger cultural associations would have helped Zane interpret Nermine’s 321

Emma Trentman

statements on religious unity as well as her insistence on describing the events of June 2013 as a “popular revolution” rather than a “coup.” The shared language also needs to include defining and analyzing larger cultural discourses, including how they are performed in interaction and shape ideologies of language learning. For example, if Juan and his local friends reflected upon how their framing of their interactions as occurring between “the Arab world and the West” actually reinforces this dichotomy, despite their multicultural and multilingual practices, they might have started to question the origin and usefulness of this distinction. Similarly, reflection upon the discrepancies between Juan’s beliefs concerning monolingual practices of study abroad and the multilingual practices that in fact allowed him to achieve his goals of language learning and cultural integration might have led Juan to question these monolingual ideologies and focus more intentionally on the value of his multilingual practices. Both of these pedagogical interventions are likely to require teacher training, as teachers may not have experience developing materials for telecollaboration and ethnographic projects, nor with developing metalanguage to describe cultural rich points. Support at the programmatic level is crucial to providing teachers with the time and training necessary to develop these projects.

Appendix: Ethnography Course and Assignments Description: This course is designed to develop your linguistic and cultural knowledge during study abroad via an ethnographic research project. Ethnographic methods emphasize understanding a group’s linguistic and cultural practices from an insider’s point of view. You will experience the various stages of an ethnographic research project, from generating research questions, to participant observation and interviewing, to data analysis and presentation. You will also reflect on the impact of this project on your linguistic and cultural learning. Objectives: By the end of this course you should: • Be able to design and carry out a small ethnographic project • Understand ethics concerns involved in ethnographic research • Have a reflexive understanding of your linguistic and cultural learning abroad

Materials: 1) Murchison, J. (2010). Ethnography Essentials. New York: Wiley 2) Other readings as assigned by the instructor

Assessment: Field Journal Weekly Tasks Preliminary Analysis and Plan Final Project Google Hangout Meetings with Instructor

20% 50% 10% 10% 10%

322

Research-based interventions for language and intercultural learning

Field Journal: In your field journal, you will focus on three major questions: 1) What are you doing (description) 2) How is your project going (reflection) 3) How is your linguistic and cultural learning impacted (reflection) 4) Other reflections (optional) You will turn in your field journal every two weeks for feedback from the instructor. Weekly Tasks: Every other week, you will complete a task that will help you in developing your ethnographic project. These tasks will be due on Friday, but you will need the entire week to complete them, so you should start them early in the week. Preliminary Analysis and Plan: On March 28, you will submit your preliminary analysis and plan for continued research. You will also submit your dataset thus far, including field notes, interview transcripts, maps, charts, and visual data. Final Project: On May 16, you will submit your final project, including the literature review, methods section, and analysis, as well as your final dataset. Google Hangout Meetings with Instructor: Each week, you will have a short (approximately 15 minute) meeting with the instructor via Google Hangouts to discuss your progress. Topic/Reading

Assignment

Chapter 1 – What is Ethnography Chapter 2 – Choosing an Ethnographic Topic Chapter 3 – Research Design

Field Journal Research Topic Research Questions/CITI Field Journal Project Plan Secondary Sources Field Journal Literature Review Project Proposal Field Journal Participant Observation Plan Participant Observation – Field Notes Field Journal Interviews Interviews Field Journal Data Analysis 1 + plan Visual Data Field Journal

Chapter 4 – Writing a proposal Chapter 11 – Archives and Secondary Sources

Chapter 5 – A guide to collecting data and taking notes Chapter 6 – Participant-Observation Chapter 7 – Interviews Chapter 8 – Analyzing along the way Chapter 9 – Ethnographic Maps Chapter 10 – Tables and Charts Chapter 12 – Sorting and Coding Data Chapter 13 – Answering Questions and Building Models

Telecollaboration Project and Assignments

323

Preliminary Analysis and plan Field Journal Final Project

Emma Trentman

Telecollaboration project guidelines For this project, you will create a series of blog and chat activities. Blog: This will allow you to practice communicating through reading and writing with your language partner. You will keep your blog on Google Blogger to make it accessible to your language partners. Every two weeks, you will write a blog entry, comment on at least two of the Egyptians’ blogs, and reply to any comments they make on your blog. You will have some class time to work on your blog and reading your partners’ blogs. To set up your blog and learn how to ad audio and video, use the instructions here: https:// support.google.com/blogger/answer/1623800?hl=en&ref_topic=3339243 Chat: This will allow you to practice informal, synchronous communication with your language partner. You will choose either voice chat or text chat (or both). Every two weeks, you will chat with your partner for 15–30 minutes. You will turn in either the transcript of your text chat or a recording of your video chat. You will email your partner to set up a time that works for both of you: do not wait till the date the chat is due to try to set it up with your partner. You will have some class time to brainstorm questions to ask your partner. If you want to use text chat, it’s recommended that you use Skype: For help with Skype, use the instructions here: Mac: https://support.skype.com/en/faq/FA12015/getting-started-with-skype-mac PC: https://support.skype.com/en/faq/FA11098/getting-started-with-skype-windows-desktop To save text chats, select the chat log, copy it, and paste it into a word processing program. If you want to use video chat, it’s recommended that you use Zoom. Instructions and a video on how to use Zoom are posted on Learn under “Telecollaboration Project.” You can also use Skype, but you will have to download extra software. This information is also posted on Learn.

Schedule (likely to change, be flexible!) Topic 1: Introduction and Family Blog (Due 24 February, 8:00 am): Introduce yourself and your family for your language partner. Include pictures! Blog Comments (Due February 26, 8:00 am): Comment on your partner’s blog and the blog of at least one other Egyptian. Chat (Due by March 7): Chat with your partner for 15–30 minutes about yourselves and your families. Topic 2: Daily Routine and Food Blog (Due 10 March, 8:00 am): Write about your daily routine, including mealtimes, and the foods you eat and/or like to eat. Include pictures! Blog Comments (Due 12 March, 8:00 am): Comment on your partner’s blog and the blog of at least one other Egyptian. Chat (Due by March 24): Chat with your partner for 15–30 minutes about your daily routines and food. Topic 3: Religious Life Blog (Due 24 March, 8:00 am): Write about the role (or lack of role) of religion in your life, or in your family, or with your friends. Include pictures to help your explanation. Blog Comments (Due 26 March, 8:00 am): Comment on your partner’s blog and the blog of at least one other Egyptian. 324

Research-based interventions for language and intercultural learning

Chat (Due by April 4): Chat with your partner for 15–30 minutes about religious life. Topic 4: Politics in Egypt Blog (Due 7 April, 8:00 am): Write about your understanding of political events in Egypt, and what matters you would like more information on. Include pictures to help your explanation. Blog Comments (Due 9 March, 8:00 am): Comment on your partner’s blog and the blog of at least one other Egyptian. Chat (Due by April 18): Chat with your partner for 15–30 minutes about politics in Egypt. Topic 5: School and university experience Blog (Due 21 April, 8:00 am): Write about your experiences in school, from elementary school to university. Include pictures to help your explanation. Blog Comments (Due 23 April, 8:00 am): Comment on your partner’s blog and the blog of at least one other Egyptian. Chat (Due by May 2): Chat with your partner for 15–30 minutes about your school experiences.

Telecollaboration portfolio Reflecting on your work is an important part of telecollaboration projects. For each topic, you should provide the following information, with evidence (a quote from the blogs or the minutes in the chat recordings) and a brief explanation (in either English or Arabic): 1) Two things you learned 2) Two things you want to improve 3) Two things you improved from last time

Due dates: Topics 1 and 2: 31 March Topic 3: 14 April Topic 4: 28 April Topic 5: 13 May

Notes 1 Texts that were originally in Arabic have been provided in Arabic and English translation. The Arabic texts are provided verbatim; however, errors were not translated into English. Grammatical errors and typos are corrected in square brackets, except in instances where speakers corrected themselves. Spelling errors (e.g., not writing initial hamza) and dialect spellings remain as is. 2 In the English translation of the transcript, italics are used to designate words spoken in English.

References Agar, M., 1991. The biculture in bilingual. Language in Society, 20 (2), 167–182. Allen, H. W., 2010. Language-learning motivation during short-term study abroad: An activity theory perspective. Foreign Language Annals, 43 (1), 27–49. Bataller, R., 2010. Making a request for a service in Spanish: Pragmatic development in the study abroad setting. Foreign Language Annals, 43 (1), 160–175.

325

Emma Trentman Belz, J., 2003. Linguistic perspectives on the development of intercultural competence in telecollaboration. Language Learning and Technology [online], 7 (2), 68–99. Available from: http://llt.msu.edu/ vol7num2/belz/default.html [Accessed 30 August 2016] Belz, J. A., 2002. Social dimensions of telecollaborative foreign language study. Language Learning and Technology [online], 6 (1), 60–81. Available from: http://llt.msu.edu/vol6num1/belz/default.html [Accessed 30 August 2016] Belz, J. A. and Müller-Hartmann, A., 2003. Teachers as intercultural learners: Negotiating German – American telecollaboration along the institutional fault line. The Modern Language Journal, 87 (1), 71–89. Belz, J. A. and Vyatkina, N., 2008. The pedagogical mediation of a developmental learner corpus for classroom-based language instruction. Language Learning and Technology [online], 12 (3), 33–52. Available from: http://llt.msu.edu/vol12num3/belzvyatkina.pdf [Accessed 30 August 2016] Bown, J., et al., 2015. Student interactions during study abroad in Jordan. In: R. Mitchell, K. Mcmanus and N. Tracy-Ventura, eds. Social interaction, identity, and language learning during residence abroad. Amsterdam: EUROSLA, 199–222. Available from www.eurosla.org/monographs/EM04/ Bown_etal.pdf [Accessed 1 September 2016] Coleman, J. A., 2013. Researching whole people and whole lives. In: C. Kinginger, ed. Social and cultural aspects of language learning in study abroad. Amsterdam: John Benjamins, 17–46. Coleman, J. A., 2015. Social circles during residence abroad: what students do, and who with. In: R. Mitchell, et al., eds. Social interaction, identity, and language learning during residence abroad. Amsterdam: EUROSLA, 33–52. Available from www.eurosla.org/monographs/EM04/Coleman.pdf [Accessed 30 August 2016] Dewey, D. P., et al., 2013. Social network development, language use, and language acquisition during study abroad: Arabic language learners’ perspectives. Frontiers: The Interdisciplinary Journal of Study Abroad [online], 22, 84–110. Available from: https://frontiersjournal.org/wp-content/ uploads/2015/09/DEWEY-BELKNAP-HILLSTROM-FrontiersXXII-SocialNetworkDevelopmentLanguageUseandLanguageAcquisitionDuringStudyAbroad.pdf [Accessed 30 August 2016] Diao, W. and Trentman, E., 2016. Politicizing study abroad: learning Arabic in Egypt and Mandarin in China. L2 Journal [online], 8 (2), 31–50. Available from http://escholarship.org/uc/item/90g8r79m [Accessed 5 September 2016] Duff, P., 2008. Case study research in applied linguistics. New York: Routledge. Flyvberg, B., 2011. Case study. In: N. K. Denzin and Y. S. Lincoln, eds. The Sage handbook of qualitative research. Thousand Oaks, CA: Sage, 301–316. Godwin-Jones, R., 2013. Integrating intercultural competence into language learning through technology. Language Learning and Technology [online], 17 (2), 1–11. Available from: http://llt.msu.edu/ issues/june2013/emerging.pdf [Accessed 30 August 2016] Helm, F., et al., 2012. Promoting dialogue or hegemonic practice: power issues in telecollaboration. Language Learning and Technology [online], 16 (2), 103–127. Available from: http://llt.msu.edu/issues/ june2012/helmguthfarrah.pdf [Accessed 30 August 2016] Jackson, J., 2006. Ethnographic preparation for short-term study and residence in the target culture. International Journal of Intercultural Relations, 30 (1), 77–98. Jackson, J., 2008. Language, identity and study abroad: Sociocultural perspectives. Oakville, CT: Equinox Publishing. Jackson, J., 2010. Intercultural journeys: From study to residence abroad. New York, NY: Palgrave Macmillan. Kinginger, C., 2008. Language learning in study abroad: Case studies of Americans in France. The Modern Language Journal, 92, 1–124. Kinginger, C., 2009. Language learning and study abroad: A critical reading of research. New York: Palgrave Macmillan. Kinginger, C., 2013. Identity and language learning in study abroad. Foreign Language Annals, 46 (3), 339–358. Kinginger, C. and Belz, J. A., 2005. Socio-cultural perspectives on pragmatic development in foreign language learning: Microgenetic case studies from telecollaboration and residence abroad. Intercultural Pragmatics, 2 (4), 369–421. Kuntz, P. and Belnap, R. K., 2001. Beliefs about language learning held by teachers and their students at two Arabic programs abroad. Al-‘Arabiyya, 34, 91–113. Mendelson, V. G., 2004. Hindsight is 20/20: student perceptions of language learning and the study abroad experience. Frontiers [online], 10, 63–83. Available from: https://frontiersjournal.org/wp-content/uploads/2015/09/MENDELSON-FrontiersX-Hindsightis2020.pdf [Accessed 30 August 2016]

326

Research-based interventions for language and intercultural learning Mitchell, R., et al., 2015. Placement type and language learning during residence abroad. In: R. Mitchell, et al., eds. Social interaction, identity and language learning during residence abroad. Amsterdam: EUROSLA, 115–138. Available from www.eurosla.org/monographs/EM04/Mitchell_etal.pdf [Accessed 12 September 2016] Murchison, J., 2010. Ethnography essentials: Designing, conducting, and presenting your research. San Francisco: Jossey-Bass. O’Dowd, R., 2003. Understanding the “other side”: intercultural learning in a Spanish-English e-mail exchange. Language Learning and Technology [online], 7 (2), 118–144. Available from: http://llt. msu.edu/vol7num2/odowd/default.html [Accessed 30 August 2016] O’Dowd, R. 2010. Online foreign language interaction: Moving from the periphery to the core of foreign language education? Language Teaching, 44, 368–380. O’Dowd, R., 2011. Intercultural communicative competence through telecollaboration. In: J. Jackson, ed. The Routledge handbook of language and intercultural communication. London: Routledge, 342–358. O’Dowd, R. and Ritter, M., 2006. Understanding and working with ‘failed communication’ in telecollaborative exchanges. CALICO Journal, 23 (3), 623–642. Roberts, C., 2003. Ethnography and cultural practice: Ways of learning during residence abroad. In: G. Alred, et al., eds, Intercultural experience and education. Tonawanda, NY: Multilingual Matters, 114–130. Roberts, C. et al., 2001. Language learners as ethnographers: Introducing cultural processes into advanced language learning. Tonawanda, NY: Multilingual Matters. Schachter, J. and Gass, S. M., 1996. Introduction. In: J. Schachter and S. M. Gass, eds. Second language classroom research: Issues and opportunities. Mahwah, NJ: Lawrence Erlbaum, vii-xiv. Schenker, T., 2012. Intercultural competence and cultural learning through telecollaboration. CALICO Journal, 29 (3), 449–470. Shiri, S., 2015. The homestay in intensive language study abroad: Social networks, language socialization, and developing intercultural competence. Foreign Language Annals, 48 (1), 5–25. Thorne, S. L., 2003. Artifacts and cultures-of-use in intercultural communication. Language Learning and Technology [online], 7 (2), 38–67. Available from: http://llt.msu.edu/vol7num2/thorne/default. html [Accessed 30 August 2016] Trentman, E., 2012. Study abroad in Egypt: Identity, access, and Arabic language learning. Thesis (PhD). Michigan State University. Trentman, E., 2013a. Arabic and English during study abroad in Cairo, Egypt: Issues of access and use. Modern Language Journal, 97 (2), 457–473. Trentman, E., 2013b. Imagined communities and language learning during study abroad: Arabic learners in Egypt. Foreign Language Annals, 46 (4), 545–564. Trentman, E., 2015a. Arabic heritage learners abroad: Language use and identity negotiation. Al-‘Arabiyya, 48, 141–156. Trentman, E., 2015b. Negotiating gendered identities and access to social networks during study abroad in Egypt. In: R. Mitchell, et al., eds. Social interaction, identity and language learning during residence abroad. Amsterdam: EUROSLA, 263–280. Available from www.eurosla.org/monographs/ EM04/Mitchell_etal.pdf [Accessed 12 September 2016] Trentman, E., 2018. Study Abroad Arabic Programs: Issues of concern, research and future directions. In: K. Wahba, et al., eds. Handbook for Arabic language teaching professionals in the 21st century. Vol 2. New York: Routledge, 151–161. Trentman, E. and Diao,W., 2017. The American gaze east: Discourses and destinations of U.S. study abroad. Study abroad research in second language acquisition and international education, 2(2), 175–205) Vande Berg, M., et al., 2012a. Student learning abroad: Paradigms and assumptions. In: M. Vande Berg, et al., eds. Student learning abroad: What our students are learning, what they’re not, and what we can do about it. Sterling, VA: Stylus, 3–28. Vande Berg, M., et al., 2012b. Student learning abroad: What our students are learning, what they’re not, and what we can do about it. Sterling, VA: Stylus. Vande Berg, M. and Paige, R. M., 2012. Why students are and are not learning abroad. In: M. Vande Berg, et al., eds. Student learning abroad: What our students are learning, what they’re not, and what we can do about it. Sterling, VA: Stylus, 29–60 Ware, P., 2005. “Missed” communication in online communication: tensions in a German-American telecollaboration. Language Learning and Technology [online], 9 (2), 64–89. Available from: http:// llt.msu.edu/vol9num2/ware/default.html [Accessed 30 August 2016]

327

PART VII

Arabic heritage learners

16 PROFICIENCY IN STANDARD ARABIC AND ITS PREDICTORS The case of heritage speakers in college-level elementary Arabic classrooms Abdulkafi Albirini A growing body of research on heritage language acquisition has suggested that, for heritage speakers, the acquisition of the standard variety of their colloquial dialects is similar to learning a new language or L3 (e.g., Polinsky 2015, Rothman 2011, Rothman and Cabrelli Amaro 2010). This study examined the proficiency in Standard Arabic (SA) that heritage Arabic speakers bring to elementary SA classrooms and its relationship to a number of linguistic, socio-affective, socio-contextual, and demographic factors. The study also investigated the factors that could predict heritage speakers’ proficiency in SA based on their pre-college linguistic and social experiences. Twenty-nine heritage speakers from two American universities completed two production tasks in SA (one oral and one written) and an oral production task in CA. In addition, the participants completed a questionnaire about factors potentially influencing proficiency in SA, including linguistic input, language attitudes, ethnic identity, religious practice, family pressure, community support, Sunday school experience, and demographics. The participants’ oral and writing productions were analyzed in terms of fluency, grammatical accuracy, and syntactic complexity, which were the three measures used to determine proficiency. The findings indicate that heritage speakers had a functional command of SA that enabled them to complete the oral and writing tasks. Thus, SA may not be described as a new language (or L3) to them because they already are exposed to it – mostly incidentally – before they start learning it formally in college. Proficiency in SA was found to correlate positively with proficiency in CA, linguistic input, language attitudes, ethnic identity, religious practice, and self-reported proficiency in SA. However, the multiple regression analysis revealed that the only factors that could significantly predict proficiency in SA were linguistic input and proficiency in CA. A discussion then follows as to why it is theoretically and pedagogically important to approach heritage speakers not as L3/L2 learners and to build on what they already know to further their language abilities in SA.

16.1 Introduction Heritage speakers are an integral part of many Arabic language programs across the United States. These learners are typically exposed to their parents’ colloquial Arabic dialects (CA) early in their lives. However, the development of their colloquial Arabic dialects is typically 331

Abdulkafi Albirini

interrupted after they are exposed to English at or before elementary school. The majority of them either gradually stop using their heritage/first language (L1) or use it in very limited contexts, mainly in the home. Their reliance on English for everyday use eventually results in reduced proficiencies in their colloquial dialects (Albirini et al. 2011). Hence, many heritage speakers come to the classroom to learn, re-learn, or enhance their skills in the “Arabic language” (Albirini 2015c). Some come to develop their Standard Arabic (SA) skills in particular for religious or literacy purposes. Heritage speakers raise a number of important pedagogical questions to Arabic language programs. One of these questions is whether heritage speakers should be placed in the same classrooms as foreign/second language learners or should be assigned to separate classrooms. This question entails considerations of instruction, learning objectives, and teaching materials targeted particularly to heritage speakers. Another question is whether heritage speakers can skip elementary SA classes because of their heritage background or their previous knowledge in SA. It is true that some heritage speakers may have been exposed to SA through Sunday school or other informal venues (e.g., TV, YouTube, etc.), but is their knowledge of SA good enough to allow them to skip one or two years of SA language instruction? This question is critical because a number of researchers have claimed that the learning of a standard variety for heritage speakers who learned the colloquial variety at home amounts to learning a third language (L3) (Polinsky 2015, Rothman 2011, Rothman and Cabrelli Amaro 2010). The claim that learning a standard variety is similar to learning a third language for heritage speakers is based on theoretical work. According to Bardel and Falk (2007, also Falk and Bardel 2011), learners acquiring an L3 may not benefit from their first language because positive L1 transfer is obstructed by L2, which is the likely source of transfer whose learning is cognitively and sociolinguistically similar to L3 learning (e.g., with respect to typical age of acquisition, learning milieu, ultimate attainment, etc.). This claim has come to be known as the L2 status factor hypothesis (Bardel and Falk 2007, Falk and Bardel 2011). Based on this hypothesis, a number of researchers argued that the colloquial varieties to which some heritage speakers are exposed early in their lives represent the L1, whereas English is the L2. When heritage speakers begin to acquire a standard variety, they cannot transfer their knowledge of the colloquial varieties to the acquisition of the standard variety. This is because the L2, i.e., English, “blocks” this type of positive transfer from the L1. In other words, learning the standard variety becomes like learning a new language, that is, an L3. The assumption that SA is an L3 for heritage speaker is also based on the notion that the social context in which heritage speakers live limits their exposure to SA in significant ways, especially when they are compared to speakers who grow up in the Arab region. While native speakers of Arabic are regularly exposed to SA in schools and other formal domains and media, SA is hardly ever used in the societies in which heritage speakers live. For example, SA has no official status and is not used in social institutions, official communications, publications, or media. Moreover, whereas SA is an important identity marker for many native speakers, the same may not always be true for heritage speakers (see Albirini 2015c for a more detailed discussion). This raises the question of whether and how social and affective factors may impact heritage speakers’ knowledge of SA. Several studies have shown that heritage Arabic speakers’ language learning experiences are influenced by a number of linguistic, socio-contextual, and socio-affective factors,1 such as linguistic input, language attitudes, sense of identity, family pressure, community support, schooling opportunities, and demographic variables (Albirini 2014, Almubayei 2007, El Aissati 1996, Martin 2009, Rieschild and Tent 2008, Rouchdy 2013, Sehlaoui 2011, among

332

Proficiency in standard Arabic and its predictors

others). For example, Almubayei (2007) found that first- and second-generation Arab Americans’ sense of ethnic and religious identity is a key factor in their keenness to learn Arabic as a heritage language. Rieschild and Tent (2008) found that religion and gender are related to Arab Australians’ attitudes to heritage Arabic and their desire to learn it. Martin (2009) reports that Arabic is maintained among heritage speakers of Arabic in Milwaukee, Wisconsin, largely due to their parents’ pressure and positive attitudes toward it. Given the importance of social and affective factors, the question is then whether and how heritage Arabic speakers’ variant social and affective experiences may affect their pre-college knowledge of SA. While the existing studies have focused mostly on “Arabic” as an umbrella term or on the colloquial dialects, the current chapter focuses on SA. One challenge for addressing these pedagogical and theoretical questions is the lack of empirical studies concerning heritage speakers’ abilities in SA. The existing studies show that heritage speakers vary in their proficiencies in the colloquial varieties (Albirini 2014, Bos 1997, El Aissati 1996). However, little is known about their skills in SA, which is the language of instruction in many Arabic programs. In many cases, Arabic programs place heritage speakers in SA classes based on their self-reported proficiency in SA or in Arabic in general, which may simply reflect their proficiency in the variety to which they are exposed from their parents, namely CA.2 This study aims to shed light on the knowledge that heritage Arabic speakers bring to elementary SA classes. In addition, the study seeks to investigate the relationship between their knowledge of SA and a set of demographic, linguistic, socio-affective, and socio-contextual factors. Lastly, the study examines the factors that may predict heritage speakers’ SA proficiency, which may be critical when no samples of their language output are available. The potential predictors of SA proficiency can also point to possible ways to develop proficiency in SA by heritage speakers. Understanding elementary heritage speakers’ knowledge of SA is critical for three reasons. First, by understanding the proficiencies that heritage speakers bring into the SA classrooms, we can be in a better position to place them in the right courses and levels. Second, instead of providing instruction that may add little to what they already know, understanding their proficiencies helps in providing them with instruction that builds on their previous knowledge and experiences. This may maximize the benefits they derive from their SA learning experience. Lastly, by measuring their SA proficiencies, we can provide more objective statements to heritage speakers about their SA abilities so that they are aware of their strengths, weaknesses, and needs. This is critical, because many heritage speakers assume that they have good knowledge of SA based on their familiarity with SA or because of their linguistic background in CA.

16.2  Statement of research questions This chapter seeks to answer the following research questions: 1 What proficiency level in SA do heritage speakers bring to elementary SA classrooms? 2 Is there a relationship between their language proficiencies in SA and a their proficiencies in CA, b self-reported proficiencies in SA, and c their linguistic, socio-affective, socio-contextual, and demographic background? 3 What are the predictors of their SA proficiencies, if any?

333

Abdulkafi Albirini

16.3 Methods Language proficiency has been measured using different methods (see Unsworth 2008 for a review). In this study, proficiency in SA is defined by three constructs: fluency, grammatical accuracy, and syntactic complexity. Fluency is measured by word production per minute. Grammatical accuracy is determined by the proportion of error-free t-units to the total number of t-units in the speakers’ output. According to Hunt (1970, p. 4), a t-unit is “a main clause plus all subordinate clauses and non-clausal structures attached to or embedded in it.” Error-free t-units have no grammatical, morphological, and lexical errors of the types shown in examples (1)–(3). In (1), for example, the speaker fails to fulfil the subcategorization requirements of the verb ʔu-ħibb “1s-like” by introducing the verb ʔa-ðhab “1s-go” without ʔan “to.” In (2), nominal agreement between the noun ṣufuufii “my classes” and the adjective mumill “boring.s.m” is violated. Example (3) features a word selection error; instead of using muʕð̣am “most,” the speaker uses the word ʕiddat “several.” Lastly, linguistic complexity is gauged by the ratio of multi-clause/complex t-units to the aggregate t-units. (1) ʔaħyaanan ʔu-ħibb ʔa-ðhab* ʔila s-sinama maʕa ʔuxt-ii . . .  wa maʕa ʔaṣdiqaaʔ-ii. . .  aaa . . . wa-maʕa ʔahlii Sometimes 1s-like 1s-go* to the-cinema with sister-my . . . and with friends-my . . .  aaa . . . and-with family.my “Sometimes I like to go to the cinema with my sister . . . and with my friends . . . aaa . . .  and with my family.” (Syrian speaker) (2) ṣufuuf-ii mumill* židdan classes-my boring* very “My classes are very boring.” (Palestinian speaker) (3) wa-fi l-bayt . . . aaa . . .  ʕiddat ʔayyaam* . . . aaa . . . aaa. . . ʕiddat ʔayyaam . . . ba . . .  ʕiddat ʔayyaam ʔa-qraʔ qurʔaan And-in the-house . . . aaa . . . several days* . . . aaa . . . aaa . . .  several days. . . [. . .] . . .  several days 1s-read Qur’an “And in the house . . . aaa . . . several days . . . aaa . . . aaa . . . several days. . . [. . .] . . . several days I read Qur’an.” (Syrian speaker). These measures of fluency, grammatical accuracy, and syntactic complexity and their units of analysis (e.g., t-units) have been widely used for assessing proficiency in L1 and L2 acquisition research (Albirini 2014, Foster et al. 2000, Housen and Kuiken 2009, Hulstijn 2010, Iwashita 2006, Larsen-Freeman 2006, among many others). The choice of these measures is driven by empirical evidence indicating that development in L1 and L2 proficiency levels is typically marked by greater fluency, grammatical accuracy, and syntactic complexity (Albirini 2014). The segmentation of the oral data into t-units was based on the data transcriptions, prosody of the speech, and overall meaning of t-units. It was also informed by the judgment of two educated native speakers of Arabic; one is a speaker of the Egyptian dialect and the other speaks the Palestinian dialect. In (4), for example, the speaker produces a lengthy t-unit consisting of two clauses, including the main clause and the subordinate clause li-nadrus “so that we study.” By contrast, example (5) contains two independent t-units which are separated by 334

Proficiency in standard Arabic and its predictors

two hesitations and two pauses followed by the connector “after that,” and a completely new thought about the same topic. For the writing task, the segmentation of t-units was based on writing conventions (e.g., punctuation) and general meaning of t-units. (4) wa fii ʕuṭlat l . . . nihaayet l-ʔusbuuʕ . . . aaa. . . ʔa-ðhab maʕ ʔaṣdiqaaʔ-ii . . . aaa. . .  ʔila maṭaaʕim . . . waaa. . . ʔila s-sinama . . . waaa. . . ʔayḍan ʔila l-žaamiʕa . . .  li-nadrus. and in holiday the . . . end the-weekend . . . 1s-go with friends-my. . . to restaurants . . . aaand . . . to the-cinema . . . aaand . . . also to university so that-1p-study “And in the weekend . . . aaa . . . I go with my friends . . . aaa . . . to restaurants. . . and . . . to the cinema . . . and also to the university . . . so that we study.” (Syrian speaker) (5) kul yoom ʔa-ðhab maʕa ʔaxi ʔila l-žaamiʕa wa-ʔa-drus mina s-saaʕa. . . mmm. . . ʕašara ṣ-ṣubħ ʔila s-saaʕa θalaaθa ḍ-ḍeher . . . aaa . . . aaa . . . wa- baʕdahaa ʔa-ruuħ ʔu-ħaḍḍer l-ʕašaaʔ maʕa ʔax-ii every day 1s-go with brother-my to the-university and-1s-study from the-hour. . .  mmm . . . ten the-morning to the-hour three the-noon . . . aaa . . . aaa . . . and- afterwards 1s-go 1s-prepare the-dinner with brother-my “Every day, I go with my brother to the university and study from . . . ten o’clock in the morning to three o’clock in the afternoon . . . aaa . . . aaa . . . and after that . . . I go and prepare the dinner with my brother.” (Egyptian speaker) Previous studies indicate that several factors may impact heritage speakers’ language proficiency (Albirini 2014, Almubayei 2007, Martin 2009, Rouchdy 2013). In this study, we focus on the following factors that may be related to proficiency in SA: language input, language attitudes, ethnic identity, religious practice, family role, community support, Sunday school experience, and demographics (gender, education, country of birth, parents’ first language, parents’ education, and visits to the Arab region). These factors were examined in relation to the heritage speakers’ proficiency in SA. As noted earlier, the study also examined the relationship between the participant’s proficiency in SA and (1) their proficiency in CA and (2) their reported proficiency in SA.

16.3.1 Participants The study involved a convenience sample consisting of 29 heritage Arabic speakers, ranging in age from 18 to 27 years (average age = 21:3). A convenience sample was used in this study for two main reasons. First, the researcher had access to these participants through his previous work experience in two American universities and his extensive network of colleagues, acquaintances, and friends in the cities where these two universities are located. Second, by allowing all interested and eligible heritage speakers to participate, the researcher sought to have an adequate number of participants in this study to be able to draw meaningful conclusions on heritage speakers with dissimilar linguistic experiences. The heritage participants spoke different dialects, including the Egyptian dialect (six subjects), Palestinian dialect (11 subjects), Syrian dialect (seven subjects), Lebanese dialect (four subjects), and Jordanian dialect (one subject).3 At the time of the study, all of the participants were undergraduate or graduate students at two different universities in the United States.4 All of the heritage speakers identified Arabic as their first and weaker language, whereas they described English as the second and stronger language. 335

Abdulkafi Albirini

Sixteen of the heritage speakers were males and thirteen were females. Only two participants were graduate students, and the remaining ones were undergraduates. Five of the participants were born in the Arab region but moved with their families to the U.S. at the ages of 2 (two participants), 4 (two participants), and 5 (one participant), whereas the rest were born in the U.S. The parents of 27 participants were of Arab descent, and two had Arab fathers and non-Arab mothers whose first language is English. With the exception of three participants, all of the heritage speakers made visits to the Arab region, and these visits ranged in lengths from 1–3 months to 10 or more months. Lastly, in terms of self-reported proficiency, the majority of the participants reported having “poor” or “fair” abilities to speak/comprehend and write/read in SA. Table 16.1 summarizes the demographic background of the heritage participants.5 The heritage group was compared in terms of SA proficiency to a control group consisting of 20 speakers of Arabic (13 males and seven females). The control group is a subset of a larger sample of participants; it only included those who arrived in the USA within four years from the time of the data collection and spoke either the Egyptian or Levantine dialects.6 All of the subjects in the control group were pursuing graduate studies after they had finished their undergraduate degrees in an Arab country. These speakers were exposed to SA through schools and other social and media channels. The average age of the control participants was 29.3 (range, 25–33).

16.3.2 Tasks The subjects completed two elicited oral production tasks (one in SA and another in CA), an elicited writing task, and a questionnaire. For the elicited oral production task in SA, the subjects were asked to talk for five minutes about their daily activities and routines. The purpose of this task was to elicit long stretches of connected discourse based on which the speakers’ language fluency, accuracy, and complexity could be assessed. Narratives are ideal for Table 16.1 Summary of heritage speakers’ (categorical) demographic variables Variable

Category

Numbers

Gender Education Countries of birth Father’s/mother’s first language

Male/female Undergraduate/graduate United States/Arab country Arabic English High school or less College Graduate school 0 times/months 1–3 times/months 4–6 times/months 7–9 times/months 10 or more times/months Unable to speak or comprehend/read or write Poor Fair Good Excellent

16/13 27/2 24/5 29/27 0/2 0/1 8/13 21/15 3/3 17/8 8/12 1/5 0/1 0/0 22/18 6/9 1/2 0/0

Father’s/mother’s education Number/length of visits to Arab region

Reported SA proficiency: ability to speak or comprehend/read or write

336

Proficiency in standard Arabic and its predictors

measuring heritage language proficiency because narratives allow for the elicitation of output that emulates real-life situations where speakers use SA to perform functions in the language. More importantly, narratives on familiar topics allow them to relate to their own lives, which increases the authenticity of their language production. By contrast, formal tests (e.g., standardized or placement tests) require analytic skills heritage speakers may never need in real-life situations, and thus they do not reflect language use in naturalistic settings. For the elicited writing task in SA, the subjects were asked to spend 5 minutes writing about their families, relatives, and friends. The purpose of the writing task in SA was to assess their ability to use SA in writing, which is an important facet of SA use in real-life contexts (documents, official papers, essays, etc.). Again, the rationale behind using these written narratives was to obtain long stretches of discourse that would allow for examining their fluency (i.e., writing speed), grammatical accuracy, and syntactic complexity. Thus, proficiency in the writing task was determined by the same criteria used in the oral task. For an assessment of their proficiency in CA, the participants were asked to talk for 5 minutes about their hobbies and favorite foods in their colloquial varieties. No written task was used to measure proficiency in CA because CA is mainly used orally in everyday conversations and rarely in writing.7 Moreover, CA does not have a standardized script. As was the case with the oral narratives in SA, the purpose of the oral narrative in CA was to elicit a relatively lengthy connected speech based on which the speakers’ language fluency, accuracy, and complexity can be assessed. The subjects also completed a 110-item questionnaire (see Appendix 1). The questionnaire was based on (1) previous theoretical and empirical accounts concerning the role of different factors in language proficiency and (2) existing psychometric scales (Albirini 2014, Chinen and Tucker 2005). The questionnaire included seven scales: 1 2 3 4 5 6 7

Language input (28 items) Language attitudes (15 items) Ethnic identity (15 items) Family role (10 items) Community support (10 items) Sunday school experience (10 items) Religious practice (10 items).

In addition to these seven scales, the questionnaire contained two sections about the speakers’ demographic background and language experience.8 The subjects filled in the questionnaire before they completed the elicited oral and writing tasks. The subjects were audio-recorded as they completed the oral tasks in SA and CA, and their recordings were transcribed verbatim in Word documents. All the tasks were completed in a single session.

16.4  Main findings The findings are organized into three sections. The first section addresses the first research question, namely the participants’ proficiency levels in SA. To assess their performance in SA, the heritage speakers are compared to speakers who acquired SA in the Arab context. The second section focuses on the second research question, namely, the relationship between the dependent variable (proficiency in SA) and a number of independent variables, including proficiency in CA, linguistic (i.e., input), socio-affective factors (i.e., language attitudes, ethnic identity, religious practice), and socio-contextual factors (family pressure, community 337

Abdulkafi Albirini

support, and Sunday school), and demographic factors (gender, education, country of birth, parents’ first language, parents’ education, number and lengths of visits to the Arab region, and reported proficiency in SA). The third, and last, section presents the results of the multiple regression analysis concerning the proportion of variance in the participants’ SA proficiency explained by the selected independent variables and the potential predictors of their proficiency in SA.

16.4.1  Dependent variable: proficiency in SA As noted earlier, the assessment of SA proficiency was based on an elicited writing task and an oral production task. The data obtained from the writing task are summarized in Table 16.2. With respect to fluency/writing speed, the heritage speakers produced fewer words than the controls. The average word production time per minute was 10.4 for the heritage speakers, compared to 21.9 for the controls. In terms of grammatical accuracy, the percentage of errorfree t-units was lower for the heritage group (78.8%) than for the control group (98.2%). As for syntactic complexity, the heritage speakers produced fewer complex/multi-clause t-units than their control counterparts (10.9% compared to 15.9%). The overall proficiency scores (calculated by adding up the scores on the three measures) presented an advantage for the control group (average = 135.5) over the heritage group (average = 96.0). An independentsample t-test displayed a significant difference between the heritage group and the control group in terms of writing proficiency in SA; t(47) = −7.420, p < .0001. The controls also outperformed their heritage counterparts on the oral production task in SA, as Table 16.3 shows. In terms of fluency, the heritage speakers produced 27.6 words per minute in their oral narratives, whereas the controls produced 60.2 words per minute. As for grammatical accuracy, the ratio of grammatically accurate t-units by the heritage speakers (67.2%) was clearly lower than that of the controls (95.3%). Similarly, the rate of complex t-units in the heritage speakers’ narratives (12.3%) was lower than the rate in the controls’ narratives (15.7%). The heritage speakers’ overall oral proficiency was poorer than that of the controls (106.5 versus 170.6). An independent-sample t-test displayed a significant difference between the heritage group and the control group in terms of oral proficiency in SA; t(47) = −10.345, p < .0001.

Table 16.2 Comparison between the heritage speakers’ and controls’ writing proficiency in SA

No. of minutes No. of words No. of t-units No. of error-free t-units No. of complex t-units Fluency (words/minutes) Grammatical accuracy (error-free t-units/all t-units) Syntactic complexity (complex t-units/all t-units) Overall writing proficiency score

Heritage Speakers (n = 29)

Controls (n = 20)

145 1462 193 152 21 10.4 78.8

100 2186 226 222 36 21.9 98.2

10.9

15.9

96.0

135.5

338

Proficiency in standard Arabic and its predictors Table 16.3 Comparison between the heritage speakers’ and controls’ oral proficiency in SA

No. of minutes No. of words No. of t-units No. of error-free t-units No. of complex t-units Fluency (words/minutes) Grammatical accuracy (error-free t-units/all t-units) Syntactic complexity (complex t-units/all t-units) Overall oral proficiency score

Heritage Speakers (n = 29)

Controls (n = 20)

145 4009 448 301 55 27.6 67.2

100 6020 578 551 91 60.2 95.3

12.3

15.7

106.5

170.6

Table 16.4 Comparison between the heritage speakers’ and controls’ overall proficiency scores in SA

Heritage speakers Controls

No. of Participants

Overall Proficiency in SA

Standard Deviation

29 20

101.2 153.0

20.2 14.0

The participants’ proficiency in SA was calculated by averaging their proficiency scores on the writing and oral tasks in SA. As Table 16.4 shows, the heritage speakers overall proficiency in SA was lower than that of the controls (101.2 versus 153.0). An independent-sample t-test revealed a significant difference between the heritage group and the control group in terms of overall proficiency in SA; t(47) = -9.927, p < .0001. One of the important features of heritage speakers’ production in SA is their frequent codeswitching to CA and English. Code-switching refers to the concurrence of elements from two or more language varieties in a single sentence or piece of discourse.9 In (6), for example, the speaker uses the CA word našuuf “see” instead of the SA word naraa. The same applies to (7), where the speaker uses the word ʕarabiyya “car” for the SA word sayyaara. This example points to the existence of cross-dialectal influence as the Syrian heritage speaker resorts to a word used mainly in the Egyptian dialect. Similarly, in (8) and (9), the speakers also resorted to the English elements “homework” and “football match” instead of their SA counterparts, waažeb and mubaaraat kurat qadam, respectively. (6) fi l. . . . l-ʕuṭal . . . aaa. . . ʕuṭal . . . l-diraasiyyi . . . na-ðhab ʔilaa New York li-na-zoor waalid-ii wa waalidat-ii . ..w . . . aaa . . . mmm . . . waaa. . . . wa-li-na-šuuf ʔaṣdiqaaʔin-aa. In the . . . the-holidays . . . aaa . . . holidays . . . academic . . . 1p-go to New York to-visit father-my and mother-my . . . and . . . aaa . . . mmm . . . aaand . . . and-to-1p-see friends-our “In the holidays . . . aaa . . . holidays . . . the academic . . . we go to New York to visit my father and my mother and . . . aaa . . . mmm . . . aaand and to see our friends.” (Palestinian speaker) 339

Abdulkafi Albirini

(7) fii . . . min l-ʔayyaam . . . zaw. . . . ʔana wa zawžat-ii . . . aaa. . . . nu-saafir ʔilaa madiina qariiba fi l- . . . l-ʕarabiyya. In . . . from the-days . . . [. . .] . . . I and wife-my . . . aaa . . . 1p-travel to city close in the . . . the-car “In . . . from the days . . . [. . .] . . . I and my wife . . . aaa . . . we travel to a nearby city in the . . . the car.” (Syrian speaker) (8) baʕdein . . . mmm. . . ʔana ʔa-ðhab ʔila l-bayt w-beʕmel l-homework Afterwards . . . mmm . . . I 1s-go to the-house and-1s-do the-homework “Afterwards . . . . mmm. I go to the house and do the-homework.” (Lebanese speaker) (9) baʔdmaa ṣalli ṣalaat l-žimʕa . . . aaa. . . ʔana ʔa-ðhab ʔila l-madrasa w-beħḍar football match After 1s.pray prayer the-Friday . . . aaa . . . 1s-go to the-school and-1s.attend football match “After I pray the Friday Prayer . . . aaa . . . I go to school and attend a football match.” (Syrian speaker) The data show that a significant proportion of the speakers’ SA-based output consisted of CA and English elements (Table 16.5).10 Both the heritage speakers and the controls produced more CA and English elements in the oral task than in the writing task, which is understandable because speakers usually spend more time planning their writing than their speech. Thus, whereas almost one-tenth of the written output of heritage speakers (9.0%) was from CA and English, almost one-third of their oral output (32.6%) came from CA and English. The CA and English elements were less prevalent in the controls’ production, with ratios of 1.5% in the writing task and 6.7% in the oral task. Since most of the switched elements were individual words, this type of code-switching could have been motivated by compensation or retrieval issues rather than by social functions (see Albirini 2015c for a more detailed discussion). In other words, the participants shift to CA and English to fill in gaps in their knowledge of SA or to overcome delays or difficulties in retrieving words that they do not hear or use frequently in everyday life.

Table 16.5 Frequency and distribution of the switched words in the writing and oral tasks Controls

CA nouns CA verbs CA adjectives Other CA words English words Total no. of non-SA elements Percentage (out of total number of words)

Heritage Speakers

Non-SA Elements

Writing Task

Oral Task

Writing Task

Oral Task

65 33 7 15 12 132 9.0%

762 229 32 226 58 1307 32.6%

20 8 0 2 3 33 1.5%

184 82 12 117 9 404 6.7%

340

Proficiency in standard Arabic and its predictors

16.4.2  Independent variables and their relationship to the dependent variable In this section, I focus on the independent variables and their relationship to the dependent variable, i.e., proficiency in SA. The purpose here is to examine the relationship between heritage speakers’ proficiency in SA on one hand and on the other their proficiency in CA and a number of linguistic, socio-affective, socio-affective, and demographic variables. Each independent variable will be presented descriptively before being correlated to the dependent variable. Given the correlational nature of this study and its focus on heritage speakers, no comparison between heritage speakers and the controls will be used in explicating the independent variables.

16.4.2.1  Language Proficiency in CA As was the case with SA proficiency, the participants’ proficiency in CA was determined by the fluency, grammatical accuracy, and syntactic complexity they displayed in their oral narratives (Table 16.6). The data analysis showed that the heritage group displayed better fluency, grammatical accuracy, and syntactic complexity in their CA narratives than in their SA narratives (see Table 16.3). Thus, whereas their proficiency score on the oral SA narrative was 106.5, their proficiency score on the oral CA narrative was 156.5. This gap in their proficiency in SA and CA appears in all three dimensions of proficiency, namely fluency (27.6 versus 61.3), grammatical accuracy (67.2 versus 79.3), and syntactic complexity (12.3 versus 16.8). A Pearson product correlation test was conducted to examine the relationship between heritage speakers’ proficiency in SA and proficiency in CA (see the correlation matrix in Appendix 2). The correlation test revealed a significant positive relationship between proficiency in SA and proficiency in CA (r = .63, p < .01). In other words, students who were proficient in SA were also proficient in CA and vice versa. This may have to do with the notable overlap between the two varieties in terms of vocabulary, morphology, syntax, and phonology. It may also be due to the notable presence of CA elements in their SA narratives, as has been explained earlier.

16.4.2.2  The linguistic variable: input The role of input in heritage language development has been widely accepted (Albirini 2014, 2015a, 2016; Anderson 1999; Silva-Corvalán 2003). In this study, language input was measured by a scale consisting of 28 items (see Questionnaire in Appendix 1). The scale included the three main types of SA input that Arabic speakers may be exposed to, including Table 16.6 Heritage speakers’ proficiency in CA writing Measure

Criteria

Count

Scores

Fluency

No. of minutes No. of words No. of t-units No. of error-free t-units No. of t-units No. of complex t-units Fluency + accuracy + complexity scores

145 8883 904 717 904 152

61.3

Grammatical accuracy Syntactic complexity Overall proficiency

341

79.3 16.8 156.5

Abdulkafi Albirini

naturalistic, incidental, and instructed input.11 Naturalistic input refers to face-to-face interactions (e.g., with family), whereas instructed input refers to SA input in instructional settings (e.g., in classroom). Incidental input refers to situations where speakers are exposed to SA neither in natural face-to-face encounters nor in instructional settings, but more indirectly (e.g., as they watch TV or surf the internet). Thus, the scale focused on the amount of L1 input that the speakers received from (1) family and relatives, (2) friends and acquaintances, (3) print materials, such as books, (4) broadcast media, such as TV, (5) electronic media, such as the internet, (6) individualized instruction, and (7) structured instruction in schools or other programs. The speakers were asked to report on these input sources in four age phases: before the age of 5, during kindergarten and elementary school, during middle school, and during high school until now (i.e., time of experiment). Language input was quantified by the score of the 28 input-related items on a 6-point scale, ranging from never (0), few times a year (1), few times a month (2), few times a week (3), and daily (4), to exclusively (5). Thus, language input comprised the frequency, range, and contexts of the language received by the participants. Table 16.7 illustrates the distribution of the mean scores on the language input scale. As Table 16.7 shows, the heritage speakers received minimal amounts of SA input from the seven main sources identified earlier, with an overall mean score of 0.87 (maximum score = 5). They received the greatest amount of input from print materials (1.72), broadcast media (1.66), electronic media (1.29), and schooling (1.13), but received very limited SA input from their families (0.16), acquaintances (0.05), and tutoring (0.09). This is not surprising because SA is used mainly in formal settings and is rarely used in casual everyday conversations between family members or friends. A Pearson product test revealed a significant positive correlation between proficiency in SA and SA input (r = .73, p < .01); heritage speakers who received greater amounts of input in SA did better on the writing and oral tasks in SA than those who received less amounts of input.

16.4.2.3  Socio-affective and socio-contextual factors Three socio-affective factors (language attitudes, ethnic identity, and religious practice) and three socio-contextual factors (family pressure, community support, and Sunday school) were examined in relation to the dependent variable. Each of the socio-affective and socio-contextual factors was measured by a 5-point, Likert-type scale, ranging from strongly disagree (1), through disagree (2), neutral (3), and disagree (4), to strongly agree (5). The responses of participants to each scale were reduced to a mean score that demonstrated how positive/ Table 16.7 Distribution of mean scores on the language input scale Type of Input

Source of Input

Mean Score (out of 5)

Naturalistic input

Parents, siblings, and relatives Friends, peers, and acquaintances Print materials (e.g., books) Broadcast media (e.g., TV) Electronic media (e.g., internet) Individualized instruction (e.g., tutoring) Structured instruction (e.g., Sunday school) All input sources

0.16 0.05 1.72 1.66 1.29 0.09 1.13 0.87

Incidental/scripted Instructed Average

342

Proficiency in standard Arabic and its predictors

negative their reaction is to each scale/variable. The mean scores of all participants were then added up and averaged for each variable. Since the respondents rated each item on these scales from highly negative (i.e., “strongly disagree” or “1”) to highly positive (i.e., “strongly agree” or “5”), the range of possible mean scores was between 1 and 5, with higher scores indicating a more positive rating of the variable in question. Prior to conducting the analysis, the scoring of all negatively stated items was reversed. Language attitudes were represented by 15 attitude-related items focusing on the speakers’ perceptions of the beauty, usefulness, and importance of SA as well as their desire to learn it, use it, and pass it on to their children. As Table 16.8 shows, more than two thirds of the participants’ responses to the attitudinal items were either positive (39.1%) or highly positive (32.0%). The mean score of the participants’ attitudes toward SA was close to positive (3.94). Pearson product correlation tests pointed to a significant positive relationship between heritage speakers’ attitudes toward SA and their proficiency in this language variety (r = .39, p < .05); speakers with more positive attitudes toward SA had higher proficiency in SA than those with less positive attitudes toward SA. The ethnic identity scale consisted of 15 items that assessed the extent to which the participants (1) felt positively about their Arab ethnic background, (2) felt attached to member groups or organizations from their ethnic descent, and (3) engaged in exploring their ethnicity and its sociocultural expressions. More than half of the participants’ responses to the ethnic identity scale were either positive (33.8%) or highly positive (19.5%). However, a sizable proportion of the responses were neutral (27.6%) or negative (17.0%), and these responses appeared particularly on items pertaining to their affiliation to Arab groups or organizations. The mean score of participants’ responses to the ethnic identity scale were midway between neutral and positive (mean = 3.52). A significant positive correlation was found between the participants’ sense of ethnic identity and their proficiency in SA (r = .41, p < .05); heritage speakers who have greater sense of pride in or attachment to their Arab ethnic identity had better proficiency in SA than those with less pride in or attachment to their Arab identity. A third socio-affective factor is religious practice.12 The religious practice scale contained 10 items that assessed the degree to which the participants (1) felt the importance of SA for their religious life, (2) were exposed to it in religious contexts, and (3) used it in their religious practices. The participants’ responses were distributed mainly among the categories of positive (28.3%), neutral (23.4%), and negative (24.5%). Fewer responses were highly positive (7.9%) or highly negative (15.9%). The heritage speakers’ mean score on the religious practice scale was close to neutral (2.88), which reflects their limited exposure to or use of SA in their religious life. This may have to do with the fact English is typically used in religious sermons and Table 16.8 Percentage distribution and mean scores on the socio-affective and socio-contextual scales Scale

Language attitudes Ethnic identity Religious practice Family pressure Community support School experience

Percent of Responses (%)

Mean

Strongly Disagree

Disagree

Neutral

Agree

Strongly Agree

(on a 5-point scale)

1.1 2.1 15.9 5.9 9.0 15.9

6.9 17.0 24.5 18.6 22.1 24.8

20.9 27.6 23.4 22.1 29.7 23.4

39.1 33.8 28.3 30.3 30.3 27.6

32.0 19.5 7.9 23.1 9.0 8.3

3.94 3.52 2.88 3.46 3.08 2.88

343

Abdulkafi Albirini

preaching in the United States. It may also be due to the fact that most of the participating heritage speakers did not have strong commitments to religious practice. A significant positive relationship was found between the participants’ reported religious practice and their proficiency in SA (r = .39, p < .05); heritage speakers who reported more positive bonds to SA for their religious practice had better proficiency in SA than those who did not. As one of the socio-contextual factors examined in this study, family role was measured by 10 items focusing on the speakers’ perceptions of the direct or indirect role of their immediate and extended family in their appreciation, learning, and use of SA. More than half of the participants expressed positive (30.3%) or highly positive (23.1%) views of the role of their families in their appreciation, learning, or use of SA. The mean score of heritage speakers’ reported perceptions of the impact of their families on their learning of SA fell between positive and neutral (3.46). The correlation tests showed no significant relationship between the role of the participants’ families and their proficiency in SA (r = .27, p > .05); greater family pressure on heritage speakers to learn and use SA does not necessarily correlate with greater proficiency in SA. Community support was represented by 10 items that assessed the speakers’ perceptions of (1) the resources available in their communities for learning SA, (2) attitudes of the community members toward maintaining and learning SA, (3) social networks available in the community, (4) participants’ own involvement in the community, and (5) the advantage of acquiring SA in the community. About two-thirds of the participants’ responses were either positive (30.3%) or neutral (29.7%) toward the level of support their communities provided for learning SA. The respondents’ perceptions of their community support were close to neutral, with a mean score of 3.08. No significant relationship was found between the participants’ perceptions of community support and their proficiency in SA (r = .09, p > .05). Lastly, Sunday school experience was determined by a 10-item scale, which asked the participants about the degree to which they enjoyed and learned from their experiences in their Sunday schools.13 All of the participants in this study had some Sunday schooling at some point in their lives. About three fourths of the participants’ responses to the Sunday school scale fell under three categories: positive (27.6%), neutral (23.4%), and negative (24.8%). The mean score of the speakers’ responses to the Sunday school experience scale were close to neutral (2.88). There was no significant correlation between the respondents’ perceptions of their Sunday school experience and their proficiency in SA (r = .16, p > .05). Overall, only the three socio-affective variables were significantly correlated to proficiency in SA. None of the socio-contextual factors was significantly related to the dependent variable.

16.4.2.4  Demographic Variables Heritage speakers vary in various respects, including their demographic background. Some demographic variables may therefore be implicated in their dissimilar levels of proficiency in SA. In this study, six demographic characteristics were examined in relation to proficiency in SA, including gender, education, country of birth, parents’ first language, parents’ education, and visits to the Arab region. Self-reported proficiency in SA was also examined with the demographic variables (see Table 16.1). The seven demographic variables were quantified by individual scores on nine questionnaire items (see Appendix 1). The scores on the question pair pertaining to the parents’ first language(s) were combined into one score (two possible points), and so were the scores on the questions about the parents’ education (six possible points). The number and length of visits to the Arab region 344

Proficiency in standard Arabic and its predictors

were also combined into one score (ten possible points). The responses to all demographic items were treated separately as descriptive information that was correlated with SA proficiency. Spearman rank correlations were used to determine the relationship between the participants’ demographic (i.e., categorical) variables and their proficiency in SA. The correlation tests yielded no significant relationships between SA proficiency and any of the demographic variables (see the correlation matrix in Appendix 2). However, the speakers’ self-reported proficiency in SA was positively correlated with their SA proficiency scores on the writing and oral tasks (r = .39, p < .05). This means that heritage speakers have a relatively good assessment of their own knowledge of SA.

16.4.3  Predictors of proficiency in SA To determine the proportion of variance in the SA proficiency of heritage speakers that could be explained by the selected independent variables, a multiple regression analysis was performed. Following Gay and Airasian’s (2000) recommendations, only independent variables that individually correlated with the dependent variable were used in the multiple regression equation, to make a more accurate prediction of proficiency in SA (the dependent variable) and to show the proportion of variance in the dependent variable explained by the selected independent variables. The independent variables that correlated significantly with proficiency in SA were: SA input, proficiency in CA, language attitudes, ethnic identity, religious practice, and reported proficiency levels in SA. The multiple regression analysis indicated that 62.3% of the variance in heritage speakers’ proficiency in SA was explained by the independent factors included in this study (Table 16.9). The test statistic was highly significant; F(6, 28) = 8.717; p < .0001. Moreover, the results indicated that the only variables significantly affecting heritage speakers’ proficiency in SA were SA input and proficiency in CA, as is reflected in the absolute values of the standardized estimate (b) of SA input (b = .415, t = 2.678, p < .05) and proficiency in CA (b = .349, t = 2.357, p < .05). In other words, SA input and proficiency in CA were the only significant predictors of the variability in heritage speakers’ proficiency in SA (Table 16.10). The remaining factors were not significant predictors of the participants’ proficiency in SA. Table 16.9 Analysis of variance (all participants) Sources

Sum of Squares

DF

Mean Square

F Value

R Square

Adjusted R Square

p

Model Residual Total

8047.768 3385.272 11433.039

6 22 28

1341.295 153.876

8.717

.704

.623

.000

Table 16.10 Multiple regression on language proficiency (all participants) Variable

Unstandardized b

Standardized b

t

p

Proficiency in CA SA input Attitudes toward SA Ethnic identity Religious practice Reported proficiency in SA

.332 32.508 14.827 9.032 1.080 1.475

.349 .415 .214 .123 .017 .080

2.357 2.678 1.468 .932 .126 .505

.028 .014 .156 .361 .901 .619

345

Abdulkafi Albirini

16.5 Discussion This study examined the proficiency levels that heritage Arabic speakers bring to elementary SA classrooms. In addition, the study investigated the relationship between heritage speakers’ proficiency in SA (i.e., dependent variable) and a number of linguistic, socio-affective, sociocontextual, and demographic factors. The relationship between the dependent variable and both the heritage speakers’ proficiency in CA and their self-reported proficiency in SA was also explored. Further, the study analyzed the factors that could significantly predict heritage speakers’ proficiency in SA, which may be critical for identifying their knowledge of SA when we do not have samples of their SA output. The study findings indicate that heritage speakers enter elementary SA classrooms equipped with a functional knowledge in SA which enabled them to complete an oral and a writing task in SA. However, their knowledge of SA was not on a par with Arabic speakers who grew up in the Arab region, mainly due to differences in the context of SA acquisition, input opportunities, and available resources that define their language development. For example, heritage speakers have fewer opportunities to be exposed to SA inside and outside their homes. However, the fact that heritage speakers display notable familiarity with SA means that SA may not be described as an L3 for many of them.14 This is partly because of the overlap between their colloquial varieties and SA in terms of vocabulary, syntax, morphology, and phonetics, but also because they are incidentally exposed to SA (see later). While several studies have shown that heritage speakers have an advantage over L2 learners in a number of linguistic areas in their colloquial varieties (e.g., Albirini and Benmamoun 2014, Montrul 2010), this study shows that heritage speakers have an advantage over L2 learners even in SA. From a theoretical perspective, therefore, heritage speakers may not be described as L2 or L3 learners of SA, and may not as such be approached as L2 or L3 learners of SA. However, the fact that heritage speakers enter elementary SA classrooms with an advantage over L2 learners does not automatically put them in a better position to further their skills in SA based on their previous knowledge of this language variety. It is well known that most elementary SA classes focus on the sounds, alphabet, and basic functional uses of the standard variety. If one takes into account heritage learners’ familiarity with SA and the phonological, lexical, and structural similarities between SA and CA, one would expect little added benefit to some heritage speakers from participation in elementary SA courses unless extra opportunities are provided for them beyond what they already know. Thus, a first step to provide a rewarding SA learning experience is to identify and then build on their already acquired skills in SA. One of the main features of heritage speakers’ use of SA is their frequent code-switching to CA and English. This is interesting for two reasons. First, unlike native speakers, who use code-switching parsimoniously for specific functions (Albirini 2011), heritage speakers use it more profusely not only to perform certain discursive functions but also for retrieval purposes or to compensate for gaps in their SA knowledge. Previous studies suggest that speakers often resort to the stronger language to compensate for any gaps in their weaker language. In this study, about one-third of the heritage speakers’ oral SA-based narratives was in CA and English (Table 16.5), which possibly reflects the existence of many gaps in their knowledge of SA. Thus, it is true that heritage speakers have an advantage over L2 learners, but they have notable gaps in their knowledge of SA. Future studies need to examine these gaps with the goal of identifying the specific areas that may be challenging to heritage speakers. The gaps that heritage speakers have in their knowledge of SA may lead them to rely on CA (e.g., code-switching or transfer) to maintain the flow of their SA speech or simply to be able to get their ideas across. In fact, heritage speakers may not always be aware of whether a 346

Proficiency in standard Arabic and its predictors

certain word, form, or structure belongs to SA or CA, as previous studies have shown (Albirini 2015b). Heritage speakers’ knowledge of CA may, therefore, have a negative impact on their acquisition of SA, because some of the forms that they learned in CA are almost automatically retrieved in their SA speech (Albirini 2015b). Thus, although heritage speakers’ knowledge of CA may provide them with a notable advantage over L2 learners, considering the notable overlap between the two varieties, it may also provide them with inflated self-confidence in SA that may hinder their feeling of the need to further their learning of the standard variety. A second goal of this study was to examine the relationship between proficiency in SA and a number of linguistic, demographic, socio-affective, and socio-contextual factors. The findings indicate that SA input correlates positively with proficiency in SA. Input has been widely accepted as a main factor in heritage language development and acquisition (Albirini 2014, 2015a, 2016; Anderson 1999; Silva-Corvalán 2003). Input is critical in the case of heritage speakers in particular because they live in an environment where SA is rarely used inside or outside their homes. The data indicate that heritage speakers are exposed to SA, mostly incidentally, through print, electronic, and broadcast media (Table 16.7). They do not feel the urgency to learn this variety because it is not required to be a functional member of the societies in which they live, unlike speakers who grow up in the Arab region, who need to learn SA for a variety of reasons, including school. This finding confirms those of previous research concerning the critical role of input in heritage language acquisition and development. Socio-affective factors (attitudes, identity, and religion) were also significantly correlated to proficiency in SA. This is also not surprising if we consider the ideological value of SA for many Arabic-speaking people. SA is often seen as a symbol of common Arab heritage and history as well as a main mark of Arab and Muslim identity. For example, SA is one of the main assets shared by Arabs from different countries. People with strong Arab ethnic or national sentiments see SA as a main emblem of Arabism (Albirini 2015c, Suleiman 2003). Similarly, SA is the main language of Islamic liturgy, and therefore, many Muslims see it as a prerequisite for understanding Islamic texts and scholarship. It is not surprising, therefore, that people who are affectively attached to SA have better proficiency in SA, because they are keener to acquire it, maintain it, and pass it on to future generations. This sentiment applies to many Arabs residing inside the Arab region and heritage Arabic speakers living outside the Arab region. In addition to SA input and the socio-affective factors, proficiency in SA was positively related to reported proficiency in SA. This suggests that heritage speakers generally have a good sense of their language abilities in SA. Their self-reported assessment of their SA proficiency may not be as accurate as actual proficiency tests, mainly because most of them underestimate their knowledge of SA, for example, by reporting that they have “poor” skills in SA (see Table 16.1). However, self-reported proficiency may be useful in providing a general idea about their language skills in the absence of more reliable proficiency measures. However, among all the six variables that were significantly related to proficiency in SA, SA input and proficiency in CA were the only predictors of proficiency in SA. In other words, SA input and CA proficiency were the only factors that may predict variance in SA proficiency. Again, in the absence of production samples in SA, data collected on SA input may help provide an assessment of students’ proficiency in SA. Likewise, based on the findings of the present study, students with greater proficiency in CA may also have good knowledge of SA, possibly because of the many similarities between SA and CA. Given the positive correlation between proficiency in SA and proficiency in CA, it is important to investigate possible ways to utilize heritage speakers’ knowledge of their colloquial varieties to enhance their SA proficiency. For example, CA is mainly an oral variety and SA is mostly a written variety of Arabic; the question therefore is whether and how heritage speakers may be given opportunities to 347

Abdulkafi Albirini

employ their oral skills in CA to enhance their reading and writing (literacy) skills in the standard variety. No significant relationship was found between SA proficiency and the three socio-contextual factors examined in this study, including family, community, and Sunday school. Previous studies indicate that Arab parents report notable commitment to their children’s acquisition of Arabic (Almubayei 2007, Al-Sahafi 2015, Martin 2009). This, however, does not mean that there is dedicated effort by Arab families and communities to maintain SA among second- and third-generation Arab Americans. This finding underlies the need for more concerted efforts from Arab parents and communities to better serve Arab American children with interest in advancing their SA skills. For example, specialized programs focusing on SA may be needed to serve students with particular interest in SA. The Sunday school experience does not seem to have a major positive impact on SA learning for the majority of the participants. This may be due to the fact that Sunday school offers SA instruction only once a week. It may also have to do with the structure, methods, and materials used in Sunday schools, which are mostly based on volunteer instructors. More dedicated and appealing programs may be needed to help heritage speakers gain the knowledge that they seek to get from attending Sunday school. Before concluding this section, it is important to mention that only 62% of the variance in the participants’ proficiency in SA was explained by the factors included in this study. This means that several other cognitive, linguistic, and social factors may be involved in heritage speakers’ proficiency in SA. Future research needs to include these factors in studying heritage speakers’ proficiency in SA.

16.6  Conclusion and pedagogical implications The study shows that heritage speakers enter elementary SA classrooms with a level of proficiency in SA that is higher than would be expected from elementary L2 learners. In other words, heritage Arabic speakers have an advantage over their L2 counterparts in elementary SA classes. Thus, SA may not be described as a new language (or L3) to heritage speakers because they already know various aspects of this variety before they start learning it formally in college. It is therefore important for Arabic programs and instructors to consider ways to build on heritage speakers’ previous knowledge of SA to maximize the benefit they derive from their SA learning experience. Proficiency in SA was found to correlate with socio-affective factors, including attitudes, identity, and religious practice. In this respect, heritage speakers have another type of advantage over L2 learners, who usually enter elementary SA classrooms with an apprehension of SA because of its perceived difficulty and its structural distance from English (Elkhafaifi 2005, Nassif 2014). Affective factors are essential for language learning, and they can have a facilitative role in the advancement of heritage speakers’ learning of SA. For example, according to Krashen’s affective filter hypothesis, learners are more likely to absorb input when their affective filter is low (Krashen 1987). This applies to many heritage speakers, who may enter SA classrooms with both passion to learn SA because of its sociocultural significance and confidence in their Arabic background. SA input and proficiency in CA were the only predictors of proficiency in SA. This finding is important because it shows that one of the key ways to build up heritage speakers’ proficiency in SA is providing sufficient input from different sources (e.g., parents, TV, books, etc.) and in different contexts (e.g., home, school, mosque/church, etc.). It also poses the question of how to utilize heritage speakers’ knowledge of their CA varieties to enhance their SA 348

Proficiency in standard Arabic and its predictors

learning. Thus, instead of disregarding CA altogether, instructors need to find ways to tap into heritage speakers’ knowledge of CA to advance their SA skills. The specific strategies for achieving this goal require more in-depth research focusing particularly on this topic. The findings revealed no significant relationship between SA proficiency and socio-contextual factors. The lack of significant relationship between these variables points out the possible lack of adequate efforts by Arab families and communities to maintain SA among second- and third-generation Arab Americans. It also highlights the need to revisit some of the current approaches to teaching SA, including those used in Sunday schools which are not viewed favorably by heritage speakers. Evidently, this study is a first step toward understanding heritage Arabic learners’ overall linguistic knowledge of SA. Future studies need to focus on more detailed analyses of their SA skills in order to uncover some of the gaps they have in their knowledge. It is also important to compare heritage speakers to native speakers of Arabic and L2 learners of Arabic so that a more comprehensive picture of their language skills may be obtained.

349

Abdulkafi Albirini

Appendix 1: Bilingual Background Questionnaire I. Background Name: ____________________ Telephone number or e-mail: ____________________ Age: _________________________ Gender: ________________________________ 1. What is your current level of education? Undergraduate Graduate 2. Were you born in an Arab country? Yes No • If yes: How old were you when you came to the U.S.? ________ years old • If no: Are you: second-generation / third-generation / Other? 3. Where are your parents from? Mother: __________ Father: ______________ 4. What is native language of your parents? Mother: ______ Father: _____________ 5. What is your father’s highest level of education? High school or less College Grad school 6. What is your mother’s highest level of education? High school or less College Grad school II. Arabic learning experience 1. At what age did you start hearing Arabic regularly? __________________________________ 2. At what age did you start hearing English regularly? __________________________________ 3. In which language are you more proficient? Arabic English 4. Have you ever visited/stayed in an Arab country while living in the US? Yes No • •

If yes, how many times have you visited/stayed in an Arab country? 1–3 times 4–6 times 7–9 times 10 or more times How long have you stayed in an Arab country if you added up all your time there? 1–3 months 4–6 months 7–9 months 10 or more months

5. How would you rate your current reading and writing proficiency level in Standard Arabic? Excellent good fair poor unable to read or write 6. How would you rate your current speaking and listening proficiency level in Standard Arabic? Excellent good fair poor unable to speak or comprehend

350

During middle school, 15. My parents, siblings or other relatives talked to me in Standard Arabic. 16. My friends, colleagues or acquaintances talked to me in Standard Arabic.

During elementary school and Kindergarten, 8. My parents, siblings or other relatives talked to me in Standard Arabic. 9. My friends, colleagues or acquaintances talked to me in Standard Arabic. 10. I read (or someone read to me) stories, books, or other written materials in Standard Arabic. 11. I heard Standard Arabic through broadcast media, such as television and radio 12. I was exposed to Standard Arabic through electronic media such as online lectures, movies, music, videos, games, websites, CDs, DVDs, etc. 13. I received tutoring or individualized instruction in Standard Arabic. 14. I was taught Standard Arabic in school or in a program, such as the Sunday school.

Before the age of 5, 1. My parents, siblings or other relatives talked to me in Standard Arabic. 2. My friends, colleagues or acquaintances talked to me in Standard Arabic. 3. I read (or someone read to me) stories, books, or other written materials in Standard Arabic. 4. I heard Standard Arabic through broadcast media, such as television and radio 5. I was exposed to Standard Arabic through electronic media such as online lectures, movies, music, videos, games, websites, CDs, DVDs, etc. 6. I received tutoring or individualized instruction in Standard Arabic. 7. I was taught Standard Arabic in school or in a program, such as the Sunday school.

Please circle the number that best describes your exposure to Standard Arabic in different contexts:

II. Language Input

1 1

0 0

1 1

1

0

0 0

1

1 1

0 0

0

1

0

1 1 1

1

0

0 0 0

1 1 1

Few times a year

0 0 0

Never

2 2

2 2

2

2

2 2 2

2 2

2

2

2 2 2

Few times a month

3 3

3 3

3

3

3 3 3

3 3

3

3

3 3 3

Few times a week

4 4

4 4

4

4

4 4 4

4 4

4

4

4 4 4

Daily

5 5

5 5

5

5

5 5 5

5 5

5

5

5 5 5

Exclusively

During high school and till now, 22. My parents, siblings or other relatives talked to me in Standard Arabic. 23. My friends, colleagues or acquaintances talked to me in Standard Arabic. 24. I read (or someone read to me) stories, books, or other written materials in Standard Arabic. 25. I heard Standard Arabic through broadcast media, such as television and radio 26. I was exposed to Standard Arabic through electronic media such as online lectures, movies, music, videos, games, websites, CDs, DVDs, etc. 27. I received tutoring or individualized instruction in Standard Arabic. 28. I was taught Standard Arabic in school or in a special program, such as the Sunday school.

20. 21.

19.

18.

1 1 1 1 1 1 1

0 0 0 0

1 1

1

1

1

Few times a year

0 0 0

0 0

0

0

0

17.

I read (or someone read to me) stories, books, or other written materials in Standard Arabic. I heard Standard Arabic through broadcast media, such as television and radio I was exposed to Standard Arabic through electronic media such as online lectures, movies, music, videos, games, websites, CDs, DVDs, etc. I received tutoring or individualized instruction in Standard Arabic. I was taught Standard Arabic in school or in a program, such as the Sunday school.

Never

Please circle the number that best describes your exposure to Standard Arabic in different contexts:

2 2

2

2

2 2 2

2 2

2

2

2

Few times a month

3 3

3

3

3 3 3

3 3

3

3

3

Few times a week

4 4

4

4

4 4 4

4 4

4

4

4

Daily

5 5

5

5

5 5 5

5 5

5

5

5

Exclusively

2 2 2 2 2 2 2 2 2 2

1 1 1 1 1 1 1 1 1 1

6. 7. 8. 9. 10. 11. 12. 13. 14. 15.

2 2 2 2 2

1 1 1 1 1

I love Standard Arabic. I think that Standard Arabic is beautiful to speak. We need to maintain Standard Arabic in the Arab community. I don’t like to listen to people speaking in Standard Arabic. If I had been given the chance to learn one language in my life, I would have chosen Standard Arabic. I don’t think Standard Arabic is useful in my daily life. Standard Arabic makes me part of the community I live in. Standard Arabic is difficult for me to learn. Standard Arabic is part of who I am. Standard Arabic is important for my personal and/or professional life. I will ask my children to learn Standard Arabic in the future. I want to improve my abilities in Standard Arabic. I would avoid being exposed to Standard Arabic as much as possible. I would try to use Standard Arabic more in the future. I will dedicate more time for learning Standard Arabic in the future.

1. 2. 3. 4. 5.

Disagree

Strongly disagree

Please circle the number that best describes how you feel about Standard Arabic:

III. Language Attitudes:

3 3 3 3 3 3 3 3 3 3

3 3 3 3 3

Neutral

4 4 4 4 4 4 4 4 4 4

4 4 4 4 4

Agree

5 5 5 5 5 5 5 5 5 5

5 5 5 5 5

Strongly agree

13. 14. 15.

8. 9. 10. 11. 12. 1 1 1

1 1 1 1 1

1 1 1 1 1 1 1

1. 2. 3. 4. 5. 6. 7.

I consider myself basically an Arab. I consider myself basically an American. I consider myself an Arab American. I would like other people to regard me as an Arab. I would like other people to regard me as an American. I would like other people to regard me as an Arab American. I like to be a member of organizations or social groups that are mostly composed of Arabs. I think a lot about how my life will be affected by being an Arab. I do not like many Arab traditions and customs, such as dress, greetings, etc. I like to learn more about my Arab cultural and ethnic background. I have a lot of pride in the Arab ethnic group. I do not like to participate in Arab cultural practices, such as special food and music. I do not feel a strong attachment to the Arab World. I like to interact with Arab people more than other groups. I have always liked to visit the Arab World.

Strongly disagree

Please circle the number that best describes how you feel about your ethnic affiliation:

IV. Ethnic Identity

2 2 2

2 2 2 2 2

2 2 2 2 2 2 2

Disagree

3 3 3

3 3 3 3 3

3 3 3 3 3 3 3

Neutral

4 4 4

4 4 4 4 4

4 4 4 4 4 4 4

Agree

5 5 5

5 5 5 5 5

5 5 5 5 5 5 5

Strongly agree

10.

5. 6. 7. 8. 9.

4.

1

1 1 1 1 1

1

1 1 1

1. 2. 3.

My parents have always wanted me to learn Standard Arabic. My parents think that Standard Arabic is not important for my career. My parents think that Standard Arabic is necessary for my relations with other Arabic-speaking people. My parents provide resources and opportunities for me to learn Standard Arabic (e.g., Arabic school; Arabic stories). My family has been involved in teaching me Standard Arabic. My family does not pressure me to learn Standard Arabic. My family sometimes urges me to go overseas to learn Standard Arabic. I was encouraged by my family to go to a Sunday school. My parents were not happy that I talked to them, my siblings, and/or my relatives in English. My parents urged me to enroll in programs or classes that teach Standard Arabic

Strongly disagree

Please circle the number that best describes your family’s involvement in your Arabic-learning experience:

V. Family Role:

2

2 2 2 2 2

2

2 2 2

Disagree

3

3 3 3 3 3

3

3 3 3

Neutral

4

4 4 4 4 4

4

4 4 4

Agree

5

5 5 5 5 5

5

5 5 5

Strongly agree

6. 7. 8. 9. 10.

5.

1. 2. 3. 4.

There are opportunities and resources to learn Standard Arabic in the community There are spaces and organizations in our community that help Arabs get together. There are a number of accessible Sunday schools in the community. The Arab community has several events and activities that build a sense of belonging to the community. I think that most of the Arabs in our community care to preserve and promote Standard Arabic. I have a lot of connections in the Arab community. I participate in different social activities in the Arab community. I need Standard Arabic to feel a sense of belonging in the Arab community. I feel I need to use Standard Arabic in certain community events or practices. There is no value for learning Standard Arabic in our community.

Please circle the number that best describes how you feel about the Arab community in your area, your relationships with other Arabs, and your involvement in the Arab community:

VI. Community support:

1 1 1 1 1

1

1 1 1 1

Strongly disagree

2 2 2 2 2

2

2 2 2 2

Disagree

3 3 3 3 3

3

3 3 3 3

Neutral

4 4 4 4 4

4

4 4 4 4

Agree

5 5 5 5 5

5

5 5 5 5

Strongly agree

9. 10.

8.

5. 6. 7.

4.

1 1

1

1 1 1

1

1 1 1

1. 2. 3.

The Sunday school was important for me. Standard Arabic was taught regularly in the Sunday school. Standard Arabic was used as the language of instruction in the Sunday school. The time allotted to learning Standard Arabic was not enough for me to develop my skills in Standard Arabic. I did not enjoy the Sunday school. I often had the chance to use Standard Arabic in the Sunday school. Studying at the Sunday school helped improve my listening and speaking skills in Standard Arabic Studying at the Sunday school helped improve my reading and writing skills in Standard Arabic If I were a mother or a father, I would send my child to the Sunday school. Studying Standard Arabic at the Sunday school helped me understand religious texts better.

Strongly disagree

If you have attended Sunday school, please circle the number that best describes your Sunday school experiences:

VII. Sunday School:

2 2

2

2 2 2

2

2 2 2

Disagree

3 3

3

3 3 3

3

3 3 3

Neutral

4 4

4

4 4 4

4

4 4 4

Agree

5 5

5

5 5 5

5

5 5 5

Strongly agree

10.

9. 1

1

1 1 1 1 1 1 1 1

1. 2. 3. 4. 5. 6. 7. 8.

I go to the mosque/church regularly. I attend most of the services offered by my local mosque/church. I use Standard Arabic in my acts of worship (e.g., prayer). I read religious texts and books in Standard Arabic. When I talk about religion, I like to use Standard Arabic. I listen to religious speeches or sermons in Standard Arabic. I like to listen to religious recitations or chants in Standard Arabic. Knowing Standard Arabic gives me an advantage in my religious community. The speeches and sermons in my local mosque/church are offered in Standard Arabic. I learned a lot of Standard Arabic words to be able to understand religious texts.

Strongly disagree

Please circle the number that best describes your religious practices:

VIII. Religious practice

2

2

2 2 2 2 2 2 2 2

Disagree

3

3

3 3 3 3 3 3 3 3

Neutral

4

4

4 4 4 4 4 4 4 4

Agree

5

5

5 5 5 5 5 5 5 5

Strongly agree

Proficiency in standard Arabic and its predictors

Appendix 2: Summary of the Correlation Matrix! SA Prof. SA Prof. CA Prof. SA Input Attitude Identity Religion Reported Prof. Family Community School Gender Education Country Parents’ Lang. Parents’ Edu. Visits

1

.633** .728** .388* .414* .385* .392* .268 .089 .164 -.091 -.033 -.142 .228 -.156 .192

CA Prof.

SA Input

1 .528** .011 .227 .399* .324 .209 .043 .253 -.174 -.049 -.284 .325 .189 .087

1 .245 .344 .441* .285 .469* .340 .277 -.062 -.041 -.159 .212 -.070 .239

Attitude

1

.194 .087 .493** .213 -.143 -.130 -.171 -.139 -.186 .065 -.305 -.170

Identity

1

.013 .254 .119 .153 .147 -.071 .123 .182 .057 -.025 .157

Religion

1 .279 .278 .223 .201 -.104 .155 -.016 .008 -.167 .284

Reported Prof.

1

.171 -.138 -.039 -.253 .084 -.151 -.084 -.300 -.026

Correlation is significant at the 0.01 level (2-tailed). Correlation is significant at the 0.05 level (2-tailed). The title row of the correlation matrix contains the dependent variable (language proficiency) and the six independent variables that were found to relate significantly to language proficiency. The rest of the matrix shows the power and direction of the relationships amongst these variables and with the rest of the independent variables.

** * !

Notes 1 In this paper, socio-affective factors refer to feelings or beliefs held by a large portion of a given society or speech community, such as attitudes and identity sentiments, which may influence SA learning and usage. Socio-contextual factors are elements of the social context that may impact SA learning, such as family, school, mosque/church, and community resources. 2 Some Arabic programs, especially the big ones, use placement tests, but this is not available in many smaller programs. 3 The data presented in this paper are part of a larger project. Only the data from speakers of the Levantine (Jordanian, Lebanese, Palestinian, and Syrian) and Egyptian dialects were included in this paper for ease of analysis. The data for the project was collected between fall 2015 and fall 2016. 4 One of these universities used to offer separate elementary sections for heritage speakers but discontinued this tradition because most heritage speakers would register in courses designed for L2 learners, mostly due to scheduling conflicts. 5 Not all of these variables are demographic in nature (e.g., self-reported proficiency in SA), but we grouped them together for organizational purposes. 6 The controls were matched to the heritage speakers in terms of the varieties they spoke. 7 In some countries, such as Egypt, some literature has been written in CA, but we cannot assume that heritage speakers are exposed to this literature. 8 It is impossible to include all demographic and social factors in this study. Therefore, only the ones that have been reported to influence heritage language proficiency were included (see Albirini 2014 for details). 9 The focus here is on code-switching in the SA narratives because the main topic is proficiency in SA. The comparison between code-switching strategies in SA versus CA narratives will be published in a separate paper (in preparation). 10 Sub-word-level switches (e.g., case and mood markers) were not counted as switches in this study.

359

Abdulkafi Albirini 11 These three types of input are the ones to which Arabic speakers in the Arab region are exposed in their lives. 12 Religious practice is based on faith, which is more affective than contextual. 13 We focused on Sunday school because it is the only program that the vast majority of Arab heritage speakers attend at some point in their lives, as previous studies (e.g., Albirini 2014) and the questionnaire data in this study indicate. 14 Heritage speakers do not form a homogenous group. Some heritage speakers are more competent in SA than others, and some may have very limited knowledge of SA, which may give them little advantage over L2 learners.

References Albirini, A., 2011. The sociolinguistic functions of codeswitching between standard Arabic and dialectal Arabic. Language in Society, 40, 537–562. Albirini, A., 2014. Toward understanding the variability in the language proficiencies of Arabic heritage speakers. International Journal of Bilingualism, 18 (6), 730–765. Albirini, A., 2015a. The role of varied input in the divergent outcomes of heritage language acquisition. In: E. Grillo and K. Jepson, eds. The 39th annual Boston University conference on language development. Somerville, MA: Cascadilla press, 27–39. Albirini, A., 2015b. The role of the colloquial varieties in the acquisition of the standard variety: The case of Arabic heritage speakers. Foreign Language Annals, 47 (3), 447–363. Albirini, A., 2015c. Modern Arabic sociolinguistics: Diglossia, variation, codeswitching, attitudes, and identity. London: Routledge. Albirini, A., 2016. The role of exposure to English in the development of Arabic as a heritage language. Language Acquisition: A Journal of Developmental Linguistics. Online first doi: 10.1080/10489223.2016.1234620. Albirini, A. and Benmamoun, E., 2014. Concatenative and non-concatenative plural formation in L1, L2, and heritage speakers. Modern Language Journal, 98 (3), 854–871. Albirini, A., et al., 2011. Grammatical features of Egyptian and Palestinian Arabic heritage speakers’ oral production. Studies in Second Language Acquisition, 33, 273–303. Almubayei, D., 2007. Language and the shaping of the Arab-American identity. UTA Working Papers in Linguistics, 2, 91–119. Al-Sahafi, M., 2015. The role of Arab fathers in heritage language maintenance in New Zealand. International Journal of English Linguistics, 5 (1), 73–83. Anderson, R. 1999. Loss of gender agreement in L1 attrition: Preliminary results. Bilingual Research Journal, 23, 319–38. Bardel, C. and Falk, Y., 2007. The role of the second language in third language acquisition: The case of Germanic syntax. Second Language Research, 23 (4), 459–484. Bos, P., 1997. Development of bilingualism: A study of school-age Moroccan children in the Netherlands. Tilburg: Tilburg University Press. Chinen, K and Tucker, G. R., 2005. Heritage language development: Understanding the roles of ethnic identity and Saturday school participation. Heritage Language Journal, 3 (1), 27–59. El Aissati, A., 1996. Language loss among native speakers of Moroccan Arabic in the Netherlands. Thesis (PhD). University of Nijmegen. Elkhafaifi, H., 2005. Listening comprehension and anxiety in the Arabic language classroom. The Modern Language Journal, 89, 206–220. Falk, Y. and Bardel, C., 2011. Object pronouns in German L3 syntax: Evidence for the L2 status factor. Second Language Research, 27, 59–82. Foster, P., Tonkyn, A. and Wigglesworth, G., 2000. Measuring spoken language: A unit for all reasons. Applied Linguistics, 21, 354–375. Gay, L.R. and Airasian, P., 2000. Educational research: Competencies for analysis and application, 6th ed. Upper Saddle River, NJ: Prentice-Hall. Housen, A. and Kuiken, F., 2009. Complexity, accuracy, and fluency in second language acquisition. Applied Linguistics, 30 (4), 461–473. Hulstijn, J. H., 2010. Measuring second language proficiency. In: E. Blom and S. Unsworth, eds. Experimental methods in language acquisition research. Amsterdam: Benjamins, 185–199.

360

Proficiency in standard Arabic and its predictors Hunt, K. 1970. Syntactic maturity in school children and adults. Monographs of the Society for Research in Child Development, 35, 1–67. Iwashita, N., 2006. Syntactic complexity measures and their relation to oral proficiency in Japanese as a foreign language. Language Assessment Quarterly, 3 (2), 151–169. Krashen, S., 1987. Principles and practice in second language acquisition. Englewood Cliffs, NJ: Prentice Hall. Larsen-Freeman, D., 2006. The emergence of complexity, fluency, and accuracy in the oral and written production of five Chinese learners of English. Applied Linguistics, 27, 590–619. Martin, N., 2009. Arab American parents’ attitudes toward their children’s heritage language maintenance and language practices. Thesis (PhD). University of North Carolina at Chapel Hill. Montrul, S., 2010. How similar are L2 learners and heritage speakers? Spanish clitics and word order. Applied Psycholinguistics, 31, 167–207. Nassif, L., 2014. Anxiety in the noticing and production of L2 forms: A study of beginning learners of Arabic. Thesis (PhD). University of Texas at Austin. Polinsky, M., 2015. When L1 becomes an L3: Do heritage speakers make better L3 learners. Bilingualism: Language and Cognition, 18 (2), 163–178. Rieschild, V. and Tent, J., 2008. Bilinguality, gender, and religion as influences on Arabic-Heritage Australian youths’ attitudes to their ambient languages and cultures. 17th sociolinguistics symposium, 3–4 April, Amsterdam. Rothman, J., 2011. L3 syntactic transfer selectivity and typological determinacy: The typological primacy model. Second Language Research, 27, 107–127. Rothman, J. and Cabrelli Amaro, J., 2010. What variables condition syntactic transfer? a look at the L3 initial state. Second Language Research, 26, 189–218. Rouchdy, A., 2013. Language conflict and identity: Arabic in the American diaspora. In: A. Rouchdy, ed. Language contact and language conflict in Arabic: Variations on a sociolinguistic theme. London: Routledge, 133–148. Sehlaoui, A., 2011. Language learning, heritage, and literacy in the USA: The case of Arabic. Language, Culture, and Curriculum, 21 (3), 280–291. Silva-Corvalán, C., 2003. Linguistic consequences of reduced input in bilingual first language acquisition. In: S. Montrul and F. Ordóñez, eds. Linguistic theory and language development in Hispanic languages. Somerville, MA: Cascadilla Press, 375–397. Suleiman, Y., 2003. The Arabic language and national identity. Edinburgh: Edinburgh University Press. Unsworth. S., 2008. Comparing child L2 development with adult L2 development: how to measure L2 proficiency. In: E.Gavruseva and B. Haznedar, eds. Current trends in child second language acquisition. Amsterdam: John Benjamins, 301–336.

361

17 EFFECT OF AGE OF ACQUISITION ON CONCEPT MEDIATION IN HERITAGE ARABIC BILINGUALS Iyad Ghanim Currently accepted models of bilingual lexical retrieval posit a single conceptual domain and two language-dependent domains for all bilinguals. Of particular interest is how the second language domain interacts with the first language domain and the conceptual domain; the L2 domain is either directly attached to the semantic domain or instead mediated by the L1. Research suggests that for highly proficient bilinguals, newly acquired words attach to their conceptual connotations directly; conversely, for beginning L2 learners, the words of an L2 lack a direct connection to this semantic information and, instead, are only accessed via L1 translation equivalents. This means that for beginning L2 learners, the words of the weaker L2 can be accessed only by translating from the L1. As the learner becomes more fluent in an L2, words in that L2 begin to be directly associated with the semantic information and are retrieved without mediating through the L1. This lexico-semantic model has been used by researchers to describe most types of bilinguals. Problems arise, however, when applying this model to heritage speakers, a population of bilinguals who learned one language in childhood but have lost fluency by adulthood. The model suggests that heritage speakers would not be able to retrieve semantic connotations from their heritage language without translating it from their second, now-dominant language. A second, though conflicting, option is that the L1 remains conceptually mediated, but those links are weakened and effectively non-functional. This would suggest that heritage speakers’ two language domains have effectively independent access to the conceptual space. I hypothesize that simply having proficiency in a language cannot be the only reliable guarantor of word-to-concept connections; instead, I propose that the age at which a language was acquired must guarantee a semantic connection even more reliably than proficiency can. The present chapter disentangles the respective effects of age of acquisition and language proficiency on developing concept mediation. In this study, 11 heritage Arabic-English bilinguals with varying degrees of Arabic proficiency were administered a picture-naming task and a translation task. The findings reveal heritage speakers’ response times generally match the predictions of the concept mediation model irrespective of proficiency, with the exception of extremely low proficiency speakers. Therefore, for all but extreme cases of language loss, having learned a language in early childhood even without having reached full fluency is enough to assure semantic connections in adulthood.

362

Effect of age of acquisition on concept mediation

17.1 Introduction Early research on the structure of bilingual lexicons begins with Weinreich (1953), who describes three “bilingualism types.” The first type of bilingualism, compound bilingualism, refers to a system where the two lexical domains are separate, but each are connected to a shared domain that houses conceptual and semantic information. Coordinate bilingualism, on the other hand, refers to a situation where each lexical domain is connected to separate language-dependent conceptual domains. Subordinative bilingualism is the case when the lexical items of the second language are only connected to their translation equivalents in the first language; in other words, the second language is only indirectly connected to the conceptual domain. By using these models, Weinreich conflates two questions: first, do two language-specific conceptual systems interact with each language-specific lexicon, or conversely, is there only one shared conceptual system? And secondly: what is the extent of the interaction or connection between a given lexicon and conceptual domain, if any interaction exists at all? Disentangling some of the implications of Weinreich’s bilingual types, some studies suggest evidence for independence between two language representations (Brown et al. 1984, Gerard and Scarborough 1989, Kirsner et al. 1984, Kolers 1963, Scarborough et al. 1984; from Kroll and Stewart 1994). Other studies contrast this with evidence for a shared conceptual memory underlying the bilinguals’ two languages (Altarriba 1990, Chen and Ng 1989, Glanzer and Duarte 1971, Schwanenflugel and Rey 1986). Works by Potter (1984) and Snodgrass (1984), among others to be discussed, propose a single, abstract memory system where concepts are stored and accessed by separate lexical memory systems (Kroll and Stewart 1994). This model of a shared conceptual domain is the currently accepted understanding of the bilingual lexicon. Yet additional research focuses on the organization of the bilingual lexicon; particularly, how the lexicons of a bilingual’s two languages interact with each other, if they interact at all. Potter et al. (1984) develop two models of bilingualism that differ on how the language domains connect to the conceptual domain or to each other: the word association model and the concept mediation model (Figure 17.1, adapted from Kroll and Curley 1988). The word association model (WAM), similar to the Weinreich’s subordinative model, posits that the second language (L2) is only connected to the first language (L1) through translation equivalents (also referred to as “lexical links,” or “word-to-word associations”) (Kroll and Stewart 1994, Jiang and Forster 2001). Critically, the L2 domain is not directly connected to the conceptual domain under this model. Instead, L2 words gain access to concepts only by mediation through the L1. The second conceptualization of cross-linguistic connection is the concept mediation model, which can be compared to Weinreich’s compound bilingualism. Unlike the word

L1

L1

L2

CONCEPTUAL

L2

CONCEPTUAL

Figure 17.1 Word association model (left) and concept mediation model (right)

363

Iyad Ghanim

association model, the concept mediation model (CMM) assumes second language words have direct access to concepts: a bilingual’s two languages are connected through shared conceptual representations (Kroll and Stewart 1994). Figure 17.1 depicts the contrasting word association model and the concept mediation model (adapted from Jiang and Forster 2001). The difference between these two models is where the second language words are associated. On the one hand, the L2 words are associated with first language words, as depicted in the word association model, or on the other, are directly associated to the concepts they denote. To test which of these models more accurately represent bilingual speakers’ lexical systems, researchers have compared response latencies in picture-naming and translation tasks (Potter et al. 1984). This insight came from prior studies, which have demonstrated that individuals can name words in a first language about 250 ms faster than naming pictures in the first language (Potter and Faulconer 1975, Smith and Magee 1980, Theios and Amrhein 1989). According to these authors, the reason participants take longer to name pictures is that picture naming requires access to concepts; word naming, on the other hand, does not require access to concepts, and therefore takes less time. These models make different predictions about the speeds of picture naming and translation for bilinguals, as shown in Figure 17.2 (Potter et al. 1984). Under the word association model, translation into the second language is a shorter sequence than picture naming in the second language, therefore taking less time. In a picture naming task, the image prompt requires retrieval of the concept before retrieving the L1 word, and finally, retrieval and production of the L2 target word. Compared to a translation task, however, an L1 prompt bypasses conceptual retrieval and can directly access the L2 word for production. Because a translation task omits two steps in the sequence, it takes a shorter length of time to translate into the L2 than name a picture in the L2 under a word association model. The concept mediation model, however, predicts little or no difference between picture prompts and L1-word prompts when the task is to produce the L2. Unlike the WAM, producing an L2 word after viewing a picture prompt has the same number of steps as the process of translating from the L1 into the L2 under the concept mediation model. Notably, while the pathways are similar, the first step is not identical. Given these models, Potter et al. (1984) compared translation and picture-naming response times in a group of highly fluent Chinese-English bilinguals. Potter et al. (1984) hypothesized that if the time to translate into the L2 was faster than picture naming in the L2, then a participant’s L2 system relied on lexical links to the L1 and did not have conceptual access (WAM). Conversely, the concept mediation model would be indicated by speeds of translation into L2

Figure 17.2 L2 picture naming and L1 to L2 translation in the two models (Potter et al. 1984:26)

364

Effect of age of acquisition on concept mediation

that were similar to picture naming into the L2, because both processes require conceptual access to retrieve the L2 word. The results of Potter et al. (1984) clearly supported the concept mediation model, with the bilinguals taking about the same amount of time to translate words from L1 to L2 as naming pictures in L2. Potter et al. (1984) interpreted these results to indicate that both the translation and picture-naming tasks followed a trajectory through a shared conceptual system: therefore, both the L1 and L2 lexical systems were conceptually mediated. To see if L2 fluency was the factor that determines the form of a bilingual’s connection, Potter et al. (1984) also tested a group of less fluent English-French bilinguals. Surprisingly, the results for the less-fluent group also supported the concept mediation model. The conclusions offered by Potter et al. (1984) suggested that concepts mediated translation equivalents for all bilinguals in any given level of fluency. Kroll and Curley (1988) challenged this conclusion, suggesting that the responses of lower proficiency bilinguals should be different from more fluent bilinguals. They suggested that the low-proficiency bilinguals in the Potter et al. study (1984) had already passed the word-associative point of early second language development, where lexical links mediate the processing of second language words. Speculating that the English-French bilinguals were already too fluent, Kroll and Curley used a wider range of bilingual participants, including some with less than two years of language experience (1988). Following a similar procedure as in the Potter et al. (1984) study, Kroll and Curley (1988) demonstrated that participants who studied an L2 for less than two years had results consistent with word association predictions. In other words, for beginning and nonproficient second language learners, translation into an L2 was faster than picture naming in an L2 (Kroll and Curley 1988, Jiang and Forster 2001). These conclusions provided support for a developmental hypothesis (also referred to as “intermediate hypothesis”) which states that a non-proficient bilingual’s L2 lexicon relies on word-to-word connections to the L1 during early stages of L2 acquisition (Kroll and Stewart 1994, Gekoski 1980, Opoku 1983, Chen and Leung 1989). According to these authors, after achieving a certain level of proficiency, words of the L2 system shift from relying on word-to-word L1 connections to relying on direct connection to their respective concepts. However, these studies confound age of acquisition with proficiency. High proficiency in a language is often associated with an early age of acquisition (AoA) in that language, while low proficiency is associated with a later AoA, when considering the typical acquisition trajectory (Deuchar and Quay 2000, Jiang 2000). Therefore, given the results of these studies, the links from the lexicon to the conceptual space (as in the CMM) may develop not with proficiency but instead as a factor of the age at which a language is acquired. In an attempt to resolve this confound, a study by Chen and Leung (1989) replicated the Potter et al. (1984) study with the same low proficiency and high proficiency groups. The researchers also included a third group of participants, child beginners, whose low proficiency was coupled with an early age of acquisition. Their rationale is clear: any differences in the performance of child beginners to adult beginners is attributable to age of acquisition of the L2 rather than proficiency. However, significant differences between the child and adult L2 learners rendered their results indiscernible. Not only were the ages of L2 acquisition different between the child and adult learners, but the children’s ages, cognitive abilities, and length of experience with the language differed significantly from the adults’ as well. The present chapter instead uses adult heritage speakers to decouple the effects of AoA and proficiency rather than child learners of a second language. Characterized by an early age of acquisition paired with low proficiency, the response rates of heritage speakers can be compared with that of late AoA/low-proficiency speakers and early AoA/high proficiency speakers 365

Iyad Ghanim

to observe the effects of each variable independently. Further, using adult heritage speakers reduces the variable effects of age, cognition, and length of language experience that made a direct comparison impossible in Chen and Leung’s study (1989). As further rationale for why early acquisition may lead to the development of lexicosemantic links, I consider the case of sequential childhood bilinguals, who acquired one language exclusively for a period of time before beginning exposure to a second language during early childhood. For these individuals, it is safe to assume that early exposure to the first language results in the development of direct links to the conceptual domain. This follows the trajectory of monolingual language acquisition, where words are directly associated with their semantic and conceptual meaning as they are acquired by children. Two scenarios are theoretically possible for the L2, which was acquired later in childhood: the L2 words can be associated to the translation equivalents of first language, as in the word association model. This is plausible, given the fact that it is the second language acquired and, in theory, is the non-dominant language for the speaker. Conversely, though, it is also possible that the words of the second language can earn a direct connection to the conceptual store if it becomes the more dominant language. This illustrates the trajectory of a heritage speaker. For heritage speakers, acquisition of the first language either halts or regresses, and the L2 becomes the dominant language (Valdés 2000). For that reason, despite acquiring it at an early age, they demonstrate minimal proficiency in their L1. Their linguistic circumstances serve as a testbed for the effects of an early AoA without high proficiency inasmuch as it relates to developing lexico-semantic links (Scontras et al. 2015). The fact that heritage speakers may be experiencing lexical retrieval difficulties is interesting, but also unaccounted for given an understanding of lexical retrieval processes. If we follow the previous studies’ understanding of proficiency as it relates to developing concept mediation, we can predict the lexicon of heritage speakers (who demonstrate markedly low proficiency) to be word-associative in nature (Kroll and Curley 1988, Chen and Leung 1989). However, this is improbable considering what is expected from childhood first language learning. During acquisition, speakers ought to have developed lexico-semantic links as the words of the heritage language were acquired. Few studies used heritage speakers to investigate the effects of AoA on developing links between lexical items and their respective concepts. One such study by Montrul and Foote (2012) looked at the individual and interactive effects of word-level AoA in addition to language AoA (typically called “global AoA”; Montrul and Foote 2012). In other words, rather than just observing the age at which an individual is first exposed to a language, this study also looked at the age at which each word is first learned and its individual conceptual connection. The two groups of the Montrul and Foote (2012) study were English-Spanish bilinguals who differed in their global AoA of Spanish but who were both dominant in English (i.e., L2 Spanish learners and heritage Spanish speakers). Each group conducted a lexical decision task in Spanish, as well as an English-Spanish translation decision task. The researchers found that the L2 learners responded just as accurately but far more quickly than the heritage speakers. Relevant to the present study, these results cannot inform which model – word association or concept mediation – better characterizes the lexical organization of L2 learners compared to heritage speakers. Only the results of the English to Spanish translation decision task indicated any difference between the heritage speakers and that of the L2 learners. It is impossible to determine whether this is attributable to the heritage speakers’ lexical retrieval failures or some other grammatical failure. Moreover, Montrul and Foote’s (2012) results do indicate a difference between speakers who acquired their language in childhood and L2 learners. The operative variable between these two groups of speakers is global AoA. 366

Effect of age of acquisition on concept mediation

17.2  Research questions The goal of this current study is to distinguish the effects of AoA from proficiency by approximating the procedures of the Chen and Leung study (1989). Chen and Leung (1989) compared the response latencies of a picture-naming and translation task with young bilinguals. Similarly to Montrul and Foote (2012), I will use heritage bilinguals as participants in a picture-naming and translation study. I predict that heritage speakers’ response times will match the predictions of the concept mediation model, with L2 picture naming taking about the same amount of time as translation to L2. This would suggest that even for individuals who have lost proficiency in a language, the fact that a language is acquired during an early age ensures lexico-conceptual links. I acknowledge that the use of the labels L1 and L2 to reference the order of the languages’ acquisition may unintentionally imply language dominance. This may be misleading in the case of heritage speaker, where the “L1” becomes the weaker language and the “L2” becomes the dominant language. For bilinguals who have acquired two languages simultaneously, the use of these labels is even more misleading. Therefore, I will henceforth refer to the heritage language as the “weaker language” (WL) for clarity, and the societal language – which is second acquired for successive bilinguals – as the language that became the “dominant language” (DL).

17.3 Methods 17.3.1 Participants The participants were 11 college-aged individuals who learned Arabic as a heritage language in United States communities where the societal language is English. Thirteen participants were recruited for the study. Inclusion criteria were: (1) having spent a majority of the first years of life in the United States where the societal language is English, (2) having a household language of Arabic, (3) beginning English exposure between 0–5 years old, (4) having once had proficiency in Arabic about as expected based on age, and (5) having at some point between the ages of 5–11 lost some proficiency in Arabic. The eligible 11 participants ranged from very low-proficiency to very high proficiency. They were six females and five males, and their ages ranged from 15 to 21 (their age mean was 18.45). They were all Arabic heritage bilinguals, all with a dominant English L1. Preference was placed on selecting participants who were exposed to a Levantine Arabic dialect. Moreover, in order to avoid dialect variation as a confounding variable, the lexical tokens involved in the study were chosen to be dialect non-specific.

17.3.2 Stimuli Both picture stimuli and spoken-word stimuli were used for the experiment. Picture stimuli consisted of a total of 46 images representing nouns from the Khawaileh et al. (2014) database of Arabic nouns with corresponding images in the Snodgrass and Vanderwart (1980) norms, as described by Sholl et al. (1995). Of the total spoken-word stimuli, 24 were from the Arabic database with corresponding images in the picture-naming norms; the remaining 22 did not have corresponding images in the picture-naming norms (see Appendix A). Extracting only the tokens from the Khawaileh database with the Snodgrass and Vanderwart norms for the picture-naming tasks resulted in consistent line-drawn images of household objects, animals, and other nouns (Appendix A). Further, use of the Arabic database allowed tokens in all tasks to be stabilized in aspects such as the age at which the word is acquired (AoA), participant agreeability on the objects name (name agreement), conceptual and visual complexity of the item being named (visual complexity/naming latency), etc. (Khwaileh, et al. 2014). 367

Iyad Ghanim

This was designed to reduce extraneous variables that could affect the duration of retrieval in the tasks. Further, because the words were matched for lexical-level AoA (i.e., they are all words that are acquired at roughly the same time during childhood), they are basic-level words that are present in the early lexicon of both bilinguals and monolinguals (Khawaileh et al. 2014).

17.3.3 Design The experiment involved four total tasks: picture naming in English, picture naming in Arabic, translation into English, and translation into Arabic. These tasks were split across two blocks, with one block containing the picture naming tasks, and the other block containing the translation tasks. The task type was blocked, and the order of the blocks were counterbalanced. Crucially, this means a participant completed one type of task in both language directions before continuing to the next task type. Each block of the picture-naming phase consisted of 22 line drawings of standard objects presented in succession. Each phase also included two “practice” items to demonstrate the task and were not included in the analysis. At the onset of the presentation of each line drawing stimulus, a 400 Hz tone was heard for 400 ms. Participants named the item out loud into a recording device and then pressed the space-bar to continue to the next token. The key press triggered an inter-stimulus interval of 500 ms before the presentation of the next token. In the translation phase, one block consisted of 22 DL items that the participant translated into the WL. The other block consisted of 22 WL items that the participant translated into the DL. Because heritage language speakers have inconsistent language reading abilities, tokens in the translation phase were auditory prompts recorded by an American college-aged male who was a fully balanced native speaker of Lebanese Arabic, who learned English and Arabic simultaneously. The rationale for the auditory stimuli is that heritage speakers often demonstrate inconsistent language reading abilities, if any at all (Polinsky 2006). Difficulties related to reading the script would add time to the response latencies, which are only intended to measure cognitive processing. Blocks and phases were separated by short filler paragraphs in the DL with one or two comprehension questions for a total of three filler tasks. This served to “reset” the participant to their dominant language and reduce order and practice effects. The total of four blocks and fillers were presented to participants on a 14.7 inch laptop monitor using PsychoPy 2 experimental display software (Pierce 2007). In order to operationalize participants’ proficiency levels relative to that of the other participants, each was asked to complete a grammaticality judgment task containing five grammatical and five ungrammatical sentences. For this task, a total of 20 sentences was recorded by a native Arabic speaker, ten of which represented one of five grammatical errors ranging from subtle to highly overt errors. In order of increasing egregiousness, the error types are: verb agreement, nominal constructions, gender agreement, preposition use, and formation of plurals, as motivated by Polinsky (2006). The complete table of grammaticality judgment sentences can be found in Appendix B. According to Polinsky, proficiency correlates with perceptibility of ungrammaticality: the subtler error types are harder to perceive for lower proficiency speakers, with higher proficiency speakers being increasingly likely to notice these subtler errors (Polinsky 2006).

17.3.4 Procedure After completing a short language history survey designed to validate and compare their language circumstances, participants completed the grammaticality judgment task (Albirini and Benmamoun, 2015). Participants then read the instructions provided and began the experiment, 368

Effect of age of acquisition on concept mediation

consisting of two picture-naming tasks and the two translation tasks. For all phases, responses were recorded via voice-recording software on a mobile device and were then uploaded into Praat for response latency measurement (Boersma, 2001). Response latencies for the picture-naming tasks were measured from the onset of the tone, which was shown concurrently with the picture, to the onset of the participant’s spoken response of the correct item. The participant’s correct response does not include fillers such as “um” or “ahh.” These articulations were ignored and counted as latency. False starts, where a participant begins an articulation, halts, and begins the word again, were also counted as latency and not included as a response onset. In addition to response latency, the accuracy of a response was also recorded and measured as either a failed response (an “I don’t know,” “pass,” or “I don’t remember”) or an erroneous (non-target) response. Failed and erroneous responses were excluded from the analysis of the response times where the participant reached the target successfully. Since there was no successful retrieval in a failed response or an erroneous response, they were not included in this analysis and had their own analysis. The survey with biographical information and the proficiency assessment confirmed the participants’ status as heritage bilinguals. An analysis of their actual responses during the experimental tasks confirmed this, being in line with observations set forth by Polinsky (2006), Sherkina-Lieber (2011), and others. Phonologically, many of the extremely low-proficiency speakers were not able to produce segments not present in English, such as the pharyngeal fricatives in words as tuffaaħa “apple” or ʕayn “eye,” though the intermediate and higher proficiency participants were able to do so easily. Low-proficiency participants often deleted such segments and compensated for them with a long vowel, glottal fricative /h/ or glottal stop (i.e., */təfaa/, */ʔayn/). In terms of the lexicon, all but the highest proficiency heritage speakers produced false starts and tip-of-the-tongue experiences, as well as indications that they could not recall the target word. These patterns appeared to correlate with proficiency, with lower proficiency speakers having longer and more frequent recall difficulties. Formal measurements of these correlations will be discussed.

17.4 Results Two dependent variables were measured for the purposes of this study: first, the length of time until the speaker’s response, and second, the accuracy of their response. Much of the analysis divides between the lower proficiency speakers and the intermediate/high proficiency speakers, given the natural contrast in the raw data. The participants’ performance on the proficiency assessment modeled after Polinsky (2000, 2006) was interpreted as an indication of the participants’ proficiency in heritage Arabic relative to the other participants in the current study. Participants 114 and 103 had one (1/10) and three (3/10) judgments correct, respectively, out of ten total judgments. In conjunction with their reaction times, these participants represent the lowest proficiency speakers. Participants number 115 (scoring 4/10), 101 (4/10), 113 (5/10), 102 (5/10), and 110 (6/10) represented more or intermediate proficient speakers. The remaining participants number 105 (7/10), 109 (8/10), 111 (9/10), and 117 (10/10) represented the speakers who scored the highest on the grammaticality judgment task (see Table 17.1 for results and ranking of the participants’ proficiency levels together with their various demographics).

17.4.1  Reaction time The hypothesis of this study was that heritage speakers’ reaction time will match the predictions of the concept mediation model, with WL picture naming taking about the same time as 369

Iyad Ghanim Table 17.1 Participants’ proficiency rankings Participant

Arabic Proficiency

Proficiency Level

Began English Exposure

Ended Arabic Exposure

114 103 115 101 102 113 110 105 109 111 117

(1/10) (3/10) (4/10) (4/10) (5/10) (5/10) (6/10) (7/10) (8/10) (9/10) (10/10)

low low intermediate intermediate intermediate intermediate intermediate high high high high

birth 0–5 y/o 0–5 y/o 0–5 y/o 0–5 y/o 0–5 y/o birth 0–5 y/o birth birth 5–10 y/o

continuing continuing 0–5 y/o continuing 0–5 y/o continuing continuing continuing continuing continuing continuing

translation from the DL to the WL. Response times to the picture naming in English task (PNE) were low, with the mean response rate attenuated for missingness in the data being the lowest of the tasks at 973 ms. This falls in line with response rates of monolinguals naming pictures, or proficient bilinguals naming pictures in their dominant language as presented in previous studies (Jiang and Forster 2001, Kroll and Stewart 1994, Sholl et al. 1995). In Figure 17.3, I present the response rates of each task across all participants. Mean response times to the picture naming in Arabic task (PN-A) were, conversely, the highest among the four tasks at 1881 ms, which is almost double the response rate of PN-E. This is also as expected, given the difficulties with the WL that the heritage speakers experience. Mean response rate for the translation into English task (Tr-E) and the translation into Arabic task (Tr-A) were similar, at 1109 ms and 1468 ms, respectively. Similarly to the results of the present study, the results of translation tasks in previous works show translation into the WL taking longer than the DL, or in some trials were equivalent in speed (Kroll and Stewart 1994). To test if heritage speakers show evidence of concept mediation, I investigated the difference between the PN-A and Tr-A tasks, which is key to indicating which lexical model describes each participant. As shown in Figure 17.4, while PN-A took an average 412.4 ms longer, the variance in the participants was large, with some participants demonstrating an additional 2653 ms lag on the PN-A task. Figure 17.4 demonstrates that for only the lowest proficiency speakers, the difference between the picture naming in Arabic task and the translation into Arabic task is far larger than that of the remaining participants, with the added reaction time possibly indicating the additional processing time for lexical mediation through the dominant language. Surprisingly, the absolute mean difference in the task speeds for intermediate proficiency speakers (34.7 ms) is closer to zero than the means of the other groups (2304.3 ms for low proficiency speakers and 550.8 ms for high proficiency speakers). The fact that the intermediate proficiency speakers’ PN-A lag is close to zero can indicate that a concept mediation model is likely characteristic of these speakers. Conversely, the large lag in the PN-A task for the lower proficiency speakers is accounted for in a word-associative model, in which the picture prompt must be mediated through the DL system first before production in the WL. For analysis, I subjected these data to a 2 (task type) × 2 (language) × 3 (proficiency) mixed-effects model to evaluate the hypothesis regarding the main effect of proficiency on reaction time, with random effects for multiple observations within the subject, and 370

Effect of age of acquisition on concept mediation

Figure 17.3 Response latencies for each task (ms)

Figure 17.4 Lag between Arabic picture naming and Arabic translation tasks

conducted a two-tailed t-test. For the lowest proficiency participants, the reaction time of the Arabic picture-naming task (µ = 4.6032 s, SE = 0.2767) was significantly slower than the reaction time for the translation into Arabic task (µ = 2.2989 s, SE = 0.2265); t(739) = 7.05, p < .0001. For the subjects with intermediate proficiency, the reaction time of the picture-naming task (µ = 1.5796 s, SE = 0.1235) was not significantly different from the translation task (µ = 1.6143 s, SE = 0.1244), indicating a concept mediation model; t(739) = −0.23, p = 0.8175. However, for the highest proficiency subjects, the reaction time of the picture-naming task (µ = 1.5690 s, SE = 0.1188) was significantly slower than that of the translation task (µ = 1.0181 s, SE = 0.1224); t(739) = 3.98, p < .0001. 371

Iyad Ghanim

I also analyzed the difference in the mean picture-naming task response rate and the mean translation task response rate between proficiency groups. The lower proficiency speakers’ difference in the picture-naming task and the translation task (i.e., the picture-naming lag) (µ = 2.3043 s, SE = 0.3270) was significantly different from that of both the intermediate proficiency speakers (t(739) = 6.50, p< .0001) and the high proficiency speakers (t(739) = 4.94, p < .0001). Additionally, the differences in the picture-naming task and the translation task for the intermediate proficiency speakers (µ = −0.03470 s, SE = 0.1503) compared to that of the highest proficiency speakers (µ = 0.5508 s, SE = 0.1383) (105, 109, 111, and 117) was also significant; t(739) = −2.87, p = 0.0043.

17.4.2  Response accuracy To quantify the accuracy of participants’ responses, I distinguish between two types of nontarget responses: (1) a failed response, which is a “don’t know” or a “pass” that indicates a failure to retrieve the word; and (2) an erroneous response, which is a response that deviates from the target. The rates of non-target responses are presented in Figure 17.5 for each task and for all participants. Throughout experimentation, participants produced a total of 272 failed responses and 23 erroneous responses.

Figure 17.5 Failed and erroneous response counts per task

372

Effect of age of acquisition on concept mediation

17.4.3  Failed responses Failing to provide a response for a certain token of the experimental tasks represents an inability to retrieve either the conceptual information for that token (picture-naming tasks) or the translation equivalent for that token (translation tasks). As shown in Table 17.2, the task with the highest number of failed responses was the Tr-A task (91, or 33.5% of all failed responses); and the task with the second highest number of failed responses was the PN-A task (85, or 31.3% of all failed responses). English tasks (PN-E and Tr-E) were relatively consistent between participants, with very low rates of response failure in the PN-E task (µ = 0.6) and a mean of 2.57 failures in the Tr-E task. In relation to participants’ proficiency, I predicted high rates of failures for low-proficiency participants, especially for the tasks that require access to the WL lexicon (PN-A and Tr-A). The failed response and erroneous response rates out of the total responses of each group are presented in Table 17.3. The fact that the failed response rate for the PN-E task was greater than zero for medium and high proficiency speakers is surprising, as I expected a value closer to zero for dominant language picture-naming failures. While producing any errors at all in a dominant language picture-naming task is unexpected, the 18 errors in the PN-E task were a result of individuals’ inabilities to identify or name the line drawings. This could be due to the nature of the line drawings, the small number of practice items, or other issues with the visual presentation of the experiment. However, the other tasks – the picture naming in Arabic task, the translation into English task, and the translation into Arabic task – appear to each show a correlation with proficiency (in that higher proficiency participants produce less failed responses than low proficiency participants). I used a standard two-tailed t-test in a mixed-effects model to test the interactive effects of proficiency and failed error rate. The failed response count made by participants with lower proficiency was significantly greater than zero (µ = 41.5, SE = 5.4618). While the intermediate participants made fewer failed responses, the number of their failed responses was also significantly greater than zero (µ = 28.2, SE = 3.454). In addition, the number of failed responses made by the high proficiency participants was significantly greater than zero (µ = 12.0, SE = 3.8621). These results are presented in Table 17.4. Table 17.2 Failed and erroneous response percentages per task

Failed Responses (272 total) Erroneous Responses (23 total)

PN English

PN Arabic

Tr. English

Tr. Arabic

6.6% 0%

31.3% 17.4%

28.3% 82.6%

33.5% 0%

Table 17.3 Percent failed and erroneous responses of all responses per participant group Low Proficiency

Medium Proficiency

High Proficiency

PN-E PN-A TR-E TR-A PN-E PN-A TR-E TR-A PN-E PN-A TR-E TR-A Total No. of Responses 44 44 44 44 % Failed Responses 0% 75% 43% 70% % Erroneous Responses 0% 0% 9% 0% % Target Responses 100% 25% 48% 30%

110 10% 0.9% 89%

373

110 110 110 88 36% 29% 40% 9.1% 1.8% 9% 0% 0% 62% 62% 60% 91%

88 9.1% 2.3% 89%

88 88 19% 17% 5.7% 0% 75% 83%

Iyad Ghanim Table 17.4 Mean failed responses by proficiency group using a mixed-effects model Parameter

Estimate

Standard Error

t Value

Pr > |t|

Low Proficiency Speakers Intermediate Prof. Speakers High Proficiency Speakers

41.500 28.200 12.000

5.4618 3.4543 3.8621

7.60 8.16 3.11

< .0001 < .0001 0.0145

Table 17.5 Mean erroneous responses by proficiency group using a mixed-effects model Parameter

Estimate

Standard Error

t Value

Pr > |t|

Low Proficiency Speakers Intermediate Prof. Speakers High Proficiency Speakers

2.0000 2.4000 1.7500

1.4990 0.9480 1.0599

1.33 2.53 1.65

0.2188 0.0352 0.1373

17.4.4  Erroneous responses Rather than the absent lexico-semantic connections that are indicated by failed responses, the number of erroneous responses indicate misdirected or weak connections to either the conceptual space (picture-naming tasks) or to translation equivalents (translation tasks). In looking at erroneous responses, the Tr-E task had a far greater number of erroneous responses, with 82.6% of all erroneous responses (19 errors), while the Tr-A task had no erroneous responses. For the picture-naming tasks, the PN-E task produced no erroneous responses, while the PN-A task accounted for only 17.4% of erroneous responses (four errors), as shown in Table 17.4. While the Tr-E task had a far greater number of erroneous responses, there was no correlation between erroneous responses and participants’ proficiency. The mean of the erroneous responses of low (µ = 2.00) and high proficiency speakers (µ = 1.7500) were non-significant compared to zero, as represented in Table 17.5. However, the mean of the erroneous responses of intermediate proficiency speakers (µ = 2.400) were significantly different from zero. The result is that the erroneous response rates are generally low and are not significantly different from zero for low and high proficiency groups, but are significantly greater than zero for the intermediate proficiency group. This, however, should take into account that the intermediate proficiency speakers produced a greater number of responses in general, and so the likelihood they would produce errors was high as well. The highest proficiency speakers, as expected, had the lowest mean error rate, indicating more robust connections. The fact that the majority of errors occur in the Tr-E task, and decidedly not in the Tr-A task, indicates that WL words are activating inaccurate DL translation equivalents. These results are also logical, considering the limited WL knowledge of heritage speakers.

17.5 Discussion 17.5.1  Discussion of general findings Analyzing the difference between the responses of the PN-A task and the Tr-A task was key to determining participants’ lexical organizations. For concept mediators, this “lag” should be very close to zero, because these individuals use the same path through the conceptual domain for both translation and for picture-naming tasks. For word-associative models, PN-A should take a significantly longer time than translation into Arabic, accounting for the extra processing time required for the L2 words to be mediated through their L1 translations. 374

Effect of age of acquisition on concept mediation

When looking at the differences in the PN-A task and the Tr-A task for each participant, it is evident that the two lowest proficiency speakers – 114 and 103, who respectively scored 1/10 and 3/10 on the proficiency assessment – had a significantly higher lag time than the rest of the participants. This contrast suggests that words may not be semantically linked for individuals of a critically low proficiency. Instead, this supports word association: that Arabic lexical items are associated to English translations. This appears to agree with the interpretation of the results of Potter et al. (1984), Kroll and Curley (1988) and Chen and Leung (1989), suggesting that higher proficiency leads to concept mediation. However, if Potter et al.’s interpretation was applicable to heritage speakers, the results should indicate a PN-A lag that gradually decreases with increased proficiency (1984). Instead, there is no gradient as proficiency increases; for all but lower proficiency speakers, the PN-A lag is insignificant. Therefore, it is not likely that language proficiency alone can lead to semantic connectedness. Based on the results of the Potter et al. (1984), Kroll and Curley (1988), and Chen and Leung (1989) studies, intermediate level second language learners can be described as word associators. We can equate the proficiencies of these intermediate second language learners to the intermediate heritage language speakers of this current study based on the descriptions provided in the previous works. However, the critical difference between the speakers described in the previous works and the participants of the current study is the age at which they acquired what became their second language. In the present study, however, we show intermediate level participants are better described as concept mediators. Therefore, this result suggests that age of acquisition does allow an individual to develop semantic connections, and also that those lexico-semantic connections are resilient in most experiences of fluency loss (i.e., except for cases of extremely low proficiency). We cannot conclude for certain that acquiring a language in early childhood is a perfect guarantor of lexico-semantic connections across all levels of proficiency. This is because the low-proficiency speakers in this study, similar to the other proficiency speakers, had an AoA in early childhood, but their processing times appeared to be different (see Figure 17.4, specifically the PN-A lag of low-proficiency speakers). Because of this, AOA alone cannot account for the pattern of word association. Instead, these results could suggest an interaction between proficiency and AoA in developing conceptual mediation. It is likely that for heritage speakers, having an early AoA helps preserve lexico-semantic links that are eroded from decreased proficiency. In contrast, for language learners acquiring an L2 in adulthood, the level of proficiency that must be reached before developing conceptual mediation is simply higher, as they do not have the effects of AoA to help lower this conceptual-mediation threshold. This would explain the difference in the results between the lower proficiency heritage speakers compared to the intermediate proficiency speakers, who differ only in AoA. This supports, therefore, that AoA has a facilitatory effect on developing concept mediation. In almost all cases, learning a language in early childhood ensures direct lexico-semantic development between that language’s lexicon and the semantic store. Further, these lexicosemantic links are a highly resilient linguistic feature that is lost late in the process of severe attrition; that is, only among individuals with a critically low proficiency.

17.5.2  Response type discussion The results of the failed responses show that lower proficiency participants more frequently fail to respond to the experimental prompts. The high rates of response failure in tasks that require access to the WL lexicon (PN-A and Tr-A) is expected for low proficiency subjects, 375

Iyad Ghanim

given these subjects’ low proficiency in Arabic. As a result, it is generally expected that proficiency and failed response rate are correlative. Specifically, a high failed response rate for picture naming in Arabic tasks indicates no conceptual connection between the weaker language (WL) and the conceptual domain, as is represented in the word association model. Conversely, an approximately equal failure rate in the translation into Arabic task can be attributable to poor proficiency in the weaker language. The low-proficiency speakers’ failed response results are in line with both the reaction time results and the results of previous studies in suggesting a pattern of word association. Similarly, a significantly smaller PN-A failed response rate for the intermediate proficiency speakers suggests a concept mediation model. The result of the accuracy measurements for the intermediate speakers is also in line with the reaction time results for the intermediate speakers. By the same rationale, the low failed response rates of the high proficiency speakers indicates a CMM, which counters the results of the reaction time results. This also counters the results of the previous studies, which attribute a conceptual-mediation model to high proficiency speakers. The discord between the accuracy results and the reaction time results indicates the need for further investigation on high proficiency subjects. The aggregated analysis of the erroneous and failed responses per task has other interesting implications. It is notable that all participants produced no erroneous responses during the Tr-A task but expressed a high failed response rate. This indicates that individuals are more inclined to fail to respond rather than to respond erroneously when responding in the weaker language. This result indicates that the participants knew when they did not know the translation for the Arabic word. This weakness in the lexical items of the weaker language is typical for heritage speakers (Polinsky 2000, 2006). Given the reaction times, I still argue that the weaker-language domain has access to the conceptual space; simply, not all words have a connection. The results of the Arabic language tasks (PN-A and Tr-A) are expected for low proficiency heritage speakers with reduced information in the weaker-language lexical entries. The fact that lower weaker-language proficiency correlates with failed responses confirms this, as individuals with lower proficiency also have a more limited Arabic lexicon. Informal observations of participants’ errors seem to agree with these findings. As an example, Participant (115), an intermediate proficient participant, produced responses that appeared to be semantically similar to the target. During the translation into Arabic task, the spoken English word prompt was “wallet” (target: žizdaan), and she produced the Arabic for the word “notebook” (i.e., daftar). Notably, this implicates the semanticity of both folding objects more so than the limited phonological relatedness of the words. During the translation into English task, the same participant (115) responded to the Arabic prompt /binaaya/ (target: building) with the English word “apartment,” for which the Arabic is šaqqa. Similarly of note, these words are far more semantically related than they are phonologically related. This appears to evidence a conceptual mediation for the translation task. For this participant, Arabic picture naming appears to take almost the same length of time as translation into Arabic (on average only about .200 ms faster), providing more evidence of conceptual mediation. However, as a counterexample, Participant (116) produced erroneous responses that were phonologically – not semantically – similar to the target. During the translation into English task, this participant responded to the Arabic prompt bərdʔaana (target: orange, as in the fruit) with curtain, the Arabic for which is bərdaaya. Yet others (110, an intermediate proficiency participant, and 114, a low-proficiency participant) responded with the word bardaana “cold” rather than the target bərdʔaana “orange.” In these cases, phonological errors are evidence for a lack of conceptual mediation. While the data of Participant116 were disqualified on the basis 376

Effect of age of acquisition on concept mediation

of incompletion, had the participant completed the study, we would predict a word association model because of the status of his errors. Crucially, no participant ever made both a semantically related error and a phonologically related error throughout the course of their experimentation, and, largely, these errors appeared to correlate with the participant’s proficiency, with higher proficiency subjects making errors based on semanticity. Based on these informal observations, the linguistic triggers that heritage speakers may rely on to retrieve words would be an interesting dimension of this research.

17.6  Conclusion and implications The present study found evidence that some heritage language bilinguals may show evidence of concept mediation even without high proficiency. The reaction time results of nine of the heritage speakers fell in line with predictions made by the concept mediation model, in that they demonstrated a small delay in picture-naming tasks relative to translation tasks. This small lag is associated with about-equal processing time for each task, which can best be explained by a conceptual-mediation model, where each language domain is directly connected to the conceptual space. Conversely, the results of the two lowest proficiency heritage speakers did not. The lower proficiency speakers showed a significantly larger lag time between the two tasks, which is associated with longer processing time for a process which requires access to a conceptual space (i.e., a picture-naming task) relative to a task that only accesses the two language domains (i.e., a translation task). Previous studies maintained that proficiency was the only factor that led to conceptual mediation. However, given the present data, speakers who were less than fully proficient appeared to conceptually mediate word items. An additional exploration of the lexical retrieval processes of heritage bilinguals as it relates to concept mediation or word association is required. In isolation, the results of the intermediate and high proficiency speakers may suggest that age of acquisition has an effect on developing conceptual mediation, even without language proficiency. However, considering the results of the low proficiency speakers, it is critical to consider a model that allows proficiency and AoA to interact (Montrul & Foote, 2012). Given the results, it appears that despite having acquired Arabic at an early age, losing the language skills leaves the low proficiency speakers dependent mainly on word-to-word connections. It seems that an early age of acquisition can facilitate lexico-semantic links only while having reached a certain level of proficiency. However, without reaching a certain level of proficiency, early AoA seems to have no effect on strengthening the lexico-semantic connections. There are several ways to expand upon this study. First, designing a study comparing the reaction times of low proficiency, low AoA heritage speakers with the response times of low proficiency, late AoA second language learners would also be helpful in determining an effect of the AoA variable without proficiency. Additionally, a major limitation of this study is its small sample size. Additional participants would help to stabilize the results, demonstrate a more definitive trend, and allow the research to be more broadly generalizable. Data from a larger number of participants would withstand more rigorous statistical analysis and indicate more definitive results. In effect, including additional participants would also require a finer gradient of proficiency testing. For example, rather than a grammaticality judgment score of 1–10, a larger number of judgments or other types of proficiency assessment would be required that would operationalize language proficiency on a continuous scale. Conducting such tasks with heritage speakers in a range of proficiencies might illuminate additional details regarding what lexico-semantic connections might have once existed during childhood and what 377

Iyad Ghanim

connections still remain. This also has the possibility of uncovering what mechanisms account for the productive deficiencies of heritage speakers and how their language skills can be maintained and improved. Taken as a preliminary study, this present work certainly highlights the need for additional research in word access for heritage speakers. Additionally, a wider range of stimuli and a repeated-measures design that re-used stimuli in different tasks would support the validity of the study and show, perhaps, that picture naming was faster than translation, even for the same exact concepts. This would reduce the effect that the semantic classification of words, frequency in childhood, phonological properties, and word-level AoA has on recall speed. Owing to their unique trajectories and the resulting linguistic outcomes, heritage speakers can serve as a great benefit to research in bilingual acquisition. What can trigger successful recall of lost or inaccessible lexical items remains to be studied, as well as how heritage speakers cognitively manage their two languages (such as choosing between them in code-mixing situations). With significantly asymmetric language abilities, heritage bilinguals remain to be more extensively studied in future research. The pedagogical implications are far-reaching for both heritage language learners and language teachers, especially those who teach heritage learners. Because of this evidence that heritage speakers retain the lexical connections to the semantic store, heritage learners may benefit from continuing to strengthen these semantic connections rather than rely on translating from the L1, as many L2 beginners do. This is easily achieved through continuing to practice speaking the language and being immersed in the language without the presence of the more dominant language. With this in mind, the work also suggests that teachers would benefit from instructing heritage language learners in a different manner than they would beginning L2 learners. Though the proficiency levels of the two types of learners may be on par, teaching heritage learners represents a better opportunity to use the language and to teach the language in an immersive, L2-only environment. These heritage learners are better equipped than beginning second language learners to use the syntactic and semantic clues to fill in the gaps of the words they do not recognize, because the language structures are retained from their childhood exposure.

378

Effect of age of acquisition on concept mediation

Appendix A: Words Stimuli Picture # (Snodgrass and Lexical Item Vanderwart, 1980)

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44

7 12 38 49 65 69 70 78 106 116 153 2 6 14 16 25 40 43 44 55 66 68 73 75 76 84 88 89 90 91 97 100 104 114 118 121 122 130 145 149 167 185 187

arm axe brush cat comb crown cup dress glove hanger necklace saucepan airplane apple ball banana bell butterfly camel candle chicken corn cow dog donkey door elephant finger fish flag flower fork frog glass hammer hat horse house knife monkey mouse pen refrigerator ring

Arabic Text

‫يد‬ ‫فاس‬ ‫فرشاية‬ ‫قطة‬ ‫مشط‬ ‫تاج‬ ‫فنجان‬ ‫ثوب‬ ‫كفة‬ ‫عُالقة‬ ‫سلسال‬ ‫طنجرة‬ ‫طيارة‬ ‫تفاحة‬ ‫طابة‬ ‫موزة‬ ‫جرس‬ ‫فراشة‬ ‫جمل‬ ‫شمعة‬ ‫جاجة‬ ‫ذرة‬ ‫بقرة‬ ‫كلب‬ ‫حمار‬ ‫باب‬ ‫فيل‬ ‫اصبع‬ ‫سمكة‬ ‫علم‬ ‫وردة‬ ‫شوكة‬ ‫ضفدع‬ ‫كاسة‬ ‫شاكوش‬ ‫طاقيّة‬ ‫حصان‬ ‫بيت‬ ‫سكينة‬ ‫قرد‬ ‫فار‬ ‫قلم‬ ‫ثالجة‬ ‫خاتم‬

379

Practice

1 1 1 1 1 1 1 1 0 1 0 1

Picture Naming

Translation

L1

L2

0 0 1 0 1 0 1 0 1 0 1 1 1 0 1 0 1 0 1 0 1 1 1 1 0 1 0 0 0 1 1 0

1 1 0 1 0 1 0 1 0 1 0 0 0 1 0 1 0 1 0 1 0 0 0 0 1 0 1 1 1 0 0 1

to L1

to L2

Picture # (Snodgrass and Lexical Item Vanderwart, 1980)

45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90

197 198 203 205 209 212 215 217 226 241 245 252 5

22 28 30 32 36 47 86 127 128 131 140 154 158 172

182

scissors screw shirt skirt snake spider spoon star table tree umbrella watermelon ant baby bag bathroom bed bird bone book bottle boy bread building car egg eye feather girl kettle key ladder lighter lion man mirror needle orange pig pigeon plate pocket police queen rabbit scorpion

Arabic Text

‫مقص‬ ‫برغي‬ ‫قميص‬ ‫تنورة‬ ‫حيّة‬ ‫عنكبوت‬ ‫معلةق‬ ‫نجمة‬ ‫طاولة‬ ‫شجرة‬ ‫شمسية‬ ‫بطيخة‬ ‫نملة‬ ‫طفل‬ ‫شنطة‬ ‫حمام‬ ‫تخت‬ ‫عصفور‬ ‫عظمة‬ ‫كتاب‬ ‫قنينة‬ ‫ولد‬ ‫خبز‬ ‫عمارة‬ ‫سيارة‬ ‫بيضة‬ ‫عين‬ ‫ريشة‬ ‫بنت‬ ‫ابريق‬ ‫مفتاح‬ ‫ِسلم‬ ‫قداحة‬ ‫اسد‬ ‫زلمة‬ ‫مراي‬ ‫إبرة‬ ‫بردقانة‬ ‫خنزير‬ ‫حمامة‬ ‫صحن‬ ‫جيبة‬ ‫شرطي‬ ‫ملكة‬ ‫ارنب‬ ‫عقربة‬

Practice

Picture Naming

Translation

L1

L2

0 1 1 0 1 1 0 1 0 0 0 0

1 0 0 1 0 0 1 0 1 1 1 1

to L1

to L2

0 1 0 1 0 1 0 0 1 1 0 1 1 0 1 1 0 0 0 0 1 0 0 0 1 1 0 1 0 1 0 1 1 0

1 0 1 0 1 0 1 1 0 0 1 0 0 1 0 0 1 1 1 1 0 1 1 1 0 0 1 0 1 0 1 0 0 1

Picture # (Snodgrass and Lexical Item Vanderwart, 1980)

Arabic Text

Practice

Picture Naming L1

91 92 93 94 95 96

199

97 98 99

250 254 257

screwdriver ship soldier tray wallet washing machine watch wheel window

Translation to L1

to L2

‫مفك‬ ‫سفينة‬ ‫جندي‬ ‫صينية‬ ‫جزدان‬ ‫غسّالة‬

1 1 1 0 0 1

0 0 0 1 1 0

‫ساعة‬ ‫عجل‬ ‫شباك‬ Totals per

10

0 1 0 22

1 0 1 22

Totals

10

22

L2

22 44

44

Nominal Construction

Preposition

Formation of Plurals

Gender Agreement

Verb Agreement

Grammar Construction

‫الدفتر الولد ما كان بِبيته‬ ʔad-daftar ʔal-walad maa kaan b-beto ‫الولد ترك اللعبة بِالبيت البنت‬ ʔal-walad tarak ʔal-luʕbe bi-l-beetə l-bint

B

‫بيّي بيشتغل مع السيارات‬ bayye bi-yištaɣil maʕa s-sayyaaraat

B

A

‫بدوّر للكتاب‬ bidawwir li-lə- ktaab

‫هي أكلت بِكل المطعمات‬ hiyye ʔaklet bi-kull ʔal-maṭʕamaat

B

A

‫أنا ب ّدي قلمات للمدرسة‬ ʔanaa biddii qalamaat la-l-madrase

‫خاف الولد من البنت فهرب منه‬ xaaf ʔal-walad min ʔal-bint fa-harab minno

B

A

‫المرة الطويل اشترت خضرة من الدكان‬ ʔal-mara ʔaṭ-ṭawiil ʔištarat xuḍra min ʔad-dukkaan

‫اشترت الزلمة سيارة حمرة و حبها‬ ʔištarat ʔaz-zalame sayyaara ħamra w-ħabhaa

B

A

‫الولد درسوا اللغة العربية‬ ʔal-walad drasuu ʔal-luɣa ʔal-ʕarabiyya

A

Ungrammatical

‫األوالد درسوا اللغة العربية‬ ʔal-ʔawlaad darasuu ʔal-luɣa ʔal-ʕarabiyya “The boys studied the Arabic language.” ‫اشترى الزلمة سيارة حمرة وحبها‬ ʔištara ʔaz-zalame sayyaara ħamra w-ħabhaa “The man bought a red car and loved it.” ‫المرة الطويلة اشترت خضرة من الدكان‬ ʔal-mara ʔaṭ-ṭawiile ʔištarat xuḍra min ʔad-dukaan “The tall woman bought vegetables from the store.” ‫خاف الولد من البنت فهرب منها‬ xaaf ʔal-walad min ʔal-bint fa-harab minnaa “The boy was afraid of the girl, so he ran away from her.” ‫أنا ب ّدي أقالم للمدرسة‬ ʔanaa biddii ʔaqlaam la-l-madrase “I need pens for school.” ‫هي أكلت بِكل المطا ِعم‬. hiyye ʔaklet bi-kull ʔal-maṭaaʕim “She ate in all restuarants.” ‫بيدوّر على الكتاب‬ bidawwir ʕalə-ktaab “He is looking for the book.” ‫بيّي بيشتغل بِالسيارات‬ bayye bi-yištaɣil fi s-sayyaaraat “My father works on cars.” ‫دفتر الولد ما كان بِبيته‬ daftar ʔal-walad maa kaan b-beto “The boy’s notebook was not at home.” ‫الولد ترك اللعبة بِبيت البنت‬ ʔal-walad tarak ʔal-luʕbe b-beetə l-bint “The boy left the toy at the girl’s house.”

Grammatical

Appendix B: Grammaticality judgment task used to assess proficiency level of participants

Effect of age of acquisition on concept mediation

References Albirini, A. and Benmamoun, E., 2015. Factors affecting the retention of sentential negation in heritage Egyptian Arabic. Bilingualism: Language and cognition, 18 (3), 470–489. Altarriba, J., 1990. Constraints on interlingual facilitation effects in priming in Spanish-English bilinguals. Unpublished dissertation. Vanderbilt University. Boersma, P., 2001. Praat, a system for doing phonetics by computer. Glot International, 5 (9/10), 341–345. Brown, K., et al., 1984. The role of script and phonology in lexical representation. Quarterly Journal of Experimental Psychology, 7, 204–215. Chen, H. and Leung, Y., 1989. Patterns of lexical processing in a nonnative language. Journal of Experimental Psychology: Learning, Memory, and Cognition, 15 (2), 316–325. Chen, H. and Ng, M., 1989. Semantic facilitation and translation priming effects in Chinese-English bilinguals. Memory and Cognition, 17 (4), 454–462. Deuchar, M. and Quay, S., 2000. Bilingual acquisition: Theoretical implications of a case study. New York: Oxford University Press. Gekoski, W. L., 1980. Language acquisition context and language organization in bilinguals. Journal of Psycholinguistic Research, 9, 429–449. Gerard, L. D. and Scarborough, D. L., 1989. Language-specific lexical access of homographs by bilinguals. Journal of Experimental Psychology: Learning, Memory, and Cognition, 15 305–315. Glanzer, M. and Duarte, A., 1971. Repetition between and within languages in free recall. Journal of Verbal Learning and Verbal Behavior, 10, 625–630. Jiang, N., 2000. Lexical representation and development in a second language. Applied Linguistics, 21 (1), 47–77. Jiang, N. and Forster, K., 2001. Cross-language priming asymmetries in lexical decision and episodic recognition. Journal of Memory and Language, 44 (1), 32–51. Kirsner, K., et al., 1984. The bilingual lexicon: language specific units in an integrated network. Journal of Verbal Learning and Verbal Behavior, 23, 519–539. Khwaileh, T., et al., 2014. A  normative database and determinants of lexical retrieval for 186 Arabic nouns: Effects of psycholinguistic and morphosyntactic variables on naming latency. Journal of Psycholinguistic Research, 43 (6), 749–769. Kolers, P., 1963. Interlingual word associations. Journal of Verbal Learning and Verbal Behavior, 2, 291–300. Kroll, J. F. and Curley, J., 1988. Lexical memory in novice bilinguals: The role of concepts in retrieving second language words. Practical Aspects of Memory, 2, 389–395. Kroll, J. F. and Stewart, E., 1994. Category interference in translation and picture naming: Evidence for asymmetric connections between bilingual memory representations. Journal of Memory and Language, 33 (2), 149–174. Montrul, S. and Foote, R., 2012. Age of acquisition interactions in bilingual lexical access: A study of the weaker language of L2 learners and heritage speakers. International Journal of Bilingualism, 18 (3), 274–303. Opoku, J. Y., 1983. The learning of English as a second language and the development of the emergent bilingual representational systems. International Journal of Psychology, 18, 271–272. Pierce, J. W., 2007. Psychophysics software in Python. Journal of Neuroscience Methods, 162 (1–2), 8–13. Polinsky, M., 2000. A composite linguistic profile of a speaker of Russian in the USA. In: Olga Kagan and Benjamin Rifkin, eds. The learning and teaching of Slavic languages and cultures. Bloomington, IN: Slavica, 437–466. Polinsky, M., 2006. Incomplete acquisition: American Russian. Journal of Slavic Linguistics, 14, 191–262. Potter, M. C. and Faulconer, B. A., 1975, Time to understand pictures and words. Nature, 253, 437–438. Potter, M. C., et al., 1984. Lexical and conceptual representation in beginning and proficient bilinguals. Journal of Verbal Learning and Verbal Behavior, 23 (1), 23–38. Scarborough, D. L., et al., 1984. Independence of lexical access in bilingual word recognition. Journal of Verbal Learning and Verbal Behavior, 23, 84–99. Schwanenflugel, P. J. and Rey M., 1986. Interlingual semantic facilitation: Evidence for a common representational system in the bilingual. Journal of Memory and Language, 25, 605–618.

383

Iyad Ghanim Scontras, G., et al., 2015. Heritage language and linguistic theory. Frontiers in Psychology, 6 (1545), 1–20. Sherkina-Lieber, M., et al., 2011. Grammar without speech production. Bilingualism: Language and Cognition, 14 (3), 301–317. Sholl, A., et al., 1995. Transfer between picture naming and translation: A test of asymmetries in bilingual memory. Psychological Science, 6 (1), 45–49. Smith, M. C. and Magee, L., 1980. Tracing the time course of picture-word processing. Journal of Experimental Psychology: General, 109, 373–392. Snodgrass, J. G. and Vanderwart, M., 1980. A standardized set of 260 pictures: Norms for name agreement, image agreement, familiarity, and visual complexity. Journal of Experimental Psychology: Human Learning and Memory, 6 (2), 174–204. Theios, J. and Amrhein, P. C., 1989. Theoretical analysis of the cognitive processing of lexical and pictorial stimuli: reading, naming, and visual and conceptual comparisons. Psychological Review, 96, 5–24. Valdés, G., 2000. Spanish for native speakers. AATSP professional development series. Handbook for teachers K-16. New York, NY: Harcourt College. Weinreich, U., 1953. Languages in contact. New York: The Linguistic Circle of New York.

384

PART VIII

The Arabic L2 teacher Teacher training and self-positioning

18 EFFECT OF USING A COLLABORATIVE VIDEOBASED SELF-EVALUATION ACTIVITY ON HELPING AFL STUDENT-TEACHERS TIE THEORY TO PRACTICE Raghda El Essawi Through content analysis of teaching self-evaluation papers written by five student-teachers, this chapter looks into the effect of a collaborative activity that manipulates a video discussion tool, namely “Vialogues,” on teachers’ depth of reflection, as indicated by a researcherdeveloped reflection classification scheme. According to this scheme, semantic units reflecting the highest level of reflection (the targeted level) were ones that incorporated theoretical knowledge in evaluating pedagogical choices. Although collective analysis of semantic units produced by participants in the study revealed the targeted levels of reflection were achieved, analysis of semantic units produced by individual student-teachers suggested a different picture. Such results indicated that a clear variation existed among student-teachers regarding the level of reflection to which each resorted during the process of self-evaluation. As indicated by the analysis, some student-teacher participants heavily relied on the targeted levels, while others minimally used them. It was also interesting to note that some participants only relied on such levels when dealing with positive aspects of their teaching but failed to do so when reviewing less successful pedagogical procedures. In an attempt to explain such variation, the study resorted to Danielson’s (2009) presentation of Grimmett’s modes of thinking, showing that participants who displayed lower levels of reflection are in fact using modes of thinking that are inappropriate for the tasks required. In other words, they fail to use modes of thinking that encourage usage of outside sources, especially theoretical, to fully understand and hence successfully evaluate a certain teaching situation. The study ends by presenting a set of suggestions deduced from previous research to help gauge student-teachers toward forms of reflection where they are likely to lead student-teachers to tie theory to practice. These suggestions include the need to train student-teachers to use modes of thinking appropriate for targeted tasks and mentor interference during the various stages of activity to encourage using targeted levels of reflection.

387

Raghda El Essawi

18.1 Introduction 18.1.1  Importance of self-reflection An extensive body of research in the field of teacher development and teacher education today regard self-reflection as a necessary practice for good teaching. According to Kong (2010), one of the most important gains of self-reflection is that it allows teachers to “externalise their thoughts on their own teaching competence. This externalisation often leads them to reinternalise and recontextualise the understanding they have gained from this self-reflection.” (p. 1772). Furthermore, self-reflection develops teachers’ ‟reasoning about why they employ certain instructional strategies and how they can improve their teaching to have a positive effect on students” (Mirzaei et al. 2013, p. 1). Successful self-reflection, however, requires enhancing teachers’ analytical skills, which are needed to make the process of self-reflection more deep and more effective. This in turn, according to Zottmann et al. (2013, p. 2100), requires “a certain level of knowledge about concepts, principles, and theories” to facilitate “professional analyses of classroom interactions”. Of particular importance “teachers must be able to apply abstract knowledge to concrete phenomena” (p. 2100). This is not an easy goal to achieve, since student-teachers (or preservice teachers) “face enormous difficulties to draw upon the professional knowledge they acquired during teacher education” when faced with complexities of classroom reality (Zottmann et al. 2013, p. 2100). This is what Stokking et al. (cited in Blomberg et al. 2013, p. 91) refers to as “practice shock”, resulting in teachers reverting to “intuitive theories of teaching and learning that correspond with their own experiences in school rather than with the researchbased knowledge from their teacher education program” (Lampert and Ball cited in Blomberg et al. 2013, p. 91). Though such knowledge is retrievable, it does not guide teachers’ pedagogical choices (Cocharn-Smith 2003 cited in Blomberg et al. 2013). Similarly, teachers with teaching experience (or in-service teachers) find this process equally challenging for reasons like “inert knowledge” and/or “washed out professional knowledge” (Zottmann et al. 2013, p. 2100). Hence the importance of self-reflection for teacher education and teacher development where teachers are encouraged to evaluate their teaching performance in light of newly gained abstract knowledge. Beside enhancing teachers’ analytical skills (as well as their self-reflection skills), encouraging in-service and student-teachers to apply newly gained abstract knowledge on concrete in-class practices is expected to help resolve one of the main problems in teacher-education and professional development, namely the theory-to-practice gap (Zottmann et al. 2013). Self-reflection in this context is defined as the ability to draw upon abstract theoretical knowledge gained in teacher-education and teacher-training programs in order to evaluate effectiveness of teachers’ pedagogical practices.

18.1.2  Effect of collaboration on the process of self-reflection The positive effect of collaboration in learning was introduced by Vygotsky’s Social Development Theory. According to Vygotsky (1978), learners’ potential development when working with peers transcends their actual development when working alone. Research in education shows that collaborative reflection gauges thinking from the least conscious formulaic level (thinking based on information from an outside source) to most conscious dialectical level (i.e., thinking to resolve problems) (Danielson 2009). In the field of teacher education and teacher professional enhancement, researchers like Manouchehri (cited in Yang 2009) point out that collaboration during reflection enhancing activities help increase professional knowledge of all individuals taking part in collaborative 388

Effect of a collaborative self-evaluation

activity. This is further emphasized by Danielson (2009), who indicates that “reflection is a skill that is best fostered with colleagues” (p. 3). Working together or collaboration with competent peers is not only expected to lead to enhancing the learning experience (as indicated by Vygotsky’s zone of proximal development) but also creates “a sense of awareness of individuality and uniqueness among participants” (Erginel 2006, p. 26). The question is, however, how and under what conditions will collaboration fulfill the aforementioned benefits. The answer, according to Mercado (2014), is that Teachers who engage in self- and peer-evaluation within a collaborative culture that fosters trust, openness and support will more likely find the motivation to develop professionally. They may be more amenable to the idea of working to improve their practice than under more traditional approaches to teacher evaluation. Furthermore, when professional development initiatives are closely tied to clearly articulated processes for self-evaluation, it can foster self-reliance for ongoing learning. (p. 65)

18.1.3 Effect of using video and video-dialogues on self-reflection Using video during the process of self-reflection is valued positively by a considerable body of research. The assumption is, unlike other tools for providing teachers with insights during self- and collaborative reflection during post-teaching evaluation, videos help teachers see or watch themselves in action. They allow teachers, peers, and mentors to re-live the teaching experience, making it easy to note all features pertaining to this experience. This assumption is supported by Orlova (2009), who stresses the positive role played by videos in enhancing student-teachers’ reflection by allowing teachers to watch (as many times as they wish), notice, and respond to points of strength and weakness in their teaching. She also highlights videos’ motivating effect. Videos were also regarded as an important means for teachers to produce evidence-based externalizations of their thoughts on their own teaching competence, in terms of deciding pedagogical contents and teaching activities, addressing learning diversity and classroom interaction, managing learning resources and classroom environment, and selecting assessment methods and pedagogical strategies. (Kong 2010, p. 1772) In a useful survey of literature on benefits of using video in teacher education, Blomberg et al. (2013) highlights the value of videos in a wide array of issues related to developing teachers’ teaching capabilities. Videos were specifically commended as a teaching tool; since they enabled teachers to watch their and others’ teaching any number of times, they provided them with the opportunity to observe their classes in small chunks, thus allowing for reflection on a lesson from different stand-points. Using videos provided a “window” that allows teachers to observe and reflect on their teaching without the psychological pressure that goes with the teaching situation (hence leading to better reflection); they were found to increase teachers’ ability to face classroom challenges, especially classroom management, and provide better chances for offering student-teachers with feedback about their teaching as well as enhancing the usage of certain teaching techniques. In a study conducted by Tripp and Rich (2012), aiming at understanding the effect of using video discussion groups on teacher change, teachers reported videos to have helped them to: “(a) focus their analysis, (b) see their teaching from a new perspective, (c) trust the feedback they received, (d) feel accountable to change their 389

Raghda El Essawi

practice, (e) remember to implement changes, and (f) see their progress” (p. 728). However, most important research has shown that using videos could help bridge the gap between theory and practice (Blomberg et al. 2013, p. 94). All of these factors make video-based technology tools like web-enabled video systems, video annotation tools, or even a video-camera on a tripod gain more importance in the process of enhancing student-teacher or in-service teacher reflection. One of the newly emergent video-based technology tools that could be used in teacher education is Vialogues or videodialogues. Google describes Vialogues as follows: “[It] is a video-based learning tool that promotes meaningful discussions with its asynchronous, time-coded commenting system. Vialogues is a video with a dialogue. To create a Vialogue, you can either upload an original video, embed one from YouTube, or use an existing video on Vialogues” (Agarwala et al. 2012, p. 629). Hence Vialogues is a tool created to facilitate group discussion of a specific video, since it allows users to add comments to a video they are watching, then share video in addition to comments with others initiating discussions. Unlike YouTube (and other videos and dialogues technologies), Vialogues is “specifically designed as a pedagogically adaptive environment for conducting collaborative conversations around video” (Agarwala et al 2012, p. 629). The fact that this tool is basically designed to be used in education (so far) is likely to take away the casual environment that students could associate with tools like YouTube. As a tool Vialogues facilitates (Agarwala et al. 2012): • Sharing video with any number of individuals. However, only the creator of Vialogues provides access to Vialogues through invitation. The creator of Vialogues could also choose to make it public. • Choosing from more than one level of participation, since the Vialogue creator can choose to allow individuals with whom video is shared to comment or just watch and read comments made by the Vialogues’ creator. • Mutually referencing portions of the video with comments made on such portions thanks to the fact that all comments are time-coded to a specific point on the video (thus making possible commenting on and/ or asking questions about specific incidents, events, even a specific moment on a video). • Saving or printing on PDF file questions and/or comments for the Vialogues’ creator as well as responses received. • Threaded discussions. All of this makes Vialogues or video-dialogues an extremely useful tool for a whole set of purposes. To begin with, it could be used in distance learning, whether learners may be student-teachers in teacher-education programs, in-service teachers in professional development programs, or even foreign language learners in AFL programs. In case of AFL learners, videos of Arabic language sessions could be uploaded on Vialogues, giving students and teachers the chance to exchange dialogues on such classes. As for pre- or in-service teachers, Vialogues could be created for self- and collaborative reflection (as is the case in this study) and/or for demonstrating certain class procedures for teacher development purposes.

18.2  Research questions The study reported here looks into the effect of a collaborative activity that manipulates a video-based web 2.0 social technology tool, namely Vialogues, on enhancing a level of 390

Effect of a collaborative self-evaluation

self-reflection that leads to relating theory to practice in student-teachers’ teaching self-evaluation papers. The study addresses the following questions: Q1: What are the dominant levels of reflection mirrored in post-video-dialogues, self-evaluation texts produced by student-teachers? Q2: What are the indications of the dominance of detected levels of reflection regarding participants’ progress toward tying theory to practice?

18.3 Methods 18.3.1  Study setting This study took place in a TAFL (Teaching Arabic as a Foreign Language) MA program at a higher education institution. Students enrolled in the program are from varying academic and career backgrounds. Although some have extensive teaching experience, others have none at all. None of the participants was specialized in Arabic language and linguistics. Two majored in English language and literature, two in commerce, and one in libraries and information. The targeted activity was conducted as part of an Arabic as a foreign language methodology course, where students were exposed to theories of teaching a second/foreign language while applying such theories to the teaching of AFL classes. As part of this activity student-teacher participants were expected to: 1 Observe for two weeks an Arabic as a foreign language class (AFL), noting pedagogical procedures used which students have read about in the course 2 Teach the class they observed for at least 45 minutes while video-taping their demo-teaching 3 Create a Vialogue where they place their comments on their teaching and share with designated peers (support group) to start a discussion or a video-dialogue about the teachingdemo session 4 Write a teaching self-evaluation paper where they highlight points of strength and weakness in their teaching and means of addressing them. The goal of the teaching evaluation paper is to encourage student-teacher participants to go through a process of systematic analysis of their teaching in order to increase student-teachers’ consciousness of why they make their pedagogical choices and the effectiveness of such choices in light of theoretical frameworks they were exposed to. In order to create an environment where active evaluation, effective communication, and supportive interaction take place during video-dialogues, student-teachers were divided into support groups. These groups were expected to work together during the process of reflection in order to help each other highlight evidence-based evaluation of their teaching in their self-reflection papers. It was hypothesized that this framework would encourage student-teachers to collaboratively work on fulfilling a process of externalizing their thoughts about their teaching while evaluating their performance in light of theoretical knowledge received, thus helping studentteachers relate theory and practice through helping them “apply abstract knowledge to concrete phenomena” (Zottmann et al. 2013, p. 2100). This process of application and the resulting teaching conceptualization (in case of positive teaching practices) and/or reconceptualization (in case of teaching practices in need of change) will stand as evidence of teachers’ capability of reaching a level of reflection that allows “drawing upon professional knowledge they have 391

Raghda El Essawi

acquired during teacher education” to justify pedagogical choices (Zottmann et al 2013, p. 2100). In other words, students would have reached a level of reflection that is “directed at an interpretive understanding of the meaning of educational experiences and choices of action” (Lee 2005, p. 703). Being unable to fulfill the framework, on the other hand, would indicate lower levels of reflection that are limited to describing the teaching event using simple observation, past experience, and/or personal experience as a frame of reference (Lee 2005).

18.3.2  Study participants Five participants (three females and two males) took part in this study. Their age range was 30–50 and age mean was 40. All five participants were native speakers of Arabic who lived in Cairo. Only one of them was a graduate of an Arabic-related major. Participants had variant academic background and years of experience. All five participants were enrolled in a methodology course taught by the researcher. Two had above ten years of experience, one had approximately three years of experience, and two had no previous teaching experience.

18.3.3  Data collection and analysis The primary source of data for this study was student-teachers’ teaching evaluation papers which the participants wrote after they watched their videos and conducted video discussions with their colleagues. Due to the limited number of participants, this study relies on a qualitative framework that aims at describing participants’ reflective behavior as a group and individually. Thus the study looks at data produced by all participants collectively (analyzing what data provided by all participants collectively indicates) as well as each participant individually (as separate cases). Like most studies that attempt to study teachers’ reflections (Kong 2010, Yang 2009, Erginel 2006), this study followed a process of content analysis of students’ teaching evaluation papers. According to Kong (2010): “Content analysis is a common research method for studying recorded human communication in varying forms” (p. 1772). The process of content analysis in this study started by reviewing the mentioned source of data, then segmenting data into semantic units (rather than syntactic ones) in order to study levels of reflection mirrored by each unit. The reason for resorting to semantic segmentation is that self-reflection reports of participants taking part in this study revealed what Srijbos et al. (cited in Kong 2010) referred to as a “tendency to use long compounded sentences to express a mix of complex messages or summations in one single entry” (p. 1776). Semantic units are therefore defined as language dealing with a “discrete theme or idea (unit of meaning)” (Kong 2010, p. 1776). To detect dominant levels of reflection displayed by semantic units noted in student-teachers’ teaching evaluation papers, a framework for measuring student-teachers’ levels of reflection was developed. This was followed by classifying semantic units detected under such levels according to a reflection classification scheme that could be objectively used to measure types of reflection mirrored by detected semantic units. After reviewing frameworks developed by Sparks-Langer et al. (1990), Lee (2005), and Erginel (2006), the researcher set out to develop a framework that is fit for fulfilling goals of the current study and the definition of self-reflection it includes. With this in mind the researcher compared frameworks used by the aforementioned studies with tasks required from study participants as well as their comments in teaching evaluation papers to reach the targeted classification scheme. Like the frameworks in mentioned studies, this scheme attempts to measure the depth of teacher reflection based on the assumption that teachers’ reflective thinking is divided into levels 392

Effect of a collaborative self-evaluation Table 18.1 The four levels of self-reflection Level #

Description of level

Example from analyzed data

Level 1

An evaluative statement with no description or justification. An evaluative statement that includes a description of what took place in class (but no justification). An evaluative statement of a certain teaching event that includes a justification of evaluation, but one that is mainly limited to contextual evidence expressed in simple layman’s language, personal opinion, and/ or past experience. An evaluative statement of a certain teaching event that includes a justification revealing an attempt to use theoretical evidence applying it to teaching context.

I used a good opening.

Level 2 Level 3

Level 4

I made use of a creative way to remember students’ names by hanging [name] tags. [Another weakness] The final pre-writing activity shows that it [would have been] more useful to give more time for the lower level student. [A strong point was that] I tried to use a variety of activities: presentational, practice, application; keeping in mind some levels of BT [Bloom Taxonomy] (remember, understand, apply).

whose simplest form is one where the student-teacher makes an evaluative statement about his/ her teaching with no attempt at justifying such a statement. The second level is one where a unit displays an evaluative judgment in addition to a description of what went on in class. This is done with the assumption that this description is in and of itself a justification. In the third level of reflection, a student-teacher participant attempts to justify what went on in class but only uses evidence from the contextual framework within which the event took place for such a justification. Justification is mainly expressed in simple layman’s language, or it uses personal opinion or past experience to justify evaluation. The highest level (Level 4) is one where conceptualization or reconceptualization of the teaching event takes place based on theory and teaching context (Lee 2005, Sparks-Langer et al. 1990, Erginel 2006). Level 4 reflection allows for abstracting targeted teaching episode through referring it to a theoretical framework that helps the evaluator to go beyond the current teaching episode to setting a framework for his/her teaching process in general. In other words, we could say that student-teacher is theorizing his/her own pedagogical behavior. This framework of reflection levels is represented Table 18.1.

18.4  Results and discussion 18.4.1  Collective results This section introduces the detected levels of reflection mirrored by semantic units of all participants participating in the study in order to detect dominant levels of reflection among participants as a group and the indications of the dominance of such levels. Table 18.2 shows the distribution of the total number of detected semantic units among the various levels of reflection. As indicated in Table 18.2, the total number of units reflected by the process of segmentation carried out as described in all five participants’ self-evaluation papers is 134 units. A quick review of figures in this table reveals that the dominant type of reflection mirrored by these units is Level 3 (38.80%); this is closely followed however by Level 4 (34.32%), while Level 1 and Level 2 units are much smaller in number (11.9% and 11.94%). This reveals that as a group participants in this study rely more on a level of reflection that does not incorporate 393

Raghda El Essawi

newly acquired knowledge in the process of justification of their pedagogical choices, using instead contextual evidence expressed in simple layman’s language as justification. However, the fact that the difference between the percentages of Level 3 and Level 4 units is only 4.48% suggests that Level 4 reflection is also reasonably represented in group results.

18.4.2  Results of individual participants Upon breaking down the mentioned collective results by participant, a different picture from the one described earlier is noted. A look at participants individually shows a great variation in levels of reflection, as indicated by their self-evaluation papers. This variation stands in need of an explanation. In subsequent sections, detected levels of reflection mirrored by semantic units of each participant are introduced in order to detect dominant levels of reflection for that participant and the indications of the dominance of such levels. Table 18.3 displays the results by individual participants

18.4.2.1  Participant 5 Participant 5 stands out as the one who, compared to all other participants, relies most heavily on Level 4 reflection (73%). Level 3 reflection follows, but in a much smaller percentage (13.33%). The same is true of Level 2 (6.66%). As for Level 1 reflections, they are totally absent. A review of the participant’s semantic units reveals that s/he makes references to comments received from his/her support group. It would seem that such comments have acted as an instigator for a higher level of reflection. For example in evaluating one of his/her activities s/he mentions: “My colleague . . . commented on this point that . . .” This is followed by his/ her colleague’s comments and the new line of reflection such a comment has instigated in Participant 5’s mind: “I should have asked myself. . . .” It is clear here that Participant 5 is reviewing his/her entire line of thought that led to the problematic event being evaluated. S/he then finds support to his/her new line of thought in abstract knowledge s/he was exposed to in the course: “And that [new line of thought] is supported by our reading. . . .” Based on this, Participant 5 decides on a change that should take place to rectify the problem: “accordingly, I have to change the idea of the third activity, maybe it could be: . . . .” Although the new Table 18.2  Total Units

Level 1 units

Level 2 units

Level 3 units

Level 4 units

134–100%

15–11.19%

16–11.94%

52–38.80%

46–34.32%

Table 18.3  Participant

Participant 1 Participant 2 Participant 3 Participant 4 Participant 5

Total Units

41 34 24 20 15

Level 1 Units

Level 2 Units

Level 3 Units

Level 4 Units

Units

%

Units

%

Units

%

Units

%

2 6 4 5 0

4.87 17.64 16.66 25 0

7 7 0 1 1

17 20.58 0 5 6.66

16 10 12 12 2

39 29.41 50 60 13.33

16 11 6 2 11

39.02 32.35 25 10 73.33

394

Effect of a collaborative self-evaluation

suggestion addresses the problem in that specific lesson/context, it is obvious through the student-teacher’s line of thought and justification of evaluation of teaching episode that s/he is developing a new awareness of the problem highlighted by peers. Such an awareness benefits from abstract knowledge and leads to reconceptualization of entire episode as well as pedagogical choices s/he made in it. This in turn results in a new pedagogical choice indicating a new vision about future change in practices. This process, according to Danielson (2009), stands as an example of deliberate thinking where teacher consciously attempts to “seek more information than immediate context”, especially theoretical information or abstract knowledge, to fully understand the situation. Deliberate thinking, as Danielson predicts, led to dialectical thinking where a teacher re-conceptualizes his/her teaching, which in turn leads to a more successful decision about change in teacher’s pedagogical decisions. Similarly, in an attempt to review strong points, Participant 5 explains his/her evaluative statement that students were actively involved in session by indicating s/he has given learners “the opportunity to think and to talk” through group and pair work (description of what went on in class), thus increasing “their motivation to learn” through a “student-centered approach” where the teacher “only performed a guiding role” (justification that relies on abstract knowledge). This again reveals how Participant 5 used Level 4 reflection by relying on theoretical evidence that helps create a framework to justify his/her current evaluative statement regarding student involvement. In the meantime, it is obvious that this would also serve as a guiding framework that helps the teacher make future judgments about pedagogical choices.

18.4.2.2  Participant 1 Participant 1 comes second in the percentage of semantic units revealing Level 4 reflection, as indicated by his/her teaching evaluation paper. There is a considerable difference, however, between Participant 1 and Participant 5 in the percentage of units showing Level 4 reflection, as indicated by Table 18.3 (the former being 39% while the latter is 73.3%). It is also important to note that the percentage of Level 4 reflection is the same as Level 3 reflection, while Level 2 and Level1 are much smaller (17.7% and 4.87%, respectively). A review of this participant’s semantic units reveals that s/he ties between theory and practice in Level 4 reflection through expressions that allude to abstract knowledge student-teachers have been exposed to rather than a clear discussion of theory, like Participant 5. For example, Participant 1 uses expressions like open versus close-ended question (alluding to types of questions), means of eliciting information, types of feedback, sequencing of activities (specifically when and how to incorporate grammar when teaching the various skills), etc. Such allusions are embedded within semantic units that justify studentteacher negative or positive evaluation of teaching event. Examples of such justifications are, as follows: • “[One of the questions I should have asked myself while planning is:] What type of questions should I use more to encourage student to express [their ideas]? This question was essential because while checking students’ comprehension, I used close-ended questions which made students [reply using] Yes or No only without encouraging them to speak by open-ended questions.” • “[While planning, I] Reminded myself that I [will] always use the most useful feedback techniques . . . [hence] elicitation, metalinguistic feedback and clarification requests [were] used.” 395

Raghda El Essawi

Both examples reveal an attempt at using theoretical information to justify evaluation of teaching procedures used, or what Danielson (2009) refers to as deliberate thinking. Thus, we can say that both show an attempt at theorizing teaching through theoretical justification which not only highlights a problem or strong point in a teaching event but also suggests that the mentioned problem or strong point will be taken into consideration in future pedagogical decisions (in the future, participant needs to use open-ended questions, and participant will continue to use a variety of feedback strategies). It is important however to note that Level 4 and Level 3 reflection units are equally represented in the participant’s self-evaluation paper. This means that the number of semantic units where justification of self-evaluation is mainly dependent on evidence extracted from a specific teaching episode (contextual evidence) using simple layman’s language are equal to units that reveal an attempt at the aforementioned process of theorizing teaching. The following are examples of Level 3 reflection: • “[Weak points of my teaching include:] Giving students 10 min. to read the article was too long as most students finished quickly. I should give them 3 or 5 min. only.” • “[Strong points of my teaching include:] Making clear in the very beginning that the students will play a game by the end of the lesson, which, in my opinion, made them attentive and enthusiastic.” In these examples, justification of evaluation of a teaching event is simply a description of what took place in class (contextual evidence) expressed in simple layman’s language. In the first example, justification of the problem reported in a teaching event (students finished too fast) is limited to a description of time given to students (“10 min.”), and suggestion for change is limited to what should have taken place in this specific incident (giving students only 3 or 5 minutes). Similarly, in the second example, justification of the evaluative statement (students were attentive and enthusiastic) is limited to a targeted teaching event (making clear that students would play a game) and is based on personal opinion. Both the strong and weak point discussed include no indication that suggested change in pedagogical behavior will affect future teaching. As for Level 2 and Level 1 reflections, they are basically evaluative statements with no attempt at justification. Examples of Level 2 reflection are: • “[Strong points of my teaching include:] I made use of a creative way to remember students’ names by hanging [name] tags.” • “[Strong points of my teaching include:] I successfully encouraged students to participate by giving them prizes.” It is hard to tell from these statements whether Participant 1 regards positive/negative evaluation of pedagogical choices (name tags, prizes) as extendable to future teaching events or only limited to the targeted teaching context. Thus, at best, it is expected that the participant will continue to use the mentioned pedagogical choices rather than detecting others that would lead to the same effect (reducing learners’ affective filter). The same is true of statements indicating Level 1 reflection, where evaluative statements stand unsupported. Examples of these are, as follows: • “I opened the class in a way that strengthened the rapport with my students.” • “I used some encouraging words, like: ‫ هايل‬، ً ‫حسنا‬. ” 396

Effect of a collaborative self-evaluation

18.4.2.3  Participant 2 Although the percentage of Participant 2 for Level 4 reflection is less than that of Participant 1 (32.35%), it is the predominant level of reflection in his/her self-evaluation paper. It is worth noting, however, that the difference between the percentage of units revealing Level 4 reflection (32.35%) and those revealing Level 3 (29.41%) is not big. The same is true of Level 3 and Level 2 (20.58%) as well as Level 2 and Level 1 (17.64%) reflections. Like Participant 1, Participant 2 made references to theoretical principles by using terms related to theoretical frameworks to which the participant was exposed in class. It was interesting to note that most the Level 4 reflection units by Participant 2 were ones that addressed strong points of the participant’s teaching. For example, in explaining teaching strong points, Participant 2 states: • “My activities were varied . . . activities covered different skills (listening and speaking), and activities covered different levels of BT [Bloom’s Taxonomy] (remember, understand, reproduce).” • “Sequencing of activities moved from pre-listening, listening, post listening and also from the simplest (matching words) to complex (produce information about famous figures using activated vocabulary).” In these examples, Participant 2 highlights certain pedagogical principles that are related not only to successful teaching in the teaching context being evaluated but also in any teaching context the participant is likely to be involved in. However, when highlighting problems in his/her teaching, Participant 2 tended toward Level 3 reflection. In other words, the participant’s justification of his/her self-evaluation relied upon contextual evidence expressed in simple layman’s language (i.e., justification concentrates mainly on a teaching event or teaching context), as indicated by the following examples: “I should not ask them to take notes at this early level because they are still not very good in writing and unable to concentrate on more than one skill at the same time. I should have given them [multiple] choices.” In this example, Participant 2 explains the problem on the bases of difficulty learners are facing in this specific context (they are still not very good in writing and unable to concentrate on more than one skill at the same time). Thus, recommendation is limited to current teaching context (multiple choice). Participant 2, however, did not relate the problem in a targeted teaching episode to a theoretical framework that highlights the negative effect of compounded high risk activities (listening to new authentic material while writing using a newly learned orthography) on beginners’ performance and the affective filter. Thus it is likely that s/he will not consider the effect of high risk versus low risk activities, as indicated by Richards and Lockhart (2000), when dealing with other students in other contexts. This variation between reflection levels used for justifying strong points (Level 4) and those used for weak points (Level 3 or less) could be one that attests to what Zottmann et al. (2013) refer to as the complexities of the classroom situation, which make it difficult for teachers to relate theory to practice. These complexities are expected to be more obvious when dealing with teaching episodes that represent problematic situations a teacher faces, as is the case here. This, in turn, explains why Participant 2 resorted to levels of reflection that do not require using theory to justify practice. 397

Raghda El Essawi

18.4.2.4  Participant 3 Unlike the previous participants, Participant 3’s teaching evaluation shows that the predominant level of reflection is Level 3 (50%), while Level 4 represents a considerably smaller percentage (25%). It is also interesting to note that though Level 2 reflection is non-existent, the percentage of Level 1 reflection (16.66%) is relatively close to Level 4 reflection (25%). This means that a clear majority of this participant’s semantic units represent evaluative statements whose justification is mainly related to a current teaching context. This in turn means that such statements do not include attempts at theorizing the participant’s teaching and his/her future pedagogical decisions. It also means that the percentage of semantic units showing an attempt at relating theory to practice is close to semantic units that show no attempt at justifying teaching evaluative statements made. Like Participant 2, this participant’s Level 4 reflection is only represented when working with positive aspects of his/her teaching, as in the following examples: • “Sequencing [of activities reflected] moving from pre-reading and vocabulary presentation to reading for comprehension, and then to analyzing grammar, and finally to free production.” • “Relying on group and pair work in most activities to motivate students and establish an interactive class environment.” On the other hand, Participant 3 relies more heavily on Level 3 reflection to justify his/her evaluation of problematic teaching events. For example, in discussing negative points in his/ her teaching, Participant 3 states: • “[During planning I should have asked myself]: how will I deal with different students’ ability levels in the class? This was very important as some students had some times nothing to do except waiting for the others to finish their work.” • “[I had to cancel part of the] vocabulary presentation activity, as students’ awareness of vocabulary was higher than I expected.” This could be attributed to the same reasons as presented in the case of Participant 2.

18.4.2.5  Participant 4 Like Participant 3, the predominant level of reflection for Participant 4 is Level 3 (60%). Examples of Level 3 reflection are: • “The final pre-writing activity shows that it was more useful to give more time for the lower level student.” • “I think more practice on connectors would have been useful especially for the weak student.” In addition, like Participant 3, the second most clearly represented level of reflection in the case of Participant 4 is Level 1 (25%), while Level 4 reflection is the last (10%). In fact, compared to all other participants, Participant 4’s work reveals the highest percentage of Level 1 reflection and the lowest Level 4 reflection. In other words, s/he had the highest percentage of 398

Effect of a collaborative self-evaluation

units where an evaluative judgment is passed with no justification, whether theoretical or contextual. The Examples of Level 1 reflection are: • “My lesson preparation was sufficient.” • “There are varieties of activities.” • “[If I were to teach this lesson again] I would give more time for the writing practice.” while an example of Level 4 reflection is: “Sequencing of activities used showed that there was a pre- and post-reading activity and a pre-writing activity, while the writing is at home.” Also, unlike other participants, the difference between the percentage of Level 3 (60%) and that of Level 4 (10%) reflection is quite big. The difference between Level 1 (second most clearly represented level of reflection – 25%) and Level 3 (most prominent form of reflection – 60%) for this participant is also big. These numbers indicate that there is a high possibility that student-teacher’s theoretical knowledge remains “inert”, as indicated by Zottmann et al. (2009), leading to an unwanted gap between theory/abstract knowledge and practice. To sum up, the results of the five participants, when considered individually, seem to suggest that the researched framework of activities and tools did not equally lead the participants to use the levels of reflection that encourage relating theory to practice. In fact, the results reveal that the study participants are set along a continuum, in which some are closer than others to the targeted process of tying theory to practice or what Lee (2005) calls using theory to inform practice, as represented in Figure 18.1. However, the question remains: why some are closer than others to the targeted goals? It is important to note here that variation in participants’ demographic factors (age, years of experience, and academic background) makes it difficult to consider any of those a reliable explanation of discrepancy in levels of reflection. It is difficult to say, for example, that high level of reliance on Level 4 reflection by Participant 5 could be attributed or related to his/her age or years of experience or academic background, since each of these is shared by at least one other participant whose performance regarding levels of reflection is different from that of Participant 5. One explanation of the mentioned behavior is presented by Danielson (2009), who highlights problems caused by resorting to modes of thinking leading to lower levels of reflection when carrying out tasks that require a high level of conscious analysis and data seeking, like evaluating teaching practices. Danielson (2009) presents Grimmett’s modes of thinking from

Participant 3

Participant 1

Theory supported evaluative statements Participant 4

Participant 2

Participant 5

Figure 18.1  Continuum of usage of theory supported evaluative statements

399

Raghda El Essawi

those requiring the lowest to those requiring the highest level of reflection; these are: formulaic, situational, deliberate, and dialectical. Formulaic thinking refers to a mode of thinking that is fit for routine decisions or ones that do not require much reflection. This includes things like “how a classroom teacher takes attendance, transitions students from subject to subject, implements emergency drills, and so on” (p. 2). As for situational thinking, it “only focuses on information embedded in a specific context at a specific time” (as we saw in examples of Level 3 reflection mentioned earlier). On the other hand, in the case of deliberate thinking, the teacher consciously attempts to “seek more information than immediate context”, especially theoretical information, to fully understand the situation. Deliberate thinking is expected to lead to dialectical thinking where the teacher re-conceptualizes his/her teaching (as in the examples of Level 4 reflection presented earlier). Danielson explains that teachers must “adjust their thinking to accommodate the level of reflection a situation calls for” (Danielson 2009, p. 1). Thus, it is possible to consider reliance on lower levels of reflection (Level 1, Level 2, and Level 3) during self-evaluation (a task requiring a high level of conscious analysis) as failure to use the appropriate mode of thinking. In other words, it could be said that for the targeted task to be effective, student-teachers must be made aware of their modes of thinking and to consciously move to modes that lead to Level 4 reflection. This in turn reflects a need to purposely foster deliberate and dialectical modes of thinking – as opposed to formulaic and contextual thinking – during the process of self-evaluation. Danielson (2009) points out that deliberate and dialectical modes of thinking are in fact “habits of mind” that could (and in the case of some participants in this study need to) be consciously fostered. Though collaborative reflection via video-dialogues did lead some student-teachers toward such modes, awareness of and actively learning about modes of thinking is expected to encourage all participants to use modes of thinking that lead to relating theory to practice.

18.5 Conclusion This chapter attempted to detect the effect of a collaborative activity that manipulated a video discussion tool, namely “Vialogues”, in enhancing student-teachers’ ability to tie theory to practice, leading to higher levels of reflection required by self-evaluation. Despite research indicating the positive effect of collaborative reflection and video-based tools in helping teachers reach higher levels of reflection and relating theory to practice, the results of the current study indicated that activity relying on collaborative reflection and use of video-based tools was only partially successful in fulfilling the targeted goal. The results here are similar to those reached by Kong (2010) regarding the effect of a web-enabled video system on student-teachers’ reflection. In Kong’s study, the results indicated that “student – teachers are capable of making reflective notes that describe and justify their general teaching behavior, but are unable to discern, associate with and then articulate abstract teaching rationales beyond specific teaching behavior in every key aspect of teaching” (p. 1781). Both studies (Kong’s and the current one) therefore suggest that a process of training through professional mentoring could have a positive effect on student-teachers’ ability to use levels of reflection that encourage relating theory to practice. It must be mentioned here, however, that two factors could have negatively affected the reliability of the results of this study. The first is the limited number of participants. The second is the fact that the study covered students’ self-evaluation of only one demo-teaching. Therefore, further studies need to be conducted to validate the results reached here. Furthermore, as Kong (2010) suggests, research also needs to be conducted to test the effectiveness of the process of training on fostering modes of thinking conducive to tying theory to practice. These 400

Effect of a collaborative self-evaluation

activities could include demonstrating the various levels of reflection to student-teachers through presenting examples that show the appropriate level of reflection for a targeted task. Another suggestion could be teaching student-teachers in support groups to “pose questions that lead their colleagues to ask productive questions themselves, to consider other sources of information that might provide additional insight, and to generate their own possible solutions” (Danielson 2009, p. 4). One final suggestion could be the presence of a more active role for the course instructor or any experienced mentor during the process of self-reflection to help provide necessary prompts that would gauge student-teachers’ thinking from formulaic or contextual to deliberate and dialectical.

References Agarwala, M., et al, 2012. Vialogues: Videos and dialogues based social learning environment. In: Proceedings of the 12th IEEE International Conference on Advanced Learning Technologies, ICALT 2012. (pp. 629–633). [6268195] 10.1109/ICALT.2012.127. Blomberg, G., et al, 2013. Five research-based heuristics for using video in pre-service teacher education. Journal for Educational Research Online, 5 (1), 90–114. Danielson, L. M., 2009. Fostering reflection. Educational Leadership [online], 66 (5). Available from: www.ascd.org/publications/educational-leadership/feb09/vol66/num05/Fostering-Reflection.aspx Erginel, S. S., 2006. Developing reflective teachers: A study on perception and improvement of reflection in pre-service teacher education. Thesis (PhD). The Graduate School of Social Sciences of Middle East Technical University. Kong, S. C., 2010. Using a web-enabled video system to support student – teachers’ self-reflection in teaching practice. Computers in Human Behavior, 55, 1772–1782. Lee, H. J., 2005. Understanding and assessing preservice teachers’ reflective thinking. Teaching and Teacher Education, 21, 699–715. Mercado, L. A. and Baecher, L., 2014. Video-based self-observation as a component of developmental teacher education. Global Education Review, 1 (3), 63.77. Mirzaei, F., et al., H., 2013. Measuring teachers’ reflective thinking skills. Procedia – Social and Behavioral Sciences. Available from: http://webapps.sps.utm.my/gsgV10/uploadFullArticle/PP103156_ Paper%20ID%202097.pdf Orlova, N. Video recording as a stimulus for reflection in pre-service EFL teacher training. English Teaching Forum. Available from: https://americanenglish.state.gov/files/ae/resource_files/09-47-2-f. pdf Richards, J. and Lockhart, C., 2000. Reflective teaching in second language classrooms. Cambridge: Cambridge University Press. Sparks-Langer, G., et al., 1990. Reflective pedagogical thinking: how can we promote it and measure it. Journal of Teacher Education, 41 (4), 23–32. Tripp, T. R. and Rich, P. J., 2012. The influence of video analysis on the process of teacher change. Teaching and Teacher Education, 28 (2012) 728–739. Vygotsky, L. S., 1978. Interaction between learning and development. In: M. Cole, V. John-Steiner, S. Scribner, and E. Souberman, eds. Mind and society: The development of higher psychological processes. Cambridge, MA: Harvard University Press, 79–91. Yang, S., 2009. Using blogs to enhance critical reflection and community of practice. Educational Technology and Society,12 (2), 11–21. Zottmann, J. M., et al., 2013. Computer-supported collaborative learning with digital video cases in teacher education: The impact of teaching experience on knowledge convergence. Computers in Human Behavior, 29, 2100–2108.

401

19 ARABIC LANGUAGE TEACHING IN THE U.S. Two Arabic language users’ views on culture and self-positioning as teachers Brahim Oulbeid This ethnographic case study builds on social identification theory (Hogg and Abrams 1998) and Tajfel’s (1978) concepts of social categorization, self-awareness, and social comparison to describe two non-native speaker (NNS) Arabic college-level teachers’ experiences in the Northeast of the United States. Specifically, this chapter explores, first, these teachers’ beliefs about culture and how those beliefs affect their classroom practices. Secondly, this study investigates how these two teachers negotiate and construct their identities as teachers of Arabic language and culture professionals and how they categorize themselves as NNS teachers. Data collection sources included classroom observations, semi-structured interviews and teaching artifacts. The analysis of the data shows how these teachers navigate their identities as learners of Arabic language and cultures, as student-teachers, and as classroom practitioners. The focus is on teacher identity and the connections between their beliefs about Arabic cultures and their implementation of these cultures in the classroom. The study findings demonstrate that these teachers still internalize the ideology of “Native-speakerism” (Holliday 2006) and position themselves as novice teachers in need of support to build selfconfidence. The data indicate that they understand culture as inherent in language, and at the same time, they acknowledge their limited linguistic and cultural knowledge. Additionally, both participants displayed multiple and simultaneous identities, specifically “non-native speaker, non-native interpreter of culture, teacher, and learner.” I argue that NNS college teachers of Arabic value their group membership as NNS practitioners despite the challenges of juggling the teaching loads within institutional contexts where they are subject to anxiety because of lack of support and mentorship. Implications for research point to the need for Arabic language programs to revisit the language-culture dichotomy and provide preservice teachers with strategies to cope with the challenges of their classrooms and the realities of their institutional settings.

19.1 Introduction This chapter examines the perceptions of two NNS teachers of Arabic and their understanding of culture in the language classroom. This study focuses on two NNS teachers of Arabic at the college level to explore, first, Arabic second language speaker teachers’ beliefs about culture 402

Arabic language teaching in the U.S.

and how they affect their classroom practice. Second, it investigates how these two teachers negotiate and construct their identities as teachers of Arabic language and culture professionals. This ethnographic case study takes place in the context of the recent interest of many young Americans in learning languages that are deemed “critical” to U.S. economic and political interests, especially after the events of 9/11. Arabic and 12 other languages have been deemed “critical-need” languages. The participants in this study are part of the wave of students who have benefited from the funds that served to stimulate U.S. students’ interests to learn these languages (Welles 2004). Competence in these languages promotes the development of intercultural communication and understanding as well as global competition. For both native and non-native language teachers, language and culture are traditionally not taught in tandem; rather, teachers focus on teaching linguistic skills more than cultural content. Culture usually takes the form of short factual vignettes in the textbook and is taught as a small and separate part of a lesson. The general practice in world language classrooms is that most language teachers recognize the significance of integrating culture in the teaching of language, but culture has always been put aside, and therefore teachers prioritize teaching linguistic forms and functional aspects of language (Byram and Risager 1999, Sercu 2006, Galeano 2014). On one hand, this tendency underlies issues connected to teachers’ abilities to teach culture. On the other hand, it questions language teacher education experiences that may not stress the significance of culture learning and/or may not provide concrete pedagogical strategies to teaching culture. Such practices consequently affect learners’ abilities and readiness for participation in intercultural encounters and the development of intercultural communication. Research on NNS teachers did not begin until the 1990s (Medgyes 1994, Braine 1999, Llurda 2005). Most of this research was on NNS teachers of English and explored the differences between native speaker (NS) teachers and NNS teachers as well as differences in teaching language and culture. While there are a few studies about the status and profiles of Arabic language teachers (Belnap 2007, Abdalla and Al-Batal 2011), there is no research addressing the status and characteristics of NNS teachers of Arabic in the United States. The present study investigates how two NNS teachers of Arabic develop their multiple identities, drawing on their sociohistories and the experiences that led them to become college Arabic teachers.

19.1.1  Literature review This section reviews the literature on language and culture in the world language classroom, NNS teachers’ characteristics, and NNS teacher identity.

19.1.1.1  Language and culture in the world language classroom This discussion examines teachers’ views of language and culture as well as the reality of their classroom practices. Three main themes are identified in the literature: definition of culture in the literature, teachers’ conceptions of culture teaching and learning, and the constraints hindering the integration of culture in the classrooms. Kramsch (2006) offers an understanding of culture in language study from two perspectives: the modernist and the postmodernist perspective. First, the modernist approach to language focuses on accuracy and describes culture as knowledge of literature and arts (Holliday 1999). This view regards culture as knowledge of facts about a social group through observation of the group. Culture is then seen as static, homogeneous, and invariable, and the learner as a distanced observer. The role of language in this case was to serve as a mere tool for 403

Brahim Oulbeid

accessing art and literature. Second, the postmodern view defines culture as “Discourse, identity, and power” (Kramsch 2006, p. 326). Culture is also thoughts, feelings, and language use (Kramsch 2006). This approach established the intercultural orientation to language learning, which aims to avoid essentializing cultures. Valdiviezo and Nieto (2015) consider cultures to be context specific and “multifaceted, heterogeneous, and eclectic” (p. 93, emphasis in the original). This perspective recognizes the fluid and dynamic aspect of culture as it adds reflection on self and the other, and membership in a discourse community (Kramsch 2014). Adding reflection highlights the necessity of responding to the different needs of learners if teachers are aware of the intercultural component of language learning. The literature reviewed indicates that world language teachers still lack a deep understanding of culture (Byram and Risager 1999). For this reason, culture has always been “essentialized” and taught from a modernist perspective. Consequently, teachers prioritize linguistic skills over cultural practices and perspectives in their classrooms. The research suggests that teachers do not seem to have a common definition of culture (Byram and Risager 1999, Schulz 2007). They find it challenging to define the construct of “culture” and frequently resort to the cultural content in their textbook units. In their study of language teachers’ views of culture in Denmark and England, Byram and Risager (1999) report that teachers could not define the concept “culture.” However, the authors indicate that some teachers have personal understandings that are related to the cultural themes that they teach. Klein (2004) conducted a study on how teachers of French, German, and Spanish in the U.S. understand their work as teachers of language and culture. This study supported Byram and Risager’s (1999) results, since it revealed that teachers acknowledge the challenge of capturing the concept of “culture”: some teachers prefer to generalize that culture is “everything” and “all embracing.” Banks (2002) and Sleeter and Grant (2002), cited in Phipps and Guilherme 2004, p. 50, indicate that other teachers understand culture as “Food, Fashion, Festivals, and Folklore.” This approach to language teaching and learning focuses on the transmission of cultural facts and results in the essentialization of the target culture (Sercu 2006), which is an obstacle for deeper understandings of other cultures. The narrow view of culture can be visible when teachers lay out objectives that foster the implementation of culture, but their teaching derails from the planned objectives. For example, some teachers have the potential and beliefs to teach culture, but in practice, it is not there. Kearney (2008) conducted a study on seven French language teachers and their engagement in teaching culture. All the teachers supported the inseparable nature of language and culture. Nevertheless, classroom practices are not determined by teachers’ views of culture and their lesson objectives but rather by their lack of common knowledge about culture, lack of time, and some teacher characteristics. These findings demand consideration of the research on teacher cognition and the divide that exists between “what teachers think, know, and believe and the relationships of these mental constructs to what teachers do in the classroom” (Borg 2003, p. 81). The disconnect between teacher knowledge and beliefs and their classroom practices is evident in world language and culture teaching due to the adoption of the communicative approach in recent decades. The communicative approach to language teaching (Brumfit and Johnson 1979) emphasized teaching linguistic skills in the world language classroom. However, because of globalization, the need to prepare globally and interculturally aware citizens for the challenges of the twenty-first century is imperative (The National Standards in Foreign Language Education Project 2006). The National Standards in Foreign Language Education Project (2006) presents culture in terms of products, practices, and perspectives; a social group’s artifacts, attitudes and values, and world views. Additionally, the report by the Modern Language Association 404

Arabic language teaching in the U.S.

(MLA) Ad Hoc Committee on Foreign Language Learning (2007), titled “Foreign Languages and Higher Education: New Structures for a Changed World,” criticizes the notion of culture as facts and sets a new goal: developing translingual and transcultural competence. Transcultural competence emphasizes the social construction of language learning and the importance of dialogue in the development of “intercultural awareness and understanding” (Mijalski 2014, p. 37).

19.1.1.2  Constraints to teaching culture In the literature, curriculum mandates and teacher characteristics seem to be the primary reasons for teachers’ inability to implement a postmodernist approach to language teaching and learning (Byram and Risager 1999, Borg 2006, Galeano 2014, Peiser and Jones 2014, Ryan and Sercu 2003, Sercu 2007). First, curriculum expectations make it difficult for teachers to devote time to culture. School missions and curriculum objectives place the four linguistic skills (speaking, listening, reading, and writing) as the most important objectives in language learning. Byram and Risager (1999) report that the purpose of culture in the curriculum is to impart facts and “break down prejudice” (p. 100). However, these authors (Byram and Risager, 1999) contend that teachers feel pressured by curriculum overload and the need to conduct frequent assessments of learners (Sercu 2007). Ryan and Sercu (2003) raise a similar issue when they report that 73.33% of the Mexican teachers in their study teach more language than culture because of lack of time. In Sercu’s (2006) study, 79.48% of 424 teachers surveyed agreed that they would like to teach intercultural competence if they were not pressured by time and curricula overload. Second, Borg (2006) describes language teachers’ characteristics as personal traits, attitudes, and knowledge of content and pedagogy. These characteristics influence the teaching and learning of language and culture, which require them to be creative, flexible, and positive (Borg 2006, Peiser and Jones 2014). Additionally, Kohler (2015) indicates that identities also shape teachers’ understandings of the relationship between language and culture, which in turn affect their practices. In this respect, Sercu (2006) argues that the focus on communication as the essence of teachers’ perspectives as learners influences their practices, limiting their implementation of culture (Byram and Risager 1999, Borg 2003, Klein 2004, Galeano 2014). The present study suggests that teachers’ experiences affect their identities in terms of how they view themselves as teachers, as role models for their students, and as interpreters of culture.

19.1.1.3  NNS teachers Cook (1999) uses the term “second language user” to refer to NNS teachers. However, throughout this study, I use the term (NNS) teachers to refer to teachers who are not native learners of Arabic because the participants in this study have used the term NNS teacher to refer to themselves during the interviews. I will use native speaker (NS) to refer to teachers who have learned Arabic as their first language at school. Medgyes (1994) identifies the differences in terms of personal traits, teaching practice, and language ability. The differences in teaching and classroom behavior relate, according to Medgyes, to the differences in language competence. Nevertheless, Medgyes acknowledges that both NS and NNS teachers could equally be effective practitioners in their own ways. The present study supports this claim despite the participants’ adherence to the ideology of “native speaker fallacy,” which entails that NS teachers are more successful than NNS teachers (Phillipson 1992). 405

Brahim Oulbeid

Braine (1999) reviewed research on teachers’ self-perceptions as NNS teachers and how the learners view them. Most of this research has been conducted by NNS teachers and in English as a second language (ESL) or English as a foreign language (EFL) contexts and underscores their empowerment and recognition of themselves as effective teachers (Braine 1999). This research provides a NS teacher’s perspective on NNS Arabic teachers’ self-perceptions and views on identity.

19.1.1.4  NNS teacher identity Language teacher identity construction in second language acquisition (SLA) has gained prominence in the last decade (Block 2007). In this study, I use Norton’s (1997) definition of identity, which she understands as “how people understand their relationship to the world, how that relationship is constructed across time and space, and how people understand their possibilities for the future” (p. 410). This definition highlights the context specific nature of identity construal and the individuals’ relationships with each other. In such relationships, they develop social identities that refer to the individual’s becoming part of a group and how this group membership helps construct his or her identities (Hogg and Abrams 1998). These social identities are developed through relationships with the “larger social world, as mediated through institutions” (Norton 1997, p. 420). Such relationships of belongings create an ingroup versus outgroup membership situation which entails both social comparisons and individual self-awareness, constructs that will be used here to understand teacher identity. Research on language teacher identities addressed the dichotomy of NS and NNS teachers in ESL and EFL fields (Medgyes 1992, Braine 1999, Moussu and Llurda 2008, Reis 2011, Zhang and Jensen 2013). Discussions on identity investigate how ESL and EFL NNS teachers construct their professional identities, how they perceive themselves, and how their learners and colleagues (Llurda 2005) perceive them. The NNS English teachers have become a movement against the seemingly widespread position that considers NS teachers of English as the “best” teachers (Phillipson 1992) at a time when the communicative approach to teaching English was prevalent in the teaching of English for speakers of other languages (TESOL) (Mahboob 2010). As for identity negotiation, NNS teachers’ classroom practices are shaped by their educational achievements, their experiences as learners, their students, their personal interests, and their future aspirations. Block (2007) noted, “[I]dentities are about negotiating new subject positions at the crossroads of the past, present, and future. Individuals are shaped by their sociohistories but they also shape their sociohistories as life goes on” (p. 27). Only a handful of studies on NNS teachers of other languages have been carried out (Ghanem 2015, Kohler 2015, Zhang and Jensen 2013). The present study contributes to the scholarship by investigating how NNS teachers of Arabic perceive themselves as NNS teachers and how they make connections between language and culture in their daily practices. It draws on theories of social identity and a sociocultural approach to language and culture teaching. This study is important for prospective teachers and teacher mentors in Arabic language programs as it provides a description of the contexts where some teachers develop their identities and how these identities position the teachers.

19.1.1.5  Theoretical framework This study draws on social identification theory, as discussed by Hogg and Abrams (1998) based on Tajfel (1978), and sociocultural theory (Vygotsky 1978). Social identification theory relates individual identity to social factors such as class, race, and national origin, categories 406

Arabic language teaching in the U.S.

that may affect their self-perceptions in various ways. For example, in world language education, university hiring committees still highly prefer NS teachers, as exemplified in online postings for jobs in higher education (HigherEdJobs. 2016, Norton 1997). This practice is still the trend in most U.S. foreign language departments, where chairs are usually native speakers, especially for less commonly taught languages such as Arabic. This tendency perpetuates the “native speaker fallacy” (Phillipson 1992), which can alienate NNS teachers, undermining their confidence and abilities to lead and direct Arabic language classrooms and departments. Therefore, social identity theories relate identity to social contexts and therefore offer ways to recognize unequal power relations. I draw on Tajfel’s (1978) concepts of social categorization, self-awareness, and social comparison to describe how NNS Arabic teachers view themselves as members of a specific group and how they compare themselves to NS teachers. The term social means belonging to a group, and categorization pertains to similarities shared with the rest of the group. Self-awareness describes the member’s awareness of belonging to a specific group. Finally, social comparison signifies distancing oneself from other group members because of differences in beliefs and ideas. A second lens for examining identity, language, and culture is the Vygotskian sociocultural theory (Johnson 2006, Reis 2011, Varghese et al. 2005). This study examines NNS Arabic teachers’ beliefs about the relationships between language and culture in the classroom and the reasons for teaching culture. Drawing on Johnson (2006), sociocultural theory considers learning a “dynamic social activity that is situated in physical and social contexts and distributed across persons, tools, and activities” (p. 237). Knowledge is therefore co-constructed through participation in social activities and mediated by culturally developed tools (Lantolf 2000). Language is one of these tools and is used to reveal how NNS teachers construct and develop their identities. In this respect, Reis (2011) notes that sociocultural theory enhances the development of NNS English teachers’ professional identities. This development “involves promoting their awareness of how they position themselves professionally and are positioned by others . . . in regards to their legitimacy and in relation to the contexts where they work and live” (p. 144). Therefore, identity development, which is a dynamic and social activity, relies both on internal and external factors linked to the surrounding environment. Identities are both a product and a process as they shift through contexts (Olsen 2008).

19.2  Research questions The present ethnographic study addresses the following questions: 1 How do NNS teachers of Arabic in the U.S. define culture? 2 What are these teachers’ perspectives regarding teaching Arab culture? What connections do they make between language and culture in the classroom? 3 What identities do they construct in relation to their learners, their Arabic language faculty, and other faculty in their departments? To address these issues, this study provides an examination of NNS teachers’ beliefs about the relationships between languages and cultures in general, an overview of previous research on NNS teachers, and finally how and to what extent being a NNS teacher affects their identities as language teachers. Answers to these questions will contribute to a better understanding of the experiences of NNS Arabic language instructors, their self-perceptions, and the challenges these teachers face in their classrooms and within their institutions. Additionally, studies like 407

Brahim Oulbeid

this may help programs in Arabic studies and Teaching Arabic as a Foreign Language (TAFL) to build programs that are sensitive to the needs of the growing population of preservice Arabic teachers and supportive of their pedagogical and emotional development in the process of becoming professional classroom practitioners. The analysis focuses on the participants’ narratives, through which they build and create several and simultaneous identities depending on the variety of contexts and situations in which they are involved. I argue that NNS teachers of Arabic at the college level are socialized to emphasize language more than culture in the classroom, even though they consider culture important. Because their teaching experiences are limited, these teachers are continuously “worried” about their linguistic and cultural expertise in comparison to NS teachers.

19.3 Methods This ethnographic case study involves the teaching of Arabic language and culture at the college level, and the participants are two NNS teachers of elementary to advanced level Arabic. Ethnography makes it possible to obtain an in-depth understanding of how NNS teachers of Arabic “behave, think, and talk” (Creswell 2012, p. 461) based on collecting and analyzing multiple sources of data. This is also a case study, given the interest in exploring NNS teacher identities as they are shaped by a variety of contexts: “[C]ase study concentrates on experiential knowledge of the case and close attention to the influence of its social, political, and other contexts” (Stake 2008, p. 120).

19.3.1  Settings of the study The first setting of the data collection is a private liberal arts’ college in the U.S. in which about 3000 students are enrolled. This college offers coed and graduate certificate programs. For students of Arabic, the college offers a B.A. in Middle Eastern studies and a minor in Arabic, which are both part of the Middle East studies concentration. The second site for this study is also a private liberal arts’ college in the U.S. At this college, there are approximately 2400 students enrolled. The Arabic language is offered as a course in the Middle Eastern studies department.

19.3.2 Participants The first participant, Adam (pseudonym), is a white U.S. citizen in his early 30s. He has curly blond hair and often dresses in a professional fashion, in contrast to his fellow teachers. Data from the questionnaire show that Adam majored in French as an undergraduate and has had seven years of formal Arabic study in the United States, Morocco, and Syria, where he served as a Fulbright fellow. Adam obtained a master’s degree in Arabic studies. He has been teaching Arabic at the college level for five years and has taught in two Arabic summer programs. In 2015, Adam taught first- and third-year Arabic. In his classes, Adam focuses on the use of both Modern Standard Arabic (MSA) and the Levantine variety. MSA is the formal variety of Arabic and Levantine is the dialect spoken in the Levant. I observed Adam in his first-year Arabic classes. The observations were conducted in the lower level of the mathematics and science building. There were 20 students in the class, which was taught in a traditional university classroom. The second participant, Thomas (pseudonym), is a white American. He is short with a light black beard and dresses casually in class. Thomas is in his mid-30s. Data from the questionnaire he completed indicate that he speaks German and Latin at the novice level. He has 408

Arabic language teaching in the U.S.

studied Arabic for five years and has been teaching for five years. During his graduate studies, he traveled to Egypt to learn Arabic. He has also taught in two Arabic summer programs. In 2015, Thomas taught two first semester Arabic classes with 12 students in each class, which is typical for his college language classes. He has been teaching at this college for the last three years after he obtained his master’s degree in TAFL. The teaching space was a traditional college classroom located in the basement of the main college library.

19.3.3  Data collection For this study, I collected data through: (a) a brief questionnaire, (b) field notes from 17 classroom observations, (c) two semi-structured interviews and an informal discussion with both participants, and (d) teaching artifacts including syllabi, teaching materials, and quizzes. The participants completed the brief questionnaire via email. The questions were all open-ended and addressed the following topics: demographic information, teaching experiences and professional information, and the challenges faced at their respective institutions. The observations took place in the classrooms during the fall semester. The 17 observations lasted for 50 minutes each and focused on the teachers in action. I chose to foreground both interviews and observations because classroom observations provide rich data to discuss and address during interviews. One of the strengths of this study is that it brings beliefs and practices together to understand teaching culture and teacher identity in the Arabic language classroom. The semi-structured interviews focused on the participants’ life histories and on the details of experiences and reflection on the meaning of the experiences (Seidman 1998). I conducted two sets of interviews (see Appendix) which lasted for 1 hour and 58 minutes with Adam and 2 hours 34 minutes with Thomas. The informal discussion with both participants lasted 1 hour and 17 minutes. The first interview is about life history or biography of the participants and the second about their daily routines in and outside of the classroom. The detailed account of the lived experiences of the participants as Arabic language learners, preservice teachers, and Arabic classroom practitioners offers insights into understanding the factors that shape the participants’ understandings of culture and the development of their multiple identities. Additionally, individual interviews offered opportunities to pose follow-up questions and observe reactions of the participants. The informal discussion allowed Adam and Thomas to discuss the challenges they encounter in their daily practices and elaborate on some issues raised during individual interviews. The interviews and the group discussion were conducted in English for purposes of ease of communication, deeper exploration of the issues covered, and confidence in using the native language of the participants. Regarding the language of the interviews, in an email communication, Adam noted “[I] am relatively confident that I would not have been able to discuss the topics we covered with the same precision and nuance had the interviews been conducted in Arabic.” Thomas concurred that “If you [i.e. the author] conducted the interviews in Arabic, we would have been less expressive (since it’s not our native language) and then you would have had to translate everything into English for the final paper.” Additionally, the participants were concerned about ambiguities and bias that might occur in translation. The teaching materials included syllabi, quizzes, and worksheets from the Arabic language courses that Adam and Thomas taught. The purpose of gathering these artifacts was to compare the teaching practices with what the participants had to say about their daily classroom experiences, especially in terms of how they implement culture in their teaching. I transcribed the interviews manually and started with an initial exploratory reading of hard copy transcripts. Then I read the interviews several times to identify emerging themes and 409

Brahim Oulbeid

patterns using “content analysis” (Patton 2002). The categories are directly derived from the data by clustering statements, which inform the research questions and provide meaningful accounts of the participants’ experiences. The analysis of the participants’ oral narratives from interviews helped me connect these narratives with other forms of data to understand the factors that contribute to the development of teacher identities.

19.3.4  Data analysis Horvat (2013, p. 117) notes, “Analysis is the process of finding the meaning in the data.” As such, the type of data gathered mandates the type(s) of analyses to use. The present study examines the experiences of two teachers of Arabic and how they construct their identities over time. As Barkhuizen et al. (2014, p. 5) posit, “Narrative inquiry can help us to understand how language teachers and learners organize their experiences and identities and represent them to themselves and others.” Thus, the context in time and space is important for exploring the life histories and identities of individuals. Making sense of such explorations can be facilitated by narrative inquiry which builds on past, present experiences, and future goals (Barkhuizen et al. 2013). This study builds on analysis of narratives, which emphasizes looking at content and generating thematic descriptions. The purpose of applying an analysis of narratives was to find out how the participants talk about issues of culture learning and teaching and how they construct and negotiate their identities. The script below is a chunk from the first interview, which shows patterns in the participants’ views on culture. Table 19.1 displays the participants’ answers to the question: “What is Table 19.1 Two teachers’ conceptions of culture Adam: “[Culture] is, it’s something that I believe can be modeled . . . , and yet, language is always a product of culture. I do not think that the two can be divorced from one another or at least in my experience. we need to be very upfront with our students that the incredible diversity they are going to encounter in the Arab World if they are fortunate enough to have the opportunity to travel and live there . . . songs and dances and particular cultural tropes from those places are valuable for my students to at least witness, engage with, and talk about . . . . I feel like that’s what it means to teach culture, to give people a wide, very wide presentation and to encourage them to do their own digging.” (11–10–2015)

Thomas: “[Arab culture]. Culture is something that, it is located in a specific point in time and it shifts . . . when I think of culture I try to think about what’s the context of whatever I am engaging in, they [language and culture] go hand in hand. You learn the culture by learning the language and, you know, I don’t think they can be separated . . . that’s where you learn it, from study abroad. You get some exposure in the classroom, from watching movies, listening to some songs and discuss it in English. I think you get some culture. But, when you travel abroad, you’re actually interacting with people in the street one on one or in groups you know, that’s really where you’re experiencing the culture first hand. . . . For me, that’s invaluable because I can relate those you know when I tell that to my students, I am giving them that through my filter, like how I viewed it . . . . They can understand it, like anthropologically; they can take a step back from your own in order to view it as hopefully unbiased as you can.” (11–30–2015)

410

Arabic language teaching in the U.S.

culture for you?” The side-by-side display of the two participants’ statements shows the multiple parallels they share in terms of how they understand the construct of culture and how they implement it in their classrooms.

19.4 Findings The data analyzed yielded insights regarding the NNS teachers of Arabic views of culture, their perspectives toward teaching culture, and their social and professional identities as NNS teachers of Arabic. Interestingly, the two participants view culture differently at first, implement culture differently, but share similar social and professional identities that they have developed through their experiences with the language and culture.

19.4.1  Two distinct views of culture “It [culture]’s something that I believe can be modeled . . . . It has a place and I think it’s worth taking valuable classroom time to offer artifacts, things like videos, things like bringing in objects like a ʕayn ‫[ عين‬eye] or kuufiyya ‫[ كوفية‬Arab head dress for men], Ramadan lantern or something like that . . . my personal belief as an educator is that we need to be very upfront with our students that the incredible diversity they are going to encounter in the Arab World if they are fortunate enough to have the opportunity to travel and live there and experience it themselves is that we cannot fully present the richness and variation of the culture that exists.” [Adam’s views of culture]

When asked about his conceptualization of culture, Adam responds that culture can be modeled in the classroom: “I think that the most important things are things that can be modeled that students can pick up on when greeting each other.” Here Adam understands culture as a concrete product that students can manipulate when interacting with each other. However, he seems aware of the diversity of culture, and for this reason, he offers his students multiple perspectives through which they can examine culture: “I present as much as I can from as many corners, social and geographical, of the Arab World as possible.” This is a way to encounter the one-sided view of other cultures in the U.S. context, given the influence and authority of the media in this country. Accordingly, Adam maintains that it is necessary to give students opportunities to do their own research: “I feel like that’s what it means to teach culture, to give people a wide, very wide presentation and to encourage them to do their own digging.” Adam offers his students such opportunities in the form of reflection papers on cultural phenomena or events in the Arab and Islamic worlds. This view of culture aligns with the view that culture is fluid and can be understood through reflection on self and other (Kramsch 2006). However, except for the culture portfolios and brief discussions on family culture in Arabic, during the classroom observations I did not record instances of culture teaching except for a song, which Adam played at the end of one class without discussing its content or meaning. As for the place of culture in Adam’s syllabi, his first semester Arabic syllabus focuses on communication and proficiency level in the four skills. There was no mention of the term “culture” in his course objectives. In his second semester Arabic course syllabus, Adam states that one of the goals is “to be familiar with many aspects of Arab culture, such as social and religious customs and traditions, family life, and education.” This statement entails that Adam conceives culture simply as performance and facts, which confirms his understanding of 411

Brahim Oulbeid

culture as products that can be modeled. However, Adam provides opportunities for his students to explore Arab cultures through culture portfolios: researching an aspect of culture through stories and movies and reflecting on what they discovered and their own cultures. Unlike Adam, Thomas talks about culture as dynamic and context bound: “I like when I talk to my students, I use the air quotes ‘Arab culture.’ Culture is something that, it is located in a specific point in time and it shifts . . . . What we mean when we say American culture, I try to make connections with my students. You know: ‘I say your culture’ ’cause someone living in Texas may have a different culture than me living in Michigan or here in Massachusetts. You know we have different experiences. So when I talk about culture and when I think of culture I try to think about what’s the context of whatever I am engaging in. You have to think about that context: the place and the temporal time period that you are looking at, that you are examining because that’s gonna influence how we read it.” [Thomas’ views on culture] In the first interview, Thomas indicated that he witnessed the events of 9/11 in the Middle East, where he served as a youth volunteer. This experience gave him a different perspective of these events than the mainstream media view in the U.S. This could be one reason for his understanding of culture as linked to context in terms of time and place and his concern about cultural inauthenticity. Thomas is concerned about cultural inauthenticity appropriation. He thinks that role-plays and skits can never be authentic: “Anything you do in a classroom is in actuality, it’s gonna be inauthentic.” Thomas claims that culture is dynamic and constrained by time and place, and that context is important when discussing it. For this reason, he does not assign roles to his students because he “want[s] [his] students to be themselves, describe themselves, their desires and give them ownership of their language.” However, he offers them opportunities for reflection on their own cultures and formation of their own unique views. Thomas wished he had such opportunities when he was learning Arabic. As a student, Thomas’ exposure to culture was basically limited to a few daily expressions that are used for greetings and congratulations: “We learn like some phrases like when someone gets a haircut, we would say ً ‫[ نعيما‬a saying to someone who had had a haircut or a shower], we would introduce ‫كل سنة وأنت طيب‬ [greetings on a holiday or someone’s birthday] so we got exposure to culture through that.” Here Thomas’ learning of culture followed the modernist perspective (Holliday 1999), which views culture as a product that is related to food, clothing, and greetings. However, he talks about culture differently. Thomas understands culture as dependent on context and connected to time and space, which influences the way we enact it. His view aligns with the sociocultural approach relating the co-construction of learning in interaction through language in different contexts (Lantolf 2000, Johnson 2006). Therefore, the language classroom is one of the contexts where learning culture occurs. Again, unlike Adam, Thomas introduces culture through movies and songs at least once a week (on Wednesdays). This is how he introduces culture in his first-year class, as recorded in my field notes: “As is the case on Wednesdays, the instructor presents a song for the students. The purpose of the song according to the instructor was to provide some cultural content to the students. Today’s song relates to the world cup song in 2014 in Arabic. It was sung by Ahmed Chawki.

412

Arabic language teaching in the U.S.

Before the students listen and watch the song, the instructor provides the students with a script and tells them to circle the words that they already know. After the students circle some of the words, the instructor reads the lyrics and explains some of the new words. The instructor told me that he intended to show this song as it talks about coexistence, peace, and love after the attacks in Paris on 11–13–2015.” [Author field notes on 11–18–2015] After he asks his students questions in Arabic to which they respond with one or two words, he would engage in a conversation in English with the students about the meanings and purpose of the song. This activity shows the discrepancy between his knowledge and classroom practice. In his third-year course syllabus, Thomas notes that one of the goals is to: “Develop an awareness of some of the important issues being discussed in the Arab World and how they are discussed/debated by Arabs. In this respect, students will learn the necessary vocabulary to comprehend and participate in these discussions.” For Thomas, participation in debates through discussion of issues related to the Arabs and their worlds helps students develop cultural awareness.

19.4.2  Language and culture teaching in the classroom As is the case with the majority of language teachers, although the participants recognize the significance of implementing cultural knowledge and understanding in the classroom (Byram et al. 2001, Sercu 2006), this study argues that they focus more on the functional aspects of language. As is the case with Thomas, this tendency derives from the participants’ language learning and preservice training experiences. Adam’s classroom routines activate the vocabulary studied the night before by having students ask and answer questions in pairs: “one of the most deliberate things that I do in the classroom is to create space as much as possible for my students to be activating [the language], for my students to be doing the talking.” This strategy offers opportunities for students who are reluctant to participate and use the language. Before the end of the session, the whole group discusses what they have learned about each other in Arabic. Then the instructor clarifies some language issues that he overheard from students. This strategy is reminiscent of the manner in which Adam acquired Arabic as a student: “there was a great deal of emphasis on again the mechanics of teaching, on when and how to correct for example and how to elicit speech from my students.” The data show that Adam was not able to offer his students opportunities for developing cultural insights because teaching culture has not been a major part of his language teaching preparation: “I can’t point to many things definitively where I felt that culture was front and center as opposed to the mechanics of the language which is frankly a deficit that I note in my own teaching.” For this reason, Adam claims that he is concerned about how to assess cultural learning. At the same time, he recognizes the importance of having what he calls “a working understanding of culture”; that is, developing cultural competence. Thus, Adam creates opportunities for his learners to have insights into the cultures through cultural portfolios: “I ask them to write what I call cultural portfolios and write a couple of hundred words on basically whether they watch a film, read a short story or research a

413

Brahim Oulbeid

particular set of traditions or cultural folkways and then present that information. But it’s up to them to do their own research.” For Adam, culture portfolios are an invitation for his students to become acquainted with some aspects of the cultures, which would stimulate their interests. Thomas’ approach to teaching is no different than Adam’s. His classroom routines emphasize vocabulary, grammar, listening, and speaking. Once a week, he introduces a song at the beginning of class or a movie, activities which he considers as spaces for students’ exposure to culture. When I asked Thomas about the place of culture in the language classroom, he hesitated before responding. Then he mentioned that culture cannot be modeled, and that, as a learner, he was taught only some basic daily expressions that would help students to make connections with their own culture. However, he indicates that exposure to culture in the classroom assists learners when they travel abroad, the only place for him where they can experience authentic culture: “You get some exposure in the classroom, from watching movies, listening to some songs and discuss them in English. I think you get some culture. But when you travel abroad, you’re actually interacting with people in the street.” Thomas also assigns cultural portfolios to help his students discover and view the cultures through their own prisms: “They’re an opportunity for them to explore on their own and reflect on it and compare between the Arab culture that has been depicted, or experiencing when talking with someone and their cultural experience.” One of the challenges that NNS teachers encounter when teaching culture is what Thomas calls “a lack of native intuition,” which he believes NS teachers have. In this regard, Thomas feels obliged to represent Arab culture correctly, but he cannot “pretend to be Arab.” This feeling of discomfort may be linked to limited cultural knowledge and insufficient teacher preparation, which foregrounds the emphasis on methods for teaching the mechanics of language.

19.4.3  NNS teacher identities Adam and Thomas display their identities in their relationships with other faculty, with their students, and in their classroom practices (see Figure 19.1). These contexts of interaction shape the numerous identities that the participants negotiate and construct. In relation to their students, for example, as NNS teachers of Arabic, the participants perceive themselves as a “role model” or “experienced fellow learner” on a journey with their students. In his interviews, Adam comments extensively on his multiple and simultaneous identities. He calls himself a “non-native interpreter of culture,” “an educator,” “a non-native speaker, learner, and teacher of Arabic and all of these things. I consider myself all three all the time.” Adam positions himself as a learner because he sometimes feels the need to answer all of his students’ questions and, seemingly, has limited experience with the culture. Furthermore, he is an interpreter of culture because he looks at it as an outsider. This self-perception confirms the social identity theory: social categorization and social comparison. Adam categorizes himself as belonging to NNS teachers, a group who is continuously learning the language and the culture. Finally, Adam claims to be privileged to teach Arabic: “as a white male American citizen, I am living in a bubble of privilege that is sooo all-encompassing that, you know, it’s difficult for me even to perceive the extent to which it exists.” However, Adam is “eternally worried” as a teacher of Arabic. In one of the interviews, Adam maintains that the source of his worry is that he is the only NNS teacher in his department, and the worries come from comparing his performance to that of another social group: NS colleagues. He forgets that his teaching abilities have been recognized and accepted by the institution that offered him the 414

Arabic language teaching in the U.S.

Figure 19.1 NNS Arabic teacher identity construction across contexts

teaching position. At some point in the interviews, he notes that he talks to his students about his ability to reach a professional level in Arabic: “I think a big part of being a non-native speaker is being a model. I definitely say, I mean I say it again to my classes, to my Arabic classes. I am not a native speaker. Yet I have achieved by certain standards professional proficiency in this language, you can also do this.” This quote seems to contradict his sense of worry when he describes the ways other faculty perceive him. Adam feels worried about how others view him, and the sense of discomfort might come from his lack of self-confidence. It can also stem from the widely spread ideological view that native speaker teachers can be more effective than NNS instructors (Phillipson 1992, Holliday 2006). The findings reveal that Thomas perceives himself as “a fellow learner, sometimes an impostor,” “not of Arab decent,” “more open and more aware,” “unique that [I] can do that,” “in a privileged situation to teach Arabic.” Thomas embraces many positive identities, which indicates the dynamic aspect of his self-perception and that can boost his confidence as a teacher. Like Adam, Thomas constructs himself as being unique and privileged to teach Arabic. Still, and similar to Adam, his status of privilege conflicts with his worries about his performance in the views of his colleagues and students: “It’s scary. It’s scary in the sense that anything I do, I don’t want to misrepresent it. I feel even though I don’t wanna be a representative of Arabic culture, I feel like I do in a way, sometimes I do, how I present it, or how I ask them about it is somehow creating misinformation about it. That’s what I worry about.” Thomas is concerned about “misrepresenting” the culture due to a lack of cultural knowledge. He also lacks confidence because he thinks native speakers can be more effective teachers than 415

Brahim Oulbeid

NNS teachers. This is evident in his statement that he is “not the person to teach but [I] find it really rewarding and enjoyable.” This sense of insecurity might stem from comparing himself to me as a NS teacher given that we taught together at a summer program, and he acknowledged that he learned a lot from me. Additionally, identifying himself as an “imposter” indicates that he is ready to abandon his position because he feels that the language and the learners “deserve a better teacher.” In terms of how the participants compare NNS teachers and NS teachers, the data indicate that they agreed that both NNS teachers and NS teachers can be effective, which questions the privileging of NS teachers and refutes Phillipson’s (1992) “native speaker fallacy.” For Thomas and Adam, the advantage of NNS teachers is that the teacher can anticipate “where the pitfalls are and . . . what’s gonna be troublesome” (Thomas). Adam concurs “a big part of being a non-native speaker is being a model” for their non-native learners. These data confirm the findings from reports on NS and NN English speaker teachers (Mahboob 2010). As for NS teachers, according to the participants, they are deemed to “have native intuition,” which NNS teachers lack. For Adam, besides innate cultural understanding, NS teachers “are the people who are going to best be able to articulate the delicate fine points of language use and the subtle shades of meaning that can arise.” Thomas expresses the same argument by saying, “I don’t have those fine tweakings of language.” Finally, Thomas argues that only NS teachers can teach advanced Arabic classes successfully. Adam heard this argument but did not necessarily agree with it. However, the participants use this issue of lack of native intuition as an impetus to continue expanding their knowledge of Arabic language and culture and, therefore, become agents of their learning and teaching. The findings in this study indicate that while Thomas and Adam view culture as important for learning the language, their experiences as learners and preservice teachers did not offer socialization in teaching culture. Therefore, they limit culture teaching to movies and songs, which support the essentialist view of culture among teachers (Holliday 1999, Sercu 2006). Even though the MLA Report (2007) and the ACTFL guidelines call for the significance of the cultural dimension in language learning, the participants’ teacher education emphasized mainly teaching the mechanics of language. What this means is that Adam and Thomas could benefit from mentorship and professional development opportunities had they been offered in their institutions. Unfortunately, these instructors have to navigate the challenges of the classrooms and the feelings of alienation in their departments on their own (Olsen 2008). When I asked Adam about his relationship with his non-Arabic instructor colleagues, he discusses his sense of alienation: “It is unfortunate that we don’t have as many opportunities to interact as I want, I have made it a goal for myself to actually attend some of their classes and lectures and I don’t have as much opportunities as I want to. Um I do feel as though they have a great source of support for the program especially because I am not a member of the department. I am a bit of an administrative orphan at least within the [College] system.” Even with his Arabic language colleagues, the interactions are limited to monthly meetings to discuss students’ progress and co-curricular events. Despite the feelings of alienation and anxiety as teachers in their classrooms and in relation to their colleagues, Adam and Thomas indicated that they would be continuing their teaching journey, which is interesting. They are up for the challenge. Thomas would like to be teaching elementary levels of Arabic until retirement. Adam does not want to be an administrator, but he would like to continue teaching Arabic and contribute to building a stronger Arabic program in his institution. 416

Arabic language teaching in the U.S.

19.5 Discussion The insights from this study point to accounts the participants constructed through narratives across the contexts of Arabic language learning, preservice teaching, and teachers in practice. In these three contexts, and building on their past, present, and future (Clandinin and Connelly 2006), the participants have accumulated experiences that contributed to the development of their multiple and sometimes conflicting social and professional identities (Johnson 2001, Varghese et al. 2005, Bukor 2015) as they interacted with their learners, Arabic language peers, and other faculty. Adam and Thomas value their group membership as NNS teachers. Additionally, they indicated that they are in a unique and privileged position to be able to teach Arabic. As language practitioners, by virtue of their teacher qualities and characteristics such as language proficiency, cultural experiences, and knowledge of language pedagogy (Johnson 2001), they position themselves as equally professional and able to better address the specific needs of their Arabic language learners in the U.S. context. For example, Adam reminds his students: “I am not a native speaker. Yet I have achieved by certain standards a professional proficiency in this language, you can also do this. This is within your power.” Here, Adam displays “positive positioning” (Søreide 2006) as he empowers his learners and places himself as a role model who can provide a better example for NNS Arabic language students in the sense that he can predict learning challenges and can offer ways to overcome them. This “positive positioning” conflicts with the participants’ acknowledgement of their limited expertise with the languages and cultures of the Arab world and their sense of insecurity as language instructors in comparison to their NS Arabic teachers. As the data have demonstrated, the participants positioned themselves in a negative fashion as well. In front of their learners but specifically in front of their NS Arabic language colleagues, Adam and Thomas claim that they need to be well spoken to prove their language competency. Adam cares about “constantly monitoring [his] pronunciation.” As for Thomas, he is concerned about being challenged by learners who are heritage speakers, which he understands as a “perceived weakness” in him. What this entails is that in their views, accent, and pronunciation are conceived as symbols of social identity that “index” the participants’ otherness (Agha 2007). Part of seeing themselves as “other” seems to be evident in their sense of insecurity as teachers. Both participants used the terms “scary” and “eternally worried” to describe what it is like for them to teach Arabic. In my interpretation, these conflicting identities seem to have come from lack of recognition of these teachers’ efforts and absence of validation of their investment in language competence and, by implication in their social identities (Norton 1997). The data show that these teachers are seeking acceptance and recognition by the target language community. Unfortunately, they are left to deal with their personal and institutional challenges on their own (Olsen 2008). What is striking is their desire to continue their teaching and learning journey because they enjoy their work. To offset the negative selfconceptions and empower themselves, the participants in this study have developed a sense of agency (Beauchamps and Thomas 2009) in the form of some strategies; e.g., constant practice of the language, meeting with an Arabic dialect tutor, attending conferences, and joining professional organizations and social media groups to share teaching resources and best practices. This study has revealed that the Arabic language classroom reality in the U.S. is gradually changing with the growing number of NNS teachers. This paper examined the experiences of two novice teachers of Arabic with a few years teaching in the classroom. As novice teachers, these participants juggle the teaching load within institutions where they are subject to anxiety and discomfort because of lack of support and mentorship. 417

Brahim Oulbeid

19.6  Conclusion and implications This study investigated the perceptions of two NNS teachers of Arabic regarding the teaching of culture in the classroom and traces the development of their professional identities and the linguistic and cultural challenges they face. The data demonstrated the importance of caring for novice teachers and guiding them through professional development opportunities and mentoring them instead of leaving them to cope with their personal and pedagogical challenges. The teachers in this study likened their participation in this work to a healing experience (Bukor 2015). As Adam indicated when he was sitting for the second interview, “the process is very interesting and, as I said, it’s like sitting with a therapist,” which is what novice teachers need: to be able to reflect on their experiences and find solutions for their dilemmas and thrive as teachers. As for Thomas, when we were talking about how even some NS teachers may face challenges in their classrooms, he was relieved to hear that. This is the kind of mentoring that all teachers need to build self-confidence, develop favorable and positive selfperceptions, and challenge the NS teacher bias. Another implication for this research is that language education programs need to revise the language-culture dichotomy and provide preservice teachers with tools and strategies for the implementation of cultures in the world language classroom. As this study demonstrates, one of the deficits of Arabic language education programs may be attributed to the constant focus on the communicative aspects of language, which influences the classroom practices of new teachers. Considering the various factors that influence teacher identity formation, this study will be valuable for preservice and novice teachers, as it provides an understanding of the processes teachers go through. In a similar vein, this study will help NS teachers to understand the challenges of their NNS colleagues and to work to support each other in a collaborative environment. While Arabic language programs’ mission statements may offer insights into the significance of language and culture teaching, this study shows that culture teaching is not an essential part of the teachers’ preservice education. I will consider this pedagogical omission in a future study with a larger sample and will specifically examine whether it is the NS–NNS dichotomy that affects the likelihood of teaching culture, insufficient preservice training, limited linguistic and cultural knowledge among some NNS teachers, or all of the above.

Appendix Sample interview questions 1 1 Tell me about yourself in relation to learning and teaching Arabic language and culture. 2 What were your very early experiences with this language and culture like? Can you describe these experiences? • How did you learn the Arab language and culture as a student? 3 How did you become an Arabic language and culture teacher? 4 Could you talk a little bit about your teacher training experiences? 5 Now as a non-native teacher of Arabic, how would you describe your classroom teaching? Can you tell me about your teaching philosophy? 6 What are your thoughts about culture in the classroom? How do you understand culture in the Arabic classroom? 7 What does it mean to teach a language and a culture that are not part of your first language? 418

Arabic language teaching in the U.S.

Sample interview questions 2 1 What does a typical workday look like for you? Can you walk me through a typical working day from morning until evening? 2 Would you mind describing how your 50/75 minutes teaching session is structured? 3 Speaking of the students, I would love to hear about the kinds of relationships you have with your students: in class, outside of class, during office hours . . . how would you describe these relationships? 4 Based on your experiences, how do you feel your students view you as a non-native teacher of Arabic? 5 Can you talk a little bit about the relationships you have with Arabic teaching faculty in your department? 6 What kinds of relationships do you have with colleagues outside of the field of Arabic teaching in your campus? 7 How do you see yourself in front of colleagues and other faculty members? How do you identify yourself as a teacher, as a non-native speaker of Arabic? 8 In what ways do you think teachers can emphasize culture in the classroom? 9 How do you find or where do you find culture instruction resources?

References Abdalla, M. and Al-Batal, M., 2011. College-level teachers of Arabic in the United States: A survey of their professional and institutional profiles and attitudes. Al-‘Arabiyya. 44/45, 1–28. Agha, A., 2007. Language and social relations. Cambridge: Cambridge University Press. Banks, J. A., 2002. Cultural diversity and education: Foundations, curriculum, and teaching. New York: John Wiley and Sons. Barkhuizen, G., et al., 2013. Locating research methods within an applied linguistics narrative framework. Te Reo, 56/57, 109–126. Barkhuizen, G., et al., 2014. Narrative inquiry in language teaching and learning research. New York: Routledge. Beauchamp, C. and Thomas, L., 2009. Understanding teacher identity: an overview of issues in the literature and implications for teacher education. Cambridge Journal of Education, 39 (2), 175–189. Belnap, R. K., 2007. Some unfortunate implications of unhappy Arabic teachers. Al-‘Arabiyya. 40/41, 87–103. Block, D., 2007. The rise of identity in SLA research, post Firth and Wagner (1997). The Modern Language Journal, 91 (1), 863–876. Borg, S., 2003. Teacher cognition in language teaching: A review of research on what language teachers think, know, believe, and do. Language Teaching, 36 (2), 81–109. Borg, S., 2006. The distinctive characteristics of foreign language teachers. Language Teaching Research, 10 (1), 3–31. Braine, G. ed., 1999. Non-native educators in English language teaching. Mahwah, NJ: Lawrence Erlbaum. Brumfit, C. and Johnson, K., 1979. The communicative approach to language teaching. Oxford: Oxford University Press. Bukor, E., 2015. Exploring teacher identity from a holistic perspective: Reconstructing and reconnecting personal and professional selves. Teachers and Teaching, 21 (3), 305–327. Byram, M., et al., 2001. Developing intercultural competence in practice. Clevedon, UK: Multilingual Matters. Byram, M. and Risager, K., 1999. Language teachers, politics, and cultures. Clevedon, UK: Multilingual Matters. Connelly, F. M. and Clandinin, D. J., 2006. Narrative inquiry. In: J. L. Green, et al., eds. Handbook of complementary methods in education research. Mahwah, NJ: Lawrence Erlbaum, 477–487.

419

Brahim Oulbeid Cook, V., 1999. Going beyond the native speaker in language teaching. TESOL Quarterly, 33 (2), 185–209. Creswell, J. W., 2012. Educational research: Planning, conducting, and evaluating quantitative and qualitative research. Upper Saddle River, NJ: Pearson. Galeano, R., 2014. Attitudes of U.S. foreign language teachers toward teaching culture. International Online Journal of Education and Teaching, 1 (2), 56–71. Ghanem, C., 2015. Teaching in the foreign language classroom: How being a native or non-native speaker of German influences culture teaching. Language Teaching Research, 19 (2), 169–186. HigherEdJobs, 2016. Lecturer, world languages and literatures (Arabic). Available from: www.higheredjobs.com/search/details.cfm?JobCode=176349033. [Accessed 10 October 2016]. Hogg, M. A. and Abrams, D., 1998. Social identifications: A social psychology of intergroup relations and group processes. London: Routledge. Holliday, A., 1999. Small cultures. Applied Linguistics 20, 237–264. Holliday, A., 2006. Native-speakerism. English Language Teaching Journal, 60 (4), 385–387. Horvat, E., 2013. Making sense of what you are seeing: analysis and writing. In: E. Horvat, ed. The beginner’s guide to doing qualitative research: How to get into the field, collect data, and write up your project. New York, NY: Teachers College Press, 105–124. Johnson, K., 2001. Social identities and the NNES MA TESOL Student. Eric Educational Resources Center 457682. https://eric.ed.gov/?id=ED457682 Johnson, K. E., 2006. The sociocultural turn and its challenges for second language teacher education. TESOL Quarterly: A Journal for Teachers of English to Speakers of Other Languages and of Standard English as a Second Dialect, 40 (1), 235–257. Kearney, E., 2008. Developing worldview(s): An ethnography of culture learning in a foreign language classroom. Thesis (Ph.D.). University of Pennsylvania. Klein, F. M., 2004. Culture in the foreign language classroom: Teachers’ beliefs, opportunities and practice. Thesis (PhD). University of Minnesota. Kohler, M., 2015. Teachers as mediators in the foreign language classroom. Buffalo, NY: Multilingual Matters. Kramsch, C., 2006. Culture in language teaching. In: H. L. Andersen, et al, eds. Culture in language learning. Oakville, CT: Aarhus University Press, 11–25. Kramsch, C., 2013. Culture in foreign language teaching. Iranian Journal of Language Teaching Research, 1 (1), 57–78. Kramsch, C., 2014. Teaching foreign languages in an era of globalization: introduction. The Modern Language Journal, 98 (1), 296–311. Lantolf, J. P., 2000. Second language learning as a mediated process. Language Teaching, 33 (2), 79–96. Llurda, E., 2005. Non-native language teachers: Perceptions, challenges, and contributions to the profession. New York: Springer. Mahboob, A., 2010. The NNEST lens: Nonnative English speakers in TESOL. Newcastle upon Tyne, UK: Cambridge Scholars. Medgyes, P., 1992. Native or non-native: Who’s worth more? English Language Teaching Journal, 46 (4), 340–349. Medgyes, P., 1994. The non-native teacher. London, UK: Palgrave Macmillan. Mijalski, M., 2014. Foreign language education for the 21st Century: Developing intercultural competence in an online intercultural exchange. Thesis (PhD). University of Southern California. MLA Ad Hoc Committee on Foreign Languages., 2007. Foreign languages and higher education: new structures for a changed world. Profession, 234–245. Moussu, L. and Llurda, E., 2008. Non-native English-speaking English language teachers: History and research. Language Teaching, 41(3), 315–348. National Standards in Foreign Language Education Project., 2006. Standards for foreign language learning in the 21st century: Including Arabic, Chinese, classical languages, French, German, Italian, Japanese, Portuguese, Russian, and Spanish. 3rd ed. Yonkers, NY: Author. Norton, B., 1997. Language, identity, and the ownership of English. TESOL Quarterly, 31 (3), 409–429. Olsen, B., 2008. How reasons for entry into the profession illuminate teacher identity development. Teacher Education Quarterly, 35 (3), 23–40. Patton, M. Q., 2002. Qualitative research and evaluation methods. 3rd ed. Thousand Oaks, CA: Sage. Peiser, G. and Jones, M., 2014. The influence of teachers’ interests, personalities and life experiences in intercultural languages teaching. Teachers and Teaching: Theory and Practice, 20 (3), 375.

420

Arabic language teaching in the U.S. Phillipson, R. 1992. Linguistic imperialism. Oxford: Oxford University Press. Phipps, A. M. and Guilherme, M., 2004. Critical pedagogy: Political approaches to language and intercultural communication. Clevedon, UK: Multilingual Matters. Reis, D. S., 2011. Non-native English-speaking teachers (NNESTs) and professional legitimacy: A sociocultural theoretical perspective on identity transformation. International Journal of the Sociology of Language, 208, 139–160. Ryan, P. M. and Sercu, L., 2003. Foreign language teachers and their role as mediators of language and culture: A study in Mexico. Estudios De Lingüística Aplicada, 21 (37), 99–118. Schulz, R. A., 2007. The challenge of assessing cultural understanding in the context of foreign language instruction. Foreign Language Annals, 40 (1), 9–26. Seidman, I., 1998. Interviewing as qualitative research. New York: Teachers College. Sercu, L., 2006. The foreign language and intercultural competence teacher: The acquisition of a new professional identity. Intercultural Education, 17 (1), 55–72. Sercu, L., 2007. Foreign language teachers and intercultural competence. In: M. J. Raya and L. Sercu, eds. Challenges in teacher development: Learner autonomy and intercultural competence. Frankfurt, Germany: Peter Lang, 41–64. Sleeter, C. E., and Grant, C. A. 2002. Making choices for multicultural education: Five approaches to race, class, and gender. Hoboken, NJ: John Wiley and Sons. Søreide, G. E., 2006. Narrative construction of teacher identity: positioning and negotiation. Teachers and Teaching: Theory and Practice, 12 (5), 527–547. Stake, R. E., 2008. Qualitative case studies. In: N. K. Denzin and Y. S. Lincoln, eds. Strategies of qualitative inquiry. Los Angeles, CA: Sage, 119–149. Tajfel, H., 1978. The social psychology of minorities. London: Minority Rights Group. Valdiviezo, L. A. and Nieto, S., 2015. Culture in bilingual and multilingual education: Conflict, struggle, and power. In: W. E. Wright, eds. The Handbook of bilingual and multilingual education. Malden, MA: Wiley-Blackwell, 92–108. Varghese, M., et al., 2005. Theorizing language teacher identity: Three perspectives and beyond. Journal of Language, Identity, and Education, 4 (1), 21–44. Vygotsky, L. S., 1978. Mind in society: The development of higher psychological processes. Cambridge, MA: Harvard University Press. Welles, E. B., 2004. Foreign language enrollments in United States institutions of higher education, Fall 2002. The Association of Departments of Foreign Languages Bulletin, 35 (2–3), 7–26. Zhang, C. and Jensen, A. A., 2013. Professional identity construction of non-native Chinese teachers. In: M. Kirkebæk, et al., eds. Teaching and learning culture: Negotiating the context. Dordrecht: Springer, 113–127.

421

INDEX

accent 39 – 40, 42, 51, 57, 114, 254, 417 acceptance 417 access: conceptual 362 – 365, 376 – 377; lexical 151, 362, 373, 375, 378; phonological 72, 86 – 87 accessibility hierarchy (AH) 182 – 183, 188, 195, 198 accuracy 3, 9, 11 – 12, 18 – 26, 28, 30, 40, 59, 62 – 63, 64 – 73, 79 – 80, 82 – 85, 90, 114, 130, 143, 153, 189, 265 – 266, 269 – 273, 275 – 279, 281, 331, 334, 336 – 341, 369, 372, 376, 403 ACTFL see American Council on the Teaching of Foreign Languages ad hoc committee on foreign language learning 405 affective factors 331 – 333, 337, 341 – 344, 346 – 348, 359n1, 360n12 affective filter 396 – 397 affective filter hypothesis 348 age: acquisition 332, 362, 365, 375, 377; arrival 57; effect 362; first encounter 96 agreement: gender 2, 201, 203 – 204, 206, 209, 215, 268, 271, 278, 368, 382; nominal 334; number 271; person 206 – 207, 215, 256, 271, 281; verb 2, 202, 203, 205, 207 – 216, 220n8, 368, 382 AH see accessibility hierarchy American Council on the Teaching of Foreign Languages (ACTFL) 119, 144, 188, 293, 416 animacy 201 – 203 anxiety 402, 416 – 417 applied linguistics see linguistics aptitude 10 Arabic L2 morphosyntax 1 – 2, 58 Arabic linguistics see linguistics article: definite 205 – 206, 228, 270, 275

articulatory: context 96; differences 98; effect 94; feedback 93; movements 94 – 95; production 107 attention 2, 29, 139 – 142, 145 – 146, 150 – 152, 212 – 213, 266, 272 attitude 253 – 254, 331 – 333, 335, 337, 342 – 343, 344 – 345, 347 – 348, 353, 359, 359n1, 404 – 405 attrition 375 automatic selective perception model (ASPM) 59, 72 awareness 29, 31, 72, 151, 193, 195, 255, 262, 389, 395, 398, 400, 407, 413; (inter)cultural 305, 405, 413; metalinguistic 3, 225 – 226, 234, 236, 239; self- 402, 406 – 407 background knowledge 158, 160; see also prior knowledge bilingual: acquisition 378; environment 300; heritage 3, 362, 367, 369, 377 – 378; identities 289, 297 – 298, 300; learners 40; lexical retrieval 362; lexical storage 3; lexicon 363, 365; programs 300; speakers 289, 299 – 300, 362, 364 – 365, 367, 370 Bosnian 189 Brazilian Portuguese 11 case 186, 201 – 203, 262n1, 271, 278, 281; accusative manṣuub 203, 217; genitive 183; nominative 203 Chinese: bilingual 364; consonants 75, 86 – 87, 118 – 119; L1 2, 9, 11, 17, 19 – 22, 24, 26 – 31, 32nn2 – 3, 57, 70 – 71, 73, 75 – 78, 81 – 83, 85 – 90, 113 – 115, 118 – 119, 121, 125, 129 – 132, 184; Mandarin 10, 17, 29, 75, 115; phonological system 115, 118, 131; resumptive pronouns 183; syllabic structure 118 – 119 clarification request 311, 395

422

Index code-switching (CS) 3, 289 – 300, 339 – 340, 346, 359n9 cognate accusative 260 cognitive: abilities 365; effects 139 – 140; effort 140, 152; factors 348; load 167; processes 142, 151, 153; processing 204, 368; psychology 5; system 181; verbs 256 communication 58, 113 – 114, 131, 233, 235, 237, 244, 273 – 274, 291 – 292, 294, 298 – 299, 304 – 306, 310, 324, 332, 391 – 392, 403, 405, 411 competence: communicative 294, 299; intercultural 289, 405, 413; linguistic 294 competition model (CM) 2, 201 – 205, 212, 215 – 216, 219 complexity 3, 86, 140, 183, 198, 265 – 270, 272, 273, 276 – 279, 331, 334, 336 – 339, 341, 367 comprehension 57, 157, 202 – 203, 395; listening 114; reading 139, 141 – 142, 144 – 145, 231, 234, 368, 398 comprehension checks 144, 395 concept mediation model (CMM) 362 – 371, 376 – 377 consonant contrasts 2, 6, 56 – 67, 71 – 72, 77, 86 – 87 consonants: affricate 4, 74 – 76, 86, 118; alveolar 4, 45, 73 – 76; coronal 76 – 77, 94, 99 – 102, 105; dorsal 77, 101 – 103; emphatic 2, 4, 6, 62, 70 – 77, 80 – 88, 90, 93 – 96, 99 – 102, 105 – 110, 116, 118; flap 75, 117; fricative 4, 6, 59, 72, 74 – 77, 80, 82 – 84, 86 – 88, 90, 95, 101 – 107, 115, 117 – 118, 132 – 135, 369; glide 4, 118; glottal 4, 6, 58 – 59, 62, 65, 67, 73 – 76, 80, 83, 86 – 89, 93 – 95, 101, 104 – 110, 116 – 118, 369; guttural 2, 93 – 96, 101, 105 – 107; (inter)dental 4, 73 – 75, 118; labial 4, 74 – 76, 116 – 118; laryngeal 86 – 89, 94; lateral 4, 73 – 76, 115, 117, 133; nasal 4, 74 – 76, 115, 117 – 118, 120, 133 – 134; obstruent 132; palatal 4, 74 – 76, 118; palato-alveolar 4, 74 – 76; pharyngeal 2, 4, 6, 58 – 60, 62, 64 – 65, 67, 70 – 73, 75 – 78, 80, 82 – 90, 93 – 95, 103 – 110, 116, 118, 369; plain 4, 6, 76, 85 – 86, 90, 100, 105 – 106; stop 4, 6, 70, 73 – 77, 80 – 86, 88, 95, 99, 100 – 101, 103, 105 – 107, 115, 117 – 118, 132 – 134, 369; tap 4, 73, 75; trill 4, 73 – 74, 118; uvular 2, 4, 59, 64, 70 – 78, 80, 82 – 88, 90, 93 – 95, 101 – 110, 116, 118; velar 4, 64, 73 – 76, 78, 80, 83, 86 – 87, 94, 101, 118; voiced 6, 42, 45, 59, 64, 72, 75, 82 – 83, 86 – 89, 95, 103, 105, 107, 115, 117, 132 – 134; voiceless 6, 42, 59, 64, 72, 74 – 75, 81 – 84, 86 – 87, 89, 95, 103 – 105, 107, 115, 117, 132 – 135 contrastive analysis 113, 231 contrastive analysis hypothesis 71 contrastive analysis of corpora 230, 233 contrastive analysis of genre 236

contrastive discourse analysis 253 contrastive rhetoric 254 convergence 228 conversational analysis 253, 290 cooperative principle (CP) 291 corpora 225 – 226, 230 – 233, 236, 238, 240, 242 – 243, 282n2 critical period 41, 44 data-driven learning (DDL) 225 – 226 deep processing 167; see also processing; shallow processing depth of processing hypothesis 141 depths of processing theory 167 dialect 5, 10, 13, 17, 32, 41, 52, 58, 73, 78, 94, 96 – 97, 99, 105, 280, 289, 304, 331 – 333, 339, 367, 388, 395 – 401, 417; Cairene 32, 41; Damascene 10, 32; Egyptian 10, 87, 98 – 99, 113, 131, 269, 271, 273, 281, 282, 320, 334 – 336, 339, 359n3; Iraqi 73; Jordanian 49, 73, 78 – 79, 99, 105, 335, 359n3; Latakian 183; Lebanese 98 – 99, 273 – 274, 335, 340, 359n3, 368; Levantine 99, 336, 359n3, 367, 408; Libyan 73, 78 – 79, 87; Najdi 39, 42, 44, 50; Palestinian 10, 17, 183, 189, 269, 334 – 335, 339, 359n3; Saudi 17 – 18, 189, 267, 269, 273; Sudanese 87; Syrian 78, 183, 199n1, 267, 334 – 335, 339 – 340, 359n3; Yamani 73 diphthong 5, 40 – 41, 118 discourse 227, 233,, 236 – 237, 239, 254 – 255, 261 – 262, 268, 270, 292, 304, 317, 320 – 321, 337, 339, 404; academic 253, 279; analysis 3, 226 – 227, 251 – 254; community 404; connected 336; cues 270; cultural 317 – 318, 321 – 322; features 258; functions 292; genre 233 – 234, 237, 272; journalistic 231 – 232, 234; marker 256 – 257, 260; meta- 254; strategies 272; written 280 distance: phonological 115, 117 – 118, 130; social 293; sociolinguistic 253; structural 348; syntactic 183 – 185 distance learning 390 divergence 279 Dutch 57, 202 emergentist accounts 1 English: agreement 202; American English 17, 38, 56, 58 – 59, 64, 94; and Arabic heritage 3, 332, 335, 336, 343, 348, 362, 367; bilingual 3, 364 – 368; British English 73; code-switching 3, 289 – 290, 292, 294 – 300, 304, 339 – 340, 346; colloquial 267; consonants 56 – 58, 64, 71 – 76, 86 – 87, 94, 96, 166, 369; discourse 253, 270; English-speaking country 40; global language 305; L1 2 – 3, 9, 11, 17, 19 – 22, 24, 26 – 30, 32nn2 – 3, 38 – 40, 42, 56 – 61, 64, 66 – 67, 70, 72 – 73, 77, 81 – 87,

423

Index 89, 94, 96, 105 – 107, 139 – 140, 142 – 143, 157, 181, 183 – 184, 188 – 189, 201 – 204, 212 – 214, 216 – 219, 251, 256, 274, 293, 336; L2 9 – 11, 29, 38 – 43, 49, 57, 71, 75, 78, 115, 159, 183 – 184, 202, 226, 253, 255, 269 – 270, 279, 293, 332, 391, 403, 406 – 407, 416; novel words 10; pronunciation 42, 94; relative clauses 183 – 184; Scottish English 73; spelling 18; stress 10 – 11, 29, 40 – 43, 50; stylistic features 251; syllable 43; voice 261; vowels 42 – 43, 57; word order 212 – 213, 220n10; words 12, 14, 40, 97, 118, 143 – 146, 158 – 159, 162, 168, 171, 314, 340; writing 252 – 258, 268 error analysis 270 errors: agreement 268; disagreement 271; gender 271, 368; grammatical 368; language user 270; lexical 334; linguistic 311; nominal construction 368; number 271, 368; phonological 65, 71, 87, 90, 106, 113 – 114, 119, 376, 377; preposition use 368; pronunciation 39, 93, 107, 113 – 115, 119 – 131; prosodic 39; relative clause 195; segmental 39; semantic 377; speech 114; spelling 271, 278; syllable structure 39 – 40; syntactical 266, 281; transfer 113, 213; vocabulary 164; word selection 334; word stress 40, 43; writing 114, 265, 270, 281 European structural semantics 227 explicit feedback 107 explicit instruction 95, 140, 201, 203, 213, 215 – 216, 252 explicit strategy 183 eye-mind link hypothesis 139, 142 eye-tracking 2, 139 – 140, 142 – 143, 145, 150 – 153 E-Z reader model 151

French 3, 10 – 11, 29, 31, 72, 225 – 227, 234 – 235, 242 – 244, 293, 365, 404, 408 frequency: cues 18, 31; effects 2, 6, 9, 11 – 12, 15 – 16, 19 – 21, 24, 26, 28 – 31, 142, 151; learning 11, 32 frequency-based accounts 1, 11 – 12, 32 future see tense gender 158, 169, 190, 203, 261, 304, 315, 333, 335 – 336, 338, 344, 350, 359; see also agreement generative grammar 1, 2, 199n3; see also universal grammar German 2, 57, 70, 73 – 78, 81, 83, 85 – 87, 89 – 90, 158, 189, 226, 306, 404, 408 grammar 40, 70 – 71, 86 – 89, 181 – 184, 188, 195, 196 – 198, 199n3, 213, 226, 245, 382, 395, 398, 414 grammaticality judgment task (GJT) see judgment Greek 2, 70 – 71, 73 – 78, 81 – 83, 85 – 90, 183 Hebrew 76 heritage language acquisition/learning 331 – 332, 337, 341, 345 – 349, 359n8, 369, 376 heritage learners/speakers 1 – 3, 98, 161, 204, 234, 269, 274, 304, 313, 331 – 333, 335 – 349, 359n4, 359nn6 – 7, 359n13 – 14, 362, 365 – 370, 374 – 378, 417 Hindi 71 HIS see initial hypothesis of syntax Hungarian 10, 29 hypergeometric model 238 identity 251 – 252, 255, 257, 261, 297 – 298, 303, 312 – 314, 317 – 318, 321, 331 – 333, 335, 337, 342 – 343, 345, 347 – 348, 354, 359, 359n1, 402 – 404, 406 – 407, 409, 414 – 415, 417 – 418 identity negotiation 406, 303, 312 ideology 225 – 226, 230 – 231, 233, 239, 271, 321 – 322, 347, 402, 405, 415 IL see interlanguage imitation 120, 268 imperfect 256, 261 implicational: pattern 183, 192, 195; scale 191 – 192, 194 – 195 incidental learning 140, 144, 150, 152 – 153; see also vocabulary learning inflection 202, 275, 281 inflectional morphology 31 initial hypothesis of syntax (HIS) 184 initial state 184 input 31 – 32, 41, 59, 88, 153, 181, 234, 237, 341, 342, 345 – 348, 360; incidental input 342; input exposure 12, 22, 26, 30, 41, 61, 78, 342; input frequency 1 – 2, 9, 12; input modalities 153; input recycling 32; input regularities 12;

feature-based model 71, 85, 90 features: aesthetic 259; default 184; dialogic 261; discourse 258; discursive 2, 251 – 252, 255, 258, 260 – 262; language 181, 184, 306; metadiscursive 253; morphosyntactic 236; phonological 58, 71 – 72, 78, 86 – 88, 90, 113; prosodic 39 – 41; sociolinguistic of written texts 251, 253; stance 252; stylistic 251, 254, 261; suprasegmental 2, 9, 39 – 40, 43; syntactic 267; unmarked 184; voice 252; written text 142, 153, 253, 256 Finnish 10, 29, 142, 183 first language acquisition 41, 107, 366, 375 first noun principle 202, 213 fluency 3, 9, 12, 18, 25, 26, 30, 143, 151, 265 – 266, 272 – 273, 276 – 279, 331, 334, 336 – 339, 341, 362, 365, 375 foreigner talk 269

424

Index 5; Arabic linguistics 71; corpus linguistics 2, 225, 227; psycholinguistics 5, 11, 144, 216; sociolinguistics 300 listening comprehension see comprehension

instructed input 342; instructional input 26, 32, 342; language input scale 342; language/ linguistic input 12, 22, 51, 142, 331, 335, 337, 341 – 342, 351, 359; naturalistic input 342; written input 153 inquiry-based mobile learning model 167 intake 140 intentional learning 140 – 141 (inter)cultural integration 305, 322, 403 (inter)cultural learning 3, 303 – 307, 309, 312, 314, 317, 321 – 323, 403 – 405, 410, 412 – 414, 416, 418 interference of L1 113, 270 interlanguage (IL) 1, 181 – 184, 188, 195 – 198, 213, 266 – 267, 270 interpretive semantics 226 – 227, 229, 235 – 236 interpretive semantic theory see interpretive semantics intonation 39, 119, 300 involvement load hypothesis 141 Italian 10, 57, 183, 202

Malay 64 Maltese 76 markedness: constraints 88; effects 71; phonological 71, 88, 115 – 116, 119; sociolinguistic (Myers-Scotton’s) 3, 289, 291 – 293, 298, 300; syntactic (Platzack’s) 184; universal 115 markedness model (MM) 3, 289, 291 – 293, 298, 300 memory 151 – 152, 166; abstract memory system 363; conceptual memory 363; lexical memory system 363; long-term memory 160; memory strategies 168, 170; working memory 160 meta-analysis 271 minimalist program (MP) 2, 181, 184 – 185, 195, 198 MM see markedness model monolingual: acquisition 366; ideologies 321 – 322; practices 317 – 318, 322; speakers 368, 370 morpheme: derivational 50; inflectional 47, 50 morphology 114, 341, 346; Arabic morphology 166; Inflectional morphology 31; pattern 144, 152, 166; root 16, 33 – 34, 42, 152, 166, 281; vocalic template 16, 33 – 34 morphosyntax see Arabic L2 morphosyntax motivation 114, 141, 152 – 153, 166 – 167, 169 – 170, 305, 389, 395 multilingual: identities 25, 317 – 318; practices 317, 322

Japanese 10, 31, 56 – 57, 71, 115, 255 judgment: evaluative 393, 399; grammaticality 181, 183, 189 – 190, 192 – 193, 195 – 196, 368 – 369, 377, 392; listeners’ 39; native speakers 334; value 251 – 254, 257; see also tasks jussive mood 271, 281 Korean 39, 71, 184 L2 status factor hypothesis 332 language skills 66, 96, 131, 225 – 226, 314, 317, 347, 349, 377 – 378 language transfer see transfer learnability 66, 183 learning: difficulties 56, 71 – 72, 87, 298, 417; experience 71, 244, 303, 332 – 333, 346, 348, 350, 355, 389; needs 88; processes 1, 233; settings/contexts 51, 166 – 167, 229, 233, 289, 292, 294, 299 – 300, 332, 417; strategies 168; style 168 lemmatization 270, 275 lexical knowledge 21 – 22, 24, 140, 145 – 146, 151, 269, 279 lexical learning 30, 140, 289, 303; see also vocabulary learning lexicon: acquisition and development 152, 240, 368, 375 – 376; bilingual lexicon 363, 365; frequency distribution 14; lexicon and conceptual domain 363, 365; lexico-semantic links 375; lexicon of heritage speakers 366; see also lexical knowledge; vocabulary learning lexico-semantic model 2, 362 linguistics 5, 189, 266, 293, 391; applied linguistics 226, 266; Arabic applied linguistics

neurology 266 noticing hypothesis 140 noticing novel exemplars 151 null subjects see pro-drop offline: exposure 150; incidental learning 140; measures 150; vocabulary learning 150 online: processing 139 – 140, 142, 150; reading 140, 144, 149, 151, 154; reading measures 146, 149, 150 ontogeny (phylogeny) model 71 operation: merge 181, 184, 196 – 198; move 181, 184, 196 optimality theory (OT) 70 – 71, 73, 88 – 89 oral proficiency interview (OPI) 119, 144, 188, 279 – 280, 293 overgeneralization 183 parsing 219n1 passive voice 212, 255

425

Index past see tense pattern see morphology pedagogical: benefits 107; challenges 418; choices 387 – 388, 391 – 392, 394 – 396; contents 389; decisions 395 – 396, 398; development 408; implications 31 – 32, 50, 66, 90, 107, 131, 153, 169 – 170, 198, 216, 240, 261, 280, 289, 300, 321, 348, 378; insights 292, 299; interventions 216, 322; practices 5, 227, 388; principles 397; procedures 387, 391; process 227; questions 332 – 333; recommendations 303, 307, 321; reflections 251; strategies 389, 403; tool 95; training 283; use 97 pedagogy 1, 3, 201, 216, 227, 245n6, 405, 417 perceptual assimilation model (PAM) 56, 59, 64, 67, 71 – 72 Persian 183 phonetics 2, 114, 346 phonological constraints: active constraints 88; constraints ranking 88; constraints re-ranking 88; faithfulness constraints 88; inactive constraints 88; markedness constraints 88; phonotactic constraints 16; universal constraints 88 phonology 114; Arabic L1 phonology 10, 341; Arabic L2 phonology 1 – 2, 9, 114 – 115; Chinese phonology 118, 131; L1 phonology 78; L2 phonology 11, 32, 38, 56, 71, 73, 118; segmental phonology 71 Polish 10, 40, 189 Portuguese 10 – 11; see also Brazilian Portuguese pragmatic: devices 299; reasons 206 pragmatics 1, 114, 158 prepositions 186, 205, 206, 271 – 272, 278, 368, 382 present see tense prior knowledge 11, 22, 58, 144, 159, 228 processing: difficulties 214; L1 processing 212, 216; L2 processing 202, 204, 216; language 11; lexical 139, 151 – 152, 167, 365; load 9, 185; real-time 140 – 141, 204; relative clauses 183 – 184; sentences 2, 142, 201 – 203, 212, 216; speech 38; strategies 204, 215; see also online, processing pro-drop parameter 187 – 188, 256 productive: deficiencies 378; knowledge 141, 143, 149; pronunciation 66; recall 151 proficiency measures 3, 308, 347 pronominal subjects 190 prosodic contrasts 38 prosodic structure 10; see also syllable structure psycholinguistics see linguistics psychology 5, 140 – 141, 266

receptive: knowledge 141, 143, 149, 152; skills 114 referential: co-referential pronouns 185; pronouns 257; semantic 227 rehearsal 159, 167 relative clauses 2, 181 – 182, 185 – 188, 190, 194, 196, 198 – 9nn2 – 3, 203, 268 relativization 184, 186 – 187, 190, 193, 196 religion 168, 318, 324, 333, 347, 358 – 359 repetition 139, 141 – 143, 150 – 151 representation: cognitive processes 142; conceptual 364; encyclopedic knowledge 228; IL grammars 196; language 363; media analysis 226; phonological 86; scalability 192; self-representation 255, 257 restructuring 267 resumptive pronouns 2, 181 – 182, 184 – 185, 188, 194 – 195, 198 root see morphology rule-based learning 32 Russian 158, 310 self-monitoring 95, 107 self-paced reading/listening 219 self-perception 406 – 407, 414 – 415 semantics 3, 228; see also interpretive semantics Semitic 58, 76, 166 shallow processing 167; see also deep processing; processing social distance see distance social interaction 114 socio-affective factors see affective factors socio-contextual factors 331 – 333, 337, 342 – 344, 346 – 349, 359n1 sociocultural theory 290, 406 – 407 sociolinguistics see linguistics sonority scale/sequencing principle (SSP) 115, 117 – 118, 120, 131 – 132 Spanish 10 – 11, 29, 40, 57, 183, 189, 215 – 216, 293, 317 – 318, 366, 404 speech learning model 71 SSP see universal sonority scale principle stance 3, 251 – 262 standardized spoken language test (SPEAK) 39 statistical learning 11 stimulated recall 151 – 152 stress contrasts 10 – 11, 16, 19, 29, 45, 50 stress correlates 10, 25, 29, 40, 43, 50 – 51 stress deafness model 10 – 11 stress typology model 11 subject-verb agreement see agreement, verb Swedish 183 syllable structure 10, 12, 14 – 16, 33 – 34, 39 – 42, 44 syntactic derivation 2, 181 – 182, 184 – 188, 196, 197 syntax 114, 141, 158, 184, 341, 346

reaction time 9, 18 – 22, 24 – 28, 369 – 371, 376 – 377 reading comprehension see comprehension

426

Index systematic variation 113, 131, 181, 188, 191, 196; see also variation

Turkish 2, 70, 73 – 78, 81 – 87, 89, 90 typological influences 10, 182 – 183

Taiwanese 57 task-based learning 153; see also learning tasks: AXB discrimination perception 60 – 61, 65 – 66, 78, 80, 83; description 268, 274, 277; evaluating teaching 399; grammaticality judgement 181, 183, 189 – 190, 192 – 193, 195 – 196, 368 – 369, 377, 392; L2-vocabulary 141; lexical decision 17, 21 – 23, 28, 30, 366; lexical discrimination AXB 58; meaning recall 145; multiple choice 145; narration/ narrative 274, 279; non-lexical discrimination AXB 58; oral production 331, 336 – 340, 342, 345 – 346; persuasive/argumentative 273, 279 – 280; picture description 181, 189, 191, 194 – 197; picture naming 3, 362, 364 – 365, 367 – 374, 377; read-aloud minimal pairs 79 – 80, 83; read-aloud production 60 – 61, 64 – 66; reading 140, 144; reading speed 32n3; recognition 145; self-evaluation 400; sentence combination 181, 189 – 190, 193, 195 – 197; stress identification 17 – 19, 21, 28 – 30; stress perception 17, 56; stress placement 40; stress production 11, 17, 26, 28 – 30, 56; subject/ agent identification 202, 216; translation 3, 362, 364 – 374, 376 – 377; writing 273 – 274; writing production 331, 335 – 340, 342, 345 – 346 tense 236, 242, 257; future 182, 417; past 237, 254, 261, 406, 410, 417; present 190, 207, 237, 256, 261, 406, 410, 417 tests: cloze 59, 267, 269, 279; form recognition 145; meaning recall/translation 145; meaning recognition 146; multiple-choice 141; paper and pencil vocabulary knowledge 139 – 140, 149; reading-aloud 120; vocabulary posttest 146, 151; vocabulary strategies 160; see also oral proficiency interview; standardized spoken language test; tasks third language acquisition 70 – 71, 332 trade-off hypothesis 266, 272 transfer: L1 transfer 2, 10, 29, 38, 40, 43, 50, 113 – 115, 119, 131 – 132, 183 – 184, 192, 201 – 204, 213 – 216, 253, 332; negative 113; positive 113, 213, 332 t-units 265, 267 – 271, 275 – 279, 281, 334 – 335, 338 – 339, 341

UG parameters: metrical parameters 39; see also pro-drop parameter UG principles: economy principle 184, 196, 198; merge-over-move 181, 184, 196; procrastinate 184 ultimate attainment 332 universal grammar (UG) 184 universal processing accounts 215 – 216; see also processing universal sonority scale principle 2, 113; see also sonority scale/sequencing principle Urdu 189 usage-based accounts 1, 9, 11 – 12, 32 variation: in communication style 305; cultural 411; dialectal 367; intercultural 303 – 304; in levels reflection 387, 394, 397; phonological 41, 46, 72 – 73, 87 – 88, 104, 125; rhetorical 226; semantic 229; in social networking 305; spelling 271, 281; stylistic 273; syntactic 201, 205 – 208, 212, 214 – 215, 237; vocabulary and lexical 149, 272; see also systematic variation video-based learning 390 vocabulary learning 2, 139 – 143, 146, 149 – 153, 157 – 160, 162 – 163, 166, 169 – 170; see also lexical learning vocalic template see morphology voice 3, 251 – 252, 254 – 258, 260 – 262 voicing see consonants vowel contrasts 43, 45, 57 vowel epenthesis 113, 121 – 131 vowel insertion 113, 121 – 131 vowels 5, 13, 18, 31, 40 – 43, 45 – 48, 57, 74, 97, 101, 105, 115 – 117, 120, 136; back 41; central 41; duration 39, 42 – 43, 46 – 47, 50, 94 – 95; front 41, 45, 118; length 18 – 20, 40, 42, 46; long 18 – 19, 42, 46, 369; pharyngealized 58; plain 58; quality 18, 40, 42; short 18, 53, 240; stressed 43, 45 – 46; unrounded 44, 50; unstressed 39, 41, 50 word association model 363 – 364, 366, 376 – 377 word order 2, 201 – 216, 220n8 working memory see memory zone of proximal development 389

427