Incomplete Acquisition in Bilingualism: Re-examining the Age Factor (Studies in Bilingualism, Volume 39)

Incomplete Acquisition in Bilingualism

Studies in Bilingualism (SiBil) The focus of this series is on psycholinguistic and sociolinguistic aspects of bilingualism. This entails topics such as childhood bilingualism, psychological models of bilingual language users, language contact and bilingualism, maintenance and shift of minority languages, and sociopolitical aspects of bilingualism.

Editors Kees de Bot

University of Groningen

Dalila Ayoun

University of Arizona

Editorial Board Michael Clyne

University of Melbourne

Kathryn A. Davis

University of Hawaii at Manoa

Joshua A. Fishman Yeshiva University

Francois Grosjean

Université de Neuchâtel

Thom Huebner

San José State University

Georges Luedi

University of Basel

Christina Bratt Paulston University of Pittsburgh

Suzanne Romaine

Merton College, Oxford

Merrill Swain

Ontario Institute for Studies in Education

G. Richard Tucker

Carnegie Mellon University

Wolfgang Klein

Max Planck Institut für Psycholinguistik

Volume 39 Incomplete Acquisition in Bilingualism. Re-examining the Age Factor by Silvina A. Montrul

Incomplete Acquisition in Bilingualism Re-examining the Age Factor

Silvina A. Montrul University of Illinois at Urbana-Champaign

John Benjamins Publishing Company Amsterdam / Philadelphia

8

TM

The paper used in this publication meets the minimum requirements of American National Standard for Information Sciences – Permanence of Paper for Printed Library Materials, ansi z39.48-1984.

Library of Congress Cataloging-in-Publication Data Montrul, Silvina.   Incomplete acquisition in bilingualism : re-examining the age factor / Silvina A. Montrul.       p. cm. (Studies in Bilingualism, issn 0928-1533 ; v. 39) Includes bibliographical references and index. 1.  Language acquisition--Age factors. 2.  Language attrition. 3.  Bilingualism.  I. Title. P118.65.M66    2008 401'.93--dc22 isbn 978 90 272 4175 7 (Hb; alk. paper)

2008027332

© 2008 – John Benjamins B.V. No part of this book may be reproduced in any form, by print, photoprint, microfilm, or any other means, without written permission from the publisher. John Benjamins Publishing Co. · P.O. Box 36224 · 1020 me Amsterdam · The Netherlands John Benjamins North America · P.O. Box 27519 · Philadelphia pa 19118-0519 · usa

Table of contents

Acknowledgements chapter 1 Foundations 1. Some outcomes of language acquisition  1 2. The critical period hypothesis  9 2.1 Cases of linguistic deprivation  12 2.2 Delay of language acquisition in the deaf  14 2.3 Creole genesis  15 3. Bilingual acquisition  17 4. Incomplete acquisition  20 5. Organization of the book  22 chapter 2 Second language acquisition 1. Characteristics of adult second language acquisition  27 2. Process and outcome  30 2.1 Prior language knowledge  30 2.2 Persistent L1 influence as fossilization  34 Is there a critical period in L2 acquisition?  41 3. The critical period position  43 3.1 3.2 The no critical period position  49 3.3 Summary  57 4. Age of acquisition and first language loss  59 chapter 3 First language attrition in adults 1. Language attrition  64 1.1 Factors contributing to attrition  65 1.2 Linguistic process and outcome of attrition  65 1.3 Baseline knowledge in language attrition research and other methodological considerations  68

ix 1

27

63

vi

Incomplete Acquisition in Bilingualism

2.

3. 4.

Theoretical models of attrition  69 2.1 The regression hypothesis  69 2.2 The generative approach  75 2.3 The activation threshold hypothesis  81 Cross linguistic influence (transfer) in L1 attrition and L2 acquisition  85 The age factor in L1 attrition  89

chapter 4 Bilingualism in early childhood 1. Types of bilinguals  94 1.1 Simultaneous bilingualism  94 1.2 Sequential bilingualism  97 2. Unbalanced development in early childhood  99 2.1 Language choices  100 2.2 The weaker language  102 3. Attrition or incomplete acquisition?  107 3.1 L1 attrition  109 3.2 Incomplete L1 acquisition  111 3.2.1 Morphosyntax  111 3.2.2 Syntax-pragmatics interface  117 4. Total L1 loss?  120 On the nature of the weaker language  123 5. 5.1 The Weaker Language as L2 Hypothesis  123 5.2 The Weaker Language as L1 Hypothesis  126

93

chapter 5 131 Bilingualism in middle and late childhood 1. Language learning at school  132 1.1 Monolingual children  132 1.2 Bilingual children  136 2. L1 attrition in minority-language speaking children  139 2.1 Inuktitut-speaking children in Eastern Canada  140 2.2 Spanish-speaking children in American schools  141 2.3 East Asian children and the role of community-based schools  147 2.4 Children with no academic support of the minority language  150 3. Age effects in L1 loss as a function of L2 acquisition  152 4. Further research  158



Table of contents vii

chapter 6 Incomplete L1 acquisition in adults 161 1. Linguistic profile of adult heritage speakers  162 1.1 L1 attrition vs. incomplete L1 acquisition  163 1.2 Methodological considerations in the study of incomplete L1 acquisition in adults  165 2. Structural characteristics of the incomplete adult L1 grammar  167 2.1 Phonology  168 2.2 Inflectional morphology  171 2.2.1 Nominal morphology: Case and gender agreement  171 2.2.2 Verbal morphology: Agreement, tense, aspect and mood  174 2.3 Syntax  182 2.4 Lexicon  187 3. Incomplete L1 acquisition and age of onset of L2 acquisition  193 3.1 Incomplete acquisition in simultaneous vs. sequential bilinguals  194 3.2 L1 Loss in adult foreign adoptees  202 4. Conclusion  205 chapter 7 207 Incomplete L1 and L2 acquisition in adults 1. Theoretical accounts of incomplete L2 acquisition  208 2. Some similarities between L2 and L1 incomplete acquisition  211 Heritage language acquisition  216 3. 4. Research findings  222 4.1 Selective advantages in grammatical knowledge  222 4.2 Written vs. oral tasks and types of knowledge  231 4.3 Reactivity to classroom instruction  240 5. Do heritage language learners have advantages over L2 learners?  247 chapter 8 Implications 1. Complete vs. incomplete acquisition  250 1.1 First language acquisition  251 1.2 Second language acquisition  255 1.3 Early bilingual acquisition  258

249

viii Incomplete Acquisition in Bilingualism

2. 3. 4.

The role of age in language acquisition and language loss  262 The role of input in language acquisition and language loss  269 Future research  275

References Index of authors Index of terms

277 303 309

Acknowledgements

As I was working on my previous book The Acquisition of Spanish (2004, John ­ enjamins) and conducting experimental work on adult Spanish heritage speakB ers, the idea for this book was beginning to shape. But materialization of that thinking could not have taken place without the help of Kees Vaes from John ­Benjamins, Kees de Bot and Dalila Ayoun, editors of the series Studies in Bilingualism, the University of Illinois at Urbana-Champaign, my home departments – the Department of Spanish, Italian and Portuguese and the Department of Linguistics – my students, my colleagues, my friends, and my family. It gives me great pleasure to take this opportunity to express my sincere gratitude to all. It was Kees Vaes’s excitement and interest in this project (upon hearing about my preliminary idea for this book in Amsterdam in August 2002) that gave me great motivation to get started on it right away. Kees de Bot favorably reviewed my proposal for this book and extended a contract in December 2005. I was then very fortunate to receive a competitive award from the Center for Advanced Study at the University of Illinois, which released me of my teaching duties for the spring semester of 2005 and provided me with the time to start writing this book. That same year, a Beckman Award from the University of Illinois Campus Research Board for projects of special merit and distinction funded most of the experimental work that forms the basis of Chapters 6 and 7. My sabbatical leave from the University of Illinois in the fall of 2006 allowed me to complete most of the manuscript. I am privileged to have the support of such a first-rate research institution. During the writing of this book, I had the opportunity to teach two graduate seminars (spring 2004 and fall 2007) exploring the differences and similarities between second language acquisition and first language loss. I thank all the graduate students (more than 30!) whom I had the privilege to teach and share my ideas with, and I thank them all for stimulating discussion that helped clarify my thinking and my writing. I also thank them all for their feedback on a preliminary version of this book. Many of my colleagues and friends at the University of Illinois and at other universities have also played a fundamental part in the completion of this book. Monika Schmid (University of Groningen), Roumyana Slabakova (University of Iowa), Tania Ionin (University of Illinois at Urbana-Champaign) and two ­anonymous reviewers from John Benjamins read the first version of the book



Incomplete Acquisition in Bilingualism

manuscript, and provided excellent suggestions for revision. Elabbas ­Benmamoun (University of Illinois at Urbana-Champaign), Melissa Bowles (University of ­Illinois at Urbana-Champaign), Rebecca Foote (Michigan State University), and Zsuzsanna Fagyal (University of Illinois at Urbana-Champaign), as well as my graduate students Silvia Perpiñán and Brad Dennison, read the revised manuscript version and helped me tremendously with their wisdom and eye for detail. I also want to thank Maria Polinsky (Harvard University) for inspiring me and encouraging me to continue working on heritage language issues. I am deeply grateful to Marc Thompson, my husband, for his stellar editorial work and for teaching me how to be a better writer. All remaining errors are my own. And last, how could I have conceivably completed this work without the implacable patience, support, and understanding from my family, especially when I had to work long hours on summers or spring breaks? I owe tremendous debt to my two children, Lea (10 years old), Olivia (4 years old), my husband Marc, and my family in Mar del Plata, Argentina. Because this book is in some sense about them, I dedicate it to all of them. 

Champaign, Illinois, April 2008

chapter 1

Foundations

This book is about non-native like attainment, and its purpose is to re-examine age as a determining factor in non-native outcomes. Because age effects have played a particularly prominent role in some theoretical perspectives on second language acquisition, in this book I re-examine these theories in light of the existence of apparently similar non-native outcomes in adult early bilinguals who, unlike adult second language learners, acquired two or more languages in childhood. Toward this end, I show that many of the characterizing features of adult second language acquisition are also true of many language minority-speaking bilinguals or heritage speakers who either never fully acquired, or lost, aspects of their first language sometime in childhood. I highlight how age of acquisition – a macro-variable that subsumes other interrelated factors (maturational state, biological age, cognitive development, degree of first and second language proficiency, amount of first and second language use, among others) – is related to the type of linguistic knowledge and behavior that emerges in the two languages under different environmental circumstances. By underscoring age of acquisition as a unifying factor in the study of L2 acquisition and L1 attrition in bilingualism, my main claim is that just as there age effects in L2 acquisition, there are also age effects, or even perhaps a critical period, in L1 attrition. To get some idea of the complexity and variability of bilingual outcomes in the first and second language, consider the following specific examples of possible profiles in adults.

1.

Some outcomes of language acquisition

Meet Kevin, a 25 year old male who was born and lived all his life in a small city in England, in a predominantly English-speaking environment. Kevin attended elementary and secondary school in English. In college, he majored in history, but also took Spanish and French as foreign languages. His command of English is native, while his linguistic ability in Spanish and French is very basic and less proficient than when he was in college three years ago.



Incomplete Acquisition in Bilingualism

Carolina, age 35, was born and schooled in Venezuela and came to pursue graduate studies in biology in the United States when she was 24. She started learning English as a foreign language in Venezuela at 13, and by the time she came to the United States, she could speak and understand English very well. However, while her Spanish skills are native, even after 11 years of living in the United States and interacting with many English-speakers, Carolina still has a heavy foreign accent in English and makes grammatical errors when she speaks and writes. Francesco, a native of Italy, immigrated to Argentina when he was 30 years old and has been living there for 15 years. His wife, whom he married two years after immigration, and his 3 children are from Argentina. Spanish is the language spoken at home, since neither his wife nor his children speak Italian. Francesco also speaks Spanish exclusively at the bank where he works. After so many years of living and working in Argentina, Francesco’s written and spoken Spanish are excellent, and most Argentines consider Francesco a native Spanish speaker. He has no foreign accent and hardly ever makes lexical or grammatical mistakes. Francesco lives in a town where there is no Italian-speaking community, and only uses Italian when he calls his family in Italy every month. Although Francesco speaks Italian fluently because he is a native speaker, his family tells him that he speaks Italian with a foreign accent. In fact, when he speaks Italian, Francesco sometimes has difficulty finding the right words and, on occasion, uses Spanish words inadvertently. These three hypothetical individuals were each monolingually raised and schooled in their countries of origin, and became native speakers of English (Kevin), Spanish (Carolina), and Italian (Francesco). They all have full command of their first language. Kevin, Carolina and Francesco have command of a second language as well, which they learned later in life, after the foundations of their first languages were in place: Kevin learned Spanish and French, Carolina English, and Francesco Spanish, but they all attained different degrees of oral and written proficiency in these languages: Kevin has basic knowledge of Spanish and French, Carolina has advanced proficiency in English, while Francesco passes as a native speaker of Spanish. This sharp difference in general mastery of the first and second language(s) by adults – full L1 acquisition but incomplete L2 acquisition – constitutes one of the tantalizing mysteries and theoretical problems guiding a great deal of research in contemporary second language studies. But is incomplete acquisition unique to the adult L2 acquisition situation, or is it also possible to have different degrees of incomplete knowledge of the L1? What are the linguistic characteristics of L1 and L2 incomplete acquisition, and what factors contribute to them? Does age play a role in incomplete L1 acquisition as well? Or is it just



Chapter 1.  Foundations

input? Before addressing these questions, take a few minutes to consider these five more other bilingual profiles. Kristi, 38 years old, was born in China and adopted by an English-speaking Canadian family when she was two years old. She has lived in Canada ever since. She went to English-speaking schools and received some instruction in French as well. Although she heard and learned some words in Chinese as a first language before adoption, she did not interact with Chinese-speakers after adoption. Kristi claims to have no recollection of Chinese. She considers English her native language, and she speaks it like a native Canadian. Elena is 24 years old and lives in France, where she was born. Her mother is Russian and her father French. When she was a child, her mother spoke to Elena exclusively in Russian while her father used French with her. By the time Elena started French-speaking pre-school, she had a balanced command of both French and Russian. However, most of her school friends spoke French. Gradually, Elena started to use French more often at home, even when her mother spoke Russian to her, and for several years she did not speak much Russian. Now at 24, Elena is very fluent in French, but her Russian did not develop to the same level as her French. Although she can presently understand and speak Russian with some proficiency, she makes many grammatical mistakes in the language. Carlos, Alicia, and Beatriz are siblings and live in the United States. They were all born in Northern Mexico and immigrated to the United States with their parents twenty years ago, when they were 9, 4 and 2 years old, respectively. Carlos attended three years of elementary school in Mexico. Upon arrival in the United States, he was enrolled in an English-speaking elementary school and spent the first two years in ESL classes because he did not speak any English when he arrived. Alicia spoke Spanish, and Beatriz was still learning it when they arrived in the United States. Alicia and Beatriz were enrolled in full-time English-speaking day-care and pre-school soon after arrival, since the two parents had to work. At home, the parents spoke Spanish with each other and to the children. In two years, the sisters were very fluent and dominant in English, while Carlos was still struggling to learn it at school. Carlos used Spanish a lot with his parents, while Alicia and Beatriz preferred to speak English among themselves and gradually spoke Spanish less with their parents. Four years later, Carlos was also speaking English to his sisters, and the parents began to use more English at home. Spanish, however, never stopped being used in the family. Now in adulthood, Carlos is very fluent in Spanish and English, even though he uses English at school and at work. He still communicates in Spanish with his parents and can use Spanish at work. Alicia and Beatriz, by contrast, only use Spanish occasionally in their daily lives with their parents and family, and prefer to use English with friends, at





Incomplete Acquisition in Bilingualism

school, and at work. They both feel English is their native language and Spanish is like a second language. Unlike the earlier examples of Kevin, Carolina and Francesco that illustrated successful cases of L1 acquisition and variable outcomes in L2 acquisition acquired later in life, the cases of Kristi, Elena, and Carlos and his siblings represent variations of the opposite situation: an L2 acquired early in life that reaches nativelike attainment in adulthood, while linguistic control of the L1 varies dramatically depending on age of acquisition of the L2 and the circumstances surrounding acquisition. In light of these other possible outcomes of early bilingualism, or when the acquisition of two or more languages takes place in childhood, I re-examine the claim that maturational constraints on the language faculty apply mostly to the adult L2 acquisition situation. I argue and demonstrate that maturational constraints play an even more decisive role in cases of L1 loss in early bilingualism than in adult L2 acquisition. A well established fact of language acquisition is that normally developing monolingual children succeed in acquiring the basic grammar of their environment (their native language or L1) in a relatively short period of time, typically 3–4 years (Crain & Lillo-Martin 1999; Guasti 2002; O’Grady 1997; Snyder 2007). Before they begin school, and without receiving any instruction, children master the basic structure of their native language, including its phonology, morphosyntax, semantics and some aspects of pragmatics and sociolinguistic conventions. Language acquisition in general is characterized as uniform because children exposed to the same language, or dialect, converge on the grammar of other members of their speech community despite variations in input. To give one example of what linguistic uniformity means in this context, mature native speakers of English perceive the ungrammaticality of the question *What did you ask who Patricia gave? and intuitively know that when there are two wh-phrases in a sentence, they cannot move the second phrase to form a question. This is the whisland constraint, which L1 acquiring children obey quite early as well (Guasti 2002). The outcome of L1 acquisition is successful, since all normally developing children (with no physiological or mental impairment) eventually master their native language completely and reach native-like attainment. Complete acquisition by a certain age is not an error free process, however. During development, language use during acquisition is prone to errors and may be deemed incomplete. In all languages, there are well-documented developmental stages in different areas of linguistic knowledge, such that some structures and sounds are mastered . Of course, there are other bilingual profiles in addition to the ones portrayed here, such as balanced and unbalanced bilinguals living in a bilingual society. Individuals of this type can be found in Canada or areas of Europe, where one or more official languages are spoken.



Chapter 1.  Foundations

earlier than others. Nonetheless, incomplete acquisition during L1 development is temporary and disappears in due course. With the linguistic foundations of the language and the essentials of native speaker competence in place by the age of 3, language acquisition continues beyond this early period. In cultures and societies where children go to school, around age 4, children’s metalinguistic ability develops through emergent literacy and continues at school, where children learn to read and write, expand their vocabulary, and acquire more complex structures. Exposure to rich oral and written input allows children to learn to communicate in different registers and styles, both orally and in writing. At the end of the process, children become educated, literate, adult native speakers capable of functioning in many social and professional contexts. And there are also children who learn their native language in a diglossia situation (Kaye 2001): these children acquire a colloquial variety at home and receive instruction in the standard variety at school (e.g. African American children in the United States, Brazilian Portuguese children in Brazil, children in the Arabic-speaking world). At least in a monolingual context, the phenomenon of fossilization (Lardiere 2007; Selinker 1972), “stabilization” (Long 2003), or arrested development, does not occur in normal L1 acquisition. Why is it important to establish the nature, course, and outcome of monolingual child language acquisition? In the study of second language acquisition and bilingualism, the notion of mature native speakers is crucial, even though the concept and definition of who exactly is a native speaker remains elusive (Davies 2003; Mack 1997; Paikeday 1985). Native speaker competence continues to be both the ideal and the benchmark against which second language performance is typically measured. As we will see in greater detail in Chapter 2, L2 acquisition by adults differs from L1 acquisition in several respects. One of the leading research questions in the study of adult L2 acquisition within the linguistic tradition is determining the extent to which child L1 acquisition and adult L2 acquisition are similar and different, and why L2 acquisition in particular appears behaviorally distinct from L1 acquisition. Differing from L1 acquisition, the outcome of the L2 acquisition process is, more often than not, variable and divergent, even when optimal conditions for learning the L2 in a naturalistic and instructed environment are available. Furthermore, since L2 learners bring knowledge of their first language to the acquisition of the second language, they frequently cannot overcome the influence of the L1 in some or many grammatical areas. Persistent L1 influence and fossilization lead to differential outcomes of the acquisition process, most of which fall short of the idealized native speaker model. . However, literacy is not a fundamental aspect of native-speaker competence, since there are also many cultures with unwritten languages and a sizable number of native speakers.





Incomplete Acquisition in Bilingualism

Due to this critical difference between L1 acquisition in children and L2 acquisition in adults, a “deficit” metaphor has guided a great deal of research on adult L2 acquisition since the inception of the field. Characterizing the outcome and endstate of L2 acquisition by adults as “lack of success” or “generalized failure” (Bley-Vroman 1989), research in this tradition seeks to explain the fact that adults hardly ever reach the level of uniform linguistic attainment observed in L1-acquiring children. At the heart of these child-adult differences is the idea that children are successful because they are exposed to human language during a biologically-determined period in early childhood that is crucial to their developing linguistic skills. Adults, by contrast, are typically exposed to the second language much later. Late onset of acquisition as a macro-variable would explain many of the observed child-adult differences in language learning, in addition to other obvious differences in input and exposure. This idea is known as the Critical Period Hypothesis. Motivation for this book arose from the collective findings of my own research program on early bilinguals or Spanish heritage speakers over the last six years, which led me to rethink several classic issues in second language acquisition, like the Critical Period Hypothesis, L1 influence and fossilization, the role of innate knowledge, and the role of the environment. In particular, if late onset of acquisition is one of the main reasons behind non native-like attainment in L2 acquisition, what do we make of similar non native-like linguistic achievement in adult early bilinguals? At first glance, the existence of incomplete acquisition of an early acquired language yields claims about maturational constraints that are questionable at best. But further reflection leads us to consider that a mature linguistic system can be incomplete for at least two reasons. The first is because the learning mechanisms themselves are subject to maturational constraints, and this is consistent with the Critical Period Hypothesis position. But the second reason is simply because the amount of input received was not optimal to fully develop and sustain the linguistic system in the first place, even when the exposure occurred early in childhood. Thus, age of acquisition is a necessary, but not sufficient, condition for complete acquisition. My primary assertion is that, hardly a hallmark of adult L2 acquisition, incomplete acquisition is also a possible result – but certainly not the only one – of language acquisition by children and adults in a dual language environment. The vexing question is whether, despite seemingly comparable outcomes, incomplete L2 acquisition and incomplete L1 . Indeed, this metaphor is also prevalent in fields other than strict second language acquisition. In his most recent book Language Interrupted, McWhorter’(2007) stresses the role of “incomplete acquisition” by L2 learners as the main driving force responsible for language change and simplification in standard languages at the macro-sociolinguistic level.



Chapter 1.  Foundations

acquisition are fundamentally distinct. In essence, this book opens an entirely new perspective on the question of age and ultimate attainment in adult L2 acquisition by bringing to the discussion other bilingual populations that speak directly to these issues. Although the question of age has been extensively discussed in the second language acquisition field, this book breaks new ground in addressing the role of age in both the acquisition of an L2 and the potential loss of the L1 in language minority immigrants immersed in an L2 environment. In sum, the discussion of L1 loss in the context of L2 acquisition makes this book substantially different from other classic and current treatments of age effects in L2 acquisition (cf. Herschensohn 2007, for example). While underscoring the role of age in both L2 acquisition and L1 loss, my goal is to also show that many of the seemingly peculiar features of second language acquisition or late bilingualism – such as transfer, fossilization and incomplete ultimate attainment – are also typical of other early bilingual situations. But even though the book is about incomplete acquisition, my intention is not to support or perpetuate the “deficit” view of bilingualism often endorsed among the nonacademic public. The term “incomplete” should be understood as a descriptive term, not a value judgment. Even when bilingual ultimate attainment in one or both languages may diverge from native speaker competence, and a bilingual may not be two monolingual individuals in one (Grosjean 1989, 1998), bilinguals have essentially the same type of abstract linguistic system, since many aspects of their languages also fully converge on the target. I maintain that many bilingual individuals are very competent speakers and users of two or more languages, and it is indeed possible to find many adult L2 learners and early bilinguals of minority languages who have reached native-like attainment in their L1 and L2. At the same time, these cases are not the focus of this book; instead, I focus on cases of incomplete acquisition for both conceptual and methodological reasons. My other theoretical goal, allied with the first, is to go beyond apparent similarities between early and late bilinguals and to explore potential differences between incomplete L1 and incomplete L2 acquisition as a function of age. If the essence of native speaker competence develops during a critical period and is not much affected by experience, it follows that once acquired, such basic linguistic knowledge should remain relatively stable and accessible throughout the speaker’s lifespan. Is there substantial evidence for the claim that despite incomplete acquisition in a number of grammatical areas, the nature of grammatical knowledge acquired before and after the critical period is still very different in nature? I suggest that even when an L1 may be partially lost or incompletely acquired in early bilingualism, it still retains the signatures of a first language because it was acquired during the Critical Period. This is the Weaker Language as L1 Hypothesis, which I introduce in Chapter 4. But questions such as whether, and how,





Incomplete Acquisition in Bilingualism

i­ ncompletely acquired L1 grammars differ from incomplete L2 grammars in fundamental ways, are a matter of ongoing and future investigation. Methodologically, to understand the process and outcome of bilingual acquisition and the variability that characterizes it, the most common research approach is comparison with a target monolingual norm, including monolingual acquisition in typically developing children and adult native speakers of the target language. Systematic departures from monolingual grammars may be due to the very nature of bilingualism itself, and to the cognitive demands of handling two independent linguistic systems under communicative pressure. What I also do in this book, is to introduce the methodological comparison of different types of adult bilinguals – adult L2 learners and adult early bilinguals – who differ on age of acquisition of their L1 and L2 and many other sociolinguistic and environmental circumstances. Ultimately, understanding the process and different linguistic outcomes of normal language acquisition under different environmental conditions sheds light on the nature, organization, and limits of the human language faculty, and this is the collective goal of linguists studying language acquisition and disruption under different circumstances and from different theoretical approaches. Beyond these theoretical and methodological concerns, systematic comparison of incomplete L1 and L2 speakers also bears great practical significance. It is widely recognized that the face of second language classrooms in Western Europe and North America has been changing dramatically in the past years, due largely to recent demographic patterns reflecting the waves of continuous immigration. Increasing numbers of bilingual speakers of minority languages like Elena, Carlos and his siblings wish to maintain and/or (re)learn their family language while they acquire second language skills in the majority language. As a result, traditional foreign/second language classes typically geared to students like Kevin, who have no background in the language, have increasingly been opening doors to speakers who were exposed to the language early in childhood and whose levels of oral proficiency in the language range from minimal to advanced. While much is known about how to best instruct typical second language learners, far less is known about the linguistic skills of bilingual speakers of minority languages (also known as heritage or ethnic community languages) and what types of gaps they have in their knowledge. Today, many institutions place these heritage language learners like Elena, Carlos, Alicia and Beatriz in foreign language classrooms that follow a traditional second language curriculum, while other institutions have created special courses to address the specific linguistic and cultural needs of these speakers. In order to better serve the needs of this immigrant population and help them reach their full linguistic potential in the heritage language, more needs to be unraveled about the nature of heritage language knowledge and its acquisition. Therefore, comparison



Chapter 1.  Foundations

of adult L2 learners (exposed to the L2 late) and bilingual speakers of minority languages (exposed to their L1 early) will eventually allow us to see whether there are long-lasting benefits of receiving input early in childhood in incomplete grammars, and how once in the classroom, such early acquired knowledge may react to formal instruction. Future curriculum development would benefit greatly from a basic understanding of these types of language learners (Brinton, Kagan & Bauckus 2008; Kondo-Brown 2004, 2006). In sum, by contributing to classic and current theoretical debates on the role of age and the nature of linguistic competence in second language acquisition, this book brings together the fields of second language acquisition, early bilingualism, and language attrition. In forging a link between child and adult bilingualism, I also hope to contribute significantly to the growing field of heritage language acquisition and teaching. One cannot talk about the role of age in language acquisition without making reference to the Critical Period Hypothesis, a topic that has been extensively addressed over the years but still continues to generate debate. Hence, § 2 introduces the concept of the critical period and illustrates how it has been argued to apply in L1 language acquisition. Since the book is about bilingualism and how age affects bilingual outcomes, a brief introduction to bilingual acquisition is presented in § 3. Section 4 characterizes the notion of incomplete acquisition. Finally, § 5 offers an overview of the organization of the rest of the book.

2.

The critical period hypothesis

One prominent explanation for why children acquire the ambient language relatively fast and successfully is biological. The innateness hypothesis states that human language is part of the human endowment. The universality and the early emergence of many structural principles of language in L1 acquisition have led many linguists to argue that such structural principles must be innate, as specified by the theory of Universal Grammar (Chomsky 1981; Crain & Thornton 1998; Guasti 2002; Pinker 1991). These grammatical principles emerge early, are universal, and appear without decisive evidence from the environment. The alternative to the innateness hypothesis is the learning hypothesis (also known as emergentism), according to which language acquisition is not as fast as the innatist position makes it out to be. This approach sees language as part of general cognition (i.e., cognition is innate but language is not a specialized, innate module of the mind), and learned largely from experience with the environment (­Tomasello & Bates 2001; O’Grady, in press).



10

Incomplete Acquisition in Bilingualism

Figure 1.1  Diagram of a critical period and ultimate attainment

One hallmark of biologically determined behavior, both in humans and animals, is the existence of a critical period (Bornstein 1987; Colombo 1982; Oyama 1976). The critical period refers to a temporal span, early in life, of heightened sensitivity to environmental stimuli. The phenomenon is physiological in nature and involves changes in the central nervous system during the course of development. For some researchers, critical periods refer to behaviors that are critical for species’ identification and survival (Brauth, Hall & Dooling 1991). Although no specific external event causes the behavior to emerge, if the organism does not receive sufficiently rich environmental stimuli during its critical period, then the behavior is not likely to develop properly. For Colombo (1982), a critical period must have an onset, a terminus, an innate maturational component, an extrinsic component that stimulates the innate component, and a system that is stimulated during the period. A graphic representation of what a critical period looks like is given in Figure 1.1. (See also Birdsong (2005) and Hyltenstam & Abrahamsson (2003) for other diagrams.) Classic examples are filial imprinting in birds, the development of birdsongs, visual development in kittens and monkeys, and the social behaviors of a variety of mammals (Bornstein 1987). As for critical periods in humans, Bornstein (1989) also refers to vision. Naturally, language is another likely candidate. The biological foundation for language and its relevance for the critical period hypothesis was originally proposed by Penfield (1953) and Penfield and ­Roberts



Chapter 1.  Foundations

(1959) based on neurological evidence. Due to progressive lateralization and loss of plasticity in cerebral functions, the neural substrate required for language acquisition is no longer available after the closure of the critical period: When a child begins to speak, there develops a functional specialization in one cerebral hemisphere, normally the left hemisphere . . . There are separate areas of the cortex on this, the dominant side which comes to be devoted to the formulation and understanding of speech . . . But, once functional localization of acquired skills has been established, the early plasticity tends to disappear.  (Penfield & Roberts 1959, pp. 203–206)

Studying human development, Lenneberg (1967) also subscribed to the view that the capacity of first language learning was lost if it was not activated or exercised during the critical period, the onset of which he set at age two (the two-word stage), the terminus at around age 13. Again, evidence from the hypothesis came primarily from the dramatic difference between young children and adults in their recovery after aphasia, the dissociation of language from other cognitive abilities (or the capacity of some children to acquire language successfully despite other cognitive deficiencies), and the neurological plateau that children reach at puberty. Like Penfield, Lenneberg also proposed brain lateralization as a likely cause for the loss of a language acquisition advantage at puberty. If the neurological basis of Lenneberg’s claims have had some acceptance (in addition to a fair amount of criticism), the specific age ranges for closure of the critical period or for loss of brain plasticity have been questioned. Corballis (1991) claims that the human brain is already lateralized from early infancy. But in the context of adult L2 acquisition Hyltenstam (1992) has shown that loss of linguistic ability starts as early as 6 years of age (see also Penfield & Roberts 1959 and Krashen 1973). There is even mounting empirical evidence that core aspects of language (phonology, morphology, syntax, lexicon) are differentially affected in their susceptibility to deterioration effects at different ages (Bialystok 1997; ­Bialystok & Hakuta 1999; Long 1990). This has already been captured in Seliger’s (1978) concept of multiple critical periods for subcomponents of language with different onsets (and offsets).

. If the onset of the critical period started at age 2 for Lenneberg, this raises the question of what happens before then. Do children lack language altogether before age 2? Today, we know that linguistic development begins soon after birth, if not in utero, and that children have linguistic knowledge and discrimination of some sort well before they put two words together (Sebastián-Gallés 2006;Werker & Yeung 2005). . On this distinction, see also Eubank & Gregg (1999).

11

12

Incomplete Acquisition in Bilingualism

While there may be different critical periods for different modules of the grammar in adult L2 acquisition, we do not know whether the same is true of L1 acquisition. Evidence for a critical period for language abilities in children can only be assessed indirectly because it is not ethically possible to set up an experiment whereby children are deprived from primary linguistic input during early life. Original claims about first language acquisition being inevitable and chronologically delimited come primarily from cases of pathological language loss (aphasia or other neurological insult) and eventual recovery in young children. More recent evidence for an optimal, maturationally determined period early in life that strongly favors efficient language acquisition comes from extra-societal children, deaf children, and more indirectly, the creation of creole languages.

2.1

Cases of linguistic deprivation

Perhaps the most cited case of linguistic deprivation in early childhood is Genie (Curtiss 1977), a child who was found in Southern California at the age of 14 and had lived most of her life under inhumane conditions. The daughter of an abusive father and a visually impaired mother, Genie was confined to a small room from the time she was a baby because her father could not stand to hear her cry. If she made any sounds, she was physically punished. When found at the age of 14, Genie was without language. After she was hospitalized, efforts were made to teach her language, and she began to acquire speech well after the onset and (possibly) offset of the critical period. Genie learned to speak in a rudimentary fashion, very much like a normal two year-old child. However, unlike two-year old children, Genie progressed more slowly and even stopped developing after several years of training. For example, Genie used primitive forms of negation (e.g., No mommy go) for over two years, and could not even produce the first stage of wh-question formation normally produced by children (e.g., Where teddy?). Instead, she produced deviant, incomprehensible questions, such as “Where is stop spitting?” Like her morphosyntactic development, her phonological development was equally rudimentary and abnormal, affecting segmental (sounds) and suprasegmental (stress and intonation) features (Curtiss, 1977; Curtiss, Fromkin, Krashen, Rigler & Rigler 1974). On the other hand, Genie outperformed normal children in her ability to learn vocabulary. She understood and produced many more words than ordinary children at her same level of morphosyntactic development. But since vocabulary

. For more details about her daily life see Curtiss, Fromkin, Krashen, Rigler & Rigler (1974).



Chapter 1.  Foundations

learning relies heavily on memory and association, such dissociation in linguistic development indicates that Genie could not handle abstract linguistic rules, or what generative linguists call the computational system. Thus, Genie’s case would provide evidence that the innate linguistic ability, which subserves rapid, implicit, and complete L1 acquisition, is subject to a critical period. Although, according to Curtiss (1977), Genie also suffered some brain damage in her left hemisphere, it is possible that she might have been functioning with the right hemisphere, which is not usually associated with language. Even if Genie suffered some brain damage, Curtiss showed that Genie’s mental age increased by one year for every year after she was found while she was studied, whereas her language abilities did not show a similar increase, stagnating at the level of a normal two-year old. Although this is fairly strong evidence for a critical period for L1 acquisition, one must still be cautious in drawing direct conclusions from this case, because Genie suffered severe psychological and physical abuse. Hence, a great deal of her poor linguistic abilities could be related to permanent damage to other systems. Other cases of linguistic deprivation, which are less confounded with other types of abuse, are the cases of Victor, the wild child of Aveyron (Itard 1801), and Isabelle (Brown 1958; Mason 1942), the child of a deaf mute. Victor was also found at the age of 13 years and is described as having no spoken language, although he could understand some. He was able to produce a few sounds, and his comprehension outperformed his production abilities. Isabelle was found in Ohio in the 1930s at the age of 6 and a half years. She had no speech and made croaking sounds. Isabelle and her mother spent most of their time in a darkened room. Yet, once found, Isabelle made significant progress when compared with Genie and Victor. According to Brown (1958, p. 192), “Isabelle passed through the usual stages of linguistic development at a greatly accelerated rate. She covered in two years the learning that ordinarily occupies six years. By the age of eight and a half, Isabelle was not easily distinguishable from ordinary children of her age.” The fact that Isabelle showed accelerated recovery of language, while Victor (and Genie), who were found much later in life, did not, is compelling evidence for the advantages of receiving linguistic exposure at a specific period early in life, and of how dramatic and irreversible the consequences of linguistic deprivation at an earlier age can be.

. Unfortunately, the information and linguistic evidence available from these two cases are not as extensive as in Genie’s case, which makes comparison between these cases and Genie difficult.

13

14

Incomplete Acquisition in Bilingualism

2.2

Delay of language acquisition in the deaf

More suggestive evidence for the critical period hypothesis in L1 acquisition comes from studies of congenitally deaf children of hearing parents, who are exposed to their first language – a sign language – at different ages (Mayberry 1993; Mayberry & Eichen 1991; Newport 1990). An extreme case of linguistic deprivation is that of Chelsea (Curtiss 1989), who was misdiagnosed as mentally retarded when she was a child, but when she was 31 years of age it was discovered that she was congenitally deaf. After the diagnosis, Chelsea was fitted with hearing aids, and immediately began to receive extensive language therapy. While she was able to acquire a large vocabulary, she did not reach the syntactic level of even a three-year old child. Despite the different reasons for deprivation, Gleitman and Newport (1995) compare the linguistic attainment of Genie, Chelsea and Isabelle to suggest that the difference between all these cases (some basic language in the case of Genie, no language for Chelsea, and full language for Isabelle) supports the view that a pronounced decrease in the ability to acquire a language completely must occur around age 7, or else Isabelle, who started learning at age 6 and a half, would not have been so successful. Early and middle childhood is the period in which human beings are biologically equipped to learn language effortlessly. The most compelling evidence to date for the upper bounds of ultimate attainment comes from studies of the acquisition of American Sign Language (ASL), which approximate experimental investigations. Newport’s (1990) seminal study examined three groups of congenitally deaf ASL learners (ages 35–70) who were exposed to the language at different ages: native learners first exposed to ASL in infancy (from birth to age 3), early learners between the ages of 4–6, and late learners after age 12 (puberty). All subjects had a minimum of 30 years of daily exposure to ASL. The three groups were tested on the comprehension and production of aspects of ASL syntax and morphology (basic and topicalized word order, subject-verb and object-verb agreement, verbal classifiers of motion verbs, aspect and number morphology, and derivational morphology distinguishing nouns and verbs). Results showed that the three groups performed equally well with ASL word order. However, there was a clear age effect for morphology: the native learners were very accurate, the early learners were good overall but had some errors, while the late learners’ performance on the tasks was quite deficient. Similar results were found by Mayberry (1993), Mayberry and Eichen (1991), and by Neville (1995).



2.3

Chapter 1.  Foundations

Creole genesis

Another type of evidence that points to the primacy of language acquisition and perhaps innateness during an optimal period in early life comes from the development of language-like behavior in children who lack normal linguistic input. One example is the formation of creole languages in cases where populations are displaced, as in plantation creoles in the Caribbean, the Indian Ocean and Hawaii (Bickerton 1981, 1999). When speakers of mutually unintelligible languages have to live and work in close proximity with one another for several years, they resort to a kind of (“simplified”) language called pidgin. Pidgin languages have simple syntax (no embedding, frequent ellipsis of arguments and verbs), and lexical items have impoverished inflectional morphology. Children whose parents, caregivers or interlocutors speak pidgin languages are not exposed to grammatically “rich” input. Because more complex grammatical items cannot be found in the input, as in the case of normal language acquisition, these children create a more complex grammar by recruiting lexical items and bleaching them of their normal lexical meaning (Bickerton 1988). This is akin to the process of grammaticalization by which lexical categories become functional categories (function morphemes). By making language more grammatically complex (by adding Tense-Aspect-Mood (TAM) inflectional morphology, embedding, movement of phrases, etc.) children create a creole, a language with a full natural language structure. In his famous Language Bioprogram Hypothesis, Bickerton (1981, 1984) proposed that a set of unmarked semantic oppositions (anterior/nonanterior, realis/irrealis, punctual/nonpunctual in TAM systems; specific/nonspecific; accomplished/unaccomplished, stative/nonstative) were like semantic parameters driving the emergence of morphology in creole languages. However, since speakers of creole languages have access to the lexicons of the source languages, it is not entirely clear whether the emergence of “structure” is purely biological, as Bickerton suggested, or emerges as a result of functional borrowing from the substrate languages (Bickerton 1984; Sankoff 1979). According to Newport (1999), the acquisition of signed languages is relevant for the issue of creole genesis, especially for teasing apart the influence of biologically programmed behavior from the influence of pre-existing languages in the environment, given that the latter are not an issue in this context. I mentioned in § 2.2 that deaf children of hearing parents begin their exposure to sign language at different ages, ranging from early childhood to adolescence. However, there are also deaf children whose deaf parents were late learners of signed languages. Late learners show variable degrees of incomplete acquisition of the signed language, and their production is usually replete with morphological errors. In other cases, the parents are not hearing-impaired, but are second language learners of ASL.

15

16

Incomplete Acquisition in Bilingualism

Newport compares these deaf children of late-ASL-learning parents with learners of creoles because she argues that both types of learners receive impoverished input. Newport shows that these children are capable of surpassing the linguistic abilities of their parents and are able to construct a more complex grammatical system by making grammatical distinctions more precise, regular, and internally consistent. If there are innate, language-specific constraints for learning languages, it appears that these constraints operate during a maturationally determined period early in life. An even more compelling case for the Bioprogram Hypothesis is made by Kegl, Senghas and Coppola (1999) in their account of the emergence, after the 1970s, of Nicaraguan Sign Language (Lenguaje de Señas Nicaragüense/Idioma de Señas Nicaragüense). This is a situation in which deaf people remained completely isolated for many years and first developed a system of idiosyncratic and unintelligible home signs. When deaf Nicaraguans had the opportunity to come together and socialize with each other (through schooling and social programs after the revolution), the system of individual home signs gave rise to a pidgin, with oral Spanish as the superstrate. The pidgin became the Lenguaje de Señas Nicaragüense, a full fledged creole with linguistic complexity. The results of Kegl and collaborators’ studies on NSL contribute, in their view, new and direct evidence in support of the critical period. Young acquirers of NSL (younger than 7) were able to go beyond the input received and grammatically improve upon it. Older signers (older than 7 years of age – younger than 16 years of age), on the other hand, were able to communicate with each other and agreed on some common signs, but unlike the younger signers, were unable to use their innately determined blueprint to surpass the linguistic model. To summarize, an optimal period for first language acquisition is empirically motivated by a variety of indirect evidence from neurology, extrasocietal children, creole genesis, and deaf populations. Existing case studies point to the conclusion that, if linguistic input and socialization occur before the ages of 6–7, chances of developing a full linguistic system are remarkably good (for example, Isabelle and the young signers). If, on the other hand, exposure to natural language becomes available after age 6–7 and before 13, some rudiments of language can be acquired, although acquisition appears to be incomplete (for example, Genie and the late signers). Finally, if input becomes available well after puberty, a first language has no chance of development (for example, Chelsea). In ­conclusion, . Although Kegl, Senghas & Coppola (1999) put it this way, creole genesis is not direct evidence for a critical period after which language acquisition is impossible. Rather, it is indirect evidence for the innateness of language and for an optimal time in early childhood for language learning.



Chapter 1.  Foundations

there is strong evidence for maturational effects in L1 acquisition, with some irreversible consequences if language is not heard or seen (in the case of signed languages) before puberty (cf. Vargha-Khadem et al. 1997). Because the Critical Period Hypothesis has been extended to explain the extent of linguistic attainment in second language acquisition, and because this book is about the role of age in bilingual outcomes, the next section introduces bilingual acquisition.

3.

Bilingual acquisition

Broadly defined, bilingualism refers to knowledge and command of two or more languages, albeit to different degrees. Due to the variety of factors that define dual language speakers and hearers (some of which do not play a role in monolingual acquisition) bilingualism comes in many shapes and sizes (Grosjean 1998). Two common parameters that distinguish bilingualism are (1) age of acquisition (early in childhood versus late after puberty), and (2) order or sequence of acquisition in childhood (two languages being acquired simultaneously versus one language being acquired successively, after the other). Even though second language acquisition is treated as a separate field of study, it is a particular case of bilingualism: early (with children) or late (with post pubescent and adults) L2 acquisition. Figure 1.2 illustrates the types of bilingualism discussed in this book. As Figure 1.2 illustrates, early bilingualism takes place before puberty and can be simultaneous or sequential. Simultaneous bilingualism occurs in early childhood, before the linguistic foundations of the languages are in place. It is also referred to as bilingual L1 acquisition (Genesee 2000; Meisel 2001) because the two languages develop together as first languages (two L1s). Sequential bilingualism, on the other hand, happens after the individual has acquired basic command of the first language, which for monolingual acquisition is typically taken to be roughly the age of 3–4. In this situation, there is a first language (L1) and a second language (L2) sequentially ordered. Sequential bilingualism can occur early, during childhood; or late, in adulthood. Early sequential bilingualism is equivalent to child L2 acquisition in the L2 acquisition field. Early child L2 acquisition probably spans about 2 years and occurs between the ages of 4–6, when spoken language is practically fully developed but the children have not yet received formal schooling. Late child L2 acquisition spans the elementary school years, when children are receiving formal instruction in one or in the two languages (depending on their sociolinguistic circumstances). Late sequential bilingualism is adult L2 acquisition. In this situation, the L1 has been acquired completely and, with the exception of vocabulary size which can increase or decrease depending on ­domains

17

18

Incomplete Acquisition in Bilingualism

Bilingual Acquisition

Simultaneous (birth-age 3)

Sequential (= L2 acquisition)

Early (= child L2 acquisition) (childhood)

Early child L2 acquisition (4–6: pre-school)

Late child L2 acquisition (7–12: elementary school)

Late (= adult L2 acquisition) (post puberty and adulthood)

Figure 1.2  Types of bilingualism by age and sequence of acquisition

of use throughout the lifespan, the L1 syntax and phonology are assumed to remain stable throughout adulthood. Two other parameters along which bilinguals vary greatly are proficiency and relative balance in the two languages. Proficiency and balance are related to the degree of ultimate attainment in the two languages. Although balanced proficiency (with native-like command of two languages) is a possible outcome of bilingualism, it is indeed very rare (Grosjean 1989, 1998). The reality is that most bilinguals are linguistically unbalanced, both functionally (in their language use) and representationally (in their linguistic knowledge). Bilingual speakers typically possess a stronger and a weaker language. The stronger language is more native-like than the weaker language. However, the relative strength of the two languages fluctuates along the lifespan depending on a variety of factors, such as age and order of acquisition, as well as what language is most often used and preferred in the community (i.e., the sociolinguistic status of the languages as majority or minority languages), and the contexts for use of each language (home, church, school, work, etc.).



Chapter 1.  Foundations

Research on late bilingualism or L2 acquisition by teenagers and adults is very extensive and is treated as a separate field of study. Since the 1990s, much research has been conducted on the process of simultaneous bilingual acquisition, while, unfortunately, research on early sequential bilingualism or L2 acquisition by children (4 to 11 year olds) is still relatively scarce in the linguistic and psycholinguistic literature. A common approach to studying the process and outcome of adult second language acquisition is to trace parallels with monolingual acquisition by children. Over the years, much of this research has shown that, with few exceptions, adult L2 learners rarely achieve the degree of linguistic success of normally developing monolingual children, falling short of native speaker performance and competence in specific areas of phonology, morphology, syntax, semantics and pragmatics. Throughout the process of acquisition, L2 learners make both developmental errors, like L1 learners, and transfer errors due to influence from their L1, especially at early stages. A key difference between L1 and L2 acquisition, however, is that while child L1 learners overcome developmental errors without need for instruction, L2 learners continue to make many errors, even after receiving formal instruction, practice and correction. Fossilization can occur at any point in L2 development. Thus, unlike L1 acquisition by children, adult L2 acquisition is typically characterized as less normal and natural. It is not universal, simply because not everybody learns a second language. In terms of outcome, L2 acquisition is variable because L2 learners do not uniformly attain the same level of linguistic competence in the second language. Furthermore, L2 acquisition is typically incomplete – most learners never reach the competence of a native speaker. Although some researchers argue that attainment of full linguistic competence in the L2 is possible, in principle, it is by no means guaranteed. The nature and outcome of adult L2 acquisition will be explored in more detail in Chapter 2. A variety of factors, and even their combination, could potentially account for the seemingly disparate outcomes of child L1 and adult L2 acquisition, like cognitive abilities, type of exposure to input, degree of socialization, previous linguistic experience, and so on. Yet the one encompassing factor singled out as most significant in predicting outcomes of acquisition and eventual success is age of acquisition. The Critical Period Hypothesis has featured prominently as a likely explanation for why adult second language learners rarely reach the level of linguistic ability of native speakers (Bley-Vroman 1989, 1990; Johnson & Newport 1989,1991; Long 1990; Schachter 1990). Following Schachter (1990) and others, non-nativelike attainment is what I term in this book incomplete ­acquisition. . This does not mean that all L2 learners are incapable of achieving native-like ability. Some certainly do, but again, these individuals are not the focus of this book.

19

20 Incomplete Acquisition in Bilingualism

­ owever, as we will see in Chapter 2, empirical support for or against the existence H of a critical period in L2 acquisition is not categorical, and this issue continues to generate much debate (Birdsong 1999, 2005, 2006; Bongaerts 2005; Singleton & Lengyel 1995).

4.

Incomplete acquisition

Working within the perspective of Universal Grammar, Schachter (1990) (following Bley-Vroman 1989 and others) proposed the Incompleteness Hypothesis to characterize late bilingualism and its non-equivalence to L1 acquisition. Simply stated, an L2 will not develop some linguistic property x completely if L2 input is received after the L1 is fully formed. Further refining this idea, Sorace (1993) distinguished between two possible non native-like outcomes in late bilingualism: incomplete and divergent representations. For Sorace, the incomplete L2 grammar lacks a given property x of the target grammar, just as Schachter proposed. As a result, L2 learners have no representation of property x in their interlanguages, and when required to use property x (in production, comprehension or judgment) they may use it inconsistently and probabilistically. A divergent grammar incorporates a version of the property x in the interlanguage (probably because the L1 grammar has a different instantiation of that same property), but linguistic behavior with respect to the property x in the target grammar is consistently different from that of native speakers. Sorace assumes that both incomplete and divergent representations are due to maturational effects. While incompleteness (and other versions of non native-like attainment) is an undeniable outcome of late bilingualism, it is questionable whether this outcome is only due to the existence of a critical period, as has often been maintained. Specifically, I argue and demonstrate in this book that incompleteness is also a possible feature of early bilingual grammars – both L1 and L2 – and of a language acquired within the critical period. Fossilization in one of the languages (and possibly even in both) is likely when rich input in the language becomes reduced or is completely interrupted before the closure of the critical period. This is typically the case of immigrant children and adults who begin to use their family language less as they embrace their new L2-speaking society. In this particular sociolinguistic situation, the family language is the minority language and the host country language the majority language. Very often, the L1 (family, heritage or ethnic community language) remains the home language, while the children are educated in the L2 – the majority language. Adults also learn the L2 to different degrees to become fully functional members in the new society.



Chapter 1.  Foundations

In my view, incomplete L1 acquisition and L1 attrition are specific cases of language loss across generations. What I broadly refer to as incomplete acquisition (for lack of a better term), is a mature linguistic state, the outcome of language acquisition that is not complete or attrition in childhood. Incomplete L1 acquisition occurs in childhood when, for different reasons, some specific properties of the language do not have a chance to reach age-appropriate levels of proficiency after intense exposure to the L2 begins. In the bilingual cases I discuss in this book, incomplete acquisition which started in childhood persists well into adulthood. Although L1 attrition can also occur in childhood, I consider attrition as the loss of a given property y of the language after property y was mastered with native-speaker level of accuracy and remained stable for a while, as in adults. It is important to clarify that attrition and incomplete acquisition in childhood are not mutually exclusive, since both processes can occur simultaneously or sequentially for different grammatical properties. For example, a child of a given age can exhibit incomplete acquisition of property A if the property was present in the input, emerged in the linguistic competence of the child, and became somewhat productive but was never fully mastered by that age, while at the same time, property B (also present with some frequency in the input) may have undergone attrition, if property B was fully mastered at an earlier age but is either produced/understood with a high degree of error or is not present at all at the current age. And one can also imagine that property A was incompletely acquired (not fully mastered) at age 4, for instance, but produced/understood with much higher error rates or not present at all a few months later.10 That is, since linguistic competence is apparently fragile before a certain age in childhood, we can also conceive of attrition and regression of an incompletely acquired property. There are also cases when a child may not develop a certain grammatical form because that form is simply not present in the input the child is exposed to. This may be common in situations when monolingual children acquire a variety at home and another one at school, but if bilingual children do not receive schooling in the standard language, they will also end up with incomplete or no knowledge of some properties of the adult grammar in educated speakers.11 Hence, to reiterate, 10. The best way to tease apart all these processes is by conducting longitudinal studies of children, some of which I discuss in Chapter 5. When we deal with adults who have incomplete knowledge of their L1, as in Chapters 6 and 7, it is impossible to determine whether the incomplete acquisition these heritage speakers manifest in adulthood is the result of one or all of these processes in childhood. 11. For examples of this situation see current work on language change and native speaker competence in Brazilian Portuguese (Kato, Cyrino & Corrêa, in press; Pires 2005; Pires & Rothman, in press; Rothman 2007).

21

22

Incomplete Acquisition in Bilingualism

incomplete acquisition occurs primarily in child bilingualism (simultaneous or sequential depending on how early or late a given structure is typically acquired in childhood), while attrition can occur in both child and adult bilingualism – if it can be shown empirically that the property in question existed and has been mastered at an earlier stage. In summary, the weaker language of adult bilinguals can be the result of developing or interrupted L2 acquisition, as in adult L2 learners, or of both L1 attrition and incomplete L1 acquisition, as in many bilinguals who speak a minority language. It is true that child L2 learners are capable of attaining full linguistic competence in their L2 when they become adults; what remains an open question is whether they are also capable of maintaining full competence in their L1 or family language. Throughout this book, I show how degree of incompleteness and L1 attrition in bilinguals are inversely related to age of acquisition and onset of bilingualism, although age alone cannot be the only factor responsible for these outcomes. I argue that the critical period is highly relevant when it comes to L1 maintenance and loss as well. In essence, I will demonstrate that there is a tradeoff in the linguistic competence of maintaining two languages, as has been recognized by other researchers in the field of bilingualism (Cook 2003; Flege 1999; Grosjean 1998). Granted that late L2 learners are less likely than early L2 learners to acquire complete nativelike linguistic competence and overall proficiency in the L2, they are at the same time, very likely to maintain full linguistic competence and proficiency in their L1. Conversely, while child second language learners may be more likely to attain native-like proficiency in the L2 as adults, they are very likely to lose linguistic ability in their first language, even to the extent of becoming monolingual in the L2. Many late bilinguals are able to maintain a high level of bilingualism. As with L2 acquisition, socio-affective factors like motivation, language identity, education, and peer pressure play a significant role in language maintenance and loss in children, but less so in adults.

5.

Organization of the book

The heart of this book is the discussion of incomplete acquisition in adult early bilinguals (heritage speakers) and how their linguistic competence compares to that of adult L2 learners, the focus of Chapters 6, 7 and 8. To understand the linguistic competence of this group, it is important to first contextualize it within L2 acquisition and early bilingualism from childhood to adulthood. Therefore, Chapter 2 and Chapter 3 are foundational: they set the stage to frame, and eventually compare, the linguistic characteristics of L2 learners with those of adult heritage



Chapter 1.  Foundations

speakers who have undergone attrition and incomplete acquisition in childhood. Chapter 2 discusses key features of incomplete ultimate attainment in adult L2 acquisition, cases like Kevin and Carolina, and samples the empirical evidence in favor of the Critical Period Hypothesis for L2 acquisition. This chapter ends with an extension of the Critical Period Hypothesis for L1 attrition, which I formulate in two hypotheses: one contrasting child and adult attrition, the other attrition in simultaneous and sequential bilingualism in childhood. Chapter 3 examines the flip side of the L2 acquisition issue: the potential language loss or L1 attrition in first generation immigrants who are late learners, very proficient in the L2, and have been living in the L2 environment for an extended period of time, like Francesco. By focusing on adults, these two chapters illustrate that just as it might be difficult to master a second language like a native speaker after a certain age, it is also highly implausible for an adult to lose his first language significantly. Because this book traces L1 attrition from childhood to adulthood, I suggest that the effects of L1 attrition in a variety of grammatical areas are really minor to negligible in adult late bilinguals, as opposed to the more dramatic effects language attrition or incomplete acquisition can have in children (early bilinguals). But we cannot understand the processes leading to incomplete acquisition in the adult early bilinguals discussed in Chapters 6 and 7 without examining closely what happens in childhood. Therefore, Chapter 4 and Chapter 5 trace early bilingualism from childhood to adulthood, discussing research on bilinguals like Kristi, Elena, Carlos, Alicia and Beatriz. Chapter 4 presents an in-depth characterization of early bilingual L1 acquisition and incomplete acquisition in language minority-speaking children (typically second generation immigrants). This chapter also introduces the Weaker Language as L1 Hypothesis, which I defend in the rest of the book. Chapter 5 examines heritage language maintenance and loss throughout the elementary school years, and points to middle childhood as a significant age for linguistic maturity. These two chapters show how childhood is a very vulnerable time for both L2 acquisition and L1 loss, and how children are more reactive to fluctuations in input than the adults discussed in Chapters 2 and 3. The studies discussed in these chapters also provide support for the hypothesis that L1 attrition is likely to be more severe in simultaneous than in sequential bilinguals. Because it is difficult to reconstruct the linguistic past of adult early bilinguals by relying on self-reports, these two chapters are crucial to understand the long-term effects of incomplete acquisition and attrition in childhood, the specific focus of Chapter 6. Incorporating results of my recent research on this topic, Chapters 6 and 7 discuss incomplete knowledge of the L1 in a variety of grammatical areas. Chapter 6 focuses on adult heritage speakers only, and stresses how incomplete acquisition leads to different degrees of proficiency in the heritage language, with retention of

23

24

Incomplete Acquisition in Bilingualism

core properties of the L1 acquired early in childhood and concomitant simplification and reanalysis of other structural properties. Furthermore, Chapter 6 considers degree of incomplete acquisition and simplification among simultaneous and sequential early bilinguals as a function of age of onset of bilingualism, stressing once again that there are age effects in L1 loss. Chapter 7 compares the linguistic outcomes of adult L2 learners and heritage speakers, revisiting the central features of adult second language acquisition presented in Chapter 2. On the basis of the existence of age effects in both L1 attrition and L2 acquisition and the context of acquisition in L2 learners and heritage speakers, in Chapter 7 I formulate and evaluate three hypotheses that predict some advantage for heritage speakers over L2 learners. These hypotheses concern early acquired grammatical knowledge, language skill and modality, and potential reactivity to classroom instruction. Finally, Chapter 8 returns to the issue of age effects in dual language acquisition and to the Critical Period Hypothesis. It revisits the concept of incomplete acquisition in the context of first, early bilingual, and second language acquisition. I argue that while the existence of a critical period is still relative, and questionable, in adult L2 acquisition, it is more of an absolute for L1 attrition in a bilingual environment. Just as early and middle childhood is the optimal period in which human beings are biologically equipped to learn language effortlessly, early to middle childhood is also the period most susceptible to language loss in the absence of sustained optimal levels of input and language use. I suggest somewhere between ages 8 and 10 as an approximate time when the possibility of achieving native-like ultimate attainment in the L2, and of undergoing significant attrition in the L1, become less likely. The extent to which age effects in L1 attrition are mainly biological, mainly environmental, or a balanced interaction of both, remains to be pursued. To illustrate how age is related to bilingual outcomes in different types of bilinguals, this book draws extensively on classic and recent empirical evidence from a variety of studies in L2 acquisition, early bilingualism, and attrition framed within cognitive, linguistic, neurolinguistic, and sociolinguistic theoretical perspectives. Because I assume that many readers, and especially graduate students, might not be familiar with all fields of inquiry brought together in this book, many of the empirical studies selected to support my arguments are discussed in some detail. Although research hypotheses and studies within the generative linguistic approach are well represented, and my particular position on the issues discussed may be apparent, my goal is to invite reflection on the issue of age in incomplete acquisition in a manner that is of interest to researchers working within a variety of theoretical traditions. Notwithstanding its linguistic focus, reference to socio-affective and environmental factors will also be brought to the



Chapter 1.  Foundations

­ iscussion, ­especially when such factors help to contextualize the situation of bid lingual speakers of ethnic-minority languages. The topics discussed in this book are of interest to graduate students and researchers – of linguistics, L2 acquisition, bilingualism, L1 attrition, education, psycholinguistics and sociolinguistics – interested in the role of age in language acquisition. Although a number of language-internal and language-external factors contribute to different outcomes in a dual language environment, this book focuses primarily on an investigation of the modularity of grammatical knowledge as currently conceived within generative linguistics. That is, the linguistic discussion focuses on different sub-systems or modules: the lexicon, phonology, morphology, syntax, and semantics and the links between these linguistic modules (or interfaces). Despite this emphasis, little technical knowledge is assumed from the reader.

25

chapter 2

Second language acquisition

For several decades, the issue of age effects on ultimate attainment has been a prominent topic of research in the second language acquisition field. Focusing on cases like those of Kevin and Carolina described in Chapter 1, this chapter characterizes adult L2 acquisition and presents a very selective overview of the theoretical and empirical evidence that supports or challenges the validity of a biological, or critical period, explanation for the outcome of late bilingualism. The purpose of the discussion in this chapter is to set the context for an in-depth examination of first language loss developed in the next five chapters, where I will show that many of the alleged hallmarks of adult L2 acquisition and behavior, including transfer from the stronger language, fossilization, and non target-like ultimate attainment – all typically ascribed to a late onset of acquisition – are also true of some bilingual speakers with an early onset of acquisition. The foundational issues presented in this chapter will be revisited in Chapters 7 and 8, where I trace parallels between adult L2 learners’ and early bilinguals’ (heritage speakers) linguistic competence in their L1 and discuss the reasons underlying incomplete outcomes in these two populations.

1.

Characteristics of adult second language acquisition

A leading problem in the study of second language acquisition is determining how L1 acquisition by children and L2 acquisition by adults are similar and different from each other, and the reasons why the outcome of L1 and L2 acquisition are behaviorally different. Naturally, the first two obvious criteria that differentiate children from adults are age and cognitive maturity. In addition to these two factors, prior linguistic knowledge, input, and affective factors also play a role in the process and outcome of L2 acquisition. These are illustrated in Table 2.1, and further discussed below. Children have no prior linguistic knowledge, while adults start the L2 acquisition process after their L1 is fully in place. In addition, critical differences between L1 and L2 acquisition involve the linguistic environment (input) and, related to that, the mode of acquisition. There are at least five dimensions of input that need to be considered: (a) timing, when exposure to input began; (b) setting, the

28

Incomplete Acquisition in Bilingualism

Table 2.1  Some differences between L1 and adult L2 acquisition. Factors

L1 Acquisition

1. previous linguistic knowledge 2. input a. timing

none L1 knowledge (fully developed) early exposure (birth) late exposure (after puberty)

b. setting

naturalistic

c. mode d. amount

aural abundant and frequent linguistically varied and rich, contextually appropriate

e. quality

3. personality and affective factors 4. target system a. types of errors b. fossilization c. outcome

Adult L2 Acquisition

naturalistic and instructed

aural and written varying (in amount and frequency) contextually restricted, less variety of structures/vocabulary, input from other nonnative speakers irrelevant relevant developmental errors developmental and transfer errors does not occur typical successful and complete

variable and typically incomplete

c­ ontext in which input is received; (c) mode, how input is received; (d) amount, how much input or frequency of input the learner is exposed to; and (e) quality, the contextual and structural variety and accuracy of input (e.g., native vs. nonnative interlocutors). Monolingual children are exposed to the language of the environment early, soon after birth (or even in utero). They mostly receive aural input through interaction and conversation with caregivers and siblings, and they hear such input all day, several hours a day, every day of the week. Therefore, it is said that L1-acquiring children are typically exposed to rich, frequent, and abundant aural input early in life. Adult L2 learners like Kevin and Carolina presented in Chapter 1, on the other hand, begin exposure to the second language later in life, after their L1 has had a chance to develop completely. They receive both aural and written input from different sources through interactions with other non-native speakers and native speakers. L1 acquisition takes place exclusively in a naturalistic context, and children do not take language lessons to learn to speak. Although L2 acquisition can take place in a naturalistic context (an L2 environment) as well, the vast majority of L2 acquisition takes place in a classroom setting. Despite efforts to emulate the natural linguistic environment that surrounds L1 acquirers (i.e., some versions of communicative language teaching or the currently popular Study Abroad Programs



Chapter 2.  Second language acquisition

for university students), a great deal of explicit instruction, grammar practice, and overt correction – negative evidence – contributes to the learning process in adults. For some researchers, negative evidence facilitates acquisition, especially in cases when the L1 of the learners and the target L2 differ in systematic ways, and exposure to positive evidence (naturally occurring speech or written text) is not sufficient for the L2 learners to take notice of and incorporate specific features of the L2 (Doughty 2003; Lightbown 1998; Russell & Spada 2006). None of this explicit instruction, correction, or grammar practice happens with L1-acquiring children before they go to school. In addition to input, environment, and mode of acquisition, affective and social factors play a role in L2 development in general. By contrast, motivation plays no role in L1 acquisition by children. It would be naïve to deny that babies and children have a strong desire to communicate with people in their environment, but it is doubtful that such need is sufficient motivation for rapid acquisition. As a matter of fact, infants are very successful at communicating their basic needs when they have no words. Children have no conscious control over their acquisition process, while many adult L2 learners make a decision to learn a second language and for different reasons: to find job opportunities in another country and immigrate, to go to college, to get a better job, to communicate with a foreign friend or loved one, to learn about another culture, to become part of a group, and so on, especially in the case of immigrants. L2 learners make a decision to study in a classroom setting or abroad, and about the length of their studies. Just as they choose to go to class or not, adult L2 learners also choose how hard to study for a test. In many ways, their conscious decisions about the when, how and how long of language learning may determine their degree of acquisition and success with the language, although these tendencies cannot be generalized to all learners because one can also find many learners, especially in instructional settings, who do not achieve high proficiency despite their best efforts. Sometimes personality plays a role as well: some L2 learners are shy and perfectionist, while others are outgoing and risk-takers. However, no research to date has clearly shown exactly how affective and social variables affect the acquisition of specific grammatical areas. Despite differences in experience and affective reasons for learning a language, L1 and L2 learners must construct a grammatical system based on the input they receive. One of the main goals of second language acquisition research . Even if this general belief were correct, when L2 learners have native-like attainment in syntax but not in phonology, for example, it is hard to understand the specific ways in which motivation contributes to native-like outcomes in one linguistic module but not in others. To the extent of my knowledge, no study of motivation has attempted to tackle this type of question.

29

30

Incomplete Acquisition in Bilingualism

is to ­ explain differences and similarities of the linguistic process and outcome between children and adults. Given this goal, some theoretical proposals focus on similarities, while others focus on differences. In the rest of the chapter, I focus mainly on differences, but not because this is the particular position within second language acquisition that I currently defend. I focus on L1-L2 differences and the “deficit” model of L2 acquisition because my goal in this book is to show that these apparent “deficits” – or differences – are also found in early learners. Nevertheless, we must then ask why these differences arise in a dual language context, and what they tell us about the nature of language acquisition at different ages.

2.

Process and outcome

2.1

Prior language knowledge

The fact that L2-acquiring adults must also build a complex and intricate implicit grammatical system based on input is known in the linguistic literature as the “logical problem of second language acquisition” (White 1989, 2003a). Critical theoretical questions within this approach are first, how children and adults start the L1 and L2 acquisition process, and what cognitive and linguistic resources they bring to the task; second, what the acquisition process is like; and third, where children and adults end in the process. It has been the assumption of generative linguists that children start their language learning process equipped with an innate Language Acquisition Device (LAD). Universal Grammar or UG (Chomsky 1986, 1995) is part of this innate component and consists of a set of grammatical elements, options, and linguistic computations. Input guides children in selecting the options of the language they are learning, and parameters are set accordingly. Since L2 learners already possess a mature linguistic system – their L1 – debate exists as to whether UG is still involved in non-native language acquisition past a critical period. The most accepted position within the generative approach is that the mature L1 constitutes the initial state in L2 acquisition, although access to other options and parameter settings allowed by UG remain available throughout the process. This is explicitly stated in Schwartz and Sprouse’s (1996) Full Transfer/Full Access Hypothesis, according to which the L1 is the blueprint for L2 acquisition, especially at early stages of development. Evidence for this view . For more information about other positions on the initial state of L2 acquisition within the framework of Universal Grammar, including partial transfer/no transfer as well as partial access/no access to Universal Grammar, see White (2003a).



Chapter 2.  Second language acquisition

comes from the fact that many errors L2 learners make can be traced back to contrastive differences between the learner’s native language and the target language, suggesting L1-based knowledge and hypotheses. However, L2 learners also make developmental errors, or errors that all learners make, including L1-acquiring children, regardless of language background or age. These errors, together with the fact that many learners overcome initial L1 influence, are taken to suggest UG involvement. To illustrate both transfer and developmental errors in L2 acquisition, I refer to Montrul’s (1999, 2000, 2001a, b) comparative studies of Spanish, English and Turkish as second languages. These studies focused on the acquisition of the causative-inchoative alternation, or verbs that alternate between transitive and intransitive (I broke the vase vs. The vase broke). These verbs exist in the three languages tested, but the three languages vary with respect to how they mark morphologically the transitive or intransitive version of the verb. For example, English has no marking (or zero morphology), Spanish marks the intransitive form with the clitic se (Rompí el jarrón vs. El jarrón se rompió ), while Turkish marks some transitive verb forms with a causative morpheme (Gemi bat-mış “The ship sank” vs. Düşman gemi-yi bat-ır-mış “The enemy sank the ship”), and other intransitive verb forms with anticausative morphology like Spanish (Pencere kır-ıl-dı “The window broke” vs. Hırsız pencere-yi kır-dı “The thief broke the window.”) English-, Spanish-, and Turkish-acquiring children go through a stage in their language development when they use intransitive verbs in a transitive sentence frame (*Don’t giggle me, *You fall the monkey). Montrul’s studies showed that intermediate-level adult L2 learners also accepted errors of this sort in the three languages tested, even when literal translations of these same sentences are not correct in the learners’ L1s. At the same time, the L2 learners made errors due to influence from their L1. For example, English-speaking L2 learners incorrectly accepted intransitive break-type verbs without reflexive morphology in Spanish and in Turkish, and incorrectly accepted transitive break-type verbs without causative morphology in Turkish. During interlanguage development, prior language knowledge can assist, and even accelerate, the L2 acquisition process (positive transfer), or act as an obstacle for progress (negative transfer) and lead to eventual fossilization. Examples of positive and negative transfer effects abound in the literature, especially in experimental studies that include L2 learner groups from different L1 backgrounds. The studies on L2 Spanish, Turkish and English discussed above are one example. Another recent example is Sabourin, Stowe and de Haan’s study (2006) of gender agreement in Dutch noun phrases. The study investigated English, German and Romance-speaking L2 learners of Dutch. Both German and Romance languages have grammatical gender, although it works differently in the two language families. The German gender system is more similar to that of Dutch (both are

31

32

Incomplete Acquisition in Bilingualism

­ ermanic languages), while English does not have grammatical gender. Results G of a gender assignment task and a gender agreement task showed that the three learner groups were significantly less accurate than the Dutch control group. However, the findings also illustrate how the structural proximity of the languages can facilitate or delay the process of L2 acquisition. The German speakers performed the closest to the native Dutch speakers; the Romance speakers were better than the English speakers but less accurate than the German speakers; the English speakers performed at chance. If, as many studies have shown, L1 knowledge guides the initial process of interlanguage development, a key question is whether such influence is eventually overcome with advanced development. This would be the Full Access part of the Full Transfer/Full Access Hypothesis. Experimental studies including L2 learners of different proficiency levels provide the best testing case to address this question. A good example of initial L1 transfer and progressive abandonment of this blueprint in response to input and grammar restructuring is Slabakova’s (2001) investigation of knowledge of telicity marking (a semantic notion related to aspect and endpoints) in English. A telic sentence like Antonia baked a cake is interpreted as completed (telic) (i.e., Antonia finished the cake), whereas the same sentence with a bare plural (Antonia baked cakes) is interpreted as habitual and incomplete (atelic). There is no indication that a cake was finished. When one compares the morphological make-up of Romance and Slavic languages to that of English, one notices that English does not mark aspectual distinctions like telic/atelic overtly, with morphology on the verb. By contrast, Bulgarian uses prefixes. Slabakova conducted two studies to assess how Bulgarian and Spanish speakers would interpret telic and atelic sentences in English: one study included Bulgarian-speaking learners of English ranging from low to high-advanced proficiency, and a follow-up study focused on low proficiency Spanish-speaking learners of English. Spanish marks telicity with count and mass nouns in object position as in English. In one of the tasks, the L2 learners were asked to judge the felicity of two clauses, which only differed on whether the object in the second clause was count or mass (one way to indicate telicity in English). An example is provided in (1). As can be seen, (1a) is a logical combination while (1b) is odd. Learners were asked to judge how well these sentences combined on a scale ranging from –3 (illogical) to +3 (logical). (1) a.   Antonia worked in a bakery and made cakes. (habitual 1st clause + atelic   2nd clause) b. #Antonia worked in a bakery and made a cake. (habitual 1st clause + telic   2nd clause)



Chapter 2.  Second language acquisition

Table 2.2  Mean judgments on telic and atelic English sentences (adapted from Slabakova 2001). Groups

Proficiency

L1

N

# telic (–3)

atelic (+3)

L2 learners

low

Spanish Bulgarian Bulgarian Bulgarian American British

21 34 50 45 16 16

0.55 1.44 0.54 0.41 0.19 0.81

2.04 1.95 1.75 2 2.09 2.41

intermediate advanced Native speakers

If the L1 constrains L2 initial representations, Slabakova predicted that Bulgarian speakers would have significant difficulty distinguishing telic from atelic sentences in English, due to the fact that the verb is morphologically unmarked, providing no clues to aspect. Eventually, to overcome L1 influence, the Bulgarian-speaking learners would have to realize that English marks telicity in the object (count vs. mass objects). Since Spanish also marks telicity on the object (and grammatical aspect with preterite-imperfect suffixes on the verb), the Spanish-speaking L2 learners were expected to perform like the English native speaker controls. Overall results are presented in Table 2.2. The results showed that the low proficiency Bulgarian-speaking group performed less accurately than all other groups, including the proficiency-matched Spanish-speaking group, as predicted. This difference was statistically significant with telic sentences (which require an overt verbal prefix in Bulgarian but not in Spanish or English). Secondly, intermediate and advanced Bulgarian speakers did not differ statistically from the British or American native speakers. Slabakova’s cross-sectional study provides an excellent example of how low-intermediate learners transfer their L1 parametric values to the L2 (Full Transfer), yet, with development, intermediate and advanced Bulgarian speakers overcome their L1 influence and reset the parameter to the target value (Full Access). In conclusion, the studies presented in this section have illustrated how developing interlanguage grammars are characterized by both developmental errors, as in L1 acquisition, and by transfer or cross linguistic errors that arise in a language contact situation at the individual level. The next section discusses how transfer errors in L2 acquisition in particular may lead to eventual fossilization and incomplete ultimate attainment in some cases.

33

34

Incomplete Acquisition in Bilingualism

2.2

Persistent L1 influence as fossilization

Although L2 learners can acquire very high levels of proficiency in the L2 given abundant exposure and opportunity to use the L2, it is not uncommon for many L2 speakers to exhibit clear non-native like features in their linguistic behavior, most notably in pronunciation and morphosyntax. It is common for adult second language learners to display synchronic variability and premature stability in some linguistic areas. In the second language acquisition literature, this phenomenon is referred to as fossilization, a term first introduced by Selinker (1972). In recent years, there has been increasing interest in investigating the nature of end-state L2 grammars, the ways in which they diverge from the target, and the reasons behind such divergence. This interest has prompted a re-examination of the concept of fossilization through a more rigorous empirical scrutiny of this phenomenon in endstate grammars longitudinally (Lardiere 2007; Long 2003; White 2003b). While many related and unrelated causes lead to fossilization (Han 2004; Selinker & Lakshmanan 1993), there is a strong sense that L1 influence is not just present in early stages of interlanguage development, but that it also exerts persistent, long-term effects in L2 ultimate attainment. In the next chapters, we will also see how dominant-language (L2) transfer also predicts patterns of L1 attrition in very advanced L2 speakers. Fossilization is not a global, or across the board, phenomenon, however: we cannot say that an L2 grammar is fossilized in its entirety. Fossilization tends to be local, in the sense that it affects some areas or modules of linguistic knowledge but not others. This is clearly illustrated by Lardiere’s (1998a, b, 2007) exemplary longitudinal case study of Patty, whose ultimate attainment in English diverges from native-like competence in some respects and completely converges in others. Lardiere followed the linguistic performance of Patty, an English L2 speaker whose first language is Chinese (actually, Mandarin and Hokkien), over a period of almost 20 years. Patty was recorded after she had lived in the USA for ten years and then again almost 9 years later. Lardiere shows that there were no changes in Patty’s grammar between the three recording sessions conducted 9 years apart. Lardiere (1998a, b, 2007) examined Patty’s use of past tense and agreement morphology, as well as other syntactic evidence of grammatical morphology in

. See Long (2003) for the validity of fossilization as a theoretical construct and its empirical testability. Long advocates using the term “stabilization” instead of fossilization, a state/stage that is, in his view, empirically more testable than fossilization. A stabilized grammar is not necessarily a fossilized grammar, if fossilization means permanent deviance. While Long makes an extremely valid point, I will continue to use the term “fossilization” in this book to describe an incompletely acquired state of linguistic knowledge.



Chapter 2.  Second language acquisition

Table 2.3  Percentage accuracy in Patty’s speech (adapted from Lardiere 2007). Recording Overt subjects

Null subjects

3rd sg. agreement on lexical main verbs

Past on finite verbs

1 2 3 Total

0.55 2.11 2.08 1.73

4.76 0 4.54 4.41

34.78 34.85 33.82 34.66

99.45 97.89 97.92 98.27

wh-questions and the nominal domain. Results show a clear dissociation between morphological and syntactic aspects of verbal inflection: Patty only supplied 4.41% of 3rd person agreement morphology and 34.66% of past tense morphology in obligatory contexts, while production of overt subjects and nominative case was above 98% accurate, as shown in Table 2.3. The first possibility that Lardiere (1998a, b) considered to explain this phenomenon in Patty’s interlanguage was that the observed dissociation between grammatical and morphological knowledge was not necessarily due to a deficit at the level of linguistic representations. Patty obviously has knowledge of the abstract feature finiteness despite morphological errors in production. Yet, Patty’s difficulty with past tense marking is amenable to other possible explanations. One explanation capitalizes on the role of the first language, since Patty’s native languages (Mandarin and Hokkien) are morphologically impoverished in this respect. On many current analyses (Lin 2003), Chinese does not have tense, so Patty’s difficulty with past tense in English could be attributed to Patty’s inability to acquire some aspects of the formal feature [ ± past] of this functional category in English, as argued by Hawkins and Liszka (2003). However, Lardiere (2007) questions at length whether this monolithic approach to features can still be upheld, given that a thorough analysis of Patty’s past-tense marking with different verbs (regular and irregular, lexical vs. auxiliaries, telic vs. atelic, in written and in oral production, etc.) is non-nativelike but still quite complex. Alternatively, Lardiere pursues a phonological explanation. According to this explanation, Patty’s performance could also be related to the fact that many past tense forms in English end in consonant clusters (-st, -zd) and Chinese (including Mandarin and Hokkien) does not allow consonant clusters. Yet another possibility shifts the burden of persistent L1 influence to the phonology-morphology-syntax interface. In agreement with Goad and White (2004), Patty’s pattern of performance with past-tense marking in English may arise from differences in the ways in which Chinese and English organize the prosodic structure that supports the production of inflectional morphology.

35

36

Incomplete Acquisition in Bilingualism

While Patty’s patterns of fossilization may be due to a number of factors, more clear evidence for persistent L1 influence contributing and determining fossilization in this area is shown by White (2003b). White studied an adult Turkish speaker who had been living in Canada for over 10 years when she was first tested at the age of 50. The participant was interviewed two times over a period of one and a half years, with no observable change in her performance. Since Turkish is a language with rich nominal and verbal inflection, White examined potential morphological variability in the verbal and nominal domain to make comparisons of this speaker and Lardiere’s Chinese-speaking subject’s idiolect. Turkish marks tense and agreement with inflection on the verb; it has no definite article. The numeral bir “one” can be used in indefinite contexts, and in specific and non-specific contexts (Yesterday I wrote a letter). Linguistic analysis of the Turkish speaker’s productions in English showed 79.75% and 80.5% accuracy on verbal morphology at Times 1 and 2, as shown in Table 2.4. However, accuracy on nominal morphology was significantly lower than on verbal morphology, especially with indefinite articles (average 60.5%). See Table 2.5. White admits that the effects of the L1 are limited in this study because the Turkish speaker appears to have knowledge of how definiteness is encoded in English. However, comparison of this Turkish speaker with the Chinese-speaking subject of Lardiere’s studies clearly shows that the Turkish speaker, whose L1 has richer verbal inflection, performs much better on verbal agreement and past tense production than the Chinese speaker, whose L1 has poor verbal inflection.

Table 2.4  Verbal morphology. Percentage accuracy in SD’s speech (adapted from White 2003b). Recording Overt Pronouns Null Pronouns 3rd p. sg. agreement in lexical main verbs

Past in Finite lexical verbs

1 2 Total

85 76 80.5

98.5 99.4 98.95

1.5 0.6 1.01

78 81.5 79.75

Table 2.5  Nominal morphology. Percentage accuracy in SD’s speech (adapted from White 2003b). Recording

Plural morphology

Definite articles

Indefinite articles

1 2 Total

87 90 88.5

73.5 71.75 72.62

59.75 61.25 60.5



Chapter 2.  Second language acquisition

Furthermore, the Turkish speaker supplied verbal morphology more reliably than nominal morphology, suggesting a clear L1 effect because Turkish lacks determiners. White (2003b) assumes the prosodic transfer account (Goad, White & Steele 2003; Goad & White 2004), concluding that the locus of persistent L1 influence at the morphological level is related to how morphology is represented prosodically in different languages. In conclusion, the studies by Lardiere (1998a, b, 2007) and White (2003b) show that L1 influence and fossilization affect specific aspects of morphosyntactic production, or how morphology is spelled out phonetically in production. Although several hypotheses have been advanced to explain this phenomenon, much research needs to be done to explore the specific locus of the disruption for this grammatical domain. But production of inflectional morphology is not the only grammatical area vulnerable to fossilization, in addition to phonetics and phonology, as we will see shortly. Other studies have shown that aspects of lexical-semantics and the syntax-discourse interface are also affected in speakers deemed to be at endstate. The earliest example of persistent L1 influence at the lexical-semantics interface is Sorace’s (1993) study of near-native speakers of Italian. In Sorace’s (1993) own words, “the term near-native seems to contain an implicit statement on the question of ultimate attainment, since it implies the no coincidence of native and non-native grammars: the near native grammar is almost the same as the native grammar, but falls short of it [emphasis mine] (p. 23).” Clearly subscribing to the deficit model of L2 acquisition, Sorace views near-native grammars as, in some sense, incomplete and divergent from native grammars. In common with how Schachter (1990) characterizes non-native like knowledge in the Incompleteness Hypothesis presented in Chapter 1, Sorace portrays an incomplete grammar as a grammar that lacks a given property P, which the native grammar has. Lack of property P in the interlanguage leads to random, variable, probabilistic, or indeterminate judgments in the L2. Another possible outcome of an incomplete L2 grammar determined by the properties of the L1 is, according to Sorace, a divergent grammar. Unlike an incomplete grammar that lacks property P altogether, a divergent grammar is a grammar that instantiates property P, but it does so in a different way than the target language. As a result, L2 speakers evidence determinate, but not probabilistic, judgments that are still different from the judgments of native speakers. To illustrate these two possible outcomes in incomplete near-native competence, Sorace investigated knowledge of Italian unaccusative verbs in near-native speakers whose L1 was English and speakers whose L1 was French, all living in Italy. A group of native Italian speakers formed the control group. All participants had started learning Italian after age 15, and their time of exposure to Italian ranged from 9–15 years.

37

38

Incomplete Acquisition in Bilingualism

Unaccusativity refers to the classification of intransitive verbs into two main subclasses: unaccusatives, such as andare “go” (verbs with non-agentive subjects), and unergatives, such as telefonare “telephone” (verbs with agentive subjects). In Italian, this semantic classification has syntactic consequences. Unaccusative verbs typically select the auxiliary essere “be” in the pasato prossimo, whereas unergative verbs select avere “have,” as shown in (2) and (3). unaccusative

(2) Mario è /* ha andato a la casa. Mario is/ has gone to the house ‘Mario went home.’ (3) Mario *è /ha telefonato a la casa. Mario is/ has phoned to the house ‘Mario phoned home.’

unergative

Some verbs, however, can take essere or avere as auxiliary, as shown in (4), but in restructuring constructions (a sequence of a modal and an infinitive) containing a clitic, if the clitic climbs to the main verb, as in (5), the auxiliary has to surface as essere obligatorily. If the clitic stays low after the infinitive, as in (6), no change of auxiliary is required. (4) Mario è /ha dovuto andare a la casa. Mario is/has must go to the house ‘Mario must have gone home’ (5) Mario ci è /*ha dovuto andare. Mario there is/has must go ‘Mario must have gone there.’ (6) Mario è /ha dovuto andarci. Mario is/has must go there ‘Mario must have gone there.’

(optional auxiliary)

(obligatory change in clitic climbing)

(optional auxiliary with no clitic climbing)

English unaccusative and unergative verbs take auxiliary have in the perfect tenses. French, like Italian, has two auxiliaries as well: avoir for unergatives and être for unaccusatives. But unlike Italian, the number of unaccusative verbs that take être in French is more restricted than the number of unaccusative verbs that take essere in Italian. Sorace tested knowledge of auxiliary verbs with five different types of unaccusative and unergative verbs, as well as auxiliary change in restructuring contexts, using a type of judgment task called the magnitude estimation technique. Sorace’s . For specific details and theoretical rationale for this task see Bard, Robertson & Sorace (1996).



Chapter 2.  Second language acquisition

Table 2.6  Mean magnitude estimation ratings on sentences with essere and avere (adapted from Sorace 1993). Groups NS (Italian) NN (L1 French) NN (L1 English)

N 36 20 24

Simple sentences

No climbing

Clitic climbing

essere

avere

essere

avere

essere

avere

9.2 3.8 7.2

9.7 9.4 6.9

8.1 4.1 6.7

8.7 7.8 6.2

8.5 8.5 6.2

3.1 4.2 6.6

results showed that the near-native speakers’ ratings were statistically different than those of the Italian native speakers, though the French- and English-speaking near-native speakers were sensitive to the notion of unaccusativity in Italian. Table 2.6 shows that the ratings for essere and avere in the three conditions tested differ for these two groups, especially in their judgments of sentences with clitic climbing and no climbing (examples (5)–(6)). The French-speaking near-natives did not accept essere in Italian, but showed a consistent pattern of acceptance of sentences with avere, including the fact that avere is not allowed in clitic climbing contexts, as in (5). According to Sorace, the French near native grammars were divergent from native Italian grammars but L1-consistent (with French). By contrast, the English-speaking near-natives showed inconsistent and indeterminate judgments overall, since they rated all sentences with essere and avere alike (with ratings between 6–7), showing no distinction. The difference between these two nonnative outcomes relates to the different L1s of the near natives, albeit indirectly, according to Sorace. French has a system of two auxiliaries to mark unaccusativity while English has only one. Therefore, the French-speaking near natives had a mental representation for the two auxiliaries, while the English-speaking nearnative speakers did not. Finally, let us turn to another recent study demonstrating L1-based incomplete representations in ultimate L2 attainment, this time on Spanish near-native speakers. To show that fossilization can also occur at the level of interpretation of formal semantic/syntactic features, Valenzuela (2006) examined the status of topic constructions in the near native L2 Spanish of L1 English speakers. Topic constructions in Spanish are expressed with clitic left dislocations. When the topic is specific, the sentence must have an obligatory clitic, as in (7). Non-specific topics, as in (8) and (9), have no clitics. (7) El libro, lo leí. /*El libro, leí. the book, CL I-read ‘The book, I read’.

specific

39

40 Incomplete Acquisition in Bilingualism

(8) Un libro, leí. /*Un libro, lo leí. a book, I-read ‘A book, I read’.

non-specific

(9) Libros, leí./*Libros, los leí. books, I-read ‘Books, I read’.

non-specific

English topics do not show a difference in structures depending on the specificity of the preposed element, since English does not have clitics. English topics appear in a construction containing a null anaphoric operator, as in contrastive left dislocations (e.g., Books, I read). Contrastive left dislocations also occur in Spanish, as shown in (11a). Both in Spanish contrastive left dislocations and clitic left dislocation constructions, the fronted elements are topics. However, the main difference between the two constructions is that the resumptive element in contrastive left dislocations is a null anaphoric operator, whereas in clitic left dislocation it is a clitic. Furthermore, the two constructions differ in the environment in which the leftdislocated phrase can occur. In clitic left dislocations (CLLD), the left-dislocated phrase can occur in either main or embedded clauses, as in (10a)–(10b), while in contrastive left dislocations (CLD), it can only occur in main clauses, as in (11a–b). (10) a.   Esas flores, las he comprado en el mercado.   these flowers CL I-have bought at the market   ‘These flowers, I bought at the market’. b.   Me pregunto que, a María, quién le dio esas flores.   me I-ask that to María who Cl gave those flowers   ‘I wonder who gave those flowers to María’. (11) a.   FLORES, compro todas las semanas.   flowers I-buy all the weeks   ‘Flowers, I buy every week’. b. *Me pregunto que, flores, quién comprará   me I-ask that flowers who will buy   ‘I wonder who buys flowers’.

CLLD

CLLD

CLD

CLD

Valenzuela tested knowledge of topic constructions in Spanish and their interpretive characteristics with respect to specificity. Participants were English speakers living in Spain who were near-natives of Spanish. For all participants, age of exposure to Spanish occurred after puberty. Monolingual Spanish speakers were tested as a control group. Valenzuela used an oral grammaticality judgment task, a sentence selection task and a sentence completion task.



Chapter 2.  Second language acquisition

Table 2.7  Oral selection task. Percentage acceptance of clitic left dislocations in specific and non-specific contexts (adapted from Valenzuela 2006). Groups

N

Native Speakers 25 L2 Speakers 15

Main Clauses

Embedded Clauses

specific (clitic)

non-specific (no clitic)

specific (clitic)

non-specific (no clitic)

  94 100

53 40

  93 100

67 22

Valenzuela found the following: the near-native group had knowledge of Spanish clitic left dislocations structures, produced and accepted clitics, and were sensitive to the syntactic distribution of contrastive left dislocations and clitic left dislocations according to clause (main vs. embedded). However, results from the three tasks indicated that the near-natives’ grammars differed from the native speakers’ grammars only with respect to specificity in embedded clauses. The near-natives appeared not to know that topicalized sentences with no clitics are non-specific in Spanish, since they only scored with 22% accuracy, as shown in Table 2.7. In this respect, the interlanguage grammar is incomplete with respect to the target and in a manner predictable from the L1 grammar. To summarize, the four studies presented in this section show that the influence of the L1 can play a deterministic, yet localized, role at very advanced stages of interlanguage development and in the endstate. It has been demonstrated at the level of inflectional morphology and syntax (morphology-syntax and morphology-phonology interface, the lexical-semantics interface, and the syntax-discourse interface). Although fossilization certainly has multiple causes (Selinker & ­Lakshmanan 1993), persistent L1 influence in near-native speakers is a form of fossilization, if the interlanguage grammars cannot restructure further in response to input. The perennial issue in second language research, however, has been, and continues to be, whether age of acquisition lies at the heart of this puzzle.

3.

Is there a critical period in L2 acquisition?

The common observation in the second language acquisition field is that, in general, native-like ultimate attainment is possible, although perhaps not probable. Native-like attainment is typically inversely related to age of acquisition. The older the learner begins exposure to the second language, the less likely he or she will be to attain a native-like level of linguistic achievement. There is ongoing debate as to whether the hypothesized critical period applies only to L1 acquisition as we

41

42

Incomplete Acquisition in Bilingualism

saw in Chapter 1, or whether it is also a factor in the acquisition of subsequent languages. Note that the claim of a critical period is not that L2 grammars are not acquirable after a certain age: there is ample evidence that many people who started acquisition in adulthood speak and use L2 languages with highly sophisticated levels of proficiency. The question is whether the linguistic means and product of L1 and L2 acquisition are the same after a certain age. If a critical period is relevant only for L1 acquisition, the sequential acquisition of additional languages later in life should be cognitively possible and easier because the blueprint for language acquisition was activated early in childhood. In other words, the early acquired linguistic mechanisms would still be in place. Since we saw that L1 knowledge may exert persistent influence on the L2, it may be the case that the processes and representations from the L1 act as a filter through which the L2 is acquired. If, on the other hand, L2 acquisition is maturationally constrained, then no language learner should ever be able to acquire an L2 completely after a certain age, at least through the same linguistic means assumed to operate in L1 acquisition. Back in the 1950s, Penfield already had his own ideas about the linguistic abilities of children and adults as they relate to the acquisition of more than one language: . . . . a child who is exposed to two or three languages . . . pronounces each with the accent of his teacher. If he hears one language at home, another at school, and a third, perhaps, with a governess in the nursery, he is not aware that he is learning three languages at all. He is aware of the fact that to get what he wants with his governess he must speak one way, and with his teacher he must speak in another way. He does not reason it out at all. There is no French, no German, no English. It is simpler than that. One secret of the success of this method is that, of course, it is employed while the child is forming the speech units in his eager little brain. A child who hears three languages instead of one, early enough, learns the units of all three without added effort and without confusion.  (Penfield & Roberts 1959, pp. 253–4)

The average lay person is familiar with the fact that many L2 learners have detectable foreign accents. Personal accounts of how adult L2 learners fall short of sounding like native speakers abound, including Penfield’s own anecdotal experience: My wife tried to reassure me by pointing out that our own children had gained a satisfactory command of two added languages. We have done no more than to have them hear German and French well-spoken in their early childhood . . . my own children learned to use German and French without apparent effort, without foreign accent, and without the long hours of toil that I had sacrificed to language study. They did well what I did badly . . . For my own part, I had heard no foreign



Chapter 2.  Second language acquisition

tongue before the age of 16. After that, I studied three modern languages for professional purposes but spoke none well. (Penfield 1965: 788)

If we accept that a foreign accent is very difficult to overcome the older the learner begins acquisition, is a critical period necessarily the cause of these experiences? Nobody denies the existence of age effects. What is debatable is the explanation behind these effects. Some maintain the age effect is biological. As human beings develop, there is loss of neurofunctioning plasticity as other cognitive abilities emerge and take over abruptly. I will refer to this position as the critical period position. The alternative account – the no critical period position – denies any maturationally-determined change, and attributes age effects to other factors, such as input and language use, affective factors, and individual differences in cognitive abilities. In the next two sections, I present the assumptions and a representative sample of empirical evidence for these two main positions in more detail, followed by an assessment of the findings in §4. The literature on maturational constraints in second language acquisition is extensive, and I cannot therefore cover it in great depth. Because the book focuses on incomplete outcomes in different grammatical modules, the studies selected illustrate ultimate attainment in phonology, morphology, syntax, and semantics, even when these studies have been discussed several times elsewhere in other publications.

3.1

The critical period position

Formulated in different ways by different theoretical approaches to second language acquisition, the critical period position posits that differential outcomes in L1 and L2 ultimate attainment are due to a maturationally determined critical period (Long 2007). Under the generative framework, one way to explain differences between children and adults is by appealing to the concept of Universal Grammar, as explicitly formulated by Bley-Vroman (1989, 1990) in the Fundamental Difference Hypothesis and by others under the no parameter resetting model (Clahsen & Muysken 1986; Meisel 1997; Schachter 1990; Hawkins & Chan 1997; Smith & Tsimpli 1995). Assuming that child L1 acquisition is constrained by Universal Grammar, Bley-Vroman claims that children use this language-specific innate system of knowledge to acquire language. Such innate knowledge explains how children solve the “logical problem,” or the gap between children’s limited or incomplete linguistic experience and the complexity and sophistication of the attained . See Birdsong 1999, 2006; Bongaerts 2005; DeKeyser & Larson-Hall 2005; Harley & Wang 1997; Herschensohn 2000, 2007; Hyltenstam & Abrahamsson 2003; Long 1990, 2007; Singleton & Lengyel 1995; Singleton & Ryan 2005, among others, for more comprehensive overviews.

43

44 Incomplete Acquisition in Bilingualism

linguistic knowledge (Chomsky 1981; Crain & Thornton 1998; Guasti 2002). In a nutshell, Universal Grammar explains the efficiency, rapidity, inevitability, and completeness of the L1 acquisition process during the critical period. Although Bley-Vroman assumes a logical problem exists in L2 acquisition as well, his specific claim is that adults approach the logical problem in a very different way, with a different set of cognitive procedures. Different outcomes entail different means. If we assume that an innate linguistic system guides language acquisition in children, this very same system is no longer available in L2 acquisition. And if it is available, its operation is highly imperfect at best because L2 development is often prone to fossilization and incomplete ultimate attainment. Another claim made by Bley-Vroman is that the only way L2 learners have access to Universal Grammar is through their already attained L1 knowledge. The implication here is that once the child sets parameters on the basis of input, Universal Grammar progressively becomes a particular language until the steady state is reached in L1 acquisition. This notion of the L1 being the only source for Universal Grammar is also explicit in another more recent version of the critical period position: Hawkins and Chan’s (1997) Failed Functional Features Hypothesis (FFFH), Hawkins and Hattori’s (2006) Representational Deficit Hypothesis and Tsimpli and Dimitrakopoulou’s (2007) Interpretability Hypothesis. The FFFH and its recent incarnations explicitly state that the L1 may act as a barrier for further development and restructuring at advanced stages of development (see also Smith & Tsimpli 1995; Tsimpli & Smith 1991; and Tsimpli & Roussou 1991). According to this hypothesis, features of functional categories are subject to a critical period. Those features not encoded in the learner’s L1 become inaccessible or impaired after a certain age: “Where parameter settings differ between an L1 and a target L2 there will be considerable restrictions on the extent to which an L2 learner can build a mental grammar like that of a native speaker” (Hawkins & Chan 1997, p. 189). Although the explanations given differ, the data on L1-based fossilization discussed in §2.2 are compatible with this position.

. Smith & Tsimpli (1995) maintain that functional categories (what they call the functional module) and their associated parameterizable features are in the lexicon. They write: “If we assume that the critical period hypothesis is correct, maturational constraints on the functional module can be interpreted as entailing its complete inaccessibility after the end of this period. The importance of this suggestion in the current context is that it has clear implications for adult second language learning: UG may still be available but parameter resetting cannot be” (Smith & Tsimpli 1995, p. 24). . White (2003a) discusses other “impairment” hypotheses within Universal Grammar, of which FFFH is one. The impairment view assumes a strong CPH position.



Chapter 2.  Second language acquisition

Approaching the issue of age from cognitive and neurolinguistic perspectives, a similar and compatible position is maintained by DeKeyser (2000, 2003), ­Ullman (2001), M. Paradis (2004) and DeKeyser and Larson-Hall (2005). ­DeKeyser and M. Paradis, who do not believe that language ability is innate, distinguish instead between implicit and explicit memory systems and learning of language. The gist of this proposal is that young children learn language implicitly, without awareness. Implicit knowledge is stored in procedural memory (see also Ullman 2001). The specific linguistic content handled by this type of memory is grammar proper (i.e., productive rules and the computational system in generative terms according to Ullman 2001 and Pinker & Ullman 2002a, b). However, the implicit learning mechanism becomes severely limited as children grow, due to both progressive loss of neurological plasticity and the emergence of other cognitive abilities. As a result of this neurocognitive change, adults learn languages explicitly, with conscious effort and awareness. Explicit knowledge, which includes knowledge of words and of irregular morphological forms in addition to other non-linguistic knowledge, is stored in declarative memory (Ullman 2001). As children age, they rely more and more on explicit learning than on implicit learning. For Ullman (2001), reliance on procedural memory for language learning is subject to a critical period. This implies that the functional lexicon and productive rules are affected. Note that the implicit learning mechanism stored in procedural memory maybe akin to aspects of Universal Grammar acquired implicitly and without awareness, whereas according to DeKeyser (2000), the explicit mechanisms stored in declarative memory parallel the domain-general cognitive mechanisms discussed by Bley-Vroman (1990). My understanding of these positions is that both implicit and explicit learning mechanisms are available to children and adults, or at least to adults, but while children acquire their L1 predominantly with implicit mechanisms, adults acquire an L2 predominantly through explicit mechanisms. This is summarized in Table 2.8, where bold typeface indicates “predominantly” used. A typical strict interpretation of the critical period hypothesis for L2 acquisition implies that there is, as Lee and Schachter (1997) put it, a “window of opportunity” for L2 acquisition. Successful and complete L2 acquisition is possible only

. DeKeyser (2000) interprets Bley-Vroman’s Fundamental Difference Hypothesis this way: implicit learning mechanisms are likened to Universal Grammar while the explicit learning mechanisms stand for general problem solving skills. But this is not what Bley-Vroman originally intended (p.c. August 13, 2007). By “general problem solving” Bley-Vroman meant the whole range of learning mechanisms that are available outside the UG/language-module. These include both explicit and implicit mechanisms. General problem solving is not restricted to explicit/conscious mechanisms as DeKeyser interpreted it.

45

46 Incomplete Acquisition in Bilingualism

Table 2.8  Learning mechanisms available to children and adults. L1 acquisition by children L2 acquisition by adults The Fundamental Difference Hypothesis Universal Grammar (Bley-Vroman 1990) Domain-specific linguistic mechanisms Neuro-cognitive approach Implicit learning (DeKeyser 2003; M. Paradis 2004) Explicit learning

L1 knowledge Domain-general cognitive mechanisms Implicit learning Explicit learning

during the critical period and impossible after its closure. The implication here, though not always explicitly recognized by these researchers, is that there should be a dramatic decline, a complete breakdown, in linguistic ability at a specific cutoff age, as we saw in Chapter 1 for L1 acquisition. Unfortunately, the cut-off age ranges in the literature from as young as 2 years of age for phonology to up to 13 years of age for morphosyntax. Studies supporting a critical/sensitive period are of two kinds. On the one hand, there are those that seek to establish a correlation between age of onset of L2 acquisition with degree of proficiency in the language. This type of research typically focuses on immigrants who arrived in the L2 environment and started learning the L2 at different ages and have resided in the host country for several years. Unlike the foreign language environment, an L2 environment makes exposure to an optimal amount of input (a condition conducive to successful acquisition in L1 learners) more likely. Holding input factors such as length and amount of exposure as constant as possible, the aim of this research design is to show that age of acquisition – operationalized as age of arrival (AoA) – is the best predictor of successful L2 acquisition. On the other hand, there are studies that focus on very advanced L2 learners or near-native speakers assumed to be at endstate who began acquisition of the L2 after puberty, without testing L2 learners with earlier ages of arrival. The goal of this approach is to show that no near-native speaker who started acquiring the L2 late can achieve the level of linguistic mastery of a native speaker. Most of these studies have focused on formal aspects of language, namely pronunciation (phonetics and phonology), morphology, syntax and semantics, either globally or targeting specific structures. Pronunciation (and phonology) has long been recognized as an obvious candidate for fossilization, due to its reliance on neuromuscular involvement (­Scovel 1988), and there is empirical evidence supporting this prediction. For example, Oyama (1976) looked at the pronunciation abilities of Italian immigrants who had lived in the United States for several years (between 5–18 years) and whose age of arrival ranged from 6–20. Subjects were asked to complete a read



Chapter 2.  Second language acquisition

aloud task and a narrative and were rated for degree of foreign accent. Results revealed a clear effect for age of arrival: young arrivals were much better at achieving an English-like pronunciation than late arrivals. There was no effect for external input factors, such as length of residence, years of exposure, or motivation to learn the language. A few years later, Patkowski (1990) replicated these results and also showed that subjects with age of arrival before age 15 were significantly better than those with age of arrival older than 15. Summarizing these findings in a critical overview of the topic, Long (1990) stated that “a native-like accent is impossible unless first exposure is quite early, probably before 6 in many individuals and by about age 12 in the remainder” (p. 266). As for morphosyntax, the most influential study advocating a critical period for L2 acquisition is Johnson and Newport (1989). Johnson and Newport tested 46 Korean and Chinese immigrants who had resided in the United States for more than 5 years, but like the Italian-speaking subjects in the Oyama study, had different ages of arrival (3–39). The participants were asked to provide grammaticality judgments on 276 English sentences presented orally. The stimuli sentences targeted regular verb morphology, irregular plurals, particle vs. preposition placement, articles, wh-questions, etc. Results showed a clear main effect for age. On all the structures tested, Johnson and Newport found that early arrivals outperformed late arrivals. There was a clear age function up until puberty, when all accuracy scores were inverse-linearly related to age of arrival. After puberty, scores were quite variable and unrelated to age of arrival, suggesting a cut-off point, or end, of a sensitive period. Many aspects of this study and its conclusions have been widely criticized by Bialystok and Hakuta (1994, 1999) and many others, as we will see in §3.2. However, Johnson (1992) and DeKeyser (2000) have partially replicated the original Johnson and Newport (1989) study by implementing several methodological modifications (including different test modalities, stimuli and even subject populations), finding, by and large, the same results. In another related study, Johnson and Newport (1991) investigated whether there were maturational effects for Subjacency, a principle of Universal Grammar. This study is important because it was one of the first to verify maturational constraints on aspects of overt syntax assumed to be regulated by universal principles. Technical details aside, Subjacency is essentially a principle that regulates the movement and distance of phrases in questions and embedded sentences (today known as locality conditions). Such a principle explains why the English sentence in (12) is grammatical, but sentence (13) is not. (12)   Who did John say that Mary saw? (13) *What did Mary meet the man who saw?

47

48 Incomplete Acquisition in Bilingualism

Chinese, Korean and Japanese do not have overt wh-movement in questions; whphrases remain in situ. It was assumed that in these languages, Subjacency does not apply in the syntax. The experiment tested 23 Chinese-speaking learners of English and 11 native speakers. The Chinese speakers were exposed to English at different ages after age 4 and had resided in the United States for at least 5 years. Participants completed an oral grammaticality judgment task consisting of declarative sentences, whquestions, ungrammatical wh-questions without subject-auxiliary inversion, and sentences with Subjacency violations. Results showed that those subjects whose age of arrival ranged from 4–7 scored the closest to the native speakers, and there was a clear decline in accuracy as a function of age. There was also a significant negative correlation between accuracy on the GJT and age of arrival in the United States. Johnson and Newport (1991) concluded that Subjacency constraints are subject to maturation, and that past a certain age L2 learners do not have access to Universal Grammar. For a different interpretation of these results, see White (2003a) and Belikova and White (in press). Also focusing on other aspects of grammar assumed to be regulated by Universal Grammar (UG), Coppieters (1987) was the first study to investigate very advanced L2 speakers exposed to the L2 after puberty. Coppieters’s goal was to assess whether non-native speakers who were deemed to pass as native speakers in performance (production), had the underlying linguistic competence of native speakers on a variety of structures, including subtle aspects of semantics. Participants were 21 near-native speakers from a variety of L1 backgrounds who had acquired French as adults, and had spent from 5.5 to 37 years living in France, with a mean of 17.4 years. Coppieters used a 107-sentence questionnaire testing a variety of French structures (some structures were deemed to fall within UG and others were not). Among these structures, the test included 5 sentences testing the past-tense contrast between Imparfait and Passé Composé. Sentences (13) and (14) are two of the five sentences included to test (examples (15) and (16) from Coppieters 1987, p. 559). Speakers were asked to indicate whether Imparfait and Passé Composé were acceptable to them, and if so, whether there was a meaning difference between the two forms. (13) Est-ce que tu {as su / savais} conduire dans la neige? ‘Did you manage/ know how to drive in the snow?’ (14) {J’ai très souvent mangé/Je mangeais très souvent} de la racine d’arnica après cette histoire. ‘I often ate arnica root after that event.’



Chapter 2.  Second language acquisition

Coppieters found clear quantitative and qualitative differences between the native speakers and the near-native speakers, with no single subject from the nearnative group performing like a native speaker. Furthermore, the near-native speakers were more accurate with structures falling within the UG umbrella (complex syntax) than with what Coppieters refers to as “cognitive” or “functional” aspects of language (i.e., semantic), such as the Imparfait /Passé Composé distinction. With these sentences in particular, Coppieters documented the most divergence, concluding that native speakers and non-native speakers did not interpret sentences in the same way. Coppieters’ study appeared to suggest that the interpretation of aspectual contrasts is vulnerable to a maturationally defined age effect. Even when near-native speakers appear to behave like native speakers, their underlying competence may reveal an entirely different picture. Similar conclusions were reached by Sorace (1993) with unaccusative verbs in Italian (§2). Coppieters’ study and conclusions have not gone unchallenged. See Birdsong (1992) for an extensive criticism of Coppieters’ study largely grounded on methodological problems. To summarize, the studies sampled above exemplify incomplete ultimate attainment in specific aspects of phonology, morphology, syntax and semantics. These results have been interpreted as evidence in support of an absolute critical period underlying age effects in L2 learning. We now turn to other studies of the same grammatical areas that challenge these findings and their conclusions.

3.2

The no critical period position

The alternative perspective – that there is no critical period for L2 learning – takes issue with the critical period position on at least two counts. On the one hand, there seems to be no evidence in the literature so far of a dramatic cut-off point for L2 acquisition as there is for L1 acquisition (see Chapter 1). On a strict interpretation of the critical period, we should see more studies with the pattern of data reported by Johnson and Newport (1989), where it was shown that before age 14 (or thereabout), there is a negative correlation between age and proficiency. On a scatterplot, the data distribution takes the shape of a slanted line, typically referred to as a linear function, or the age function. After age 14, there is no correlation between age and proficiency. There is a series of random dots, all over the plane, but no linear pattern, as is illustrated in Figure 2.1. Bialystok (1997) reanalyzed and re-plotted the original data from Johnson and Newport (1989) and found that the break in performance was not as dramatic as originally stated. What we see instead is a pattern of declining proficiency – a

49

50

Incomplete Acquisition in Bilingualism

Figure 2.1  Graphic representation of patterns of responses on a proficiency test assuming a strict version of the critical period hypothesis for L2 acquisition.

linear function – that extends beyond puberty, probably ad infinitum, as shown in Figure 2.2. This pattern is called the age effect. Nobody denies the existence of age effects, or the observed fact that the younger the age of onset of L2 acquisition the more native-like an individual is likely to be. The point of contention is a maturational explanation for these effects. Since age is a macro-variable that cannot be isolated from other co-occurring factors, many other researchers argue instead that apparent age effects are due to other confounding factors, including different environmental input factors and frequency of use of the L1 and L2 (Bialystok & Hakuta 1999; Bialystok & Miller 1999; Flege, Yeni-Komshian & Liu 1999; Jia & Aaronson 1999), maturation and decay of cognitive abilities related to aging (Bialystok & Hakuta 1999), and socioaffective factors (Gardner & Lambert 1972; Krashen 1981). Disagreement on a strict interpretation of the critical period within the generative approach exists as well. All the Full Access accounts fall within the age effects position (Epstein, Flynn & Martohardjono 1996; Schwartz & Sprouse 1996; White 1989, 2003a, among others). Given the fact that the generative approach to second language acquisition embraces the innateness hypothesis, the issue is not whether or not there is a biological basis for language acquisition. What Full Access accounts contend is that access to Universal Grammar does not deteriorate with age. L2 learners have access to Universal Grammar when learning a second



Chapter 2.  Second language acquisition

Figure 2.2  Graphic representation of patterns of responses on a proficiency test with no random performance after the critical period.

language past the critical age, and they can overcome the influence from their L1 and achieve native-like knowledge (i.e., the parameter resetting position). For the Full Transfer/Full Access Hypothesis (FT/FA), for example, complete native-like attainment is in principle possible, but not guaranteed for reasons unrelated to the mal-operation of the innate computational system. Epstein, Flynn and Martohardjono (1996), who downplay (or even deny) the influence of the L1 in the L2 learning process, maintain that access to Universal Grammar is constant from the initial state up to ultimate attainment. Although not explicitly stated, Epstein, Flynn and Martohardjono’s Full Access position, unlike the FT/FA hypothesis, guarantees complete ultimate attainment. Such an absolute outcome is incompatible with the available empirical evidence to date. And in her discussion of Sorace’s (1993) proposal about different types of near-nativeness, White (2003b) even suggests that incompleteness or divergence do not entail lack of UG access in adult L2 acquisition. An L2 grammar can be different from the target and still fall within the hypothesis space of Universal Grammar. The studies selected for discussion in §3.1 all found that age of arrival plays a significant role in the L2 ultimate attainment of pronunciation, morphosyntax, syntax, and semantics, and were interpreted as corroborating maturational constraints for L2 acquisition. Many other studies, however, have contested these conclusions on the basis of the methodology employed in these studies and on

51

52

Incomplete Acquisition in Bilingualism

the interpretation of the results. Furthermore, quite a few other studies appear to support that, while there is a clear effect of age, it is not unlikely to find a few nonnative speakers who perform on a variety of tasks like native speakers in many grammatical areas. A strict critical period explanation predicts there should be none. Unlike the studies discussed in §3.1, these studies tend to focus on individual results, case studies, and on subjects who are pre-screened for nativeness (the “cream of the crop” to use Birdsong’s 1999 descriptor). The logic of this research approach is as follows: First, let’s find late learners who are very proficient and sound like native speakers. Then let’s scrutinize these individuals in more detail under experimental conditions and see how native they are. And this is exactly what Coppieters did, and most recently Abrahamsson and Hyltenstam (in press). By selecting speakers assumed to be at endstate, it is possible to measure their ultimate attainment more accurately, without running the risk of testing subjects who are advanced but are still in the process of development. Studies on age effects that include immigrants with different ages of arrival do not typically control for ultimate attainment. Therefore, it is possible that many of the subjects in these studies, who did not perform at a very high level of proficiency, may still be in the process of learning the language, even after 5 years or more of immersion in the L2-speaking community. If Scovel (1988), Oyama (1976) and Patkowski (1990) all support a critical period for pronunciation, then Neufeld’s studies (1977, 1978, 1979) contradicted the general claim that it is unlikely for L2 learners who have acquired the L2 past a certain age in childhood to exhibit native-like pronunciation. Neufeld tested late L2 learners of Chinese and Japanese in one study (1977) and L2 learners of French in two others (1978, 1979). The participants were asked to practice a few sentences and were later tape-recorded. Native speaker judges rated native and non-native speaker speech samples for accentedness. Neufeld found that a handful of individuals in his studies were rated as native speakers and had an accentfree performance. While Long (1990) found serious methodological flaws in the design of Neufeld’s studies, similar findings were later reported by Flege, Munro and MacKay (1995). Flege et al. (1995) tested 240 Italian participants, who had been living in Ontario, Canada for at least 15 years and most of them indicated that they spoke more English than Italian in their daily lives. The participants were asked to read aloud five short English sentences which were recorded and digitized for acoustic analysis. Twenty-four English native speakers rated the sentences for accentedness. Results showed a clear age effect (earlier arrivals were rated as having more native-like accents in English), but contrary to what Johnson and Newport reported, there was no discontinuity in the ratings of age of arrival after 15 years, or at any other later age. Similar results were found by Yeni-Konshian, Flege and Liu (2000) with Korean-English bilinguals living in the United



Chapter 2.  Second language acquisition

States. The fact that these data do not yield a sharp decline in pronunciation abilities, appears to be a problem for the strong critical period position. In another series of studies, Bongaerts (1999) attempted to find out whether some very advanced late learners can ultimately attain a native-like accent in an L2. One study tested 10 very advanced Dutch-speaking learners of English, 12 Dutch learners of English of different proficiency levels (but not near-natives), and a control group of 5 British English speakers. Four speech samples were taken from each speaker, consisting of an oral report on a recent holiday, reading aloud a short text, 10 sentences, and 25 words. Four native speakers of British English, who were not linguists or teachers, were asked to rate all the randomized speech samples for native accent, on a 1–5 scale. All the very advanced learners passed as native speakers and even received higher scores than the true native speakers in the study. In a second experiment, 10 native speakers of British English, 11 very advanced Dutch speaking learners of English, and 20 learners of different proficiency levels read six English sentences. The speech samples were rated by 13 judges. Next, mean ratings that fell within 2 standard deviations of the native speakers’ scores were calculated. Five of the highly successful learners met the criterion for nativeness with all the sentences. A third study tested whether Dutch learners of French would also attain native-like pronunciation. Four out of 9 successful learners of French scored within the range of nativeness of the native speakers. In conclusion, Bongaerts identified a few individuals who could be considered a counterexample for the claim that there is a critical period for pronunciation. Turning now to morphosyntax, Bialystok (1997), Bialystok and Hakuta (1994, 1999) and Birdsong and Molis (2001) clearly disagree with the conclusions reached by Johnson and Newport (1989) and Johnson (1992) on maturational effects. Bialystok (1997) and Bialystok and Hakuta (1994, 1999) re-plotted Johnsonn and Newport’s original data in a single analysis and graph and found that the linear decline covered up to age 20 years, and not 15 as Johnson and Newport originally reported. After the reanalysis, there was no change in the magnitude of the correlation between age of immigration less than 20 and age of immigration after 20 years. Once the analysis changed, Bialystok and Hakuta (1994, 1999) found no defined cut-off point or sharp decline in learning, but instead discovered a linear function. In addition to issues of proficiency, length of residence and of learning English, Bialystok (1997) questioned the performance of individual subjects on different grammatical structures, raising the possibility that differential performance is related to the L1 of the immigrants (Korean and Chinese), and not to a critical period effect. Could language transfer and linguistic proximity between the L1 and the L2 mask maturational effects? Continuing this line of reasoning, Birdsong and Molis

53

54

Incomplete Acquisition in Bilingualism

(2001) used the same task used by Johnson and Newport (1989) with Korean speakers to study the performance Spanish native speakers. And what Birdsong and Molis found was that a few successful Spanish-speaking individuals did indeed score within the range of the English native speakers. This, they claimed, was probably due to the typological similarity between Spanish and English as opposed to the proximity of Korean and English. Birdsong and Molis also made a case for degree of language use as a reason why these individuals who scored within the native-speakers’ range were so successful. It turned out that amount of English use was positively correlated with the accuracy scores of the late arrivals. The counterpoint to Johnson and Newport’s study on Subjacency comes from White and Genesee (1996) and White (2003a), who dismissed an absolute critical period explanation for the conclusion that failure to reject Subjacency violations indicates that L2 grammars are no longer constrained by Universal Grammar. Based on a different syntactic analysis and on problems with the original test sentences, White (2003a) considers an alternative explanation for Johnson and Newport’s findings. If syntactic movement is not always implicated in wh-movement (Hawkins & Chan 1997), White suggests that learners may represent questions without movement (by generating the wh-phrase in topic position and co-indexed with a pro in the lower clause). Some of the sentences used by ­Johnson and Newport corroborate such analysis. White considers that the Chinese and Korean learners of English do not behave like the English native speakers because they have adopted instead the syntactic analysis of Chinese. The fact that the L2 learners adopt a Chinese-based analysis at some stage is no indication that they have no access to Universal Grammar. Rather, it shows that they have UG-constrained divergent representations. Hence, while accepting Johnson and Newport’s results that there are age effects in the rating of Subjacency violations in English, White disputes Johnson and Newport’s interpretation that inaccurate responses indicate lack of access to UG. White and Genesee (1996) also tested knowledge of Subjacency violations in English. The participants were native speakers of French. French represents Subjacency syntactically like English. Disagreeing with most of the methodological procedures used in studies of maturational accounts, White and Genesee asked whether near-native competence is like native-like competence, and tested very advanced participants who were deemed to perform like native speakers on a number of independent measures (vocabulary, syntax, morphology, etc.) as rated by native speaker judges. The main task was a grammaticality judgment administered on a computer. Both accuracy scores and reaction times were calculated for

. See also Belikova and White (in press).



Chapter 2.  Second language acquisition

each subject. Results showed no significant differences between the native speakers and the near-native speakers on grammatical and ungrammatical sentences, or in reaction times, even when some of the near-natives had acquired English as adults. They found no evidence for maturational effects or a critical period. However, as many others pointed out (Birdsong & Molis 2001; Eubank & Gregg 1999; Long 2007), the fact that the non-native speakers spoke French – a language with overt wh-movement like English – may explain the results as L1 transfer. Finally, does a maturational explanation hold for aspects of semantics? Recall that according to Coppieters (1987), aspectual judgments, among other structures of French, are subject to a critical period and fall outside the phenomena typically covered by Universal Grammar. Montrul and Slabakova (2003) specifically focused on the preterite/imperfect semantic contrast in Spanish and used three instruments: a morphological recognition task, a sentence conjunction judgment task, and a truth value judgment task with stories. One of the instruments included different types of verbal predicates (states, accomplishments, achievements), while the other focused on stative predicates which changed aspectual value depending on the form of the verb, habitual vs. one-time readings of the preterite and imperfect and the generic vs. specific subject interpretation of impersonal constructions with preterite and imperfect forms. As with the Bongaerts (1999) and White and Genesee (1996) studies, a group of 64 non-native speakers of Spanish were subclassified in different levels of proficiency and ultimate attainment based on the combined results of a standardized proficiency test and speech samples rated for nativeness on a variety of grammatical areas by two independent linguistically naïve native speaker judges. As a result, the participants were classified as advanced (n = 24), superior (n = 23), and near native (n = 17). Results showed no statistically significant differences on any of the sentences and conditions of the three tasks between the control group of Spanish speakers (n = 20) and the near-natives. At the individual level, 30% of the entire sample of L2 learners performed within the range of variation of the native speakers, and 70% of the near-natives were indistinguishable from the natives. In contrast to Coppieters’s claims about potential fossilization of aspectual contrasts in the aspectual domain, this study showed that native-like performance (and competence) in the realm of semantic interpretations is possible, even in individuals who are not totally immersed in the language and in individuals who are not necessarily at endstate.10 See Table 2.9 for a summary of all these studies.

10. For extensive evidence that there is no critical period for semantics in L2 acquisition see Slabakova (2006) and Slabakova (2008).

55

Representative Studies Oyama (1976)

Italian immigrants, AoA 6-20 Read aloud task

Method

Syntax Johnson & 23 Chinese (wh-movement) Newport (1991) immigrants 11 native English speakers Grammaticality judgment task Semantics Coppieters 21 near native speak(tense-aspect) (1987) ers of French (all late learners) Questionnaire 20 French native speakers None of the NN speakers scored within range of variation of NS.

Montrul & Advanced and near native Slabakova (2003) speakers of Spanish Sentence conjunction judgment task Truth Value Judgment Task

30% of L2 speakers met criterion for nativeness in all experimental conditions.

Performance of subjects with early arrival was at ceiling. Performance of subjects with late arrival was correlated with AoA. Accuracy on GJT White & Genesee Near native speakers of No differences between correlated with (1996) English (French L1) native speakers and AoA: earlier arGrammaticality judgment near native speakers on rivals better than task reaction times. later arrivals.

5 (25%) non-native speakers met criterion for nativeness in all experimental conditions.

findings

No Support for Critical Period in L2 Acquisition

Representative Method Studies Effect for AoA. Bongaerts (1999) 20 advanced English learnYounger arrivals ers of Dutch had more native11 British English speakers like pronunSpeech samples rated for ciation than older accentedness by 13 judges. arrivals. Effect for AoA. Birdsong & Molis Replication of Johnson & Early arrivals (2001) Newport (1989) with 61 outperformed Spanish-speaking subjects late arrivals. (early and late arrivals)

findings

Support for Critical Period in L2 Acquisition

Morphology Johnson & 46 Korean and (verbal, nominal Newport (1989) Chinese immigrants morphology) AoA 3-39 Oral grammaticality judgment task

Pronunciation (phonology)

Linguistic Domain

Table 2.9  Summary of sample of studies in favor of and against the Critical Period Hypothesis in adult L2 acquisition.

56 Incomplete Acquisition in Bilingualism



3.3

Chapter 2.  Second language acquisition

Summary

Age effects on ultimate attainment in L2 acquisition appear to be quite robust in different grammatical areas. However, there is no consensus that these age effects are necessarily related to or suggestive of a critical period as in L1 acquisition since, with the exception of the findings of Johnson and Newport (1989, 1991), there is little evidence of abrupt cut-off points or sharp declines in L2 learning ability after a certain age. Rather, age effects appear to occur ad infinitum throughout the lifespan. Furthermore, other studies have shown that there are quite a number of successful individuals who can perform within the range of variation of native speakers. Criticizing the counterarguments and counterexamples to the strong critical period position, Long (2000, 2007) has questioned the methodologies, the narrowness of the linguistic foci, and the interpretations and the generalizability of results in support of the view that overall native-like proficiency is possible in many grammatical areas. But the debate is not over yet. At present, what we know about maturational constraints in adult L2 acquisition and the open questions that remain are as follows: 1. Very often, the degree of ultimate attainment in L2 acquisition diverges from that of native speakers: it can be incomplete, divergent or indeterminate (variable) in a number of grammatical areas. 2. L1-influence can be overcome, but it can also persist and lead to fossilization in ultimate attainment. 3. Degree of ultimate attainment in L2 acquisition is selective: it differentially affects phonology or pronunciation as opposed to morphosyntax, syntax, and semantics. 4. Although native-like competence and performance is not guaranteed, age of arrival (or age of acquisition) is the strongest predictor of native-like ultimate attainment by adults. 5. Available empirical evidence appears to be more consistent with a strong main effect for age extending linearly throughout the lifespan and beyond puberty but not with a critical period: there is little evidence of an abrupt cutoff point or decline in the ability to learn a second language past a certain age (e.g., 7 years as in L1 acquisition). 6. There are many high proficiency individuals (up to 30% depending on the study) whose linguistic behavior on a variety of specific tasks and grammatical domains falls within the range of variation of native speakers, even when these individuals started learning the L2 late, after the critical period.

57

58

Incomplete Acquisition in Bilingualism

7. Among studies focusing on near-natives, few have shown that native-like attainment is possible for the same individual(s) in a number of grammatical areas and under different experimental conditions (i.e., different tasks).11 The sample of studies just discussed shows that the issue of maturational effects in L2 acquisition is far from settled. For some researchers, the glass is half full, for others half empty. There is a sense that both maturational and input factors contribute to linguistic success in L1 acquisition, but it is not clear how one or the other explains variability of outcomes in adult L2 acquisition. All research discussed in this chapter was conducted exclusively with late bilinguals. But very little is known, for example, about child L2 acquisition during childhood, within the critical period. Linguistically-oriented research investigating how child L2 differs or not from adult L2 acquisition and child L1 acquisition has not received the same attention as adult L2 acquisition, although this situation is currently changing (Gavruseva & Lardiere 1996; Haznedar & Schwartz 1997; Ionin & Wexler 2002; Lakshmanan 1995; Prévost 1997; Prévost & White 1999; White 1996; Schwartz 2003, 2004; Unsworth 2005, 2008). If children start L2 acquisition before the ages of 7–8 years – before the critical period – it is believed that they often achieve native-like proficiency, especially in pronunciation, and do not fossilize, unlike their adult counterparts. However, very recent research puts these long held ideas into question. McDonald’s (2000) study of the acquisition of L2 morphosyntax by young Vietnamese children, Flege, Munro and MacKay (1995) study of L2 English pronunciation by Italian children and Flege, Yeni-Komshian and Liu (1999) study of English L2 pronunciation by Korean children all show that children do not perform at the same level of native speaker counterparts, even after 4 years of immersion. Hulk and Cornips’s (2006) and Unsworth’s (2008) studies of gender agreement in Dutch suggest that the generalization that fossilization at the morphosyntactic level does not occur in child L2 acquisition may also be overstated. The study of child L2 acquisition is another missing link in our understanding of age effects in L1 and L2 acquisition. Yet, a line of research that captures very nicely potential differences between L1 and L2 acquisition as a function of age is the study of second language acquisition in the deaf (Mayberry, in press; Mayberry & Squires, in press; Mayberry & Locke 2003; Mayberry, Locke & Kazmi 2002). Recall from Chapter 1 that sign language research provides the closest occurring approximation to an empirical study of the critical period. This is because many deaf individuals are exposed to sign language as a first language at different ages in childhood. At the same time, sign language is also learned as a second language, when hearing family ­members 11. See Marinova-Todd (2003).



Chapter 2.  Second language acquisition

wish to communicate with a non-hearing relative, or when people become deaf after they have learned their language orally (i.e., cases of acquired deafness). To give just one example of this fascinating type of work, Mayberry (1993) and Mayberry and Eichen (1991) compared age of onset of acquisition in L1 and L2 learners of American Sign Language. Following Newport (1990), Mayberry tested knowledge of morphology and syntax in three groups of signers: a group that had acquired ASL between ages 0–3 (native group), a group whose onset of acquisition was between the ages of 5–8 (child signers), and a group that started learning ASL much later (9–13) (late signers). There was also a group of L2 signers: people who knew spoken English but became deaf between the ages of 9–15. All groups had been using ASL as a first or second language for at least 20 years. The task used was an elicited imitation task targeting several syntactic and lexical aspects of ASL. Among the L1 acquirers, the results were very similar to those of Newport (1990): the native signers outperformed the child signers and the late signers, and the child signers were more grammatically accurate than the late signers. In light of the focus of this book, the most interesting finding from this study is that the L2 learners of ASL received higher scores on morphology and syntax than the child and late L1 learners, yet they did not match the performance of the native group. These findings suggest that previous knowledge of a language facilitates acquisition of another one, but complete success, as in L1 acquisition, is not guaranteed. The study also suggests that critical periods are more “critical” if an opportunity to acquire an L1 is missed, but less dramatic in the acquisition of an L2. In many ways, the results of the L2 studies discussed in this chapter point to a similar conclusion. L2 acquisition is possible and doable, not a hopeless endeavor, as with Genie and Chelsea discussed in Chapter 1. L2 learners can reach high command and very sophisticated knowledge of the second language, yet few will fall within the range of performance of a native speaker on a number of grammatical and language use measures. Nonetheless, age effects are much more dramatic and irreversible when input is received later for L1 acquisition. Both maturational and input factors account for this performance.

4.

Age of acquisition and first language loss

Up until now, most research on the critical period in L2 acquisition has focused exclusively on late bilinguals immersed in a second language environment. The usual theoretical focus has been a contrastive analysis of child L1 and adult L2 acquirers. Yet the Critical Period Hypothesis has implications for early bilingualism as well, both simultaneous and sequential, and until recently this area of study has received very little attention. If we wish to make progress in our ­current ­understanding of

59

60 Incomplete Acquisition in Bilingualism

age effects in L1 and L2 acquisition, we must look at the problem from different angles. In this book, I expand the empirical base by asking whether there are age effects, or a critical period, for L1 loss in early bilingualism as well. If early age of acquisition is crucial for full linguistic development, this also raises the question as to the fate of the L1 in a dual language context. The next three chapters focus on the flip side of the second language acquisition coin: the loss of the first language, or one of the first languages, in simultaneous acquisition in early and late bilinguals. By definition, a bilingual is an individual exposed to two or more languages. As a result, bilinguals receive less input in the two languages than a monolingual child receives in one. How does reduced input affect the linguistic development of the two languages? If reduced input affects degree of language knowledge or access to that knowledge, how does it interact with age? Under a view of the critical period according to which the cortical representations subserving early L1 and late L2 acquisition are very different, exposure to a language early in childhood should have long-lasting traces in the neural circuits subserving language processing, representation, and use. This means that once completely acquired and stabilized (i.e., crystallized), knowledge of the L1 should be very difficult, if not impossible, to lose under normal circumstances (i.e., in the absence of pathology) in a dual language context, even when input is no longer available. However, in cases of early bilingual acquisition and development into adulthood, if input to the first languages is severely reduced or interrupted altogether during childhood, degree of language loss should also be a function of age. Therefore, the following two hypotheses will be examined in the next three chapters: Hypothesis 1 If L1 attrition occurs in children, it will be more severe than L1 attrition in adults. That is, language loss should be more dramatic in early than in late bilingualism. Hypothesis 2 If language attrition occurs within early (pre-puberty) bilingualism, it will be more severe in simultaneous bilinguals (exposed to the two languages very early) than in sequential bilinguals (when the L1 was acquired before the L2). These two hypotheses suggest that we should find an age function in the opposite direction, as implied by the progression observed in the case studies of Genie, Isabelle and Chelsea: the earlier the deprivation of the L1, the greater the loss. And is there a specific age range after which susceptibility to loss is entirely unlikely, signaling an absolute critical period?



Chapter 2.  Second language acquisition

The next three chapters will test the predictions of the critical period for language loss in bilingual immigrants and examine some empirical evidence. Chapter 3 explores L1 attrition in adult L2 learners who are first generation immigrants. Chapter 4 investigates language loss and incomplete acquisition in early bilingualism, both simultaneous and sequential bilinguals of pre-school age. Chapter 5 turns its attention to simultaneous and sequential bilingualism during the school-age years, tracing the gradual loss of the L1 as the L2 gets stronger through literacy. Chapter 6 focuses on incomplete acquisition in adult early bilinguals. Finally, Chapter 7 returns to the issue of a critical period in both early and late bilingualism, to assess the effects of age in incomplete acquisition under L2 acquisition and L1 loss. In Chapter 8, I attempt to delineate how the age functions for acquisition and loss intersect at a particular age in middle childhood.

61

chapter 3

First language attrition in adults

The studies on L2 ultimate attainment discussed in the previous chapter, and especially studies of near-native speakers, make it clear that adult L2 learners ­acquire complex grammatical knowledge of the L2 and are able to reach quite advanced levels of attainment. Nevertheless, highly complex near-native grammars may turn out to be non-identical, incomplete, or divergent from native-speaker grammars in several respects. I demonstrated that L1 influence can persist at this stage, and may even be a cause for the apparent fossilization of linguistic features that some L2 speakers manifest in their linguistic behavior. Most of the studies discussed in Chapter 2 involved adult L2 speakers who immigrated to the host countries (the L2 context) relatively late in life. Some others compared early and late arrivals, however. Before the late arrivals became fully immersed in an L2 environment, their respective native languages had a chance to develop completely. L2 acquisition is a particular type of bilingualism, and a critical question in bilingualism is to understand how the two languages interact and relate to each other in the bilingual mind. This chapter, therefore, investigates what happens with the mature L1 after extensive and prolonged immersion in the L2 environment, regardless of whether these L2 speakers achieve a high level of proficiency in the L2. Recall Francesco, one of the hypothetical bilingual profiles introduced in Chapter 1. Can a language acquired completely in childhood during the critical period and reinforced through literacy at school, well before the onset of bilingualism, be lost or forgotten in some sense? This is the phenomenon of L1 attrition. In this chapter, I will only discuss L1 attrition in adults, while the following chapters will focus on children. In Chapter 2, I spelled out two hypotheses regarding age effects and language loss. I will now show how Hypothesis 1 – language loss should be more dramatic in early (children) than in late bilingualism (adults) – is largely supported by the empirical evidence. My main claim is the following: although there are attrition effects in a variety of grammatical areas, the extent of L1 loss is relatively minor in adults, especially when compared to the extensive attrition observed in children, as I show in the following chapters. The grammatical system per se is not deeply compromised under attrition. And the mild effects observed at the level of discourse-pragmatic and phonological knowledge are often easily explained by the influence of the L2 onto the L1. The conclusion I will ultimately reach is that the

64 Incomplete Acquisition in Bilingualism

extent to which the grammar is affected by L1 attrition in adults and in children may indicate different mechanisms responsible for attrition according to age of onset of bilingualism, as the Critical Period Hypothesis predicts. This chapter is organized into 4 sections. Section 1 defines L1 attrition and discusses methodological issues related to the empirical study of this phenomenon. In §2, I discuss three existing theoretical approaches to L1 attrition and present some empirical evidence. In §3, I relate L1 attrition to L2 acquisition by focusing on the role of crosslinguistic influence, or L2 influence on the L1. Finally, in §4, I return to the issue of age to make sense of the patterns of attrition attested in the literature.

1.

Language attrition

Language attrition is a particular case of language loss at the individual level. An individual can suffer language loss as a result of neurological insult (aphasia or other neuro-debilitating conditions such as Alzheimer’s disease or other forms of dementia), or more gradually in other non-pathological circumstances, as a result of the normal aging process. Non-pathological language loss, however, is most often discussed in the context of bilingualism and language contact situations, and in this book, I will only be concerned with such cases. Attrition in individuals occurs during the first generation of immigration, according to de Bot (1991). Attrition appears to affect structural aspects of the L1 as a result of language shift, or a change in the relative use of the L1 and the L2. At the same time, I must clarify that just as in the case of second language acquisition, where extensive naturalistic exposure to the L2 in an L2 environment does not guarantee complete, native-like, ultimate attainment, language shift from the L1 to the L2 does not inevitably lead to L1 attrition either. Seliger (1996, p. 616) defines attrition as “the temporary or permanent loss of language ability as reflected in a speaker’s performance or in his or her inability to make grammaticality judgments that would be consistent with native speaker (NS) monolinguals of the same age and stage of language development.” For Pavlenko (2003), attrition can only be applied to permanent, rather than temporary, linguistic deficits in the L1. Note that this pattern is similar to the concept of fossilization in L2 grammars, and the debate over whether fossilization is stabilization or not (Han 2004; Lardiere 2006; Long 2003). Implicit in Seliger’s characterization of attrition, but explicitly addressed by Sharwood Smith (1983, 1989, 2007) and Sharwood Smith and Van Buren (1991), is the competence/performance theoretical dichotomy. Does attrition involve qualitative changes in the actual L1 system at the level of



Chapter 3.  First language attrition in adults

representation, or does it just involve difficulties with access and control of the linguistic system during linguistic performance? My own take on this question is that attrition in adults affects primarily performance (retrieval, processing, and speed), but does not result in incomplete or divergent grammatical representations, as may be the case in adult L2 acquisition or children undergoing L1 attrition.

1.1

Factors contributing to attrition

If fossilization is not a guaranteed outcome of late L2 acquisition, bilingualism itself does not guarantee or determine that language attrition will occur in a given speaker either. L1 attrition depends on a variety of interacting external factors typically studied within sociolinguistic perspectives, such as age and length of immigration, level of education, amount of contact with L1, availability of a speech community (L1 input), and degree of L1 and L2 use, among others (­Ammerlaan 1996; de Bot 2007; de Bot, Gommans & Rossing 1991; Gürel 2007; Hulsen 2000; Hutz 2004; Köpke 2007; Schmid 2007; Yağmur 1997). Affective factors related to language identity and attitude, motivation, and emotion (Pavlenko 2003; ­Pavlenko & ­Jarvis 2002; Schmid 2002), also seem to play a role, although it is not yet clearly understood how these factors individually and/or collectively affect the process and outcomes of attrition directly. Of all the factors mentioned above, I will consider age of onset of bilingualism in more detail because age seems to be the strongest predictive factor for attrition, even though age itself, being a macrovariable, cannot be properly isolated from the other encompassing factors mentioned above as well as neurocognitive factors including brain plasticity, long and short term memory capacity, declarative and procedural memory, among others. Age of onset of bilingualism and degree of attainment in the L1 are intrinsically related to the extent of attrition in a given grammar. If an L1 has been acquired completely and has remained relatively “stable” for a while, as in late childhood or adolescence, L1 attrition at the competence level is unlikely (de Bot & Clyne 1989). But if the L1 grammar did not develop fully, especially in early to middle childhood, age of onset of bilingualism and extensive exposure to the L2 matter significantly for L1 attrition, as I show in subsequent chapters.

1.2

Linguistic process and outcome of attrition

Granting that there are sociolinguistic and affective factors contributing to L1 attrition, the most theoretically appealing aspect of attrition is that it has

65

66 Incomplete Acquisition in Bilingualism

c­ onsequences for specific areas of grammatical knowledge. Just like fossilization in the L2, L1 attrition affects the form and structure of language, but it does so selectively. Attrition can affect different grammatical areas and linguistic abilities, including dysfluency and inability to retrieve words, pronunciation of the L1 with foreign accent, acceptance and production of morphological errors (inflectional morphology, case, gender agreement, etc.), and acceptance and/or use of syntactic structures that are deviant in the L1 (Seliger 1996). For some researchers, code-switching may be a sign of language loss, when used indiscriminately without regard for interlocutor and context, as a result of difficulty retrieving lexical items on-line in the L1 (Pavlenko 2004). From a linguistic perspective, the selectivity of attrition raises fundamental questions about the nature of mature linguistic competence of the idealized native hearer-speaker, and about the nature of language. For example, why does attrition target certain grammatical areas and not others? Which are the grammatical areas that are most vulnerable to attrition and why? The selectivity of attrition by grammatical domain and language skills, much like the selectivity of fossilization in L2 acquisition, supports the modularity of grammatical subsystems (e.g., knowledge of phonology, morphology, syntax, semantics, pragmatics) and the links or integration of these structural areas of knowledge with one another and with other grammar- external systems (also known as interfaces). Recent work within the generative framework – notably Sorace (2000a) and Tsimpli, Sorace, Heycock and Filiaci (2004), and Tsimpli (2007) – sought to address these questions in adult L1 attrition. Seliger (1991) suggests that attrition is a “rule-governed” process. In many ways, these are the same processes attested in early stages of L1 and L2 acquisition and in creole genesis, such as transfer or crosslinguistic influence from the L2 onto the L1, and rule generalization and preservation of unmarked over marked parameter settings and forms. Lending support to the modular architecture of language, simplification and loss of restrictions can happen at the level of phonology, morphology, syntax, semantics, and pragmatics. In general, due to crosslinguistic influence or transfer from the L2 onto the L1, lack of use, and the operation of some sort of leveling mechanism, alleged attrited grammars display restructuring of the L1 in a number of grammatical areas. Restructuring typically occurs in morphosyntax, and refers to the deletion of certain grammatical elements of the L1 or the addition of elements of the L2 into the L1. For example, a three-way case system may be reduced to a two-way system, as in the loss of structural dative case in the history of English (Lightfoot 1991) and retention of nominative and accusative, or the loss of VS



Chapter 3.  First language attrition in adults

word order and the emergence of dislocated subjects in Brazilian Portuguese due to the ­erosion of the agreement paradigm related to the null subject parameter. Restructuring can also lead to convergence, when the linguistic systems of the L1 and the L2 come together with mutual influence from one language onto the other, or when the two languages influence each other but the transfer is not unidirectional, as defined by Bullock and Toribio (2004). A typical case of convergence is observed in the phonological system of bilingual grammars, where some sounds can acquire some intermediate values between L1 values and the L2 values, as with Voice Onset Time (VOT) in stop consonants (Flege 2002), or the adoption of sounds that combine phonetic characteristics of similar L1 and L2 sounds (Bullock & Gerfen 2004). Given these linguistic manifestations of attrition that affect various formal aspects of language, let me reiterate that, in the case of adults, there is no total or global loss of the L1. Here again, the phenomenon is akin to fossilization discussed in the context of L2 acquisition. Attrition, like fossilization, is localized. Many highly proficient L2 speakers have near-native knowledge of many aspects of the L2, yet some aspects may remain non-native like. Similarly, attrited L1 speakers preserve a very high overall command of their L1, while some aspects of their L1 become non native-like, typically pronunciation and lexical retrieval. Because most attrition effects have been found in these areas, many researchers believe attrition is governed by cognitive processes rooted in memory (see review in Köpke 2007). L1 attrition amounts to L1 forgetting, as captured precisely by Polinsky’s (1997) use of the term “L1 forgetters.” Given this view of attrition, the crucial question concerning adult mature speakers, therefore, is whether non native-like features in their L1 indicate difficulty with access and retrieval of linguistic elements, as Sharwood Smith (1983, 1989) asked, or whether they are indicative of erosion at the level of linguistic competence (i.e., representation). In §2 I review, and critically assess, existing theoretical approaches that attempt to account for the observed effects. But before I do so, let me state explicitly some of the empirical problems that are currently pervade this fascinating field of inquiry.

. Although these are examples from diachronic change, diachronic change is the aggregate of individual grammars across generations. Many of the processes involved in attrition are the same as the processes that take place in language acquisition, bilingualism and diachronic change. For Lightfoot (1979), language acquisition is the medium through which a language change is transmitted over time.

67

68 Incomplete Acquisition in Bilingualism

1.3

Baseline knowledge in language attrition research and other methodological considerations

Although L1 attrition research in a bilingual situation started almost three decades ago, sustained interest in understanding the phenomenon from a ­variety of theoretical perspectives has only emerged recently (de Bot 2004; Köpke 2004; Köpke, Schmid, Keijzer & Dostert 2007; Schmid & Köpke 2004). Presently, there is vivid discussion concerning the empirical methods and instruments that should be applied to investigate the phenomenon. Language attrition research presents some of the same challenges posed by the investigation of fossilization in advanced stages of L2 acquisition. To study attrition under ideal conditions, a diachronic (longitudinal) research design is warranted. This is because the best way to establish that a certain grammatical feature of a given speaker has fossilized in the L2, or has been lost in the L1, is to conduct a longitudinal investigation of the same speaker over an extended period of time, and to have at least two measures: one at Time 1 and another at Time 2, several years apart. In fossilization studies, it is important to establish that the speaker has reached a stable state of competence, with no change or development from Time 1 to Time 2. In Chapter 2, I discussed one study on fossilization that met this ideal extended period requirement, namely Lardiere’s (1998a, b, 2007) studies. In language attrition studies, on the other hand, one is looking for a perceptible and measurable change between Time 2 and Time 1, such that the L1 was at native speaker norms at Time 1, but less native-like at Time 2. Studies reporting meaningful measurable attrition effects have typically found them in speakers who have been living in the L2 environment for more than 10 years. However, except for de Bot and Clyne (1989, 1994) and Hutz (2004), rare are the L1 attrition studies that follow an individual for such an extended period of time. According to Seliger (1996), and as his definition of primary language attrition implies, monolingual control groups can provide a valid and accurate representation of the L1 state before the onset of bilingualism and potential attrition. Most language attrition studies that include a baseline group resort to this solution. Thus, for a number of practical reasons, the idealized monolingual speaker is still the norm, or comparison group, for both L2 acquisition studies and L1 attrition studies, even though it is widely recognized that bilingual grammars are different from monolingual grammars and should not be regarded as the combination of two encapsulated and autonomous linguistic systems (Cook 2003; Grosjean 1989; Mack 1997). Because of this monolingual-comparison methodological necessity, the “incompleteness,” “divergence” and “deficit” metaphors are pervasive characterizations of the bilingual state of knowledge, both in the context of L2 acquisition and L1 attrition.



Chapter 3.  First language attrition in adults

Another important consideration when investigating potential L1 attrition effects in L2 speakers is whether attrition is defined at the level of language performance and production, at the level of competence (i.e., represented linguistic knowledge), or at the level of access to knowledge (automatic and fast on-line processing). We know that tapping linguistic competence directly is impossible; we can only rely on tasks that measure different types of linguistic behavior suggestive of implicit linguistic competence. Keeping task difficulty constant in a given study, in order to actually claim that a certain linguistic domain is affected by language attrition, one must demonstrate that the change occurs at different levels of knowledge (representation, performance and processing), or skill (production, comprehension, judgment). Most often, however, attrition effects in adults are documented at the level of processing and production, especially with vocabulary. Without comparable effects in comprehension, for example, production and processing effects may simply reflect a superficial loss of automaticity and fast retrievability rather than actual loss of deep linguistic knowledge. Finally, if the subjects studied are too old (older than 65), it is hard to tell whether the attrition effects observed are the result of lack of L1 use in an L2 context, part of the normal aging process, or some combination of the two. In the next section, I discuss classic and more recent studies focusing exclusively on adults and meeting some of the basic methodological standards mentioned above. Potential attrition effects are examined through the lens of current theoretical approaches: the Regression Hypothesis (Jakobson 1941), the generative approach (Gürel 2002; Sorace 2000a; Tsimpli 2007), and the Activation Threshold Hypothesis (M. Paradis 2004, 2007).

2.

Theoretical models of attrition

2.1

The regression hypothesis

Jakobson’s (1941) Regression Hypothesis is the earliest and most controversial model discussed in language attrition research. This hypothesis attempts to explain the selectivity of attrition by specifically relating language attrition to the process and developmental stages of L1 acquisition. The central claim is that acquisition is the mirror image of attrition: structures acquired late in L1 acquisition will be the first affected in L1 attrition. This is also known as the last in, first out principle. Several researchers agree that sequences of L1 acquisition and L1 attrition appear related in reverse (Andersen 1982; Pan & Berko-Gleason 1986). With some already established stages of development and processes in L1 acquisition of many languages, it is easy to verify whether these same stages of ­acquisition

69

70 Incomplete Acquisition in Bilingualism

and processes are also affected under L1 attrition, a condition which makes this hypothesis empirically appealing and testable. The Regression Hypothesis captures, in some sense, the well-observed processes of simplification, redundancy reduction, or the resort to unmarked universal values, involved in language attrition. Children’s acquisition of language is assumed to be governed by universal learning principles and markedness considerations. Unmarked sounds, features, and parameter values are typically learned before more marked ones (e.g., coronal sounds are learned before non-coronal sounds, regular paradigms are learned before exceptions, simple sentences are learned before complex sentences, etc.). Most basic universal features of language are learned before language-specific features. Thus, in attrition, we observe the reverse: loss of marked language-specific features, regularization of irregular morphological patterns, difficulty processing complex sentences. Like the Critical Period Hypothesis, the Regression Hypothesis was formulated on the basis of neurological evidence from aphasia, although it has not survived empirical scrutiny with aphasia research (Berko-Gleason 1982; de Bot et al. 1991; Caramazza & Zurif 1978; Grodzinsky 1990). This is because language acquisition by children is a gradual process consisting of well-delineated stages or linguistic milestones. If, by contrast, language loss due to aphasia is abrupt, and the nature of linguistic knowledge retained must depend on the specific, localized area of the brain injured, it is not surprising that the structures affected in acquisition and attrition have not always coincided. If the gradualness of the process in the absence of pathology is relevant, it stands to reason that normal language loss in a bilingual environment provides a more suitable testing ground for the Regression Hypothesis. Yet, surprisingly, very little research on L1 attrition has been conducted with normal language loss of this kind (Keijzer 2007). If the Regression Hypothesis can indeed be extended to investigate the normal process of attrition, its level of applicability still needs further, more precise, clarification. What does the Regression Hypothesis predict for the attrition of different grammatical domains: namely phonology, morphosyntax, lexicon, semantics, pragmatics? Within a single domain, how does order of acquisition interact with structural complexity? How does regression apply to linguistic abilities, such as perception, comprehension, and production? Although Jakobson’s evidence for the Regression Hypothesis was based largely on phonological features and phonological processes from Slavic, it remains to be seen how this hypothesis would explain the attrition of phonology (or pronunciation) in a bilingual environment, for example. Phonology and phonological contrasts are acquired before the emergence of the lexicon and grammar in L1 acquisition. If pronunciation is a lowerorder process dependent on early maturation and less adaptive macroneural circuits, early maturation would explain why it is very difficult to achieve native-like



Chapter 3.  First language attrition in adults

pronunciation in the acquisition of an L2 after childhood. But if phonetics/phonology is one of the earliest acquired aspects of language in infancy, it follows that it should also be the most resistant to attrition. However, there is actually empirical evidence for the development of a non-native accent in the L1, due to changes in the L1 phonetic values under influence from the L2 (Major 1992). The Regression Hypothesis may hold some promise to understand morphosyntactic attrition, since in many languages there are well-documented stages of development in this grammatical domain. As I showed in Chapter 2, inflectional morphology is a highly problematic area for complete mastery in adult L2 acquisition, and the question is whether it is equally affected under L1 attrition. A few studies have investigated potential attrition of the complex gender and number inflectional suffixes in German adults (Altenberg 1991; Gross 2004; Hutz 2004; Köpke 1999; Schmid 2002), and found some overregularization errors like those produced by German-acquiring children (e.g., the most frequent plural morpheme (e)n was overextended to other irregular forms). In general, the percentage error rate in all these studies was lower than 5% (15 out of 1,000 words for all 35 German Jew speakers combined, or 1.5%, in Schmid’s 2002 study, for example). As for syntax, Yağmur (1997, 2004) tested production of relative clauses in two groups of adult Turkish immigrants in Australia who differed in levels of education. Since relative clauses are acquired after age 5 by Turkish-speaking children, Yağmur predicted that these constructions would be vulnerable to attrition (indirectly assuming the Regression Hypothesis). He found that the immigrant groups (regardless of level of education) did not produce object and subject relative clauses in an elicited production task to the same extent and with the same accuracy as two control groups of Turkish speakers in Turkey. . German nouns belong to one of three genders: masculine, feminine and neuter. Plural morphology is complex and full of irregularities. Some linguists claim that German has only five plural allomorphs (Clahsen, Marcus, Bartke & Wiese 1996), but others contend there are nine (Köpcke 1988). There is great variation in the frequency of these allomorphs, but overall -en is the most frequent and -s is the least frequent. Plural allomorphs are assigned to nouns on the basis of gender and phonological ending. Although the German nominal inflectional system is replete with irregularities, the most robust rule appears to be that feminine nouns ending in /e/ take (e)n. Other plural allomorphs vary as a function of the type of derivational affix preceding them. Plural marking in German is also difficult for German-speaking children to master, and even 6 year olds still make errors (Schmid 2002, p. 117). The most common error is to overgeneralize (e)n, the most frequent morpheme, to other less frequent regular and irregular forms (Clahsen, Rothweiler & Woest 1990). . Clearly, complexity and order of acquisition are confounded in the case of relative clauses, so it is not possible to conclude that difficulty with these structures are due to order of acquisition alone, as the Regression Hypothesis would predict.

71

72

Incomplete Acquisition in Bilingualism

But in order to test the Regression Hypothesis at least two main conditions have to be met. First, the study must focus on individuals assumed to have ­completely acquired the language, typically adults. Second, at least two structures must be tested, one acquired earlier than the other in L1 acquisition. A recent study meeting these minimum methodological requirements designed to specifically test the Regression Hypothesis is Keijzer (2007). The study focused on adult Dutch immigrants in Ontario, Canada, and tested experimentally 15 morphological and syntactic features of Dutch, specifically chosen on the basis of the developmental sequences attested in the published literature on the L1 acquisition of Dutch. In the realm of inflectional morphology, the study targeted nominal and verbal inflection (gender agreement in determiners and adjectives, plural, agentive morpheme, diminutive, simple present, simple past, past participle, auxiliary selection, future tense). Command of morphology was assessed through an elicited oral production task with nonsense words (the Wug test) and a film retelling task. As for syntax, knowledge of negation, passives with transitive and intransitive verbs, V2, subordination and word order with particles and auxiliaries were assessed with the film retelling task (for those structures that are frequent in production) and with a written, untimed, grammaticality judgment task. For each grammatical property tested, Keijzer gives a table illustrating an approximate order of acquisition, frequency of the construction in Dutch and equivalence in English. To give just one example, passives are acquired between the ages of 2;6 and 5;00 and are of mid frequency. But the passive with transitive verbs is acquired earlier than the passive with intransitive verbs (impersonal passives). Whereas English has passives with transitive verbs, it does not have passives with intransitive verbs. Thus, the prediction is that passives with intransitive verbs will be more affected by attrition in the Dutch speakers living in Canada than passives with intransitive verbs. Keijzer formulated similar implicational predictions for each structure tested. The participants were 45 speakers of Standard Dutch who emigrated to ­Canada after age 15 and had all lived there for at least 20 years. Their mean age was 66.4 years (range 41–79). This experimental group was compared to two control groups of Dutch speakers living in The Netherlands; a group of age-matched adult Dutch speakers (mean age 66.20) and a group of 35 teenagers (ages 13–16). Interestingly, this last group is labeled the “acquisition” group. Even though all the structures tested in this study are allegedly acquired by age 5–6 or later (with later never being defined in the text), Keijzer motivates the inclusion of this younger group because they represent Dutch “language users who were on the brink of . Keijzer also used a cloze test to establish overall proficiency, a sociolinguistic questionnaire, and a self-rating questionnaire including “can do” lists.



Chapter 3.  First language attrition in adults

Table 3.1  Overall performance of the three groups on different measures (adapted from Keijzer 2007). Groups

N

Cloze

Wug Test

Attriters Adult controls Acquirers

45 45 35

61.07 89.95 45.25

Noun phrase 71.68 89.8 79.17

GJT Verb phrase 80.9 95.5 87.26

69.15 78.53 71.19

Note: For comparability across tasks, all scores have been converted to percentages.

complete L1 mastery, but who might nonetheless show optionality where mature grammars do not” (Keijzer 2007, p. 150). The most significant pattern of results of this study is captured in the overall scores shown in Table 3.1. In general, the attrited group performed on all measures at the level of the teenagers, and both groups differed from the control group of adult Dutch speakers. Overall, Keijzer found that level of education was a significant predictor of accuracy in all tests for all subjects, including the adult Dutch controls. Those with a lower level of education produced more errors than those with a higher level of education. How does this study support the Regression Hypothesis since normal adults under attrition have never been shown to have regressed to the linguistic knowledge of a 3–5 year old? In general, errors in this study – as in most other attrition studies already referred to – were minimal, but when attested, errors affected morphology more than syntax. Results of the Wug test, which tested the morphology of the noun phrase, revealed significant differences between the attriters and the adult control only on feminine agentive formation and diminutives and adjectival agreement (with the most irregular and infrequent forms as in child language). Errors with verbal morphology were even fewer. As for syntax, only V2 and subordination in the GJT were found to differ between the attriters and the adult controls. In general, since very few differences between the groups were found, Keijzer observes that “at no point were dramatic changes perceived in the attrition of Dutch in Dutch Canadian émigrés” (Keijzer 2007, p. 266). Are the few morphological errors attested evidence for the Regression Hypothesis? The errors produced involved regularization of irregular forms, such as the replacement of the agentive suffix -aar by -er resulting in the nonsense word *linderer for linderaar in the Wug test, the use of the common determiner de for neuter het, or attaching the default diminutive -tie instead of -pje to nouns. Although these are the same types of errors produced by normally developing Dutch children, the mild attrition effects found here are also amenable to another explanation. If morphological irregularity is part of the lexicon, while ­morphological

73

74

Incomplete Acquisition in Bilingualism

r­ egularity is handled by a productive rule, idiosyncratic lexical information is more prone to erosion than productive morphological rules. This dissociation between regular and irregular forms may point to a retrieval problem (related to memory) rather than a competence problem. It also speaks to the theoretical distinction later made by the Dual Mechanism Model of inflection (Pinker 1999; Prasada & Pinker 1993; Pinker & Ullman 2002a, b). According to this model, regular inflections (e.g., talk-ed) are handled by a symbol-manipulating rule generated by the computational system, whereas irregular inflections (e.g., spoke) are part of the lexicon, learned by associative memory. Ullman (2001) proposed that two different types of memory (procedural and declarative) and different neural substrates are involved in processing regular and irregular morphology: regular inflection is stored in procedural memory, irregular inflection in declarative memory. The underlying neural substrates subserving declarative memory have been shown to be negatively impacted by maturational effects in L2 acquisition (Birdsong & Flege 2001). Similarly, declarative memory appears affected under L1 attrition. This suggests that while the lexical retrieval of morphological forms is vulnerable to attrition effects, the computational system itself (grammar) is not. And this is also consistent with the findings of the GJT in Keijzer’s study, where practically no attrition was observed with the syntactic structures tested. And the few errors observed with V2 – in oral production, for example – can be attributed to influence from English, a language without V2. Keijzer admits that none of the errors attested in the attrition group can be due only to restructuring of the L1. Most of the errors can also be explained by influence from English. Finally, it is not clear why the teenagers in this study were considered the acquisition group, when by that age, acquisition of the L1 is typically assumed to be complete. Since level of schooling correlated with accuracy on all tasks when all participants were pooled together, it is possible that what we have here is an effect . Since not all structures were tested in the elicited production and in the grammaticality judgment task, it is very hard to make a case for attrition affecting grammatical competence in this study. Most of the errors observed were in production. These errors could reflect problems with fast retrievability and automaticity under communicative pressure, rather than lack of relevant linguistic knowledge at the level of representations. This is the same argument as the proponents of Missing Surface Inflection Hypothesis (Prévost & White 1999, 2000) have made to describe and explain morphological variability in adult L2 acquisition. See Chapter 2. . The number of test items per grammatical structures was not balanced in the GJT. For example, under subordination there were only 4 sentences, 2 of which tested temporal/causal clauses and 2 tested dat subordination. Similarly, V2 tested 4 sentences, 2 subject initial and 2 with wh-interrogatives. And for passives, there were also 4 sentences, 2 testing transitive verbs and 2 testing intransitive verbs. Results are presented by each subtype within each type, but 2 items is hardly enough to run statistics or reach any conclusion about attrition.



Chapter 3.  First language attrition in adults

for literacy and not for incomplete acquisition in the teenagers. Several errors in child language, such as irregular morphology or overgeneralization of transitive patterns to intransitive verbs, vanish once children go to school (Pinker 1999). Coincidentally, the effect for education was most visible in the cloze test and in the GJT, the tasks requiring more metalinguistic awareness than the Wug test or the film retelling task. The subjects with elementary and high school level of education performed much lower than those subjects with post-secondary education, and these analyses also included the control group of adult Dutch speakers. In conclusion, support for the Regression Hypothesis is hard to come by because attrition in adults is relatively minor. The few errors observed are consistent with the Regression Hypothesis, but they are equally consistent with a variety of other explanations: online assembly due to declarative memory slow-down, L2 influence on the L1, and literacy.

2.2

The generative approach

The phenomenon of attrition raises important questions about the nature of native speaker grammars acquired during the critical period. To date, the theory of Universal Grammar has played a central role among other competing theories in explaining the nature of the mature native speaker’s linguistic competence – the realm of theoretical linguistics – as well as how first, second, and bilingual first language acquisition unfold. More recently, this theoretical framework has been extended to L1 attrition (Sharwood Smith & van Buren 1991; Sorace 2000a; ­Tsimpli 2007). Its emphasis on linguistic universals, together with its central goal of explaining language acquisition, renders this linguistic approach compatible with the basic tenets of the Regression Hypothesis. At the same time, the theory of Universal Grammar can potentially account for other features of the attrition process and outcomes not adequately captured by the Regression Hypothesis, such as the fact that attrition is selective both among and within grammatical domains, and the fact that many attrition effects are also due to crosslinguistic influence from the L2 to the L1. A potentially problematic assumption of the generative framework, however, is that once acquired, linguistic competence is relatively stable, as frequently illustrated by Chomsky’s example of the idealized speaker-hearer:

. The generative approach has never been concerned with explaining how external variables like age, sex, level of education, language use affect the internal grammar. Therefore, criticisms based on the fact that these extralinguistic factors are not considered in this framework are misguided.

75

76

Incomplete Acquisition in Bilingualism

Linguistic theory is primarily concerned with an ideal speaker-listener, in a completely homogeneous speech community, who knows its language perfectly and is unaffected by such grammatically irrelevant conditions as memory limitations, distractions, shifts of attention and interest, and errors (random or characteristic) in applying his knowledge of the language in actual performance.  (Chomsky 1965, pp. 3–4).

If this is the case, there should be no cases of true attrition, at least at the level of linguistic representation or competence. If several formal features underlying grammatical categories are selected, and parameter values are set in early childhood, can these same parameters, for example, be unset or reset to the L2 value altogether as a result of L1 attrition? To date, there are no studies documenting complete reversal of parameter values in L1 attrition by adults, or loss of linguistic competence in a particular grammatical area. Here again, what has been observed are apparent minor non-native (L2) effects in native grammars. An interesting attempt to explain non-native effects in native speaker grammars and to predict how and why they select particular areas of linguistic knowledge has been made by Sorace (2000a), Tsimpli, Sorace, Heycock and Filiaci (2004), and Tsimpli (2007). In a series of studies on near-native competence in the L2, Sorace (1999, 2000b) has been documenting the phenomenon of what she calls “residual optionality” or persistent influence of the L1 on the L2, which leads to incomplete/divergent end-state L2 grammars. Although the focus has been mainly on the L2 of late bilinguals, Sorace (2000a) brings to our attention that similar effects are manifested in the L1 of near-native speakers of an L2. Sorace observed, for example, that Italian native speakers who are near-native speakers of English, and who have been immersed in an English-speaking environment for a long time, produce overt subjects as in (2b) in cases where an Italian native speaker would produce a null subject, as shown in (2c) (from Sorace 2000b: p. 718, ex (1)). (2) a. b. c.

Perchè Maria è uscita? why Maria left ‘Why did Maria leave?’ Lei ha deciso di fare una passeggiata. she has decided to do a walk ‘She decided to go for a walk.’ Ha deciso di fare una passeggiata. has decided to do a walk ‘She decided to go for a walk.’

In the Minimalist Program (Chomsky 1995), a distinction is made between the computational system, which includes operations like, Merge, Move and Agree,



Chapter 3.  First language attrition in adults

Table 3.2  Feature specification in Italian/Spanish and English subject expression. Language

Features

Italian/Spanish

uninterpretable (strong) interpretable uninterpretable (weak)

English

Subjects agreement topic/focus agreement

overt/null overt

and the lexicon. The lexicon includes lexical items and features. Features are of two kinds: interpretable or uninterpretable. Articulatory-perceptual interpretable features are visible at the level of Phonetic Form, whereas conceptual-intentional interpretable features have semantic content and are visible at the level of Logical Form. Phonetic Form and Logical Form are external interfaces. Internal interfaces link all the submodules of the grammar, and include the syntax-morphology, syntax-semantics, morphology-semantics, morphology-phonology, phonologysyntax interfaces (White, in press). Uninterpretable features play an important role in grammatical derivations and are deleted in the syntax, before reaching the interfaces. In short, uninterpretable features are formal syntactic features in the syntax; while interpretable features operate at the external interfaces. Adopting Cardinaletti and Starke (1994) and Grimshaw and Sameck-Lodovici’s (1998) syntactic analysis of pronouns, Sorace claims that overt subject pronouns in Italian (and in other null subject languages) carry the interpretable feature [+ topic shift], because the distribution of overt and null pronouns in prodrop languages is governed by pragmatic considerations. That is, in (2) we see that since Maria is a topic, the answer requires a null pronoun because there is no shift of reference. What happens both in attrited L1 grammars and in incomplete L2 endstate grammars, according to Sorace, is that overt pronouns lose the interpretable feature [+ topic shift], or the feature becomes unspecified. Table 3.2 illustrates the feature composition of null subject and non-null subject languages with respect to agreement and subject expression. The shaded cell is the extra layer of complexity that null subject languages present, and which is somewhat eroded under attrition. Thus, the prediction is that overt subjects are found more often in situations where null subjects should occur. If overt subjects in null subject languages are structurally more complex than in non-null subject languages, and if attrition

. The opposite situation – use of null subjects in places where overt subjects should occur – is unattested and should not happen on theoretical grounds according to Sorace. But see Montrul (2004a), Montrul & Rodríguez Louro (2006) and Lafond, Hayes & Bhatt (2001) for findings that contradict this strong claim.

77

78

Incomplete Acquisition in Bilingualism

implies loss of restrictions (or marked values), Sorace’s observations and account are compatible with this phenomenon. Tsimpli, Filiaci, Heycock and Sorace (2004) conducted an experiment on the overt/null subject pronoun distribution with Greek and Italian near-native speakers of English assumed to be undergoing attrition in their respective L1s. Specifically, they wanted to test whether speakers undergoing L1 attrition would produce more SVO than other allowable word orders in focused contexts in Italian and Greek, and whether they would produce more overt than null subjects in Topic contexts. Unlike the majority of attrition studies that focus on individuals who disuse or underuse their L1, participants in this experimental study were very proficient (near-native) adults living in England for more than 6 years, but who continued using Italian and Greek in their daily lives. The two experimental groups were 20 Italians and 19 Greeks living in the UK. The two control groups consisted of 20 Italians and 20 Greeks living in Italy and Greece, respectively, and with limited knowledge of English. A comprehension task and a production task were used in this study. The production task presented words in random order (noun, verb and adverbial expression), and participants were asked to make sentences as if the sentences were newspaper headlines. Participants were prompted to produce a sentence starting with the phrase Did you hear that . . .? In half of the cases, the subject was indefinite (e.g., … a building collapsed because of an explosion); in the other half, definite (… the Wimbledon tournament started in bad weather). The task manipulated preverbal and postverbal subjects in focused contexts, and the prediction was that the Greek- and Italian-speaking control groups would produce more postverbal subjects than the attrited groups. The comprehension task tested pronoun reference and was a picture verification task in which participants were presented with three pictures and a sentence containing two clauses. The pictures depicted two characters and at most three, so that a third referent could potentially be an antecedent for the pronoun in discourse. In some cases, the main clause introduced a referent, and the preposed subordinate clause used a null or overt pronoun, as in (3a). In other cases, the main clause introduced a referent and the postposed subordinate clause contained a null/overt pronoun, as in (3b). These were cases of forward and backward anaphora. Participants were to judge which of the pictures better matched the sentence. Although there are three possible antecedents (and pictures) for (3a), the preferred option is the pronoun coreferential with l’anziana (the subject). For (3b), the preferred antecedent is la ragazza.



Chapter 3.  First language attrition in adults

(3) a. b.

Quando leik/l/proi attraversa la strada, l’anziana signorai saluta while she crosses the street the old woman greets la ragazzak. the girl ‘While she/pro crosses the street, the old woman greets the girl’. L’anziana signorai saluta la ragazzak quando leik/l/proi attraversa the old woman greets the girl when she crosses la strada. the street ‘The old woman greets the girl when she/pro crosses the street.’

The results of the production task in Greek revealed a significant difference between the control group and the attrited speakers. The attrited Greek speakers produced more preverbal subjects than the Greek controls, and were also statistically different from the controls in their interpretation of preverbal overt indefinite subjects as “new” information in the picture verification task. The results of similar sentences with postverbal subjects also showed increased indeterminacy on the part of the attrited group as compared with the control group in the interpretation of postverbal subjects. As for the Italian part of the study, Tsimpli et al. (2004) only discussed the picture verification task. In sentences of forward anaphora with a null subject in the subordinate clause, such as (3a), the preferred choice of referent was the subject for both groups (85% vs. 87%), as shown in Table 3.3. However, the Italian control group differed statistically from the Italian attrited group when the subject of the subordinate clause was an overt pronoun: here, the Italian controls strongly preferred a “new” referent for the interpretation of the overt subject pronoun (63.6%), while the Italian attriters accepted a variety of referents, choosing the new referent for the overt subject only 39.9%. The results of sentences with backward anaphora further demonstrated differential treatment of null and overt subjects by the control and experimental

79

80 Incomplete Acquisition in Bilingualism

Table 3.3  Mean percentage responses by Italian attrited and non-attrited control subjects on overt and null subject preferences in a picture verification task (adapted from Tsimpli et al. 2004). Groups

N

Forward anaphora null subject overt subject

Backward anaphora null subject overt subject

control attrition

20 20

85 87

50.75 69.8

63.6 39.9

  7.6 21.15

groups, since the control group only chose the subject referent (l’anziana in 3b) for the overt pronoun 7.6% of the time, while the attrited speakers chose it 21.15% of the time. This result is consistent with Tsimpli et al.’s prediction that the interpretation of overt pronouns would show attrition effects, since the attrited group showed a significantly greater tendency to allow an overt pronoun to be interpreted as a continued topic. Tsimpli et al. concluded that there is no attrition at the level of syntax, in the sense that the morphosyntactic properties of the Null Subject Parameter are in place. Attrition affects only the distribution of subjects regulated by parameterizable interpretable features operating at the pragmaticssyntax interface, but not the syntax proper. One question that Sorace’s proposal raises is whether the loss of interpretable features is caused by transfer from the L2, or is simply a consequence of the complexity associated with syntax-pragmatics interface, independent of language transfer. Unfortunately, these two potential explanations are difficult to tease apart in the domain of subject expression in Italian, Greek and English, because lacking the pragmatic layer, overt subjects are less complex in English, the dominant language. Furthermore, as we will see in the next section, there is convincing evidence from other studies that the L2 plays a crucial role in shaping L1 grammars under attrition. Rather than actual loss, L1 attrition effects also result from the restructuring and incorporation of L2 elements and rules into the L1 grammar, as reflected in a bilingual speaker’s production, comprehension and/ or acceptance of syntactically deviant L1 sentences under the influence of L2 rules and constraints. L2 effects in L1 attrition have been reported in other linguistic domains, including syntax and morphology (Keijzer 2007), lexical-semantics (Balcom 2003), and phonetics-phonology (Major 1992). I further expand on the

. Major (1992) examined English VOT values in 5 Americans living in Brazil, and found that the values were closer to the Portuguese values in casual speech but not in formal style. This suggests that the English speakers retained knowledge of the articulatory and acoustic properties of English voiceless stops /p, t, k/, the realm of phonetics. Although no perception tasks were used in this study, it is reasonable to assume that the 5 bilingual subjects retained



Chapter 3.  First language attrition in adults

issue of L2 influence on the L1 in §3 when I discuss Gürel’s studies, but before I do so I will close this section by discussing M. Paradis’s Activation Threshold Hypothesis because this hypothesis is also addressed in Gürel’s work.

2.3

The activation threshold hypothesis

If the Regression Hypothesis and, to some extent, the generative approach, predict that order of acquisition and markedness play a deterministic role in attrition, other researchers consider that it is not only what was acquired first, but what was acquired best – through frequency and reinforcement – that is likely to be retained under attrition (de Bot 2002). It may be the case that the first language is not lost, but becomes less accessible, and processing is more demanding. Crucial to this view are the psycholinguistic mechanisms behind memory and forgetting due to crosslinguistic influence in bilinguals and to the cognitive processes of activation and inhibition (Green 1986). Although not grounded in any particular linguistic theory, M. Paradis’s (2004, 2007) Activation Threshold Hypothesis is an example of a cognitive and neurologically-based approach to attrition, which emphasizes the role of inhibition and frequency in bilingual language use. Based on an analogy with neuron action potential, the Activation Threshold Hypothesis states that a linguistic item is activated when a sufficient amount of positive neural impulses – its activation threshold – has reached its neural substrate. If the item is inactive or unselected due to disuse, its threshold of activation raises. The more an item is used, selected, or activated, the lower its activation threshold. “Attrition is the result of long-term lack of stimulation. Intensive use/exposure to one of the languages in a bilingual environment leads to a lower activation threshold for that language (i.e., it requires fewer resources), even in early, fluent, behaviorally balanced bilinguals” (M. Paradis 2004, p. 28). An advantage of the Activation Threshold Hypothesis over the Regression Hypothesis and the generative approach, is that Paradis’s hypothesis also predicts that production (or recall) of a linguistic item will be more difficult than comprehension of the same item, because production requires a higher level of activation than comprehension, due to the neurological mechanisms involved.

categorical knowledge of VOT boundaries for English stops. Major’s results are compatible with the findings of many studies conducted by Flege (Flege 2002; Flege et al. 2006), which show that in bilinguals, the phonetic/phonological categories of the L1 and the L2 influence each other. However, there is no evidence to date that a categorical phonological contrast (b vs. p) can be lost at the level of abstract representation.

81

82

Incomplete Acquisition in Bilingualism

Another prediction of the Activation Threshold Hypothesis is that the less the L1 is used, the more attrition there should be because competition from the L2 would be much higher. However, just as parameter values cannot be completely unset or reset once acquired, it is also not the case under this model that one linguistic system of the bilingual is completely switched off due to inhibition or a high activation threshold (Green 1986). In other words, while the activation threshold of the L1 of bilinguals living in an L2 environment is generally high, this does not lead to generalized L1 loss. There is recent empirical evidence showing that degree and frequency of L1 use does not correlate with degree of lexical attrition or fluency (Schmid 2007, but cf. Gürel 2007). As for the selectivity of the attrition phenomenon, M. Paradis claims that this depends on frequency of use: different linguistic items within the same language and grammatical areas might require different degrees of stimulation in order to become activated. However, specific discussion of just how much activation is needed for different grammatical elements, or why, is not provided. M. Paradis explains that susceptibility of L1 grammatical properties to attrition is determined by their activation threshold level relative to the corresponding L2 properties with which they compete. This predicts that structural selectivity will be closely linked to frequency of use. Strong support for this model has been hard to find in the attrition of morphosyntax (Köpke 2002), although the few errors with irregular plural morphology in German, which are suggestive of a memory problem, are potentially compatible with this hypothesis. On the other hand, retrieval difficulties are often reported in the domain of the lexicon, and lexical knowledge does depend on frequency effects. As a matter of fact, one of the first signs of language attrition manifests itself as the inability to remember words; as a consequence, speakers borrow and adapt words from the L2 into the L1. Word finding difficulties and lexical errors in production in the L1 have been documented by Ammerlaan (1996), Köpke (2002), Olshtain and Barzilay (1991), Waas (1996) and Yağmur (1997), among others. One of the earliest attrition studies documenting lexical retrieval difficulties in adults immersed in an L2 environment is Olshtain and Barzilay (1991). The purpose of this study was to describe features of language attrition at the lexical level, by asking subjects to do a task that required a great deal of conscious effort to remember meanings of words. The participants were 15 Americans who had been living in Israel for an extended period of time (range 8–25 years), whose age at time of testing ranged between 23–55, and 6 Americans living in the United States and who had never been to Israel (age range 28–56). Participants were asked to narrate two stories from Mayer (1969) (Frog, where are you, and A boy, a dog and a frog). All participants (the control group and the



Chapter 3.  First language attrition in adults

“attrited” group) had no difficulty retelling the stories with fluency. The authors focused on the production of five words in the stories that seem problematic for the “attrited” group but not for the control group, namely the words jar, cliff, pond, gopher and deer. All attrited subjects could not recall these words and used three strategies: (a) use of circumlocution or paraphrase (e.g., body of water for pond), (b) replacement of the target word by a word of similar meaning (e.g., puddle, water or swamp for pond), or (c) activation of a conscious retrieval process for the right word (think aloud). They would often say “I forgot the word” or “I forgot my English” (Olshtain & Barzilay 1991, p. 145). By contrast, the American control group had no difficulty producing these words naturally and fluently during the course of the stories. Hesitations with these specific words were minor to nonexistent. Olshtain and Barzilay concluded that language loss may affect specific meanings of words. As a result, attrited speakers resort to more neutral words to express related meanings, such as body of water for pond. Hence, in the realm of lexical semantics, complex concepts are replaced by simpler concepts. This is consistent with the loss of restrictions or the resort to unmarked options observed in other grammatical areas. Perhaps the best way to really test lexical retrieval and access difficulties is by undertaking psycholinguistic studies that test processing on-line. Hulsen (2000), for example, tested lexical knowledge and processing in three generations of Dutch speakers living in New Zealand (30 speakers of each generation). Although Hulsen did not test the Activation Threshold Hypothesis directly, her findings are entirely compatible with the predictions of this hypothesis. (I only focus on the results of the psycholinguistic tasks of the first generation speakers, and will defer discussion of the findings with second and third generation speakers for Chapter 6, when we discuss incomplete acquisition in adult bilinguals.) The first generation speakers immigrated to New Zealand between 1950 and 1965. Their mean age of arrival was 24.4, and their mean age at time of testing was 68. Ten Dutch speakers residing in the Netherlands (mean age 63.1) were the comparison group for the first generation group. A picture-naming task and a picture-matching task were used to investigate Dutch speakers’ knowledge of words in Dutch and English. Selection of the set of stimuli was based on morphological complexity of the word, frequency, cognate status, cultural bias and visual complexity. Participants had to name the picture as quickly as they could. Accuracy and response times were recorded on a computer. In the picture-matching task, participants saw a picture with a word underneath. They were to indicate as quickly as possible whether the word was the correct name for the picture displayed. Half of the words matched the picture presented, for example a picture of a kangaroo with the word kangaroo (yes condition); while the other half did not

83

84

Incomplete Acquisition in Bilingualism

Table 3.4  Percentage accuracy and reaction times (RTs) in the picture naming and picture-word matching tasks in English and Dutch (adapted from Hulsen 2000). Language

Group

Picture Naming Task

Picture-Word Matching Task

Dutch

control Gen 1 Gen 1

Accuracy 95% 89% 92%

Accuracy 99% 98% 98%

English

RT 1163 ms 1205 ms 1168 ms

RT 1126 ms 1150 ms 1228 ms

match, showing the picture of the kangaroo with the word toe (no condition). The stimuli included low, high frequency words and cognate, non-cognate words, plus fillers and practice items. Hulsen predicted that self-ratings for proficiency in Dutch would be lower than or equal to those of English in the first generation speakers. For the processing experiments, Hulsen expected the speakers to perform better in Dutch than in English. High frequency and cognate words would be remembered best and recognized fastest, and shorter reaction times were expected in the matching task than in the production task. First generation speakers would not differ from the Dutch control group in terms of accuracy, but they would be slower than the Dutch control in their reaction times. Results showed that 53.3% of the first generation speakers felt that their Dutch language proficiency had changed after so many years of residing abroad. On a scale of 1–5, the mean rating of Dutch and English for this generation was 4.58 and 4.69, respectively. Dutch was rated lower than English, and this difference was statistically significant. These ratings clearly show that both their L1 (Dutch) and L2 (English) are self-perceived as near-native rather than native, and that Dutch may already be undergoing some kind of loss. Table 3.4 summarizes the overall accuracy and reaction times means for the three groups. Overall, and consistent with the Activation Threshold Hypothesis, high frequency words and cognate words were recognized best and fastest in both the Dutch and English language experiments. Performance on the picture matching task was better than on the picture naming task, suggesting that production is more affected by attrition than comprehension. This trend was also confirmed in the English tasks, where accuracy was better on the picture matching than on the picture naming task. The comparison with the Dutch control group showed that in the Dutch language experiments, the first generation group scored significantly lower than the Dutch control group, but the words they remembered were retrieved equally fast. The two groups performed better, and reacted faster, to



Chapter 3.  First language attrition in adults

high frequency words in the two tasks. The cognate status of words only affected accuracy in the picture-word matching task for the first generation speakers. While the results of the study are compatible with the frequency effects predicted by the Activation Threshold Hypothesis, the advanced age of many of the participants in the first generation group raises questions about the nature of the slower speed in lexical processability. According to Goral (2004), slowed lexical retrieval due to cognitive decline is very common in elderly individuals past the age of 60 or 70. Thus, while L1 attrition due to competition with the other language is very likely in bilinguals, slower retrievability is also compatible with the normal process of aging. Individual analysis of subjects by age might be necessary to tease apart these two possibilities. In the next section, we will discuss a study testing both the Activation Threshold Hypothesis and the generative approach in aspects of syntax.

3.

Cross linguistic influence (transfer) in L1 attrition and L2 acquisition

In the previous chapter, we discussed how the L1 grammar can have a persistent, and even deterministic, influence, not only in early but also in advanced and near-native L2 grammars. And all the studies showing mild attrition effects in morphology, syntax and syntax-discourse discussed in this chapter also point to the conclusion that the L2 influences the L1 to a large extent under attrition. If patterns of cross linguistic influence in L1 attrition and L2 acquisition are theoretically linked, what is the relation between cross linguistic effects from the other language in L1 attrition and L2 acquisition? One study comparing L2 acquisition and L1 attrition of the same language is Gürel (2002). Gürel investigated the impact of English – the dominant language as L1 and as L2 – as a possible factor in patterns of language loss and restructuring in the less dominant language – Turkish – as L1 and L2. In particular, Gürel examined whether L1 attrition and L2 acquisition of Turkish were affected by English, the other language of the bilinguals tested. The specific grammatical focus was the acquisition and attrition of pronoun interpretations in Turkish, which falls within the domain of binding principles of Universal Grammar. Given that certain binding properties are subject to language specific constraints, binding properties make an excellent testing candidate for investigating crosslinguistic influence. Turkish is a pro-drop or null subject language, whereas English is not. In English, pronouns can take referential (Mary, Peter, the man) or quantified antecedents (somebody, nobody, everybody), as in (4)–(6):

85

86 Incomplete Acquisition in Bilingualism

(4) Briani believes that [hei/j is intelligent]. (5) Briani ate [hisi/j cake]. (6) Nobody believes [that hei/j is intelligent].

The pronouns he and his above must be free in their governing category, either a nominal clause or a possessive noun phrase (marked with square brackets in examples (7) and (8)). What this means is that pronouns can be coreferential with the subject of the main clause or with some other person in discourse. Turkish has two overt pronouns – o “s/he” and kendisi “self ” – as well as null pronouns. Null pronouns and kendise can be correferential with the matrix subject or another subject in the discourse, as shown in (7). (7) a. Murati [proi/j zeki olduğ-u]-nu düşün-üyor Murat pro intelligent be-nom-3sgposs-acc think-prog ‘Murat thinks that he is intelligent.’ b. Murati [kendi-si-nini/j zeki olduğ-u]-nu düşün-üyor Murat pro intelligent be-nom-3sgposs-acc think-prog ‘Murat thinks that self is intelligent.’

Unlike the English pronouns he and she, the Turkish pronoun o must be disjoint from the matrix subject, as shown in (8). (8) Murati [o-nun*i/j zeki olduğ-u]-nu düşün-üyor Murat pro intelligent be-nom-3sgposs-acc think-prog ‘Murat thinks that he is intelligent.’

The binding properties of these pronouns in English and Turkish are regulated by Binding Principle B, which states that pronouns must be free in their governing category. The difference between the binding possibilities of Turkish o and English he/she is related to the size of the governing categories. Nominal embedded clauses in English function as binding domains. By contrast, nominal embedded clauses in Turkish do not constitute binding domains, and the obligatory disjoint reference for the overt pronoun o is expected under Binding Principle B. Thus, in this sense, Turkish constitutes a subset of English, because the Turkish governing category is smaller than the English one. Given this particular subset relation between English and Turkish, and assuming the generative framework, Gürel predicted that transfer would be persistent at the endstate in L1 attrition and L2 acquisition, albeit selectively. That is, transfer would be predictable from the specific subset relation holding between the two languages of the bilinguals. Gürel hypothesized that, in L2 acquisition of Turkish, English-speaking learners would incorrectly allow binding of the overt



Chapter 3.  First language attrition in adults

pronoun o within the same sentence, out of the embedded clause or out of a possessive noun phrase. By contrast, L2 learners would be accurate with null subjects and the pronoun kendise. The same predictions were formulated for the potential L1 attrition of Turkish under the influence from English. Two related studies, employing the exact same methodology, were conducted. Before discussing further details of these two studies below, it is important to understand that Gürel (2004) recast the predictions of the attrition part of the study within M. Paradis’s Activation Threshold Hypothesis. The selectivity of attrition within the Activation Threshold Hypothesis, according to Gürel, can only be expected with L1 properties that have analogous forms in the L2 (hence in competition with the L2 forms). These grammatical properties will undergo attrition due to inhibition (or a high activation threshold). L1 elements that have no corresponding forms in the L2 will not be vulnerable to attrition, as they will not be in competition with the L2 elements. Accordingly, the attriter’s L1 grammar under the influence of English is expected to allow noun phrases (either embedded clauses or simple possessive DPs) to function as governing domains for overt pronouns, since the L2 grammatical option has a lower activation threshold. In situations where a particular L1 grammar rule/item does not have any equivalent form in the L2, no competition between the L1 and the L2 is expected. Therefore, the overt pronominal kendisi and the null pronoun, together with their relevant binding features, are predicted to be well-preserved in the grammars of Turkish attriters because they do not have equivalent and competing forms in the L2. The first study discussed in Gürel (2002) is on L2 acquisition, in which 28 native speakers of English living in Istanbul, Turkey were tested. Their mean age was 46 years, and they had been living in Turkey for at least 10 years. Age of first exposure to Turkish coincided with their age of arrival in Turkey as adults. All participants were assumed to be at end-state in learning Turkish, although they were classified as intermediate and high intermediate on the basis of a proficiency test. Thirty native speakers of Turkish were tested in Turkey as the control group. The L1 attrition study tested native speakers of Turkish (mean age 47) who had immigrated to Canada and the United States as adults (mean age of immigration 25.5) and had been living in the English-speaking environment for at least 10 years. With the exception of two participants, all other participants had learned English after puberty in Turkish schools. All speakers worked in English-speaking environments and used English in their everyday lives. Turkish was only spoken with some friends and relatives. The native speakers tested in Turkey also acted as a control group for the attrition study.

87

88

Incomplete Acquisition in Bilingualism

Table 3.5  Mean percentage acceptability on overt and null subjects in embedded clauses (adapted from Gürel 2002). Groups

N

Interpretation Overt embedded subjects

Null embedded subjects

native speakers

30

bound disjoint

 o  4 96

kendisi 79 21

pro 76 24

L1 attriters

28

L2 learners

24

bound disjoint bound disjoint

38 62 30 70

73 27 74 26

79 21 82 18

Three tasks were used to test binding interpretations of the pronouns in question: a written interpretation task, a truth value judgment task with stories, and an oral picture identification task. Table 3.5 summarizes overall results from the truth-value judgment task on stories with referential antecedents. The results of all these tasks confirmed the predictions that 1) L2 learners and L1 attriters had difficulty with the acquisition and preservation of the overt pronoun o, yet 2) they were able to successfully acquire and retain the binding properties of null pronouns and the pronoun kendisi at native-speaker level. In the two cases (L2 acquisition and L1 attrition), the experimental participants treated the Turkish pronoun o as identical to the English overt pronouns he and she. While the L2 acquirers extended the English governing category to the Turkish binding situation, the L1 attriters expanded the binding domain of Turkish to include the English values as well. Thus, in both the L2 acquisition and L1 attrition situations, English was the influencing language. As such, English binding properties were imposed on the Turkish pronoun o in these bilingual grammars. If the predictions of the generative approach and those of the Activation Threshold Hypothesis were upheld in this study, they were not entirely upheld in a recent follow up study testing potential attrition in English as L1 under contact with Turkish as L2 in the same binding domains (Gürel 2007). This time around, Gürel found no effects of attrition in the syntax of the L1-English speakers. She suggests that no L1 loss was found because these speakers continue to use English in their daily lives. Since the Activation Threshold Hypothesis predicts that decreased L1 use will result in more attrition, this is how Gürel explains the results. But in light of Schmid’s (2007) findings of no correlation between L1 use and degree of attrition in German immigrants in Canada, it is difficult to assess at this point where amount of L1 use stands in adult L1 attrition. In conclusion, Gürel’s most recent study suggests that there is no attrition at the level of syntax, and if there is, the mild attrition effects can be explained by L2 influence on the



Chapter 3.  First language attrition in adults

L1 (Gürel 2002, 2004). This is similar to what Keijzer found, since the few errors attested in the Dutch immigrants could also be accounted for by the influence of English on Dutch. In short, once the core L1 grammatical system is acquired and becomes stable – syntax especially – it is impermeable to shifts in input and use, as the Chomskyan view of linguistic competence implies. This is also consistent with the Critical Period Hypothesis.

4.

The age factor in L1 attrition

This chapter has led the foundations for the discussion of attrition and incomplete acquisition in children that follows in the rest of the book. I have examined different aspects of linguistic knowledge and use in adults: pronunciation, lexicon, morphosyntax, syntax and other interfaces. The general picture that emerges is that the effects of L1 attrition in adult L2 learners, if any, are relatively small and hard to interpret as true loss of linguistic competence, at the level of representation. In phonetics/phonology, L1 VOT values can move toward L2 VOT values as a function of proficiency in the L2 (Major 1992). At the level of the lexicon, attrition usually manifests itself in problems with lexical access and retrieval (including inflectional morphology), specific meanings of words, and lexical production. By contrast, the effects of attrition in morphosyntactic rules affecting mostly irregular forms are almost negligible. In syntax, the salient pattern that emerges is one of crosslinguistic evidence and convergence with the influencing language. But parameters are not certainly unset or wiped out altogether, even when the distribution and frequency of use of certain structures changes slightly under contact. Some of the patterns of attrition attested are comparable to similar effects in L2 acquisition. Throughout this chapter, I have been invoking inverse parallels between L1-induced fossilization in the L2 on the one hand, and L2-induced attrition in the L1, on the other – two possible outcomes of fluent bilingualism. However, it is too early to tell, in my view, whether fossilization and attrition are two sides of the same phenomenon. At this point, I suspect they are not, but this issue certainly deserves more empirical investigation. For example, within the generative paradigm, a principled distinction is made between interpretable and uninterpretable (formal) features (Chomsky 1995, 2000, 2001, 2004), as discussed in §2.2. Interpretable or semantic features (including gender, person and number agreement on nouns, and tense, aspect and mood in verbs), contribute to meaning and interpretation. Uninterpretable features are formal features that do not contribute to interpretation and are assigned in the lexicon (case, wh, ­finiteness, agreement on verbs). Smith and Tsimpli (1995), Hawkins and Chan (1997) and

89

90 Incomplete Acquisition in Bilingualism

Hawkins and Hattori’s (2006) syntactic account of L2 acquisition and fossilization briefly presented in Chapter 2 restrict the critical-period based deficit to uninterpretable (formal) features that have never been selected in the learners’ L1. By contrast, Sorace’s (2000a) account of L1 attrition states that it is only interpretable (semantic) features, which are legible at the interfaces between grammatical domains that are affected in near-native speakers’ L1s. Furthermore, if uninterpretable features are selected in childhood as part of the functional lexicon (Smith & Tsimpli 1995), while interpretable features acquire their values during the course of the derivation and are legible at the interfaces, then only uninterpretable (formal) features may be affected by a critical period while interpretable (semantic) feature may be spared. If uninterpretable features are integral part of native grammars acquired in childhood, then L1 attrition can never affect linguistic competence in adults.10 Rather than invoking observed changes in the L1 as “deficiencies” analogous to arrested development due to fossilization, minor quantitative departures from monolingual values may be construed as an artifact of the nature of bilingualism, where each language affects the other (Cook 2003). If these hypotheses are confirmed by further empirical evidence, they will provide a principled and theoretically based explanation for both fossilization and attrition effects in bilingual grammars. Humans acquire knowledge of language at a very early age, and what develops is their capacity to implement that knowledge during production, comprehension and processing. The results of the studies discussed at length in this chapter suggest that once acquired completely, linguistic knowledge cannot be lost in adults. Rather, attrition in adults may decrease or simply slow down the resources necessary for the implementation of the available knowledge. That is why attrition affects the level of performance and use, rather than competence. In children, as we will see in the next chapters, attrition appears to reach the level of competence, especially because the L1 grammatical system did not develop fully. As in L2 acquisition, there is a clear indication that age is a critical determinant of the depth and degree of attrition following a change in linguistic environment and habits. Based on current theories of acquisition, the rate and extent of decay of language is assumed to depend, among other things, on the age of acquisition of the various language systems, and the duration of language maintenance of a certain level of proficiency (also known as entrenchment). After puberty, or even earlier, loss of the L1 under attrition is highly unlikely. This indicates that a biologically or cognitively-determined maturational period unquestionably links both acquisition and attrition. 10. This raises the question of whether only uninterpretable features are part of grammatical (syntactic) competence and interpretable features are not.



Chapter 3.  First language attrition in adults

De Bot and Weltens (1991) asked whether the process of acquisition and the context in which a language was acquired determined the storage or representation of that language. Storage and representation of the first language might, in turn, influence the process of its decline. Assuming the distinction between implicit and explicit language learning and the procedural/declarative model (M. Paradis 2004; Ullman 2001), the attrition data discussed in this chapter suggest that whatever has been acquired through implicit learning and stored in procedural memory remains intact and is not affected by attrition in adulthood. If attrition effects are observable in adults, these appear to target aspects of language handled by explicit learning stored in declarative memory, aspects that require conscious control and effortful retrieval of lexical items (including irregular morphology). In conclusion, age is related to degree and depth of attrition: the earlier the onset of bilingualism the more severe the loss is likely to be, as hypothesis 1 stated. If it is true that implicit knowledge acquired early in childhood is less vulnerable to attrition than explicit learned knowledge acquired with the development of other cognitive skills, then the question is when in childhood such early acquired linguistic knowledge becomes resistant to attrition. We now turn to an investigation of these issues in children.

91

chapter 4

Bilingualism in early childhood

As foundation for the issues to be discussed in subsequent chapters, Chapters 2 and 3 traced theoretical links between L2 acquisition and L1 attrition in late bilingualism, after the critical period. The common theme between Chapters 2 and 3 was the focus on adults, and the fact that just as it may be difficult to attain high levels of proficiency in an L2 in adulthood it is also highly implausible to lose significant aspects of one’s L1 after extensive contact with the L2. My main claim in the previous chapter was that despite minor processing, performance, and what could be considered L2-induced effects in mature L1 systems in contact with an L2, L1 attrition does not substantially affect linguistic competence in adults, regardless of amount of L1 exposure and use. But as I will show in the rest of the book, when language exposure and use is reduced in childhood, the grammatical system of bilingual children in either language, and in some cases of their family language, can be dramatically compromised, especially at the level of morphosyntax. This will become evident in Chapters 6 and 7, when I discuss the long-term effects of incomplete acquisition in adult early bilinguals of minority languages, called heritage speakers in the United States and Canadian contexts, background speakers in Australia, or ethnic community speakers in the European context. Because it is difficult to reconstruct the linguistic past of adult early bilinguals in the absence of longitudinal data, to understand the validity of my claim, I must now show how language acquisition and loss interact in early bilingualism as a function of age and input, when exposure to the two languages happens during the critical period. Although full linguistic attainment is possible in the two languages when bilingualism starts early, in this book I am interested in cases of unbalanced development and the fate of the weaker language, particularly when the weaker language is the family language not fully supported in the community. Because many minority language-speaking children typically fail to develop age-appropriate levels of vocabulary and grammar in the family language, key questions arise as to the linguistic nature of these delayed and underdeveloped grammatical systems. For example, what are the structural characteristics of the weaker language? Does the weaker language have the signatures of a first language acquired in childhood, or does it develop like a second language acquired in adulthood, under the influence of different learning mechanisms? In this chapter, I discuss bilingualism in

94 Incomplete Acquisition in Bilingualism

early childhood, while the focus of the next chapter is on L1 maintenance, shift, and loss during the school years. Since the effects of delayed acquisition in childhood may persist in adulthood, the discussion and empirical evidence presented in this chapter are crucial to understanding the root of incomplete acquisition in adult heritage speakers, the focus of Chapters 6 and 7.

1.

Types of bilinguals

In order to discuss unbalanced development in bilingual children, it is important to reiterate that bilingual children differ in many ways. Recall from Chapter 1 (Figure 1.2) that a distinction is made between simultaneous or bilingual first language acquisition and sequential bilingualism, based on sequence of acquisition of the two (or more) languages. In addition to sequence and timing of acquisition, bilingual children differ with respect to the sociopolitical context of the languages they are learning. That is, the status of the family language as a minority or majority language in the society has important implications for the linguistic development of bilingual children. Of particular importance in this book is the status and development of the minority language in children born to immigrant families.

1.1

Simultaneous bilingualism

In many parts of the world, bilingual children are typically children of first generation immigrants (discussed in Chapter 3), often called second generation immigrants. In other cases, they are born in bilingual or multilingual communities (Montreal, Catalonia, Switzerland, Papua New Guinea, etc.). Simultaneous bilingual children are exposed to two languages since birth or before the age of 3, the latter being the approximate age when basic syntactic knowledge is assumed to be in place. Sequential bilingual children are children who first acquired one language, then a second language after the rudiments of the first language were established, say after the age of 4 years. Many simultaneous bilingual children are born to parents who speak different languages and are raised under the one-parent/one-language strategy (by which each parent speaks almost exclusively his/ her native language to the child). Sometimes the two parents speak the same language, but the child is under the care of a caregiver who speaks another language. Even in these cases, the child is exposed very early to two languages. A frequent observation in the child acquisition literature is that by roughly 3 to 4 years of age, children succeed in acquiring the basic grammar of their speech community. Even if innate mechanisms for language explain the ­rapidity



Chapter 4.  Bilingualism in early childhood

and ­inevitability of the acquisition process, it stands to reason that language acquisition depends on input. If a child is exposed to one language, for example Japanese, the child learns Japanese, but if the child is exposed to, and uses, two languages, for example Japanese and French, the child then learns and retains both Japanese and French. The generative approach to language acquisition holds that the robustness and flexibility of the human language faculty enables human beings to acquire more than one language at the same time (Crain & Lillo-Martin 1999; Genesee 2000; Meisel 2001, 2004a, 2007). Researchers like Jürgen Meisel and Fred Genesee, among others, consider the simultaneous acquisition of two or more languages by a single individual under the age of 3 years a case of bilingual first language acquisition (BL1). The term bilingual first language acquisition implies that the acquisition of two languages simultaneously in early childhood is similar, if not identical, to the acquisition of only one language by monolingual children. For generative linguists, this means that the two languages are acquired through Universal Grammar, and for neurolinguistis like Michel Paradis and Michael Ullman, the two languages are acquired with implicit learning mechanisms and stored in procedural memory. The collective findings of Fred Genesee and Jürgen Meisel’s research programs have shown that bilingualism does not present a cognitive challenge to children, nor are bilingual children linguistically confused. Contrary to the claims of the Unitary Linguistic System Hypothesis (Volterra & Taeschner 1978), bilingual children have been found to ­differentiate

. Even within academic circles, today’s positive outlook on the bilingual capacity has not always been upheld, especially in the United States where it is often believed that bilingual children are at an academic disadvantage in comparison to monolingual children (Arsenian 1937; Macnamara 1966; Ruiz 1988; Smith 1935). But in 1962, Peal and Lambert conducted a study of proficient French-English bilingual children in Montreal, matched for SES, and found cognitive advantages for the bilingual children pertaining to metalinguistic awareness, or the ability to analyze language and its different subcomponents independent of its meaning. And the most recent research of Ellen Bialystok (Bialystok & Martin 2004; Bialystok & Shapero 2005) shows that bilingual children between the ages of 4 and 8 are more advanced than monolinguals of comparable intelligence in solving problems of attentional control, particularly when the problems include misleading or conflicting perceptual cues. These advantages for bilingual children, as well as their ability to do translation and simultaneous interpretation, are discussed by Bialystok (2001) and Valdés and Angelelli (2003). The apparent cognitive and academic disadvantages of bilingual children in the U.S. context can be related to other factors, including poverty, low level of parental education, inability of parents to speak the second language, and an inability of parents to support their children academically at home (Hakuta, Butler & Witt 2000; Oller, Eilers, Urbano & Cobo-Lewis 1997; Oller & Eilers 2002). Naturally, all of these factors have negative consequences on the self-esteem, acculturation, and academic progress of bilingual children, as they do on any monolingual or bi-dialectal children who live in similar disadvantaged circumstances.

95

96 Incomplete Acquisition in Bilingualism

s­ yntactically and phonologically the two languages from the very beginning, showing autonomous linguistic representations from the outset of syntactic acquisition (­Genesee 1989; Meisel 1989, 1990; J. Paradis & Genesee 1996, 1997). Furthermore, and because they are guided by Universal Grammar, simultaneous bilingual children progress through the same developmental milestones as monolingual children in each language, eventually arriving at the same kind of grammatical knowledge as monolingual children. Even when the linguistic systems develop independently, the languages of the bilingual child can still influence each other in systematic ways. This has been observed in phonology (Kehoe, Lleó & Rakow 2004; J. Paradis 2001), morphology (Nicoladis 1999), and syntax (Gawlitzek-Maiwald & Tracy 1996; Hulk & ­Müller 2000; Müller 1998; Müller & Hulk 2001; J. Paradis & Genesee 1996; Yip & Matthews 2000). Furthermore, like proficient bilingual adults, bilingual children mix the two languages within and between utterances. Initially, code switching may be a strategy bilingual children resort to in order to compensate for gaps in lexical development. But research has shown that, as their linguistic competence in the two languages develops, bilingual children’s mixed utterances progressively . Under the linguistic approach to acquisition, children learning parametrically different languages set the parameters for each language accordingly, early on. Meisel (1994b), for example, analyzed the emergence of the functional category IP (the inflectional phrase comprising agreement and tense) in French-German bilingual children. He observed that in both languages subject agreement emerged before tense and at roughly the same age. The complementizer phrase (needed for embedding, questions, and word order), on the other hand, developed later in the two languages. Similarly, Koehn (1994) and Müller (1994) demonstrated that French-German bilingual children first go through a stage where no nominal inflection (gender agreement and number) is produced, and when inflection appears, it does so in the two languages. Taken together, these studies show that bilingual children acquire the two languages as independent linguistic systems and do not show evidence of persistent delay or full transfer from one language to the other. . Volterra and Taeschner’s (1978) Unitary Linguistic System Hypothesis was partially based on the fact that bilingual children, like adults, display mixed utterances such as these, suggesting at that time that children had one underlying linguistic system and two sets of vocabulary. For many, code-switching is a sign of linguistic incompetence, cognitive confusion, or inability to speak either language competently. Since the seminal work of Poplack (1980) and many others afterwards, the case has in fact been made for the exact opposite interpretation: intrasentential code switching is a sign of sophisticated linguistic competence in the two languages, given that code switches are rule governed, and bilinguals hardly ever violate the grammatical constraints of the two languages. Furthermore, code switching is context-specific, typically occurring in conversations with other bilingual speakers who share the code. Finally, code switching serves a variety of pragmatic purposes and adult bilinguals have been shown to adjust competently to the socio-pragmatic situation.



Chapter 4.  Bilingualism in early childhood

reflect a highly sophisticated and independent command of the grammatical and pragmatic rules of the two languages (Allen, Genesee, Fish & Crago 2002; ­Genesee, Nicoladis & J. Paradis 1995; Genesee, Boivin & ­ Nicoladis 1996; ­ Lanza 1997; Meisel 1994b; J. Paradis, Nicoladis & Genesee 2000; Vihman 1998). In essence, bilingual children’s code switching competence is linked to their linguistic development in the two languages. To summarize, research evidence to date suggests that very young bilingual children are fully capable of acquiring two languages with no adverse cognitive consequence. Simultaneous bilingual children posit two independent linguistic systems from the initial state of language acquisition. Furthermore, they use the same linguistic mechanisms as those used by monolingual children, and they follow the same developmental schedule as monolingual children in each language.

1.2

Sequential bilingualism

In sequential bilingualism – or child L2 acquisition – one language is acquired during the age of early syntactic development, and the second language is acquired after the structural foundations of the first language are in place. If the child of immigrant parents is exclusively exposed to the family language until he or she goes to pre-school/school and the second language is introduced at age 4 or 5, then the child is considered a sequential bilingual. There are also sequential bilingual children (or child L2 learners) who are immigrant children. They were born abroad and they moved to the majority language-speaking country between the ages of 4–5. They learned their first language at home in a monolingual environment, and some may have even attended pre-school before immigration. The second language is learned in the new country. Not all sequential bilingual children are born to immigrant families. Other children are members of the majority language, which they speak at home in a monolingual household, and acquire a second language at school. In parts of Canada, for example, many English-speaking children attend bilingual EnglishFrench schools or French immersion schools in order to learn the other official language of the country. There are also many children all over the world who speak the official language of their country and learn English or any other nonofficial language as a foreign language at school, for example, children in South America who attend private Spanish-English bilingual schools. These are situations of additive bilingualism (Lambert 1977), whereby bilingualism is a sign of prestige because only children from upper and middle classes go to these expensive schools. Spanish, French, Chinese, and other languages are also taught as a foreign language in some North American elementary public schools that offer an

97

98 Incomplete Acquisition in Bilingualism

early foreign language curriculum. For children of immigrant families who speak a minority language at home, bilingualism is neither a choice nor a sign of prestige, but a matter of survival. In addition to L2 acquisition, these children must cope with adaptation and assimilation into the mainstream culture. The difference between L2 acquisition in a foreign, as opposed to a second, language context is relevant to theoretical debates on native-like attainment, degree of bilingualism, and maintenance of the L1. In the foreign language context, degree of L2 ultimate attainment will depend on the number of hours of exposure per week. Typically, native-like attainment is unlikely because children may not be immersed in the L2. In these cases, L2 acquisition does not have detrimental effects on the L1. The situation is very different for minority language-speaking children, who acquire a second language in the L2 environment (community and school). Several theoretical issues arise in the study of child L2 learners in an L2 context. These concern the nature of the L2 acquisition process in childhood, rate of acquisition, and ultimate attainment. If L2 acquisition occurs during the critical period, the acquisition of an L2 in childhood is presumably also guided by Universal Grammar (Bley-Vroman 1990), although how child L2 acquisition differs from L1 acquisition and adult L2 acquisition is a matter of ongoing investigation (Lakshmanan 1995; Schwartz 2003; Unsworth 2005, 2008). Most of what we know about child L2 acquisition comes from studies that have compared adult and child L2 acquisition in terms of ultimate attainment, as was discussed in Chapter 2. If L2 acquisition for children immersed and educated in the L2 environment is likely to end up in native-like ultimate attainment, such potential success in the L2 brings a downside: the loss of linguistic skill in their first language, also known as subtractive bilingualism (Lambert 1977). The distinction between simultaneous and sequential bilingualism in childhood, and the sociopolitical status of the languages, is significant for incomplete acquisition and language attrition in childhood. If degree of L1 loss is related to age, as I argued in the previous chapter, there should be age effects as a function of input and use within the critical period. Hypothesis 2, stated in Chapter 2, is repeated here for convenience: Hypothesis 2 If L1 attrition occurs in early (pre-puberty) bilingualism, it should be more severe in simultaneous bilinguals (exposed to the two languages very early) than in sequential bilinguals (when the L1 was acquired before the L2).



Chapter 4.  Bilingualism in early childhood

Note that age and input factors are difficult to tease apart. Because some simultaneous bilingual children learn the majority language along with the minority language from birth, they have the longest possible period of exposure to the majority language. Accordingly, they receive less input in their family language than sequential bilingual children whose L1 is the family language. Furthermore, when the minority language does not share the prestige of the majority language, this could also lead to imbalanced acquisition, with a trajectory toward eventual lower proficiency in the minority language than in the majority language. Sequential bilinguals – those children who are native speakers of the minority language and acquire the majority language predominantly at school – are also subject to strong societal and peer influences toward a preference for the majority language. In this chapter and in the next one, I examine available empirical evidence from studies on children consistent with Hypothesis 2.

2.

Unbalanced development in early childhood

An important difference between monolingual and bilingual acquisition is the potential outcome of the process, and this is related to input and use. Normally developing monolingual children receive sufficient language exposure to attain full linguistic competence in their native language. Knowledge of language is later reinforced at school through literacy. In contrast, many bilingual children receive much less exposure in each of the languages in early childhood, and may even have unequal opportunities to use the two languages. As a result, full linguistic attainment in the two languages is theoretically possible, but not guaranteed. Typically, children will have unbalanced development reflected in a stronger language and a weaker language, and language dominance can change during development. While it is difficult to characterize precisely, let alone quantify what counts as “insufficient” exposure in terms of quantity and regularity, one typically assesses the nature of the exposure by considering the bilingual child’s particular family situation. . Yang (2002, p. 103–105) provides hypothetical modeling of first language acquisition and approximate quantification of how much input is needed to set parameters in early child language. Based on an analysis of the CHILDES corpus for French, Yang established that sentences with a finite verb and postverbal negation constitute 7% of all French sentences heard by French-speaking children. He suggests that 7% of unambiguous evidence like this constitutes a lower bound that suffices for early acquisition of the verb movement parameter. He further suggests that any aspect of grammar with at least 7% of unambiguous input should also be acquired very early. In V2 grammars, the evidence to set the parameter comes from OVS sentences. Since OVS sentences are rare in the input to Germanic-speaking children, occurring only 1.3%

99

100 Incomplete Acquisition in Bilingualism

Children who speak a language that is not represented in the community, for example, German in Argentina or Australia, Finnish and Hungarian in the United States; and children who speak an ethno-linguistic minority language, such as Turkish children in Germany, Latino, Middle Eastern, African and Asian children in the United States and parts of Europe, typically develop weaker proficiency in the minority language. In these cases, the minority status of the language, and its relatively limited availability beyond the home, are linked to specific language choices young children make. Over time, these choices gradually affect bilingual children’s patterns of language use and, eventually, their linguiatic competence, as I show next.

2.1

Language choices

At a very young age, bilingual children become aware of their linguistic environment. They realize that one language is spoken by the majority speech community and the other may be spoken just by the parents. An example of this is Döpke (1992a, b), who investigated the language choices of a simultaneous bilingual child raised under the one-parent/one-language rule. While the family lived in Australia, he was exposed to German (the minority language) through his mother, and to English (the majority language) through his father. The child was reported to be fairly balanced up to the age 1;11, after which he started English daycare and input in English significantly increased. When the child was 2;7, the mother introduced the concept of “mommy words” and “daddy words,” making the one-parent/one-language strategy explicit to him. The child was tape-recorded once a month, and patterns of language choice were examined. The data was coded for use of German in German environments (i.e., when speaking to the mother), use of English in German environments, and language mixing in German environments. Similar coding was done for English. Results showed that, in the earliest stages of development, the child used English in German environments and German in English environments, but that overall, there was significantly more use of German in German environments and of English in English environments (see also Genesee, Nicoladis & J. Paradis 1995). Even though the child was responsive to his mother’s interventions to use German when spoken to in German, since the earliest stages, the child had a strong preference, nonetheless, for English. Similar preference for the majority language

of the time, V2 grammars are a relatively late acquisition (or are acquired later than a non-V2 grammar). Meisel (1986) shows that monolingual German children acquire V2 between MLU 1.75 and MLU 2.25, quite early in child language.



Chapter 4.  Bilingualism in early childhood 101

by very young bilingual children is documented by Lanza’s (2001) case study of an English-Norwegian child living in Norway. Although it is hard to generalize from a couple of case studies, what emerges from this type of research is that, from a very young age (2 years), bilingual children are already very aware of the status of their languages, and tend to have a preference for the majority language. Most often, the responsibility for maintenance of the minority language rests on the minority language-speaking parent. Lanza (1997) suggests that discourse strategies encouraging children to use the two languages (e.g., instructional techniques, repetitions in the other language, refusals to answer if not spoken in the right language, etc.), are decisive for establishing active bilingualism, particularly when employed by the minority-language speaking parent. But the fact is that, even if effective, parental strategies vary significantly from culture to culture and even among families that speak the same languages. Kasuya (1998), for example, followed 4 pre-school children of Japanese-English bilingual families living in the United States. Although all the Japanese-speaking parents in this study believed they were using Japanese exclusively and consistently with their children, Kasuya showed that the children used a lot of English with the Japanese-speaking parents: between 60% and 90% in the first recordings, between 60% and 80% in the second recordings, and between 20% and 73% in the third recordings. Except for one parent who used 30% English on the three occasions, all other parents used some English instead of only Japanese (up to 10%). In general, the children were not consistent in choosing Japanese when addressing the Japanese-speaking parent, because the children knew the Japanese-speaking parent could also communicate in English. Kasuya also showed that parental code switching always led to the child continuing the conversation in English. Children used more Japanese when the parents used explicit discourse strategies (e.g., instructing the child to say something in Japanese, correcting the child’s utterance, and translating the child’s utterance). Implicit or indirect discourse strategies (e.g., recasting the utterance in Japanese, moving on with the conversation, or code-switching to English) were not conducive to the child using the minority language more. To summarize, bilingual children who speak a minority language at home become aware very early in life of the political and social status of the languages they speak once they start to socialize beyond the home through childcare, friends, and other social interactions. Even when children are encouraged to use the minority language at home, preference for the majority language is very strong. When the family language is used less frequently than the majority language by simultaneous bilingual children, it runs the risk of becoming weaker, eventually affecting its vocabulary and grammar.

102 Incomplete Acquisition in Bilingualism

2.2

The weaker language

In early childhood bilingualism, the weaker language is defined as the less proficient or structurally underdeveloped language, as measured by mean length of utterance (MLU), morphosyntactic development, word types, functional elements, and degree of language mixing. In young bilingual children, bilingual balance is constantly in flux because the relative command of the two languages is highly susceptible to abrupt changes in the input, as shown by Kravin’s (1992) case study of an English-Finnish bilingual child living in the United States. The mother spoke Finnish and the father English, and although the parents adhered to the one-parent/one-language method with the child, they spoke English with each other. Kravin documents the progressive weakening of the child’s use and competence in Finnish in relation to different input factors, as well as the progressive preference and strengthening of English during the ages of 5;11–6;9. For example, at age 1;5 the child began childcare in English, and the mother noticed a drastic shift towards English. The family spent summers in Finland when the boy was 2;1–2;4 and later when the boy was 5;1–5;4. By the time the child was 6 years of age, his Finnish was quite weak. Although the data is not quantified in relation to English or to what age-matched monolingual Finnish children would do, the child used many English words in Finnish sentences, evidenced command of only 11 of 15 Finnish grammatical cases, and produced several omission errors with words and morphemes, and transfer errors from English. By and large, disfluency in Finish was quite generalized, according to Kravin. Underscoring the role of input in bilingual language development, this study shows how, in the absence of a broader speech community, input from only one parent does not seem sufficient to help the child develop and maintain full linguistic ability in the family language. The child evidenced omission of obligatory elements, problems with inflectional morphology, and word order errors (due to transfer). While some aspects of language may have been acquired, they were subsequently lost (attrition), whereas others aspects of the language may have not been acquired at all, showing delayed and, eventual incomplete acquisition. Fluctuations in input affect structural knowledge in young bilingual children, in particular the syntax and the morphology, which, as I argued in Chapter 3, are largely unaffected in adult attrition. For example, Schlyter’s (1993) study traced the structural characteristics of the weaker language in a longitudinal study of six Swedish-French bilingual children living in Stockholm. In each family, one parent spoke Swedish and the other one spoke French. The children were followed longitudinally since they were 3 years of age and recorded every 6 months for half an hour in each language during play sessions. In three of the six children, French (the minority language) was the weaker language, but in the other three cases the



Chapter 4.  Bilingualism in early childhood 103

Table 4.1  Percentage of correct finiteness per comprehensible utterance (adapted from Schlyter 1993). Child

Stronger language

Anne

French age

stage

MLU

2;7 2;11 3;3

III III IV

2.7 3.2 2.9

% correct finiteness 40% 55% 47%

2;0 2;2

III IV

2.1 3.2

Swedish 2;2 III 2;6 IV

1.7 2.7

Mimi

Jean

Weaker language Swedish age stage

MLU

2;11

III

2.5

% correct finiteness   6%

3;3

IV

2.9

34%

18% 52%

2;2

IV

2.6

14%

12% 58%

French 2;6 3;1

III IV–V

2.0 3.0

  9%   5%

weaker language was Swedish (the majority language) because the children spent more time with their French-speaking parent. Schlyter looked at the development of subjects, finiteness, and word order. The speech samples in each language were coded for MLU (mean length of utterance) and types of errors, and were divided into different developmental stages based on number of words (stage I one word, stage II two words), appearance of inflectional morphology (stage III), and emergence of more complex syntactic structures (stage IV). Table 4.1 shows accuracy percentage of finiteness per utterance (calculated on the basis of correct tense and person marking on the verb, explicit subject and correct word order) for 3 of the 6 children (Anne, Mimi and Jean). According to the results in Table 4.1, the accuracy percentage in the stronger language is considerably higher than in the weaker language, even when it does not reach 100% correct. Thus, these children are still in the process of acquisition in the two languages. Schlyter argues that the stronger language followed the developmental path and schedule of monolingual acquisition, whereas the weaker language developed like a second language. (I will expand, and critically assess, this particular interpretation of weaker language development in §5.) Errors of omission of obligatory elements were frequent and persistent in the weaker language, which Schlyter likens to developing and fossilized adult interlanguages in L2 acquisition. The results of the children whose weaker language is French are consistent with the general finding that it is usually the minority language which lags behind. This study also shows that the majority language (Swedish) can develop slower when input in this language is restricted as well.

104 Incomplete Acquisition in Bilingualism

The potential structural similarities between the weaker language of simultaneous bilingual children and a second language are further pursued by Schlyter and Håkansson (1994). Schlyter and Håkansson investigated the acquisition of V2 in Swedish in three types of children: 5 monolinguals, 5 sequential bilinguals or child L2 learners of Swedish (4–5 year olds), and 6 simultaneous French-Swedish bilinguals (3 with Swedish as weaker language and 3 with Swedish as stronger language). Like all Germanic languages except English, Swedish is a verb-second (V2) language; that is, subject-verb inversion is obligatory when the sentence has a topic (adverbial or object in sentence initial position), and in yes-no questions, as in (1) and (2) below. If there is no adverbial, or if the adverbial is not a topic, the order is S-V, as in (3) and (4). In subordinate clauses, there is no S-V inversion (5). Thus, Swedish has sentences with (X)-V-S order and sentences with S-V-(X) order. (1) Nu kommer han (X-V-S) now comes he ‘He comes now.’ (2) Kommer han? (V-S) comes he ‘Does he come?’ (3) Han kommer (S-V) he comes ‘He comes.’ (4) Han kommer he comes ‘He comes now.’

nu (S-V-X) now

(5) . . . eftersom han inte since he not ‘Since he is not coming . . .’

kommer (S-Neg-V) comes

Schlyter and Håkansson focused on sentences in which finite verbs either followed (SV) or preceded (VS) the subject and found that, like adult native speakers, the monolingual children produced V2 sentences with almost 98% accuracy, while the L2 learners used many sentences with ungrammatical XSV (V3) order (between 3% and 11.8%, average 7.5%). Furthermore, while the adult and child monolinguals used SV around 60% of the time, the L2 children used SV 90% of the time. The 3 simultaneous bilingual children with Swedish as the stronger language patterned with the monolingual children, whereas the 3 children with Swedish as the weaker language patterned with the L2 learners, both in their incorrect use of V3 and in their preference for SV over VS. Meisel (2007) correctly points out that although the percentage of use of word order in L1 and bilingual



Chapter 4.  Bilingualism in early childhood 105

children with Swedish as a weaker language differ, this does not necessarily indicate, at least at the level of syntax, that the children who speak Swedish as a weaker language have not set the V2 parameter. In fact, they do know V2 and VS: they just do not use those options as much as the monolingual children. It is hard to argue, solely on the basis of percentage of word order use, that these simultaneous bilingual children with Swedish as the weaker language lack the relevant knowledge of Swedish V2. Yet, the stronger argument for a different grammatical system in this case is not actually the rate of use of V2, but the fact that the children with Swedish as a weaker language use a substantial percentage of V3 structures, not apparently present in the productions of the children with stronger competence in Swedish. Another way in which the structural characteristics of the weaker language have been observed is in code switching patterns, where the stronger and weaker languages interact. Although code switching is a sign of sophisticated linguistic knowledge in balanced bilinguals, unbalanced bilinguals may resort to code switching to compensate for structural or lexical deficits in the weaker language (Bolonyai 1998; Kaufman & Aronoff 1991; Pfaff 1996). Bernardini and Schlyter (2004) proposed the Ivy Hypothesis to account for a particular pattern of code switching observed in simultaneous French-Swedish and Italian-Swedish bilingual children (ages 2–4) with unbalanced development. Following Radford’s (2000) Gradual Structure Building Hypothesis of L1 acquisition, according to which functional structure in the syntax is built incrementally, Bernardini and Schlyter proposed that the weaker and stronger languages differ on the availability of higher functional projections like DP (determiner phrase) and CP (complementizer phrase), both available in the stronger language but not in the weaker language. In other words, more structure is projected in the stronger language than in the weaker language. Bernardini and Schlyter found that when the very young children wanted to communicate in the weaker language, they often code switched sentence-internally. However, the switches were of a particular type. Children mixed or introduced functional elements from the stronger language into the weaker language when using structures assumed to be higher up in the syntactic tree (DP, IP, CP), which were undeveloped (or lagged behind) in the respective weaker languages of the children. Examples from Bernardini and Schlyter are shown in (6) and (7) below. (Bolded words represent the stronger language). (6) Titta på le rasoir ‘Look at the razor.’

(Alex 2;6 Swedish strong, French weak)

(7) var är la mucca? ‘Where is the cow?’

(Lukas 2;3 Swedish strong, Italian weak)

106 Incomplete Acquisition in Bilingualism

Table 4.2  Overall patterns of code switching in a Hungarian-English bilingual child (adapted from Bolonyai 1998). Stronger language in code switching Age 3;7 4;2 4;10

Total # CPs 143 116 188

CPs with CS   4.8% (7/143) 43.1% (50/116) 57.9% (109/188)

Hungarian 71% (5/7) 52% (26/50) 77% (84/109)

English   0% (0/7) 38% (19/50)   2% (2/109)

Thus, in (6) and (7), the weaker language clings to the structure of the stronger language for support – like an ivy. Rather than a sign of linguistic proficiency in the two languages, code switching in these cases is seen as a compensatory strategy for filling lexical gaps in the developing children’s linguistic competence. Fluctuating patterns of code switching as a function of bilingual imbalance and input in a young bilingual child are also discussed by Bolonyai (1998), who presents a case study of an English-Hungarian bilingual child (Hanna) living in the United States since age 1;5, and observed during the ages 3;7, 4;2 and 4;10. The child’s parents were Hungarian and spoke Hungarian to each other and to Hanna at home. At 1;8, Hanna started English nursery school (3 hours a day 4 times a week). At age 3;5, she started attending another school for 8 hours a day, 5 days a week. At age 4;3, Hanna spent 2 months in Hungary, where she spoke Hungarian exclusively with the family and other Hungarian-speaking children. After her return from Hungary at age 4;5, Hanna was back at school. The linguistic analysis of Hanna’s speech focused on CPs (Complementizer phrases). Bolonyai observed a significant increase of CPs over the three periods studied, with an increase of code switching over time. Interestingly, the matrix language (or the language providing the syntactic structures for “the ivy” described above), which was typically the stronger language, shifted over time, as shown in Table 4.2. At age 3;7, Hungarian is the stronger language, and the switches reflect that change, as in (8). (8) mi-t cook-sz? what sg. acc cook 2sg pres ‘What are you cooking?

At age 4;2, Hanna had spent more than two years in the United States interacting with English-speaking children at school. As a result, her social network and identity became more associated with English, and these socio-affective changes are reflected in the structure of code-switched utterances. English use and ­proficiency



Chapter 4.  Bilingualism in early childhood 107

must have improved at this time because it now competes with Hungarian as the matrix language in code switching 38% of the time (from 0% in the previous period), as shown in (9). (9) But I want the lion to sit by me, I want, I want some, I want you to put this mosó-gép somewhere. ‘But I want the lion to sit by me, I want, I want some, I want you to put this washing machine somewhere.’

At age 4;10 – six months after the child’s visit to Hungary – the pattern of code switching observed in the previous period is reversed, where Hungarian is again the preferred stronger language in code switching (77%) and English declines (2%). (10) és akkor egy még cracker vol-t and then one more cracker be-3sg past ‘and then there was one more cracker.’

In conclusion, the collective results of all these case studies of very young children show that bilingual balance, as evaluated from spontaneous production, is a function of input, even though bilingual children have the biological and cognitive potential to become fully competent in the two languages. Children exposed to family bilingualism, or who are members of an ethno-linguistic minority culture, do not usually get the full benefits of substantial linguistic exposure and use of the two languages. Typically, the less frequently used language lags behind in development and becomes functionally, psycholinguistically, and structurally weaker. The fact that bilingual balance at this early age fluctuates with abrupt changes in input suggests that linguistic knowledge, as reflected in language use, must not be very stable at this early age. Although the weaker language displays delayed development, the effects of reduced input at this early age can also take the form of attrition or incomplete acquisition, compromising the children’s linguistic competence in specific grammatical areas.

3.

Attrition or incomplete acquisition?

Because adults have reached their mature linguistic state, if language loss occurs in the L1 after disuse and extensive exposure to the L2, this is typically characterized as L1 attrition (see Chapter 3). But when disuse of the L1 and extensive exposure to the L2 occurs in children during the age of linguistic development, L1 loss can be due to attrition, incomplete acquisition, or both.

108 Incomplete Acquisition in Bilingualism

To illustrate these distinctions, imagine a 3 year old Spanish-speaking child, who has acquired that Spanish is a pro-drop language with flexible word order, uses the preterit-imperfect aspectual distinction in the past, produces all nouns with correct gender agreement, uses subjunctive in commands and volitional contexts, but does not use or comprehend subjunctive in presuppositional contexts, such as in relative clauses – does not use or comprehend verbal passives

– – – –

The child is introduced to English at day-care, but uses Spanish at home with family. Two years later, at age 5, the child’s Spanish exhibits these changes: – significantly more use of overt than of null subjects and word order is less flexible (SVO), – sometimes confuses preterit for imperfect and vice versa, – has more than 20% gender agreement errors with nouns, – does not produce or understand subjunctive in relative clauses, – does not use or understand the verbal passive. One can safely assume that the first three properties (subjects, agreement, aspect), which underwent changes and regressed, are instances of attrition. But the last two properties, subjunctive in relative clauses and passives, were not present at age 3 and are no present at age 5, even when normally developing monolingual 5 year olds have already acquired those two properties. These, I claim, can be taken as cases of incomplete acquisition. And yet another possibility is to imagine a child who does not fully master a given grammatical domain, producing a high error rate. Two years later the errors rates are even higher, and the given property remains underdeveloped into adulthood. In this case, there is partial loss of something that was partially acquired. This would be a case of attrition on top of incomplete acquisition. Hence, children can exhibit attrition, incomplete acquisition, or both in the same or different aspects of their grammar. Unlike L1 attrition in adults, L1 attrition in young children occurs in a relatively short period of time, and the effects are much more extensive than what is observed in adults. My specific claim is that although attrition in children may also start as difficulty retrieving lexical items and assembling items together due . This could be because the structures are available in the input, but are of very low frequency, or because the child is not exposed to input that contains the relevant structures, as in the case of speakers who learn vernacular and standard varieties of a language. See Pires and Rothman (in press) mentioned in Chapter 5.



Chapter 4.  Bilingualism in early childhood 109

to lack of use, it eventually leads to fairly substantive gaps in areas of grammatical knowledge largely unaffected in adults, such as morphology and syntax-related interfaces. Because the language looks as if parts of the grammatical system did not reach full development and stabilized at a more simplified stage, I refer to the outcome of attrition in childhood as incomplete acquisition. And as I will demonstrate in the next chapters, incomplete L1 acquisition persists into the adult mature state. Since fluctuations in input and experience alter very rapidly the degree of command of the two grammars in bilingual children, the ideal way to tease apart cases of incomplete acquisition from cases of L1 attrition in early childhood is by conducting longitudinal studies by which a child’s linguistic development or regression in the weaker language is documented over an extended period of time. Another way to indirectly assess incomplete acquisition and attrition is by comparing the developmental stage of a bilingual child’s weaker language at a given age to that of a fluent bilingual or monolingual child of the same age and cognitive development. But most often, the published chronological schedules available from the L1 acquisition literature are used as benchmarks (see Keijzer 2007, for example). Still others have used competent adults as the baseline (Silva-Corvalán 2003). In the next two sections, I review the few existing, yet very important, longitudinal studies of children documenting attrition and incomplete acquisition. If L1 attrition of inflectional morphology in adults was minimal (Chapter 3), these studies show that, contrary to this, morphosyntactic development is highly volatile in children.

3.1

L1 attrition

Kaufman and Aronoff (1991) and Turian and Altenberg (1991) are two early studies of fast attrition in young children immersed in an English-speaking environment. Kaufman and Aronoff ’s study focuses on the “disintegration” of Michal’s verbal system, a fluent speaker of Hebrew who immigrated with her parents to the United States at the age of 2;6. According to what is reported, her command of Hebrew was age-appropriate at the time of immigration. At age 2;7, one month after her arrival to the new country, Michal started attending an English-speaking nursery school three hours a day and decline of her Hebrew proficiency soon ensued. Kaufman and Aronoff distinguish four stages of attrition, as evidenced in code-switched utterances: (1) onset of attrition; ages 2;9-3;1, after 4–7 months in the L2 environment, (2) bilingual period: ages 3;1–3;2, after 7–8 months in the L2 environment, (3) disintegration of L1: ages 3;2–3;5, after 8–11 months in the L2

110 Incomplete Acquisition in Bilingualism

environment, and (4) idiosyncratic template: ages 3;5–4;6, after 12–24 months in the L2 environment. Kaufman and Aronoff claim the onset of attrition began manifesting itself in the lexicon, when the child started to insert English nouns in Hebrew sentences. Although English verbs were also borrowed, they were not integrated with Hebrew verbal morphology at this stage. During the bilingual period, Kaufman and Aronoff believe the child to have balanced command of the two languages, although there is no quantification of data or independent measures of proficiency attesting to this assumption. Even though Michal continued to insert English verbs and nouns into Hebrew sentences, the syntax is reported to be intact at this stage. The third stage, labeled “disintegration of the L1,” coincides with the child’s reduced use of Hebrew and extensive use of English. At this stage, Michal continues to borrow English verbs, but the verbal forms used in Hebrew have now become non target like. Apparently, the child’s knowledge of Semitic roots is still intact, but the templates and inflections are now incorrectly used. By the fourth stage, Michal has created a hybrid verbal pattern that shows convergence towards English. Adopting the most colloquial and productive template in Hebrew child language, she overextends it to other idiosyncratic templates. Similar decline in morphosyntactic abilities is illustrated by Turian and Altenberg (1991) in a Russian-speaking child followed longitudinally from the ages of 3;00–3;7, and a year later, between the ages of 4;3–4;4. At time 1 (3;00–3;07), the child had fluent and grammatically accurate command of Russian. Turian and Altenberg report correct and age-appropriate use of verbal forms, negation, nominal cases, agreement morphology, and conjunctions. By time 2 (4;03–4;04), the child exhibited loss of the nominative-accusative distinction and present tense verbal forms. Further evidence for attrition came from the fact that the child borrowed words from English into Russian, began to code switch more frequently, and evidenced syntactic transfer from English onto Russian. Because the two children in these two studies were reported to have more knowledge of Hebrew and Russian at the beginning of the data collection than at the end of the data collection – more than a year later – one must assume that their language underwent some form of attrition. However, since no comparable data on age-matched monolingual children acquiring Hebrew and Russian are provided, there is no way to tell with more certainty whether the children had indeed completely acquired the structures and forms in question. Furthermore, with the evidence based exclusively on isolated examples, none of the studies provides adequate quantification of data attesting to the magnitude of attrition revealed in the two children. Therefore, the researchers’ conclusions about the extent of attrition present in these children, although suggestive, need to be taken



Chapter 4.  Bilingualism in early childhood

with extreme caution, since it is also very common for young children to experience incomplete development of the minority language, as I show next.

3.2

Incomplete L1 acquisition

In this subsection I discuss two longitudinal studies conducted by Anderson (1999, 2001) and one by Silva-Corvalán (2003) that illustrate attrition and incomplete acquisition at the level of inflectional morphology. In §3.2.2 I discuss similar effects at the syntax-discourse interface.

3.2.1 Morphosyntax Anderson (1999, 2001) looked at the loss of nominal and verbal inflectional morphology in two normally developing Spanish-speaking siblings – Beatriz and Victoria. The siblings came to live to the United States at the ages of 3;6 (Beatriz) and 1;6 (Victoria). Data collection started almost three years after, when the children were 4;7 and 6;7, and ended two years later, when the children were 6;5 and 8;5. The two parents were from Puerto Rico, and Spanish was spoken at home. At the beginning of the study, the siblings spoke Spanish and English with each other, but toward the end of the data collection period, the children shifted to using only English with each other and continued using Spanish with their parents. When data collection started, the older sibling had some literacy skills in Spanish, and during the course of the study the two children developed literacy skills in English in daycare and pre-school. The siblings were videotaped in play sessions with their parents every 1–2 months over a period of 22 months. The results of gender agreement in noun phrases are reported in Anderson (1999); results of verb morphology in Anderson (2001). Spanish nouns are inflected for gender (masculine or feminine) in the lexicon. Most masculine nouns end in the vowel -o and feminine nouns in -a, although there are several exceptions. Gender is also a syntactic property. Within the noun phrase, there is concord between the head noun, the determiner and the adjective, as in (11) and (12). (11) La bufanda the-fem scarf-fem ‘The white scarf ’

blanca white-fem

(12) El coche the-masc car-masc ‘The yellow car’

amarillo yellow-masc

111

112 Incomplete Acquisition in Bilingualism

Table 4.3  Percentage of gender agreement errors in the Spanish of two bilingual siblings (adapted from Anderson 1999).

Beatriz Victoria

Age first recording

% errors on gender agreement

Age last recording

% errors on gender agreement

6;7 4;7

0% 8%

8;5 6;5

  5.8% 18.2%

Longitudinal studies based on spontaneous production data suggest that gender agreement with all the elements of the determiner phrase emerges around ages 1;5–1;6 in normally developing monolingual Spanish-speaking children (Hernández Pina 1984; Lleó 1998; López Ornat 1997). Common errors at this age are regularization of invariant adjectives or non-canonical feminine or masculine nouns (e.g., *tierra azula for tierra azul or *mota roja for moto roja), or the overextension of masculine to feminine nouns ending in -e (*un leche, *un llave) (Hernández Pina 1984). By the age of 3 years, children produce gender and number agreement morphology with 100% accuracy, and some as young as 2;01 (Snyder, Senghas & Inman 2001). Anderson (1999) found that the two siblings made errors with gender, as shown in Table 4.3. At the time of the first recording, Beatriz – the older sibling – produced gender agreement 100% correctly (0% error), like a typical monolingual child her age. Two years later, she was producing 5.8% errors. Victoria produced 8% errors during the first recording. By the end of the data collection period, there was a steady increase in the production of gender errors to 25%, 17.4% and 18.2% in the last three sessions. While the percentage of masculine and feminine errors were comparable in Beatriz’s production, Victoria produced more errors with feminine nouns, ranging from 16.7% (sessions 7 and 9) to 50% at sessions 10 and 11. Errors with masculine nouns reached 7.1% by the last session. Judging by the ages, the older child could be considered a sequential bilingual (AoA in the U.S. 3;6), while the younger sibling would be a simultaneous bilingual (AoA in the U.S. 1;6). The results of this longitudinal study show that after only two years of contact with English, gender agreement in Spanish was affected, more in the younger child that in the older child. The older sibling – Beatriz – represents a case of L1 attrition, since at the time of first recording she appeared to control agreement with 100% accuracy, and two years later she was producing almost 6% errors. Even when 6% looks like a low percentage to claim attrition, for gender agreement it is not low because, once acquired, children and adults produce gender agreement with 100% accuracy with words they know. And more than 5% in gender agreement errors in adult L2 learners has been taken as a sign of fossilization (Franceschina 2001, 2005). Recall that the error rates for plural



Chapter 4.  Bilingualism in early childhood 113

morphology in German adults under L1 attrition discussed in Chapter 3 were less than 2%. On the other hand, the younger sister – Victoria – is a case of incomplete acquisition and attrition, since she produced 8% errors at the time of first recording and more than doubled this error rate in a two-year period. Consistent with the general tendency of marked forms to be more affected than unmarked forms under attrition, more errors were observed with the feminine than with the masculine forms of nouns, especially in the younger sibling. If gender agreement is affected under attrition and incomplete acquisition in these bilingual siblings, is verbal agreement a vulnerable area as well? Spanish verbs are inflected for person, number, tense, aspect and mood. Verbs are classified into three groups – -ar, -er and -ir – depending on the thematic vowel of the infinitive ending. Inflected forms have a stem, consisting of the root plus a thematic vowel (a, e or i). The affixes for tense-aspect and mood, and for person and number, are added to the base form of the verb, as shown in (13) with the verb cantábamos (we sang-imperfect). (13) [cant + á]stem sing ‘We sang’

+ baAf1 + past imperfect indicative

+ mosAf2 + first person plural

The paradigms for person and number include three persons (1st, 2nd and 3rd) and two numbers (singular and plural). With respect to the tense, aspect and mood (T-A-M) system, Spanish distinguishes between present, past, future and conditional tenses, perfective (preterite) and imperfective (imperfect) aspect in the past, and between indicative, subjunctive and imperative mood. There are simple and compound finite verb forms. Table 4.4 shows four conjugations of the verb hablar “speak” in the present the past (preterite and imperfect) and the present subjunctive. Verb conjugations can also be regular and irregular in Spanish. Regular verbs, like hablar in Table 4.4, have no change in their stems across conjugations. Irregular verbs, on the other hand, can show changes in the stem (e.g., pedir “ask”, Table 4.4  Spanish verbal agreement paradigm. Person/number

present

preterite

imperfect

subjunctive

yo tú el, ella

habl-o habl-as habl-a

habl-é habl-as-te habl-ó

habl-a-ba habl-a-bas habl-a-ba

hab-le hab-le-s hab-le

nosotros vosotros ellos

habl-a-mos habl-á-is habl-a-n

habl-a-mos habl-á-steis habl-a-ron

habl-á-bamos habl-ás-steis habl-a-ban

habl-e-mos habl-éi-s habl-e-n

114 Incomplete Acquisition in Bilingualism

yo pido “I ask”), the verbal suffix (tener “have”, yo tengo “I have”), or both stem and suffix (ser “be”, yo soy “I am”, tú eres “you are”). Irregularities can affect all persons, or just some. Unlike English-speaking children who produce a high percentage of uninflected verbal forms at early stages of acquisition, normally developing Spanish-speaking children produce verbal inflections from the onset of verb production before the age of 2 years (Bel 2001; Mueller Gathercole, Sebastián & Soto 1999; ­Grinstead 2004; López Ornat 1997; Torrens 2002). Children distinguish between 1st and 3rd person and between present and past forms. The preterite emerges before the imperfect, and the present subjunctive in imperative forms is already present by age 2;00 (Montrul 2004b). Clahsen, Aveledo & Rocca (2002) show that Spanish-speaking children make overgeneralization errors with irregular forms until they reach almost perfect command of verbal morphology at the age of 3 years. In comparison to the results for gender, Anderson (2001) found that, from all verbs for each child, the error rate was quite low. Still, the younger sibling (Victoria) produced more errors with verbal morphology than the older sibling (Beatriz), as shown in Table 4.5 (all periods combined). The siblings were not very different from each other with respect to overall error rate with different morphology (person/number, tense-aspect and mood), as shown in Table 4.6. The siblings frequently used the 3rd person singular for the 1st person singular (*va a cocinar instead of voy a cocinar “I am going to cook”), indicative for subjunctive mood (habla for hable), and use of preterite for imperfect (fue for era). Analysis of verb types showed that for both children, the number of verbs used with person and mood contrasts decreased with time, more drastically in Victoria (the younger sibling) than in Beatriz (the older sibling).

Table 4.5  Percentage of verbal agreement errors in the Spanish of two bilingual siblings (adapted from Anderson 2001).

Beatriz Victoria

Age first recording

% errors on verbal agreement

Age last recording

% errors on verbal agreement

6;7 4;7

2.3% 3.8%

8;5 6;5

  5% 13.9%

Table 4.6  Error rates with different verbal morphemes (adapted from Anderson 2001).

Beatriz (older) Victoria (younger)

Person-number

Tense-aspect

Mood

42.3% 44.7%

9.6% 8.5%

44.7% 36.2%



Chapter 4.  Bilingualism in early childhood 115

Contrastive morphological forms were reduced in person-number distinctions and in mood distinctions, but not so much in aspect distinctions. Overall, this study showed attrition in person-number but incomplete acquisition of mood morphology, since the error rates for person-number increased towards the end, whereas mood errors remained stable across all the taping sessions. This differential pattern of errors between the two siblings can be explained by age of onset of bilingualism and time of exposure to Spanish as a first language, strength and/or degree of acquisition of their first language before L2 acquisition of English began. Beatriz could be considered a sequential bilingual child, whereas Victoria is a simultaneous bilingual child. Consistent with Hypothesis 2 (§1.2) incomplete acquisition appears to be more dramatic in simultaneous than in sequential bilingualism, due to the reduced amount of input received in one language while the language was not yet fully developed. Further empirical evidence for incomplete acquisition of the verbal inflection system of Spanish in simultaneous and sequential bilingual children comes from Silva-Corvalán (2003). Silva-Corvalán’s study suggests that many of the features of incomplete acquisition and simplification of the verbal system observed in adult bilinguals (Chapter 6) can indeed be traced back to childhood attrition. Silva-Corvalán compared the results of the bilingual adults studied in Silva-Corvalán (1991, 1994) with those of seven pre-school children (ages 5;1–5;11) who acquired Spanish from birth, and English, either from birth or later. Table 4.7 presents information about the children, classified according to their perceived dominant language, language spoken at home, and language spoken at school. Table 4.7  Linguistic characteristics of the bilingual children studied by Silva-Corvalán (2003). child

Dominant Language

Home Language

School Language

Daisy Cindy Bryan Mike Chris Nico Bren

Spanish Spanish Spanish/English English English English English

Spanish Spanish Spanish/English Spanish/English English English English

Spanish/English English Spanish/English English Spanish/English Spanish/English English

. Five of the children were of Mexican-American origin and the other two were ChileanAmerican children (Silva-Corvalán’s grandchildren) who were studied longitudinally from the time they were 11 months old.

116 Incomplete Acquisition in Bilingualism

Oral production samples from the children were recorded on three different occasions, for a total of approximately 6 hours each (3 hours in each language). Overall results showed that the children were at a fairly similar stage of development for the Spanish tense-mood-aspect system, regardless of language dominance and home language. Except for Cindy, who did use the pluperfect subjunctive (hubiera + past participle), the rest of the children never produced the future or conditional, the perfective infinitive, the pluperfect indicative, the imperfect subjunctive, the present perfect subjunctive, or the pluperfect subjunctive, even when the context called for these forms. All these verbal forms were produced by the adult Spanish monolingual speaker used as baseline. The children exposed to English and Spanish since birth – the simultaneous bilinguals – showed less ability with the preterite-imperfect aspectual contrast than the sequential bilinguals. Again, this is consistent with Hypothesisis 2: degree of attrition will be more severe in simultaneous than in sequential bilinguals. Some of the children made overregularization and person agreement errors with the preterite, while others used imperfect forms in perfective discourse contexts and perfective forms in imperfect discourse contexts, especially with stative verbs. Silva-Corvalán states that the bilingual children with less exposure to Spanish at home displayed the same problems with the perfective/imperfective distinction as the 2nd and 3rd generation adults studied in Silva-Corvalán (1994), suggesting long-lasting effects of incomplete acquisition into adulthood. Interestingly, the two children from whom longitudinal data over 3 years was available – Nico and Bren – showed very little development in their verbal system from ages 2;11–5;11. This suggests L1 fossilization at a stage of incomplete acquisition. With respect to the use of the subjunctive, Silva-Corvalán also noted a few differences that appear to correlate with degree of exposure to Spanish and type of bilingualism. Again, simultaneous bilinguals have less knowledge of Spanish than sequential bilinguals, supporting Hypothesis 2 (§1.2). For example, Daisy and Cindy, who spoke only Spanish at home, used preterite, imperfect, and present subjunctive as the adult monolingual did, and they were beginning to use imperfect subjunctive. In contrast, the children from bilingual and English-only homes did not use imperfect subjunctive, and Nico and Bren, the children from an English-only home, did not use the present subjunctive at all in the recordings corresponding to ages 5;5 and 5;6. That the children do not have subjunctive is not surprising, since in monolingual acquisition all the complexities of subjunctive use . The recordings included free conversations, play sessions with puppets, answers to a set of questions about hypothetical situations, and story retelling guided by a picture book (Frog, where are you? Mayer 1969). Results of the seven children’s verb systems were compared with the results of one monolingual adult Mexican Spanish speaker.



Chapter 4.  Bilingualism in early childhood 117

are mastered quite late (Blake 1983; Pérez-Leroux 1998). As Silva-Corvalán correctly points out, it is possible that when many bilingual U.S.-born children start kindergarten, they have not yet acquired the complete Spanish TAM system. To see potential attrition in these early bilingual grammars, Silva-Corvalán examined in more detail the data from the two children studied longitudinally from age 2;10 living in an English-speaking home. Very little progress in almost 3 years was observed, since the children’s verb system throughout all the ages seemed incomplete, as compared with the verbal systems of age-matched monolingual children. When the bilingual children were 5 years of age, no other new forms were consistently used beyond those acquired by age 2;10-2;11, suggesting again incomplete acquisition and probably fossilization. For example, Nico produced imperfect subjunctive forms sporadically before age 5;6, but after this age, all simple and compound subjunctive forms were absent in the data. Indeed, Nico and Bren’s tense-mood-aspect system at age 5;6 appears to be further reduced than the system they had acquired by age 3;0–3;3, when the present subjunctive was attested in their speech. The results of Anderson’s and Silva-Corvalán’s longitudinal studies show that pre-school bilingual children who speak a minority language run the risk of developing incomplete proficiency in the minority language as a result of intense contact with the majority language and reduced exposure and use of the minority language. Very often, minority language-speaking children do not reach age-appropriate levels of morphological development. This raises the question of whether incomplete acquisition in bilingual children is also severe in other aspects of syntax and its interfaces with other domains.

3.2.2 Syntax-pragmatics interface Except for the studies already mentioned in this chapter, research on incomplete acquisition of syntax and other interfaces in young bilingual children is scarce, but there are two studies on subject expression suggestive of this possibility in simultaneous bilingual children. Recall from Chapter 3 that subject expression, handled by the discourse-syntax interface, was a vulnerable area in adult L2 acquisition and L1 attrition, as studies of Spanish-English and Italian, Greek and English bilinguals showed (Sorace 2000a; Tsimpli et al. 2004). In this section, I show how this interface is also affected in early bilingual children, and that the error rates are higher than those reported in adult attrition. Being pro-drop languages, Spanish and Italian can omit subject pronouns because the morphosyntactic information about nominative case, person, and number can be recovered from the rich person and number inflection on the verb. Still, it is not the case that subject omission in these languages is entirely

118 Incomplete Acquisition in Bilingualism

Table 4.8  Percentage of overt subjects in a bilingual and two monolingual Spanish children (adapted from J. Paradis & Navarro 2003). Child

Stage I

Manuela (bilingual) Emilio (monolingual) María (monolingual)

% overt 30.8  5  7

Stage II range 14–71   0–6   0–18

% overt 39 24 44

range 32–57 12–30 33–61

optional; rather, it is regulated by discourse-pragmatic considerations, as we saw in Chapter 3. Overt subjects are used when a new referent is introduced in discourse, while null subjects are used in cases where the subject has already been introduced and carries low informative value in discourse. J. Paradis and Navarro (2003) conducted a study of the discourse-pragmatic distribution of null and overt subjects in a Spanish-English bilingual child – Manuela (CHILDES database, Deuchar corpus). Manuela was exposed to Cuban Spanish from the father and to L2 Spanish from her English-speaking mother. J. Paradis and Navarro compared the production of null vs. overt subjects in this child with the data of the two Mexican children studied by Grinstead (1998), to see whether the two-stage development of subjects Grinstead argued for in monolingual acquisition carried over to bilingual acquisition. Due to potential influence from English, J. Paradis and Navarro predicted that Manuela might produce more overt subjects than monolingual children, and might not exhibit the two-stage developmental pattern attested in Grinstead’s monolingual children. For comparison, Table 4.8 shows mean percentages and rate of subjects use by Manuela (bilingual) and Emilio and Maria (monolingual). Manuela’s pattern of development does not clearly fit the two-stage model (i.e., early stage with no overt subjects, late stage with overt subjects) evidenced by the two monolingual children, since she produces a high rate of overt subjects from the very beginning (30.8% compared to 5% and 7%). Manuela used more overt pronoun subjects than the other children. With respect to word order variation, J. Paradis and Navarro noticed that Manuela used fewer postverbal subjects in stage two (5%) than Emilio (19%) and María (42%). Therefore, Manuela’s syntax of subjects is clearly different from that of the monolingual children and seems influenced by English, a language with little word order flexibility in which overt pronouns are the norm. J. Paradis and Navarro looked at the pragmatic constraints on the use of overt subjects by Manuela, identifying three categories:



Chapter 4.  Bilingualism in early childhood 119

Table 4.9  Pragmatic contexts of overt subject use in bilingual and monolingual children (adapted from J. Paradis & Navarro 2003). Child

Pragmatically necessary Pragmatically odd Indeterminate

Manuela (bilingual) Emilio (monolingual) María (monolingual)

44.6% 48.5% 65.8%

32.2% 20.8%   9.8%

23.2% 30.7% 24.4%

pragmatically necessary, pragmatically odd, and indeterminate. These results appear in Table 4.9. Of all the children, Manuela had the highest percentage of pragmatically odd subjects, and although María also produced many overt subjects, their pragmatic distribution is clearly different from those of Manuela: only 65.8% of María’s overt subjects were pragmatically necessary and only 9.8% were pragmatically odd, while 44.6% of Manuela’s overt subjects were pragmatically necessary and 32.2% – 3 times the percentage for María – were pragmatically odd. An examination of parental speech revealed that Manuela’s parents used more overt subjects than the other two parents. Nevertheless, J. Paradis and Navarro concluded that there is crosslinguistic interference from English in Manuela’s speech because Manuela does not show the Stage I to II pattern of overt subject emergence documented in monolingual children. A potential problematic aspect of this study is that Manuela was exposed to Caribbean Spanish, a Spanish variety that is moving away from the pro-drop parameter and exhibits higher rates of overt subjects Toribio (2000), while the monolingual comparison children spoke Mexican Spanish, a variety that is still strictly pro-drop. Thus, their conclusion about crosslinguistic influence from English is not the only explanation for the patterns of subject use found in Manuela. Furthermore Silva-Corvalán’s (2007) case study of a 2-year-old Chilean Spanish-English bilingual child living in the U.S. does not support Paradis and Navarro’s conclusion either, since this child’s use of overt/null subjects in Spanish and English matched the input. Yet, the crosslinguistic influence explanation advanced by J. Paradis and ­Navarro finds support from another study of subject expression in a simultaneous English-Italian bilingual child. Serratrice, Sorace and Paoli (2004) analyzed the subject and object use of Carlo, an Italian-English bilingual child living in the UK, during the ages 1;10-4;6. Carlo’s spontaneous production data were ­ compared

. Pragmatically necessary overt subjects included subjects that introduced new information, were used for emphasis, or for disambiguation/clarification. Pragmatically odd subjects were clearly redundant. The indeterminate category included subjects that did not fit the other two categories.

120 Incomplete Acquisition in Bilingualism

with data from age-matched monolingual English and Italian children. Results showed that in terms of subject omission, Carlo’s proportion of null subjects was well outside the monolingual range at all stages, since he used significantly more overt subjects than the Italian monolingual children, especially with 3rd person pronouns. Results of the pragmatic analysis of subject discourse functions showed that, like the monolingual Italian children, Carlo was sensitive to the discoursepragmatic status of referents. That is, he used null subjects in cases where the subjects carried low informative values. But when it came to overt subjects, Carlo produced 9% of overt subjects in cases where a null subject was pragmatically more appropriate. In all these cases, the overt subject was coreferential with a topic antecedent and as such should have been realized as null. If cross linguistic influence lies at the heart of this overt subject phenomenon in simultaneous bilingual children who speak a minority language at home, it may be the case that overuse of overt subjects due to cross linguistic influence from English never goes away in adulthood, as we will see in Chapter 6. In conclusion, the experimental evidence presented in this section shows that in minority-language speaking children, the weaker language very often lags behind with respect to the language of other fluent bilingual or monolingual children of the same age and cognitive development. Due to reduced input and use, children seem to experience attrition and incomplete acquisition in specific grammatical areas, depending on age and level of grammatical attainment. Unlike adults assumed to be cases of attrition, children show incomplete mastery of nominal and verbal morphology, in addition to showing divergent outcomes at the discourse syntax interface as well. The error rates seen in children are much higher than what has been reported in adults to date.

4.

Total L1 loss?

So far we have looked at cases of minority language-speaking immigrant children who continue to use the family language as they begin exposure to and acquire the second language spoken by the speech community. Due to diminished opportunity to hear and use the minority language, these children exhibit a marked degree of attrition and incomplete acquisition in morphosyntactic aspects of the heritage language. Although more severe than in adults, attrition seems to be local and selective. In this section, we turn to cases in which input to the first language is interrupted altogether. In Chapter 3, I showed that lack of exposure to or use of the L1 does not affect linguistic competence in adults. Adults can speak the L1 fluently and correctly, even if they had not used their L1 much for several years. However,



Chapter 4.  Bilingualism in early childhood 121

if input is completely interrupted in childhood, it seems to affect linguistic competence in the L1 quite dramatically, and even globally. Furthermore, the younger the child when exposure to the minority language stops altogether, the more severe the loss. What happens to the first language when exposure stops abruptly and completely and is replaced by the second language? Internationally adopted children represent an ideal population to test the relationship between lack of input, early age, and L1 loss directly. It has been documented that, albeit anecdotally, internationally-adopted children quickly lose their first language within the first year following adoption. Language loss occurs more rapidly in infants than in toddlers, who had less contact with the L1 (Glennen 2005). Recent research has shown that the majority of children adopted from China and Eastern Europe by American families learn English rapidly and with few delays (Pollock 2005). Contrary to popular belief, there is little evidence for the claim that the sudden and complete language shift in language environment faced by internationally adopted children causes them to experience language delays with the acquisition of the new language. However, this is related to age: the older the child at the time of adoption, the more difficulty the child may have learning English. Studies by Krakow, Tao and Roberts (2005) and Geren, Snedeker and Ax (2005) showed that a group of infants and toddlers demonstrated remarkable lexical and syntactic development after 6–12 months in the United States. By 3 years of age, nearly all these young children performed at the level of comparable, non-adopted, monolingual English peers (Pollock & Price 2005). However, none of these studies specifically report how severe the loss of the L1 can be. A study that is unique in following an internationally adopted child’s L1 and L2 development soon after adoption is Nicoladis and Grabois (2002). Nicoladis and Grabois provide us with a compelling case study of a Chinese infant adopted into an English-speaking family in Canada at the age of 17 months. The infant was abandoned 3 days after birth and lived in an orphanage until she was picked up by the Canadian family at age 1;5. The girl was exposed to Chinese through Chinesespeaking caregivers until that age. Nicoladis and Grabois wanted to know how soon the adopted infant would lose Chinese, how soon she would learn English, and how she would adapt to the sudden language shift. The infant was followed longitudinally for two months (ages 1;6-1;8). She was videotaped during free play sessions interacting with English- and Chinese-speaking adults (five 30-minute sessions in each language approximately two to three weeks apart). Nicoladis and Grabois examined production and comprehension of words as well as prosodic features of the infant’s English productions. Figure 4.1 shows the results of lexical knowledge, calculated from production, comprehension, and imitation of words in the two languages.

122 Incomplete Acquisition in Bilingualism

Figure 4.1  Adopted Child’s vocabulary by session (adapted from Nicoladis & Grabois 2002).

While acquisition of English was fast for this child, her loss of Chinese was equally, if not even more, rapid. After six weeks in Canada, the infant no longer produced Chinese words, and after three months in Canada, she showed no understanding of Chinese words anymore. In terms of the prosodic features of Chinese, there was already prosodic influence from English in her Chinese at this young age. The child’s knowledge and development of English was age-appropriate. In summary, abrupt language shift does not seem to affect internationally adopted children in terms of their second language acquisition. Yet, when input in the L1 is interrupted, there is concomitant rapid and global language loss. We will return to this topic of language loss in international adoptees when we discuss the long-lasting effects of child L1 attrition and incomplete acquisition in adulthood in Chapter 6. Now that I have addressed in detail the structural characteristics of the weaker language and the possible circumstances under which these characteristics may come about, I must next turn to the other key theoretical question in the study of the weaker language posed in the introduction to this chapter: namely, Does the weaker language have the signatures of a first language acquired in childhood or does it develop like a second language acquired in adulthood, under the influence of different learning mechanisms?



5.

Chapter 4.  Bilingualism in early childhood 123

On the nature of the weaker language

A generalized belief in current bilingual research within a linguistic tradition is that all children are born with the capacity to learn one or more languages with ease, as amply documented by the case of simultaneous bilingual children during the age of early syntactic development (Meisel 2001, 2004, 2007; J. Paradis & Genesee 1996). At the same time, in some bilingual situations, children rarely achieve balanced development in the two languages. Indeed, in many cases, one of the languages lags behind. It is usually children who speak a minority language that is not well represented in the larger speech community who are at high risk of language attrition and eventual incomplete acquisition. There seems to be a very close relationship between language dominance, exposure to, and use of languages in young children, more so than in adults. As the studies discussed show, a young child’s bilingual balance fluctuates at the rhythm and pace of abrupt changes in input. These fluctuations are reflected in the structural properties of the languages, at several levels of linguistic analysis. This suggests that, even though the human language faculty is prepared to handle, and facilitates, monolingual and bilingual acquisition at an early age, whatever competence is acquired by the ages of 4–5 years may not be very stable, at least for some grammatical modules and interfaces. If input is reduced in one of the languages, children suffer attrition or experience incomplete acquisition, failing to reach critical levels of development and categorical knowledge in key ­grammatical areas. If input is completely interrupted – as in the case of international adoptees – the young child appears to replace her first language, quickly and inadvertently, with the language of the new speech community. This state of affairs suggests that not only do age of acquisition and level of attainment in the L1 play a role in L2 acquisition, but that they also play a substantial role in L1 attrition and incomplete acquisition. I believe these findings are theoretically related to the Critical Period Hypothesis, or to maturational constraints on language learning.

5.1

The Weaker Language as L2 Hypothesis

An interesting theoretical proposal about the nature of the weaker language and eventual incomplete acquisition in simultaneous bilingual children is mentioned, although not fully developed, by Schlyter (1993) and Schlyter and Håkansson (1994) (see §2). Based on the uneven development of French and Swedish inflectional morphology and word order (syntax), and noting that the development of the stronger language matches the achievement of monolingual children, Schlyter

124 Incomplete Acquisition in Bilingualism

Table 4.10  Learning mechanisms available to children and adults. L1 acquisition by children L2 acquisition by adults The Fundamental Difference Hypothesis (Bley-Vroman 1990)

Universal Grammar

L1 knowledge

Domain-specific linguistic Domain-general mechanisms cognitive mechanisms Neuro-cognitive approach (DeKeyser 2003; M. Paradis 2004)

Implicit learning

Implicit learning

Explicit learning

Explicit learning

claims that while the stronger language develops like a first language, the weaker language develops like a second language. But what does this mean theoretically? It is not clear what Schlyter means by a “second language” in the context of these young children, in light of the theoretical approach to L1-L2 acquisition differences she explicitly assumes. Adopting Clahsen and Muysken’s (1986), Bley-Vroman’s (1990), and Meisel’s (1997) “deficit” view of adult L2 acquisition, Schlyter (1993) claims that L2 acquisition is qualitatively and fundamentally different from L1 acquisition, as we saw in Chapter 2. Recall that for Bley-Vroman (1989) and Schachter (1990), among many others, child L1 learners rely on Universal Grammar and linguistic mechanisms to acquire their L1, whereas adult L2 learners rely on their L1 and general problem-solving cognitive skills. Table 2.8 from Chapter 2 is repeated as Table 4.10 here for convenience. Implicit in Schlyter’s assumption is the claim that children acquiring their L1 are guided by Universal Grammar, whereas adult L2 acquisition is not. Parameter setting (or resetting) is not involved, and second language systems are “rogue” grammars that violate principles of Universal Grammar (Meisel 1997). However, as I explicitly discussed in Chapter 2, the deficit, or the “fundamental difference” view of L2 acquisition refers to adults, who are allegedly not able to reset parameters because they are acquiring a language past the critical period. It is due to their age that adults must resort to other learning mechanisms, different from those primarily used by children during the period of early linguiatic development. Surprisingly, Schlyter extends this deficit view of adult L2 acquisition to children, who are all presumably still acquiring the two languages well within the span of the critical period, and who are supposedly guided by Universal Grammar. For example, Schlyter and Håkansson (1994) compare simultaneous bilingual children with sequential bilingual children who acquired Swedish between the ages of 4–5 years. At least for morphosyntax, 5 years of age would still fall within the critical period. Unlike Ionin and Wexler (2002), Schwartz (2003, 2004), and ­Unsworth (2005, 2008), among many others, Schlyter and ­Håkansson



Chapter 4.  Bilingualism in early childhood 125

(1994) seem to assume that already at age 5 child L2 acquisition must no longer be guided by Universal Grammar. But how can the weaker language of a 2 or 3 year old bilingual child be similar to the L2 grammar of an adult L2 learner assuming the Fundamental Difference Hypothesis like Schylter does? How can simultaneous bilingual children – at the same age and stage of cognitive development – use Universal Grammar (or implicit learning mechanisms) to acquire the language for which they receive more input (the majority language perhaps), while shutting off Universal Grammar and using general problem-solving cognitive mechanisms (or explicit learning mechanisms) – which are clearly not developed at this age – to acquire the language for which they receive less input? This theoretical position strikes me as entirely implausible, if not contradictory in light of critical period proposals for both L1 and L2 acquisition. Even if the weaker language lags behind in development in a simultaneous bilingual child, it could not have been acquired with different cognitive and linguistic mechanisms than the the ones used to acquire the stronger language at the same age, simply because claims about lack of access to Universal Grammar in adult L2 learning rest on late age of acquisition. But under the assumption that there is full access to Universal Grammar in L1, early bilingualism and adult L2 acquisition (Schwartz 2003, 2004; White 2003a), the weaker language in young simultaneous bilingual children can be perfectly likened to the second language of postpuberty learners. While Meisel (2007) does not criticize the questionable theoretical underpinnings of Schlyter’s proposal, he nevertheless casts doubt on the validity and tenability of the Weaker Language as L2 Hypothesis strictly on empirical grounds. For Meisel, unequivocal evidence showing that young simultaneous bilingual children fail to set the parameters of the weaker language successfully must come from syntax, and Meisel is particularly interested in Schlyter and Håkanson’s (1994) discussion of word order (V2, SV and VS) in Swedish. After examining the patterns of language use reported, along with other relevant data, Meisel reaches the conclusion that the children appear to display “delayed” acquisition of the language which they avoided speaking during an extended period of time. There is no evidence that children ever set the parameters of the weaker language to the parameter of the stronger language (suggesting language fusion at the structural level). The evidence shows that the children set the parameters of the weaker language correctly but then do not use some word orders with the same frequency as the monolingual and balanced bilingual children. Because the children still use the correct word order in the weaker language (though not all the time), in Meisel’s view, this constitutes strong evidence against the claim of wrongly set parameters or failed acquisition. While I agree with Meisel’s argument regarding acquisition of word order (what may be considered syntax proper), it cannot go unnoticed that the bilingual

126 Incomplete Acquisition in Bilingualism

children sampled in this chapter still experience significant regression and delays (as judged from high error rates) with the inflectional morphology of the weaker language. This may suggest that reduced input and use affects the morphologysyntax interface but not perhaps the syntactic module per se, and we will come back to this generalization when we discuss adults. In short, I believe that Schlyter’s Weaker Language as L2 Hypothesis is untenable on conceptual grounds, particularly given her assumptions about the fundamentally different nature of L2 acquisition. And if the available empirical evidence is also subject to other possible interpretations, then the null hypothesis is tenable. I offer instead the Weaker Language as L1 Hypothesis.

5.2

The Weaker Language as L1 Hypothesis

The Weaker Language as L1 Hypothesis Even when it may lag behind in development due to insufficient exposure to input and use, the weaker language in simultaneous bilingual acquisition is acquired as a first language, through the same cognitive and linguistic means used to acquire the stronger language available in early childhood. If age determines the way children and adults approach the language learning process, I assume that simultaneous bilingual children, like L1 acquiring children, unquestionably use Universal Grammar and implicit learning mechanisms to acquire the two languages. Quantitative differences between the weaker language of simultaneous bilinguals and the language of monolingual and balanced bilinguals arise from reduced input and opportunities of language use. They can not possibly arise from inability to deploy the relevant language learning mechanisms, as may be the case in adult L2 acquisition. Even if the weaker language turns out to be a simplified and reanalyzed version of what it could have been had it reached its full development, I will show in Chapters 6 and 7 that the adult weaker language still retains the signatures of an L1 acquired early in childhood. In this context, there are two other points raised by Meisel (2007) that deserve further consideration. The first is that the weaker language in very young simultaneous bilingual children (younger than 4 years), shows delayed development. Unfortunately, most available studies of the weaker language in simultaneous bilingual children stop observations before age 4 or 5, so not much can be inferred about further development of the two languages. But is delayed development incomplete acquisition or attrition at this young age, as I have characterized it in this chapter? Although the studies conducted by Anderson and Silva-Corvalán suggest as much, perhaps following young children for two or three years is not



Chapter 4.  Bilingualism in early childhood 127

really sufficient time to answer this question with any certainty. The question can be resolved if we look at simultaneous bilingual children older than age 4 and into adulthood. The data I will present and discuss in the next three chapters suggest that initial delayed development in the weaker language of some simultaneous bilinguals is not temporary but permanent, and easily characterized as incomplete acquisition in adulthood. The second point Meisel makes is that these simultaneous bilingual children have grammatical competence in the two languages but do not exercise all the grammatical possibilities in the weaker language. Is it linguistic competence at the level of representations that is affected, or simply difficulty putting the available linguistic knowledge into use (i.e., a retrieval and assembly problem during production)? After all, all the case studies discussed in this chapter focused on the spontaneous oral production of very few children, which is the most common methodology used in the study of simultaneous bilingual acquisition. Similar to what happens in monolingual acquisition, it is often problematic to make claims about linguistic competence based solely on patterns of language use. Ideally, experimental studies with higher numbers of subjects and using comprehension measures as well as production would be better suited to understand the extent of language attrition, incomplete acquisition, or simply delay at the level of linguistic competence. Taking into account what we know about language acquisition and loss in childhood, my position regarding language development in young bilingual children is the following: Universal Grammar provides the initial blueprint for acquisition, but input shapes the extent of development. Obviously, early exposure to the two languages is not enough for full linguistic development. Sustained input contributes significantly to the molding of linguistic representations at this early age, until children reach age-appropriate levels of linguistic achievement. Perhaps the close relationship between knowledge of language and use of language as a function of input is easier to see in bilingual acquisition than in monolingual acquisition, since input in bilingual children is the variable factor. Comparison of bilingual children who are continuously exposed to input, as opposed to children who are completely cut off from it (adoptees), suggests that both early age of onset and a minimum threshold level of input are required to acquire and maintain a critical mass of language, that eventually leads to full linguistic ability. The fact that international adoptees can, apparently, completely replace the L1 with the L2 suggests that the underlying cognitive and linguistic mechanisms for language learning are there, but it is the input received – during a maturationally determined critical time – that determines which language will be best acquired and eventually maintained.

128 Incomplete Acquisition in Bilingualism

Table 4.11  Effects of attrition by grammatical domain and age. Grammatical domain

L2 acquisition Adult Attrition Child Attrition

Grammar and internal interfaces Uninterpretable formal features (e.g., agreement, gender, case, etc.)

affected

not affected

affected

External interfaces Interpretable semantic features (e.g. topic, specificity, etc.)

affected

mildly affected

affected

Ideally, the Weaker Language Hypothesis (as L1 or L2) can only be empirically testable if it states exactly which areas of the grammar are likely to be affected. One possibility is to extend the linguistic feature-based proposal for L1 attrition and L2 fossilization discussed in the previous chapter to cases of unbalanced bilingualism. The generalization emerging from the emprical evidence discussed in Chapter 2 was that fossilization in adult L2 acquisition appears restricted to uninterpretable formal features at the morphology-syntax interface (an internal interface) as well as the syntax-discourse interface (an external interface). L1 attrition in adults is mostly restricted to interpretable features regulating the syntax-discourse interface. The data from attrition and delayed/incomplete acquisition in children presented in this chapter suggests that, by contrast, both types – interpretable and uninterpretable features – can be affected in childhood when insufficient input is available at a critical time in language development. In that sense, the weaker language in simultaneous bilingual children resembles an L2. This proposal is summarized in Table 4.11. If similar domains are affected in adult L2 acquisition and very early child attrition/delayed acquisition, this supports the Weaker Language as L2 Hypothesis. Under the view that L1 and L2 acquisition, and simultaneous bilingualism, are fundamentally similar and constrained by Universal Grammar (Montrul 2004a, White 2003a), there is no contradiction. If, like Meisel (1997, 2007) and Schlyter (1993), we assume that L1 and L2 acquisition are fundamentally distinct, then we are forced to claim that bilingual L1 acquisition is not governed by Universal Grammar, which seems unlikely. In conclusion, the questions of whether delayed acquisition and attrition in simultaneous bilinguals lead to long-term incomplete acquisition, and whether permanent incomplete L1 acquisition is like L2 adult acquisition, cannot be answered satisfactorily by looking at the available empirical evidence in a few case studies of very young children in the process of development. But I believe these two questions can be addressed in at least two possible ways. One way is by looking at how bilingual children continue to develop beyond the stage of early



Chapter 4.  Bilingualism in early childhood 129

syntactic development during the school years and into adulthood, as I do in Chapters 5 and 6. The other is by comparing the weaker language in adult early bilinguals with developing L2 adult grammars at the same level of proficiency, as I do in Chapter 7. Only the latter comparison will allow us to begin to answer the question of whether the weaker language in simultaneous bilinguals is fundamentally different from an L1 and fundamentally similar to an L2 acquired past the critical period.

chapter 5

Bilingualism in middle and late childhood

The previous chapter discussed the grammatical development of both simultaneous bilingual children, and of children who are still in the process of acquiring their L1 when exposure to the L2 begins. As a result of receiving early input in the L2, the minority language can lag behind and develop incompletely. Attrition and incomplete acquisition in childhood affect vocabulary, morphosyntax, and the syntax-discourse interface, appearing to have deeper consequences for grammatical competence than attrition in adults. Furthermore, degree of attrition/incomplete acquisition is related to age: the younger the child when use of the minority language is reduced, the more severe the effects of attrition and incomplete acquisition appear to be. As predicted by the hypothesis formulated at the end of Chapter 2, the minority language as weaker language is more affected in simultaneous bilinguals with less exposure to the L1, than in sequential bilingual children who acquired the majority language as an L2 after age 4 or 5. Because one of the question posed in Chapter 4 – whether delayed acquisition and attrition in very young simultaneous bilinguals leads to long-term incomplete acquisition – cannot be properly answered when case studies focus on a very short period in childhood, in this chapter I continue to trace the weaker language and degree of L1 loss during the school age period. I focus specifically on the impact of the school environment and literacy on the grammatical development of the minority language in simultaneous and sequential bilingual children, both those born in the majority-language speaking country and those who immigrate in middle to late childhood. The main claim I make in this chapter is that empirical evidence supports the notion of a sensitive, perhaps even critical, period for L1 loss in bilingual children: vulnerability to L1 attrition in L2-acquiring children gradually decreases with age, tapering off at around age 8–10. After the age of 10 years, language loss is unlikely, especially in minority language-speaking children who start L2 acquisition at this age and who are still exposed to the minority language.

132 Incomplete Acquisition in Bilingualism

1.

Language learning at school

1.1

Monolingual children

Even though linguists consider that the basis of native speaker competence is acquired between the ages of 3 and 4, children’s knowledge of language in all areas continues to develop and goes through significant changes after the age of 4, while, at the same time, children undergo further cognitive, social, and emotional growth. Children experience increasing development of more complex vocabulary and structures, as well as acquisition of pragmatic knowledge and grammatically expressed social conventions of the language (e.g., the use of honorifics in Asian languages and different forms of address in Spanish). At about the age of 4, metalinguistic ability – the ability to talk about and analyze language – also emerges, and is further developed in literate children. At school, children develop phonological awareness, which is a prerequisite for reading alphabetic writing systems, narrative skills, and knowledge of textual conventions. In comparing the speech of a 6 year old and a 8 year old, Nippold (1998, p. 2) notes that both children already speak in long and complex sentences, produce clear articulation, use a variety of words, and contribute actively to a conversation. Yet, when linguistic markers are analyzed more closely, especially in their writing, it becomes clear that the 8 year old is far ahead of the 6 year old in language development, as judged by MLU (a measure of morphosyntaxtic complexity) (Loban 1976) and number of words per communicative unit (a measure of sentence length) (­Leadholm & Miller 1992). For most school-age children and adolescents, in addition to spoken input, written communication plays an increasingly important role. Once children

. Gombert (1992) argued that behavior that has been interpreted as metalinguistic functioning before the age of 6 to 7 years is merely “epilinguistic,” that is, not based on systematically represented knowledge which can be intentionally applied (Gombert 1992, p. 9). By contrast, Doherty and Perner (1998) argue that metalinguistic awareness arises at the same time children begin to succeed on the false belief task, roughly at the age of 4 years. False belief tasks require understanding that language is a formal system carrying meaning and the ability to intentionally apply systematically represented knowledge. These tasks must be based on simple, basic, and familiar formal aspects of language in order to be sure that children do not fail for reasons other than lacking the competence the task aims to assess. For instance, de Villiers and de Villiers (1972, 1974) and Smith and Tager-Flusberg (1982) examined children’s ability to identify ungrammatical utterances as odd or silly. The results suggest that children begin to succeed on these tasks around the age of 4 years. Smith and Tager-Flusberg (1982) found that while 78% of 4 year olds successfully judged the grammaticality of sentences, only 22% of 3 year olds could do so.



Chapter 5.  Bilingualism in middle and late childhood 133

­ egin to read, their language development becomes more individualized, guidb ed more by personal interests. This is most obvious in vocabulary development. Children interested in animals will read more and learn words for a variety of animals, whereas children interested in sports will have a larger vocabulary in that particular semantic field. Not only are metalinguistic awareness and increased ability to think abstractly reflected in children’s vocabularies, but they are also reflected in children’s use and understanding of complex and less frequent syntactic structures like relative clauses, conditionals, and counterfactuals, among others. In short, during the early years of elementary school, children’s automaticity, fluency and structural development of language increase. At the same time, children’s pragmatic awareness and use of language become more complex. Representative changes in children’s language outlined by Menyuk and Brisk (2005) are illustrated in Table 5.1. One of the main functions of school is to teach children how to read and write and to turn them into mature and literate speakers/writers. Not only do reading and writing contribute to the reinforcement and expansion of children’s lexical Table 5.1  Structural and pragmatic development in 6–8 year old children (adapted from Menyuk and Brisk 2005). Category Grammar

Pragmatics

Change Syntax

Sentence length increases (e.g. I see the boy who I played with yesterday) Combining structures becomes more frequent (through complementation, conjunction, subordination) (e.g., She likes me to do homework before watching television) Morphology Prefixing and suffixing increases (e.g. unhappiness, disapprove, discussion) Lexicon Use of abstract categories increases (e.g., liberty, vast, imagination) Synonyms and antonyms used more widely (e.g., large, big, huge, small, little, minute) Multisyllabic words appear more frequently (e.g., disappointment, unhappiness) Phonology Stress rules of language acquired (e.g., history, historical, influence, influential) Morpho-phonological rules (e.g., a car, an apple) Conversation Begin to take perspective of others Begin to make relevant responses Storytelling Begin to be listener friendly Begin to follow story grammar Explanation Begin to move from personal reference to abstract knowledge

134 Incomplete Acquisition in Bilingualism

and grammatical growth, but these activities also make children aware of different uses and functions of language, for example, the difference between academic, non-academic, and prestige varieties of the language. Schools typically teach the standard variety of the language, and children who speak other (vernacular) dialects become aware of, and must learn, the standard variety through schooling. If the written, standard variety also includes grammatical structures and vocabulary that are not typical in the spoken varieties, it is obvious that native speaker knowledge can hardly be considered complete by the age of 3–4 years as is often assumed in language acquisition literature. Contrary to this, it is suggested in educational literature that complete mastery of phonology, productive control of most syntactic structures, and early literacy, are achieved at about age 8 (Smit, Hand, Freilinger, Bernthal & Bird 1990; Snow, Burns & Griffin 1998). Even if few studies have documented it, a substantial amount of linguistic knowledge and the fixation of that knowledge occur until adolescence (Braine, Brooks, Cowan, ­Samuels & Tamis-LeMonda 1993; Jia & Aaronson 2003). For instance, Braine et al. (1993) examined age-related changes in the acquisition of the syntactic category (as opposed to the semantic category) subject in English transitive and intransitive verbs, finding that 4 and 5 year olds did not show adult-level performance on a metalinguistic task. Only the 9 year olds performed at the level of the adult comparison group and showed evidence of having knowledge of subjects as a syntactic category. Other studies show that typical child language grammatical errors do not go away until children go to school. It must be recalled from Chapter 3 (L1 attrition in adults) that Keijzer (2007) tested a group of 45 13 year olds (the L1 acquisition group) in her study of attrition of Dutch, which included tests of nominal morphology, verbal morphology, negation, passives, V2 and subordination. Keijzer (2007) found that the control group of adult Dutch speakers and the teenagers differed statistically in a number of structures, including plural -s, neuter article selection, adjectival inflection, diminutive formation, irregular present tense, and some instances of subordination. Based on these results, she claimed that “the largest part of native language acquisition is complete by early adolescence, but there continue to be areas characterized by residual optionality, where even 13 and 14 year olds do not perform at the same level of mature native speakers” (­Keijzer 2007, p. 284). It is true that this trend of later language development of morphology and complex syntax is not generally reported in the L1 acquisition literature. Similar to this, Jia and Aaranson (2003) showed that on a morphosyntactic task, monolingual Mandarin speaking children in China did not ­perform at

. This is true in a monolingual setting as well.



Chapter 5.  Bilingualism in middle and late childhood 135

adult native speaker level until they reached the age of 16, concluding that ­Chinese speakers in their study who immigrated to the United States before that age had incomplete knowledge of their L1 when English L2 acquisition began. One recent exception in the L1 acquisition literature indirectly addressing the potential role of literacy, however, is Guasti and Cardinaletti’s (2003) study of relative clauses. Guasti and Cardinaletti examined Italian and French-speaking monolingual children’s production of subject relative clauses in an elicited production task, especially with respect to the absence of prepositional pied-piping relatives and the use of resumptive pronouns (e.g., Il ragazzo a cui ho dato il libro “The child to whom I gave the book” vs. Il ragazzo che gli ho dato il libro “The child that to him I gave the book”). The children ranged in ages from 4;5 to 10 years. Guasti and Cardinaletti found that children’s relative clauses are syntactically derived in the same way as adults’ (i.e., children posit the same wh-movement analysis or by base generation of a null operator). However, children do not use the same types of relative clauses as adults until the relative pronouns (Italian cui, French laquelle) are learned at school. Only the children older than 7 years of age produced these structures correctly. Because pied-piping relatives are typical of the more formal variety of the language, Guasti and Cardinaletti concluded that children learn relative pronouns and thus pied-piping relatives only during the school years, through explicit teaching. Perhaps a better example of how literacy impinges on later language development comes from a recent study of Brazilian Portuguese. Pires and Rothman (in press) investigated the acquisition of inflected infinitives in by monolingual school age children. While inflected infinitives are present in the standard written Brazilian Portuguese and in European Portuguese, inflected infinitives are being replaced by other structures in colloquial, spoken varieties of Brazilian Portuguese. Pires and Rothman tested 87 upper income Brazilian Portuguese children and adolescents ranging from ages 6–15 on a morphological recognition task and a context match task. The results showed that the 13–15 year old children had acquired both the morphology and semantic interpretation of these forms, but all the children younger than 9 had not. The 10–12 age group was on their way to acquisition, since they were statistically better than the younger groups but still less accurate and different from the 13–15 year olds on both tasks. According to Pires and Rothman, inflected infinitives are not acquired until ages 10–12 (fourth grade), which coincides with the age when Brazilian Portuguese children are instructed on verbal paradigms at school, and receive more exposure to the written and oral standard variety. Hence, both Guasti and Cardinaletti (2003) and Pires and Rothman (in press) show that complex aspects of syntax, morphology, and semantics, like conditionals, counterfactuals, relative clauses and other types of subordination are not fully

136 Incomplete Acquisition in Bilingualism

mastered by monolingual children until children go to school and receive explicit teaching and exposure to the written register. But how does literacy acquisition affect the linguistic development of preschool and school age bilingual children, especially when instruction takes place in their second language?

1.2

Bilingual children

As simultaneous and sequential bilingual children acquire basic proficiency and literacy in the L2, their use of and proficiency in the L1 and L2 change over time. When the minority language is not fully supported at school, it runs the risk of being lost. Research on dominant language shift indicates that when loss occurs, it takes place more gradually at this age than during the age of early syntactic development discussed in Chapter 4. Furthermore, as I will show, there is a threshold for vulnerability to language loss in sequential bilinguals. Minority-speaking children younger than 10 years of age show a more rapid shift to the L2 and a larger degree of L1 loss than children older than 10. This gradual decline in vulnerability to loss is consistent with the Critical Period Hypothesis for L1 acquisition and loss. The ability of language minority children to acquire the L2 and maintain proficiency in their L1 is in part related to the type of support the minority language receives in the school environment. At least in the United States, the predominant educational model does not help language maintenance, unfortunately. For many years, emphasis has been placed on having immigrant children acquire English as rapidly as possible. Currently, three models of education are available to some bilingual children in the United States, mostly those of Spanish-speaking origin. The English immersion system places immigrant children and heritage speakers in English-speaking schools, and typically does not offer support for the other language. Transitional bilingual programs are available for some languages but not for others. Many of these programs offer language instruction in the heritage language together with English during the first two years of elementary school. In third grade, bilingual children are mainstreamed into English-only classes. The goal of . Köpke and Schmid (2003) have suggested that changes in vulnerability to L1 attrition emerge at around age 9, and it appears that schooling and literacy may have a lot to do with it. Schooling and literacy somehow reinforce knowledge of that native language by providing written input, exposing children to other structures not frequent in oral language, expanding and reinforcing children’s vocabulary, and teaching the written code. From an affective perspective, literacy raises awareness about language prestige. These factors may contribute to the hypothesized “stability” of linguistic competence acquired before a critical period and its resilience to loss later in life.



Chapter 5.  Bilingualism in middle and late childhood 137

transitional bilingual schools is to make use of instruction in the heritage language temporarily to ease the transition to an English-only curriculum. Both English immersion and transitional bilingual schools promote English monolingualism. The third type of educational model is the double immersion, or two-way bilingual, school. In these schools, 40% of the time and curriculum are offered in another language, most commonly Spanish, and 60% in English until the fifth grade. English-speaking children from English-only families learn Spanish as a second language, whereas simultaneous and sequential Spanish-English bilingual children are offered the opportunity to develop and maintain both Spanish and English. Despite the limited availability of these models in some states, to speed up the process of L2 acquisition, early education in the majority language has been identified as the ideal solution to solve the academic problems of immigrant children. There is no doubt that becoming part of the L2 society is a priority of immigrant children and their parents. Indeed, children have a strong motivation to fit in. However, Wong-Fillmore (1991) strongly articulates the dangers of starting L2 instruction as early as 3 or 4 years of age. Wong-Fillmore (1991) reports on a national large-scale survey of 1,100 Spanish, Korean, Japanese, Chinese, Khmer and Vietnamese families whose children had attended preschool programs in the United States. All these families formed the main sample (67.8% of the families were Latinos and 13. 6% were East Asian). A group of 311 Spanish-speaking families who had not sent their children to English pre-school programs served as baseline or control group. The purpose of the study was to determine the extent to which early total or partial English instruction was affecting children’s language patterns in the family language. The data was based exclusively on parental interviews. All the participating parents were asked, among many other things, whether they perceived a change in patterns of language use in the family once children started daycare, and if so, whether the change was positive (more heritage language use), negative (less heritage language use, more English), no change, or neutral (less English). (Positive and negative changes were defined by the researchers and not used as such in the surveys but added later for data analysis.) Wong-Fillmore showed that 50.6% of the main sample reported a negative change, or a shift from the home language to English, as opposed to 10.8% of the families who reported a negative change in the comparison group. Responses according to language of early education also showed interesting differences between the two groups, as illustrated in Table 5.2. Parents whose children attended English-only and bilingual programs reported more “negative” changes than parents of children in the comparison sample. By contrast, parents in the comparison sample reported more “positive” changes than parents in the main sample, especially for bilingual and heritage language programs. The parents also rated their children as having less knowledge of the

138 Incomplete Acquisition in Bilingualism

Table 5.2  Changes in language use depending on language of schooling (adapted from Wong-Fillmore 1991).

No change Negative change Positive change Neutral

Main Sample

Comparison Sample

English only 31.3% 64.4%   2.8%   1.4%

Bilingual Heritage Language only 34.4% 16% 12.5% 10.3% 46.9% 72.8% 6.2%   0.8%

Bilingual Heritage Language only 34.4% 31.6% 47.2% 26.3% 18.6% 42.1%   3.7%

heritage language than monolingual children of the same age. Important differences were also found between older and younger siblings in the families: younger children, both in the main and comparison groups, showed greater loss than the older children in the families. This is because younger siblings are exposed to English earlier, through their older siblings. Wong-Fillmore concluded from this study that bilingual education does not offer bilingual children protection against language shift and loss. The younger the immigrant children when they learn English, the more minority language loss there was. Before age 3, Wong-Fillmore asserts, children have not yet reached a stable command of their native language, and as a result this knowledge is likely to be lost. Although the younger children are when they start their L2 acquisition, the higher their chances of attaining full linguistic competence in the L2, WongFillmore still warned that L2 acquisition at such a young age may also result in L1 attrition, especially when the foundations of the L1 are not fully in place. In Wong-Fillmore’s opinion, learning a second language by itself is not harmful, but timing is crucial. The methods, findings, and rhetoric of the Wong-Fillmore study (1991) have not gone unchallenged. With respect to the methodology used, many of the language groups included in Wong-Fillmore’s main sample are not spoken in the community, whereas Spanish – the comparison group – is linguistically supported by the large number of people who speak the language in the United States. So, it is possible that Spanish-speaking children receive language support beyond the family and are thus more likely than other linguistic groups to maintain language skills. Another major problem with the study is that language loss was exclusively based on parental perceptions, and there was no independent measure of linguistic proficiency used to show that children indeed underwent language loss in the heritage language. Rodríguez, Díaz, Durán and Espinosa (1995) conducted a follow-up study of Wong-Fillmore’s original study, by focusing exclusively on Spanish-speaking



Chapter 5.  Bilingualism in middle and late childhood 139

children and families. They also used comprehension, production, and vocabulary measures of proficiency and language complexity in English and Spanish to quantify potential changes in linguistic competence in the two languages and concomitant language loss in Spanish. The researchers also conducted class observations of the children at school. Some of the children had attended English pre-school while others had not. They found that early onset of L2 acquisition was not as dramatic as Wong-Fillmore stated, and did not affect L1 maintenance. The English-preschool group did not perform worse in Spanish than the stay-at-home group. Indeed, many children maintained, and many others even improved, their Spanish proficiency. In both stay-at home and preschool children, proficiency in English increased progressively, and gains were made more rapidly and sooner in comprehension than in production, but only children enrolled in the bilingual programs showed increased proficiency in both Spanish and English. The studies by Wong-Fillmore (1991) and by Rodríguez et al. (1995) underscore how early L2 education impinges on language use and cultural loss of the minority language, even when they do not directly address how L2 acquisition and language shift to the majority language affect specific aspects of the L1 grammatical system. Although these studies did not document language loss directly, I will now focus on evidence from other studies that older Spanish-speaking, and other ethnic minority children, undergo a measurable level of language loss during the school age period.

2.

L1 attrition in minority-language speaking children

In this section, I discuss studies showing overall proficiency changes as well as specific grammatical deficits in school-age minority-language speaking children. I show that some of the same patterns of attrition and incomplete acquisition of morphosyntax exhibited by pre-school children in Chapter 4, also manifest themselves with older children. Language loss with this age group is not just a matter of lexical retrieval and processing, as in attrition in adults (Chapter 3), but it affects the grammar and its interfaces (including interpretable and uninterpretable features, as shown in Table 4.11, Chapter 4). An advantage of investigating language development in school-age children is that it is possible to use research methodologies other than spontaneous production data. Thus, the extent of incomplete acquisition and attrition can be inferred on other sources of evidence, including oral and written language. To illustrate the extent of attrition and incomplete acquisition in this age group, I focus on different minority-language speaking groups receiving different types of academic support for the heritage language.

140 Incomplete Acquisition in Bilingualism

2.1

Inuktitut-speaking children in Eastern Canada

Inuktitut is an aboriginal language (Eskimo-Aleut family) spoken in Alaska, Siberia, Labrador, Greenland and Eastern Canada. While the language has declined significantly in the westernmost regions, Allen (2007) reports that a large number of Canadian Inuit (above 70%) still learn and use their native language. Indeed, several studies have attested the fluent acquisition of Inuktitut by pre-school children (Allen 1996; Crago & Allen 1998; Swift 2004) and the growth of the language into the early school years (Allen, Crago & Pesco 2006; Crago, Annahatak, ­Doehring & Allen 1991). Inuktitut is presently taught in the schools until Grade 2. In Grades 3 and 4, English and French – the two official majority languages in Canada – are introduced in the school system. Inuktitut instruction may continue or may not in some content areas, depending on the availability of qualified teachers. As I stated in Hypothesis 2 formulated in the previous chapter, simultaneous bilingual children are more vulnerable to language shift and loss than sequential bilingual children who are language-minority dominant until they go to school. Simultaneous bilinguals have the longest exposure to the majority language and receive less input in the minority language. Allen (2007) reviews several studies of very young simultaneous Inuktitut-English speaking children and shows that the bilingual children do not show delays or declining command of Inuktitut. On the contrary, these very young bilingual children (1;8–2;11 years) possess age-appropriate native-like knowledge of the parametric properties of both English and Inuktitut, two languages very different typologically. What happens once children start exposure to English or French at school, especially when the children were predominantly monolingual speakers of Inuktitut before English/French instruction began? Wright, Taylor and Macarthur (2000) followed a group of 62 Inuit children for four years, since kindergarten. Some children received instruction in Inuktitut, some in English, and some in French. The children completed a battery of tests of language ability in the three languages at the beginning and at the end of each academic year. All tests were averaged to obtain a composite proficiency score. Table 5.3 summarizes the general proficiency scores in Inuktitut for all the children, depending on language of instruction (Inuktitut, English, French). Table 5.3 shows that at the time of first testing (Kindergarten, fall), the three groups of children did not differ in their overall proficiency in Inuktitut, which ranged between 36.6% and 39.04%. However, from the spring of kindergarten on, there are significant differences in Inuktitut proficiency scores for the Inuit children receiving instruction in English and French in comparison to the children who continued to receive instruction in Inuktitut, reaching more than 20% difference by the spring of Grade 2. The results of this study, as well as those of



Chapter 5.  Bilingualism in middle and late childhood 141

Table 5.3  Proficiency scores (in percentages) in Inuktitut for Inuit children receiving language instruction in Inuktitut, English or French (adapted from Wright, Taylor & Macarthur 2000). Language of Instruction

N

Kindergarten

Grade 1

Inuktitut English French

31 14 17

Fall 39.04 37.38 36.56

Fall 57.55 46.25 46.84

Spring 58.53 48.12 48.20

Grade 2 Spring 72.05 56.13 57.54

Fall 75.70 55.41 56.30

Spring 82.99 60.14 65.18

Allen, Crago and Pesco (2006), suggest that there is native language decline, slow development, and even stagnation in Inuit-speaking children as a result of school or community exposure to an L2 that is a majority language.

2.2

Spanish-speaking children in American schools

The patterns of attrition and incomplete acquisition of inflectional morphology discussed with young Spanish-English bilingual children in Chapter 4 manifest themselves in older bilingual children as well. One of the earliest research projects documenting interrupted acquisition and loss of morphosyntax in the minority language in school-age bilingual children is Merino (1983). Merino (1983) reports the results of two studies on language loss in Spanish-speaking children of Mexican origin, attending English school, with no support for Spanish. The first study followed 41 bilingual children ranging from kindergarten to fourth grade (5–10 years old), and measured degree of language loss in comprehension and production in a variety of morphosyntactic features of Spanish (gender and number, tense, word order, relative clauses, conditional and subjunctive) and their equivalents in English. Results showed steady chronological development between kindergarten and the upper grades in English (both in comprehension and production), but a significant decline in Spanish comprehension in third and fourth grade. These findings are summarized in Table 5.4. Table 5.4 also shows that performance in the older children dropped dramatically in Spanish production (from 84% in first grade to 65% in fourth grade), while comprehension remained relatively stable. With respect to the structures most affected in production, the children had significant difficulty with the subjunctive . However, despite declining proficiency with age in the Inuit population as interaction in the majority language starts, Allen (2007) reports active bilingualism and high linguistic vitality of Inuktitut in this area.

142 Incomplete Acquisition in Bilingualism

Table 5.4  English L2 acquisition and Spanish L1 decline in Mexican-American children (adapted from Merino 1983). Production Grade K 1 2 3 4

N  9  4  9 10  9

Spanish 56% 84% 71% 77% 65%

Comprehension English 52% 86% 75% 81% 86%

Spanish 73% 86% 84% 75% 80%

English 76% 89% 85% 86% 88%

Table 5.5  Spanish production accuracy by grammatical category (adapted from Merino 1983).

Past tense Relative clauses Subjunctive

Time 1

Time 2

  87% 100%   70%

74% 44% 55%

and the conditional verb forms, the most complex forms. With the subjunctive in particular, the fourth graders performed at the level of the kindergarteners, and with the past tense, the fourth graders performed below the level of first graders. A second study was conducted two years later with 32 children of the original sample. The children were again administered the Spanish and English production and comprehension tests. The results showed that performance in English continued to improve for all the Spanish-speaking children, while performance in Spanish deteriorated dramatically: 50% of the children showed loss of some sort, while another 25% did not show any progress. The locus of the differences between the first and second test administrations was found in the past tense, the subjunctive, and relative clauses, while performance on the conditional remained fossilized (no development). Merino suggests that this pattern of language loss is consistent with the Regression Hypothesis (see Chapter 3), even when the Spanishspeaking children did not have full mastery of these structures two years earlier. This is shown in Table 5.5. To summarize, this study attested significant loss in the subjunctive and relative clauses, especially in children who attend transitional bilingual and Englishonly programs. A look at the personal history of the children revealed that use

. But no quantification of this lack of development between the first and second administration of the tests is given See Merino (1983, pp. 288–9).



Chapter 5.  Bilingualism in middle and late childhood 143

of language at home in early childhood was a strong predictor of language maintenance and loss. Children who used Spanish only with parents, relatives, and friends had stronger Spanish language skills. Children who tended to use both languages with their parents, relatives, and friends demonstrated the greatest loss. Thus, use of English in the home, and lack of academic support at school, contribute to language shift and eventual language loss. The bilingual children studied by Merino, like the vast majority of Spanishspeaking children in the United States, were of low socioeconomic status (SES). SES has been shown to be an important predictor of literacy and academic performance of elementary school children, independent of bilingualism. Degree of language loss in these children may also be linked to SES. In the Miami area, however, Spanish enjoys a different prestige because many of the first Cuban immigrants were from middle and high socioeconomic status. Thus, Spanish is used and maintained by Spanish speakers from a variety of SES. In a large scale study of Spanish-English bilingual children from the MiamiDade area, Mueller Gathercole (2002a, b, c) investigated the effects of the school environment, language spoken at home, and SES in bilingual children’s performance in morphosyntactic aspects of Spanish and English. The research design included simultaneous bilinguals, who typically use Spanish and English at home, and sequential bilinguals, who tend to speak only Spanish at home. The research design also permitted the investigation of the three current models for bilingual instruction in the United States: English-only, transitional bilingual, and dual-immersion (two-way bilingual) programs. As part of a larger research project (see details in Oller & Eilers 2002), Mueller Gathercole tested 294 second and fifth grade English monolingual and SpanishEnglish bilingual children from high and low SES enrolled in English or dual immersion bilingual programs. The control group for the Spanish parts of the study consisted of 32 monolingual Peruvian children (matched for age and SES). Three morphosyntactic properties of Spanish and English were tested in three experiments: the mass/count distinction in English nouns, gender agreement in Spanish nouns, and the that-trace phenomenon in Spanish and English. In the three experiments, the children were asked to give grammaticality judgments of correct and incorrect oral sentences spoken by a puppet. If sentences uttered by the puppet sounded wrong, children were asked to tell the puppet how to say it correctly. Because the focus of this chapter is on potential L1 loss in these children, I will only focus on the Spanish experiments (experiments 2 and 3). Experiment 2 (Mueller Gathercole 2002b) investigated knowledge of gender agreement in Spanish noun phrases in the same children (their L1). While Spanish does not make a grammatical distinction between mass and count nouns, it does classify words by gender. Prototypical masculine nouns end in -o (el niño

144 Incomplete Acquisition in Bilingualism

“the boy”, el libro “the book”), while prototypical feminine nouns end in -a (la niñar “the girl”, la mesa “the table”), although there are several exceptions to this central tendency of the language. Mueller Gathercole focused on exception cases, such as words ending in -e (la calle “the street,” el valle “the valley”), feminine-o words (la mano “the hand”, la modelo “the model”) and masculine -a words (el mapa “the map,” el problema “the problem”). The task consisted of two sets of 8 sentences (4 grammatical and 4 ungrammatical). The results of the Spanish gender experiment showed that the children performed more accurately with grammatical sentences than with ungrammatical sentences; the monolingual children were more accurate than the bilingual children; and the fifth graders were more accurate than the second graders on ungrammatical sentences. Performance on ungrammatical sentences varied according to language spoken at home: the children who spoke Spanish only at home did better than the children who spoke Spanish and English or only English at home. There was no effect for SES in Spanish (although SES was a factor in the experiment on the mass/count distinction in English). The results also showed that children attending two-way bilingual schools performed better than the children schooled in English-immersion, especially in the fifth grade. In conclusion, while many Spanish-English bilingual children lag behind their Spanish monolingual peers in the acquisition of gender in Spanish, bilingual children who receive the greatest amount of input in Spanish – at school and at home – have an advantage over bilinguals exposed to English and less Spanish. Montrul and Potowski (2007) further confirmed that a dual immersion school environment is conducive for language maintenance in Spanish-speaking children. Their study also tested command of gender agreement in common, canonical-ending nouns in two oral production tasks. Although simultaneous bilingual children produced more gender errors with feminine words (60%) in the elicited production task than sequential bilingual children (36%) and than monolingual children (0%), the error rate did not increase with age, as the study by Merino (1983) showed. Overall, the children did not necessarily improve, but maintained the same level of Spanish proficiency in the language, at least in the area of gender agreement between nouns and adjectives, as shown in Table 5.6. Taking into account age of onset of bilingualism, the older simultaneous bilingual children (ages 9–11) made more errors than the younger simultaneous bilingual children (ages 6–8), although the difference between the two groups was not statistically significant. By contrast, the older sequential bilingual children (ages 9–11) were overall more accurate than the younger sequential bilingual children (ages 6–8). If morphology is affected to some extent in later childhood bilingualism, are areas of syntax equally affected? Mueller Gathercole (2002c) investigated the acquisition of the that-trace phenomena in Spanish and English. In English, when



Chapter 5.  Bilingualism in middle and late childhood 145

Table 5.6  Percentage accuracy on gender agreement between nouns and adjectives in an elicited production task (adapted from Montrul & Potowski 2007).

masculine feminine overall

Monolinguals

Simultaneous bilinguals

Sequential bilinguals

younger 100 100 100

younger 98.5 70.7 84.6

younger 95 38 66.5

older 100   98   99

older 93.2 61.8 77.5

older 98 70 84

forming questions, subjects cannot be extracted past a complementizer, as in (1a), whereas in Spanish, extraction from subjects is perfectly acceptable, as in (2a). In English, complementizers are optional, as in (1b), whereas in Spanish they are required, as in (2b). (1) a. *Who do you think that [t] has green eyes? b.   Who do you think [t] has green eyes? (2) a. ¿Quién crees que [t] tiene ojos verdes? b. *¿Quién crees [t] tiene ojos verdes?

Early accounts of this phenomenon in generative grammar were referred to as the that-trace filter (Chomsky & Lasnik 1977). The that-trace effect was first assumed to be part of the cluster of properties associated with the null subject parameter (Chomsky 1981; Jaeggli 1982), and later with the Empty Category Principle (ECP) (Chomsky 1981). Work by Thornton (1990) with pre-school English-speaking children suggests that, overall, children know about these constraints in English. Following this claim, Mueller Gathercole hypothesized that if the that-trace effect was part of innate linguistic knowledge, it should be well established in the two languages of bilingual children. The children were asked to judge orally 4 grammatical and 4 ungrammatical sentences in Spanish and in English (8 in each language), and to provide corrections, following Thornton’s (1990) methodology. The bilingual children were administered the two tests (in Spanish and English), while the monolinguals were given the test in their native language. Results of Experiment 3, the that-trace phenomenon, showed that the Spanish monolingual children from Peru were more accurate at rejecting ungrammatical sentences than the English monolingual children from the US, but the two groups were comparatively correct on grammatical sentences. The results of the bilinguals showed that, as with the gender sentences, the children were more accurate on grammatical than on ungrammatical sentences. In terms of age, the fifth graders were more accurate than the second graders, and with respect to language, accuracy in Spanish was higher than accuracy in English, following the monolingual pattern. The children who were in the English-immersion school performed better

146 Incomplete Acquisition in Bilingualism

on the English sentences than on the Spanish sentences, whereas the effect was reversed for the children attending the Two-way bilingual school: performance in Spanish was statistically higher than on English. When correction results were examined, the monolingual Spanish speakers corrected more sentences (2.5 out of 4) than the monolingual English speakers (1.44 out of 4). In general, children rarely deleted que in Spanish sentences, while all groups changed some grammatical sentences to ungrammatical by inserting that in English. The English monolingual subjects did so more than the two groups of bilingual children. While this study shows some differences between monolingual and bilingual children, it does not show that bilingual children schooled in English have lost knowledge of the that-t effect in Spanish. Thus, while Mueller Gathercole (2002b) and Montrul and Potowski (2007) showed that nominal inflectional morphology is still quite vulnerable to incomplete acquisition and language loss in school-age Spanish-English bilingual children who receive academic support for the minority language at school, “innate” aspects of syntax, like the that-trace effect, are not affected. The collective findings of Mueller Gathercole’s (2002a, b, c) experiments showed that monolinguals and bilinguals differ in their command of morphosyntactic features of Spanish and English. Bilinguals lag behind monolinguals in recognizing and correcting grammatical errors. Among the bilingual groups, SES, age, and school environment play a role in English accuracy. Low SES 2nd graders in a two-way bilingual school performed at lower levels of accuracy than bilingual children in English-immersion and bilingual children from high SES. By contrast, bilingual children attending two-way bilingual schools who speak Spanish only at home (sequential bilinguals) performed better in Spanish than bilinguals attending English-immersion and speaking English at home. Overall, although . A question that arises in this study is why judging that-t structures in English was more problematic than in Spanish for both monolinguals and bilinguals. One possibility is that there is dialectal variation with the acceptance or lack of acceptance of that-trace sentences in English. Another possibility may be related to language-particular properties of the complementizer. In Spanish, the complementizer que is used in many structures, including relative clauses, and is obligatory. By contrast, that in English is used in fewer structures and is often optional. Perhaps children of this age are still figuring out the optionality of that and the constraints on the optionality. . A recent reanalysis of the Miami study carried out by Oller, Pearson and Cobo-Lewis (2007) suggests that these bilingual children’s language and literacy is stronger in some domains than in others. Profile effects indicated comparable performance of the monolingual and bilingual children in basic reading, but lower vocabulary scores for the bilingual children in the two languages. Lower vocabulary scores are explained by the distributed characteristics of bilingual lexical knowledge. Typically, bilingual children learn the words for the two languages in different contexts (home vs. school).



Chapter 5.  Bilingualism in middle and late childhood 147

Spanish-speaking children in the United States maintain a certain command of Spanish as they learn English throughout elementary school, quality and quantity of input in each of the languages is crucial for bilingual outcomes, in addition to language status, access to literacy, family language use and community support (Pearson 2007).

2.3

East Asian children and the role of community-based schools

In the absence of minority-language education through the public school system, one way East Asian families have sought to transmit their language and culture to their children is by the creation of community- and church-based schools for Korean and Chinese and the hoshuukoo or Japanese supplementary schools. These schools operate on weekends, or as after school programs. One of the aims of these schools is to retard the process of intergenerational language loss. However, even when community schools are great vehicles for language socialization and ethnic identity formation, studies of bilingual children attending these schools have failed to demonstrate a positive correlation between proficiency in the minority language and length of instruction at heritage language schools (KondoBrown 2004, 2006; G. E. Kim 2006). An example of Asian children’s minority language morphosyntactic development is provided by the Song, O’Grady, Cho and Lee (1997) study of Korean monolingual children between the ages of 3–8 and age-matched Korean-English bilingual children born and raised in the United States (Hawaii). All the US-born Korean-English bilingual children had a very good level of oral proficiency in Korean. (It is not clear whether the children were simultaneous or sequential bilinguals, since no details are given about the family and linguistic history of the children.) In order to maintain and preserve proficiency in Korean, the bilingual children attended regular English language school during the week and a Korean community school on Saturdays. While the bilingual children had advanced oral proficiency skills and were receiving some schooling in Korean, their Korean still did not reach age-appropriate levels of development. Song et al. (1997) conducted two small-scale experiments to find out which areas of linguistic knowledge were still incomplete in these children and how they could be addressed with instruction. The first study looked at Korean case marking. In Korean, subjects are casemarked with the particle -ka and direct objects with the particle -lul. These case markers allow tracing subjects and objects in a sentence. Therefore, Korean is not a strict SOV language but allows variable word order instead, as shown in (3) and (4).

148 Incomplete Acquisition in Bilingualism

Figure 5.1  Percentage accuracy on SOV sentences (adapted from Song, O’Grady, Cho & Lee 1997).

(3) Yeca-ka namca-lul mil-ess-ta girl-subj boy-obj push-past-decl ‘The girl pushed the boy.’

SOV

(4) Namca-lul yeca-ka mil-ess-ta boy-obj girl-subj push-past-decl ‘The girl pushed the boy.’

OSV

To investigate knowledge of Korean case markers, children were shown 16 pairs of pictures depicting two characters and an action, such as hugging, pushing, kicking, etc. For example, in one picture a girl was hugging a boy while in the other picture the action was reversed and the boy was hugging the girl. Half of the sentences were presented with a context (4 SOV and 4 OSV), and the other half (4 SOV and 4 OSV) without a context. An experimenter tested each child individually. After the experimenter said each sentence, the children were asked to point to the picture that matched the sentence. Figures 5.1 and 5.2 summarize the main results. (The results of the monolingual 2-year olds, who behaved below chance with all sentence types, are omitted.) Figure 5.1 shows the results of the SOV sentences. The 3–4 year olds performed below or slightly above chance. By contrast, the older children performed



Chapter 5.  Bilingualism in middle and late childhood 149

Figure 5.2  Percentage accuracy on OVS sentences (adapted from Song, O’Grady, Cho & Lee 1997).

at close to 90% accuracy (Korean monolinguals) and close to 80% accuracy (bilinguals). Results of the OSV sentences, shown in Figure 5.2, present a very different picture, however. Only the older monolingual children performed close to 90% accuracy with sentences for which a context was provided, whereas the bilingual children performed at or below chance. The older bilingual children performed worse than the younger children, providing indirect evidence for language loss (probably incomplete acquisition) in this morphosyntactic domain. In their second experiment, Song et al. (1997) also tested knowledge of the Korean reflexive pronouns caki and caki-casin in sixteen 6–14 year-old Korean bilingual children attending the same Korean community school. Results of this study showed that only 3 of the 16 children knew the meaning of the Korean reflexives. The combined results of the two experiments suggest that even when there is some support for the heritage language through school, reduced input has long-lasting effects on older bilingual children. The inability to use case markers or to recognize reflexive pronouns in age-appropriate ways is related to the fact that these bilingual children are exposed to much less Korean than their monolingual peers.

150 Incomplete Acquisition in Bilingualism

In the case of East Asian languages, quality of input and socialization matter a great deal to develop aspects of language that crucially depend on the sociopragmatic context, such as politeness markers. One example is the acquisition of honorific verb suffixes, a grammatical domain that straddles morphology and pragmatics. Many second generation Korean heritage language speakers never fully develop the complex system of honorifics, because they simply do not have the opportunity to interact with grandparents or older relatives and learn these forms (Jo 2001). Undoubtedly, this is an aspect of the Korean language that is incompletely acquired by many Korean-American children if they do not grow up with a grandparent at home (see Park 2006).

2.4

Children with no academic support of the minority language

Children who speak Spanish and Korean have some linguistic support from the heritage language at school in the United States, but what happens when children do not have support and access to a heritage language speech community is not available? Bolonyai (2007) examined the linguistic abilities of 6 simultaneous Hungarian-English bilingual children, ages 7–9. Four of the children were born in the United States and two others immigrated at the age of two years. The parents were Hungarian, but the children attended English-only schools, 5 of them since day-care. As a result, English was the stronger language and Hungarian the weaker language. Bolonyai wanted to find out whether nominal and verbal inflectional morphology was vulnerable to incomplete acquisition in these children who were well beyond the period of early syntactic development (unlike the children discussed in Chapter 4). In particular, Bolonyai examined the children’s accuracy on verbal agreement and possessive agreement in noun phrases in oral production. Hungarian, an agglutinative language, has a complex system of agreement. Table 5.7, from Bolonyai (2007), shows examples of possessive agreement in noun Table 5.7  Possessive agreement and object verb- agreement in Hungarian (adapted from Bolonyai 2007). Possessive agreement

Object-verb agreement definite

indefinite

1SG 2SG 3SG

kalap+om kalap+od kalap+ja

‘my hat’ ‘your hat’ ‘his hat’

lat+ok lat+sz lat+O

‘I see’ ‘you see’ ‘he/she sees’

lat+om lat+od lat+ja

1PL 2PL 3PL

kalap+unk kalap+otok kalap+juk

‘our hat’ ‘your hat’ ‘their hat’

lat+unk lat+tok lat+nak

‘we see’ ‘you see’ ‘they see’

lat+juk ‘we see’ lat+ja-tok ‘you see’ lat+jak ‘they see’

‘I see’ ‘you see’ ‘he sees’



Chapter 5.  Bilingualism in middle and late childhood 151

Table 5.8  Errors in possessive agreement in Hungarian by 6 Hungarian-English bilingual children (adapted from Bolonyai 2007). Construction

Total

% Correct

% Incorrect

Possessor in NP Possessor in be-clause

205   80

96.6 33.8

  3.4 66.2

phrases and definite and indefinite object agreement in the verb phrase. Interestingly, possessive and verbal inflections are very similar. According to Bolonyai, the Hungarian-English children produced 3.3% errors with verbal agreement and 21.1% with possessive agreement. All errors with possessives involved omissions, while all errors with verbal agreement involved substitutions. There was a clear difference also between possessive agreement in noun phrases versus possessive agreement in clauses, as summarized in Table 5.8. The data revealed differential vulnerability to verbal and nominal morphology, even when the actual surface morphemes are very similar: possessive be-clauses were the most affected, whereas possessive noun phrases and verbal morphology in general were largely error free (only roughly 3% error rate). Although data from monolingual Hungarian children are not provided, Bolonyai claims these bilingual children still show significant delay and imperfect acquisition of grammatical inflections that are not particularly problematic for monolingual Hungarian children of the same age. Thus, this study shows that delayed acquisition of the syntaxmorphology interface in early bilingualism as described in Chapter 4, eventually ends up as incomplete acquisition in older simultaneous bilingual children. Errors which disappear in monolingual acquisition persist well after the period of primary linguistic development and into the school-age period in simultaneous bilingual children who have no academic support in the minority language. All the cases of bilingual children discussed so far suggest that language loss and incomplete acquisition is possible also in bilingual school age-children who are primarily exposed to the L2 and majority language, even when these children have different degrees of access to the minority language at home, school, or the community. But can a language be lost entirely in middle childhood in circumstances where input to the L1 is completely cut-off? Isurin (2000) reports a case study of a Russian girl who was adopted by an American family when she was 9 years of age. Her Russian was well-developed and reflected the proficiency and fluency of Russian-speaking children of the same age. Upon arrival in the Unites States, the adopted girl knew no English. Six months after arrival, the girl was enrolled in the third grade at an English-only school, and in one semester she had acquired the linguistic skills of a first grader. Isurin followed the rapid acquisition of English and even more rapid decay of this child’s Russian for a year, using three

152 Incomplete Acquisition in Bilingualism

picture-naming tasks targeting both objects and actions. Accuracy and reaction times in the two languages were measured. Results showed that although L1 loss is related to L2 acquisition, the rate of forgetting in Russian during a year was slower than the rate of L2 acquisition of English. Nouns were acquired and lost more rapidly than verbs. Two years after arrival, and a year after the last testing session, loss of vocabulary in the Russian adoptee was even more severe, since about 60% of the vocabulary used in the study was no longer retrieved, But Isurin did not report total language loss. This suggests, perhaps, that unlike the case study of the infant Chinese adoptee discussed in Chapter 4, who showed dramatic loss of Chinese in less than three months, complete loss of the L1 is less likely in children adopted after the L1 has had a chance to develop fully. In conclusion, all these studies show that incomplete acquisition in simultaneous bilingual children is also manifested in the school age period, even when some children may have access to some instruction of the standard variety in the minority language. Sequential bilingual children experience language loss once instruction in the majority language begins, but the degree of loss in sequential bilinguals may turn out to be less severe than in simultaneous bilinguals, due to amount of exposure to the L1 earlier in childhood. These studies also show that incomplete acquisition affects some aspects of morphosyntax more than others (nominal vs. verbal morphology, tense vs. mood morphology, etc.), as well as structures that are different in the languages in contact (word order with relative clauses, binding dependencies). Other presumably innate aspects of language, like the that-trace effect in Spanish, is not vulnerable to loss. In all cases, the minority language is susceptible to influence from the dominant language at the structural level. In the next section I examine how language shift at a particular age in childhood contribute to degree of language loss.

3.

Age effects in L1 loss as a function of L2 acquisition

Both simultaneous and sequential bilingual children are very successful at acquiring the majority language as L1 or L2 (sequential bilinguals), and their degree of L1 loss depends on a variety of input factors. Still, age of onset of bilingualism during childhood plays an even more fundamental role. Several studies show that immigrant children who arrive at the country early eventually switch their primary and dominant language from L1 to L2, while children who arrive later in childhood keep their L1 as their primary, strong and dominant language (Flege, Munro & MacKay 1995; Jia & Aaronson 2003; Jia, Aaronson & Wu 2002; Kohnert, Bates & Hernández 1999; McElree, Jia & Litvak 2000; Yeni-Komshian, Flege & Liu 2000). Two questions that arise in this situation are first, how long does it



Chapter 5.  Bilingualism in middle and late childhood 153

Table 5.9  Accuracy and reaction times in Spanish and English lexical access during production (blocked conditions) (adapted from Kohnert, Bates & Hernández 1999). Age Group 5–7 8–10 11–13 14–16 College

N 20 20 20 20 20

Spanish (L1) accuracy speed

English (L2) accuracy

speed

60% 78% 79% 71% 87%

38% 70% 75% 87% 96%

1360 1125 1117   999   983

1307 1198 1148 1154 1109

take minority-language dominant children to switch dominance to the majority language? And second, what is the approximate age at which L2 acquisition and L1 loss become less likely in childhood? Kohnert, Bates and Hernández (1999) investigated how long it takes sequential bilingual children who are Spanish-dominant to shift language dominance to English while learning English as an L2 at school. They investigated lexical acquisition, which is often used as a predictor of language proficiency and academic success in bilingual children. The study tested bilingual children on both languages as they gained proficiency in English by attending English-only schools in the United States. Rather than focusing on the relative size of the bilingual vocabulary, as many previous studies have done, this study investigated the speed of lexical access and the ability to resist interference from the other language when naming nouns in English or Spanish. Participants were 100 school-age SpanishEnglish bilingual children between the ages of 5 and 22, divided into 5 age groups: 5–7 years, 8–10 years, 11–13 years, 14–16 years, 18–22 years. All were sequential bilingual or child L2 learners residing in the United States. All came from Spanish-speaking families (Mexican) and began intensive exposure to English at school, some at different ages. All participants were asked to name pictures presented on a computer screen, in the language indicated by a simultaneous auditory cue (“say” or “diga”). A total of 100 pictures were divided into 3 conditions or modes of presentation: Spanish only (25 words), English only (25 words), and mixed (50 words). Pictures described objects like bed, cat, moon, shoe, key, etc. Accuracy and reaction times were measured. Results, summarized in Table 5.9, showed an overall increase in accuracy at naming pictures in each language by age, although the overall gains were greater in English (58%) than in Spanish (27%). Similar results were obtained for reaction times: mean gains in English were of 377 milliseconds from the youngest to the oldest group, while gains in Spanish were 198. In the two languages, the main differences were between the 5–7 and 14–16 and college ages, but there were no differences for children in the 8–10 and

154 Incomplete Acquisition in Bilingualism

Table 5.10  Accuracy and reaction times in Spanish and English lexical access during comprehension (blocked conditions) (adapted from Kohnert & Bates 2002). Age Group

Spanish (L1) accuracy speed

English (L2) accuracy speed

5–7 8–10 11–13 14–16 College

85% 93% 95% 96% 98%

83% 94% 97% 99% 99%

2058 1666 1493 1319 1262

2021 1689 1417 1242 1194

11–13 age groups (see Kohnert 2002 for a one-year follow up study of the 8–10 age group). While gains were made in the two languages, there were no significant decrements in Spanish (the L1). Rather than documenting language loss, this study documents shift in dominance: The greater gain in English than in Spanish suggests a developmental cross-over in proficiency and speed (i.e., fluency) indicative of a shift in relative language dominance from L1 to L2. The younger children were faster and more accurate in Spanish than in English, while the next two age groups (8–10 and 11–13) were balanced in the two languages. The older adolescents and young adults were stronger in English than in Spanish, the language in which they received academic literacy, supporting a shift in balance from L1 to L2. Kohnert and Bates (2002) investigated developmental changes in lexical comprehension in the same English-Spanish bilingual children, this time with a timed picture-word verification task. Results, illustrated in Table 5.10, were very similar to those of Kohnert et al. (1999): gains were made in the two languages across age groups, but more gains were made in English than in Spanish. In this study, however, the younger children were already balanced in the two languages (as measured by accuracy and speed) and the overall pattern was one of English dominance in middle childhood. According to this study, language shift shows up earlier in comprehension than in production, which was evident after 10 years of schooling. The findings of Kohnert and her collaborators raise a number of important questions, such as which factors influence the speed at which immigrant children and adults acquire the language of the host country, and which mechanisms lead to the dominant language shift or maintenance processes among younger and older arrivals. But with respect to our research question, this study suggests that L1 lexical proficiency clearly continues to grow and solidify until adolescence, and even beyond. The older the children at age of immigration, the less likely they are to lose skill in their L1 while they gain proficiency in their L2. On the issue of age and vulnerability to L1 loss and the question of what the approximate age is, Hakuta and D’Andrea (1992) also suggested that exposure to



Chapter 5.  Bilingualism in middle and late childhood 155

English before age 10 contributes to language loss in Latino children, but that language use at home continues to be crucial for language maintenance even during the High School years. Hakuta and D’Andrea investigated maintenance and loss of Spanish in Mexican-background high-school students (n = 308, mean age 16;4). A high-school population was chosen because at this age bilingualism should have stabilized while the influence of the linguistic home environment should still be strong, since most high school students live at home with their parents. Self-reported proficiency in each language was compared to performance on a response latency task for vocabulary production and recognition, a grammaticality judgment task, and a cloze test. Another variable of interest was language attitude. Therefore, all students also completed a linguistic background questionnaire with questions about personal and linguistic information, language choice in a variety of settings, and language attitudes towards Spanish. Results of the proficiency measures were converted into a composite score (one for each language) and several correlations were run with composite scores for the different groups of variables. Results showed that for those students who were born in Mexico and immigrated to the U.S. after age 10, and those who were born in the U.S. of Mexican parents, Spanish proficiency was related to the age at which they started speaking English: the earlier they started speaking English, the lower their proficiency in Spanish, and there was a continuous drop until age 10. For teenagers, maintenance of Spanish proficiency was more strongly associated with language practices by the parents in the home than with language use outside the home or their attitude toward Spanish. While language attitude was not related to proficiency, it was related to language shift. In all cases, the high-school students showed shift from Spanish to English outside the home. Patterns of language shift and dominance in sequential bilinguals at the level of morphosyntax have also been reported. Jia and Aaronson (2003) looked at morphosyntax (globally defined) in a longitudinal study of 10 Chinese children as they became bilingual in English. In particular, Jia and Aaronson investigated how degree of L1 ultimate attainment affected L1 loss and L2 attainment in immigrant children with different ages of arrival. Six children had arrival ages of 9 or younger, and the other 4 had arrival ages of 12 or older. Jia and Aaronson investigated changes in linguistic competence as a function of shift, as well as how degree of language maintenance and loss interacted with the children’s linguistic environment and their linguistic preferences. Data from all the children were collected every month during year 1, every 3 months during year 2, and twice a year during year 3 (total 13 data points per child). During each session, each child completed a series of linguistic tasks and an interview. All L1 tasks and interviews were conducted in Mandarin Chinese, and the L2 tasks were conducted in English. Language proficiency was measured in English (L2) through a

156 Incomplete Acquisition in Bilingualism

Table 5.11  Changes in Chinese children’s linguistic competence in Chinese after exposure to English as reported by parent (adapted from Jia & Aaronson 2003). subject age listening A B C D E F G H J K

 5  6  7  8  9  9 12 12 15 16

Y1  5 –1  0  0  0  0  0  0  0  0

Y2 3 4 0 0 1 0 0 0 0 0

speaking Y3 0 0 0 0 3 0 0 1 0 0

Y1 4 0 0 0 0 0 0 0 0 0

Y2 3 4 0 0 1 0 3 0 0 0

reading Y3  0  0  0  0  3 –2  0  0  0 –1

Y1  1 –2 –1 –1 –1 –1  3 –1  0  0

Y2  0  0 –2 –2 –1 –4  3  2  1  0

writing Y3  2 –3 –2 –5 –2 –2  2  0  0  0

Y1  1 –2  1 –2 –2 –2  1 –2  0  0

Y2  0 –1 –1 –3 –2 –5  1 –1  0 –1

Y3  1 –2 –4  0 –1 –3 –1 –1  0  0

grammaticality judgment testing several morphosyntactic features of English (the same instrument used by Johnson & Newport 1989). The children also translated 36 orally-presented sentences from Chinese into English. The parents were asked to complete a questionnaire and an interview, where they rated how they perceived their children’s proficiency in English and Chinese, as well as the children’s changes in Chinese proficiency in reading, speaking, listening and writing. Overall results showed what Jia and Aaronson (2003) interpret as L1 attrition among early arrivals and L1 maintenance among late arrivals. At the beginning of the study, all children were Chinese-dominant in terms of language preference, use, and proficiency. All participants knew Chinese but little or no English at the beginning of the study. During the first two years of study, there was an emerging switch toward English as the younger children gradually acquired, began to use, and preferred, more English than Chinese, while the older arrivals maintained Chinese as their dominant language. These language preferences and the alleged results of these preferences (i.e., attrition vs. maintenance) were associated with L1 and L2 proficiency changes. The study examined gains in L2 proficiency but unfortunately did not include a similar proficiency measure for Chinese. According to parental reports, the younger children lost proficiency in Chinese, especially in reading and writing, while the older arrivals maintained it. This is shown in Table 5.11, where 0 indicates no change, a score close to 5 indicates significant improvement, and a score of –5 indicates greatest decline. In terms of language preference, a clear switch was also evident in the younger children’s reports as early as 12 months after exposure for both speaking and reading, as shown in Table 5.12. Thus, even though this study only shows results of parental reports and does not offer concrete examples of grammatical areas



Chapter 5.  Bilingualism in middle and late childhood 157

Table 5.12  Changes in Chinese children’s language preferences after exposure to English (adapted from Jia & Aaronson 2003). Subject

A B C D E F G H J K

Age

Speaking preference

5 6 7 8 9 9 12 12 15 16

3 months L1 L1 L1 L1 L1 L1 L1 L1 L1 L1

12 months L2 L2 L2 L2 Both Both L1 L1 L1 L1

24 months L2 L2 L2 L2 Both Both L1 L1 L1 L1

Reading preference 36 months L2 L2 L2 L2 Both L2 L1 L1 L1 L1

3 months – L1 L1 L1 L1 L1 L1 L1 L1 L1

12 months L2 L2 L2 L2 L2 L2 L1 L1 L1 L1

24 months L2 L2 L2 L2 L2 L2 L1 Both L1 L1

36 months L2 L2 L2 L2 L2 L2 L1 Both L1 L1

affected by switches in L1 and L2 proficiency, it still suggests that the earlier the exposure to the L2, the more L1 loss of some sort in both structural aspects of the language, language skill, or modality. Further evidence for the claim that the younger the exposure to the L2, the more severe the loss in the L1, or at least the perception of one’s L1 ability, comes from other recent studies of East Asian immigrants. Luo and Wiseman (2000) compared self-reported Chinese language proficiency and use in 250 school age and university Chinese heritage language learners. The bilingual children were divided into two groups based on age of immigration before/after age 5. Those who immigrated after age 5 (sequential bilinguals) rated their Chinese proficiency higher than those who immigrated before age 5 or were born in the US (simultaneous bilinguals). Kataoka, Koshiyama and Shibata (in press) found similar results for age of immigration predicting degree of language maintenance in 1,591 first to ninth grade children attending Japanese supplementary school (hoshuuko). They found that when the children’s age of immigration was younger than 10 (fifth grade), their English proficiency level was likely to surpass their Japanese proficiency level within two years of immersion in the US, even when the children continued to receive Japanese instruction at the supplementary school. Finally, E. J. Kim’s (2004) survey of 70 Korean college age students who had arrived in the United States at different ages in childhood (birth to 20), showed that those who arrived before age 12 did not rate their Korean proficiency as high as those who arrived later in childhood. H. S. Kim (2005), who examined the reported use of the L1 at early ages in 88 school-age Korean-English bilinguals, found that those who used Korean as their L1 performed better on a listening test of relative

158 Incomplete Acquisition in Bilingualism

clauses than those who used less Korean because they also used English (simultaneous bilinguals). The generalization that emerges from all these studies is that age of arrival/ immigration is a highly predictive factor in minority language loss and maintenance. The earlier the age of arrival and onset of bilingualism, the lower the eventual proficiency in the L1 is likely to be. Vulnerability to loss seems to cease around the age of 10 for children who are sequential bilinguals. Unlike the studies of adult attrition reported in Chapter 3, which did not find a correlation between frequency of language use and degree of attrition (Schmid 2007), studies of children show that language use and degree of attrition, especially in early childhood, are related.

4.

Further research

Having looked at data from various research studies that examine language ability and language use in minority language school-age bilingual children, I am now in a position to draw some conclusions about the fate of the weaker language in these children. Research on older children is scarce and scattered, and a vast majority is based on self-ratings and reports rather than on actual linguistic measures. The few linguistically-oriented studies available and discussed in this chapter show that there are important changes in L1 knowledge when school-age bilingual children socialize and receive academic support in the majority language. But the specific nature of these changes should be pursued with further research. Unfortunately, however, we also know very little about language acquisition during the school age-period in monolingual children. Even though the language acquisition literature states that basic linguistic knowledge is acquired by age 3–4, I see this as an overstatement. Obviously, there are many complex and subtle aspects of grammar, meaning and pragmatics that children do not fully master and comprehend until much later. But these have not received much attention. The field of L1 acquisition needs to expand its inquiry to the adolescent years, since some recent research suggests that children are not done with language learning that early (Guasti & Cardinaletti 2003; Jia & Aaronson 2003; Keijzer 2007; Pires & Rothman, in press). In this context, it is crucial to understand the precise impact of literacy on language development in general and how native speaker knowledge develops at school. Until very recently, the field of child L2 acquisition within the context of education has been mostly concerned with how fast minority-speaking and L2 acquiring children can achieve native-like levels of achievement, so that they can reap the benefits and opportunities the L2 society has to offer. If an early onset



Chapter 5.  Bilingualism in middle and late childhood 159

of L2 acquisition ever leads to native-like command of the L2, it has increasingly become more apparent that such a high degree of native-like attainment may also come at the cost of reduced L1 ability in language minority children. Progressive loss of L1 skill is likely in both simultaneous bilingual children and in sequential bilingual children. Among the most salient input factors responsible for these outcomes in children are reduced use and support of the minority language at home and at school. We need to keep exploring why this is so. Learning more about later language development in monolingual children is crucial to understanding what exactly bilingual children at this age miss when they do not get the full advantage of receiving academic training in one of the languages. Are complex syntactic structures incompletely acquired later in life because they are more frequent in written input, which many bilingual children do not get? Whereas use of the minority language at home does not contribute to slower progress in the majority language for L2 children, the entrenchment of the majority language in the home (through majority language speaking parents and siblings), coupled with limited to no support for the heritage or family language at school, has cumulative detrimental effects on the linguistic ability in the L1. Thus, during the school-age period, we see a trade-off between L2 acquisition and L1 loss in children: as L2 development advances steadily and progressively until about 10–11 years of age, L1 development may gradually deteriorate. To conclude, if there is language delay in very early bilingualism, as discussed in Chapter 4, the age function for language loss extends to the school-age period as well. Research so far seems to point to 8–10 years of age as the critical threshold age after which vulnerability to L1 loss is less likely. However, considerably more linguistic and psycholinguistic research is needed to understand the interplay of L1 loss and L2 acquisition in bilingual children between the ages of 4 and 18 years, at all levels of grammatical development, from lexicon to complex syntax and pragmatics. In the next two chapters, I address the long-term effects of language loss throughout childhood, by focusing on adult early bilinguals who have different degrees of incomplete acquisition of the minority language.

chapter 6

Incomplete L1 acquisition in adults

In the previous two chapters, we saw how reduced exposure to the family language may cause language-minority children to progressively lose linguistic ability in the heritage language. Furthermore, degree of L1 attrition in minority language-speaking children is intrinsically related to age of onset of bilingualism and extensive exposure to the majority language: the earlier the exposure to the L2, the more severe the L1 loss is likely to be. This chapter addresses the following questions. First, what is the fate of the heritage language (in some cases their weaker language) when these children reach adulthood? In other words, what are the long-term effects of incomplete acquisition? And second, what structural characteristics does the heritage language have? Thus, the focus of this chapter is on adult early bilinguals at time of testing, who acquired the two languages (simultaneously or sequentially) in childhood, like the cases of Kristi (the Chinese adoptee), Elena (the Russian girl who grew up in France), and Carlos, Alicia, and Breatriz (the Mexican siblings who immigrated to the United States in childhood), introduced in Chapter 1. Throughout this chapter and the rest of the book, I will use the term heritage speaker to refer to adult early bilinguals of minority languages. I will show how many of the characteristics typically ascribed to adult L2 acquisition presented in Chapter 2, such as non-targetlike ultimate attainment, morphological variability, and cross-linguistic influence, are not unique to L2 acquisition as often portrayed in the L2 acquisition field. In fact, these characteristics also apply to this early bilingual population. We will see that the range of proficiency in the L1, or one of the L1s, varies considerably in adult heritage speakers as a result of many environmental variables. Degree of proficiency is also deeply related to onset and extent of L1 attrition at different ages in childhood. The language of these speakers has been frequently discussed in the context of L1 attrition, on par with the cases of adult attrition discussed in Chapter 3. My position with respect to childhood attrition is different from that of other researchers. I maintain that childhood attrition often results in incomplete or fossilized L1 acquisition, which is different than the type of L1 attrition many adults experience after living in an L2 environment for an extended period of time. The effects of attrition in childhood seem to affect linguistic competence deeply and not just performance and retrieval, as we saw in adults (Chapter 3). In keeping with what is frequently observed in adult L2 acquisition, but unlike L1 attrition in

162 Incomplete Acquisition in Bilingualism

adults, inflectional morphology, as we will see, is the grammatical domain most vulnerable to incomplete acquisition and reanalysis. I begin with a description of the characteristics of this adult population.

1.

Linguistic profile of adult heritage speakers

The linguistic profile of adult L2 learners assumed to be undergoing L1 attrition after extensive exposure to, and use of, the L2 was discussed in Chapter 3. These are first generation immigrants who immigrate to the L2 society in adulthood. They are typically monolingual speakers of the minority language who learn the majority language late in life (and imperfectly). Command of the heritage language is strong in this immigrant group, although there can be some attrition after more than 10 years of intense exposure to the majority language. However, in this population, attrition affects mostly retrieval and access of lexical items rather than linguistic competence per se. If there is attrition at the syntactic, semantic and pragmatic level, it takes the form of L2 influence. Although some L2 effects were observed in areas of syntax related to binding interpretations and the expression of subjects in null subject languages, these were relatively minor. Morphology is more affected than syntax, but in general, error rates are very low in these studies. More severe effects of language loss, however, are observed in second generation speakers, that is, the children of first generation immigrants. But there is a great deal of variation within the second generation group, as discussed in Chapters 4 and 5. This group is very heterogeneous. It includes the children of first generation immigrants born in the host country to at least one first generation parent. It also includes immigrant children who come to the host country at different ages in childhood. In terms of types of bilingual profile, this group may include simultaneous bilinguals, those exposed to the heritage and the majority language before the age of 3–4; sequential bilinguals or child L2 learners, those exposed to the heritage language at home until age 4–5 and to the majority language once they start pre-school; and late child L2 learners, children who are monolingual in the heritage language, received some elementary schooling in their home country, and immigrated around ages 7–8 or later. Internationally adopted individuals are also a type of heritage speaker. Language shift in the home typically occurs in the second generation of speakers when children are schooled in the majority language and have a strong desire to fit in with the new society. As the majority language begins to be used more than the home language, some aspects of the heritage language may be incompletely acquired, others may undergo attrition, and yet others may undergo ­attrition when



Chapter 6.  Incomplete L1 acquisition in adults 163

they were not fully mastered. Many of these children are either monolingual or dominant in the heritage language in early childhood. As bilingualism progresses during the elementary school period, the children can be balanced in the two languages (typically at ages 10–11) and eventually become dominant in the majority language, as we saw in the previous chapter. When they reach adolescence, minority language speaking-children are already dominant in the majority language, and by the time they are adults, the majority language is both stronger and dominant in proficiency and domains of use. Due to this rapid shift, heritage speakers are native speakers of the majority language by the third generation (the grandchildren of the first generation immigrants). Some may have limited knowledge of the heritage language while most do not. This declining pattern of bilingualism shows how minority languages do not typically survive intergenerational transmission, even when they are supported by continuous immigration.

1.1

L1 attrition vs. incomplete L1 acquisition

When early bilinguals reach adulthood, their competence in the heritage language varies considerably. Depending on their ethnic group, many, but by no means all, heritage speakers possess good speaking and listening abilities, home-based vocabulary, native-like levels of pronunciation and fluency, and familiarity with the cultural norms in the target language and culture. What is less clear is how proficient these speakers are with other aspects of grammar. Describing the typical profile of Spanish-English bilinguals in the United States, Valdés (2000) states that while they are fluent in English, some produce errors similar to those of adult L2 learners and child L1 learners in Spanish; others are as proficient as native speakers in written and spoken Spanish (Lipski 1993). Second generation speakers have often been studied as cases of L1 attrition (Håkansson 1995; Hulsen 2000; Silva-Corvalán 1994; Vago 1991). Some longitudinal studies discussed in Chapter 4 showed that very young bilingual children can display more complex and accurate knowledge of morphosyntax at age 3 than at age 5 (­Anderson 1999; Silva­Corvalán 2003; Kaufman & Aronoff 1991; Turian & Altenberg 1991). Findings such as these suggest that something already in place at an earlier age was lost or weakened (i.e., produced with higher error rates) a few months later. ­However, . My experience working with Spanish heritage speakers in the United States is that many early bilinguals retain intermediate to advanced levels of proficiency as adults. Yet, Maria Polinsky (Harvard University, personal communication), who has worked extensively with the Russian-speaking population has found that most American Russians have very low to intermediate levels of proficiency. For Korean as a heritage language in the United States, it is possible to find individuals ranging from very low to very advanced proficiency in the language.

164 Incomplete Acquisition in Bilingualism

when it comes to assessing the linguistic characteristics of heritage language knowledge in adults, it is very difficult, in fact, to tell whether their apparent incomplete L1 knowledge is actually due to L1 attrition in childhood, to incomplete L1 acquisition, or to both, especially in the case of more complex structures and vocabulary assumed to be acquired after age 3 by monolingual children. If, as we saw in the previous chapter, language acquisition spans more than twelve years when we take into account language development at school, it is quite possible that many adult heritage speakers – especially all those who were schooled almost exclusively in the majority language – missed the opportunity to acquire the standard variety of their heritage language fully in childhood. Unable to properly distinguish between childhood attrition and incomplete acquisition in the absence of longitudinal data, I have argued at length elsewhere that most of these adults, especially those who began exposure and use of the L2 in early childhood, may be better characterized as cases of incomplete L1 acquisition (Montrul 2002, 2004a; Polinsky 1997, 2006, 2008). It is hard to take these adult early bilinguals as strict cases of L1 attrition because it is impossible to assess the route of their language development in childhood. All we know about these adults when they are tested is that the outcome of attrition or incomplete acquisition in childhood is a linguistic system which, as we will see, varies to some extent from the outcome of adult L1 attrition in fully developed native speakers discussed in Chapter 3. With these caveats in mind, I will still refer to these adult heritage speakers as incomplete L1 learners to distinguish them from first generation adult bilinguals assumed to have undergone L1 attrition. The crucial distinction between incomplete learners and L1 attriters has to do with the nature of their bilingualism and command of the two languages. Thus, while learning a second language after puberty – after the first language is fully developed – does not affect proficiency in the L1, learning a second language before puberty, and becoming highly proficient in it, may have more dramatic effects on the L1. In the next section, we will explore the effects of stabilized incomplete acquisition in different areas of grammatical knowledge. Why is it important to understand the linguistic competence and patterns of language use of adult heritage speakers? Research on the linguistic abilities of adult heritage speakers has been gaining momentum lately. Although for many years the study of heritage speakers has been the realm of bilingual education and sociolinguistics, the last decade has seen studies tackling these issues from other approaches and disciplines: linguistic, psycholinguistic, second language acquisition, bilingualism, language education, and language disorders. Understanding the type of linguistic knowledge these speakers have retained from childhood has both theoretical and practical implications. From the perspective of linguistics, psycholinguistics, and second language acquisition, this population raises



Chapter 6.  Incomplete L1 acquisition in adults 165

i­ mportant questions about the nature of language, the role of input (quantity and nature), and the role of age in the acquisition and loss of language in a bilingual situation. As stated in Chapter 2, many features of adult L2 acquisition which do not occur in monolingual L1 acquisition – L1 transfer, incomplete ultimate attainment and fossilization – have traditionally been explained by the age factor in adult L2 acquisition. I argue that many of these features also characterize incomplete L1 acquisition in heritage speakers, and I suggest that the age of L1 acquisition cannot be the entire explanation behind incomplete L1 knowledge in adulthood. Rather, it is age of onset of bilingualism and reduced exposure to the L1 input in these cases that matters most. But there are other more practical reasons for studying this particular bilingual population. Today more than ever before, many of these speakers enroll in second language classes at the post-secondary level because they have a strong desire to relearn, maintain or expand their knowledge of the language for both personal and professional reasons. Depending on the minority language they speak and their level of proficiency in it, these speakers are often placed in classes specifically designed for post-puberty second language learners with no previous knowledge of the L2, but who have very advanced literacy and academic skills in the L1 and the L2. Other institutions are able to offer special programs for heritage speakers. The particular pedagogical approach assumed in either program clearly demands a thorough understanding of the type of linguistic knowledge heritage speakers have or lack, and how this knowledge is similar to, or different from, that of L2 and L1 learners of the language. The language education of heritage language learners is also an area of current intense research interest and significance (Brinton, Kagan & Bauckus 2008; Kondo-Brown 2006; Valdés, Fishman, Chávez & Pérez 2006). Other educators, medical doctors, and speech pathologists need to understand what normal interrupted L1 acquisition looks like in children and adults with these linguistic characteristics in order to make proper diagnoses and referrals about language disorders in bilingual speakers. We will revisit to these and related issues in the next two chapters.

1.2

Methodological considerations in the study of incomplete L1 acquisition in adults

Before discussing specific studies documenting the long-term effects of incomplete acquisition in adult early bilinguals, let me briefly state some of the methodological challenges these studies face. As in cases of L1 attrition and L2 fossilization, the ideal research design would be a longitudinal study following the same individuals and documenting changes in their linguistic behavior from time 1 to

166 Incomplete Acquisition in Bilingualism

time 2 or time n. Unfortunately, for a variety of practical reasons, this is not always possible. Consequently, many of the available studies use competent monolingual speakers who speak the “full” variety (Polinsky 1997, 2006) as the baseline, control, or comparison group. The language of these monolingual speakers is compared to the family language of heritage speakers, most often their weaker language. If significant quantitative and qualitative differences are found between these groups, then incomplete acquisition is assumed from the heritage speakers. Other researchers have also used first generation immigrants as the baseline group (even if they have also suffered some sort of attrition) because the degree of loss in second and third generation immigrants is typically more severe than in first generation immigrants (e.g., Hulsen 2000; Silva-Corvalán 1994). Since many heritage speakers speak different regional dialects, or come from different socioeconomic backgrounds, it is also important to control for language variety. In other words, not much can be concluded about incomplete acquisition if, for example, middle class native speakers of peninsular Spanish are compared with Spanish heritage speakers who come from low socio-economic class and were exposed to Northern Mexican Spanish in the home. The same situation applies to the linguistic and dialectal diversity found in heritage speakers of South Asian languages (Hindi, Urdu, etc.) and Arabic. In addition to the adult monolingual or L1 dominant speaker, another important methodological comparison is that between adult heritage speakers and monolingual children of the heritage language. In the absence of longitudinal data from the same speaker(s), one indirect way to assess incomplete L1 acquisition is to establish when in childhood a given grammatical phenomenon is typically mastered and controlled with almost 100% accuracy. If adult heritage speakers are shown not to control the grammatical phenomenon to the same degree as monolingual or very fluent bilingual children – especially if onset of bilingualism occurred before the grammatical phenomenon is assumed to be fully acquired – then incomplete acquisition can be indirectly inferred in the adult heritage speaker. Finally, another methodological factor that must be considered when investigating the linguistic knowledge of heritage speakers is the degree of proficiency in different language skills. Heritage speakers typically have more advanced oral and aural skills than reading and writing skills because they were exposed to the . See also Seliger (1996) for the same approach. . I am referring to accuracy rates here for the sake of argument. It is important to also note whether heritage speakers make the same grammatical distinctions as full speakers or are sensitive to similar grammatical constraints, regardless of whether or not they reach 90% or 100% accuracy on a given task.



Chapter 6.  Incomplete L1 acquisition in adults 167

language aurally at home. The levels of academic literacy in the heritage language (reading and writing) also vary considerably depending on the availability of the language and schools that teach the language in the speech community, the family socio-economic class, whether or not they speak a heritage language with a different writing system, and other cultural differences. Thus, many experimental tasks widely used in L2 acquisition and bilingual studies might not be entirely suitable for tapping linguistic knowledge in some heritage speakers. This, however, is a matter that also needs further investigation.

2.

Structural characteristics of the incomplete adult L1 grammar

Most of the available studies on adult incomplete acquisition have focused on the linguistic characteristics of the heritage language on the one hand, or on how to teach heritage languages on the other. In this section, I focus on the former. There are two questions that have guided this research so far. First, are adult heritage speakers fossilized L1 learners (“frozen children” as Polinsky 2008a puts it)? And second, are incompletely acquired adult L1 grammars systematic linguistic systems (i.e., UG-governed)? According to the Weaker Language as L1 Hypothesis I offered in Chapter 4, my position is that even if there are missing elements, adult weaker languages have the signatures and linguistic systematicity of both developing and full L1 systems acquired in early childhood. An incomplete grammatical system is not necessarily UG-incompatible (see White 2003a for a similar position with respect to adult L2 acquisition). Hence, my claim is that incomplete L1 grammars are fully constrained by Universal Grammar. Even when adult heritage speakers display high error rates in several grammatical domains, and especially in morphology, there is substantial evidence for the claim that adult heritage speakers have set the syntactic and phonological parameters of their native language early in childhood. We saw in Chapter 3 that, if there was attrition in first generation immigrants, it manifested itself as lexical retrieval difficulties and minor quantitative changes in pronunciation (phonetics/phonology), and L2 effects at the syntax-pragmatics interface. Errors with inflectional morphology, which are most common in adult L2 acquisition, were negligible. By contrast, the studies on attrition and incomplete acquisition in bilingual children discussed in Chapters 4 and 5 showed that the effects of attrition can be quite extensive in childhood, and most notable in aspects of morphosyntax. In the present chapter, I examine the long-lasting effects of childhood attrition in adult grammars. We will see that incomplete acquisition affects all aspects of grammar, and morphosyntax in particular.

168 Incomplete Acquisition in Bilingualism

2.1

Phonology

How does incomplete acquisition affect phonology? One of the first studies documenting the effects of childhood attrition into adulthood in phonological processes is Vago’s (1991) case study of a 36-year-old Hungarian speaker who immigrated to Israel with her parents at age 5;10. The study examined potential loss of exceptions (rule simplification), rule reordering, and loss of rules. Although not explicitly stated by Vago, these are the linguistic processes predicted by both the Regression Hypothesis and the generative approach to attrition, since both theories predict some sort of simplification or loss of non-core aspects of grammar. At the time of testing, Hebrew (the L2) was the subject’s primary language, while use of Hungarian (her L1) was severely reduced to weekly conversations with her parents (who remained Hungarian-dominant speakers like most first generation immigrants). Vago interviewed the speaker in Hungarian and elicited nominal and verbal paradigms to see whether she simplified morphophonological rules. To complement this procedure, the participant was also asked to give grammaticality judgments on individual word tokens. The standard dialect of Hungarian spoken by the speaker’s parents was used as the baseline data, and this seems appropriate because, presumably, the parents’ language must have been the dominant source of Hungarian input when the speaker was growing up. Any deviation from these data was interpreted by Vago as a sign of attrition (incomplete acquisition in my view). In the Hungarian possessive paradigm, 3rd person singular and plural nouns ending in consonant or vowel show an epenthetic glide right after the stem and before the possessive suffix, as in the examples with the word for “stick” in Table 6.1. The incomplete learner’s version of the singular possessive paradigms had a glide before any vowel initial suffix including 1st and 2nd plural forms, with the word for “house”. Thus, the rule of /j/ epenthesis was simplified and overapplied to other contexts to achieve paradigmatic regularity, such that /j/ precedes all vowel initial morphemes following a nominal stem. Another rule of Hungarian is that short low vowels become long at the end of a morpheme, except word-finally. The vowel ‘a’/"/ is lengthened to ‘á’ /a:/ and . Although the Regression Hypothesis was formulated on the basis of aphasic children, I agree with an anonymous reviewer that it is hard to see how the Regression Hypothesis relates to cases of attrition in childhood. If attrition in childhood is incomplete acquisition, then how can one lose something that has never been acquired? But the Regression Hypothesis has two parts to it: one is the acquisition part, which follows a given chronological order of acquisition. The other is the regression part. The Regression Hypothesis as applied to incomplete acquisition can only relate to the acquisition part of the hypothesis and whatever structures have been acquired (i.e., point of truncated acquisition), but not to the regression part.



Chapter 6.  Incomplete L1 acquisition in adults 169

Table 6.1  Examples of overapplication of the /j/-epenthesis rule in Hungarian possessive agreement by one adult incomplete learners (adapted from Vago 1991). Possessive agreement in Hungarian

Incomplete learners’ version

3rd sing. 1st pl. 2nd pl. 3rd pl.

bot-j-a bot-otok bot-unk bot-j-u-k

‘his stick’ ‘your stick’ ‘our stick’ ‘their stick’

*bot-j-otok

3rd sing. 1st pl. 2nd pl. 3rd pl.

hazá-tok hazá-j-a hazá-unk hazá-j-u-k

‘your homeland’ ‘his homeland’ ‘our homeland’ ‘their homeland’

*hazá-j-átok

*bot-j-unk

*hazá-j-unk

‘e’/ε/ is lengthened to /e:/, as in the term haza “home or homeland,” which becomes hazá-m “my house” when inflected, while another rule applies to words ending in /h/. In word final position, and before a consonant initial suffix, the /h/ is not pronounced (although it is retained in spelling). For example, the word for “bee” méh [me:] in the dative is méh-nek [me:nεk], but in the 3rd sg. possessive it is méh-e [mehε]. In standard Hungarian, these two rules interact in such a way that the vowel lengthening rule precedes the final /h/ rule. Thus, the word “cseh + nek” /twεnεk/ “Czech, dative” is pronounced cse + nek or /twεnεk/ in the dative and cseh-e /twεkε/ in the 3rd singular possessive. According to Vago, the subject appeared to have knowledge of the two rules operating independently, but had trouble when the two rules interacted, as in the Czech example. She incorrectly produced /twεnεk/with a long vowel in the dative. The final type of attrition phenomena that Vago discusses is rule loss. Hungarian has some words that end in a sequence of an /h/ and an /r/ as in “teher” /tεhεr/ for load. When inflected, the sequence of /h/ and /r/ (or a liquid) is interrupted by a vowel (as in teher load-nom. or teher-nek, load, dat.) or metathesized in prevocalic position, as in 3rd sg. possessive terh-e. The subject’s elicited data did not contain instances of the metathesized process. This rule and two others (t-palatalization and assimilation of the instrumental suffix) were assumed to be lost (or actually never acquired in my opinion) in the grammar of this speaker. These three morphophonological processes and others pertaining to lexical restructuring led Vago to conclude that they were attrited in the Hungarian-Hebrew speaker interviewed for this case study. While these simplification processes are attested in the speaker, the data is, unfortunately, not quantified, so we do not know what percentage of errors these examples represent, nor are they presented in comparison with the baseline data from the parents. Vago acknowledges in a footnote that some of these processes could be due to transfer from the L2 and not

170 Incomplete Acquisition in Bilingualism

to attrition, but no relevant data from Hebrew is discussed to allow the reader a proper evaluation of this hypothesis. Vago also rules out what he calls “imperfect acquisition” because the morphophonemic alternations discussed are in general “under control by monolingual Hungarian children by the time they reach first grade, the approximate age of the subject when she emigrated from Hungary” (Vago 1991, p. 250). However, no data exists from the subject at that time, so even if these rules were already acquired, they were not acquired robustly enough to remain as part of her competence. Although Vago claims that none of these “deviant” processes were observed in the subject’s parents, the study does not include the actual results of the data elicited from the parents. A more recent study documenting phonetic changes in vowel production as a result of incomplete acquisition is Godson (2004), who investigated how the age at which English becomes dominant in Western Armenian immigrants in the United States affects their production of vowels in this language. Ten participants learned English before the age of 8, while the other ten learned English in adulthood. Following Polinsky (1997), Godson refers to the bilinguals exposed to English before the age of 8 as “incomplete language” learners and to the late bilinguals as “uninterrupted acquirers.” Western Armenian has five vowels: front high /i/, front mid /ε/, central low /a/, back mid /o/, and back high /u/. The experimental subjects in Godson’s study were asked to read out loud 86 sentences containing target words likely to be acquired by age 5 in Western Armenian, and which contained the target vowels in stressed closed syllables (CVC). Responses were audio recorded and submitted to acoustical and statistical analysis. Results of the adult bilinguals were compared to those of an adult monolingual Western Armenian speakers and showed that, in general, the Western Armenian vowels were influenced by English in all the bilingual speakers, but more so in the incomplete learners than in the uninterrupted speakers. Furthermore, incomplete acquisition was selective, only affecting the production of the two front vowels /i/ and /ε/ and the central vowel /a/. The values for these vowels were closer to English values than those of /o/ and /u/. The incomplete L1 learners showed more signs of an “incomplete” or different system than the uninterrupted acquirers. In conclusion, both Vago’s and Godson’s studies show that incomplete adult grammars show signs of simplification, restructuring, and probable L2 influence at the phonetic and phonological levels. Godson’s study further confirms that the age ate which the family language was demoted to minority status in heritage speakers is a determining factor in the degree of divergence from fully acquired systems in immigrants who moved to the L2 environment much later in life.



2.2

Chapter 6.  Incomplete L1 acquisition in adults 171

Inflectional morphology

As with adult L2 acquisition, inflectional morphology is the grammatical domain most vulnerable to incomplete L1 acquisition. Adult heritage speakers also display morphological variability. In this section, I discuss data from a comprehensive study of Russian and several recent studies (many of my own) on Spanish heritage speakers.

2.2.1 Nominal morphology: Case and gender agreement Spanish nouns are inflected for gender and number, while Russian nouns are marked for gender, number and case. Both gender and case morphology are affected under incomplete acquisition in children exposed to English as a majority language, as we saw in Chapters 4 and 5, and persist into adulthood. Polinsky (1997, 2006) offers a comprehensive description of the linguistic system of American Russian. Among the 20 speakers studied, 16 immigrated to the United States in childhood (ages ranged from birth to 10 years). The other 4 were 11–18 years old at age of immigration and, in my view, would be first generation speakers undergoing L1 attrition rather than cases of incomplete acquisition. Full Russian (the standard Russian spoken in Russia) has a 6-way case distinction in nouns: nominative, accusative, dative, instrumental oblique and genitive. In general, the case system is severely reduced in speakers of American Russian. Dative is replaced by accusative and accusative by nominative in many constructions with subjects, direct, and indirect objects. Thus, while full Russian has nominative, accusative, and dative, American Russian speakers only manifest nominative and accusative case in their speech. The dative is retained with some pronouns, while the instrumental, genitive and oblique are also affected. For example, some verbs of existence (byt “be,” stanovit’sja “become,” ostavat’sja “remain,” umirat “die”) and other motion verbs assign nominative or instrumental case to the predicative nominal or adjective. Verbs in the future also take nominals or adjectives in the instrumental case. Polinsky reports that American Russian speakers do not produce the instrumental case in all these instances and inflect all nominals and adjectives with nominative case. In many of the structures that require genitive case (lexically governed genitive, genitive of negation, genitive of possession, and count form), American Russian speakers of higher proficiency use accusative, while those of lower proficiency use the nominative. The genitive of negation, obligatory in Russian with unaccusative verbs, is entirely absent in American Russian speakers. However, the genitive is preserved in genitives governed by a numeral. As for prepositional obliques, these are ­replaced by

172 Incomplete Acquisition in Bilingualism

nominative after prepositions, which never occur in Russian. Retention of prepositional obliques occurs in some frozen forms. Gender agreement is also simplified in adult bilingual speakers who are incomplete learners. We saw in Chapters 4 and 5 that Spanish-English bilingual children lag behind their monolingual peers in their control of gender agreement. We also discussed a case of incomplete acquisition at a very young age (Anderson 1999) and reviewed two other studies of school-age children (Mueller ­Gathercole 2002b; Montrul & Potowski 2007). Studies of adult early bilinguals suggest that these errors are rarely overcome and may remain fossilized into adulthood. For example, Lipski (1993: 161) writes that “vestigial Spanish speakers are aligned with second language learners as regards adjective inflection, for errors of gender and number concord are quite frequent.” Examples (1)–(3) are from Lipski (1993). (1) *Mi blusa es blanco. (should be blanca)   my blouse-fem is white-masc   ‘My blouse is white.’ (2) *Tenemos un casa allá. (should be una casa)   we have a-masc house-fem there   ‘We have a house there.’ (3) *¿Cuál es tu favorito parte? (should be parte favorita)   which is your favourite-masc part-fem   ‘Which is your favorite part?’

Similarly, Polinsky (2008b) shows that gender agreement errors are frequent in heritage speakers of American Russian. Compared to Spanish, Russian exhibits a complex interaction between the three-way gender system (feminine, masculine and neuter) and the complex six-way case system discussed above (nominative, accusative, dative, genitive, instrumental, locative). As in Spanish, gender agreement is manifested syntactically in adjectives, participles, possessive pronouns, demonstratives, and some numerals. In addition, Russian gender agreement also surfaces in past tense verbs. The sources Polinsky (2008b) cites for L1 acquisition indicate that children acquiring L1 Russian control the gender and case agreement systems of the language between the ages of 2;5–2;7, which is approximately the same age Spanish-speaking children control gender. By the age of 3, Russian-speaking children already produce correct gender agreement with adjectives, pronouns, and other . For more errors with case marking, see also Polinsky’s (2008a) case study of child and adult incomplete heritage speakers of Russian.



Chapter 6.  Incomplete L1 acquisition in adults 173

elements. But Russian-speaking children older than 3 years have residual difficulty with non-canonical endings, such as masculine nouns ending in a vowel (e.g., papa “daddy”), feminine nouns ending in a palatal consonant (myš “mouse”), and stem-stressed neuter nouns (čudo “miracle”). In general, Polinsky states that children’s gender errors reflect regularizations of the morphophonological rules of the language, akin to the morphophonological processes attested in the Hungarian incomplete learner reported by Vago (1991) in §2.1. Some of these gender errors persist in Russian-speaking children until the ages of 6–7, and Polinsky assumes they only go away with schooling. The research question guiding Polinsky’s (2008b) study was whether adult heritage speakers would show the same problems with gender as young Russian-acquiring children who have not completed their language learning process. In other words, are adult heritage grammars fossilized L1 grammars? Two experiments were conducted to answer this question. The first tested 12 speakers of American Russian (average age 27) brought up in Russian-speaking families. Nine subjects were born in the United States, while the other three came to live in the United States between the ages 3–5 years. Five monolingual Russian speakers acted as the baseline group. For this experiment, participants were given a list of masculine, feminine and neuter inanimate nouns. Each gender category included words with canonical and non-canonical endings, especially those forms that are problematic for monolingual Russian-acquiring children. Participants heard each word and were asked to repeat each noun using a given possessive pronoun or an adjective. According to the results, the control group of Russian speakers performed at ceiling (almost 100% correct), while the heritage speakers showed error rates ranging from almost 5% with masculine nouns to more than 20% with feminine and neuter nouns. The specific word types that caused more problems were stemstressed neuters and feminine nouns ending in a palatalized consonant. Except for the treatment of stem-stressed neuter forms, which were assimilated to the feminine instead of the masculine gender, all other patterns of errors found reflect the errors made by Russian-acquiring children. However, there is a crucial difference between the L1 children and the adult heritage speakers. The L1 children typically vacillate with the classification of neuter nouns, even though they have the representation for the neuter category. By contrast, adult low proficiency heritage speakers apparently lack neuter gender altogether, and are left with a two-gender system (feminine-masculine). Thus, adult incomplete L1 grammars are simplified versions of the full grammars: the three-way gender classification is reduced to a two-way classification. The second experiment tested the same individuals with a comprehension measure. Feminine, masculine and neuter nouns were presented with a correctly or incorrectly inflected adjective (gender-matching vs. gender incongruent ­conditions).

174 Incomplete Acquisition in Bilingualism

Participants were presented the pairs of words orally through a computer (a noun and an adjective) and were instructed to press a key for YES or NO if the pairs of words heard were an acceptable combination in Russian. Accuracy and reaction times were measured. The Russian control group showed almost perfect accuracy (above 97%). The heritage speakers, by contrast, showed error rates ranging from 11% with masculine to 30% with feminine to 55% with neuter. According to Polinsky, the findings of the second experiment corroborated the split of the heritage speakers into two groups: those with a three-way gender system different from that of L1 children and mature Russian native speakers, and those of lower proficiency with a two-way system that excludes neuter. Since the low proficiency heritage speakers also lacked knowledge of the basic declensional system, it appears that they have reanalyzed the gender system in a different way from speakers who still have some knowledge of case distinctions. Polinsky explained that gender assignment is then assigned by these speakers in a probabilistic fashion, on the basis of morphophonological regularities. Once again, we see in this study how incomplete acquisition and child L1 attrition affect adult linguistic competence and may end up as a reduced, perhaps even different, grammatical system altogether. Reduced input in the heritage language allows these speakers to acquire core, regular, properties of the language (accusative/nominative, feminine/masculine distinctions), while non-core marked properties (neuter gender) are not acquired or simply lost. This is reflected in the loss of marked forms and overapplication of regular rules.

2.2.2 Verbal morphology: Agreement, tense, aspect and mood Let us now examine the long-term effects of incomplete acquisition in the verbal system, namely person and number agreement, tense and aspect marking, and mood morphology. Here again, I will refer chiefly to recent work with Russian and Spanish heritage speakers, since these groups have received the most attention in recent literature. Polinsky’s (1997, 2006) comprehensive description of the grammar of Russian heritage speakers also mentions gradual loss of verbal agreement. The more proficient speakers produce at most 66% of correct subject-verb agreement, while those with the lowest proficiency in the language only produce 30% correct agreement. When speakers make errors, they most often use the third person singular in any tense, infinitive and third person, as shown in (4) and (5). These forms are typically considered defaults in theoretical treatments of morphology (Mc ­Carthy 2007). (4) *moi roditeli oni kupil rugoj dom.   my parents reflexive 3rd pl bought-past 3rd sg another house   ‘My parents bought another house.’



Chapter 6.  Incomplete L1 acquisition in adults 175

(5) *deti guljat’ tam instead of deti guljali tam.   children walk-inf there children walked-pl-past there   ‘The children went for a walk there.’

The main error in (4) is in using the singular verb form in place of the plural. The other error is phonetic: the word for “another” should be drugoj. The correct form (or Full Russian counterpart) is shown in (6). (6) moi roditeli oni kupili drugoj dom. my parents they bought-pl-past another house ‘My parents bought another house.’

In addition to overt agreement morphology, Russian has a complex aspectual system instantiated in preverbal prefixes, and according to Polinsky (1997), the perfective-imperfective morphological opposition is lost in Russian heritage speakers. Most verbs become lexicalized perfectives or lexicalized imperfectives, depending on lexical class. That is, the telic classes – achievements and accomplishments – are lexicalized in the perfective (with a perfective prefix), while the atelic classes – activities and states – are lexicalized in the imperfective (with an imperfective prefix). Examples of these tendencies in American Russian are shown in (7) and (8). (7) kogda my žili v Louisiana ja smogla pročitat’ russkie knigi. when we lived in Louisiana I could-perf read-perf Russian books ‘When I lived in Louisiana I could read Russian books.’ (8)

esli ja xotel exat’ v East Coast moj mat’ ne. if I wanted-subj go-imperf in East Coast my mother not razrešat’ menja let-imperf me-acc. ‘If I wanted to go to the East Coast my mother would not let me.’

American Russian speakers do not use the subjunctive form with by, typical of Full Russian, as in (9). (9) ja ne xochu, čtoby papa tak govoril. I not want-1sg, that-by daddy so spoke-past-sg-masc ‘I don’t want my dad to say that.’

. I thank Tania Ionin (University of Illinois at Urbana-Champaign) for providing the correct Russian forms and explaining the errors to me.

176 Incomplete Acquisition in Bilingualism

They use verbal forms in the past, present, imperative introduced by the complementizer čto instead of the subjunctive complementizer čtoby, as in (10). (10) ja ne xočet čto papa skazat’ tak. I not want-3sg that daddy say-inf so ‘I don’t want my dad to say that.’

Similar effects of incomplete acquisition in the verbal system are reported by Silva-Corvalán (1991, 1994) in her pioneering study of three generations of Mexican-American bilinguals (including adolescents and adults) living in the Los Angeles area. As in Polinsky’s work, the Mexican-Spanish heritage speakers studied by Silva-Corvalán are defined and classified by generation, although there is some overlap of ages. First generation speakers (group I) were born in Mexico and had immigrated to the United States after age 11. These speakers had native command of Spanish, and their command of English ranged from near-native to poor. Speakers of group II were either born in the United States or had arrived before the age of 11. All the U.S. born-bilinguals had been exposed to the two languages from birth, had native command of English, and ranged from near-native to poor in their Spanish ability. All speakers were interviewed orally and were asked to complete fill-in-the-gap questionnaires designed to elicit supplementary information about their choice of tenses. As introduced briefly in Chapter 4, the Spanish verbal system includes three tenses (past, present and future), two aspectual distinctions in the past (perfective and imperfective), two moods (indicative and subjunctive), and the imperative (defective mood). The conditional is the past of the future. There are simple and compound tenses (progressive forms formed with ser/estar, and a present participle and perfect forms formed with haber plus a past participle). Silva-Corvalán documented seven implicationally ordered stages of simplification and loss of this complex verbal system. The forms most affected and lost from the grammars of

. I thank Tania Ionin (University of Illinois at Urbana-Champaign) for providing me with the example in (9). This example also shows that it’s more natural to use ‘govoril’ (spoke) than ‘skazal’ (said) as the American Russian example in (10). The latter does not sound right in this particular context. . In a way, this particular group is problematic for the distinction I intend to make between incomplete learners and L1 attriters, because this group includes individuals with very different onsets of bilingualism (some very early, before the age of 3, and some much later). Furthermore, those adults who arrived to the United States at the age of 11 already received some schooling in their family language in Mexico, whereas those adults born in the United States typically receive schooling in English, unless they attend bilingual schools. Group III speakers were also adults born in the United States, but at least one parent had to be classified as a speaker of group II.



Chapter 6.  Incomplete L1 acquisition in adults 177

heritage speakers – especially those in groups II and III (our incomplete learners) – are the future perfect and the conditional (as a tense), followed by the present subjunctive. There is also simplification, but not complete loss, of the preterite and imperfect tenses in the indicative mood (the perfective-imperfective aspectual ­contrast). Simplification of the preterite (with a close class of stative verbs) occurs before simplification of the imperfect. Some evidence presented by Silva-­Corvalán is the use of imperfect for preterite in perfective contexts, as in (11), and the use of preterite for imperfect in imperfective contexts, as in (12) (11) Yo fui el único hombre que *tenían. (should be tuvieron) ‘I was the only son they had.’ (12) En la casa mi mamá era la única que *habló español y las demás *hablaron en inglés.(should be hablaba and hablaban) ‘At home my mom was the only one who spoke Spanish and the other ones only spoke English.’

Spanish expresses modality in the grammar by means of inflectional morphology for mood (indicative vs. subjunctive). If tense/aspect distinctions are affected under incomplete acquisition, the subjunctive mood is affected even more. Note that this fact is predictable from the order of acquisition aspect of the Regression Hypothesis because subjunctive is acquired by monolingual Spanish children at around age 2;6–3;00, after the past tense is in place, but not fully developed until the age of 12 (Blake 1983), probably because the syntactic, semantic and pragmatic rules that govern the use of indicative vs. subjunctive mood selection in Spanish are very complex. For the purposes of this chapter, I will only focus here on two facts about Spanish subjunctive discussed in the bilingualism literature. A frequent distinction is made between obligatory and optional (or variable) uses of subjunctive in Spanish clauses. The first fact is that the use of subjunctive or indicative depends on the lexical meaning of the subordinating verb. In the case of volitional predicates, as in (13), or of other nominal expressions, as in (14), subjunctive is obligatory, while in the example in (15), indicative (rather than subjunctive) is obligatory. (13) Quiero que vengas/*vienes. I want that you come-subj/*come-indic ‘I want you to come.’

subjunctive

(14) Es importante que tengas/*tienes cuidado. it is important that you have-subj/*have-indic care ‘It is important that you be careful.’

subjunctive

178 Incomplete Acquisition in Bilingualism

(15) Creo que *sea/es verdad. I believe that it *is-subj/is-indic true ‘I believe it is true.’

indicative

The second fact is that, in many other cases, subjunctive selection is optional and grammatical, except that the difference between indicative and subjunctive carries a difference in meaning. One example is the case of restrictive relative clauses in which the choice of subjunctive or indicative varies as a function of presupposition: with indicative, the presupposition of the embedded clause is asserted, whereas with subjunctive there is no presupposition, as in (16). (16) a. b.

Busco un estudiante que habla japonés. indicative-presupposition I am looking for a student that speak-indic Japanese ‘I am looking for a student that speaks Japanese.’ Busco un estudiante que hable japonés. subjunctive-no presupposition I am looking for a student that speak-subj Japanese ‘I am looking for a student that may speak Japanese.’

The Los Angeles speakers studied by Silva-Corvalán frequently replaced subjunctive with indicative in cases where subjunctive is obligatory, as in (17). Furthermore, in contexts where subjunctive use is variable, speakers used indicative exclusively, as in (18) (Silva-Corvalán 1994: 42, ex [33]). (17) I hope que no me toca (PI) la misma problema. (exp.: toque PS) ‘I hope I don’t run into the same problem.’ (D39, f28, 3, ELA42) (18) Quizás vengo mañana. (= venga (PS)) Maybe I come (PI) tomorrow ‘Maybe I come tomorrow.’

Silva-Corvalán (2000) summarizes the percentage use of subjunctive in Los Angeles bilinguals, as shown in Table 6.2. The use of subjunctive (and its replacement by the indicative) decreases dramatically by generation, and this pattern is typically taken as a sign of a simplified linguistic system due to language attrition or loss. When stating that Los Angeles bilinguals have a simplified mood system, ­Silva-Corvalán claimed that “this statement applies to Spanish language production [sic]. Indeed, I have enough evidence to assume that most of the bilingual speakers at the lower levels of the Spanish proficiency continuum understand . PI means present indicative and PS means present subjunctive. The codes given in parenthesis for example (16) are the codes used by Silva-Corvalán to refer to the speaker’s first initial and number, sex, age, generation (1, 2 or 3) and tape (ELA number) where the example occurred.



Chapter 6.  Incomplete L1 acquisition in adults 179

Table 6.2  Frequency of subjunctive use in Los Angeles bilinguals (from Silva-Corvalán 2000).

Group 1 Group 2 Group 3

Obligatory context

Optional context

93.8% 75% 52.5%

30.9% 23.3% 12.4%

Note: Group = generation

(emphasis mine) the meaning of these tense forms as they are used by their interlocutors” (Silva-Corvalán 1994: 26). While Silva-Corvalán calls these cases attrition or loss, I maintain that these simplified grammars may be the result of incomplete acquisition in childhood, since many of the second generation and all of the third generation bilinguals in her study were exposed to Spanish and English from birth. I base this claim on the age of onset of bilingualism of many of the subjects in Silva-Corvalán’s study and on Silva-Corvalán’s (2003) own interpretation of the empirical data from the bilingual children she followed longitudinally. Further recall that in Chapters 4 and 5, I demonstrated how degree of incomplete acquisition correlates with age of onset of bilingualism in childhood. Silva-Corvalán (2003) documented patterns of incomplete acquisition of the tense-aspect and mood systems in bilingual children between the ages of 3 and 5 years. Zentella (1997) also reports similar error patterns with preterite and imperfect in five Puerto Rican children (ranging from ages 6 to 8 and 11) growing up in New York City. Merino (1983) demonstrated lack of development and loss of subjunctive in school-age Chicano children. Since the studies of children show that young bilingual children already lag behind in their command of the tense, aspect and mood system, I assume this type of incomplete knowledge extends well into adulthood. Two recent studies confirm and expand Silva-Corvalán’s original claim that the subjunctive mood is much more affected than the preterite-imperfect aspectual contrast. The first study is Lynch (1999), who followed closely SilvaCorvalán’s methodology to investigate the potential loss of the subjunctive in the speech of 30 Cuban-Americans (first to third generation) in the Miami region. Results showed that all speakers produced subjunctive between 80–100% of the time in obligatory contexts. The lowest percentages come from the second and third generation groups who performed 89% and 81% accurate in volition contexts, respectively. The frequency of subjunctive use was substantially reduced in variable/optional contexts, especially in the second and third generation speakers who used mostly indicative in these contexts. The first generation speakers produced subjunctive between 71–100%, second generation between 52–100% and

180 Incomplete Acquisition in Bilingualism

third generation 0–60%. The use of subjunctive and conditional in hypothetical discourse was even further reduced. Lynch concluded that while there is semantic simplification of the subjunctive in second generation speakers, the subjunctive is clearly lost in all those bilinguals born in the United States, exposed to English and Spanish since an early age, and schooled in English. This confirms the claim about age effects in attrition that I am advancing in this book. The available studies of children as well as Silva-Corvalán’s and Lynch’s studies of adults based their claims on oral production data. This raises the question of whether the lack of subjunctive use by Spanish heritage speakers is just related to patterns of use and preferences rather than lack of linguistic knowledge of subjunctive, at the level of competence. Montrul (2007) is a study that goes beyond documenting productive ability with morphological forms, and probed into the interpretations adult heritage speakers assign to sentences with indicative and subjunctive forms. By investigating whether bilinguals not only produce or fail to produce, but also understand or fail to understand, the meanings of mood morphology, we can obtain a more comprehensive picture of how serious, or dramatic, the extent of subjunctive loss in bilingual speakers really is in given grammatical contexts. Participants were 20 Spanish heritage speakers (mean age 20.1) and a control group of 15 monolingually-raised native speakers of Spanish. The main experimental tasks were a morphology recognition task and a sentence conjunction judgment task. The first task consisted of a passage with blanks for some verbs and tested knowledge of subjunctive and indicative forms in obligatory contexts. The second task focused on the interpretation of subjunctive/indicative morphology in variable contexts (when the use of indicative or subjunctive carries a different meaning) in three conditions: temporal clauses with cuando “when” in habitual contexts, as in (19), purpose clauses with de manera que “such that,” as in (20), and restrictive relative clauses, as in (21). Half of the sentences were in the indicative, and the other half in the subjunctive, and were presented in randomized order together with a 5-point scale (not in minimal pairs as in the examples). Participants were asked to indicate on the scale (where –2 = contradictory and 2 = logical) whether each sentence made sense or not. (19) a. Cada año, Ana se alegra cuando le aumentan el sueldo. b. Cada año, Ana se alegra cuando le aumenten el sueldo. ‘Every year, Ana is happy when her salary is increased.’

logical contradictory

–2      –1      0      1      2

(20) a. El profesor siempre explica ese teorema de manera que todos los estudiantes lo entienden, pero unos pocos estudiantes no lo entienden.  contradictory



Chapter 6.  Incomplete L1 acquisition in adults 181



b.

El profesor siempre explica ese teorema de manera que todos los estudiantes lo entiendan, pero unos pocos unos estudiantes no lo entienden. logical ‘The professor explained the theorem such that all students understand it, but some students don’t understand it.’

(21) a. b.

Necesito un libro de cuentos para niños que tiene ilustraciones de Miró, pero no sé si hay uno. contradictory Necesito un libro de cuentos para niños que tenga ilustraciones de Miró pero no sé si hay uno. logical ‘I need a children’s book with illustrations by Miró, but I don’t know if there is one.’

Results of a proficiency test and the morphology recognition task were highly correlated. Spanish heritage speakers with the lowest proficiency in the language scored the lowest with subjunctive forms in obligatory contexts (r = .85, p < 0.01). Results of the task testing variable uses of subjunctive as a function of pragmatic context showed that while the monolingual Spanish speakers discriminated between logical and contradictory sentences that differed only on the subjunctive or indicative verb form, the heritage speakers showed no statistical discrimination between indicative and subjunctive, as illustrated in Table 6.3.10 The results of this experimental study show that second generation speakers, who may know when to use subjunctive or indicative in obligatory contexts, do not necessarily have the ability to discriminate semantically between subjunctive and indicative in variable contexts, when there is a subtle meaning difference. Findings such as these suggest that problems with subjunctive morphology are not merely superficial, at the level of production. Since comprehension is ­compromised Table 6.3  Semantic and morphological discrimination between indicative and subjunctive forms in Spanish heritage speakers (adapted from Montrul 2007). Group

N

Mood

Cuando

Relative clauses De manera que

Native speakers

15

indicative subjunctive

  1.68 –1.55

–1.4   1.8

–1.61   1.41

Heritage speakers

20

indicative subjunctive

  1.48   0.92

  1.44   1.39

  0.06 –0.16

Note: a mean rating of 2 or close to 2 indicates that those sentences were judged logical or felicitous. A rating of –2 indicates that the sentences were judged as contradictory.

10. See Montrul (2008) for a replication of this study with similar results.

182 Incomplete Acquisition in Bilingualism

even more, it appears that incomplete acquisition may affect linguistic competence as well, not just performance or retrieval of lexical form as in L1 attrition by adults. In conclusion, incomplete L1 acquisition in adult early bilinguals is like fossilization in the L1. It can deeply compromise the domain of grammatical morphology. Incomplete learners display simplification of rules and forms in both nominal and verbal morphology, at the level of production, judgment, and comprehension. The extent of language loss is more dramatic when bilingualism started in childhood, as in the second and third generation speakers, than when it started in adulthood, as in the first generation speakers. If Sorace (2000a) is correct in her proposal that L1 attrition only affects interpretable (semantic) features, apparently, incomplete L1 acquisition affects both interpretable (tense, mood) and uninterpretable (case, gender agreement) features as well. See Table 4.10 in Chapter 4.

2.3

Syntax

Does the simplification attested in the morphology and phonology of incompletely acquired L1 grammars also occur at the level of syntax? We saw in Chapter 3 that the expression of subjects is a feature of the syntax-pragmatics interface that seems affected in L1 attrition. Similarly, we saw in Chapter 4 that English-Spanish and Italian-English simultaneous bilingual children also showed a tendency to overproduce overt subjects in Italian and Spanish, the minority languages, under the influence of English, the majority language. What I will show now is that these effects of apparent cross linguistic transfer in early bilingualism persist well into adulthood under incomplete acquisition. Once again, we turn first to studies of Russian and Spanish heritage speakers. Polinsky (1997, 2006) suggests that the agreement system in American Russian speakers is severely reduced, and this reduction has consequences for the interpretation of pronouns because Russian is a null subject language. In such languages, null subjects are possible because the syntactic and morphological information can be recovered from the rich agreement morphology on the verb.11 By contrast, null subjects in English are typically not possible (except for imperative sentences and diary registers) because the identification information cannot be recovered from the weak or impoverished agreement morphology. Thus, syn11. Despite this tendency, there is no strict correspondence between richness of agreement inflection (or inflectional paradigms) and availability of null subjects cross linguistically. This applies only to some null subject languages (Spanish, Italian, Russian) but not others (East Asian Languages, German, Icelandic, etc.).



Chapter 6.  Incomplete L1 acquisition in adults 183

tactically, subjects in English must be realized overtly. While null subjects are licensed syntactically in null subject languages (and identified morphologically by the verbal inflection), the option to use a null or an overt subject in these languages is also guided by discourse pragmatics. Furthermore, Spanish and Russian allow a variety of word orders in addition to SVO, and these too are regulated by lexical and discourse factors. As a result of the deterioration of the agreement system in heritage Russian speakers, Polinsky (1997) reports that American Russian speakers overuse overt subjects (pronominal or NP) instead of the corresponding null subject in discourse, just like the cases discussed in Chapters 3 and 4 with Spanish, Greek, and Italian bilinguals who speak English. Furthermore, American Russians overuse SV and SVO order, as shown in (22) (22) Oni smotrel kino I oni dumal . . . he watched movie and he thought ‘He was watching the movie and he was thinking about this.’

Coincidentally, Spanish heritage speakers display the same tendencies in their incomplete grammars. Although the agreement paradigm is not as severely reduced in Spanish heritage speakers as in the Russian speakers described by Polinsky (1997), Silva-Corvalán (1994) documents comparable overuse of overt subjects and stricter SVO order in the Spanish of the Los Angeles speakers. Especially those from the second and third generations overused SVO and SV order between 91% and 94% of the time, in clear violation of semantic and discourse pragmatic rules of the language. Further evidence for overuse of overt subjects in Spanish heritage speakers comes from Montrul’s (2004a) experimental study of subject expression in a production task. The overall aim of this study was to show how the incomplete grammars of Spanish heritage speakers show syntactic convergence toward English by losing the pragmatic and semantic layers that regulate the use of subjects (and objects) in Spanish. Based on Sorace’s (2000a) linguistic account of attrition (see Chapter 3), it was hypothesized that if incomplete acquisition affects interface areas of language (semantics and pragmatics) more than the purely syntactic domain, then Spanish heritage speakers should display robust knowledge of null subjects even though English, the dominant language in these speakers, does not share these syntactic properties. Therefore, exclusive use of overt subjects, which would indicate direct transfer from English or complete loss of the null subject parameter, was not expected. If the syntax-semantics and syntax-pragmatics interfaces are vulnerable to change, heritage speakers should display variable behavior with the discourse related distribution of null vs. overt subjects.

184 Incomplete Acquisition in Bilingualism

Table 6.4  Percentage production of null and overt subjects in an oral narrative task by Spanish monolinguals and heritage speakers (adapted from Montrul 2004a).

Monolinguals Advanced heritage speakers Intermediate heritage speakers

N

Overt subjects

Null subjects

20 14 10

40 47 68.6

60 53 31.4

Monolingual Spanish (not from the Caribbean dialect regions) and adult Spanish heritage speakers of Mexican-American background (raised bilingually in the United States since an early age and schooled in English), took a Spanish proficiency test and were then asked to retell a famous children’s narrative in the past. The range of variability of the heritage speakers proved to be quite extensive in the proficiency test (56%–96% accuracy). Overall, the Spanish heritage speakers in this study displayed robust knowledge of the syntax of subjects – a signature of very early and robust L1 acquisition – and there was no evidence for loss of agreement. Furthermore, all heritage speakers produced overt and null subject pronouns, as seen in Table 6.4. However, low proficiency heritage speakers were clearly different from monolinguals and the advanced heritage speakers, both in the percentage of overt and null subjects and in the discourse-pragmatic distribution of null and overt subjects. In particular, these heritage speakers produced more overt subjects than null subjects, whereas the monolinguals averaged fewer overt subjects and more null subjects. The advanced heritage speakers were closer to the monolinguals, producing similar rates of overt subjects and null subjects. Furthermore, there were also errors in the pragmatic use of null subjects. In addition to investigating subject expression in Spanish heritage speakers, Montrul (2004a) also investigated object expression, another grammatical domain vulnerable to incomplete acquisition. Spanish has a very rich system of pronominal object clitics which are marked for case, and agreement (gender and number). Accusative clitics (la) are direct objects, as in (23b), and dative clitics les in (23b) are indirect objects. Both animate and inanimate objects are replaced by the accusative clitics lo and la, as (23b) shows. Animate and specific direct object NPs are preceded by the dative preposition a, as in (23a), which is an instance of differential object marking or, according to Torrego (1998), marked accusative case. The distribution of the preposition a with animate direct objects appears to be sensitive to definiteness/specificity and lexical aspect of the verb and animacy/ agentivity of the subject (Torrego 1998).



Chapter 6.  Incomplete L1 acquisition in adults 185

(23) a. b.

Patricia vio a mi abuela/la película. Patricia saw to my grandmother/the movie ‘Patricia saw my grandmother/the movie.’ Patricia la vio. Patricia it saw ‘Patricia saw it.’

(24) a. b.

Mariano mandó una carta a mis amigos. ‘Mariano sent a letter to my friends.’ Mariano les mandó una carta (a mis amigos). Mariano them sent a letter (to my friends) ‘Mariano sent them a letter (to my friends).’

Indirect objects (24a, b) are prepositional phrases, always preceded by the dative preposition a, and replaced by the dative clitic le/les. Clitic doubling, the co-occurrence of the clitic and the prepositional phrase, is a possible option in all Spanish dialects (24b). With indirect objects, dative case is an instance of structural case. Other Spanish dative constructions take obligatory clitic doubling. The clitic in these cases is not related to syntactic dative case (indirect object), but appears to have a semantic function (related to aspect or other semantic notions). In some of these cases, the prepositional phrase bears the role of locative (25), possessor (26), as in possessor-raising constructions. Semantically-based case is an instance of inherent dative case. (25) Olga les echó agua a las plantas./*Olga echó agua a las plantas. Olga dat-cl threw water to the plants ‘Olga watered the plants.’ (26) Cecilia le lavó las manos a María/ *Cecilia lavó las manos a María. Cecilia dat-cl washed the hands to María ‘Cecilia washed María’s hands.’

To summarize, structural dative and accusative case (for indirect and direct objects) are syntactic, while realization of the preposition a with animate direct objects and the obligatory dative clitic of inalienable possession constructions are cases of inherent (lexical, semantically based) accusative and dative case, respectively. Montrul (2004a) looked at the production of accusative and dative clitics, clitic doubling with inalienable constructions, and the use of a-personal in the oral production of Spanish heritage speakers. With respect to the production of accusative and dative clitics, the results showed solid command of this aspect of Spanish. Monolinguals and heritage speakers produced comparable rates of accusative and dative lexical and pronominal objects, as shown in Table 6.5.

186 Incomplete Acquisition in Bilingualism

Table 6.5  Percentage production of accusative and dative clitics in an oral production task (adapted from Montrul 2004a).

Monolinguals Advanced heritage speakers Intermediate heritage speakers

N

Accusative Clitics Dative Clitics Clitic doubling

20 14 10

97.8 98.6 96.7

79.3 80.9 80

20.6 19.1 20

These results replicated those reported by Silva-Corvalán (1994) with Los Angeles speakers, and confirm the theoretical observation that aspects of syntax are resilient to change, transfer of some sort, or parameter unsetting despite the fact that the contact language, English, has no clitics. Thus, this study shows how adult incomplete L1 learners have the parameters of their L1 set since childhood, despite error rates in agreement morphology or pragmatic distribution of subjects and objects. This is consistent with the Weaker Language as L1 Hypothesis that I offered in Chapter 4, §5.2. With respect to differential object marking (the preposition a with direct objects) and inalienable possession constructions, Montrul’s results confirmed the original observations of Luján and Parodi (1996) and Silva-Corvalán (1994). Spanish heritage speakers displayed high rates of omission of the preposition a with animate direct objects (6% advanced speakers, 21.3% intermediate speakers), and produced few structures with inalienable possessor NPs containing a clitic (9.5% vs. 29.2% produced by the monolinguals and 30.5% by the advanced heritage speakers).12 Unlike the monolinguals, low proficiency heritage speakers used significantly more structures with no doubled PP. Both the realization of the preposition a with animate direct objects and the possibility of clitic doubled possession constructions are regulated by lexical or semantic accusative or dative case, respectively. These are the semantic features that presumably undergo erosion in these grammars, leading to morphosyntactic convergence with English in the grammatical expression of direct objects and possession constructions. By losing inherent case, the incomplete grammars of adult heritage speakers are, once again, simplified with respect to the grammars of full speakers. In conclusion, the patterns of incomplete acquisition observed in Russian and Spanish heritage speakers suggest that subject and object expression are affected in these reduced L1 grammars. While the null subject option of the parameter is not unset altogether, and heritage speakers have solid knowledge of clitics, there is erosion of pragmatic rules of subject expression and semantic conditions 12. Montrul & Bowles (in press a) is another recent study documenting incomplete knowledge of differential object marking in other groups of Spanish heritage speakers.



Chapter 6.  Incomplete L1 acquisition in adults 187

that regulate inherent, as opposed to structural, case. The question that arises is whether these effects can be the result of direct transfer, structural convergence with English, or whether it is limited input which allows only unmarked, less complex options to be fully developed, regardless of the nature of the contact majority language. Since the contact language is English in the cases discussed above, the two explanations converge on the same result. Clearly, more work is needed on heritage languages in contact with languages other than English, and which are equally or more morphologically and syntactically complex than the heritage language.13 In essence, these studies show that core syntax is acquired quite solidly with reduced input in childhood, while non-core properties of language at the interfaces may need more sustained amounts of input throughout middle childhood. Still, basic linguistic knowledge in adult heritage speakers, even if not fully developed into adulthood, was acquired in childhood like an L1. This is the claim of the Weaker Language as L1 Hypothesis. Having now covered structural aspects of incomplete acquisition in phonology, morphology and syntax, I turn to the lexicon next.

2.4

Lexicon

There are two assumptions in the studies presented so far. The first is that the heritage language in early bilingual speakers of a minority language is incomplete or reduced in comparison to the grammars of fully competent speakers. The second is that competence in the minority language is weaker than competence in the majority language in these adult early bilinguals. In many of the studies presented so far, however, it was assumed that the minority language of these speakers was weaker (in terms of linguistic competence and complete acquisition) than their majority language, although this was not actually demonstrated empirically. In this section, I discuss two studies examining the linguistic and psycholinguistic status of the two languages in the minds of these speakers. First, I return to the study conducted by Hulsen (2000), who tested lexical knowledge and processing in three generations of Dutch speakers living in New Zealand. The results of first 13. A recent example of this line of work is J. H. Kim (2007), who tested binding interpretations in Korean heritage speakers in the U.S. and Korean heritage speakers in China. Even though the Korean system of long-distance anaphors is closer to the Chinese system than to the English system, Kim found no overall differences between the Korean heritage speakers with respect to Chinese and English as contact language. The patterns of simplification attested were similar in the two groups. This suggests that heritage language may undergo universal linguistic processes of simplification under reduced input conditions.

188 Incomplete Acquisition in Bilingualism

generation speakers and a control group of Dutch speakers from the Netherlands were discussed in Chapter 3 in the context of language attrition in adult immigrants. Second, I will discuss a study that investigated knowledge and processing of lexical semantics in Spanish heritage speakers. What emerges from these two studies is that the weaker language is indeed weaker than the stronger language at the level of retrieval and access, but less so at the level of lexical syntactic competence, confirming, once again, that aspects of syntactic knowledge acquired during the period of early syntactic development are unlikely to be disrupted by fluctuations in input later on. To be part of the second and third generation groups in Hulsen’s (2000) study, participants had to have arrived in New Zealand before the age of 4, or had to have been born in New Zealand. As a result, 86.7% of the second generation speakers were born in New Zealand, and many of the second and third generation speakers were related to the first generation speakers. The mean age at the time of testing for the second generation speakers was 38 (17–45) and 17 for the third generation speakers (range 12–27). A picture-naming and a picture-matching task were used to investigate the heritage Dutch speakers’ knowledge of words in Dutch and English. In the picture-matching task, participants saw a picture with a word underneath. They were to indicate as quickly as possible whether the word was the correct name for the picture displayed. Half of the words matched the picture presented, for example, a picture of a kangaroo with the word kangaroo (yes condition). The other half did not match, showing the picture of the kangaroo with the word toe (no condition). (For more details on the experiment refer to Chapter 3.) Predictions for the second and third generation speakers were that self-ratings for proficiency in Dutch would be lower than for English. For the processing experiments, Hulsen predicted that second and third generation speakers would perform better in English than in Dutch (while it must be recalled from our discussion in Chapter 3 that first generation speakers would perform better in Dutch than in English). High frequency and cognate words would be remembered best and recognized the fastest, while shorter reaction times were expected in the matching task than in the production task. Young arrivals would differ from the Dutch control and first generation groups in terms of accuracy in the Dutch experiments, and they would be slower in their reaction times. By contrast, these speakers would be more accurate and faster in the English experiments. Results showed important differences in self-perceived Dutch proficiency in all the groups. On a scale of 1–5, the mean rating of Dutch and English for the first generation speakers was 4.58 and 4.69 respectively. Self-rating proficiency for the second generation speakers was 2.60 for Dutch and 4.96 for English, while those for the third generation were 1.29 for Dutch and 4.74 for English. Table 6.6



Chapter 6.  Incomplete L1 acquisition in adults 189

Table 6.6  Percentage Accuracy and RTs in the picture-naming and picture-word matching tasks in English and Dutch (adapted from Hulsen 2000).

Language Dutch

English

Group control Gen 1 Gen 2 Gen 3 Gen 1 Gen 2 Gen 3

Picture naming task

Picture-word matching task

Accuracy 95% 89% 52% – 92% 98% 97%

Accuracy 99% 98% 90% 73% 98% 93% 98%

RT 1163 ms 1205 ms 1910 ms – 1168 ms   905 ms   932 ms

RT 1126 ms 1150 ms 1328 ms 1676 ms 1228 ms   951 ms   947 ms

s­ ummarizes the overall accuracy and reaction time means for all the groups, including the first generation and Dutch control group for comparison. First, it is important to point out that the picture-naming task proved too difficult for the third generation speakers, and the results had to be discarded. This already indicates that Dutch proficiency in this group was seriously compromised. Performance on the picture-matching task was better than on the picture-naming task, suggesting that production is affected more by incomplete acquisition than comprehension. The results of the Dutch experiments showed main effects for generation (first generation better than second and third), cognate status of words (cognate easier and faster than non-cognate), and frequency (high frequency words recognized faster than low frequency words). The results of the English experiments also showed significant main effects by generation, cognate status and frequency. Compared to the first generation and compared with their own English language proficiency, the second generation group showed a decrease in Dutch productive skills but stable knowledge in receptive skills. Only a few subjects in the third generation group were able to complete the picture-naming task in Dutch, due to their weak command of the language. As for the picture-matching task, the third generation speakers were better in English than in Dutch. Thus, at least for lexical retrieval and access, Hulsen found that, in fact, Dutch is the weaker language both in terms of use and speed of access in incomplete learners of Dutch. Incomplete acquisition, as demonstrated from the second and third generation speakers, affects lexical retrieval and production more dramatically than L1 attrition in first generation speakers. Similar findings with low proficiency heritage speakers are reported by Polinsky (2005). This is consistent with my claim about age effects in L1 attrition. But do difficulties retrieving and producing lexical items reach the level of grammatical competence at the lexical-semantic level in the adult weaker language? Investigating other aspects of lexical competence, Montrul (2006) focused

190 Incomplete Acquisition in Bilingualism

on knowledge and processing of unaccusative and unergative verbs in Spanish heritage speakers born and schooled in the United States. Potential incomplete acquisition in Spanish was assessed with respect to the grammar of monolingual Spanish speakers on the one hand, and with respect to the bilinguals’ own competence in English – their perceived stronger language – on the other. Unaccusativity classifies intransitive verbs into unergatives (27a) and unaccusatives (28a), depending on the semantic characteristics of the sole argument (more or less agentive). unergative

(27) a. John walked. b. [John [VP walked ]]

unaccusative

(28) a. John arrived. b. [ e [ VP arrived John]] c. [Johni [VP arrived ti]]

The unaccusative/unergative distinction is universal (i.e., all languages have unaccusative and unergative verbs), but languages vary as to the syntactic reflexes of unaccusativity, or the specific constructions in which these verbs can appear. In Chapter 2 we discussed Sorace’s (1993) study of unaccusativity in Italian and saw that Italian uses different perfective auxiliaries (essere and avere) with these verbs. For generative linguists, the distinction is syntactic (Burzio 1986; Rosen 1984) because the argument of unaccusative verbs behaves syntactically like the object of a transitive verb. A number of constructions in English appear to be sensitive to the syntactic distinction between the two classes of verbs. Since resultatives can only be predicated of the object, they are possible with transitive (The blacksmith pounded the metal flat) and unaccusative verbs (which have an object), as in (29), but impossible with unergatives, as in (30), whose sole argument is a subject. unaccusative

(29) The book broke apart. (30) *At his wedding, Peter sang

sore.14

unergative

In Spanish, a syntactic construction that discriminates between unaccusative and unergative verbs, is the use of postverbal bare plural subjects (generics), which are grammatical with unaccusative verbs, as in (31), but ungrammatical with unergative verbs in (32). (31) Han pasado camiones. have passed trucks ‘Trucks have passed by.’

unaccusative

14. This example is ungrammatical with a resultative meaning: Peter is sore as a result of singing. It may be acceptable with a depictive meaning: Peter was already sore when he was singing.



Chapter 6.  Incomplete L1 acquisition in adults 191

(32) *Han dormido animales.   have slept animals   ‘Animals have slept.’

unergative

A second syntactic test discriminating between unaccusative and unergative verbs in Spanish is the absolutive construction (de Miguel 1992). Absolutive constructions are grammatical with unaccusative participles, as in (33), but clearly ungrammatical if the participle is an unergative verb, as in (34). (33)   Muerto el perro, se acabó la rabia.   dead the dog the rabies stopped   ‘Once the dog was dead, the rabies stopped.’ (34) *Nadado Juan, se sintió mejor.   swam Juan he felt better   ‘Once Juan swam he felt much better.’

unaccusative

unergative

The participants in Montrul’s (2006) study were monolingual Spanish native speakers (mean age 31.03), adult Spanish heritage speakers (adult simultaneous bilinguals) (mean age 21.54), monolingual American English speakers. Knowledge of the semantic and syntactic distinction between the two classes of verbs was tested through a grammaticality judgment task and an on-line visual probe recognition task in the two languages. The heritage speakers were tested in English and in Spanish (in separate sessions for each language, two weeks apart) using the same instruments but with different sentences in the two languages. The grammaticality judgment task consisted of 9 unaccusative and 9 unergative verbs presented in all the constructions that were sensitive to the syntactic distinction between the two types of verbs. Within the unaccusative and unergative verbs were 3 verbs of different semantic subclasses representing Sorace’s (2000c) unaccusative/unergative semantic hierarchy. Subjects were asked to judge grammatical and ungrammatical sentences on a 5-point Likert scale. The English version of the test used some of the same verbs in the Spanish test, except for constructions that required specific verbs, such as resultatives, cognate object constructions and pseudopassives. The heritage speakers completed a proficiency test in each language, and the results confirmed that ability in Spanish (86.72% accuracy) was slightly weaker than in English (91.52%). On a 5-point self-rating scale for each language, the bilinguals perceived Spanish as weaker than English (mean ratings 4.01 for Spanish and 4.88 for English). The prediction was that accuracy on the two main tasks testing unaccusativity would be lower in Spanish than in English for the heritage speakers, reflecting the proficiency scores and self-ratings. Tables 6.7 and 6.8

192 Incomplete Acquisition in Bilingualism

Table 6.7  Mean grammaticality judgments on Spanish unaccusative and unergative verbs (adapted from Montrul 2006). Groups

N Absolutive construction Bare plural postverbal subjects Unaccusative Unergative Unaccusative Unergative (gram.) (ungram.) (gram.) (ungram.)

Spanish native speakers 28 3.93 Spanish heritage speakers 18 2.79

1.25 1.56

4.77 4.4

2.98 2.92

Table 6.8  Mean grammaticality judgments on English unaccusative and unergative verbs (adapted from Montrul 2006). Groups

N Cognate Objects Pseudopassives Resultatives Unacus. Unerg. Unacus. Unerg. Unacus. Unerg. (ungram.) (gram.) (ungram.) (gram.) (gram.) (ungram.)

English native speakers 21 1.95 Spanish heritage speakers 18 2.32

4.6 4.53

1.33 1.31

2.92 2.58

3.76 3.68

2.7 3.26

summarize critical experimental sentences in the English and Spanish grammaticality judgment tasks. Contrary to expectations, results showed that syntactic knowledge of unaccusativity was quite robust in the bilinguals’ two languages: bilinguals distinguished reliably between unaccusative and unergative verbs in the different constructions, largely patterning with the adult monolinguals in the two languages. A likely reason for why the syntactic knowledge underlying the lexical distinction between these two types of intransitive verbs might be so resilient is that it is assumed to develop quite early in life, before age 4, and might not be highly dependent on fluctuations in the amount of input or the development of literacy skills during the school years. However, there was individual variation from monolingual Spanish speakers in the semantic classification of some verbs along the unaccusativity continuum (Sorace 2000c).There were also crosslinguistic effects with some constructions and some verbs in the two languages. See the study for details. Montrul’s (2006) results differ somewhat from those of Hulsen’s in that Hulsen found much more reduced command of Dutch than English in her second and third generation Dutch heritage speakers. Several reasons may account for the differences between the two studies. One has to do with amount of contact with the language. From what Hulsen described, most of the Dutch speakers she tested were married to English speakers and did not use Dutch very frequently. By contrast, Spanish heritage speakers in the United States continue to use Spanish in their daily lives due to the size of the Spanish-speaking community. Second, the



Chapter 6.  Incomplete L1 acquisition in adults 193

results of the proficiency test showed that even though the bilinguals in Montrul (2006) scored lower in Spanish than in English, they still scored at the advanced level in Spanish, well above 80% accuracy. Finally, the main distinction relates to the grammatical domain investigated. While Hulsen looked at access and retrieval of nouns, Montrul instead investigated a semantic and syntactic universal distinction. Syntactic knowledge is more resistant to incomplete acquisition than lexical knowledge. The critical period for acquisition and loss applies to the grammar (computational system) but not to the lexicon.

3.

Incomplete L1 acquisition and age of onset of L2 acquisition

Up until this point, we have been describing the grammatical endstate of the minority language in adult heritage speakers who are second generation immigrants. Yet, it is in this second generation where heritage speakers display the greatest variability, and they are a very heterogeneous group. While some are very proficient in the heritage language, others are receptive bilinguals who seldom speak the language, let alone read it. As already discussed at length, there are myriads of sociolinguistic and family circumstances that certainly account for the wide range of variability in linguistic skill observed in this group. In addition to whether only one or both parents speak the heritage language at home, one factor that appears to play a crucial role in determining degree of proficiency attained in adulthood is, once again, age of onset of bilingualism. At the end of Chapter 2, I formulated the following hypothesis: Hypothesis 2 If L1 attrition occurs in early (pre-puberty) bilingualism, it would be more severe in simultaneous bilinguals (exposed to the two languages very early) than in sequential bilinguals (when the L1 was acquired before the L2). In Chapters 4 and 5, we saw that simultaneous bilingual children from immigrant families run a higher risk of developing incomplete linguistic knowledge of the family language than children whose exposure to the majority language starts after the age of 5 or later. The longer a child had a chance to receive more extensive amounts of input and to use the language, the higher the degree of language maintenance later on. Time/length of exposure and language use both contribute to solidify linguistic competence at a young age. I will now show how the overall validity of this hypothesis carries over into adulthood. To illustrate the long-lasting effects of early L2 exposure into adulthood, I turn to three studies that included bilinguals with different ages of onset

194 Incomplete Acquisition in Bilingualism

of bilingualism in childhood. In §3.2, I revisit the case of foreign-born adoptees, or adults who were completely cut off from their L1 early in childhood.

3.1

Incomplete acquisition in simultaneous vs. sequential bilinguals

Flege, Yeni-Komshian and Liu (1999) and Yeni-Komshian, Flege and Liu (2000) are two related studies that examined pronunciation and morphosyntactic proficiency in both the first (Korean) and second language (English) of 240 Korean heritage speakers. The Korean participants arrived in the United States between the ages of 1 and 23 years, and age of arrival (AoA) was used as the index of age of L2 learning. They all spoke Korean at home. They were divided into 10 subgroups (n = 24) based on AoA (1–3, 4–5, 6–7, 8–9, 10–11, 12–13, 14–15, 16–17, 18–19, 20–23). Groups 1-3 and 4-5 had not received schooling in Korea, while the rest of the groups had. All participants were asked to produce 5 English and 5 Korean sentences and repeat each sentence three times. Ten monolingual native listeners from each language group (English and Korean) listened to the pronunciation of the sentences and assigned ratings from 9 (no accent) to 1 (very strong accent). As all studies of age effects have found, mean English pronunciation as judged by native speaker judges was negatively correlated with AoA. Seventeen individuals (7%) were rated as native speakers, and all of them had AoA between 1 and 8 years. As AoA increased, mean pronunciation in English decreased. But what is important for the purposes of this chapter is what happens with the Korean pronunciation of the earlier and later Korean arrivals in the United States, and the opposite trend obtained. The poorest pronunciation ratings were assigned to the earliest arrivals (AoA 1–7), whose scores were significantly lower than all other groups. These are the individuals who also did not receive schooling in Korean. However, the scores of all other Korean groups were also significantly lower than those of the Korean monolingual group. Only bilinguals whose age of arrival was greater than 12 years were judged to pronounce Korean with native-like ability. When the scores for each participant in the two languages were compared, results showed that for most Korean heritage speakers, pronunciation ratings in Korean and English were inversely related (the better their English, the worse their Korean, and vice-versa). Interestingly, the group that did not show a significant difference between English and Korean pronunciation were those with AoA 10–11. Recall from Chapter 5 that this is the approximate age at which bilingual balance begins to tip, as shown from studies of lexical production and comprehension (Kohnert, Bates & Hernández 1999). Table 6.9 shows subjects z-scores on L1 and L2 pronunciation (where the mean is 0, positive scores are above the mean, and negative scores below the mean).



Chapter 6.  Incomplete L1 acquisition in adults 195

Table 6.9  Four groups representing different patterns of L1 and L2 proficiency (mean z-scores)(adapted from Yeni-Komshian, Flege & Liu 2000).

English Korean

E– K+ (n = 104)

E+ K– (n = 83)

E+ K+ (n = 37)

E– K– (n = 16)

–0.90   0.73

  1.02 –1.12

0.55 0.65

–0.77 –0.43

E+/K+ = above average in English/Korean pronunciation; E–/K– below average in English/ Korean pronunciation

The 83 participants whose English was excellent but whose Korean was below average (E+ K–) were all about 6 years old when they came to the United States, had missed schooling in Korea but received 15 years of schooling in the US and had resided in the US for 17 years. These are incomplete learners. The 37 individuals who had excellent pronunciation in the two languages immigrated around 11 years of age, when the L1 is considered solidified, perhaps, and had 5 years of schooling in Korea and 12 in the United States. It should be noted that while pronunciation in their L1 (Korean) and their L2 (English) may have been excellent, none of these 37 individuals showed native-like pronunciation. Only two participants (AoA 5 and AoA 8) had L1 and L2 pronunciation scores within the range of scores given to monolinguals (Yeni-Komshian, Flege & Liu 2000, p. 145). As with the studies on children by Kohnert, Bates and Hernández (1999) and Jia and Aaronson 1999, 2003) discussed in Chapter 5, not only does this study show differences between degree of affectedness in the L1 of children exposed to the second language early in life, but it also shows bilingual balance at the same time (9–11 years of age). Furthermore, this study shows that bilinguals older than 12 are not susceptible to attrition, incomplete acquisition, or L2-induced changes in the grammar of their L1. At least for the realm of pronunciation, this study does not mesh well with the Critical Period Hypothesis, since once acquired in early childhood, pronunciation should not be affected by learning other languages later in life. However, if pronunciation, as Flege et al. describe it, is a surface phonetic phenomenon, this study does not show that actual phonological categories and contrasts acquired early in childhood (i.e., phonological mental representations) are no longer detectable in adulthood, in bilinguals with early exposure to the L2. This study shows that the earlier the onset of bilingualism, the more affected the pronunciation of the L1. Let us now consider how some aspects of morphology and syntax relate to age of onset of blingualism. Montrul (2002) attempted to tease apart cases of attrition vs. incomplete acquisition in Spanish heritage speakers in aspect marking. Work by Polinsky (1997) and Silva-Corvalán (1994) suggests that incomplete

196 Incomplete Acquisition in Bilingualism

a­ cquisition manifests itself in the non-targetlike production of aspecual morphology. But rather than focusing exclusively on production, Montrul (2002) examined the morphological and semantic interpretations that heritage speakers with different onsets of bilingualism assign to tense/aspect morphemes. If incomplete acquisition is manifested at the level of interpretation, then this might suggest that incomplete acquisition affects linguistic representation and not just retrieval during production. Participants were native Spanish speakers and Spanish-English bilinguals from the Chicago area, along with bilinguals who were born in Latin America and moved to the United States in late childhood (8–12). The mean age for all bilinguals at the time of testing was 22.43. For the purposes of this study, all the bilinguals were divided into 3 groups based on onset of bilingualism: simultaneous bilinguals (n = 16), with onset of bilingualism between birth and age 3; early child L2 learners (n = 15), with onset of bilingualism between ages 4–7; and late child L2 learners (n = 8), with onset of bilingualism between ages 8–12. If age of onset of bilingualism determines degree of incomplete acquisition of the heritage language, then simultaneous bilinguals should display more incomplete knowledge of Spanish (and transfer from English) than early child L2 learners and late child L2 learners. Two tasks tested oral production and written recognition of preterite and imperfect verbal forms in narratives. The other two tasks tested the interpretation of aspect morphology in a sentence conjunction judgment task and in a truth value judgment task. In one of the written tasks, participants were presented with a cloze passage in the past and had to choose the correct form of the verb (preterite or imperfect). The monolingual control group scored 96.5% on this task and there were significant differences between that group and the simultaneous bilinguals, as predicted, who scored 85%. The early child L2 learners and the late child L2 scored 91.33% and 92.06% accurate, respectively. In the oral narratives, the percentage distribution of verb tenses was calculated by lexical class (use of preterite and imperfect with accomplishment, achievement, activities and stative verbs). The frequency distribution of tenses by verbs was nearly identical in the 4 groups. Around 95% of imperfect forms were used with states and activities, and preterite forms were mostly used with accomplishment and achievements. As for accuracy, the monolinguals performed at ceiling (100%) as shown in Table 6.10, while the bilingual groups performed above 90%. The bilinguals performed close to 100% in many conditions, but those prone to errors were stative verbs in the preterite and achievements in the imperfect. The simultaneous bilinguals were 77.14% accurate with stative verbs in the preterite



Chapter 6.  Incomplete L1 acquisition in adults 197

Table 6.10  Oral production task. Percentage accuracy on preterite and imperfect by predicate type (from Montrul 2002)

aspect

Native speakers Spanish heritage speakers (n =20) Simultaneous Early child L2 Late child L2 bilinguals (n = 16) learners (n = 14) learners (n = 8) Pret. Imperf. Pret. Imperf. Pret. Imperf. Pret. Imperf.

Predicates states activities accomplishments achievements total

100 100 100 100 100

Overall total

100

100 100 100 100 100

  77.14 100   91.4   91.1   89.9   92.66

  98.42 100 100   83.3   95.43

  98.27 100 100 100   99.56   95.62

100   95.34   85.71   85.71   91.69

  71 100 100 100   92.75

100 100 100 100 100

  96.37

and 83.3% accurate with achievements in the imperfect. The late child L2 learners were also 71% accurate with stative verbs in the preterite, while the early child L2 were 85% accurate with achievements in the imperfect. Semantic interpretations of the Spanish aspectual markers were tested in the sentence conjunction judgment, where subjects had to judge on a 5-point scale the combinatorial felicity of two conjoined clauses. The example in (35a) shows that the imperfect in the first clause is compatible with the negation of the event expressed in the second clause, since the event is not viewed as bounded. However, the preterite in the first clause (35b) makes the second clause contradictory. The test included minimal pairs with accomplishments, achievements, state predicates and distracters presented in randomized order. (35) a.  b.

Pedro corría la maratón de Barcelona pero no participó. non-contradictory Pedro ran-impf the marathon of Barcelona but he did not participate Pedro corrió la maratón de Barcelona pero no participó. contradictory Pedro ran-pret the marathon of Barcelona but he did not participate ‘Pedro ran the Barcelona marathon but he did not participate in it.’

The truth value judgment task focused on other subtle differences between preterite and imperfect forms. Namely, the aspectual shift of some stative verbs with the use of preterite and imperfect illustrated in (36); the one-time event vs. habitual meaning of preterite and imperfect, as in (37); and the generic or specific interpretation of impersonal pronouns as a function of the preterite or imperfect form of the verb in (38). These sentences were presented in the contexts of stories.

198 Incomplete Acquisition in Bilingualism

(36) a. b.

Juan sabía la verdad. Juan know-impf the truth ‘Juan knew the truth.’ Juan supo la verdad. Juan know-pret la verdad ‘Juan found out the truth.’

stative

(37) a. b.

Pedro robó algo en el autobús. Pedro robbed-pret something in the bus Pedro robaba algo en el autobús. Pedro robbed- impf something in the bus ‘Pedro robbed something in the bus.’

one-time event

(38) a. b.

Se comía bien en este restaurante. se eat-impf well in this restaurant ‘One/We would eat well in that restaurant.’ Se comió bien en este restaurante. se eat-pret well in this restaurant ‘We ate well in that restaurant.’

generic/specific

eventive

habitual or progressive

#generic/specific

Group results for the sentence conjunction judgment task showed a statistically significant difference by group, and by aspectual form (preterite and imperfect) with the three classes of verbs for all the bilingual groups. Achievements in the imperfect and statives in the preterite received the lowest ratings, especially by the simultaneous and early child L2 groups. Individual results (Table 6.11) showed that 50% of the late child L2 learners showed a significant contrast with all the predicates like the monolingual native speakers, while only 13.33% of the early child L2 and 12.5% of the simultaneous bilingual group did so. Another 13.3% of the early child L2 and 12.5% of the simultaneous bilingual groups did not have the semantic contrast with any of the predicates, a sign of language loss or incomplete acquisition. The results of the truth value judgment task showed that the simultaneous bilinguals and the early child L2 learners were less accurate than the monolinguals and the late child L2 learners. Conditions most problematic for the early bilinguals were change of meaning stative predicates in the preterite (sabía vs. supe “I knew” vs. “I found out”), the imperfect in habitual stories, and the specific interpretation of the imperfect with impersonal subjects. According to the individual results, a third of the subjects in the early bilingual groups (simultaneous and early child L2) did not perform within the range of variation of native speakers, whereas most of the late child L2 bilinguals did. In conclusion, the main finding of this study was that age of onset of bilingualism plays a crucial role in incomplete acquisition or maintenance of the ­family



Chapter 6.  Incomplete L1 acquisition in adults 199

Table 6.11  Sentence conjunction judgment task. Number of individuals who had statistically significant contrasts between preterite and imperfect sentences (Montrul 2002) Native speakers (n =20) count percent 17 85%

Simultaneous bilinguals (n = 16) count percent 2 12.5%

Early child L2 learners (n =14) count percent 2 13.33%

accomp and achiev. sig accomp and states sig. states and achiev sig

 1

1

1

  6.66% 2

25%

1

  6.66% 1

12.5% 12.5%

accomplishments sig. achievements sig. states sig. no single contrast sig.

all contrasts sig.

  5%

 0  2

  6.25%

0 10%

Late child L2 learners (n = 8) count percent 4 50%

3

18.75%

3

20%

1

 0

5

31.25%

2

13.33% 0

 0  0

1 2

  6.25% 12.5%

1 3

  6.66% 0 20% 0

 0

2

12.5%

2

13.33% 0

language. There is evidence of incomplete acquisition in the early bilingual groups (simultaneous and early child L2) which affects the production and interpretation of tense/aspect contrasts with certain verbs: statives in the preterite and achievements in the imperfect. Studies of monolingual acquisition suggest that these are the last stages of acquisition of tense/aspect in children. These results show that this stage of acquisition is perhaps never completed or reached in early childhood and remains incomplete into adulthood. It appears then that this is a grammatical domain likely to remain fossilized in early bilinguals. By contrast, results of the late child L2 learners, who arrived in the United States after age 8 and had opportunity to receive schooling in Spanish in their home countries, were indistinguishable from adult Spanish native speakers. This finding underscores the roles of L1 proficiency and schooling in potential vulnerability to loss, as discussed in Chapter 5. Another recent study that speaks to the relationship between age of onset of bilingualism and degree of incomplete acquisition is Kim, Montrul and Yoon’s (under review) experiment with Korean heritage speakers living in the United States. The study focused on the interpretation of anaphors (reflexive pronouns), since previous findings on heritage speakers suggest that anaphor binding is also affected under incomplete acquisition (Polinsky 1997). Korean is an interesting language for investigating reflexive anaphors under incomplete acquisition

200 Incomplete Acquisition in Bilingualism

­ ecause it possesses three types of reflexives – caki, casin and caki-casin – which b differ in their distributional and interpretive properties. According to a corpus study by Kang (1998), caki is subject-oriented and prefers long-distance (LD) antecedents. On the other hand, caki-casin, which is less frequently used than ­caki, requires a local antecedent. The third reflexive casin is used less frequently than caki and allows both local and LD- antecedents. Studies on L1 transfer in L2 acquisition of binding showed L2 learners whose L1 has only local binding (such as English) have difficulty interpreting reflexives that can be bound long-distance in languages like Japanese and Chinese (Hirakawa 1990; Thomas 1995; White, Hirakawa & Kawasaki 1996; Yuan 1998). In Chapter, 2 we also discussed how L1 binding interpretations can be affected in L1 attrition in postpuberty immigrants, as shown by Gürel’s (2002) study of Turkish. Kim, Montrul and Yoon asked whether the influence of English, a language with local anaphors, would be apparent in the binding interpretations of long-distance reflexives by Korean heritage speakers, such that Korean-English bilinguals whose Korean was weaker than their English would show a heightened preference for local, as opposed to LD, binding in Korean as well. It was also hypothesized that because English has only one type of reflexive (himself), while Korean has three (caki, casin and caki-casin), the interpretive distinctions among the three Korean reflexives, might be lost or neutralized to some degree in the bilinguals, regardless of reflexive type. Participants were 50 Korean Heritage speakers living in the United States (mean age: 21.4). Twenty-one were late bilinguals (age of arrival 11–15 yrs; length of residence 4–9 yrs) and the other twenty-nine were early bilinguals exposed to both languages simultaneously since birth. In addition, 68 monolingual speakers of Korean were tested as the control group in Seoul, Korea. A Korean proficiency test showed that early bilinguals performed at around 56% accuracy, while the late bilinguals and the monolingual controls performed at over 90% accuracy. The main task was a truth value judgment task with pictures (as used in White et al., 1996). Target items consisted of bi-clausal sentences illustrating each type of reflexive (caki, casin, caki-casin). According to the main findings, long-distance binding preferences are affected in early bilinguals. With the reflexive caki, which allows both local and long distance binding, early bilinguals preferred the local interpretation, while the late bilinguals and the control group accepted both interpretations. The results of the other two anaphors, casin and caki-casin, did not differ significantly among the groups. However, while the monolinguals and the late bilinguals retained a threeway anaphor system, results of the early bilinguals showed that they treated casin and caki-casin alike. They seem to have a reduced two-anaphor system. This is illustrated in Figure 6.1.



Chapter 6.  Incomplete L1 acquisition in adults 201

Figure 6.1  Acceptance percentage of local and long-distance binding with three Korean anaphors (adapted from Kim, Montrul & Yoon, under review).

While binding interpretations in Korean are not entirely lost, Kim, Montrul and Yoon found that the grammar of early bilinguals is more affected than that of late bilinguals. The late bilinguals retain a three-way Korean anaphor system, the early bilinguals appear to have a two-way system (with caki and caki-casin collapsed). This, again, is evidence for the simplification and reanalysis processes ­attested in heritage language grammars under reduced input conditions, discussed in §2.1, 2.2 and 2.3. In conclusion, the results of the studies discussed in this last section demonstrate that when heritage languages are used less than the majority language in bilingual children, less than optimal exposure can contribute to incomplete acquisition with long-lasting consequences into adulthood, such as reduced grammatical system. Degree of incomplete acquisition, and of simplification of the grammatical system, varies with age of onset of L2 acquisition: simultaneous bilinguals are more likely to display more severe L1 incomplete acquisition into adulthood than sequential bilinguals. Yet the cases discussed so far also show that early bilinguals who are still exposed to, and use, the heritage language have a good overall command of the heritage language, even when they fall short of native speaker norms in several areas of morphosyntax. And there are also cases of very fluent heritage speakers, indistinguishable from full speakers, although these cases have not been the focus of recent research and are not the focus of this book. But what happens when input is lacking altogether? In the next section, we will focus on two studies

202 Incomplete Acquisition in Bilingualism

documenting the dramatic effect of L1 loss and L2 replacement when input to the L1 is completely interrupted.

3.2

L1 Loss in adult foreign adoptees

So far, we have seen that the effects of reduced input in the family language in childhood have repercussions for the developing grammar into adulthood, and that age of reduced exposure to input matters significantly: adult simultaneous bilinguals exhibit more incomplete acquisition than sequential bilinguals. This suggests that age of acquisition and input are both responsible for this state of affairs. I will now show that there is a difference between reduced input to and use of the language in childhood versus completely interrupted input to and use of the language in childhood. And, to reiterate the main argument of this book, age is a crucial factor, determining the extent of language retention and loss into adulthood. In Chapter 3, we discussed the case study of a Chinese infant adopted by an English-speaking Canadian family at the age of 1;5 (Nicoladis & Grabois 2002). In two months, knowledge of Chinese (as measured by productive and receptive vocabulary) had severely declined. Isurin (2000), reviewed in Chapter 5, also showed that a Russian-speaking girl adopted at age 9 showed progressive decreasing access to the L1 lexicon in a one-year longitudinal study. On the basis of these case studies, one must ask whether such dramatic L1 loss in early childhood is permanent or whether there are recoverable remnants somewhere. In order to trace remnants of the L1 in adulthood, Pallier, Dehaene, Poline, LeBihan, Argenti, Dupoux and Mehler (2003) tested eight Korean adults born in Korea and adopted by French families between the ages of 3–8 years. For all of them, contact with Korean ceased immediately after adoption. Mean age at the time of testing was 26.8. Interestingly, while all adoptees had become very fluent in French (their L2), none of them reported remembering any Korean, which is highly surprising if some children were adopted as late as 8 years of age. The purpose of Pallier et al.’s study was to test a particular version of the critical period hypothesis for L1 and L2 acquisition, which stresses loss of cerebral plasticity in childhood. Despite having received mixed experimental support, the crystallization hypothesis predicts that the later an L2 is learned, the more different the cortical representations of the L1 and the L2 will turn out to be. Another prediction of this particular version of the hypothesis is that early exposure to the L1 between the ages of 3 and 8 should have long-lasting traces in the neural circuits subserving language processing. The study included behavioral and neuroimaging tests. The behavioral tests consisted of a language identification test and a word recognition task. In the language identification test, 8 adult adoptees and 8 adult monolingual French



Chapter 6.  Incomplete L1 acquisition in adults 203

s­ peakers heard 60 sentences in five different unfamiliar languages (Korean, Japanese, Polish, Swedish and Wolof). After hearing each sentence, subjects were to provide a judgment on a 7-point scale as to whether each sentence was Korean (7 = I am very sure it is Korean, 1 = I am sure it is not Korean, 4 = unsure). The second task was a translation task. A French word appeared first followed by two Korean words. Subjects were to indicate which one of the two Korean words was a correct translation for the French word. Finally, an fMRI was performed while the participants listened to sentences in four different languages: French, Korean, Japanese and Polish. French and Korean were the experimental languages, while Polish and Japanese (both unknown to all participants) were the control languages. Results of the behavioral tests revealed that the Korean adoptees did not differ significantly from the French native speakers in their identification of Korean sentences, which were rated close to 4 (= I don’t know). In the word translation test, the Korean adoptees and the French native speakers performed alike in recognizing Korean translations of French words (56% and 52% success, respectively). The fMRI data did not reveal any differences in brain activation between the Korean adoptees and the French native speakers either. The cortical regions that responded more to French (the known language) than to the other unknown languages were also similar in the adopted Koreans and in the French native speakers. Since no detectable L1 traces were found by these researchers, they concluded that an L1 can be completely lost, even when input to the L1 is completely severed and input in the L2 begins between the ages of 3–8. Since the L2 has completely replaced the L1 in these cases, this finding casts doubt on the crystallization view of the Critical Period Hypothesis. However, if a child is considered to have complete command of their L1 by age 8 according to some researchers (Smit, Hand, Feilinger, Benthal & Bird 1990; Snow, Burns & Griffin 1998), it is hard to believe that no traces of the L1 can be found in these adoptees. Moreover, the fact that something is not found does not necessarily imply it is not there. This study may simply show that the methods and approach used to find L1 traces may not be the right ones, or that many other psychological and affective factors may be responsible the apparent total inhibition of the first language in these cases. If Korean adoptees cannot recognize Korean words and sentences, can they at least detect Korean phonological contrasts? Studies of phonological discrimination in early exposed versus late exposed learners appear to show an advantage for early exposed learners, even when exposure to the L2 began as early as two years of age (Werker & Tees 1983; Tees & Werker 1984). In follow-up studies, Ventureyra, Pallier and Yoo (2004) and Ventureyra (2005) tested 18 adult Korean adoptees between the ages of 22 and 36 raised by French-speaking families in France, Belgium and Switzerland, and adopted at different ages in early childhood (range 3–9, mean age 5.8). Nine of them had never been exposed to Korean

204 Incomplete Acquisition in Bilingualism

after adoption, while the other nine had been to Korea on vacation for a short period of time. There were two control groups, a Korean native speaker group and French-native speaker group. The focus of the investigation was discrimination of the Korean stop consonants /p, t, k/ which differ along three parameters – voice, ­laryngeal activity and supralaryngeal articulation (plain, tense, aspirated). Twenty-four CVCV Korean pseudo words were created and presented aurally in minimal pairs differing in whether the initial consonant was plain, aspirated, or tense. Subjects were asked to perform a phoneme discrimination task, by indicating whether the two stimuli in each pair were the same or different. According to the results, discrimination of Korean stop consonants was significantly different between the Korean native speakers and the French speakers. The Korean adoptees and the French native speakers did not differ from each other, indicating that the Korean adoptees – just like the French native speakers – were completely unable to discriminate Korean stop consonants. The group of Korean adoptees was subsequently split into those who had been completely cut off from Korean and those who had been re-exposed to Korean later in life. Results showed that the adoptees re-exposed to Korean were able to discriminate Korean stop consonants only in one of the conditions tested (tense-aspirated pairs). Although the researchers suggested that such advantage was minimal, since this is the pair that is easiest to recognize due to the large difference in VOT, the advantage still suggests that re-exposure to the language in adulthood has the potential to bring back some aspects of the “lost” language. Inasmuch as there is a re-exposure effect, even if minimal, this effect suggests that there are traces in the L1 somewhere in memory. Furthermore Ventureyra (2005) reports on other studies of word recognition of Korean early acquired words and phrases, numbers, days of the week, months of the year and the Korean alphabet. She found that the Korean adoptees re-exposed to Korean recognized many of these aspects of Korean much better than the non re-exposed subjects and the French native speakers. And in a smallscale study which trained Korean adoptees (re-exposed and non re-exposed) and French native speakers at discriminating the stop contrasts, Ventureyra (2005) found that the re-exposed Korean adoptees made more gains after training than the other groups. Therefore, an effect for re-exposure indicates that the dormant L1 has been reactivated. If it were entirely lost, as the early research of this group suggests, then there would be no activation whatsoever. The fact that early acquired languages are not totally lost in cases of interrupted input, but may instead be hidden or temporarily inaccessible, is supported by the case studies of recovery under hypnosis reported by As (1962), Fromm



Chapter 6.  Incomplete L1 acquisition in adults 205

(1970) and, most recently, Footnick (2007).15 Footnick reports on a case study of a 21-year-old Togolese speaker raised in France (although the subject was not an internationally adopted child). From the ages of 2 1/2 to 6 years, he learned the Mina language in Togo, and at age 6 he returned to France with his family. However, the language ceased to be spoken in the family. When the subject was tested, he reported not speaking Mina since age 6. In several hypnosis sessions, the subject was regressed to ages 4–5 and was able to both speak and understand Mina in the presence of an interpreter. The results of this study indicate that it is possible for memory traces of a “lost” language to remain in the brain even when the language is consciously inaccessible to the speaker. In conclusion, results of adult adoptees who stopped receiving input in childhood before age 8 suggest that an L1 can be forgotten to a large extent, as measured by behavioral and brain imaging measures. These findings, as well as those of incomplete acquisition surveyed in this chapter, support my claim that the earlier the onset of L2 acquisition, and the more severe the reduction of L1 input, the more extensive the degree of incomplete L1 acquisition. While reduced exposure to the language appears to result in localized fossilization, and complete cut-off from the language appears to result in more generalized L1 loss, further research on interrupted input and re-exposure in adoptees is still needed.

4.

Conclusion

I have shown that incomplete L1 acquisition and stabilization is very common and likely in adult early bilinguals of ethnic minority languages who are very fluent in the majority language. Furthermore, degree of incomplete acquisition is inversely related to age: the earlier the age of onset of bilingualism or extensive exposure to, and use of, the majority language, the more severe the effects of incomplete acquisition. Unlike L1 attrition discussed in Chapter 3, incomplete acquisition clearly affects language at the level of competence, not just performance, since non-target like linguistic behavior is evident from a variety of behavioral tasks targeting production, comprehension, and grammaticality or semantic judgments. With respect to the specific structural properties of these incompletely acquired grammars, many heritage language speakers, and especially those with the lowest proficiency in the language, have linguistic systems that appear simplified at the levels of phonology, morphology, syntax, and lexicon, with respect to the grammatical systems of fully competent native speakers.

15. See also cases of recovery in bilingual aphasia (M. Paradis 1983).

206 Incomplete Acquisition in Bilingualism

Incomplete L1 acquisition appears to share many linguistic features with incomplete L2 acquisition. Grammatical areas that are typically difficult by L2 learners to master (in terms of protracted acquisition and high error rates), and prone to arrested development in the L2, are also problematic for adult heritage speakers. Adult heritage speakers lag behind first generation or monolingual speakers in their control of nominal (agreement, case) and verbal morphology (tense, aspect and mood), word order, anaphoric expressions, and many other language specific properties of particular languages (e.g., differential object marking or a personal in Spanish, genitive of negation in American Russian). Furthermore, many of the errors observed in L2 acquisition and in incomplete heritage language acquisition can be accounted for by transfer from the dominant language (or L2 influence on the L1). For example, the areas of incomplete knowledge in the Spanish, Russian and Korean heritage speakers discussed in this chapter may easily be attributed to English. Because English has no gender agreement in nouns, it has much poorer verbal inflection that does not distinguish morphologically between tense-aspect and mood, stricter word order, and no null subjects, clitics, differential object marking or long-distance binding of anaphors. To really determine whether it is contact with English which is driving such linguistic simplification in these heritage languages, other language contact situations should be examined (Korean in contact with Spanish, or Arabic in contact with Spanish in Spain, for example). Overall, more crosslinguistic studies comparing adult heritage language systems are needed to understand how much of incomplete acquisition is driven by universal linguistic tendencies and how much is induced by language contact (L2 influence on the L1). I have also claimed that the simplified grammatical systems of heritage speakers have the signatures of L1 grammars acquired within the critical period, since core aspects of the language are retained while other complexities are lost or remain underdeveloped due to reduced input conditions. As in L2 acquisition, inflectional morphology appears to be more affected than syntax proper (and we will return to this distinction in Chapter 7), and inherent, semantically-based case is more affected than structural or syntactic case. Similarly, the syntax-semantics interface is more affected than syntax proper. Since all these grammatical areas also cause persistent difficulties to adult L2 learners, in the next chapter I compare the linguistic outcomes of L1 and L2 incomplete acquisition.

chapter 7

Incomplete L1 and L2 acquisition in adults

After having traced bilingual acquisition from childhood into adulthood in the preceding three chapters, we now come back to the facts and theoretical issues presented at the beginning of the book. In Chapter 2, I discussed at length the observation that adult L2 acquisition very often results in incomplete and variable ultimate attainment. This fact has formed the basis for postulating a critical period for L2 learning, as assumed by several theories of L2 knowledge. However, in light of the data on early bilingualism (children and adults) brought to light in this book, it becomes obvious that this view of incomplete acquisition and its theoretical significance for theories of L2 competence may need reconsideration. Chapters 4, 5 and 6 revealed that incomplete acquisition, transfer, and fossilization feature prominently in bilingual speakers of minority languages, or heritage speakers, who were exposed to the language either exclusively during their first few years, or together with the second language since birth. If incomplete acquisition of an early acquired language due to reduced input and use is possible, we are led to question whether age of acquisition is the main reason why post puberty L2 learners rarely reach the degree of ultimate attainment of a native speaker. The question one must ask, naturally, is whether the means, processes, and outcomes of L1 and L2 incomplete acquisition in adults are the same or not. This chapter revisits some of the theoretical perspectives on L2 acquisition presented in Chapter 2 in order to understand how early and late bilinguals may differ from each other. I then engage in a systematic comparison of L1 and L2 incomplete acquisition by looking at empirical studies that directly compare post puberty L2 learners and early bilinguals who are heritage speakers of a minority language. In essence, how does the grammatical competence achieved by Kevin in Carolina, the hypothetical L2 learners introduced in Chapter 1, compare with the grammatical competence of Elena, Carlos, Alicia and Beatriz, all of whom exemplified different degrees of incomplete L1 acquisition? I maintain that, both theoretically and methodologically, comparing these two bilingual populations may provide a more illuminating perspective on differences and similarities between L1 and L2 acquisition than the traditional child L1-adult L2 comparison. Why is this comparison compelling and unique? Not only do L1 children and L2 adults differ in age, but they also differ in their cognitive maturity, the existence

208 Incomplete Acquisition in Bilingualism

of a previously acquired language, the initial state of acquisition, and the context of acquisition. By comparing adult bilinguals who vary in age of onset of bilingualism under the same experimental conditions, we easily control for age and cognitive maturity at time of testing. If the critical period hypothesis holds, adult early bilinguals with incomplete knowledge of the family language should still have some advantages over L2 learners because they received input in the target language early in childhood. This comparison allows us to test more directly whether the weaker language in adult early bilinguals is truly an L1 (the Weaker Language as L1 Hypothesis that I offered in Chapter 4, §5.2) or resembles an L2 with fundamentally different properties from an L1 (the Weaker Language as L2 Hypothesis proposed by Schlyter (1993), see Chapter 4, §5.1). In addition to these theoretical considerations, there are practical reasons for engaging in a systematic comparison of these two types of adult language learners. In many institutions of higher education in Western Europe and North America, heritage speakers of languages like Chinese, Korean, Spanish, etc. enroll in classes for second language learners. Very often, these speakers have not spoken their family language since childhood, but when in college, they are interested in reacquiring, maintaining or expanding their linguistic knowledge in a classroom setting. Most of the studies examined in this chapter compare college-age L2 learners and heritage speakers enrolled in language classes. Uncovering the type of linguistic knowledge these speakers do or do not have in common is crucial for the education of heritage language learners, which continues to generate significant challenges for language teaching practitioners due to the scarcity of systematic research in this area (Brinton, Kagan & Bauckus 2008). Further research in this area will also expand our current understanding of the limits and possibilities of bilingualism under different circumstances.

1.

Theoretical accounts of incomplete L2 acquisition

Chapter 2 introduced several theoretical positions that either deny or downplay the existence of a critical period for L2 acquisition and access to Universal Grammar (Epstein, Flynn & Martohardjono 1996; Prévost & White 2000; Schwartz & Sprouse 1996; White & Genesee 1996; White 1989, 2003a) or stress other confounding factors that give the impression of an age effect (Birdsong 1999, 2005; Flege 1999; Bialystok 2001; Bialystok & Hakuta 1994, 1999). In the rest of this chapter, however, I will focus on theories that assume a critical period in order to explain differences between L1 and L2 acquisition. The reason for this particular emphasis is that the Critical Period Position makes explicit claims about the



Chapter 7.  Incomplete L1 and L2 acquisition in adults 209

e­ tiology of “deficits” in incomplete L2 acquisition, and my goal in this chapter is to uncover how incomplete L1 and L2 acquisition may relate to each other. Bley-Vroman’s (1989) Fundamental Difference Hypothesis (FDH), Schachter’s (1990) Incompleteness Hypothesis, Meisel’s (1997, 2001) no parameter resetting account, Hawkins and Chan’s (1997) Failed Functional Features Hypothesis, Hawkins and Hattori’s (2006) Representational Deficit Hypothesis, and Tsimpli and ­Dimitrakopoulou’s (2007) Interpretability Hypothesis, are all theoretical positions that directly link incompleteness and variable outcomes of the L2 learning process to an age effect. In Chapter 2, I have referred to this group of accounts as the deficit view of L2 acquisition, or the Critical Period Position. It must be recalled from Chapter 2 that for Bley-Vroman (1989), child L1 acquisition apparently happens so rapidly and efficiently because the process is guided largely by innate, deductive, implicit and specific linguistic mechanisms assumed to be part of Universal Grammar. To explain the apparent differences in outcome between L1 and L2 acquisition, the main claim of the FDH is that access to Universal Grammar is subject to a critical period, such that when learning a second language, post puberty L2 learners can only rely on their L1 knowledge (a particular instantiation of Universal Grammar, but not the full spectrum of linguistic options) and the principles and parameters active in their L1. Unable to utilize deductive, domain-specific (i.e., purely linguistic) mechanisms like L1 learners, L2 learners resort to domain-general, inductive problem-solving skills like analogy or pattern-matching instead. In a similar vein, Schachter (1990) argues that differences between L1 and L2 achievement, transfer effects, and fossilization indicate that “the notion of UG in its entirety” (p. 95) is not available to post puberty L2 learners: “If UG were totally available, adult second language learner outcomes would be more uniform and they would generally be more proficient” (Schachter 1990, p. 96). More specifically related to the issue of age, Schachter states that If . . . some principles require certain types of input in order for them to be incorporated into the L1 and that input is not available at the maturationally appropriate time (emphasis mine), then all that remains as part of the knowledge state of an adult native speaker of a language is a language-specific instantiation of UG which does not incorporate the aforesaid principles.  (Schachter 1990: 99–100)

Both Hawkins and Chan (1997) and Meisel (1997) claim that principles of Universal Grammar remain available past puberty, but parameterized features do not. Bley-Vroman, Schachter, Meisel and Hawkins and Chan are all in agreement that L2 learners cannot overcome the structural limits imposed by their mature L1s.

210 Incomplete Acquisition in Bilingualism

For Schachter (1990), the Incompleteness Hypothesis is the “correct hypothesis” because it is consistent with many linguistic facts, serves as a partial explanation for the lack of completeness so often found in adult L2 learners, and predicts the non-equivalence between L1 and L2 grammars of the same language. Similar versions of this particular position within generative linguistics can also be found within cognitive and neurolinguistic perspectives on L2 acquisition. These approaches do not necessarily view language and language learning as innate and domain specific (i.e., inherently linguistic), or assume a particular linguistic theory of grammatical organization, but take into account the distinction between procedural and declarative knowledge and implicit and explicit language learning (DeKeyser 2003; N. Ellis 2005; M. Paradis 2004, 2007; Ullman 2001). “Implicit knowledge” refers to that which is learned without awareness of what is being learned, and is learned incidentally or not (depending on the author). Implicit knowledge is stored in procedural memory, and when this knowledge is accessed or recalled, it is executed automatically and quickly. By contrast, “explicit knowledge” is acquired with awareness of what is being learned, and with conscious effort. Because explicit knowledge is learned explicitly, individuals can verbalize this knowledge on demand. It is stored in declarative or episodic memory, where our world knowledge is stored. It is accessed on demand, not necessarily automatically. Adult native speakers have both systems of learning available and use them as needed. According to M. Paradis (2004), when very young children and illiterate adults speak or comprehend language, they use implicit competence (or knowledge) only. By contrast, incipient L2 learners use explicit knowledge of the L2 when producing or understanding the L2, and steadily, and in tandem, develop implicit competence of it. Much debate exists as to whether explicit knowledge can become implicit after substantial exposure, practice, and automatization (­Hulstijn 2005; Jiang 2007) or how the two types of knowledge interact (N. ­Ellis 2005). Although different in approach, DeKeyser (2000, 2003) agrees with the spirit of the Incompleteness Hypothesis and the Fundamental Difference Hypothesis. He contends that adult L2 learners use a different cognitive system to learn an L2 because maturational constraints apply to implicit linguistic competence acquired early in childhood (see also Abrahamsson & Hylstenstam, in press). The decline of procedural memory and the attenuation of implicit cognitive mechanisms somewhere in childhood for language – what others take roughly to be the domain of Universal Grammar and specific linguistic mechanisms – force late L2 learners to rely on explicit learning. With this theoretical background in mind, let us turn to a potential counterexample to the Incompleteness Hypothesis and a strong version of the critical period position as formulated for adult L2 acquisition: the heritage language speaker.



2.

Chapter 7.  Incomplete L1 and L2 acquisition in adults 211

Some similarities between L2 and L1 incomplete acquisition

If incomplete acquisition in adult L2 acquisition were merely due to a late onset of bilingualism which missed the critical period for the L2, early bilinguals (heritage speakers) who learned the two languages simultaneously or sequentially in childhood should, in principle, acquire complete knowledge of the two languages, provided they received sufficient input. To the extent of my knowledge, I am not aware of any studies of ultimate attainment in adult early bilinguals showing this specific outcome. One could argue that incomplete acquisition of an L1 in early bilinguals is not due to age but to insufficient input. But if short of showing complete grammatical competence, age of acquisition is still a key factor in degree of linguistic ability, early bilinguals (heritage speakers) should still have a considerable advantage over late bilinguals (L2 learners) because they received exposure to the family language early in childhood. What appears problematic is that some recent studies comparing adult heritage speakers and L2 learners on a variety of morphosyntactic properties have shown that heritage speakers and typical second language learners are not very different from each other (Au et al. 2002). In other words, when compared to native speakers, both L2 learners and heritage speakers whose L1 is the weaker language show distinctive and comparable patterns of incomplete acquisition, either due to fossilization or transfer from the dominant language. In one of my own studies, Montrul (2004c), I focused on the acquisition of the Spanish aspectual system. This study statistically compared the results of three groups of L2 learners (intermediate, advanced and superior) from Montrul and Slabakova (2003) and a group of heritage speakers (some simultaneous bilinguals and others child L2 learners of English or sequential bilinguals) from Montrul (2002) (see details in Chapter 6). Participants in these studies judged the meanings of the preterite and imperfect aspectual markers with accomplishment, achievement and stative predicates in a sentence conjunction judgment task and in a truth value judgment task. At issue was whether L2 learners and Spanish heritage speakers were aware of the semantic contrast between the preterite and imperfect with different predicates. Although all groups showed overall discrimination of the preterite-imperfect contrast in the two tasks, in general, advanced L2 learners and heritage speakers did not differ statistically from each other, but differed significantly from monolingual Spanish speakers in some areas of semantic interpretations. When compared to the target, this may be taken as a sign of incomplete knowledge. If differences between these two groups of bilinguals and the monolinguals is related to the influence of English as L1 (in the L2 learners) and to the dominant L2 (in the heritage speakers), then I hypothesized that incompleteness would be observed with stative verbs, and with the habitual, generic, and specific meanings of the

212 Incomplete Acquisition in Bilingualism

Table 7.1  Mean judgment on preterite and imperfect sentences by predicate type (adapted from Montrul 2004c). Groups

N

Monolinguals 20 L2 learners 24 Heritage speakers 31

accomplishments imperfect preterite

achievements imperfect preterite

states imperfect preterite

1.51 1.32 1.06

1.28 0.44 0.63

1.55 0.53 1.17

–1.38 –0.91 –0.78

–1.78 –1.39 –1.58

–1.41 –0.91 –0.66

Note: a mean close to 2 indicates that the sentences were judged as logical, a mean close to –2 contradictory.

Table 7.2  Accuracy percentage on preterite and imperfect sentences by story type (adapted from Montrul 2004c). Condition

Story Aspectual context form

Change of meaning stative preterites

Habitual vs. One time event

Monolinguals L2 learners Heritage speakers (n = 20) (n = 24) (n = 31)

preterite (F) 89.16 imperfect (T) 86.6

71.5 82.5

54.16 83.33

eventive preterite (T) 78.33 imperfect (F) 90.83

68 81.33

65.33 71.5

habitual preterite (F) 97 imperfect (T) 93

74 68.2

91 68.8

onetime

99 60.8

96.2 60.4

81.16 86.6

84.33 88.33

81.8 52.66

85.83 56.66

preterite (T) 99 imperfect (F) 94

Generic vs. Specific generic preterite (F) 85.6 interpretation of se imperfect (T) 92.5 specific preterite (T) 96.6 imperfect (T) 73.33

imperfect. Tables 7.1 and 7.2 summarize the results of the advanced L2 learners and heritage speakers. Overall, stative verbs in the imperfect were only problematic for the L2 speakers, while stative verbs in the preterite were problematic for both the heritage speakers and the advanced L2 speakers. However, the heritage speakers and the L2 groups gave similar judgments on achievements in the imperfect, which were rated much less acceptable as compared with the ratings of the monolinguals. . These conditions were related to the [–perfective] feature that English was assumed to lack, according to Giorgi and Pianesi’s (1997) account of the functional projection aspect phrase in Germanic and Romance.



Chapter 7.  Incomplete L1 and L2 acquisition in adults 213

Change of meaning preterites (saber “know”, poder “be able to”, etc.) tested in the truth value judgment task were equally problematic for the heritage speakers and the advanced L2 learners, who recognized that these verbs are stative in the imperfect, but did not readily accept the stative-preterite combination in eventive contexts. Results are summarized in Table 7.2. Similarly, the same groups were inaccurate with the imperfect sentences in habitual stories and in one-time event stories. Finally, while all speakers recognized that the imperfect has a generic reading, neither the L2 speakers nor the heritage speakers recognized that the imperfect can also have a specific reading. Even though all the early and late bilinguals compared in this study were quite proficient in Spanish, their performance was still quantitatively very different from that of monolingual speakers, suggesting that the grammars of L2 learners and heritage speakers are similarly incomplete with respect to the target. Furthermore, the two groups displayed non target-like performance in exactly the same combinations and conditions, namely achievements in the imperfect, states in the preterite, and specific interpretation of the preterite in impersonal sentences. Another study that appears to go against the predictions of the Strong Critical Period position for incomplete L2 acquisition is Bruhn de Garavito (2002). Bruhn de Garavito investigated syntactic knowledge of the verb movement parameter in Spanish, which refers to the position of finite verbs with respect to adverbs in transitive sentences. Spanish, like French, is a verb-movement language. Because agreement is strong, verbs move in the syntax past adverbs, quantifiers and negation, as shown in (1) with the manner adverb cuidadosamente “carefully”. In English, agreement is weak and verbs do not move, as shown in (2). (1) Lucía escribe cuidadosamente sus tareas. (2) a. *Lucy writes carefully her homework. b.   Lucy writes her homework carefully.

Furthermore, in most Spanish varieties, the verb also moves in some wh-questions producing subject-verb inversion, as in (3); In English, as in (4), there is do-support:

. I have recently replicated this study and the study on subjunctive (Montrul 2007) with larger groups of heritage speakers and L2 learners of low to advanced proficiency levels. When proficiency is taken into account, different patterns are observed between the two groups. At lower levels of proficiency, the heritage speakers show similar pattern of responses (although still far from targetlike) as the native speakers, while the L2 learners show a different pattern altogether.

214 Incomplete Acquisition in Bilingualism

(3) a.   ¿Qué escribió Lucía? b. *¿Qué Lucía escribió? (4) a.   What did Lucía write? b. *What wrote Lucía?

Bruhn de Garavito tested knowledge of verb movement in Spanish through a grammaticality judgment task administered to 15 L2 learners and 15 Spanish heritage speakers and a monolingual control group. Overall results showed significant differences between the monolingual control group and the two experimental groups, indicating that heritage and non-heritage learners did not have the same level of attainment of native speakers. However, there were no statistically significant differences between the heritage speakers and the L2 learners, suggesting overall similarity in level of attainment. Bruhn de Garavito concluded that age of acquisition does not necessarily provide an advantage for heritage speakers who are relearning the family language in the L2 classroom, as opposed to typical L2 learners who started learning the language after puberty and in a formal setting. Therefore, according to Bruhn de Garavito, it is possible to reset parameters in an L2, regardless of age of acquisition (cf. Meisel 1997). Two other studies comparing Korean heritage speakers and L2 learners of Korean report very similar findings: no differences between L2 learners and heritage speakers. O’Grady, Lee and Choo (2001) investigated knowledge of morphosyntactic cues in the comprehension of Korean subject and object relative clauses, as exemplified in (5) and (6). (5) [ _ namca-lul cohaha-nun] yeca man-acc like-Prs woman “the woman who likes the man” (6) [namca-ka _ cohaha-nun] yeca Man-nom    like-Prs woman “the woman who the man likes”

subject relative clause

object relative clause

Korean is a head-final language. The head of the relative clauses in (5) and (6) is the word yeca “woman” and appears to the left of the relative clause. In order to correctly interpret these relative clauses, speakers need to be sensitive to the overt case markers (ka vs. lul) on the word namca “man.” O’Grady et al. (2001) used a picture selection task to test comprehension of these two types of relative clauses. Two pictures with reversible arguments were presented (one depicting a woman who likes a man and the other one a man who likes a woman), and participants were asked to choose the picture that matched the sentence. Results are summarized in Table 7.3.



Chapter 7.  Incomplete L1 and L2 acquisition in adults 215

Table 7.3  Accuracy percentage selecting subject and object relative clauses in a picture selection task (adapted from O’Grady, Lee & Choo 2001). Groups

N

Relative clause Correct

Reversals

Other

L2 learners (4th semester)

25

L2 learners (2nd semester)

20

Korean heritage learners

16

subject object subject object subject object

  6.4 38.4 14 31 13.75 23.75

17.6 52 16 23 21.25 35

76 17.6 70 46 65 41.3

The lack of statistical difference between the L2 learners and the Korean heritage learners was interpreted as no difference in their linguistic knowledge. There was a main effect for relative clause type, since all groups were more accurate with subject than with object relative clauses. As can be seen from Table 7.3, accuracy for object relatives was below chance. Most of the errors involved reversals, or the comprehension of an object relative as a subject relative. This is the same type of error that children learning Korean and English as L1 frequently make with relative clauses (Cho 1999; O’Grady 1997), as well as adults learning English as L2 (Gass 1979; Wolfe-Quintero 1992). The conclusion reached by O’Grady et al. (2001) is that heritage Korean learners do not appear to have an advantage over L2 learners when making use of overt case markers to interpret complex sentences in Korean. Finally, we turn to J.-H.Kim, Montrul and Yoon (in press) as our last example of apparent similarities between L2 learners of Korean and Korean heritage speakers. The study tested knowledge of the structural conditions for the binding possibilities of the Korean reflexive caki, especially conditions allowed in Korean but not in English (sub-commanding antecedent). Among the four groups of participants tested, there were 22 adult early Korean-English bilinguals residing in the United States, 18 English-speaking L2 learners of Korean, and a Korean monolingual group (n = 30). At the outset of the study, there were no statistically significant differences on the performance of the L2 learners and the Korean heritage speakers in a Korean proficiency test (64% vs. 58% respectively), but the two bilingual groups were statistically less accurate than the monolingual Koreans (mean percentage accuracy 92%), suggesting some incomplete knowledge on the part of both the early and late bilinguals. The results of a truth value judgment task investigating the size of the governing category in binding interpretations of the Korean anaphor caki showed that heritage speakers assigned a lower degree of acceptability to Korean sentences where caki is bound outside the English governing category, but inside the Korean governing category. The L2 learners showed the exact same problems

216 Incomplete Acquisition in Bilingualism

with governing category size. These findings indicate that there are transfer effects from English both in L1 and L2 incomplete acquisition. To summarize, the conclusion that can be reached by examining the linguistic competence of adults is that bilingual children can, in principle, become native speakers of their two first languages, but this does not necessarily imply that they will, when input and opportunities to use one of the languages are reduced. What starts as a pattern of delayed acquisition of the weaker language early in childhood (Chapter 4), persists into the school years (Chapter 5) and stabilizes at an incomplete state of knowledge into adulthood (Chapter 6). The four studies discussed in this section demonstrate that while the grammars of L2 speakers still fall short of native-speaker performance, starting early with the acquisition of a language in a bilingual environment does not automatically bring an added advantage to heritage speakers in some aspects of inflectional morphology and syntax, even if input is more or less available. What factors, then, lead to comparable outcomes in L1 and L2 incomplete acquisition? Let us examine more closely the characteristics of heritage language acquisition.

3.

Heritage language acquisition

In many respects, heritage language acquisition has characteristics of both L1 and L2 acquisition. One possible way of characterizing it is to suggest that it is incomplete L1 acquisition that takes place in a bilingual environment rather than a monolingual one (the Weaker Language as L1 Hypothesis). Because it takes place in a dual language context, heritage language acquisition shares the developmental path (or lack thereof) and characteristics of L2 grammars as well. To show how heritage language acquisition – a particular case of early bilingualism – is a hybrid type of acquisition which combines features of child L1 and adult L2 acquisition, part of Table 2.1 from Chapter 2 is repeated and modified here as Table 7.4. Shaded cells represent the intersecting subset between L1 and L2 acquisition which characterizes heritage language acquisition. Table 7.4 shows that similar outcomes of acquisition, as estimated from degree of variable proficiency, fossilization, and cross linguistic influence in heritage language and L2 acquisition may be due to at least two factors: 1) variable amount of input; and 2) socio-affective factors. Unlike L1 learners, who receive abundant input early in childhood, early bilinguals of minority languages receive less input and have fewer opportunities to use the family language once they begin exposure to the majority language. Although input is not entirely disrupted, as in foreign adoptees discussed in Chapters 4, 5 and 6, early bilinguals vary significantly in the amount of exposure and use of the family language due to different family and



Chapter 7.  Incomplete L1 and L2 acquisition in adults 217

Table 7.4  Characteristics of L1, L2 and heritage language acquisition. Factors

L1 acquisition

Adult L2 acquisition

1. previous linguistic knowledge

none

2. input

a. timing b. setting c. mode d. amount

early exposure (birth) naturalistic aural abundant and frequent

e. quality

linguistically varied and rich, contextually appropriate

L1 knowledge (fully developed) late exposure (after puberty) naturalistic and instructed aural and written varying (in amount and frequency) contextually restricted, less variety of structures/ vocabulary, input from other non-native speakers literate in the L1 and L2

f. literacy

more complex structures and vocabulary continue to be acquired and reinforced after age 5 when metalinguistic skills develop 3. personality and affective factors irrelevant 4. target system a. types of errors developmental errors b. fossilization c. outcome

does not occur successful and complete

relevant developmental and transfer errors typical variable and typically incomplete

Note: Shaded cells represent the intersecting factors between L1 and L2 acquisition that characterize heritage language acquisition.

sociolinguistic factors. In this regard, L2 learners also receive variable amounts of input depending on how many classes they take, how many opportunities they seek to use the L2 beyond class, whether they travel abroad or not, etc. Other factors that may play a role in early and late bilingual outcomes – and in the particular case of heritage speakers – are socio-affective, which appear to have no impact in monolingual development. Like adult L2 learners for whom personality and degree of motivation to learn the language play a role in how much exposure to the language they will seek, very young bilingual children also become aware of their social context and make language choices accordingly (Chapter 4). When children perceive that their family language is not as valued as the majority language, they begin to speak it less and less. Less willingness to speak the language throughout childhood and negative feelings about the practicality of using the heritage language may eventually cause adult early bilinguals

218 Incomplete Acquisition in Bilingualism

to seek less input, use the language less, and end up with weaker competence in the heritage language. Despite these similarities between heritage speakers and L2 learners, which could explain the overall patterns of findings discussed in §2, Table 7.4 also clearly shows that heritage learners and L1 learners share many characteristics which set them apart from L2 learners. In addition to age, L2 learners and heritage speakers differ in the nature, mode, and timing of input received. Having been exposed to the family language since birth, either by itself or simultaneously with the majority language, heritage speakers and L2 learners may differ significantly in their knowledge of principles of grammar typically acquired during the period of early linguistic development, including phonology and some aspects of morphology and syntax. If age of acquisition plays a role in linguistic development, then one would expect heritage speakers with incomplete knowledge of the heritage language would still control basic aspects of the heritage language grammar that are implicitly acquired early in childhood by monolingual and bilingual children of that language. That is, if learning mechanisms are different in childhood and adulthood, there should be some advantages in some linguistic domains for heritage speakers (child acquirers) over L2 learners (adult acquirers). Note that the prediction for later language development may be different, especially if certain structures are very frequent in written language and less so in spoken varieties. If heritage speakers receive little to no academic instruction in the heritage language and miss the opportunity to be exposed to and acquire aspects of grammar that may only be imparted through the standard variety taught at school (e.g., subjunctive clauses, conditionals, relative clauses, etc.) then heritage speakers will probably be at a disadvantage in comparison to adult L2 learners, who are typically highly literate in the standard varieties of the language taught in instructed L2 acquisition. Unfortunately, maturational and environmental factors have been extremely difficult to tease apart in normal early linguistic development. In addition to age, L2 learners and heritage speakers differ in the type of input received, which varies according to context of acquisition. Adult heritage speakers, like L1 children, were exposed to the heritage language naturalistically at home, through the aural medium. They learned the vocabulary typical of pre-school children to talk about their environment and communicate with their caregivers. The amount of input . This hypothesis is confirmed by Rothman’s (2007) study of knowledge of inflected infinitives by adult Brazilian Portuguese heritage speakers and L2 learners. While the heritage speakers lacked morphological and semantic knowledge of these verbal forms like Brazilian Portuguese monolingual children reported in Pires & Rothman (in press), the L2 learners behaved like the educated monolingual Brazilian Portuguese speakers.



Chapter 7.  Incomplete L1 and L2 acquisition in adults 219

received in the heritage language by these children varies greatly depending on whether the two parents speak the language, whether other members of the family live in the house, whether the child has siblings who speak English, and how many hours a day the child hears and uses the heritage language. By contrast, postpuberty L2 learners typically begin exposure to the second language in highschool and college in a classroom environment through the written medium. They receive approximately one hour of instruction a day, or less, with emphasis on reading, writing and grammatical explanations. Some learners also study abroad for a period of time. In these contexts, L2 learners also learn age-appropriate basic vocabulary suitable for their communicative needs. A third important difference between heritage speakers and L2 learners is literacy, discussed in Chapter 5. Many heritage speakers are schooled in English, especially in the United States. Access to schooling in the heritage language also varies enormously depending on the heritage language, the availability of a sizable speech community, and whether the members of this community have developed cultural activities to transmit their language to their children. It is common for Korean, Chinese, and other heritage speaker groups to have weekly lessons in their language for school-age children (as in community or church schools). In the case of Spanish in the United States, some states and school districts offer bilingual education, and dual immersion schools, but this is not available everywhere. In general, the great majority of heritage speakers in the United States receives schooling in English and is typically semiliterate in the heritage language. While it is true that basic aspects of grammar are learned before the ages of 3–4, many more complex aspects of language are learned during the school-age period and are reinforced by literacy training (see Table 5.1, Chapter 5). Furthermore, academic metalinguistic knowledge, or knowledge about the structure of language, develops after the age of 4 in monolingual children and changes the way children look at language and operate within it. From a learnability perspective, Birdsong (1989, p. 45) states that this change, which is perhaps related to other parallel developments in Theory of Mind “is not a necessary condition for acquiring the language, since the essentials of native speaker competence are acquired before then.” Yet, as shown in chapter 5, literacy and written input continue to feed and shape language development. Little to no literacy in the heritage language would explain why some aspects of language appear incomplete in adulthood. As a result of this situation, heritage speakers will probably retain aspects of language acquired implicitly before 4 years of age, but may lack aspects of the heritage language that are acquired late by monolingual children and reinforced through language arts instruction at school, such as complex tenses and complex constructions. Returning to the issue of age, if heritage speakers appear to have an age advantage over post puberty L2 learners, this given advantage is masked by other

220 Incomplete Acquisition in Bilingualism

intervening and confounding factors, including context of acquisition, amount and type of input, and even type of learning. Extending DeKeyser’s (2003) position on adult L2 acquisition to the heritage speaker situation, an age advantage as implied in the Fundamental Difference Hypothesis would still hold for this population because the Critical Period Hypothesis applies to “implicit” linguistic competence. It is the decline of procedural memory, and loss of implicit mechanisms for language somewhere in childhood that forces late L2 learners to rely on explicit learning when learning an L2. As a result, DeKeyser contends that L2 learners use a different cognitive system to learn language. If that is the case, then heritage speakers should show an advantage over L2 learners with early, and implicitly acquired, aspects of grammar, and a disadvantage with late and explicitly acquired aspects learned during the school years. If an advantage is found for heritage speakers assuming this theoretical position with respect to L1 and L2 acquisition, then Schlyter’s Weaker Language as L2 Hypothesis for early bilinguals would be disconfirmed. Given these significant differences in age and context of acquisition, type, amount and timing of input, and literacy development, we can now extend the Strong Critical Period Position – originally formulated in the context of L1 and L2 acquisition – to formulate specific hypotheses regarding differences between L2 and heritage language learners. Prediction 1:

Heritage speakers should have advantage over L2 learners with early acquired grammatical knowledge

If heritage learners’ knowledge of the language (prior to instruction) has been acquired implicitly and primarily through access to Universal Grammar in childhood (as an L1), the Weaker Language as L1 Hypothesis predicts they should have implicit knowledge of core aspects of phonology and morphosyntax which emerge very early in childhood and are not dependent on a significant amount of sustained input, including phonology, core functional projections, basic word order, some early set parameters (e.g. pro-drop, word order, clitic pronouns in Spanish). By contrast, aspects of language that are context-dependent, acquired after the age of 5 years in monolingual children, and reinforced through reading and formal instruction at school, should be either missing or remain imperfectly acquired depending on the amount of input received in late childhood. Examples include specific vocabulary, forms of address and honorifics (in Asian languages), complex structures (i.e., relative clauses), semantically and pragmatically conditioned uses of subjunctive in Spanish and Russian, etc. If parameter resetting does not take place in L2 acquisition because acquisition is primarily achieved through the deployment of explicitly acquired knowledge and problem-solving cognitive



Chapter 7.  Incomplete L1 and L2 acquisition in adults 221

skills, then heritage speakers should be better than L2 learners, especially with aspects of language acquired early in childhood. Prediction 2:

Heritage speakers should be more accurate and faster than L2 learners in oral than in written production and comprehension tasks that minimize metalinguistic knowledge

Differences in mode of acquisition and degree of metalinguistic command of the language (effects of literacy) predict differential performance for L2 learners and heritage speakers depending on task and modality. Heritage speakers should be more accurate than L2 learners in oral than in written production and comprehension tasks. Furthermore, they should be less accurate than L2 learners in written tasks that allow time for reflection and require more metalinguistic knowledge, especially when comparing L2 learners and heritage speakers of low proficiency in the target language. Care should be taken in interpreting results from studies failing to find differences, however. If adults predominantly use their explicit learning skills for learning language beyond childhood as the Fundamental Difference Hypothesis claims, explicitly learned knowledge may mask implicit linguistic competence in L2 learners, especially if the tasks used to tap that knowledge are very metalinguistic (e.g., fill in the blanks with a given grammatical form vs. a free composition, or a judgment task). Prediction 3:

If there is a re-exposure effect, heritage speakers should react faster and better to instruction than L2 learners

For many researchers, it is not clear that L2 learners eventually develop implicit competence in their L2, although this remains controversial. If L2 learners have decreased access to implicit, deductive linguistic competence – as the Fundamental Difference Hypothesis implies – then they should not, in general, be able to perform like native speakers, even after several years of instruction and immersion. If heritage speakers are interrupted L1 learners who received some crucial input during the critical period when those mechanisms were still operative, instruction should conceivably be able to turn incomplete native speakers into complete educated native speakers, given optimal amounts of input and sufficient time to develop the underdeveloped skills. That is, there should be a re-exposure effect.

. See chapters in N. Ellis (1994) and special issue of Studies in Second Language Acquisition 27, 2, 2005 guest edited by Jan Hulstijn and Rod Ellis. . One can also imagine a situation in which heritage speakers do not make the projected progress, but not because their early acquired knowledge is not activated, but because being now adults, the implicit learning mechanisms are now attenuated or inaccessible for them as

222 Incomplete Acquisition in Bilingualism

Table 7.5  Summary of predictions assuming the Weaker Language as L1 Hypothesis for heritage speakers and the Fundamental Difference Hypothesis for L2 acquisition. Predictions

Heritage speakers

1. early acquired   grammatical knowledge

advantage

2. task and modality

advantage in oral and in less metalinguistic written tasks overall advantage

3. reactivity to instruction

L2 learners

advantage in more metalinguistic written tasks

In principle, it should be faster for these heritage language learners to reach certain linguistic milestones than for L2 learners. By contrast, regardless of amount of input and experience with the target language, L2 learners are not necessarily guaranteed to attain native-speaker competence in the L2. Thus, the Strong Critical Period position predicts early bilinguals have the potential to reach nativelike competence in the heritage language, while late learners do not. Table 7.5 summarizes these predictions, and in §4 I discuss relevant research findings from emerging research in this area.

4.

Research findings

4.1

Selective advantages in grammatical knowledge

Although early exposure to the L1 in heritage speakers would predict an advantage in several aspects of grammar in adulthood when compared with late L2 learners, the few studies on different aspects of Spanish and Korean morphosyntax reviewed in §2 appear to go against this general prediction. As we will see, this issue is far from clear since other research findings suggest the opposite, i.e., that there are advantages for heritage speakers. However, these advantages are selective rather than global in the sense that they do not surface in all linguistic domains, structures, and task types to the same extent. Furthermore, since heritage speakers are such a heterogeneous group, the proficiency level of early and late bilinguals interacts with accuracy in different grammatical domains. For example, with respect to phonological perception and production, research so far has demonstrated a clear advantage for heritage speakers, especially when their proficiency level is quite low. As for morphosyntax, some studies also show advantages for low proficiency well. Of course, this is a question that needs to be pursued empirically, which has not been investigated.



Chapter 7.  Incomplete L1 and L2 acquisition in adults 223

heritage speakers, while others have found advantages at advanced levels of development only. Crucially, proficiency level was not controlled in all the studies that found no differences between L2 learners and heritage speakers discussed in §2. On this note, it is also true that it is sometimes difficult to compare L2 learners and heritage speakers directly by using proficiency measures (oral or written) developed for, and normed with, second language learners, since as Valdés (1995) has shown, heritage speakers do not fall neatly in proficiency categories used with L2 learners. Still, I believe that to have a measure of proficiency (oral, written, selfreport, or all of the above combined) at the outset of a study is better than to have nothing at all. And to carry out controlled experimental studies, the same measure must be used with the two experimental groups. Oh, Jun, Knightly and Au (2003) looked at the benefits of early language experience in phonology. The study compared production and perception of the Korean dento-alveolar stop consonants (plain /t/, aspirated /th/ and tense /t’/ and plain /d/, aspirated /dh/ and tense /d’/). One of the aims of the study was to evaluate the effects of childhood linguistic memory under reduced input conditions. Tees and Werker (1984) showed that English-speaking adults who had heard Hindi in early childhood could discriminate Hindi speech sounds better than English speakers who had never been exposed to Hindi. In this study, Oh et al. (2003) tested two groups of Korean heritage speakers. The first group, called Korean childhood speakers in the study, were adults who had spoken Korean regularly until the age of 5, before becoming almost English monolinguals. The second group, called Korean overhearers in the study, consisted of adults exposed to Korean aurally in childhood but who hardly spoke it. In addition to the two heritage speaker groups, the study included a group of adult L2 learners of Korean and a control group of 12 Korean native speakers. Except for the Korean native speakers, all the other participants (n = 31) were enrolled in an elementary Korean language class and were tested after 4 months of instruction. All participants also completed a language background assessment questionnaire. Assuming there is a critical period effect for the acquisition of phonology, Oh et al. (2003) hypothesized that the heritage speakers should be better than L2 learners in both perception and production, but the heritage speakers who only heard, but did not speak, Korean should only be better in perception (and not in production). All participants were asked to complete a phoneme perception task and a phoneme production task. From the production task, VOT measurements were taken from the target words. This task was also used to estimate an overall accent rating for each subject. The phonemic perception task revealed differences between the two groups of heritage speakers, who were not different from the Korean monolinguals, and the L2 learners. Thus, in phonemic perception, very low proficiency heritage

224 Incomplete Acquisition in Bilingualism

s­ peakers have native-like abilities, whereas the L2 learners do not. In accent rating, there was a decreasing native-like trend from the Korean native speakers being the most native-like, followed by the Korean childhood speakers, followed by the Korean childhood overhearers, and the Korean L2 learners being the least native-like. The production task revealed that the native speakers and the Korean heritage speakers produced a contrast between each consonant pair (aspiratedplain, plain-tense, aspirated-tense), whereas the Korean overhearers and the L2 learners did not produce a contrast for any pair. This study is important because it stresses the role of childhood linguistic experience even when input and use become severely restricted upon beginning elementary school in the majority language. It also suggests that contact with the language and use of it – even if minimal – is important when these speakers come to the L2 classroom to relearn their heritage language later in life. Thus, in the realm of phonology, which is problematic in L2 learners, it is obvious that early childhood experience brings an added advantage to heritage speakers of very low proficiency in the heritage language. In light of findings such as these, are such advantages also measurable in other grammatical areas? In Prediction 1, I stated that heritage speakers should have advantage over L2 learners with early acquired grammatical knowledge. Au, Knightly, Jun and Oh (2002) and Knightly, Jun, Oh and Au (2003) examined potential childhood overhearing advantage in both phonology and morphosyntax. Participants were undergraduate students enrolled in second year Spanish language classes. From an original pool of 238 students, the researchers identified 15 U.S.-born heritage speakers who had been exposed to Spanish in childhood through parents and other relatives but who spoke Spanish poorly as adults, together with 15 typical late L2 learners of Spanish from English-speaking families. Since all students were taking the same classes, I assume they were roughly at the low-intermediate level of proficiency in the language. A group of 15 adult native speakers (all proficient users) of Spanish acted as the control group. For the phonology part of the experiments, all subjects were asked to pronounce a list of 36 Spanish words containing the voiced and voiceless stops /p, t, k, b, d, g/, half in utterance initial and half in word medial positions. Results showed that the heritage speakers – or childhood overhearers – produced word initial /p, t, k/ more native-like than the L2 learners. Acoustic measurements of . As Brad Dennison (University of Illinois at Urbana-Champaign) correctly pointed out, the L2 learners in this study were of very low proficiency, enrolled in an elementary Korean class. These learners have clearly not had as much input in Korean as the heritage speaker groups, even if the input the Koreans received was limited. Regardless of age of input, the Korean heritage speakers groups would be expected to outperform the L2 groups, at least in phonology.



Chapter 7.  Incomplete L1 and L2 acquisition in adults 225

Table 7.6  Accent ratings assessment results (maximum 5) (adapted from Knightly et al. 2003). Rated speech type Phonemic task /p, t, k/ /b, d, g/ Narrative task

Native speakers (n = 15)

Heritage speakers (n = 15)

L2 learners (n = 15)

4.4a 4.4 a 5.0 a

3.6b 3.4 b 3.0 b

3.0c 2.8 c 2.4 c

Note: within a row, means with different superscripts are statistically different from each other.

the voiced stops confirmed that the heritage speakers were better than the L2 learners, although the differences were not obvious across the three sounds in different positions. The results of accent ratings of these 6 stops replicated the advantage for the heritage speakers over the L2 learners of the Knightly et al. (2003) study, as shown in Table 7.6. The morphosyntax part of the experiment used a grammaticality judgment task, a noun phrase production task to test gender agreement, a verb-phrase production task eliciting verbal morphology (tense, aspect, person, and number agreement), and an oral narrative to test overall proficiency in the language. The grammaticality judgment task included 60 grammatical and ungrammatical sentences testing gender agreement in nouns, tense, aspect, person and number agreement in verbs, negation, and indirect object. Results showed statistically significant differences between the native speakers and the other two experimental groups. Unlike the results reported for phonology and pronunciation, the heritage speakers and the L2 learners were not different from each other in any of these four measures. Scores are shown in Table 7.7. On the basis of findings such as these, Knightly et al. (2003) concluded that a childhood overhearing advantage is domain-specific; that is, it shows up in phonology but not in morphosyntax. In my view, the conclusion that heritage speakers do not have advantages in morphosyntax is perhaps too strong. After all, morphosyntax is a very broad grammatical domain that includes knowledge of syntax proper, inflectional and derivational morphology, and the interaction of syntax with semantics and pragmatics. Not only do many aspects of morphosyntax vary on linguistic complexity, but they also vary in their frequency in the input, and in their developmental schedule in monolingual children. Furthermore, there are other studies comparing L2 learners and heritage speakers that show selective advantages in morphosyntax as well but, as we will see, these depend on the structures tested, the tasks,

226 Incomplete Acquisition in Bilingualism

Table 7.7  Participants’ performance on Spanish morphosyntax assessment tasks (adapted from Knightly et al. 2003). Measure

Native speakers

Heritage speakers

L2 learners

Grammaticality Judgment Task Percentage correct

91.8a

63.6b

62.5 b

Noun phrase production (in %) Gender agreement Number agreement

94.8 a 93.8 a

66.3 b 82.5 a

72.7 b 92.3 a

Verb phrase production (in %) Tense/aspect person number

94.1 a 98.2 a 98.6 a

50.4 b 68.6 b 70.7 b

50.0 b 72.3 b 80.3 b

  4.93 a

  2.5 b

  2.6 b

Narrative production (max = 5)

and the proficiency of the speakers. Thus, this study does not necessarily disconfirm Prediction 1. One of the first studies comparing morphosyntactic knowledge of heritage speakers and L2 learners and showing an advantage for heritage learners is Håkansson’s (1995) study of five Swedish bilingual “expatriates.” Three of these Swedish speakers grew up, and were schooled, in the United States, while the two others grew up in France and Sweden but were schooled in French. At age 20, the five bilingual Swedish speakers returned to Sweden to start college and were taking Swedish as a foreign language class at the time of testing. As Håkansson described, these speakers had the advanced oral skills of a Swedish native speaker, yet they failed the vocabulary and grammar parts of a written Swedish placement test. At first glance, this observation is consistent with Prediction 2: Heritage speakers should be more accurate and faster than L2 learners in oral than in written production and comprehension tasks that minimize metalinguistic knowledge. Although the focus of Håkansson’s study was to assess degree of attrition in the bilingual speakers, the results of this small-scale study also speak to a potential advantage in knowledge of syntax for heritage language learners. The main purpose of this study was to tease apart the potential selectivity of attrition, or whether it affects linguistic subsystems (morphology, phonology, syntax, vocabulary) in different ways. Framed within generative grammar, and consistent with the Weaker Language as L1 Hypothesis, Håkansson hypothesized that parameter values set early in L1 acquisition should not undergo attrition, and set out to analyze knowledge of verb second (V2) and gender and number agreement on NPs.



Chapter 7.  Incomplete L1 and L2 acquisition in adults 227

From all the subjects, Håkansson collected written tests and free compositions, as well as spontaneous production data from recorded interviews. The results of the Swedish heritage speakers were compared to those of 6 L2 learners of Swedish whose native languages were English, Chinese, Swahili and Persian. Examples of Swedish word order were provided in Chapter 4, §2.2. Swedish nouns are inflected for gender (common vs. neuter), number, and definiteness. Determiners and adjectives agree with the noun. Swedish also marks definiteness on the article and noun when there is an attributive adjective in the noun phrase. (7) en god kaka a-common good-common cake ‘a good cake’ (8) ett gott vin a-neuter good-neuter wine ‘a good wine’ (9) mina goda kakor my-pl good-pl cake-pl ‘my good cake’ (10) mitt goda vin my-neut good-def wine ‘my good wine’ (11) kakorna är goda. cake-pl-def are good-pl ‘These cakes are good.’

Results showed revealing differences between the L2 learners and the Swedish heritage speakers. The V2 pattern is regarded as one of the most difficult structures for L2 learners of non-V2 languages to acquire. Here, the Swedish heritage speakers were significantly better than the L2 learners. While all the L2 learners made errors with XVS (producing XSV), it is reported that none of the Swedish heritage speakers made errors with these structures. With respect to the ratio of SV and VS order, the results also revealed differences between the two groups. The Swedish heritage speakers were more like Swedish native speakers in their production rate of SV and VS order (which is in the 60% and 40% range for native  . Although we do not expect these L2 learners to speak like Swedish native speakers, it is not clear whether or not the L2 learners and the Swedish heritage speakers were comparable in their written proficiency, since no information is provided.

228 Incomplete Acquisition in Bilingualism

speakers), while the L2 learners produced SV well above 82% of the time. Thus, in terms of syntax, not only have the Swedish heritage speakers acquired and retained aspects of their L1, but they were also superior to the L2 learners. The same cannot be concluded from the results of gender agreement on NPs, which were very different. Here, the L2 learners outperformed the Swedish heritage speakers, who produced up to 88% errors. Errors were made with gender and number, with possessives, and in predicative structures with adjectives after a verb. Rather than errors of omission, as is common in L1 acquisition, the heritage speakers made more errors of commission (using wrong morphemes in different contexts), common in L2 acquisition. It seems that while the bilinguals have not lost the syntactic rule for agreement, they had trouble retrieving the right affixes for a given function. Although the number of subjects participating in Håkansson’s study is very small, the findings are relevant to the claim that incomplete acquisition affects different aspects of morphosyntax selectively. Clearly, the V2 phenomenon is a purely syntactic rule, whereas gender agreement is a syntactic rule that depends on lexical knowledge. It may be the case that the reason why heritage speakers do not display advantages in gender has to do with lexical retrieval and frequency effects rather than with the abstract syntactic rule of agreement. This study also shows that heritage speakers set parameters for the weaker language (their L1), whereas L2 learners do not. In German, Meisel (1986) showed that V2 is acquired between MLU 1.75 and MLU 2.25, quite early in child language. That heritage language speakers set the parameters of their weak L1 early in childhood whereas L2 learners may take longer, is further confirmed by two other studies I carried out: Montrul (2004a) and Montrul and Rodríguez Louro (2006). In these two studies, I compared L2 learners and heritage speakers’ knowledge of syntactic aspects of the null subject parameter in Spanish. As a pro-drop language, Spanish has rich verbal agreement that licenses and identifies subjects, and pronominal subjects can be omitted. However, in addition to this morphosyntactic requirement, subject omission is regulated by discourse-pragmatic factors, such as topic, switch reference, and focus or contrast, as in examples (12)–(15) (see also discussion on Italian in Chapter 3). (12) Pepe no vino hoy a trabajar. *Pepe/?él/Ø estará enfermo. Pepe no came today to work Pepe/?él/Ø will be sick ‘Pepe did not come to work today. He must be sick.’

same referent

(13) Hoy no fui a trabajar. Pepe/él/*Ø pensó que estaba enferma. today I no went to work Pepe/él/*Ø thought that I was sick ‘Today I did not go to work. Pepe/he thought I was sick.’

different referent



Chapter 7.  Incomplete L1 and L2 acquisition in adults 229

(14) ¿Quién vino? Él/Mario/*Ø vino. ‘Who came? He/Mario/*Ø came.’

topic

(15) El periodistai dijo que éli no había escrito ese reporte. the journalist said that he (himself) not had written that report ‘The journalist said that he had not written that report.’

focus

The participants in Montrul (2004a) were 14 heritage speakers of advanced proficiency in Spanish and 10 of intermediate proficiency. There was also a control group of 20 monolingual speakers. Montrul and Rodríguez Louro (2006) tested 16 near-native speakers, 16 advanced and 16 intermediate L2 learners of Spanish, using the same tests as Montrul (2004a). Results of the two studies are summarized in Tables 7.8 and 7.9. In Montrul (2004a) I found that bilingual speakers have robust knowledge of the morphosyntax of subjects in Spanish: they produced both overt and null subjects as well as postverbal subjects. However, Table 7.8. shows that the intermediate heritage speakers were significantly different from the advanced heritage speakers and the monolinguals in the production of overt and null subjects: they overproduced overt subjects in

Table 7.8  Combined results of Montrul (2004a) and Montrul & Rodríguez Louro (2006). Percentage correct agreement, null, overt and postverbal subjects. Groups

N

% Correct agreement

% Overt subjects

% Null subjects

% Postverbal subjects

Monolinguals

16

99

42.8

57.2

20.2

Advanced L2 learners Advanced Heritage speakers

16 14

97.35 99

42.5 47.2

57.5 52.7

19.1 31.8

Intermediate L2 learners 16 Intermediate Heritage speakers 10

87.45 97.45

68.3 68.6

31.7 31.4

 0 24.3

Table 7.9  Combined results of Montrul (2004a) and Montrul & Rodríguez Louro (2006). Percentage correct, redundant overt, and illicit null subjects. Groups

N

% Overt subjects

% Null subjects

Monolinguals Advanced L2 learners Advanced Heritage speakers Intermediate L2 learners Intermediate Heritage speakers

16 16 14 16 10

total 42.8 42.5 47.2 68.3 68.6

total 57.2 57.5 52.7 31.7 31.4

correct 100   92   99.3   77.1   91.5

redundant  0  8   0.7 22.9   8.5

correct >99 90 92 > 99 84.5

illicit <1 10  8 <1 15.5

230 Incomplete Acquisition in Bilingualism

same referent contexts (redundant subjects) and also overproduced null subjects in switch reference contexts, although to a lesser degree. My interpretation of these results is as follows. The intermediate L2 learners were still in the process of resetting some morphosyntactic aspects of the Null Subject Parameter because they displayed significantly more agreement errors than the intermediate heritage speakers (12.55% vs. 2.55%) and produced no postverbal subjects (0% vs. 24.3%) (Table 7.8). They also had a significant amount of errors with the pragmatic distribution of null and over subjects (Table 7.9). Similar advantages for heritage speakers over L2 learners on clitic climbing, which is related to the null subject parameter according to Kayne (1989), are demonstrated in Montrul (under review). I further examined the interaction between proficiency and grammatical knowledge in L2 learners and heritage speakers in Montrul (2005), where I found advantages for heritage speakers of low proficiency in the language. The study was about semantic and syntactic knowledge of unaccusativity. Some basic background on unaccusativity in Spanish was already given in Chapter 6 in the context of Montrul (2006) (see §2.4 in Chapter 6). Constructions that in Spanish discriminate between unaccusative and unergative verbs are postverbal bare plural generic subjects (grammatical with unaccusative verbs), and the absolutive construction. Participants were a control group of 28 Spanish native speakers from a variety of Spanish-speaking countries, 71 English-speaking L2 learners of Spanish (ranging from low-intermediate to advanced), and 36 early bilingual speakers of Mexican background living in the United States. The main instrument was the same written grammaticality judgment task used in Montrul (2006) discussed in Chapter 6, §2.4. Results of this study showed that intermediate and advanced L2 learners and heritage speakers displayed robust knowledge of the syntax of unaccusativity, revealing the exact same pattern of discrimination and response as the monolingual control group with all the sentence types tested. Yet an important qualitative difference between L2 learners and the early bilingual speakers emerged at the low proficiency level, the results of which are shown in Table 7.10. While the low proficiency heritage speakers discriminated syntactically (and statistically) between unaccusative and unergative verbs in sentences with postverbal subjects, with postverbal bare plural subjects, and in the participial absolutive construction, the low proficiency L2 learners did not show any type of semantic or syntactic discrimination, rating unaccusative and unergative verbs statistically alike in all these constructions. In conclusion, it appears that at lower levels of proficiency, and despite the fact that their grammars are still quite incomplete with respect to the target grammar, the study on Swedish V2 and Spanish null subjects show that heritage speakers are superior to L2 learners with parameters that are assumed to be set very early



Chapter 7.  Incomplete L1 and L2 acquisition in adults 231

Table 7.10  Mean ratings on sentences with unaccusative and unergative verbs by L2 learners and heritage speakers of low proficiency in Spanish (adapted from Montrul 2005). Groups

N

L2 learners Heritage speakers

Postverbal subjects Bare plural postverbal subjects

Absolutive construction

unaccus. unerg. (gram.) (gram.)

unaccus. unerg. (gram.) (ungram.)

unaccus. unerg. (gram.) (ungram.)

3.14 3.88

3.11 2.85

25 3.29   5 3.94

3.25 3.55

3.08 3.21

3.03 2.05

Note: 5 = grammatical, 1 = ungrammatical

in monolingual and early bilingual grammars. Since knowledge of unaccusativity emerges very early in monolingual children, it is not surprising that the Spanish heritage speakers have an advantage in this area as well. These findings support my first prediction – that heritage speakers have advantages over L2 learners in aspects of grammar acquired early in childhood – reinforcing the idea that the weaker language in these adult early bilinguals was acquired as a first language and with the same linguistic mechanisms available for all children. Just like an L2, reduced input makes the L1 look incomplete with respect to the target.

4.2

Written vs. oral tasks and types of knowledge

Not only do heritage and non-heritage learners vary on age of acquisition and proficiency, but they also vary on mode and context of acquisition, as summarized in Table 7.4. Heritage learners acquired the language aurally and in a naturalistic setting, whereas L2 learners acquire the language both written and orally, typically in a classroom setting. As such, their literacy skills in the target language are much more developed than in heritage speakers, even when heritage language learners can transfer literacy skills from their stronger language onto the weaker . Lee, Kim, Kong, Hong and Long (2008) is another recent study investigating the linguistic profiles of advanced Korean heritage speakers and Korean L2 learners (with L1 English). All learners had reached the advanced or intermediate high level of ACTFL OPI standards, and were asked to perform a guided oral narrative. Secondly, they completed a written test with a fill in the blanks part and a cloze part. The test assessed structural knowledge of particles, connectives, complex predicates, lexical knowledge and idiomatic expressions and mimetics. Lee et al. found substantial differences between the heritage and the non-heritage Korean learners, with the heritage Korean learners performing well above the L2 learners on most items and grammatical structures tested, except for idiomatic expressions, where the groups did not differ.

232 Incomplete Acquisition in Bilingualism

Table 7.11  Percentage accuracy on Japanese oral production and reading comprehension (adapted from Matsunaga 2003).

Heritage Learners L2 Learners (kanji) L2 learners (no kanji)

N

Oral production

Reading comprehension

14 11 14

94% 77% 75%

80% 88% 60%

language. Therefore, Prediction 2 states that heritage learners should be more accurate than L2 learners in tests of oral production and in written tasks that minimize explicit knowledge of grammatical rules. Although an advantage in oral production is an observation that many teachers make (often anecdotally), few empirical studies have been conducted to test these impressions. Matsunaga (2003) is one such study offering confirming evidence for this prediction. Matsunaga tested Japanese heritage speakers’ oral and reading skills against those of L2 learners (some of whom were Chinese and knew kanji, while others did not know any kanji, like the Japanese heritage speakers). The participants were 30 intermediate and advanced students of Japanese: 14 were heritage speakers of Japanese, 11 were Chinese-speaking L2 learners with knowledge of kanji, and the remaining 14 were L2 learners with no knowledge of kanji. All participants were asked to read three written passages with some words in kanji. The oral task consisted of an oral interview in which participants were asked to describe the picture of a family. The interviews were tape recorded and transcribed for analysis. Oral performance was rated on the following categories: pronunciation, vocabulary, grammatical accuracy, fluency and communication. Results revealed that the Japanese heritage language learners had significantly higher oral proficiency than the Japanese L2 learners, but their reading proficiency was about the same as that of the L2 learners with knowledge of kanji, as summarized in Table 7.11. While Matsunaga showed global results for overall performance on oral and written performance, Montrul, Foote and Perpiñán (2008) focused on knowledge of gender assignment and agreement in Spanish which, as we have seen throughout this book, is a highly volatile area in child language attrition and incomplete acquisition. Furthermore, several recent studies within the generative framework have been advancing the claim that gender agreement is also vulnerable to incomplete acquisition and fossilization in L2 learners whose L1 lack an uninterpretable grammatical feature for gender, such as English (Carroll 1989; Hawkins & Franceschina 2004; Franceschina 2005). By contrast, other studies have supported the . Provided the two languages share a similar writing system.



Chapter 7.  Incomplete L1 and L2 acquisition in adults 233

opposite view, namely that gender features are acquirable in a second language past puberty (Bruhn de Garavito & White 2003; White, Valenzuela, KozlowskaMacgregor & Leung 2004). Since gender agreement is problematic for early and late bilinguals, therefore, the purpose of Montrul et al.’s (2008) study was to test whether heritage speakers of low-intermediate proficiency in the language would have advantages over proficiency-matched L2 learners (Prediction 1). But what is relevant for my second prediction, Prediction 2 (see Table 7.5), is that the study included two written tasks and an oral task. The first written task tested comprehension of gender agreement and adapted the task developed by White et al. (2004). Subjects were presented with 48 sentences preceded by a short context and 3 pictures. The sentences capitalized on the syntactic phenomenon of noun drop in Spanish, or the possibility to elide nouns, as shown in (16). It is believed that noun drop is possible in Spanish because the gender and number agreement on the determiner allows identification of the noun referred to in context. (16) ¿Qué manzana quieres? La roja.   which apple you want the red   ‘Which apple do you want? The red one.’

The task included sentences targeting masculine and feminine nouns, followed by three pictures each. Participants had to read the sentence and choose the picture the sentence referred to, as shown in (17) (17) “No quiero llevar las de ese color.” “I don’t want to take those ones of that color”

A bufanda B maletas C pantalones ‘scarf ’ ‘suitcases’ ‘pants’

The second written task was a passage with deleted determiners and adjectives (like a cloze test). Participants were to complete the passage by choosing the correct feminine or masculine form of the determiner or adjective. Finally, the oral task was a picture naming task consisting of 65 photographs of objects, people

234 Incomplete Acquisition in Bilingualism

Table 7.12  Spanish heritage speakers and L2 learners’ overall percentage accuracy in three tasks testing gender agreement (from Montrul, Foote & Perpiñán 2008). Groups

N

Written gender recognition task

Written picture Oral picture interpretation task description task

Native speakers Heritage Speakers L2 Learners

20 66 65

97.4 84.6 89.5

99.1 83.3 88.5

99.9 89.7 72.1

or animals. Participants were asked to describe each picture by using the carrier sentence Veo det N adj “I see determiner noun adjective.” Results of the two written tasks (Table 7.12) showed that the L2 learners were statistically more accurate on gender marking than the heritage language learners, while the heritage language learners were significantly better and faster than the L2 learners in the spontaneous oral picture naming task. Without going into the specific details of each task, suffice it to say that, in common with all other existing studies of gender marking and agreement, L2 learners and heritage speakers were less accurate on feminine than on masculine, agreement with determiner was better than agreement with adjective, agreement with canonical words (-a feminine and -o masculine) better than with non-canonical and exceptional forms. This confirms the same trends reported in many previous studies. Furthermore, while the heritage speakers had comparable performance in the three tasks (close to 80% accuracy), the L2 learners had a more skewed performance of above 85% accuracy in the two written tasks and only 72.1% in the oral task. The results of the L2 learners are fully compatible with the Missing Surface Inflection Hypothesis because L2 learners appear to have knowledge of features of Spanish gender as revealed by the two written tasks, but have difficulty producing correct agreement in oral production. Interestingly, a critical-period-based theory like the Failed Functional Features Hypothesis (Hawkins & Chan 1997) does not appear to explain the results of either group, if we take the results of all the measures together. If we make specific assumptions about what type of knowledge each task reveals, then the results are compatible with other theoretical positions. If the results of the oral task are taken as the most representative measure of implicit linguistic knowledge (R. Ellis 2005), then a critical period effect can be maintained. Under this assumption, the results of the oral task may be more representative of fast, implicit, and automatically processed knowledge (typically acquired early in childhood according to DeKeyser 2000, 2003; M. Paradis 2004 and Ullman 2001), while the results of the written tasks could reflect metalinguistic, explicit knowledge (typically acquired later). Since metalinguistic knowledge can compensate



Chapter 7.  Incomplete L1 and L2 acquisition in adults 235

Table 7.13  Spanish heritage speakers and L2 learners’ overall mean ratings on a grammaticality judgment task testing gender agreement in Spanish Groups

N

Grammatical

Ungrammatical

Native speakers Heritage Speakers L2 Learners

20 66 65

4.92 4.52 4.72

1.03 3.06 2.91

Note: 5 = grammatical, 1 = ungrammatical

for gaps in implicit competence (M. Paradis 2004), L2 learners appear to be better than the heritage speakers with gender in the written tasks when in fact they may not be better with gender in general after all.Thus, both L2 learners and heritage speakers know something about gender in Spanish, but such knowledge may be stored and represented differently, at least at this level of proficiency. The results of this study may be questioned on the grounds that the written tasks used are too metalinguistic and cannot be used to claim that L2 learners are using implicit knowledge primarily. Therefore, we need to look at other less metalinguistic tasks. Birdsong (1989) cites Bialystok and Ryan’s (1985) metacognitive model of linguistic and metalinguistic functions, according to which the cognitive dimensions of control (low and high) and analyzed knowledge (low and high) are orthogonal to each other. Metalinguistic skills require high control and high analyzed grammatical knowledge, whereas conversation, for example, requires low analyzed knowledge and low control. While the two written tasks used in Montrul, Foote and Perpiñán would require high control and high analyzed knowledge, according to Birdsong’s hierarchy of tasks, grammaticality judgment tasks require moderate analyzed knowledge. In fact, as part of a larger study, these same participants took a written untimed grammaticality judgment task, which also taps explicit knowledge to some extent, but perhaps less so than cloze-type tasks. Among the filler items were grammatical and ungrammatical sentences testing gender. The results displayed in Table 7.13 showed no statistically significant differences between the L2 learners and the heritage speakers on gender errors in this task. Not only are results such as these very revealing, but they also lead us to question the results of many studies that have used untimed grammaticality judgment tasks to compare the linguistic ability of L2 learners and heritage speakers, finding no difference in group performance (Au et al. 2002; Bruhn de Garavito 2002). Grammaticality judgment tasks may have to be supplemented with other measures to gain a better understanding of the linguistic command of heritage speakers and proficiency-matched L2 learners. Further evidence of adult early bilinguals having more automatized, implicit, or integrated knowledge than adult L2 learners – Prediction 2 – is provided

236 Incomplete Acquisition in Bilingualism

by Guillelmon and Grosjean’s (2001) study. Focusing on spoken word recognition, Guillelmon and Grosjean investigated how early and late English-French bilinguals reacted to gender marking when processing French sentences. Their assumption was that early bilinguals, who acquired the second language before puberty, would be sensitive to gender matches and mismatches, whereas late bilinguals, who acquire the second language after puberty, would find gender error recognition in spoken language extremely difficult, especially when their first language is not a gender-marking language. In particular, Guillelmon and Grosjean conducted two experiments to find out whether the same trends would hold for speech perception and oral comprehension. The first experiment tested a group of 32 monolingual French speakers and a group of 32 early French-English bilinguals (simultaneous and sequential) exposed to English and French before age 13 (mean age of French acquisition 4 years). Based on self-assessed scales, the participants reported a very high command of English and French production and comprehension, although the researchers report that these participants were slightly dominant in French. Participants heard short noun phrases consisting of a determiner, a noun and an adjective, and were asked to repeat the noun. In some conditions, the gender marking on the determiner agreed with the noun or was neutral (same for feminine and masculine nouns), while in other conditions the gender marking on the determiner was the opposite of the noun (an error) or neutral. (e.g. le joli bateau (correct), *la jolie bateau (incorrect), leur joli bateau (neutral). Reaction times were recorded for each participant. Results of the first experiment showed that the early bilinguals and the monolinguals were equally sensitive to gender marking since they named correct gender-marked nouns faster than neutral nouns in the correct-neutral condition and named incorrect nouns slower than neutral nouns in the incorrect-neutral condition. The second experiment used the same materials and experimental procedures as experiment 1, except for another control group of 32 French monolingual speakers and an age-matched group of 32 late bilinguals, with a mean age of 15.11 of first exposure to French as a second language. Although the late bilinguals had advanced proficiency in the two languages, self-reported ratings indicated slightly stronger English than French skills. The results of the monolingual group confirmed a gender marking effect in the two conditions (neutral-correct, neutral incorrect). By contrast, the late bilinguals showed no effect in either condition. They took the same time to name nouns in the neutral, correct and incorrect contexts. Results of the two studies are summarized in Table 7.14. Therefore, this study showed that while early bilinguals are like monolinguals in both speed and sensitivity to gender marking, the late bilinguals are very different: they are both slower in naming words and do not recognize gender marking



Chapter 7.  Incomplete L1 and L2 acquisition in adults 237

Table 7.14  Naming speed (in ms) in a spoken word recognition study (adapted from Guillelmon & Grosjean 2001). Experiment 1

Correct Neutral Incorrect Neutral

Monolingual A (n = 32) 479 498 513 483

Experiment 2 Early bilinguals (n = 32) 481 525 574 519

Monolingual B (n= 32) 521 545 594 547

Late bilinguals (n = 32) 620 620 632 626

errors. Guillelmon and Grosjean interpret these findings to suggest that a gender feature cannot be triggered after a critical period, following Carroll (1989). Thus, so far we have seen that Prediction 2 – Heritage speakers should be more accurate and faster than L2 learners in oral than in written production and comprehension tasks that minimize metalinguistic knowledge – appears to hold. Type of task matters when we compare early and late bilinguals with different onsets of bilingualism and mode of acquisition. Not only do important differences arise when we test bilinguals in written and oral tasks that require different degrees of metalinguistic analysis and control, but we also find that in psycholinguistic oriented tasks measuring reaction times, heritage speakers may also have a processing speed advantage over L2 learners. Guillelmon and Grosjean’s study also showed that the early bilinguals were faster overall than the late bilinguals in naming nouns. Montrul, Foote,Perpiñán, Thornhill and Vidal (2006) investigated knowledge of accusative object clitics and word order in Spanish, confirming a processing advantage for heritage speakers over age-matched L2 learners in a written on-line picture-sentence matching task.10 In one experiment, the two groups and a control group of 15 native speakers completed a written grammaticality judgment task. The task included sentences with correct and incorrect placement of clitics in Spanish. Clitic placement varies depending on the finiteness of the verb: it is preverbal when the verb is finite, as in (17), and postverbal when it is non-finite, as in (18). (17) Juan la vio. vs. *Juan vio la. Juan it saw ‘Juan saw it.’

10. Cagri (2008) reported similar results with on-line oral tasks and Farsi heritage speakers and L2 learners: comparable results for accuracy but heritage speakers are faster than L2 learners in all the processing tasks used.

238 Incomplete Acquisition in Bilingualism

(18) Juan quiere verla. vs. *Juan quiere la ver. Juan wants see it ‘Juan wants to see it.’

Furthermore, Spanish has flexible word order and allows sentences with preverbal objects, either with clitics, as in (19), or with clitic left dislocations, as in (20). (19) a. b.

La mujer lo besa. the woman him kisses ‘The woman kisses him.’ Lo besa la mujer. him kisses the woman

(20) a. b.

Juan tiene las carpetas en la oficina. ‘Juan has the folders in his office.’ Las carpetas las tiene Juan en la oficina. the folders them has Juan in the office ‘The folders, Juan has them in the office.’

Results of the grammaticality judgment task (with acceptability ratings ranging from 5 = grammatical to 1 = ungrammatical) showed no significant difference between L2 learners and heritage speakers recognizing the correct placement of clitics with respect to the verb, as in examples (19a,b). However, the heritage speakers were more accepting than L2 learners of sentences with clitic left dislocations, as in (20), as shown in Table 7.15. To pursue the strength of this result, a second experiment tested the on-line processing of sentences with clitics and alternative word orders, such as (20a,b), with a visual sentence-picture matching task. The same participants were presented with two pictures, A and B, on a computer screen with a sentence underneath. The pictures depicted the same action, but with the participants reversed (e.g., a boy calling his parents vs. the parents calling their son). Participants had to decide as quickly as possible which picture the sentence described, by pressing A or B on the keyboard. The time (in milliseconds) was measured from the presentation of the slide until the participant made a decision and pressed the A or B key. Table 7.15  Spanish heritage speakers’ and L2 learners’ overall mean ratings (max = 5) on a grammaticality judgment task testing gender agreement in Spanish (adapted from Montrul et al. 2006). Groups

N

S-V-O

Clitic left dislocations

Monolinguals L2 learners Heritage speakers

19 27 23

5 4.72 4.86

4.73 1.72 3.38



Chapter 7.  Incomplete L1 and L2 acquisition in adults 239

Figure 7.1  Visual picture-sentence matching task. Mean reaction times (in ms.) on sentences with clitics (adapted from Montrul et al. 2006).

(21) a. b. c. d.

Lo llaman por teléfono sus padres. him call by phone his parents Juan llama por teléfono a sus padres. Juan calls by phone his parents. Sus padres lo llaman por teléfono. his parents him call by phone A sus padres los llama Juan por teléfono. his parents them call Juan by phone

(Cl-V-S) (S-V-O) (S-Cl-V) (O-Cl-V-S)

In terms of overall speed and accuracy, the results of this study showed that the heritage speakers and the L2 learners made comparable amounts of errors (80.7% vs. 81.8%) but in terms of speed, the heritage speakers were faster than the L2 learners (native speaker control 3766 ms, L 2 learners 5055ms, heritage speakers 4120ms). Figure 7.1 shows the results by sentence types. The results of the on-line visual picture matching task showed an advantage for heritage speakers over L2 learners with all sentences with clitics in speed but not in accuracy. The heritage speakers were as fast as the monolingual control group with all sentences containing clitics, and especially with sentences containing objects in sentence initial position (cl-V-S and O-cl-V-S). In short, the substantial advantage for early bilinguals in speech and sentence processing, strongly suggests that heritage speakers appear to have more implicit, native-like knowledge of clitics than L2 learners, even when they have low-to-intermediate proficiency in the language. Or at least, that such knowledge is more automatic, integrated and retrieved faster than in L2 learners. This may be due to

240 Incomplete Acquisition in Bilingualism

the fact that the clitic system was acquired by the heritage speakers early in childhood, before the age of 4, as in monolingual children. By contrast, L2 learners, who have been exposed to Spanish in a classroom setting, may have more metalinguistic knowledge of Spanish word order at this point in their development, and are still in the process of acquiring the Spanish clitic system. Although entirely possible according to Green (2003) and Abutalebi and Green (2007), whether such metalinguistic knowledge in L2 learners can, and will, become more automatic as proficiency increases is an open question that needs to be tested empirically with more advanced L2 learners and heritage speakers. In conclusion, Prediction 2 is largely confirmed so far: early bilinguals with incomplete knowledge of the heritage language are better than proficiency matched L2 learners of the same language in oral tasks and in tasks that require fast and less consciously analyzable linguistic knowledge.11 To the extent that performance on these types of tasks can be taken as indirect evidence for implicitly learned and stored linguistic knowledge, the heritage speakers appear to possess more of that type of knowledge in their L1 than L2 learners have in their L2.

4.3

Reactivity to classroom instruction

If heritage language learners received early input in the heritage language, and this early input is crucial for language learning even if exposure to and use of the language were reduced later in childhood, will there be a re-exposure effect when heritage speakers seek to re-learn their language later in adulthood? Prediction 3 stated that heritage speakers should react faster and better to instruction than L2 learners if that early input really matters. In recent years, traditional foreign/second language classrooms in Western Europe and North America have seen an increasing number of speakers of minority languages who want to reacquire, expand or maintain their heritage language skills. As a result, and depending on the status of the minority language in the community, heritage speakers and second language learners either share the same classrooms, or special programs are being created to better serve the linguistic, cultural and professional needs of heritage language speakers (Brinton, Kagan & Bauckus 2008; Montrul, in press).

11. However, in a production study of agreement attraction errors in early and late SpanishEnglish bilinguals, Foote (2006) found that the advanced late bilinguals (L2 learners of Spanish) made overall fewer errors than the advanced heritage language speakers. Foote attributed this finding to enhanced awareness and monitoring on the part of the L2 learners, most of whom were teachers of Spanish.



Chapter 7.  Incomplete L1 and L2 acquisition in adults 241

For many language educators, there is a sense that heritage speakers possess substantial knowledge of the language due to early exposure in childhood, despite the fact that many of these speakers also exhibit distinctive gaps in their language skills. These impressions are largely confirmed by the empirical evidence discussed so far. In essence, many heritage learners (but not all, and depending on the heritage language) possess more native-like levels of pronunciation and fluency, large vocabularies, some command of core syntax, good listening and speaking abilities, and familiarity with cultural norms of the language and culture. However, heritage speakers have non-native accents and do not display full command of several areas of morphosyntactic knowledge, in particular areas that interface with other grammatical domains, including some aspects of inflectional morphology, discourse-pragmatics and semantics. Once we identify specific grammatical areas that seem incompletely acquired by heritage speakers, the question is whether these learners will be sensitive and reactive to focused instruction and practice in the classroom. According to my third prediction, Prediction 3, if early exposure makes a difference for degree of ultimate attainment, then heritage speakers should be able to reactivate that linguistic memory and somehow learn the grammatical property in question faster and better than L2 learners who did not receive such early exposure. The studies on adult Korean adoptees discussed in Chapter 6 appear to suggest, albeit very weakly, that this prediction is on the right track. Those adoptees who were re-exposed to Korean through trips to Korea showed a higher sensitivity to a specific phonetic contrast and early Korean vocabulary than those who had not been re-exposed to Korean at all. Since this is a very new area of research, with the exception of Song, O’Grady, Cho and Lee (1997), a study on Korean heritage children discussed in Chapter 5, and Bowles and Montrul (in press), Montrul and Bowles (in press b), and Potowski, Jegerski and Morgan-Short (in press), there are virtually no studies addressing these important issues, but hopefully future research will continue to emerge. Song et al. (1997) studied knowledge of overt case markers in Korean with a written sentence comprehension task and found that the bilingual children (ages 3–8) performed above chance on this task. The first goal of this study was to understand how monolingual and bilingual Korean children differed on their grammatical knowledge of certain morphosyntactic aspects of Korean. The other goal was to identify possible deficits in the children’s verbal skill in order to address them through focused instruction. Therefore, the results of the main study were addressed by a follow-up study. A different group of Korean bilingual children (ages 7–12) born and raised in Hawaii completed the sentence comprehension task as a pre-test. Results indicated very low performance by the children on OSV sentence patterns as compared with SVO patterns, confirming the trends of the main study with a comparable group of bilingual speakers.

242 Incomplete Acquisition in Bilingualism

Figure 7.2  Korean children’s performance on a sentence interpretation task before and after instruction (adapted from Song et al. 1997).

Song et al. examined the textbooks used in class and found that there were hardly any examples of OSV sentences in the materials. Furthermore, most of the sentences exemplified in the textbooks were SVO sentences where the case markers did not contribute to meaning, since the sentence could be interpreted on semantic or pragmatic grounds, as in John throws the ball. Song et al. (1997) therefore designed a set of instructional materials and activities focusing on sentences with OSV and SVO orders, where case markers are crucial for interpreting who is doing what. Furthermore, the children received explicit information about case markers in Korean with different kinds of sentences. The children demonstrated actions with stuffed animals and were asked to describe the action using a sentence. Other classroom activities included a comprehension task with cards and an act-out task with the children in costumes. During all activities, the teacher provided corrective feedback when errors were made. Children received this training for 50 minutes of class for two weeks after the pretest. The children were then administered a post-test and nine weeks later a delayed posttest. Results of pre-test, immediate post-test and delayed posttest are shown in Figure 7.2. The results showed a noticeable improvement with OSV sentences from pretest (25% accuracy) to immediate post-test 66.37% accuracy. Long-term instruction had positive effects, because the magnitude of the improvement remained relatively stable nine weeks later. (Similar improvement after instruction, practice and feedback is reported for another experiment with the reflexive anaphor caki.)



Chapter 7.  Incomplete L1 and L2 acquisition in adults 243

The results of this experiment suggest that heritage learners are very sensitive to corrective feedback, and that they can bring back, albeit still far from completely, complex aspects of morphosyntax to which they were exposed in childhood but which were incompletely acquired. Although highly promising, this study is about children and did not compare the magnitude of the change in a comparable group of L2 learners. A study of adult heritage speakers on this issue is Montrul and Bowles (in press b). Recall from Chapter 6 that Spanish heritage speakers sometimes omit the preposition a with animate/specific direct objects in Spanish (a-personal), a well-known instance of the phenomenon of Differential Object Marking (­Montrul 2004a; Montrul & Bowles, in press a). L2 learners of Spanish at early stages of development also do not typically produce this case marker with direct objects in written or oral discourse, and if they do, their performance seems random (*Marisa conoce mi hermana instead of Marisa conoce a mi hermana ‘Maria knows my sister.’). Perusal of textbooks at this level reveals that this structure, although highly frequent in the input, is not explicitly taught or, consequently, practiced in L2 classrooms to the same extent that other structures are. One reason why a-personal may be difficult at this stage is because there is no such marker in the learners’ L1 (English), and the preposition marks dative case in many other constructions, like indirect objects (María le dio flores a su madre “Maria gave flowers to her mother”) and some psychological verbs (A Juan le gusta bailar “Juan likes to dance”). Even if the learners do not produce the structure, since it is salient in the input, one question is whether they can still recognize its grammaticality. Bowles and Montrul (in press b) investigated whether explicit, focused instruction and practice can help learners at this early stage of development to incorporate a personal into their interlanguage, in terms of both recognition and production. Montrul and Bowles (in press b) used the exact same methodology with Spanish heritage speakers. Like Song et al. (1997), Bowles and Montrul (in press) is also a classic preposttest design. Bowles and Montrul tested 30 fourth-semester college level (low-intermediate) L1 English-speaking students of Spanish who had not been formally exposed to a personal. In the first session, participants completed a language background questionnaire and a pre-test consisting of two tasks specifically targeting the use of a with different classes of verbs and objects. The first task was a written elicited production task consisting of 25 sentences. The second task was a written grammaticality judgment task (GJT). Learners were asked to rate grammatical and ungrammatical sentences consisting of transitive, ditransitive, and psychological verbs on a five-point Likert scale ranging from 1= ungrammatical to 5 = grammatical.

244 Incomplete Acquisition in Bilingualism

Figure 7.3  Spanish L2 learners’ judgments on grammatical and ungrammatical sentences with differential object marking after instruction (from Bowles & Montrul in press).

One week after the pre-test, learners completed the instructional treatment, which consisted of a web-based explicit grammatical explanation of the uses of a with all three classes of verbs and three practice exercises which required learners to recognize the use or non-use of a. For all practice exercises done on computers, learners received immediate, explicit feedback in the form of a metalinguistic explanation, whether their answer was correct or incorrect. In the next class session, learners completed the production and grammaticality judgment task as the immediate post-test. Three weeks later, learners completed the delayed post-test, which consisted of the two tasks again. The tests were also administered to a control group of 10 native speakers, whose mean acceptability judgments were as follows: animate grammatical (with a) 4.95, animate ungrammatical (no a) 1.1, inanimate grammatical (no a) 4.96, inanimate ungrammatical (with a) 1.2. Figure 7.3 shows that at the pre-test the L2 learners do not recognize a personal, but results from the immediate post-test reveal that at least some learners made gains in recognition. If L2 learners have problems with a-personal, what about Spanish heritage speakers? Montrul and Bowles (in press b) assessed the effects of instruction on Differential Object Marking in 14 adult heritage language learners. The instructional treatment was the same as in Bowles and Montrul (in press): an on-line delivered, explicit grammatical explanation of the uses of “a” followed by three practice exercises, for which participants received immediate, explicit feedback (including negative evidence). Results of the pre-test and post-test are shown in Figure 7.4.



Chapter 7.  Incomplete L1 and L2 acquisition in adults 245

Figure 7.4  Spanish heritage speakers’ judgments on grammatical and ungrammatical sentences with differential object marking after instruction (from Montrul & Bowles in press b).

Results of the pre-test confirmed that the heritage language learners’ recognition and production of Differential Object Marking (a personal) with animate objects was probabilistic. Post-test results, however, revealed highly significant gains, suggesting that negative evidence facilitates classroom heritage language acquisition, as it does with L2 acquisition. Although the heritage learners’ performance on the post-test still does not match the degree of acceptability assigned by the monolingual native speakers to these sentences, the magnitude of the gains appears to be higher in the heritage speakers than in the L2 learners, especially with ungrammatical sentences with inanimate objects. Figure 7.5 plots the magnitude of the changes (positive values for grammatical sentences and negative values for ungrammatical sentences) in the ratings of grammatical and ungrammatical sentences with a-personal after instruction in the L2 learners from the Bowles and Montrul (in press b) study and the heritage speakesr from the Montrul and Bowles (in press) study. But because the L2 learners and the heritage speakers in these two studies did not complete a proficiency test, and the sample sizes in the two studies were also not comparable, we were reluctant to compare these two groups directly in a statistical analysis. Yet another recent study of Spanish heritage language learners and L2 learners conducted by Potowski, Jegerski and Morgan-Short (in press), found gains

246 Incomplete Acquisition in Bilingualism

Figure 7.5  Comparison of the gains made after instruction in the L2 learners tested in Bowles and Montrul (in press) and the heritage speakers tested in Montrul and Bowles (in press b).

as a result of instruction for both groups, but greater gains for the L2 learners than the heritage language learners on all tasks, contrary to prediction 3. The study tested the efficacy of two types of instruction (processing instruction vs. traditional instruction) in production, comprehension and grammaticality judgments of the imperfect subjunctive with indefinite referents. Potowski et al. (in press) suggest that the advantage of the L2 learners in this study may be related to the fact that they have received many more years of language instruction with focus on forms than the heritage language learners. Another possibility is that the imperfect subjunctive is a more common in the written, standard variety, than in spoken Spanish. It is very common for native speakers to use the present subjunctive even when the verb of the main clause is in the past: e.g., Dijo que venga (present subjunctive) vs. dijo que viniera (imperfect subjunctive) “He said that I come.” Thus, limited literacy with an aspect of later language development may be the reason why heritage speakers are not as reactive to this form as they appear to be to the cases of differential object marking tested by Montrul and Bowles (in press b), which is acquired in early infancy by monolingual and bilingual children. Certainly, how heritage speakers and L2 learners react to different grammatical constructions under instruction needs to be investigated. In conclusion, heritage speakers appear to react positively to instruction, but to address whether they are better and faster than L2 learners at learning or



Chapter 7.  Incomplete L1 and L2 acquisition in adults 247

r­ eactivating missing or not fully developed grammatical knowledge (Prediction 3 in Table 7.5), we need to conduct further studies. Considerably more research is needed on the effects of instructional intervention in these two groups, focusing on learners of different proficiency levels, and on different structures that are acquired at different ages in childhood.

5.

Do heritage language learners have advantages over L2 learners?

This chapter has shown that both L2 learners and early bilingual speakers of minority languages may end up with incomplete knowledge of the target language, especially in many aspects of inflectional morphology and syntax. This may be due to the fact that L2 learners and heritage speakers share some characteristics, like reduced and variable exposure to the target language, and degree of motivation needed to seek opportunities to use the language. However, crucial differences between the two groups, such as timing of exposure (age), and mode of acquisition, make heritage speakers very similar to child L1 learners as well. If these latter characteristics are really significant for language development, I predicted that despite incomplete knowledge of the language, heritage speakers would have an advantage over L2 learners in several grammatical areas (Prediction 1), in different tasks (Prediction 2), and in response to instruction (Prediction 3). The results of several studies showed that while Predictions 1 and 2 were borne out, a clear understanding of incomplete acquisition in early bilinguals remains very complex because at least two other factors appear to interact with degree of linguistic ability in different grammatical areas – proficiency and type of task. Although research on the effects of instruction in these two types of language learners is very scarce at this point, the few studies available suggest that this is a promising area of research. Table 7.16 summarizes the findings to date. Table 7.16  Linguistic advantages of heritage speakers over L2 learners by grammatical module. Heritage speakers phonology morphology syntax lexical-semantics sentence processing (written and spoken)

advantage advantage in less metalinguistic tasks advantage advantage advantage

L2 learners advantage in more metalinguistic tasks

248 Incomplete Acquisition in Bilingualism

In conclusion, at present, there seems to be a weak advantage for age of exposure, which would support the Critical Period Position for L2 acquisition. The decline in the efficiency of the learning mechanisms not only holds for phonology, but for morphology, syntax and lexical semantics as well. The Weaker Language as L1Hypothesis for early bilingualism is also consistent with the advantages found in heritage speakers, at least with aspects of grammar assumed to develop in early childhood, despite incomplete acquisition. However, these advantages are selective and depend on level of proficiency of participants for different grammatical areas, degree of use of the heritage language, and type of task. All these findings need to be pursued with further research utilizing a variety of research methodologies for the assessment of production, comprehension, and processing of spoken and written language. Whether the incomplete grammars of heritage speakers are fundamentally different from the incomplete grammars of L2 learners, or whether the Fundamental Difference Hypothesis applies to L2 learners but not to heritage language learners, is a question that I posed, and have been addressing, in my current research with these two types of language learners, but one that I cannot yet answer with certainty at this stage of my research program.

chapter 8

Implications

Now that I have traced the linguistic development of two languages from childhood to adulthood, it is time to reflect upon what it means for a grammar to be acquired incompletely, on the etiology of incomplete acquisition as a function of age; and how incomplete L1 acquisition fits within our current understanding of L2 acquisition, L1 attrition, bilingualism, and the nature of human language in general. I started the book with a consideration of critical periods in L1 (Chapter 1) and L2 acquisition (Chapter 2). The empirical evidence gathered to date suggests that the critical period has robust effects on the development of L1 acquisition, while the effects on L2 acquisition are not nearly as robust or clear cut. Hence, we can talk about a critical period for L1 acquisition, after which complete and successful acquisition is unlikely, but only of age effects in L2 acquisition, because different degrees of L2 acquisition are possible in adults depending on age of acquisition. My main claim in this book is that a critical period is also relevant for L1 loss in a dual language environment. After a certain age in middle childhood, vulnerability to L1 loss is unlikely in immigrant children fully immersed in an L2 environment. The extent and degree of L1 loss is related to age effects within the critical period and to the availability of input. I supported this claim with two main bodies of empirical evidence: first, younger bilingual children are more vulnerable to language loss than older bilingual children; and second, internationally adopted children with no access to input experience faster and more dramatic effects of attrition than simultaneous bilingual children. Simultaneous bilingual children, in turn, tend to show more attrition or incomplete acquisition than sequential bilingual children exposed to their L1 predominantly until they start school. The second central claim of this book is that attrition in childhood ends up as incomplete L1 acquisition, since many adult early bilinguals who are heritage speakers rarely reach the level of grammatical competence of age-matched and educationally matched fully competent native speakers in their L1, especially with grammatical properties typical of later language development. Yet, although these incomplete L1 grammars may superficially resemble incompletely acquired adult L2 grammars, they still retain the signatures of L1 grammars because they were acquired within the critical period. I captured this idea in the Weaker Language as L1 Hypothesis (see Chapter 4, §5.2), in opposition to Schlyter’s (1993)

250 Incomplete Acquisition in Bilingualism

Weaker Language as L2 Hypothesis, which assumes the deficit view of L2 acquisition. Unlike Schlyter, I maintain that incompletely acquired L1 grammars diverge from the target because they are simplified, but are nonetheless constrained by Universal Grammar, just like the temporarily incomplete L1 grammars of monolingual children. In this concluding chapter, therefore, I first elaborate on the concepts of complete and incomplete acquisition introduced in the opening chapter in light of the research findings presented in this book to arrive at a linguistic characterization of incomplete acquisition in L1 and L2 as a function of age (child vs. adult). I then turn to a consideration of the role and interaction of age and input, and how they speak to the perennial debate on the etiology of age effects in L2 acquisition and the argument that there are also age effects in L1 loss.

1.

Complete vs. incomplete acquisition

As I am sure most readers would agree, the precise definitions of a native speaker and of native speaker competence remain elusive to this day (Davies 2003). Nonetheless, virtually everyone intuitively recognizes a native speaker when she sees or hears her. To begin with, a prototypical (educated) native speaker has a “native” pronunciation, and a sizable and comprehensive vocabulary. He or she speaks in grammatical sentences (except for the occasional slip of the tongue), does not omit or misplace morphemes, recognizes ambiguity and/or multiple interpretations and pragmatic implications of words and sentences, and is attuned to his or her sociolinguistic environment (social class, social context, gender, register, etc.). When judging speakers, we all have an intuitive idea of a norm. At the same time, we know there is predictable variation among native speakers. Not only can we recognize immediately whether somebody is, or sounds like, a native speaker as opposed to a non-native speaker, but we can also tell a great deal by the way a native speaker speaks, how he or she pronounces certain sounds, and the words and syntactic structures he or she uses. Based on these “unspoken” parameters, we can also readily identify native speakers as members of specific and distinct social groups, social classes and races, and geographical regions. Thus, on the one hand, we have a sense that linguistic competence is stable, categorical, and deterministic, fitting some idealized norm, as readily expressed in the Chomskyan linguistic tradition; while on the other hand, we also know there is predictive variation among native speakers in their use of language, as studied in the Labovian sociolinguistic tradition, for example. It is precisely this notion of categorical linguistic knowledge, together with some degree of predictable variability, which characterizes the complete and successful outcome of the acquisition of a first (or



Chapter 8.  Implications 251

native) language in a (predominantly) monolingual environment. In this context, how should we then characterize the notion of incomplete acquisition?

1.1

First language acquisition

As stated in Chapters 1 and 2, incomplete acquisition is a feature of both monolingual and bilingual grammars. In monolingual acquisition, incomplete acquisition is a temporary stage (or a series of stages) in the L1 acquisition process. A developing grammar is, by definition, incomplete. When two-year-old children produce CV syllables as a substitute for CCV or CVC syllables in languages that allow such syllable structures (e.g., [kimi] for /krimi/ “creamy” in English, or [fo] for [flor] flor “flower” in Spanish), the child at that same stage is deemed to possess an incomplete grammar that does not allow complex onsets or codas. When two-year-old children do not produce tense and agreement morphology in English, for example, their grammars are characterized as incomplete in their instantiation of finiteness (agreement, tense, aspect) and verb placement. And when four-year-old children are claimed not to produce or interpret different types of passive sentences in English (Horgan 1978; Sudhalter & Braine 1985), their grammars are also assumed to be incomplete with respect to the grammars of older children or the adult English grammar that includes passives with agentive byphrases and different types of verbs and arguments. Incompleteness in these cases is inferred on the basis of lack of use or high error rates in omission, substitution, and interpretation of the target forms, as compared to older children and adults. Even when we measure and quantify errors as evidence of non-target linguistic behavior, an important question in language acquisition research is to decide the criterion for acquisition, or what error percentage counts as evidence of a complete or incomplete grammar. Typically, a certain percentage is established as a cut-off, above which a child is said to have acquired the structure under investigation. Is 90% accuracy on a given structure or suppliance of correct inflectional morphology in obligatory contexts, as considered by Brown (1973), the best criterion for complete acquisition? Or is it 75%, 60%, or any other number above chance performance (50%)? Yet, in the context of longitudinal data, Stromswold (1996) and Snyder (2007) consider that children should be credited with knowledge of a given grammatical property at the point of first clear use followed soon after by regular productive use (Snyder 2007, p. 71). While researchers recognize that estimates in percentages are highly arbitrary, it is still possible to imagine that 50%–60% accuracy on a given form required categorically in a specific context reflects unstable knowledge of that form. That is, the form has emerged in the grammar of the child but that knowledge is not solid or stable yet, thus the optionality.

252 Incomplete Acquisition in Bilingualism

By contrast, an accuracy rate of 90% and above may indicate complete mastery with occasional performance errors, which are also typical in adults performing a task under experimental conditions. But error rates are not the only measure of an incomplete grammatical system. The specific structures used, as well as those not used by a speaker when the context calls for them, can also indicate the current state of grammatical development. Once again, it is an empirical question to determine how often a child must use a particular structure to consider that she has acquired it. There are cases in which mastery measures cannot be used to determine whether a construction has been acquired because their use is optional (subject to stylistic, contextual or pragmatic rules) rather than obligatory, as in the case of passive sentences (The bear pushed the lion vs. The lion was pushed by the bear), or double object constructions in English (Mary sent a package to her mother vs. Mary sent her mother a package). That is, the same events can be expressed in the active voice or with a dative prepositional phrase, respectively. Furthermore, reduced vocabularies and a more restricted and simple structural repertoire may also indicate that a grammatical system has not yet reached its full potential. Spanish, for example, allows different positions for subjects and objects. Sometimes two-yearold children may use SVO sentences exclusively. Even if these are produced with no errors, children do not vary the position of the verb’s arguments according to verb type or pragmatic context. Since children at this age do not exercise all the grammatical options available in the mature grammar, they may also be said to possess an incomplete knowledge of word order. Similarly, even if overt and null subjects are grammatical in Spanish, a child who does not use overt subjects initially (­Grinstead 2004), or who tends to overuse overt subjects, is considered less able to control the verbal and pronominal system than a native speaker with complete knowledge of the language. In principle, production data is problematic in cases such as these because absence of evidence for a particular form or construction in the language of a child is not reliable evidence for the absence of such form and construction in the child’s mentally represented grammar. However, an incomplete state of knowledge in these cases can only be assessed longitudinally by comparing the child’s speech at two different times, or by conducting specific experimental tasks demonstrating that a given form or structure is also absent from comprehension or judgment. Both in syntactic and morphological development, evidence of productivity is another factor that must be taken into account when evaluating incomplete or complete linguistic knowledge by children. For example, children may know . On the issue of whether frequency of use is isomorphic with knowledge of different word order patterns in children see Meisel (2007).



Chapter 8.  Implications 253

some wh-expressions and questions they have learned as chunks (What’s that?), but are unable to use the wh- word what in other newly generated sentences (What color is the truck?). In morphology, children initially produce irregular past tense forms, such as went, which researchers assume are learned by rote. But it is only when regular forms emerge (talk-talked) and are then overgeneralized to irregular forms (goed), that a child is considered to have productive (albeit erroneous) control of the past tense rule. Thus, while overgeneralization errors show knowledge of a productive rule, they also show lack of knowledge of, or difficulty retrieving, the specific lexical entries for irregular forms. Another primordial characteristic of children’s linguistic development is that errors are constrained and correlate with, or depend on, other structural developments. Consider the well-studied phenomenon of root infinitives in child language (Radford 1990; Rizzi 1993/1994; Wexler 1994). In many languages, most notably English, Swedish and German, children go through a stage when they produce optional tense and agreement morphology in finite clauses. That is, verbs appear in the infinitive or bare stem form (e.g., *Mommy come to the park). However, even when children show a high number of tense and agreement omission errors (close to 60%, according to Wexler 1998), they appear to know the distinction between finite and nonfinite clauses, as judged by the position of verbs in the sentence. That is, when inflected forms are produced in English, they are always in raised positions (such as copula, auxiliaries and modals), whereas non-finite forms always occur in non-raised positions. This is even more striking in a language like French or German, in which inflected and uninflected forms appear in different positions in the clause structure in both the child and adult renditions (Child: non-finite form after negative marker pas as in pas manger la poupée; Adult: finite form before negative marker pas, as in La poupée ne mange pas “the doll does not eat”). As noted by Meisel (1997), finite forms may fail to move on occasion, giving rise to optional root infinitives, but non-finite forms never move. In addition to verb-raising, optional root infinitives also correlate with overt subjects. In general, it appears that in Germanic languages, root infinitives appear in much higher proportions in sentences that also have omitted subjects, especially when modals and auxiliaries are present (Hyams 1996; Phillips 1995; Pierce 1992; Poeppel & Wexler 1993). Finally, root infinitives hardly ever occur in questions or embedded clauses (sentences containing a CP) (Rizzi 1993/1994). In sum, certain grammatical errors in child language are systematic and correlate with the development (or lack thereof) of other abstract aspects of the grammar. The distinction between incomplete versus divergent outcomes in second language acquisition originally proposed by Sorace (1993), and discussed in Chapter 2, may be appropriately extended to describe some temporary stages of first language acquisition as well. If errors of omission, substitution, or misplacement of

254 Incomplete Acquisition in Bilingualism

some obligatory grammatical element, together with restricted use of some structures, are indicative of incomplete knowledge in developing child grammars, some other errors children make may be taken as examples of a divergent grammar. It has been noted that sometimes children adopt parameter values that are not target like in the adult system but are attested in other human languages. The first example of this sort discussed in the acquisition literature, I believe, was Hyams’s (1986) early account of null subjects in child English, although further research by Hyams herself and others later concluded that this view of the null subject phenomenon in child language was inadequate as it did not capture all the facts (cf. Yang 2002). Since English is a non-pro-drop language, Hyams accounted for the high proportion of null subjects produced by two-year-old children by arguing they had adopted the grammar of Spanish or Italian, which are pro-drop languages. This was taken as evidence of parameter mis-setting, because English is non pro-drop. Another widely discussed example of this sort comes from Englishacquiring children when asked to produce long-distance wh-questions in elicited production tasks. The sentence in (1) is ungrammatical in English because children produce a wh-word instead of a trace. McDaniel, Chiu, and Maxfield (1995) argue that the following type of elicited question resembles a possible adult question-type found in certain German dialects and in Romany, as shown in (2): (1) *What do you think what’s in that box? (2) Wer glaubst du wer nach Hause geht? who think you who to home goes ‘Who do you think is going home?’

Hence, these examples show that incomplete and divergent acquisition in child language, as assessed from patterns of errors and structures used in spontaneous production and experimental settings, is a temporary stage in normally developing monolingual children. These errors eventually go away. Most people would agree that L1 grammars may be temporarily incomplete and divergent, but are always UG-constrained. An interesting idea of these developing states is Yang’s (2002) variationist model. Yang claims that in L1 acquisition there are competing grammars, just as in bilingualism. As the target grammar gains prominence (through processing of input), the child entertains co-existing, competing grammars, and this is reflected in the non-uniformity and inconsistency of the language they produce (e.g., optional root infinitives, null subjects and overt subjects co-existing in English child language, etc.). That is, English-acquiring children entertain a Chinese-type gram. I thank William Snyder (University of Connecticut) for these examples.



Chapter 8.  Implications 255

mar and an English-type grammar for a while until the input evidence leads them to select the English-type grammar for the expression of subjects. If cues or signatures for parameter setting are rare (less frequent) in the input, the development of a particular aspect of the grammar (or a parameter) will be relatively late. Yang explains that null/overt subjects are acquired relatively late in child English because the input signature for acquisition is expletives in existential there constructions, and these occur only 1.2% of the time in child-directed speech. By contrast, any aspect of grammar with at least 7% of occurrence in the input should be acquired quite early, such as verb movement with respect to adverbs and negation in French. In the end, children select alternatives from within the range of possibilities allowed by Universal Grammar, on the basis of input. So, during the period of development, L1-grammars are always UG-constrained, even if non-targetlike. Yang further states that “a speaker’s linguistic knowledge is stable (emphasis mine) after the period of language acquisition” (Yang 2002, p. 133), but does not specifically state when that period of language acquisition is actually over. While most acquisition literature may place basic grammatical acquisition by age 3–4, I discussed in Chapter 5 how this estimation is an overstatement. Children’s language development continues throughout the school years, or else we would not observe the high levels of language loss in bilingual children who continue to be exposed to their first language during their life.

1.2

Second language acquisition

That post puberty L2 learners rarely behave like native speakers in spontaneous production and in a variety of experimental tasks, has been interpreted time and again as meaning that their grammars diverge in fundamental ways from that of native speakers. As discussed in Chapters 2 and 5, incomplete acquisition is also the quintessential identifying feature of child and adult L2 acquisition. In the first place, just like in L1 acquisition, incomplete acquisition in the L2 is part of the developmental process: despite differences in approximate rates of acquisition (see discussion in Chapter 5), both children and adults take some time to reach a stable (and advanced) level of linguistic attainment. Before reaching that steady state, children and adults make many of the same omission, substitution, and misplacement errors in phonology, morphology, syntax, and semantics attested in L1 acquisition, all of which are a normal part of the developmental process. Like L1 learners, L2 learners also fail to use, or (unconsciously) avoid, structures they find difficult or that have not become internalized in their interlanguages. Schachter (1974) found that Chinese and Japanese speakers appeared more accurate on English relative clauses than Arabic and Persian speakers, when

256 Incomplete Acquisition in Bilingualism

in fact Chinese and Japanese speakers produced far fewer relative clauses than the other two groups, thus giving the illusion they were more accurate when in fact they were not. Similar avoidance of structures by L2 learners is reported by Oshita’s (2001) study of unaccusatives and subject position in English interlanguage grammars. And, L1 and L2 learners of different languages have been shown to exhibit morphological variability with the production of tense and agreement morphology, the so called optional root infinitive phenomenon described above (Haznedar & Schwartz 1997; Lardiere 1998a, b, 2006; Prévost 1997; Prévost & White 2000). But there are also important differences between incomplete acquisition in adult L2 learners and temporary incomplete L1 acquisition in child learners with respect to the patterns of errors found. Prévost (1997) and Prévost and White (1999) examined production of root infinitives in L2 learners of German and French and observed optionally inflected verbs, like those attested in L1 acquisition. However, with respect to the nature and distribution of errors, there were some important differences. The behavior of the adults differed from the behavior of L1-acquiring children. Recall that in L1 acquisition, non-finite forms occur in specific structural environments. The adult data, on the other hand, showed finite and non-finite forms occurring in the same contexts (substitutions), such as in questions and embedded clauses, with subject clitics, preceding negative markers, etc. An explanation for such distinct behavior is that while in child L1 and L2 acquisition optional verbal morphology is intrinsically linked to the development (or maturation) of other parts of the grammar (probably Tense), morphological variability in adult L2 learners is better characterized as instances of missing inflection in production. There is no impairment at the level of the abstract representation for Tense in adults, but there is some difficulty in the spell out of morphology (or phonology) (White 2003a). Unlike L1 learners, L2 learners make transfer errors during their L2 development due to influence from their native language. The groundbreaking work of Donna Lardiere on fossilization discussed in Chapter 2 showed that Patty, the Chinese (Mandarin and Hokkien) speaker of English as a second language followed longitudinally for almost 20 years, displayed a mere 35% incidence of tense morphology and less than 17% of 3rd person singular agreement in oral production (and a higher, but still non-target rate, in written production). Meanwhile, Patty shows 100% correct incidence of nominative case assignment and complete knowledge of the fact that lexical verbs do not raise in English. Clearly, Patty’s production of verbal morphology is far from native-like and has stabilized (fossilized) at a very . The study also included child L2 learners, and their pattern of errors was similar to that of the L1 learners in that they were also structurally constrained.



Chapter 8.  Implications 257

low accuracy level. While several theoretical accounts have been proposed for the specific etiology of fossilization in this case study, persistent L1 transfer, either at the syntactic level (since on many accounts Chinese is assumed not to have tense) or at the phonological level (since Chinese imposes different licensing conditions on prosodic structure than English), seems a plausible explanation. In a similar vein, Franceschina (2005) looked at knowledge of gender agreement, another volatile area for successful acquisition discussed throughout this book, in advanced to near-native Spanish L2 learners. Most of the speakers tested by Franceschina on a variety of oral and written tasks spoke languages that mark gender grammatically like Spanish (the [+ gender] group), whereas the other subjects spoke a [–gender] language (mostly English). Results showed that the [+gender] group did not differ significantly from a control group of native speakers in their accuracy rates, producing correct gender agreement between 95–100% depending on task. The [–gender] group was also fairly accurate, producing overall error rates in the range of 81–99% depending on task. However, according to Franceschina the [–gender] group behaved systematically different from the other two groups. Similarly, different error rates, based on the availability of gender in the L1 of the learners tested, were found by Sabourin, Stowe and de Haan (2006) in German, English and French-speaking L2 learners acquiring Dutch (see Chapter 2). In sum, L1-induced fossilization has been identified as a hallmark of adult L2 acquisition, and is believed not to occur if L2 acquisition started in early childhood. However, recent research on child L2 acquisition of Dutch gender reviewed in Chapter 5, indicates that fossilization is possible in child L2 acquisition as well. As in L1 acquisition, one has to question what percentage accuracy (20%? 60%?) should be taken as clear signs of complete and incomplete and fossilized knowledge in adult L2 acquisition. Answers to this question may vary, since cut-off points of this sort are arbitrary, as most researchers acknowledge. It is interesting to note that while Lardiere (1998 a, b, 2007) takes a robust 65% error rate in past tense omission as a clear sign of fossilization in Patty, White (2003b) takes error rates between 15% and 30% in the Turkish subject of her study, while Franceschina considers a 1%–19% error rate with gender accuracy in advanced L2 learners as a sign of fossilization and divergent linguistic representations. If these error rates are not attested in adult native speakers (or native speakers perform at above 95% accuracy) but are attested in L2 speakers assumed to be at endstate and who have had ample opportunity to use and acquire the language, then any systematic error rate above 5% may be considered an instance of an incomplete grammar. The key difference between L1 acquiring children and L2 adults is that in child language, errors eventually disappear, while in adult L2 speakers they persist. One of the purposes of this book was to reconsider fossilization and incomplete acquisition as uniquely defining characteristics of adult acquisition, a

258 Incomplete Acquisition in Bilingualism

­ attern which “deficit” views of second language acquisition purport to explain. p Since very little research on the linguistic development of children learning an L2 exists to date, I believe it is premature to conclude that fossilization hardly ever occurs in child L2 acquisition. In fact, the findings of Unsworth (2007, 2008) in morphosyntax and of Flege, Birdsong, Bialystok, Sung and Tsukada (2006) in pronunciation suggest that this notion needs revision as well. Flege et al. (2006) found that after more than four years of living immersed in the L2 environment, Korean-speaking L2 learners of English still had detectable foreign accents when compared to age-matched English-speaking children. (Of course, this study is not about endstate in children, but of an extended period of development during childhood). Clearly, this is an area in need of further empirical development and theoretical refinement.

1.3

Early bilingual acquisition

Contrary to the commonly held idea that permanent, rather than temporary, incomplete acquisition and fossilization are restricted to adult L2 acquisition, discussion of the empirical evidence presented in Chapters 4–7 demonstrated that permanent incomplete acquisition and fossilization are also typical of L1 acquisition in a dual language context. This is particularly the case in some, but by no means all, bilingual situations as a result of second language acquisition. When bilingual children who speak ethnic-minority languages are exposed to the majority language, use and ability in the command of their first language begins to decrease. Different aspects of their grammars may, in turn, cease to develop at different non-native levels of proficiency. These non-target effects, many of which may be caused by L2 influence on the L1, persist in their weaker language well into adulthood. Results of several studies discussed in Chapter 4 with simultaneous or very early bilingual children reported difficulty with words and many errors in verbal and nominal morphology (Kaufman & Aronoff 1991; Turian & Altenberg 1991). Schlyter’s (1993) study of the weaker language in French and Swedish identified between 65% and 95% omission of inflectional morphology in verbs (tense and agreement). Anderson’s (1999) study of two Spanish-speaking siblings reported between 6% and18% errors in gender agreement and an average of 40% error rate with agreement and mood in verbs. Paradis and Navarro (2003) discuss a much higher rate of overt subjects in a Spanish-English bilingual child, and incomplete knowledge of the pragmatic distribution of subjects. Older bilingual children do not seem to overcome many of these morphosyntactic errors, as discussed in Chapter 5. The few available studies to date show



Chapter 8.  Implications 259

that Spanish-speaking bilingual children schooled in English continue to exhibit more than 30% error rates with verbal morphology (Merino 1983) and systematic errors with gender agreement (Mueller-Gathercole 2002a, b, c). Work by William O’Grady and collaborators showed that Korean children attending communityschools also lag behind monolingual peers in their production and comprehension of case markers, relative clauses, and binding properties of anaphors. Finally, Chapters 6 and 7 showed that incomplete knowledge of the family language in early bilinguals persists into adulthood. Adult heritage speakers cease development of their L1 at different levels of proficiency, and continue to produce high error rates in nominal and verbal morphology (Montrul 2002; Montrul, Foote and Perpiñán, in press; Polinsky, 2007, 2008b; Silva Corvalán 1994, 2000), and errors with subject expression in languages like Spanish and Russian, and in several other areas of syntactic knowledge, as with relative clauses and binding in Korean (Song, O’Grady, Cho & Lee 1997; H.-J. Kim, Montrul & Yoon, in press). Because incompletely acquired L1s are still acquired within the critical period, it follows that they have been acquired through the same cognitive and linguistic mechanisms utilized by normally developing children acquiring their L1. For some researchers this means through Universal Grammar. The only differences between complete and incomplete L1 acquisition in normally-developing children are degree and availability of input and use of the language. If a certain percentage of unambiguous signatures are needed to acquire or set a parameter (Yang 2002), many bilingual children may not reach the minimum threshold. Thus, the weaker language is like a first language, only with non-target features and simplified. An interesting question, which I started to explore in Chapter 7, is how incomplete L1 and L2 acquisition do or do not differ from each other. Is there evidence that incomplete L1 acquisition is UG-constrained while incomplete L2 acquisition acquired after the critical period is not, as Schlyter (1993) contends? Research on incomplete L1 acquisition in bilinguals is relatively new and scarce in comparison with the existing body of research in L1 acquisition and L2 acquisition. At this point, it is very difficult to evaluate more precisely whether the types of errors found in adult early bilinguals are more similar to what is found in L1 acquisition or in L2 acquisition. Judging from the available evidence, it appears that in the realm of morphology and morphophonology, omission and overegularization errors are more common than substitution errors, which is the overall trend found in developing L1 grammars. Naturally, this preliminary observation needs to be substantiated through further empirical research. At the same time, and in common with L2 acquisition, some patterns of L1 simplification (maybe even “fossilization”) take the apparent form of influence from the dominant language, which is English in most of the cases discussed in

260 Incomplete Acquisition in Bilingualism

this book, such as the adoption of an English-type governing category for the licensing of reflexives in Korean (J.-H. Kim, Montrul & Yoon, in press); difficulty with stative verbs in the Spanish preterite (Montrul 2002); the omission of differential object marking in Spanish (Montrul 2004a; Montrul & Bowles, in press a, b); omission of case markers and reliance on canonical SVO order in Korean (Song, O’Grady, Cho & Lee 1997), and even the overuse of overt subjects in null subject languages (Paradis & Navarro, 2003; Serratrice, Sorace & Paoli 2004; Montrul 2004a). To summarize thus far, I have shown in this book that permanent incomplete acquisition is not unique to adult L2 acquisition. It is also common, and likely, in many cases of early bilingualism, including incomplete L1 acquisition in a minority language situation. Table 8.1 summarizes the key linguistic characteristics of L2 acquisition and L1 loss by children and adults discussed in this book. The first column describes general patterns of acquisition and loss in adults. What is important to notice about adults is that if L1 loss occurs under attrition, it does not really seem to affect linguistic competence to the extent that incomplete L2 acquisition can. In other words, attrition effects more readily reflect minor difficulties with lexical access and processing. In the realm of syntax, phonetics and other linguistic interfaces, L1 attrition in these cases takes the form of L2-induced changes in the L1. True L1 loss that is not L2-transfer could only be assessed if the individual were completely cut-off from the L1 and had no access to any other language, which is unlikely. As for the column on children, we see that we do not know enough about L2 acquisition by children, and what we do know so far may have to be revised in light of recent research showing age effects in late childhood, and that some younger children may follow the route of L1 acquisition while older children may be more like adult L2 learners. If there are individual differences in L2 acquisition by children at specific ages, this has yet to be determined by more empirical research. But these distinctive patterns may also occur in different grammatical domains or modules (syntax, morphology, phonology). Whether L2 children also stabilize at non-native levels of development is yet another question open to further investigation. The bulk of this book focused on tracing incomplete acquisition of one of the languages of bilingual children from childhood into adulthood. One aim was to show that many characteristics attributed to adult L2 acquisition are also true to some extent of incomplete L1 acquisition (L1 loss) in some bilingual situations. As adults, incomplete L1 learners show many of the developmental and ­fossilization



Chapter 8.  Implications 261

Table 8.1  Some linguistic characteristics of L2 acquisition and L1 loss by children and adults. Adult (late bilingualism)

Child (early bilingualism)

L2 Acquisition • • • •

• Developmental errors • L1 influence • Morphological variability (some), but eventually overcome (open to investigation) • Less incidence of fossilization (open to investigation) • More likely to achieve native-like accent

L1 Loss

Incomplete acquisition • Developmental errors and regularization patterns • May exhibit a more simplified system than target grammar (e.g., case marking and gender in Russian, binding in Korean) • L2 induced changes in L1 • Depending on age and input, both interpretable and non-interpretable features may be affected • Depending on age and input, core aspects of syntax may be retained (e.g., V2 in Swedish, pro-drop, word order, clitics in Spanish) • Structures with a later developmental schedule in L1 acquisition and non-core aspects of grammar may not be acquired • Foreign accent (although better than that of L2 adults)

Developmental errors L1 influence Morphological variability Fossilization affects both interpretable (semantic) and non-interpretable features (formal) • Various degrees of foreign accent Attrition • L2-induced changes in the L1 • If affected, only interpretable features (interfaces) • uninterpretable features (core grammar) remain intact • Lexical retrieval delays • Minor changes in L1 accent

patterns observed in adult L2 acquisition. Important systematic differences between incomplete L2 acquisition by adults and incomplete L1 acquisition by children remain to be explored in further research. Although incomplete outcomes in these two situations cannot be due to a late age of L2 acquisition, as it has been claimed time and again for adult L2 learners, my claim in this book has been that age is the macro factor that links these two learning situations. Age, and possibly a critical period, is very relevant to L1 attrition in language minority speakers.

262 Incomplete Acquisition in Bilingualism

2.

The role of age in language acquisition and language loss

With respect to theoretical models of second language acquisition, this book has emphasized the deficit view of adult L2 acquisition. I chose to focus on this position rather than on a Full Access view of L2 acquisition because the deficit view specifically identifies age as the main reason behind the frequently observed phenomena of fossilization and permanent incomplete acquisition so common in this context. For all these researchers and many others (Long 2000, 2007; ­Coppieters 1987; DeKeyser 2000; Johnsohn & Newport 1989), different endpoints for L1 and L2 acquisition are due to maturationally related age effects, as stated in the Critical Period Hypothesis. It has been my contention throughout this book, however, that the deficit view of L2 acquisition does not adequately represent or explain the nature and outcome of second language acquisition. Not only have I shown that permanent incomplete acquisition is hardly unique to, and characteristic of, adult L2 acquisition, but I have also shown that early exposure to a language – while certainly necessary – is not a sufficient condition for complete L1 (and probably L2) acquisition either. In other words, once a native language is acquired in childhood, it must be stimulated by rich input and must be used in order to be kept intact. Similar observations have been made by Hyltenstam (1992). To be clear, I am not in disagreement with the notion that permanent incomplete acquisition and fossilization are common in L2 acquisition; what I hope to have uncovered in this book is that incomplete acquisition and age effects are not restricted only to adult L2 acquisition as has long been argued in the field. Age effects are more pervasive in bilingualism in general since they also correlate with L1 acquisition and loss in some early bilingual situations, a population that has not received much attention in the critical period literature. Theories of second language acquisition that rest on the observation that these features are only true of adults are in need of revision. Once we begin to consider the other side of the L2 acquisition coin – L1 loss – as I have done in this book, our conception of the nature of L2 acquisition is likely to change as well. It has become a truism by now that children seem endowed with a gift for learning one or more languages in early childhood. But biology does not work in a vacuum. When children do not receive linguistic stimulation early in life, . The deficit view includes Bley-Vroman’s (1990) Fundamental Difference Hypothesis; Schachter’s (1990) Incompleteness Hypothesis; Tsimpli & Roussou (1991) and Meisel’s (1997) No parameter resetting hypothesis and Hawkins and Chan’s (1997) and Hawkins and Hattori’s (2006) Failed Formal Features Hypothesis; and Tsimpli and Dimitrakopoulou’s (2007) Interpretabilty Hypothesis, among many others.



Chapter 8.  Implications 263

the likelihood of developing full linguistic ability in the ambient language is seriously compromised. A variety of indirect evidence from neurology, extrasocietal children, creole genesis, and deaf populations surveyed in Chapter 1 suggests that a maturationally determined critical period for L1 language acquisition may be justified. If a child receives linguistic input through socialization before the ages of 6–7, chances of developing a full linguistic system are remarkably good, as exemplified by the case of Isabelle (Brown 1958; Mason 1942) and young ASL signers (Mayberry 1993; Mayberry & Eichen 1991; Newport 1990). If, on the other hand, exposure to natural language becomes available after age 6–7 and before age 13, some rudiments of language can be acquired, although acquisition appears incomplete, as exemplified by Genie (Curtiss 1977) and late learners of ASL. Finally, the case study of Chelsea (Curtiss 1989) clearly illustrates how, when initial exposure to input starts well after puberty, a first language has no chance of development. These collective findings are often taken as strong evidence for maturational effects in L1 acquisition, with some irreversible consequences if language is not heard before puberty. Despite divergent opinions to date as to whether a critical period exists in adult L2 acquisition as well, the undeniable fact is that given ample input and linguistic opportunities to use the language, late (post puberty) onset of acquisition is a high predictor of fossilization, even if the effects are not as dramatic and irreversible as in L1 acquisition. Early onset (childhood) predicts complete L2 acquisition. The typical characterization of the age effect in L2 acquisition is a negative correlation and linear function, as schematized in Figure 8.1. Figure 8.1 also includes the populations of L2 learners studied in this book. Although the graph is a mere approximation, it shows that of all L2-acquiring populations examined in this book, young adoptees (with no L1 input) attain native-like command of the L2 as adults, and simultaneous bilinguals who receive full academic support in the L2 are also very likely to become L2 native speakers. The chances diminish gradually as bilingual children get older, and the L2 is learned after the structure of the L1 is fully in place (sequential bilingualism or child L2 acquisition). Finally, adults are less likely to attain full grammatical competence (proficiency) in the L2. If age is crucial for acquisition, it seems to play a significant role in language loss as well. At the end of Chapter 2, I spelled out specific predictions the critical period hypothesis makes for language loss as a result of L2 acquisition in early bilingual acquisition. Under a version of the critical period that considers the type of knowledge and the cortical representations subserving early L1 and late L2 acquisition are very different, exposure to a language early in childhood should have long-lasting traces in the neural circuits subserving language processing, representation, and use (Long 2000; DeKeyser 2000; M. Paradis 2004; Ullman 2001;

264 Incomplete Acquisition in Bilingualism

Figure 8.1  Typical shape of a negative correlation between age and proficiency for L2 acquisition.

Pallier et al. 2003; Hyltenstam & Abrahamsson 2003). Thus, once completely acquired and fixed in the brain, knowledge of the L1 should be very difficult, if not impossible, to lose in the absence of pathology under less than optimal access to input, as it happens in a dual language context. In Chapter 3, I presented findings from L1 attrition studies in adults who have immigrated to the L2 environment in adulthood, had more than 10 years of extensive exposure to the L2, and hardly ever used the L1. I concluded that language erosion in a variety of grammatical areas is very unlikely in adulthood, at the level of linguistic competence (see also Table 8.1). At the morphosyntactic level – a highly vulnerable area for both temporary and permanent incomplete acquisition in L1 and L2 – adults under

. However, recent articles reviewing past research in this area come to very different conclusions about the neural substrate for L1 and L2 acquisition (Abutalebi, Cappa & Perani 2001, 2005; Abutalebi & Green 2007). For example, Abutalebi et al. (2005, p. 497) conclude that attained L2 proficiency, and perhaps L2 exposure, are more important than age of L2 acquisition as a determinant of the cerebral representation of language in bilinguals: very proficient L2 speakers appear to engage the same neural network subserving the L1 within the dedicated language areas. And in another recent review, Abutalebi and Green (2007, p. 242) conclude: “Neuroimaging data, we argue, support the notion that the neural representation of a second language converges with the representation of that language learned as a first language.”



Chapter 8.  Implications 265

attrition made very few errors, well below 5%. If anything, findings from these studies demonstrate that some aspects of pronunciation together with lexical retrieval and production and interpretation of pronouns are mildly affected by the L2, but these findings do not show that the mature linguistic competence attained has been destabilized. Similar findings are reported and discussed by de Bot and Clyne (1994), Köpke (1999, 2002, 2004), Jaspaert and Kroon (1992), and Schoenmakers Klein-Gunnewiek (1998). By contrast, if input to the first language(s) is reduced, or cut-off altogether, in early bilingualism, the critical period hypothesis predicts that the degree of language loss should also be a function of age. Not only will children be more susceptible to L1 attrition or incomplete L1 acquisition than adults, but there should be gradient effects during childhood as well. For example, within early bilingualism, simultaneous bilinguals should exhibit more severe loss than sequential bilinguals. Once again, the earlier the onset of bilingualism and the earlier a child starts using the L2 more than the L1, the more dramatic the loss in the L1 is likely to be. (This should not be interpreted that onset of bilingualism immediately marks the onset of attrition, however, since children can, and often do, exhibit balanced bilingualism for some time before attrition as well.) If this view is correct, the empirical evidence should yield a hypothetical age function in the opposite (positive) direction from the one illustrated in Figure 8.1, as suggested by the progression observed in the case studies of Genie, Isabelle and Chelsea: i.e., the earlier the deprivation of the L1, the greater the loss. The shape of the age function for L1 loss, which is steeper than that of L2 acquisition, is illustrated in Figure 8.2. Despite the fact that there is significantly less research on L1 loss in bilingualism than there is on age effects in L2 acquisition, the empirical evidence presented in Chapters 3–7 is highly consistent with this prediction. Several studies discussed in Chapter 4 showed that bilingual children can lose linguistic competence in the L1 quickly and easily, particularly during the age of early language development (birth to age 3–4), and especially when the family language is not supported by a speech community outside the home (Kaufman & Aronoff 1991; Turian & ­Altenberg 1991; Anderson 1999; Silva-Corvalán 2003; Nicoladis & ­Grabois 2002). Immigrant children who arrive at the L2 country with two or more years of schooling in their first language, on the other hand, experience slower and less dramatic effects of L1 loss (Jia & Aaronson 2003; Montrul 2002; Yeni-Komshian, Flege & Liu 2000). L1 loss after age 9 appears to be less likely (Montrul 2002) in normal bilinguals who are not completely cut off from the language. The case for adoptees may be different. The claim that there are age effects in both L2 acquisition and L1 attrition raises two very important questions in language acquisition and bilingualism

266 Incomplete Acquisition in Bilingualism

Figure 8.2  Positive correlation between age and proficiency for L1 loss.

that have not received much attention. First, how do the two languages interact from childhood into adulthood (birth to age 18) in bilingualism (some exceptions are Flege, Yeni-Komshian & Liu 1999; Jia & Aaronsson 1999; Kohnert, Bates & Hernández 1999; Yeni-Komshian, Flege & Liu 2000); and second, at approximately what age does a heightened sensitivity to acquire an L2 with more accuracy, and propensity to lose ability in the L1 more easily, cease to exist? If we cross the linear functions illustrated in Figures 8.1 and 8.2, the hypothetical scenario would be like that shown in Figure 8.3. The intersection in the age axis could indicate the approximate age at which, hypothetically, the linguistic capacity for language, or the L1 in monolingual acquisition, is about to reach, or has reached, its mature and steady state. It may be at this point – approximately – when knowledge of the native language in monolingual speakers takes its permanent form, i.e., “crystallizes” due to maturational reasons. In a bilingual situation, this point would mark the age after which the capacity for acquiring a second language like a native speaker becomes, in general, irreversibly disabled and L1 loss is less likely, and the age before which it is possible to develop native-like knowledge in a second language while vulnerability to L1 loss is still enhanced. . See Birdsong (2005, 2006) for recent summaries of relevant literature.



Chapter 8.  Implications 267

Figure 8.3  Intersection between L2 acquisition and L1 loss functions.

What would that approximate age be? Research suggests that the hypothetical functions in Figure 8.3 must intersect at a time that could be between 8–10 years old (see Jia & Aaranson 2003, Table 5.11 in Chapter 5). Lenneberg’s proposed explanation for the critical period was that of hemispheric lateralization, which he said became complete at the age of puberty. In the context of L2 acquisition, Schachter (1996) mentions results of some dichotic listening test studies conducted by Satz, Bakker, Tenunissen, Goebel, and Van der Vlugt (1975) that found right ear superiority (indicating left-hemisphere dominance for language) becoming significant around age 9 and leveling off at age 11. Although, as Schachter correctly warned us, lateralization as a putative cause for the critical period needs further study, 9 years of age may be a significant age for native speaking ability. Consistent with the importance of this approximate milestone in childhood, Hyltenstam and Abrahamsson (2003, p. 575) make the following claim: At least up to AOs 6 or 7, all (L2) learners will automatically reach levels that allow them to pass as native speakers – provided that there is sufficient input and that the learning circumstances are not deficient . . . 8-year-old starters must have  . In a recent meta-analysis of 66 published studies, Hull and Vaid (2007, p. 1987) concluded that bilinguals who acquired the two languages before age 6 showed bilateral hemispheric involvement for both languages. Those sequential bilinguals who acquired their second language after the age of 6 showed left hemisphere dominance for both languages.

268 Incomplete Acquisition in Bilingualism

a certain, albeit small, degree of extraordinary motivation (and/or positive affect, input, instruction, aptitude, etc.) in order to reach the same levels automatically reached by their 6-year-old friends.

And according to Ullman’s (2001) declarative/procedural model briefly described in Chapters 2 and 3, over-reliance on the declarative system in L2 processing is attributed to maturational changes that occur during childhood and adolescence, such as increasing estrogen levels in both males and females. Although a specific age is not stated, I suspect that 8–10 years of age may be the transition between childhood and adolescence that Ullman has in mind. With respect to L1 attrition, there is much less information to go by. Citing evidence from L2 loss in children, Köpke (2004) and Köpke and Schmid (2003) suggest that vulnerability to attrition closes off at age 9. There is some evidence for this purported age from some studies on bilingual children discussed in this book. The studies conducted by Pallier and collaborators on loss of the L1 in foreign-born adoptees (see Pallier 2007) showed that, surprisingly, up until age 8, total language loss is possible when input is completely cut-off. As for cases of bilinguals who continue to be exposed to their L1, Montrul’s (2002) study of aspectual interpretations in Spanish heritage speakers also found age effects. Montrul’s study tested simultaneous bilinguals (birth-3), child sequential bilinguals (ages 4–7), and late child L2 learners (ages 8–12), and found that effects of incomplete acquisition were more evident in the first two groups (and more severe in the simultaneous than in the early child L2 learners) than in the last group. Indeed, as a group, the late child L2 learners with age of intense of exposure to English after the age of 8 performed on all tasks and conditions like the adult native speakers. Similarly, Jia and Aaranson’s (2003) study of Chinese children who immigrated to the U.S. reported more perceived negative changes in the L1 before, or up to, age 9. The conclusion that emerges from all this is that, after age 9, vulnerability to attrition seems unlikely, at least for bilingual children who continue to be exposed to their L1. Cases of adoptees completely cut-off from input may be different. In conclusion, it appears that a series of biological, physiological, sociocognitive, neuroanatomical and neurofunctionally determined changes occur at, or about, age 9, which may be critical for the solidification or crystallization of native or dominant language skills, in both monolinguals and bilinguals. However, if biology is involved, it certainly does not work in isolation. I turn to an examination of the role of input and context of acquisition next.



3.

Chapter 8.  Implications 269

The role of input in language acquisition and language loss

It is a trivial fact that input plays a fundamental role in language acquisition. Even if one maintains a nativist view of language, an innate capacity for language acquisition cannot do its job without appropriate input. But what exactly is input? How is it measured, quantified and categorized into good, adequate or poor quality? What is abundant input? What is reduced or impoverished input? Researchers working within empiricist and social-interactionist perspectives tend to portray input as something external to the learner that is out there in the environment. Carroll (2001, 2005), however, explains that input is the product of language processing. In order for input to be used by the learner, it must be converted into intake, or the analyzed part of input that feeds the language learning mechanism. In essence, there is both an external-physical and an internal-mental element of input. Physically, aural input is some type of noise, while written input is a bunch of strokes. It is our mind, through a set of cognitive procedures, which perceives and organizes aural and visual stimuli into sounds or letters, morphemes, words, phrases, sentences and discourse. Processing mechanisms parse, analyze, categorize and internalize such data. Raw linguistic data out there in the environment is useless for the language faculty in the absence of mental mechanisms that convert these physical data into intake. Although many researchers often refer to quality and quantity of input in the environment as crucial for developing critical levels of linguistic skill in the L1 or L2, others have stressed the close connection between the external and internal (mental) components of input. In the context of the critical period debate, for example, Hyltenstam and Abrahamsson (2003) argue that maturation and stimulation work together since the “language learning mechanism” would be designed in such a way that it requires “immediate triggering from the environment” in order for it to develop and work appropriately; that is, the learning mechanism inevitably and quickly deteriorates from birth if not continuously stimulated. The cases of the feral children Genie and Chelsea discussed in Chapter 1 may provide indirect evidence for this claim. However, what is not clear is precisely how much input or what type is needed to develop language, since we also saw in Chapter 1 that in creole genesis – when access to input is less than optimal – children’s language learning patterns contribute significantly to creole formation. We have seen in Chapters 4 and 5 that many of the patterns of acquisition and language loss observed in children can be related to fluctuations in the input. Bilingual learners’ access to input and their propensity to learn and/or lose the language(s) also vary according to context and mode of acquisition. At the same time, amount of input

270 Incomplete Acquisition in Bilingualism

before or after a certain age has different effects on the outcomes of acquisition. Child and adult differences in cognitive ability are intrinsically linked to the issue of how input is processed or used for language acquisition. A problem with using input as a key variable to explain much of language acquisition is that it is difficult to operationalize. Some studies consider number of hours, days, months and years of exposure to the language as criteria. Assuming a normally developing monolingual child sleeps 10–12 hours a day, she is exposed to the language most of the remaining waking 12–14 hours, seven days a week for at least 3–4 years to develop basic grammatical competence. Bilingual individuals, by definition, must split their waking hours for linguistic exposure into two or more languages, and the hourly, daily, monthly or annual exposure to each language varies considerably. Bilingual children who attend school in only one language receive at least 6–7 hours of exposure just in that language. But what are the lower bounds of achievement? How much exposure to input is necessary for the presence of a given structure to be recognized and incorporated into the grammar? There must be some minimum threshold of quantity of input that a learner must be exposed to. Furthermore, the learner needs to be at the right developmental stage in order to be sensitive to, notice, and thus acquire a given property. Different grammatical areas may require different thresholds of input to develop categorical knowledge, or to lose categorical knowledge, but what those thresholds are has not been clearly specified in the research on incomplete acquisition discussed in this book. In the historical linguistics literature, Lightfoot (1999) asks when the input (triggering experience) of two generations triggers a different grammar. Lightfoot’s analysis of V2 in Middle English shows this structure was produced in about 30% of matrix clauses with an initial non-subject in sentence initial position, but due to the co-existence of a dialect with V2 and another without V2, the children of the Southwest Midlands were exposed to only 17% of those structures in the input (estimated from the frequency found in a text from the Middle ages). Lightfoot concluded that 17% of initial non-subjects were not enough evidence to trigger the V2 grammar, and as a result many children did not have it. Lightfoot also wonders what happens when input to this particular structure is between 17% and 30%, what he calls transition phase. (A problem with historical data of this sort is that it is based on written texts and not actual language produced by adults and directed to children.) Yang (2002) suggests that heterogeneity in the linguistic input and ambiguous evidence lead learners to attain a mixture of co-existing grammars rather than to converge on one. Even when input is quantified in this manner, amount of input measured in time is not enough. Quality is another criterion. Who are the interlocutors, and how do they speak? Is watching TV the same as talking to people? Is just listening to language the same as speaking it? Do incomplete learners talk to each



Chapter 8.  Implications 271

other in the target language? Few are the studies that actually describe the speech of interlocutors in detail or even analyze parental speech samples. In explaining incomplete acquisition of L2 Dutch by ethnic minority children in the Netherlands, Hulk and Cornips (2006) concluded that the families of the children have different patterns of language use: they speak non-standard varieties and use both the minority and the majority language in the home to different degrees. Since the parents are non-native speakers of Dutch, the minority speaking children are exposed to input replete with non-native features, which is qualitatively different from the more “grammatical” and native-like input that monolingual children are assumed to be exposed to. In the context of explaining why older bilingual children may reach higher levels of L2 attainment than adults, Jia and Aaronson (1999) claim that children are exposed to a “rich environment,” which they define and quantify in terms of watching more hours of TV, having more friends who speak the target language, and reading more books in the language. But Paradis and Navarro (2003) is an example of a study which actually quantified speech samples from the parents of the Spanish-English bilingual child they studied. They found that the parents of this bilingual child used more overt subjects than parents of the monolingual children used for comparison. Therefore, the increased use of overt subjects in Spanish used by the child could be explained by the rates of overt subjects also observed in the parents. All these studies have made efforts to provide a more tangible measure of what quality of input really looks like. In all these studies, quality appears to mean exposure to “native-like” input. Input to the L2 learners may be different. When post-puberty L2 learners start their L2 acquisition process, they do so very often in an instructed environment. Exposure to instances of the target language can range from 2 hours a week to 3 hours a day, more or less. In a classroom setting, L2 learners are exposed to oral input from a native or non-native speaking teacher, and several other learners who either speak the same native language as the learner, or may come from other language backgrounds. The L2 learner is exposed to a variety of discourse: written input from books and textbooks, aural input from videos, songs, etc., as well as non-native input from their classmates or the teacher (if the teacher is a non-native speaker). Studies also indicate that teacher talk can be structurally different (e.g., containing more SVO and overt subjects in Spanish than in oral corpus data) than native speaker talk (Santilli 1996; Sanz 1999). In a second language context, the learner may be exposed to input at the workplace and at school, but again, he or she may be exposed to both native and non-native input, depending on who the learner socializes with. In these situations, the quantity and quality of input alone may explain why adult L2 learners achieve different competence in the L2 than L1- and L2-acquiring children.

272 Incomplete Acquisition in Bilingualism

Immersion situations provide adult L2 learners the rich linguistic environment that children are assumed to have access to. Most L2 learners continue to use their L1 on a daily basis, which remains their stronger language throughout their lifespan. But even when adult L2 learners are completely cut-off from their L1 and fully immersed in the L2, total immersion and interaction with native speakers exclusively does not unequivocally guarantee full linguistic competence in the L2. Failure to achieve native-like linguistic behavior under optimal conditions for learning is documented by Lardiere’s (2007) case study of Patty discussed in Chapter 2. Lardiere tells us that Patty appeared to be “fully integrated” linguistically, as judged by her linguistic behavior and daily interaction with those who lived and worked with her in an English-speaking environment. Patty had graduate and undergraduate degrees from the United States, worked in corporate America, had been married to two native-speakers of English, and her daily interactions had been in English for the last 25 years. Despite this input flood and her daily use of English, Patty’s English is native-like in many respects, yet far from native-like in many others. In essence, input to the L2 learner for certain grammatical form or rule can be rich, abundant, and frequent, or it can be variable and infrequent. In the generative tradition, two learnability scenarios are often discussed. The first is the poverty of stimulus argument, which states that learners are able to acquire a complex system (or extract a certain abstract rule), when evidence from the input is either poor, infrequent, or ambiguous. It is believed that many aspects of syntactic and semantic acquisition fall within this learnability scenario. The other problem common during stages of incomplete acquisition in L1 and most common in the adult L2 acquisition, is the opposite situation: the abundance of stimulus problem. In this case, input is extremely frequent and reliable yet learners do not seem to take notice because they continue to produce the structure and rule in question with a high rate of errors. An example of the abundance of stimulus problem is the case of inflectional morphology, including optional root infinitives in child language, or the acquisition of gender agreement by L2 learners in Spanish. Tense marking and gender agreement are produced in every single sentence uttered. Yet control of this morphology is not always accurate in children and adults. When it comes to the L1 loss situation, we also saw that lack of linguistic exposure to the L1, or exposure to non-native features of the L1, does not affect L1 linguistic competence in the adult. All the German Jews interviewed by Schmid (2002), for example, chose not to use German for many years, but their linguistic competence remained there when tapped in an experimental situation. In sharp contrast, both quality and quantity of L1 input dramatically affect acquisition, maintenance, and degree of loss of an L1 in childhood. The fact that the quantity and quality of input does not matter that much for L2 acquisition and L1 loss in



Chapter 8.  Implications 273

adults while it is so deterministic in children suggests to me that the main difference for different outcomes in children and adults is not in the input out there but in how the external input is attended to, processed and internalized. Both degree of entrenchment or proficiency in the L1 and length of relatively stable competence in it, together with other language processing mechanisms, may explain divergent patterns of ultimate attainment. A substantial body of research in both the syntactic (Schwartz & Sprouse 1996; Hawkins & Chan 1997), morphological (Montrul 2001a, b) and phonological literature (Brown 1998; Flege et al. 2003) suggests that the crystallized L1 system may funnel, inhibit, or distort the way physical input is perceived and eventually analyzed by the adult L2 learner, especially with aspects of grammar that differ categorically in the two languages. The L1 imposes its structural architecture when incorporating and analyzing incoming input from the L2. Johnson and Newport (1989), Long (1990), Carroll (2001), MacWhinney (1992) and Lardiere (2007), among others, all reached the conclusion that either “sensitivity” to input, or the way in which adults analyze and internalize input, may explain their incomplete or divergent patterns of acquisition. For Newport (1988, 1990), maturational changes in the adult learner play a role in how input is perceived and statistically computed to derive categorical generalizations. In her Less is More Hypothesis, Newport (1988) stated that the very limitations of the child’s information processing abilities provide the basis on which successful language acquisition occurs. This hypothesis assumes that children and adults vary in their abilities to perceive, analyze and remember complex stimuli. Specifically, the hypothesis assumes that due to their larger short term memory capacity, adults are able to perceive and remember larger chunks of stimuli, which may make it more difficult later for computation and analysis of its components. By contrast, children’s more limited short term memory capacity allows them to store less information and to locate its components more easily, especially in grammatical areas that require componential analysis, such as morphology. Johnson and Newport (1989) explain morphological errors in adult and child learners by stating that children must perform fewer computations to develop categorical knowledge of a morpheme, while adults, having an exponentially larger job at sorting out the components, learn the morphology inconsistently. Further support for age related differences related to input analysis comes from a more recent study by Hudson Kam and Newport (2005) with an artificial miniature language experiment. The goal of this study was to model the situation of creole genesis, where input to the learners is variable (i.e., containing inconsistent grammatical morphemes), and to see whether L2 learners of different ages modeled the input veridically or regularized it. Results showed that adults modeled the input veridically, matching the frequencies of suppliance and omission of determiners provided in the input, whereas 5–7 year old children picked up a

274 Incomplete Acquisition in Bilingualism

form in the input (either omission or suppliance of determiner) and overgeneralized it. Thus, Hudson, Kam and Newport concluded that children and adults do not treat input in the same way and, as a result, they learn differently. Literacy acquisition is another related aspect of language learning which may differ among children and adults in how they perceive and internalize linguistic input. According to Bialystok and Hakuta (1999, p. 171), “the acquisition of literacy inflicts permanent change on children’s conception of language… All children, irrespective of the language they read, advance rapidly in their metalinguistic concepts as literacy is established.” Even when the bulk of research on adult L2 acquisition tests literate adults, the role of literacy in adult L2 acquisition has not been extensively researched. With the exception of the migrant workers who might have no academic skills in the L2, most adult L2 learners receive instruction in the L2 through reading and writing, and many of these instructional forms are not available to preliterate or younger learners. Bialystok and Hakuta (1999) believe that literacy may explain why many children of immigrants are successful second language learners, whereas their parents may not be. Still, if we think of the student who starts acquisition of an L2 in college, is very literate, and finds himself surrounded by texts of all sorts in the target language, like Kevin and Carolina in the opening chapter, no amount of literacy and exposure to the language can make that student achieve the same level of proficiency of the child of an immigrant, like Alicia and Graciela portrayed in Chapter 1. On the other side of the question, bilingual children schooled in the majority language who develop little to no literacy in their L1 (Alicia and Graciela) fail to reach age-appropriate levels of linguistic competence in the L1, whereas adults cut off from L1 texts completely for a long period of time will not lose ability in their L1 (Francesco). If metalinguistic skills help L1- and L2-acquiring children strengthen linguistic knowledge in their L1 or L2, these same skills do not have the same effect in adult L2 acquisition and L1 loss. Further research on how literacy contributes to later language development and stabilization in monolingual and bilingual children and adults is necessary. Finally, in a recent review of how cognitive mechanisms change through the lifespan, Craik and Bialystok (2006) distinguish different levels of the cognitive representation system that are hierarchically organized: general, conceptual, context-free knowledge are at the higher level whereas episodic instances, lexical and phonological information and category exemplars are the lower level. Access to these levels operates asymmetrically during the lifespan: children first develop the lower levels and gradually gain access to the higher levels. By contrast, adults are good at retaining the higher levels and progressively lose access to the lower levels. Extending this explanation to the language situations discussed in this book, it seems that children are better with bottom-up linguistic processing while adults



Chapter 8.  Implications 275

rely more on top-down processing. Bottom-up processing takes into account specific linguistic details; top-down processing takes as input larger units and uses contextual information for processing as well. This cognitive processing account is consistent with Elissa Newport’s Less is More Hypothesis. In conclusion, input plays a role in language acquisition and language loss. Since children are highly reactive to input whereas adults are less so, this suggests that both input and age-related linguistic and cognitive mechanisms interact closely to bring about different linguistic outcomes. It is optimal amounts of input of good quality during childhood that matters for complete first and second language acquisition.

4.

Future research

I have argued in this book that there are age effects in both L2 acquisition and L1 loss. This follows from the Critical Period Hypothesis. While the Critical Period Hypothesis cannot entirely explain age effects in L2 acquisition, it appears to be eminently more applicable and predictive in the case of L1 loss at this early stage of research in this area. This conclusion stems from the fact that the shape of the age function may be different in the two situations. In L2 acquisition there is a linear decline ad infinitum, whereas language loss occurs during a specific period in childhood. Apparently, after age 9, there is practically no vulnerability to loss. The extent to which age effects in L1 attrition are mainly biological, mainly environmental, or both needs to be pursued empirically. Clearly, more research needs to be done to understand how incomplete L2 acquisition in adulthood and incomplete L1 acquisition in childhood differ from each other as a function of age. We need to learn more about how language learning and processing mechanisms develop and change with age in both monolingual and bilingual individuals. We need to learn more about how these mechanisms interact with the environment in guiding learners to detect, encode and filter linguistic stimuli. We need to learn more about how the interaction of these factors results in complete and incomplete grammars in different bilingual situations. And we also need to learn whether and how child and adult incomplete acquisition can be overcome with re-exposure and instruction. Research on language maintenance and loss in bilingualism is in its infancy, but combining the richness of contemporary grammatical theory and the growing sophistication of language processing research would allow this field to grow and provide more conclusive answers to many of the basic research questions posed in this book.

References

Abrahamsson, N. & Hyltenstam, K. In press. The robustness of aptitude effects in near-native second language acquisition. Studies in Second Language Acquisition 30, 4 (December 2008). Abutalebi, J., Cappa, S. & Perani, D. 2001. The bilingual brain as revealed by functional neuroimaging. Bilingualism: Language and Cognition 4, 179–190. Abutalebi, J., Cappa, S. & Perani, D. 2005. What can functional neuroimaging tell us about the bilingual brain? In Handbook of Bilingualism: Psycholinguistic Approaches, J. Kroll & A. de Groot (eds.), 497–515. Oxford: OUP. Abutalebi, J. & Green, D. 2007. Bilingual language production: The neurocognition of language representation and control. Journal of Neurolinguistics 20, 242–275. Allen, S. 2007. The future of Inuktitut in the face of majority languages: Bilingualism or language shift? Applied Psycholinguistics 28, 515–536. Allen, S. 1996. Aspects of Argument Structure Acquisition in Inuktitut. Amsterdam: John Benjamins. Allen, S. Crago, M. & Presco, D. 2006. The effect of majority language exposure on minority language skills: The case of Inuktitut. International Journal of Bilingual Education and Bilingualism 9, 578–596. Allen, S., Genesee, F., Fish, S. & Crago, M. 2002. Patterns of code mixing in English-Inuktitut bilinguals. In Proceedings of the 37th Annual Meeting of the Chicago Linguistic Society, M. Andronis, C. Ball, H. Elston & S. Neuvel (eds.), Vol. 2, 171–188. Chicago IL: Chicago Linguistic Society. Altenberg, E. 1991. Assessing first language vulnerability to attrition. In First Language Attrition, H. W. Seliger & R. M. Vago (eds.), 189–206. Cambridge: CUP. Ammerlaan, T. 1996. “You get a bit Wobbly” Exploring Bilingual Lexical Retrieval Processes in the Context of First Language Attrition. PhD dissertation, University of Nijmegen. Andersen, R. 1982. Determining the linguistic attributes of language attrition. In The Loss of Language Skills, R. Lambert & B. Freed (eds.), 83–118. Rowley MA: Newbury House. Anderson, R. 1999. Noun phrase gender agreement in language attrition. Preliminary results. Bilingual Research Journal 23, 318–337. Anderson, R. 2001. Lexical morphology and verb use in child first language loss. A preliminary case study investigation. International Journal of Bilingualism 5, 377–401. Arsenian, S. 1937. Bilingualism and Mental Development. New York NY: Columbia University Press. As, A. 1962. The recovery of forgotten language knowledge through hypnotic age regression: A case report. The American Journal of Clinical Hypnosis 5, 24–29. Au, T., Knightly, L, Jun, S. & Oh, J. 2002. Overhearing a language during childhood. Psychological Science 13, 238–243.

278 Incomplete Acquisition in Bilingualism

Balcom, P. 2003. Crosslinguistic influence of L2 English on middle constructions in L1 French. In Effects of the Second Language on the First, V. Cook (ed.), 168–192. Clevedon: Multilingual Matters. Bard, E., Robertson, D. & Sorace, A. 1996. Magnitude estimation of linguistic acceptability. Language 72, 32–68. Bel, A. 2001. Teoría Lingüística i adquisició del llenguatge. Analisi comparada del trets morfologics en catala i en castella. Barcelona: Institut d’Estudis Catalans. Belikova, A. & White, L. In press. Evidence for the Fundamental Difference Hypothesis or not? Subjacency revisited. Studies in Second Language Acquisition. 2009, 31, 2. Berko-Gleason, J. 1982. Insights from child language acquisition for second language loss. In The Loss of Language Skills, R. Lambert & B. Freed (ed.), 13–23. Rowley MA: Newbury House. Bernardini, P. & Schlyter, S. 2004. Growing syntactic structure and code-mixing in the weaker language: The Ivy Hypothesis. Bilingualism: Language and Cognition 7, 49–70. Bialystok, E. 1997. The structure of age. In search of barriers to second language acquisition. Second Language Research 13, 116–137. Bialystok, E. 2001. Bilingualism in Development. Language, Literacy and Cognition. Cambridge: CUP. Bialystok, E. & Hakuta, K. 1994. In Other Words. The Science and Psychology of Second Language Acquisition. New York NY: Basic Books. Bialystok, E. & Hakuta, K. 1999. Confounded age. Linguistic and cognitive factors in age differences for second language acquisition. In Second Language Acquisition and the Critical Period Hypothesis, D. Birdsong (ed.), 161–181. Mahwah NJ: Lawrence Erlbaum. Bialystok, E. & Martin, M. 2004. Attention and inhibition in bilingual children. Evidence from the dimensional change card sort task. Developmental Science 7, 225–339. Bialystok, E. & Miller, B. 1999. The problem of age in second language acquisition. Influences from language, task, and structure. Bilingualism: Language and Cognition 2, 127–145. Bialystok, E. & Shapero, D. 2005. Ambiguity benefits. The effects of bilingualism on reversing ambiguous figures. Developmental Science 8, 595–604. Bialystok, E. & Ryan, E. 1985. A metacognitive framework for the development of first and second language skills. In Metacognition, Cognition and Human Performance, D. L. ForresterPressley, G. MacKinnon & T. Walker (eds.), 207–252. Orlando FL: Academic Press. Bickerton, D. 1981. Roots of Language. Ann Arbor MI: Karoma. Bickerton, D. 1984. The language bioprogram hypothesis. Behavioral and Brain Sciences 7, 173–221. Bickerton, D. 1988. Creole languages and the bioprogram. In Linguistics. The Cambridge Survey, F. Newmeyer (ed.), 267–284. Cambridge: CUP. Bickerton, D. 1999. How to acquire a language without positive evidence. What acquisitionists can learn from creoles. In Language Creation and Language Change, M. DeGraff (ed.), 49–74. Cambridge MA: The MIT Press. Birdsong, D. 1989. Metalinguistic Performance and Interlinguistic Competence. Berlin: Springer. Birdsong, D. 1992. Ultimate attainment in SLA. Language 68, 706–755. Birdsong, D. (ed.). 1999. Second Language Acquisition and the Critical Period Hypothesis. Mahwah NJ: Lawrence Erlbaum. Birdsong, D. 2005. Interpreting age effects in second language acquisition. In Handbook of Bilingualism. Psycholinguistic Perspectives, J. Kroll & A. de Groot (eds.), 109–127. Oxford: OUP.



References 279

Birdsong, D. 2006. Age and second language acquisition and processing. A selective overview. Language Learning 56, 9–49. Birdsong, D. & Flege, J. 2001. Regular-irregular dissociations in the acquisition of English as a second language. Proceedings of the 25th Boston University Conference on Language Development, 123–132. Somerville MA: Cascadilla. Birdsong, D. & Molis, M. 2001. On the evidence for maturational constraints in second language acquisition. Journal of Memory and Language 44, 235–249. Blake, R. 1983. Mood selection among Spanish-speaking children, ages 4 to 12. The Bilingual Review 10, 21–32. Bley-Vroman, R. 1989. The logical problem of second language learning. In Linguistic Perspectives on Second Language Acquisition, S. Gass & J. Schachter (eds.), 41–68. Cambridge: CUP. Bley-Vroman, R. 1990. The logical problem of foreign language learning. Linguistic Analysis 20, 3–49. Bolonyai, A. 1998. In between languages. Language shift/maintenance in childhood bilingualism. The International Journal of Bilingualism 2, 21–43. Bolonyai, A. 2007. (In)vulnerable agreement in incomplete bilingual L1 learners. The International Journal of Bilingualism 11, 3–21. Bongaerts, T. 1999. Ultimate attainment in L2 pronunciation. The case of very advanced late learners. In Second Language Acquisition and the Critical Period Hypothesis, D. Birdsong (ed.), 133–160. Mahwah NJ: Lawrence Erlbaum. Bongaerts, T. 2005. Introduction. Ultimate attainment and the critical period hypothesis for second language acquisition. IRAL 43, 259–267. Bornstein, M. 1987. Sensitive periods in development. Definition, existence, utility, and meaning. In Sensitive Periods in Development. Interdisciplinary Perspectives, M. Bornstein (ed.), 3–17. Hillsdale NJ: Lawrence Erlbaum. Bornstein, M. 1989. Sensitive period in development. Structural characteristics and causal interpretations. Psychological Bulletin 105, 179–197. Bowles, M. & Montrul, S. In press. Instructed L2 acquisition of differential object marking in Spanish. In Little Words. Their History, Phonology, Syntax, Semantics, Pragmatics and Acquisition [Georgetown University Round Table], R. Leow, H. Campos & D. Lardiere (eds.). Washington DC: Georgetown University Press. de Bot, K. 1991. Language attrition, competence loss or performance loss. In Sprache und Politik, B. Spillner (ed.), 63–65. Frankfurt: Peter Lang. de Bot, K. (2002, August). Language attrition. Where are we and where are we going? Plenary talk. First International Conference on First Language Attrition, Vrije Universiteit, Amsterdam. de Bot, K. 2004. Introduction to special issue on language attrition. International Journal of Bilingualism 8, 233–237. de Bot, K. 2007. Dynamic system theory, lifespan development and language attrition. In Language Attrition. Theoretical Perspectives, Köpke, B., Schmid, M., Keijzer, M. & Dostert, S. (eds.), 53–67. Amsterdam: John Benjamins. de Bot, K. & Clyne, M. 1989. Language reversion revisited. Studies in Second Language Acquisition 11, 167–177. de Bot, K. & Clyne, M. 1994. A 16-year longitudinal study of attrition in Dutch immigrants in Australia. Journal of Multilingual and Multicultural Development 15, 17–28.

280 Incomplete Acquisition in Bilingualism

de Bot, K. & Weltens, B. 1991. Recapitulation, regression, and language loss. In First Language Attrition, H. Seliger & R. Vago (eds.), 31–51. Cambridge: CUP. de Bot, K, Gommans, P. & Rossing, C. 1991. L1 loss in an L2 environment. Dutch immigrants in France. In First Language Attrition, H. Seliger & R. Vago (eds.), 87–98. Cambridge: CUP. Braine, M. D., Brooks, P., Cowan, N., Samuels, M. & Tamis-LeMonda, S. 1993. The development of categories at the semantics/syntax interface. Cognitive Development 8, 465–494. Brauth, S., Hall, W. & Dooling, R. (eds.). 1991. Plasticity of Development. Cambridge MA: The MIT Press. Brinton, D., Kagan, O. & Bauckus, S. (eds.). 2008. Heritage Language Education. A New Field Emerging. London: Routledge. Brown, C. 1998. The role of the L1 grammar in the acquisition of the L2 segmental structure. Second Language Research 14, 136–193. Brown, R. 1958. Words and Things. New York NY: The Free Press. Brown, R. 1973. A First Language: The Early Stages. Cambridge MA: Harvard University Press. Bruhn de Garavito, J. 2002. Verb raising in Spanish. A comparison of early and late bilinguals. Proceedings of the 26th Annual Boston University Conference on Language Development, 84–94. Somerville MA: Cascadilla. Bruhn de Garavito, J. & White, L. 2003. The L2 acquisition of Spanish DPs. The status of grammatical features. In The Acquisition of Spanish Morphosyntax. The L1/L2 Connection, A. T. Pérez-Leroux & J. Liceras (eds.), 151–176. Dordrecht: Kluwer. Bullock, B. & Gerfen, C. 2004. Phonological convergence in a contracting language variety. Bilingualism: Language and Cognition 7, 95–104. Bullock, B. & Toribio, A. J. 2004. Convergence as an emergent property in bilingual speech. Bilingualism: Language and Cognition 7, 91–93. Burzio, L. 1986. Italian Syntax. Dordrecht: Reidel. Cagri, I. 2008, January. Heritage speakers and Persian complex predicates. Paper presented at the Linguistic Society of America Meeting, Chicago, Illinois, January 6, 2008. Caramazza, A. & Zurif, E. 1978. Comprehension of complex sentences in children and aphasics. A test of the regression hypothesis. In Language Acquisition and Language Breakdown. Parallels and Divergences, A. Caramazza & E. Zurif (eds.), 145–161. Baltimore MD: Johns Hopkins University Press. Cardinaletti, A. & Starke, M. 1994. The typology of structural deficiency in the three grammatical classes. Working Papers in Linguistics, Vol 4, 2. Venezia: Centro Linguistico Interfacoltà, Università degli Studi di Venezia. Carroll, S. 1989. Second language acquisition and the computational paradigm. Language Learning 39, 535–594. Carroll, S. 2001. Input and Evidence. The Raw Material of Language Acquisition. Amsterdam: John Benjamins. Carroll, S. 2005. Input and SLA: Adults’ sensitivity to different sorts of cues to French gender. Language Learning 55 Suppl, 79–138. Cho, S. 1999. The Acquisition of Relative Clauses. Experimental Studies on Korean. PhD dissertation, University of Hawaii. Chomsky, N. 1965. Aspects of the Theory of Syntax. Cambridge MA: The MIT Press. Chomsky, N. 1981. Lectures on Government and Binding. Dordrecht: Foris. Chomsky, N. 1986. Knowledge of Language: Its Nature, Origin and Use. New York NY: Praeger. Chomsky, N. 1995. The Minimalist Program. Cambridge MA: The MIT Press.



References 281

Chomsky, N. 2000. Minimalist inquiries. The framework. In Step by Step. Essays on Minimalist Syntax in Honor of Howard Lasnik, R. Martin, D. Michaels & J. Uriagereka (eds.), 9–155. Cambridge MA: The MIT Press. Chomsky, N. 2001. Derivation by phase. In Ken Hale, a Life in Language, M. Kenstowitz (ed.), 1–52, Cambridge MA: The MIT Press. Chomsky, N. 2004. Beyond explanatory adequacy. In Structures and Beyond. The Cartography of Syntactic Structures, A. Belletti (ed.), Vol. 3, 104–131. Oxford: OUP. Chomsky, N. & Lasnik, H. 1977. Filters and control. Linguistic Inquiry 8, 425–504. Clahsen, H., Aveledo, F. & Rocca, I. 2002. The development of regular and irregular verb inflection in Spanish child language. Journal of Child Language 29, 591–622. Clahsen, H. & Muysken, P. 1986. The availability of Universal Grammar to adult and child learners. A study of the acquisition of German word order. Second Language Research 2, 93–109. Clahsen, H., Marcus, G., Bartke, S. & Wiese, R. 1996. Compounding and inflection in German child language. In Yearbook of Morphology 1995, G. Booijand & J. van Marle (eds.), 115–142. Dordrecht: Kluwer. Clahsen, H. Rothweiler, M. & Woest, A. 1990. Lexikalische Ebenen und morphologische Entwicklung. Eine Untersuchung zum Erwerb des deutschen Plural-systems im Rahmen der Lexikalischen Morphologie. Linguistische Berichte Suppl. 3, 105–126. Colombo, J. 1982. The critical period concept. Research, methodology, and theoretical concerns. Psychological Bulletin 91, 260–275. Cook, V. (ed.). 2003. The Effects of the Second Language on the First. Clevedon: Multilingual Matters. Coppieters, R. 1987. Competence differences between native and near-native speakers. Language 63, 544–573. Corballis, M. 1991. The Lopsided Ape. The Generative Mind. Oxford: OUP. Cornips, L. & Hulk, A. 2006. External and internal factors in bilingual and bidialectal language development. Grammatical gender of the Dutch definite determiner. In L2 Acquisition and Creole Genesis. Dialogues, C. Lefebvre, L. White & C. Jourdenais (eds.), 355–377. Amsterdam: John Benjamins. Crago, M. & Allen, S. 1998. Acquiring Inuktitut. In Language Acquisition in North America: Cross-cultural and Cross-linguistic Perspectives, L. Leonard & O. Taylor (eds.), 245–280. San Diego CA: Singular. Crago, M., Annahatak, B., Doehring, D. & Allen, S. 1991. Evaluation of Inuit children’s first language by native speakers. Journal of Speech-Language Pathology & Audiology 15, 43–48. Craik, F. & Bialystok, E. 2006. Cognition through the lifespan: Mechanisms of change. Trends in Cognitive Sciences 10, 131–138. Crain, S. & Lillo-Martin, D. 1999. An Introduction to Linguistic Theory and Language Acquisition. Oxford: Blackwell. Crain, S. & Thornton, R. 1998. Investigations in Universal Grammar. A Guide to Experiments on the Acquisition of Syntax and Semantics. Cambridge MA: The MIT Press. Curtiss, S. 1977. Genie. A Linguistic Study of a Modern Day “Wild” Child. New York NY: Academic Press. Curtiss, S. 1989. The case of Chelsea. A new case test of the critical period for language acquisition. Ms. University of California, Los Angeles. Curtiss, S., Fromkin, V., Krashen, S., Rigler, D. & Rigler, M. 1974. The linguistic development of Genie. Language 50, 528–554.

282 Incomplete Acquisition in Bilingualism

Davies, A. 2003. The Native Speaker. Myth and Reality. Clevedon: Multilingual Matters. DeKeyser, R. 2000. The robustness of critical period effects in second language acquisition. Studies in Second Language Acquisition 22, 499–534. DeKeyser, R. 2003. Implicit and explicit learning. In The Handbook of Second Language Acquisition, C. Doughty & M. Long (eds.), 313–348. Malden MA: Blackwell. DeKeyser, R. & Larson-Hall, J. 2005. What does the critical period really mean? In The Handbook of Bilingualism. Psycholinguistic Approaches, J. Kroll & A. de Groot (eds.), 88–108. Oxford: OUP. Doherty, M. & Perner, J. 1998. Metalinguistic awareness and theory of mind: Just two words for the same thing? Cognitive Development 13, 279–305. Doughty, C. 2003. Instructed SLA, constraints, compensation and enhancement. In The Handbook of Second Language Acquisition, C. Doughty & M. Long (eds.), 256–310. Oxford: Blackwell. Döpke, S. 1992a. One Parent, One Language. An Interactional Approach. Amsterdam: John Benjamins. Döpke, S. 1992b. A bilingual child’s struggle to comply with the one parent-one language rule. Journal of Multilingual and Multicultural Development 13, 467–485. Ellis, N. (ed.). 1994. Implicit and Explicit Learning of Languages. New York NY: Academic Press. Ellis, N. 2005. At the interface: Dynamic interactions of explicit and implicit language knowledge. Studies in Second Language Acquisition 27, 305–352. Ellis, R. 2005. Measuring implicit and explicit knowledge of a second language: A psychometric study. Studies in Second Language Acquisition 27, 141–172. Epstein, S., Flynn, S. & Martohardjono, G. 1996. Second language acquisition. Theoretical and experimental issues in contemporary research. Behavioral and Brain Sciences 19, 677– 758. Eubank, L. & Gregg, K. 1999. Critical periods and (second) language acquisition. Divide et impera. In Second Language Acquisition and the Critical Period Hypothesis, D. Birdsong (ed.), 65–93. Mahwah NJ: Lawrence Erlbaum. Flege, J. E. 1999. Age of learning and second language speech. In Second Language Acquisition and the Critical Period Hypothesis, D. Birdsong (ed.), 101–132. Mahwah, NJ: Lawrence Erlbaum. Flege, J. E. 2002. Interactions between the native and second language phonetic systems. In An Integrated View of Language Development. Papers in Honor of Henning Wode, P. Burmeister, T. Piske & A. Rohde (eds.), 217–244. Trier: WissenschaftlicherVerlag Trier. Flege, J. E., Munro, M. & MacKay, I. 1995. Factors affecting strength of perceived accent in a second language. Journal of the Acoustical Society of America 97, 3125–3134. Flege, J. E., Schirru, C. & MacKay, I. 2003. Interaction between the native and second language phonetic subsystems. Speech Communication 40, 467–491. Flege, J. E., Yeni-Komshian, G. & Liu, S. 1999. Age constraints on second language learning. Journal of Memory and Language 41, 78–104. Flege, J., Birdsong, D., Bialystok, E., Mack, M., Sung, H. & Tsukada, K. 2006. Degree of foreign accent in English sentences produced by Korean children and adults. Journal of Phonetics 34, 153–175. Foote, R. 2006. A Psycholinguistic Investigation of Agreement in Spanish and English Monolinguals and Bilinguals. PhD dissertation, University of Illinois at Urbana-Champaign.



References 283

Footnick, R. 2007. A Hidden language. Recovery of a “lost” language is triggered by hypnosis. In Language Attrition. Theoretical Perspectives, Köpke, B., Schmid, M., Keijzer, M. & S. Dostert (eds.), 169–187. Amsterdam: John Benjamins. Franceschina, F. 2001. Morphological or syntactic deficits in near-native speakers? An assessment of some current proposals. Second Language Research 17, 213–247. Franceschina, F. 2005. Fossilized Second Language Grammars. The Acquisition of Grammatical Gender. Amsterdam: John Benjamins. Fromm, E. 1970. Age regression with unexpected reappearance of a repressed childhood language. The International Journal of Clinical and Experimental Hypnosis 18, 79–88. Gardner, R. & Lambert, W. 1972. Attitudes and Motivation in Second Language Learning. Rowley MA: Newbury House. Gass, S. 1979. Language transfer and universal grammatical relations. Language Learning 29, 327–344. Gavruseva, E. & Lardiere, D. 1996. The emergence of extended phrase structure. Proceedings of the 20th Boston University Conference on Language Development, 225–236. Somerville MA: Cascadilla. Gawlitzek-Maiwald, I. & Tracy, R. 1996. Bilingual bootstrapping. Linguistics 34, 901–926. Genesee, F. 1989. Early bilingual development. One language or two? Journal of Child Language 6, 161–179. Genesee, F. 2000. Syntactic aspects of bilingual acquisition. Bilingualism: Language and Cognition 3, 167–172. Genesee, F., Boivin, I. & Nicoladis, E. 1996. Talking with strangers. A study of bilingual children’s communicative competence. Applied Psycholinguistics 17, 427–442. Genesee, F., Nicoladis, E. & Paradis, J. 1995. Language differentiation in early bilingual development. Journal of Child Language 22, 611–631. Geren, J., Snedeker, J. & Ax, L. 2005. Starting over. A preliminary study of early lexical and syntactic development in internationally adopted preschoolers. Seminars in Speech and Language 26, 44–53. Giorgi, A. & Pianesi, F. 1997. Tense and Aspect. From Semantics to Morphosyntax. Oxford: OUP. Gleitman, L. & Newport, E. 1995. The invention of language by children. Environmental and biological influences on the acquisition of language. In Language. An Invitation to Cognitive Science, L. Gleitman & M. Liberman (eds.), 2nd ed. Vol. 1, 1–24. Cambridge MA: The MIT Press. Glennen, S. 2005. New arrivals, speech and language assessment for internationally adopted infants and toddlers within the first months home. Seminars in Speech and Language 26, 10–21. Goad, H. & White, L. 2004. Ultimate attainment of L2 inflection, effects of L1 prosodic structure. In Eurosla Yearbook 4, S. Foster-Cohen, M. Sharwood Smith, A. Sorace & M. Ota (eds.), 119–145. Amsterdam: John Benjamins. Goad, H., White, L. & Steele, J. 2003. Missing surface inflection in L2 acquisition. A prosodic account. Proceedings of the 27th Annual Boston University Conference on Language Development, 264–275. Somerville MA: Cascadilla. Godson, L. 2004. Vowel production in the speech of Western Armenian heritage speakers. The Heritage Language Journal 2, 1–26. Gombert, J. 1992. Metalinguistic Development. London: Harvester Wheatsheaf.

284 Incomplete Acquisition in Bilingualism

Goral, M. 2004. First language decline in healthy aging. Implications for attrition in bilingualism. Journal of Neurolinguistics 17, 31–52. Green, D. 1986. Control, activation and resource. A framework and model for the control of speech in bilinguals. Brain and Language 27, 210–223. Grimshaw, J. & Sameck-Lodovici, V. 1998. Optimal subjects and subject universals. In Is the Best Good Enough? Optimality and Competition in Syntax, P. Barbosa, D. Fox, P. Hangstrom, M. McGinnis & D. Pesetsky (eds.), 193–219. Cambridge MA: MIT Press. Grinstead, J. 2004. Subjects and interface delay in child Spanish and Catalan. Language 80, 40–72. Grodzinsky, Y. 1990. Theoretical Perspectives on Language Deficits. Cambridge MA: The MIT Press. Grosjean, F. 1989. Neurolinguists, beware! The bilingual is not two monolinguals in one person. Brain and Language 36, 3–15. Grosjean, F. 1998. Studying bilinguals. Methodological and conceptual issues. Bilingualism: Language and Cognition 1, 131–149. Gross, S. 2004. A modest proposal. Explaining language attrition in the context of contact linguistics. In First Language Attrition. Interdisciplinary Perspectives on Methodological Issues, M. Schmid, B. Köpke, M. Keijzer & L. Weilemar (eds.), 281–298. Amsterdam: John Benjamins. Guasti, M. T. 2002. Language Acquisition. The Growth of Grammar. Cambridge MA: The MIT Press. Guasti, M. T. & Cardinaletti, A. 2003. Relative clause formation in Romance child’s production. Probus 15, 47–89. Guillelmon, D. & Grosjean, F. 2001. The gender marking effect in spoken word recognition. The case of bilinguals. Memory and Cognition 29, 503–511. Gürel, A. 2002. Linguistic Characteristics of Second Language Acquisition and First Language Attrition. Turkish Overt versus Null Pronouns. PhD dissertation, McGill University. Gürel, A. 2004. Selectivity in L2 induced L1 attrition. A psycholinguistic account. Journal of Neurolinguistics 17, 53–78. Gürel, A. 2007. (Psycho)linguistic determinants of L1 attrition. In Language Attrition. Theoretical Perspectives, Köpke, B., Schmid, M., Keijzer, M. & Dostert, S. (eds.), 99–120. Amsterdam: John Benjamins. Håkansson, G. 1995. Syntax and morphology in language attrition. A study of five bilingual expatriate Swedes. International Journal of Applied Linguistics 5, 153–171. Hakuta, K. & D’Andrea, D. 1992. Some properties of bilingual maintenance and loss in Mexican background high-school students. Applied Linguistics 13, 72–99. Hakuta, K., Butler, Y. G. & Witt, D. 2000. How long does it take English learners to attain proficiency? The University of California Minority Research Institute, policy report 2000–1. Online. Available at http//www.stanford.edu/hakuta/Docs/HowLong.pdf Han, Z. 2004. Fossilization. Five central issues. International Journal of Applied Linguistics 14, 212–242. Harley, B. & Wang, W. 1997. The critical period hypothesis. Where are we now? In Tutorials in Bilingualism, A. de Groot & J. Kroll (eds.), 19–51. Mahwah NJ: Lawrence Erlbaum. Hawkins, R. & Chan, C. 1997. The partial availability of Universal Grammar in second language acquisition. The ‘failed functional features hypothesis’. Second Language Research 13, 187–226.



References 285

Hawkins, R. & Liszka, S. 2003. Locating the source of defective past tense marking in advanced L2 English speakers. In The Lexicon-Syntax Interface in Second Language Acquisition, R. van Hout, A. Hulk, F. Kuiken & R. Towell (eds.), 21–44. Amsterdam: John Benjamins. Hawkins, R. & Franceschina, F. 2004. Explaining the acquisition and non-acquisition of determiner-noun gender concord in French and Spanish. In The Acquisition of French in Different Contexts, P. Prévost & J. Paradis (eds.), 175–205. Amsterdam: John Benjamins. Hawkins, R. & Hattori, H. 2006. Interpretation of English multiple wh-questions by Japanese speakers. A missing uninterpretable feature account. Second Language Research 22, 269–301. Haznedar, B. & Schwartz, B. 1997. L2 acquisition by a Turkish-speaking child. Evidence of L1 influence. Proceedings of the 21st Boston University Conference on Language Development, 245­–256. Somerville MA: Cascadilla. Hernández Pina, F. 1984. Teorías psicosociolingüísticas y su aplicación a la adquisición del español como lengua materna. Madrid: Siglo XXI. Herschensohn, J. 2000. The Second Time Around. Minimalism and L2 Acquisition. Amsterdam: John Benjamins. Herschensohn, J. 2007. Language Development and Age. Cambridge: CUP. Hirakawa, M. 1990. A study of the L2 acquisition of English reflexives. Second Language Research 6, 60–85. Horgan, D. 1978. The development of the full passive. Journal of Child Language 5, 65–80. Hudson Kam, C. & Newport, E. 2005. Regularizing unpredictable variation: The roles of adult and child learners in language formation and change. Language Learning and Development 1, 151–195. Hulk, A. & Cornips, L. 2006. Neuter gender and interface vulnerability in child L2/2L1 Dutch. In Paths of Development in L1 and L2 Acquisition. In Honor of Bonnie Schwartz, S. ­Unsworth, T. Parodi, A. Sorace & M. Young-Scholten (eds.), 107–134. Amsterdam: John Benjamins. Hulk, A. & Müller, N. 2000. Bilingual first language acquisition at the interface between syntax and pragmatics. Bilingualism: Language and Cognition 3, 227–244. Hull, R. & Vaid, J. 2007. Bilingual language lateralization: A meta-analytic tale of two hemispheres. Neuropsychologia 45, 1987–2008. Hulsen, M. 2000. Language Loss and Language Processing. Three Generations of Dutch Migrants in New Zealand. PhD dissertation, University of Nijmegen. Hulstijn, J. 2005. Theoretical and empirical issues in the study of implicit and explicit second language learning: Introduction. Studies in Second Language Acquisition 27, 129–140. Hutz, M. 2004. Is there a natural process of decay? A longitudinal study of language attrition. In First Language Attrition. Interdisciplinary Perspectives on Methodological Issues, M. Schmid, B. Köpke, M. Keijzer & L. Weilemar (eds.), 189–206. Amsterdam: John Benjamins. Hyams, N. 1986. Language Acquisition and the Theory of Parameters. Dordrecht: Reidel. Hyams, N. 1996. The underspecification of functional categories in early grammar. In Generative Perspectives on Language Acquisition, H. Clahsen (ed.), 91–128. Amsterdam: John Benjamins. Hyltenstam, K. 1992. Non-native features of near-native speakers. On ultimate attainment of childhood L2 learners. In Cognitive Processing in Bilinguals, R. Harris (ed.), 351–368. Amsterdam: Elsevier. Hyltenstam, K. & Abrahamsson, N. 2003. Maturational constraints in SLA. In The Handbook of Second Language Acquisition, C. Doughty & M. Long (eds.), 539–588. Oxford: Blackwell.

286 Incomplete Acquisition in Bilingualism

Ionin, T. & Wexler, K. 2002. Why is “Is” easier than “-s”? Acquisition of tense/agreement morphology by child L2-English learners. Second Language Research 18, 95–136. Isurin, L. 2000. Deserted island or a child’s first language forgetting. Bilingualism: Language and Cognition 3, 151–166. Itard, J.-M. 1801. De l’éducation d’un homme sauvage ou des premiers développements physiques et moraux du jeune sauvage de l’Aveyron. Paris: Gouyon. Jaeggli, O. 1982. Topics in Romance Syntax. Dordrecht: Foris. Jakobson, R. 1941[1962]. Kindersprache, Aphasie, und allgemeine Lautgezetze. Uppsala, Almqvist & Wiksell. Also in R. Jakobson, Selected Writings, 1, Phonological Studies. Den Haag: Mouton. Jaspaert, K. & Kroon, S. 1992. From the typewriter of A.L.: A case study of language loss. In Maintenance and Loss of Minority Languages, W. Fase, K. de Bot & T. van Els (eds.), 137– 147. Amsterdam: John Benjamins. Jia, G. & Aaronson, D. 1999. Age differences in second language acquisition. The dominant language switch and maintenance hypothesis. Proceedings of the 23rd Annual Boston University Conference on Language Development, 301–312. Somerville MA: Cascadilla. Jia, G. & Aaronson, D. 2003. A longitudinal study of Chinese children and adolescents learning English in the United States. Applied Psycholinguistics 24, 131–161. Jia, G., Aaronson, D. & Wu, Y. 2002. Long-term language attainment of bilingual immigrants. Predictive variables and language group differences. Applied Psycholinguistics 23, 599– 621. Jiang, N. 2007. Selective integration of linguistic knowledge in adult second language learning. Language Learning 57, 1–33. Jo, H.-Y. 2001. Heritage language learning and ethnic identity: Korean-Americans’ struggle with language authorities. Language, Culture and Curriculum 14, 26–41. Johnson, J. 1992. Critical period effects in second language acquisition. The effect of written versus auditory materials in the assessment of grammatical competence. Language Learning 42, 217–248. Johnson, J. & Newport, E. 1989. Critical period effects in second language learning. The influence of maturational state on the acquisition of English as a second language. Cognitive Psychology 21, 60–99. Johnson, J. & Newport, E. 1991. Critical period effects on universal principles of language. The status of subjacency in the acquisition of a second language. Cognition 39, 215–258. Kang, B.-M. 1998. Grammar and the use of language: Korean reflexives ‘caki, ‘casin’, and cakicasin’. Kwuk-e-hak 31, 165–204. Kasuya, H. 1998. Determinants of language choice in bilingual children. The role of input. The International Journal of Bilingualism 2, 327–346. Kato, M. Cyrino, S. & Corrêa, V. In press. Brazilian Portuguese and the recovery of lost clitics through schooling. In Minimalist Inquiries into Child and Adult Language Acquisition: Case Studies Across Portuguese, A. Pires & J. Rothman (eds.). Berlin: Mouton de Gruyter. Kaufman, D. & Aronoff, M. 1991. Morphological disintegration and reconstruction in first language attrition. In First Language Attrition, H. Seliger & R. Vago (eds.), 175–188. Amsterdam: John Benjamins. Kaye, A. 2001. Diglossia: The state of the art. International Journal of the Sociology of Language 152, 117–129. Kayne, R. 1989. Null subjects and clitic climbing, In The Null Subject Parameter, O. Jaeggli & K. Safir (eds.), 239–261. Dordrecht: Reidel.



References 287

Kegl, J., Senghas, A. & Coppola, M. 1999. Creation through contact. Sign language emergence and sign language change in Nicaragua. In Language Creation and Language Change, M. DeGraff (ed.), 179–238, Cambridge MA: The MIT Press. Kehoe, M., Lleó, C. & Rakow, M. 2004. Voice onset time in bilingual German-Spanish children. Bilingualism: Language and Cognition 7, 71–88. Keijzer, M. 2007. Last in First Out? An Investigation of the Regression Hypothesis in Dutch Emigrants in Anglophone Canada. Vrije Universiteit Amsterdam, LOT: Netherlands Graduate School of Linguistics. Kim, E. J. 2004. Korean-English bilinguals and heritage language maintenance. In Korean Language in America 9: Papers from the eighth Annual Conference and Professional Development Worskshop, E. J. Kim (ed.), 315–326. Honolulu HI: American Association of Teachers of Korean. Kim, G. E. 2006. Heritage language maintenance by Korean-American college students. In Heritage Language Development. Focus on East Asian Immigrants, K. Kondo-Brown (ed.), 175–208. Amsterdam: John Benjamins. Kim, H.-S. 2005. Processing Strategies and Transfer of Heritage and Non-heritage Learners of Korean. PhD dissertation, University of Hawai’i at Manoa, Honolulu. Kim, J.-H. 2007. Binding Interpretations in Korean Heritage Speakers and L2 Learners. PhD dissertation, University of Illinois at Urbana-Champaign. Kim, J.-H., Montrul, S. & Yoon, J. In press. Dominant language influence in acquisition and attrition of binding, Interpretation of the Korean reflexive caki. Bilingualism: Language and Cognition, 2010. Kim, J.-H, Montrul, S. & Yoon, J. Under review. Binding interpretation of anaphors in Korean heritage speakers. Knightly, L., Jun, S., Oh, J. & Au, T. 2003. Production benefits of childhood overhearing. Journal of the Acoustic Society of America 114, 465–474. Koehn, C. 1994. The acquisition of gender and number morphology within NP. In Bilingual First Language Acquisition, German and French, J. Meisel (ed.), 29–51. Amsterdam: John Benjamins. Kohnert, K. 2002. Picture naming in early sequential bilinguals. A 1-year follow-up. Journal of Speech, Language and Hearing Research 45, 759–771. Kohnert, K. & Bates, E. 2002. Balancing bilinguals II. Lexical comprehension and cognitive processing in children learning Spanish and English. Journal of Speech, Language, and Hearing Research 45, 347–359. Kohnert, K., Bates, E. & Hernández, A. 1999. Balancing bilinguals, Lexical-semantic production and cognitive processing in children learning Spanish and English. Journal of Speech, Language and Hearing Research 42, 1400–1413. Kondo-Brown, K. (ed.). 2006. Heritage Language Development. Focus on East Asian Immigrants. Amsterdam: John Benjamins. Kondo-Brown, K. 2004. Do background variables predict students’ scores on a Japanese placement test? Implications for placing heritage language learners. Journal of the National Council of Less Commonly Taught Languages 1, 1–19. Köpcke, K. 1988. Schemas in German plural formation. Lingua 74, 303–355. Köpke, B. 1999. L’attrition de la première langue chez le bilingue tardif, Implications pour l’étude psycholinguistique du bilinguisme. PhD dissertation, Université de Toulouse-Le Mirail.

288 Incomplete Acquisition in Bilingualism

Köpke, B. 2002. Activation thresholds and non-pathological L1 attrition. In Advances in the Neurolinguistics of Bilingualism. Essays in Honor of Michel Paradis, F. Fabbro (ed.), 119– 142. Undine: Forum. Köpke, B. 2004. Neurolinguistic aspects of attrition. Journal of Neurolinguistics 17, 3–30. Köpke, B. 2007. Language attrition at the crossroads of brain, mind, and society. In Language Attrition. Theoretical Perspectives, B. Köpke, M. Schmid, M. Keijzer & S. Dostert (eds.), 9–37. Amsterdam: John Benjamins. Köpke, B. & Schmid, M. 2003. Language attrition. The next phase. In First Language Attrition. Interdisciplinary Perspectives on Methodological Issues, M. Schmid, B. Köpke, M. Keijzer & L. Weilemar (eds.), 1–143. Amsterdam: John Benjamins. Köpke, B., Schmid, M., Keijzer, M. & Dostert, S. (eds.). 2007. Language Attrition. Theoretical Perspectives. Amsterdam: John Benjamins. Krashen, S. 1973. Lateralization, language learning and the critical period. Some new evidence. Language Learning 23, 63–74. Krashen, S. 1981. Second Language Acquisition and Second Language Learning. Oxford: Pergamon. Kravin, H. 1992. Erosion of a language in bilingual development. Journal of Multilingual and Multicultural Development 13, 307­–325. Krakow, R., Tao, S. & Roberts, J. 2005. Adoption age effects on English language acquisition. Infants and toddlers from China. Seminars in Speech and Language 26, 33–43. Lafond, L., Hayes, R. & Bhatt, R. 2001. Constraint demotion and null subjects in Spanish L2 acquisition. In Romance Syntax, Semantics and L2 Acquisition, J. Camps & C. Wiltshire (eds.), 121–135. Amsterdam: John Benjamins. Lakshmanan, U. 1995. Child second language acquisition of syntax. Studies in Second Language Acquisition 17, 301–329. Lambert, W. 1977. Effects of bilingualism on the individual. Cognitive and sociocultural consequences. In Bilingualism, Psychological, Social and Educational Implications, P. A. Hornby (ed.), 15–28. New York NY: Academic Press. Lanza, E. 1997. Language Mixing in Infant Bilingualism. A Sociolinguistic Perspective. Oxford: Clarendon Press. Lanza, E. 2001. Bilingual first language acquisition. A discourse perspective on language contact in parent-child interaction. In Trends in Bilingual Acquisition, J. Cenoz & F. Genesee (eds.), 201–230. Amsterdam: John Benjamins. Lardiere, D. 1998a. Case and tense in the fossilized steady state. Second Language Research 14, 1–26. Lardiere, D. 1998b. Dissociating syntax from morphology in a divergent end-state grammar. Second Language Research 14, 359–375. Lardiere, D. 2007. The Fossilized Steady State in Second Language Acquisition. Mahwah NJ: Lawrence Erlbaum. Leadholm, B. & Miller, J. 1992. Language Sample Analysis: The Wisconsin Guide. Madison WI: Wisconsin Department of Public Instruction. Lee, D. & Schachter, J. 1997. Sensitive period effects in binding theory. Language Acquisition 6, 333–362. Lee, Y.-G., Kim, H.-S., Kong, H., Hong, J.-M. & Long, M. 2008. Variation of the linguistic profiles of advanced English-speaking learners of Korean. In Heritage Language Education. A New Field Emerging, D. Brinton, O. Kagan & S. Bauckus (eds.), 165–180. London: Routledge.



References 289

Lenneberg, E. 1967. Biological Foundations of Language. New York, NY: Wiley. Lightbown, P. 1998. The importance of timing in focus on form. In Focus on Form in Classroom Second Language Acquisition, C. Doughty & J. Williams (eds.), 177–196. Cambridge: CUP. Lightfoot, D. 1979. The Principles of Diachronic Syntax. Cambridge: CUP. Lightfoot, D. 1991. How to Set Parameters: Arguments from Language Change. Cambridge MA: The MIT Press. Lightfoot., D. 1999. The Development of Language. Acquisition, Change and Evolution. Oxford: Blackwell. Lin, J. 2003. Temporal reference in Mandarin Chinese. Journal of East Asian Linguistics 12, 259–311. Lipski, J. 1993. Creoloid phenomena in the Spanish of transitional bilinguals. In Spanish in the United States, A. Roca & J. Lipski (eds.), 155–173. Berlin: Mouton. Lleó, C. 1998. Proto-articles in the acquisition of Spanish. Interface between phonology and syntax. In Models of Inflection, R. Fabri, A. Ortman & T. Parodi (eds.), 175–195. Tübingen: Niemeyer. Loban, W. 1976. Language development: Kindergarten through grade twelve [Research Report No. 18]. Urbana IL: National Council of Teachers of English. Long, M. 1990. Maturational constraints on language development. Studies in Second Language Acquisition 12, 251–285. Long, M. 2003. Stabilization and fossilization in interlanguage development. In The Handbook of Second Language Acquisition, C. Doughty & M. Long (eds.), 487–536. Oxford: Blackwell. Long, M. 2007. Problems in SLA. Mahwah NJ: Lawrence Erlbaum. López Ornat, S. 1997. What lies in between a pre-grammatical and a grammatical representation? Evidence on nominal and verbal form-function mapping in Spanish from 1;7 to 2;1. In Contemporary Perspectives on the Acquisition of Spanish, A. T. Pérez-Leroux & W. Glass (eds.), 3–20. Somerville MA: Cascadilla. Luján, M. & Parodi, C. 1996. Clitic-doubling and the acquisition of agreement in Spanish. In Perspectives on Spanish Linguistics, Vol.1, J. Gutiérrez-Rexach & L. Silva Villar (eds.), 119– 138. Los Angeles CA: UCLA, Department of Linguistics. Luo, S. H. & Wiseman, R. 2000. Ethnic language maintenance among Chinese immigrant children in the United States. International Journal of Intercultural Relations 24, 307–324. Lynch, A. 1999. The Subjunctive in Miami Cuban Spanish. Bilingualism. Contact and Language Variability. PhD dissertation, University of Minnesota. Mack, M. 1997. The monolingual native speaker. Not a norm, but still a necessity. Studies in the Linguistic Sciences 27, 113–146. Macnamara, J. 1966. Bilingualism and Primary Education. Edinburgh: EUP. MacWhinney, B. 1992. Transfer and competition in second language learning. In Cognitive Processing in Bilinguals, R. Harris (ed.), 371–390. Amsterdam: Elsevier. Major, R. 1992. Losing English as a first language. The Modern Language Journal 76, 190–209. Marinova-Todd, S. 2003. Native, Near-Native or Non-Native: Comprehensive Analysis of the Linguistic Profiles of Highly Proficient Adult Second Language Learners. Ms. Harvard Graduate School of Education. Mason, M. 1942. Learning to speak after six and one half years of silence. Journal of Speech Disorders 7, 295–304.

290 Incomplete Acquisition in Bilingualism

Matsunaga, S. 2003. Instructional needs of college-level learners of Japanese as a heritage language. Performance based analysis. Heritage Language Journal 1, 1 www.heritagelanguages. org Mayberry, R. 1993. First language acquisition after childhood differs from second language acquisition. The case of American Sign Language. Journal of Speech and Hearing Research 36, 1258–1270. Mayberry, R. I. In press. Second language learning of sign languages. In Sign Language. Encyclopedia of Language and Linguistics, 2nd edn, B. Woll (ed.). Oxford: Elsevier. Mayberry, R. & Eichen, E. 1991. The long-lasting advantages of learning sign language in childhood. Another look at the critical period in language acquisition. Journal of Memory and Language 30, 486–512. Mayberry, R. I. & Squires, B. In press. Sign language acquisition. In Language Acquisition. Encyclopedia of Language and Linguistics, 2nd edn, E. Lieven (ed.), Oxford: Elsevier. Mayberry, R. I. & Lock, E. 2003. Age constraints on first versus second language acquisition: Evidence for linguistic plasticity and epigenesis. Brain and Language 87, 369–383. Mayberry, R. I., Lock, E. & Kazmi, H. 2002. Linguistic ability and early language exposure. Nature 417, 38. Mayer, M. 1969. Frog, where are you? New York NY: Dial Press. McCarthy, C. 2007. Morphological Variability in Second Language Spanish. PhD dissertation, McGill University. McDaniel, D., Chiu, B. & Maxfield, T. 1995. Parameters for wh-movement types: Evidence from child language. Natural Language & Linguistic Theory 13, 709–753. McDonald, J. 2000. Grammaticality judgments in a second language: Influences of age of acquisition and native language. Applied Psycholinguistics 21, 395–423. McElree, B., Jia, G. & Litvak, A. 2000. The time course of conceptual processing in three bilingual populations. Journal of Memory and Language 42, 229–423. McWhorter, J. 2007. Language Interrupted. Signs of Non-native Acquisition in Standard Language Grammars. Oxford: OUP. Meisel, J. 1986. Word order and case marking in early child language: Evidence from simultaneous acquisition of two first languages: French and German. Linguistics 24, 123–183. Meisel, J. 1989. Early differentiation of languages in bilingual children. In Bilingualism Across the Lifespan, Aspects of Acquisition, Maturity and Loss, K. Hyltenstam & L. Obler (eds.), 13–40. Cambridge: CUP. Meisel, J. 1990. INFL-ection, subjects and subject-verb agreement. In Two First Languages, Early Grammatical Development in Bilingual Children, J. Meisel (ed.), 237–300. Dordrecht: Foris. Meisel, J. 1994a. Code-switching in young bilingual children. The acquisition of grammatical constraints. Studies in Second Language Acquisition 16, 413–441. Meisel, J. 1994b. Getting FAT, finiteness, agreement and tense in early grammars. In Bilingual First Language Acquisition. French and German Grammatical Development. J. Meisel (ed.), 89–130. Amsterdam: John Benjamins. Meisel, J. 1997. The acquisition of the syntax of negation in French and German. Contrasting first and second language development. Second Language Research 13, 227–263. Meisel, J. 2001. The simultaneous acquisition of two first languages. Early differentiation and subsequent development of grammars. In Trends in Bilingual Acquisition, J. Cenoz & F. Genesee (eds.), 11–42. Amsterdam: John Benjamins.



References 291

Meisel, J. 2004. The bilingual child. In The Handbook of Bilingualism, T. Bhatia & W. Ritchie (eds.), 91–113. Oxford: Blackwell. Meisel, J. 2007. The weaker language in early child bilingualism: Acquiring a first language as a second language? Applied Psycholinguistics 28, 495–514. Menyuk, P. & Brisk, M. E. 2005. Language Development and Education. Children with Varying Language Experience. Basingstoke: Palgrave. Merino, B. 1983. Language loss in bilingual Chicano children. Journal of Applied Developmental Psychology 4, 277–294. de Miguel, E. 1992. El aspecto en la sintaxis del español. Perfectividad e imperfectividad. Madrid: Ediciones de la Universidad Autónoma de Madrid. Montrul, S. 1999. Se o no Se. A look at transitive and intransitive verbs. Spanish Applied Linguistics 3, 145–194. Montrul, S. 2000. Transitivity alternations in second language acquisition. Toward a modular view of transfer. Studies in Second Language Acquisition 22, 229–274. Montrul, S. 2001a. Causatives and transitivity in L2 English. Language Learning 51, 51–106. Montrul, S. 2001b. The acquisition of causative and inchoative verbs in L2 Turkish. Language Acquisition 9, 1–58. Montrul, S. 2002. Incomplete acquisition and attrition of Spanish tense/aspect distinctions in adult bilinguals. Bilingualism: Language and Cognition 5, 39–68. Montrul, S. 2004a. Subject and object expression in Spanish heritage speakers. A case of morpho-syntactic convergence. Bilingualism: Language and Cognition 7, 125–142. Montrul, S. 2004b. The Acquisition of Spanish. Amsterdam: John Benjamins. Montrul, S. 2004c. Convergent outcomes in second language acquisition and first language loss. In First Language Attrition: Interdisciplinary Perspectives on Methodological Issues, M. Schmid, B. Köpke, M. Keijzer & L. Weilemar (eds.), 259–280. Amsterdam: John Benjamins. Montrul, S. 2005. Second language acquisition and first language loss in adult early bilinguals. Exploring some differences and similarities. Second Language Research 21, 199–249. Montrul, S. 2006. On the bilingual competence of Spanish heritage speakers. Syntax, lexicalsemantics and processing. International Journal of Bilingualism 10, 37–69. Montrul, S. 2007. Interpreting Mood distinctions in Spanish as a heritage language. In Spanish Contact. Policy, Social and Linguistic Inquiries, K. Potowski & R. Cameron (eds.), 23–40. Amsterdam: John Benjamins. Montrul, S. 2008. Incomplete acquisition in Spanish heritage speakers: Chronological age or interface vulnerability? Proceedings of the 32nd Annual Boston University Conference on Language Development. Somerville MA: Cascadilla. Montrul, S. In press. Heritage language programs. In The Handbook of Second and Foreign Language Teaching, C. Doughty & M. Long (eds.). Oxford: Blackwell. Montrul, S. (under review). Dominant language transfer in Spanish L2 learners and heritage speakers. Special issue of Second Language Research, 2009. Montrul, S. & Bowles, M. In press a. Back to basics: Differential Object Marking under incomplete acquisition in Spanish heritage speakers. Bilingualism: Language and Cognition. Montrul, S. & Bowles, M. In press b. Is grammar instruction beneficial for heritage language learners? Dative case marking in Spanish. The Heritage Language Journal. Montrul, S. & Potowski, K. 2007. Command of gender agreement in school-age Spanish bilingual children. International Journal of Bilingualism 11, 301–328.

292 Incomplete Acquisition in Bilingualism

Montrul, S. & Rodríguez Louro, C. 2006. Beyond the syntax of the null subject parameter. A look at the discourse-pragmatic distribution of null and overt subjects by L2 learners of Spanish. In The Acquisition of Syntax in Romance Languages, L. Escobar & V. Torrens (eds.), 401–418. Amsterdam: John Benjamins. Montrul, S. & Slabakova, R. 2003. Competence similarities between native and near-native speakers. An investigation of the preterite/imperfect contrast in Spanish. Studies in Second Language Acquisition 25, 351–398. Montrul, S., Foote, R., Perpiñán, S., Thornhill, D. & Vidal, S. 2006. Full access and age effects in adult bilingualism: An investigation of accusative clitics and word order. In Selected papers from the Hispanic Linguistics Symposium, N. Sagarra & J. Toribio (eds.), 217–228. Somerville MA: Cascadilla. Montrul, S., Foote, R. & Perpiñan, S. 2008. Gender agreement in adult second language learners and Spanish heritage speakers: The effects of age and context of acquisition. Language Learning 58, 3, 503–553. Mueller Gathercole, V. 2002a. Command of the mass/count distinction in bilingual and monolingual children. An English morphosyntactic distinction. In Language and Literacy in Bilingual Children, D. K. Oller & R. Eilers (eds.), 175–206. Clevedon: Multilingual Matters. Mueller Gathercole, V. 2002b. Grammatical gender in monolingual and bilingual acquisition. A Spanish morphosyntactic distinction. In Language and Literacy in Bilingual Children, D. K. Oller & R. Eilers (eds.), 207–219. Clevedon: Multilingual Matters. Mueller Gathercole, V. 2002c. Monolingual and bilingual acquisition. Learning different treatments of the that-trace phenomena in English and Spanish. In Language and Literacy in Bilingual Children, D. K. Oller & R. Eilers (eds.), 220–254. Clevedon: Multilingual Matters. Mueller Gathercole, V., Sebastián, E. & Soto, P. 1999. The early acquisition of Spanish verbal morphology. Across the board or piecemeal knowledge? International Journal of Bilingualism 3, 133–182. Müller, N. 1994. Gender and number agreement within the DP. In Bilingual First Language Acquisition, German and French, J. Meisel (ed.), 53–88. Amsterdam: John Benjamins. Müller, N. 1998. Transfer in bilingual first language acquisition. Bilingualism: Language and Cognition 1, 151–171. Müller, N. & Hulk, A. 2001. Crosslinguistic influence in bilingual language acquisition. Italian and French as recipient languages. Bilingualism: Language and Cognition 4, 1–21. Neufeld, G. 1977. Language learning ability in adults. A study on the acquisition of prosodic and articulatory features. Working Papers on Bilingualism 12, 45–60. Neufeld, G. 1978. On the acquisition of prosodic and articulatory features in adult language learning. Canadian Modern Language Review 34, 163–174. Neufeld, G. 1979. Towards a theory of language learning ability. Language Learning 29, 227– 241. Neville, H. 1995. Developmental specificity in neurocognitive development in humans. In The Cognitive Neurosciences, M. S. Gazzaniga (ed.), 219–231. Cambridge MA: The MIT Press. Newport, E. 1988. Constraints on learning and their role in language acquisition: Studies of the acquisition of American Sign Language. Language Sciences 10, 147–172. Newport, E. 1990. Maturational constraints on language learning. Cognitive Science 14, 11–28. Newport, E. 1999. Reduced input in the acquisition of signed languages. Contributions to the study of creolization. In Language Creation and Language Change, M. DeGraff (ed.), 161– 178, Cambridge MA: The MIT Press.



References 293

Nicoladis, E. 1999. ‘Where’s my brush-teeth?’ Acquisition of compound nouns in a FrenchEnglish bilingual child. Bilingualism: Language and Cognition 2, 245–256. Nicoladis, E. & Grabois, H. 2002. Learning English and losing Chinese. A case study of a child adopted from China. International Journal of Bilingualism 6, 441–454. Nippold, M. 1998. Later Language Development. The School-age and Adolescent Years. Austin TX: Pro-ed. O’Grady, W. 1997. Syntactic Development. Chicago IL: The University of Chicago Press. O’Grady, W., Lee, M. & Choo, M. 2001. The acquisition of relative clauses by heritage and nonheritage learners of Korean as a second language. A comparative study. Journal of Korean Language Education 12, 283–294. O’Grady, W. (in press). Emergentism. In The Cambridge Encyclopedia of Language Sciences, P. Hogan (ed.), Cambridge: CUP. Oh, J., Jun, S., Knightly, L. & Au, T. 2003. Holding on to childhood language memory. Cognition 86, B53–B64. Oller, D. K. & Eilers, R. E. (eds.). 2002. Language and Literacy in Bilingual Children. Clevedon: Multilingual Matters. Oller, D. K., Eilers, R. E., Urbano, R. & Cobo-Lewis, A. B. 1997. Development of precursors to speech in infants exposed to two languages. Journal of Child Language 27, 407–25. Oller, D. K., Pearson, B. & Cobo-Lewis, A. 2007. Profile effects in early bilingual language and literacy. Applied Psycholinguistics 28, 495–514. Olshtain, E. & Barzilay, M. 1991. Lexical retrieval difficulties in adult language attrition. In First Language Attrition, H. W. Seliger & R. M. Vago (eds.), 139–150. Cambridge: CUP. Oshita, H. 2001. The unaccusative trap in second language acquisition. Studies in Second Language Acquisition 23, 279–304. Oyama, S. 1976. A sensitive period for the acquisition of a non-native phonological system. Journal of Psycholinguistic Research 5, 261–283. Paikeday, T. M. 1985. The Native Speaker is Dead! Toronto: Paikeday. Pallier, C. 2007. Critical periods in language acquisition and language attrition. In Language Attrition. Theoretical Perspectives, Köpke, B., Schmid, M., Keijzer, M. & S. Dostert (eds.), 99–120. Amsterdam: John Benjamins. Pallier, C., Dehaene, S., Poline, J., LeBihan, D., Argenti, M.-A., Dupoux, E. & Mehler, J. 2003. Brain imaging of language plasticity in adopted adults. Can a second language replace the first? Cerebral Cortex 13, 155–161. Pan, B. & Berko-Gleason, J. 1986. The study of language loss, models and hypotheses for an emerging discipline. Applied Psycholinguistics 7, 193–206. Paradis, J. 2001. Do bilingual two-year olds have independent phonological systems? International Journal of Bilingualism 5, 19–38. Paradis, J. & Genesee, F. 1996. Syntactic acquisition in bilingual children. Autonomous or interdependent? Studies in Second Language Acquisition 18, 1–25. Paradis, J. & Genesee, F. 1997. On continuity and the emergence of functional categories in bilingual first language acquisition. Language Acquisition 18, 1–15. Paradis, J. & Navarro, S. 2003. The use of subjects by a Spanish-English bilingual child. Crosslinguistic influence or the influence of the input? The Journal of Child Language 30, 371–393. Paradis, J., Nicoladis, E. & Genesee, F. 2000. Early emergence of structural constraints on codemixing. Evidence from French-English bilingual children. Bilingualism: Language and Cognition 2, 245–261. Paradis, M. (ed.). 1983. Readings on Aphasia in Bilinguals and Polyglots. Quebec: Didier.

294 Incomplete Acquisition in Bilingualism

Paradis, M. 2004. A Neurolinguistic Theory of Bilingualism. Amsterdam: John Benjamins. Paradis, M. 2007. L1 attrition features predicted by a neurolinguistic theory of bilingualism. In Language Attrition. Theoretical Perspectives, Köpke, B., Schmid, M., Keijzer, M. & S. Dostert (eds.), 121–134. Amsterdam: John Benjamins. Park, E. 2006. Grandparents, grandchildren, and heritage language use in Korean. In Heritage Language Development. Focus on East Asian Immigrants, K. Kondo-Brown (ed.), 57–88. Amsterdam: John Benjamins. Patkowski, M. 1990. Age and accent in a second language. A reply to James Emil Flege. Applied Linguistics 11, 73–89. Pavlenko, A. 2004. L2 influence and L1 attrition in adult bilinguals. In First Language Attrition. Interdisciplinary Perspectives on Methodological Issues, M. Schmid, B. Köpke, M. Keijzer & L. Weilemar (eds.), 1–43. Amsterdam: John Benjamins. Pavlenko, A. 2003. ‘I feel clumsy speaking Russian.’ L2 influence on L1 in narratives of Russian L2 users of English. Effects of the Second Language on the First. In Cook, V. (ed.), 32–61. Clevedon: Multilingual Matters. Pavlenko, A. & S. Jarvis. 2002. Bidirectional transfer. Applied Linguistics 23, 190–214. Peal, E. & Lambert, W. 1962. The relation of bilingualism to intelligence. Psychological Monographs, General and Applied 76, 1–23. Pearson, B. 2007. Social factors in childhood bilingualism in the United States. Applied Psycholinguistics 28, 399–410. Penfield, W. 1953. A consideration of the neurophysiological mechanisms of speech and some educational consequences. Proceedings of the American Academy of Arts and Sciences. 82, 199–214. Penfield, W. 1965. Conditioning the uncommitted cortex for language learning. Brain 88, 787– 798. Penfield, W. & Roberts, L. 1959. Speech and Brain Mechanisms. Princeton NJ: Princeton University Press. Pérez-Léroux, A. T. 1998. The acquisition of mood selection in Spanish relative clauses. The Journal of Child Language 25, 585–604. Pfaff, C. 1996. Earlier and later developmental tendencies in language mixing and convergence in bilingual children. Paper presented at the Conference on Processes of Language Change in an Immigrant Context. Tilburg University, The Netherlands. Phillips, C. 1995. Syntax at age two: Cross-linguistic differences. In Papers on Language Processing and Acquisition. C. Schütze, J. Ganger & K. Broihier (eds.), MITWPL 26, 225–282. Pierce, A. 1992. Language Acquisition and Syntactic Theory: Comparative Analysis of French and English Child Grammars. Dordrecht: Kluwer. Pinker, S. 1991. Rules of language. Science 253, 530–535. Pinker, S. 1999. Words and Rules. The Ingredients of Language. New York NY: Basic. Pinker, S. & Ullman, M. 2002a. The past and future of the past tense. TRENDS in Cognitive Science 6, 456–463. Pinker, S. & Ullman, M. 2002b. Combination and structure, not gradedness, is the issue. TRENDS in Cognitive Sciences 6, 472–474. Pires, A. 2005. Verb movement and clitics: Variation and change in Portuguese. In Grammaticalization and Parametric Change, M. Batllori, M. L. Hernanz, C. Picallo & F. Roca (eds.), 48–59. Oxford: OUP. Pires, A. & Rothman, J. In press. Acquisition of Brazilian Portuguese in late childhood: Implications for syntactic theory and language change. In Minimalist Inquiries into Child and



References 295

Adult Language Acquisition: Case Studies Across Portuguese. A. Pires & J. Rothman (eds.). Berlin: Mouton de Gruyter. Poeppel, D. & Wexler, K. 1993. The full competence hypothesis of clausal structure in early German. Language 69, 1–33. Polinsky, M. 1997. American Russian. Language loss meets language acquisition. Proceedings of the Annual Workshop on Formal Approaches to Slavic Linguistics [The Cornell Meeting 1995], 370–406. Ann Arbor MI: Michigan Slavic Publications. Polinsky, M. 2005. Word class distinctions in an incomplete grammar. In Perspectives on Language and Language Development, D. Ravid & H. Bat-Zeev Shyldkrot (eds.), 419–436. Dordrecht: Kluwer. Polinsky, M. 2006. Incomplete Acquisition: American Russian. Journal of Slavic Linguistics 14, 191–262. Polinsky, M. 2008a. Heritage language narratives. In Heritage Language Education. A New Field Emerging, D. Brinton, O. Kagan & S. Bauckus (eds.), 149–164. London: Routledge. Polinsky, M. 2008b. Russian gender under incomplete acquisition. The Heritage Language Journal, 6, 1. Pollock, K. 2005. Early language growth in children adopted from China. Preliminary normative data. Seminars in Speech and Language 26, 22–32. Pollock, K. & Price, J. 2005. Phonological skills of children adopted from China. Implications for assessment. Seminars in Speech and Language 26, 54–63. Poplack, S. 1980. ‘Sometimes I’ll start a sentence in English y termino en español.’ Toward a typology of code switching. Linguistics 18, 581–618. Potowski, K., Jegerski, J. & Morgan-Short, K. In press. The effects of instruction on linguistic development in Spanish heritage language speakers. Language Learning Prasada, S. & Pinker, S. 1993. Generalization of regular and irregular morphological patterns. Language and Cognitive Processes 8, 1–56. Prévost, P. 1997. Truncation in Second Language Acquisition. PhDdissertation, McGill University. Prévost, P. & White, L. 1999. Truncation and missing inflection in adult second language acquisition. In The Acquisition of Syntax, M. A. Friedman & L. Rizzi (eds.), 202–235. London: Longman. Prévost, P. & White, L. 2000. Missing surface inflection or impairment in second language acquisition? Evidence from Tense and Agreement. Second Language Research 16, 110–133. Radford, A. 1990. Syntactic Theory and the Acquisition of English Syntax. Oxford: Blackwell. Radford, A. 2000. Children in search of perfection. Toward a minimalist model of acquisition. Essex Research Reports in Linguistics 34, 57–89. Rizzi, L. 1993/1994. Some notes on linguistic theory and language development: The case for root infinitives. Language Acquisition 3, 371–393. Rodríguez, J., Díaz, R., Duran, D. & Espinosa, L. 1995. The impact of bilingual preschool education on the language development of Spanish-speaking children. Early Childhood Research Quarterly 10, 475–490. Rosen, C. 1984. The interface between semantic roles and initial grammatical relations. In Studies in Relational Grammar. D. Perlmutter & C. Rosen (eds.), 38–77. Chicago IL: University of Chicago Press. Rothman, J. 2007. Heritage speaker competence differences, language change, and input type: Inflected infinitives in heritage Brazilian Portuguese. The International Journal of Bilingualism 11, 359–389.

296 Incomplete Acquisition in Bilingualism

Ruiz, A. 1988. Orientation in language planning. In Language Diversity, Problem or Resource? A Social and Educational Perspective on Language Minorities in the United States, S. L. McKay & S. C. Wong (eds.), 3–25. Boston MA: Heinle & Heinle. Russell, J. & Spada, N. 2006. The effectiveness of corrective feedback for the acquisition of L2 grammar: A meta-analysis of the research. In Synthesizing Research on Language Learning and Teaching, J. M. Norris & L. Ortega (eds.), 133–164. Amsterdam: John Benjamins. Sabourin, L., Stowe, L. & de Haan, G. 2006. Transfer effects in learning a second language grammatical gender system. Second Language Research 22, 1–29. Sankoff, G. 1979. The genesis of a language. In The Genesis of Language, K. Hill (ed.), 23–47. Ann Arbor MI: Karoma. Santilli, M. 1996. Teacher Talk and Written Materials in The Spanish as a Second Language Classroom and Their Importance as Input for Second Language Acquisition. Dissertation Abstracts International: The Humanities and Social Sciences, 57(4). Sanz, C. 1999. What form to focus on? Linguistics, language awareness, and the education of L2 teachers. In Form and Meaning: Multiple Perspectives, J. F. Lee & A. Valdman (eds.), 3–23. Boston MA: Heinle & Heinle. Satz, P., Bakker, D., Tenunissen, J., Goebel, R. & Van der Vlugt, H. 1975. Developmental Patterns of the ear asymmetry: A Multivariate approach. Brain and Language 2, 171–185. Schachter, J. 1974. An error in error analysis. Language Learning 24, 205–213. Schachter, J. 1990. On the issue of completeness in second language acquisition. Second Language Research 6, 93–124. Schachter, J. 1996. Maturation and the issue of Universal Grammar in second language acquisition. In Handbook of Second Language Acquisition, W. Ritchie & T. Bhatia (eds.), 159–193. New York NY: Academic Press. Schlyter, S. 1993. The weaker language in bilingual Swedish-French children. In Progression and Regression in Language. Sociocultural, Neuropsychological and Linguistic Perspectives, K. Hyltenstam & A. Viberg (eds.), 289–308. Cambridge: CUP. Schlyter, S. & Håkansson, G. 1994. Word order in Swedish as the first language, second language and weaker language in bilinguals. Scandinavian Working Papers on Bilingualism 9, 49–66. Schmid, M. 2002. First Language Attrition. Use and Maintenance. The Case of German Jews in Anglophone Countries. Amsterdam: John Benjamins. Schmid, M. 2007. The role of L1 use for L1 attrition. In Language Attrition. Theoretical Perspectives, Köpke, B., Schmid, M., Keijzer, M. & S. Dostert (eds.), 135–154. Amsterdam: John Benjamins. Schmid, M. & Köpke, B. 2004. Language attrition. The next phase. In First Language Attrition. Interdisciplinary Perspectives on Methodological Issues, M. Schmid, B. Köpke, M. Keijzer & L. Weilemar (eds.), 47–60. Amsterdam: John Benjamins. Schoenmakers-Klein Gunnewiek, M. 1998. Taalverlies door taalcontact? Een onderzoek bij Portugese migranten. Tilburg: Tilburg University Press. Schwartz, B. 2003. Child L2 acquisition. Paving the way. Proceedings of the 27th Boston University Conference on Language Development, 26–50. Somerville MA: Cascadilla. Schwartz, B. 2004. Why child L2 acquisition? In Proceedings of Generative Approaches to Language Acquisition 2003 [Lot Occasional Series], J. Van Kampen & S. Baauw (eds.), 47–66. Utrecht: LOT. Schwartz, B. & Sprouse, R. 1996. L2 cognitive states and the full transfer/full access hypothesis. Second Language Research 12, 40–72.



References 297

Scovel, T. 1988. A Time to Speak. A Psycholinguistic Inquiry into the Critical Period for Human Speech. New York NY: Newbury House. Sebastián-Gallés, N. 2006. Native language sensitivities: Evolution in the first year of life. TRENDS in Cognitive Sciences 10, 6, 239–241. Seliger, H. 1978. Implications of a multiple critical periods hypothesis for second language learning. In Second Language Research, Issues and Implications W. Ritchie (ed.), 11–19. New York NY: Academic Press. Seliger, H. 1991. Language attrition, reduced redundancy, and creativity. In First Language Attrition, H. Seliger & R. Vago (eds.), 173–184. Cambridge: CUP. Seliger, H. 1996. Primary language attrition in the context of bilingualism. In Handbook of Second Language Acquisition. W. Ritchie & T. Bhatia (eds.), 605–625. New York NY: Academic Press. Selinker, L. 1972. Interlanguage. International Review of Applied Linguistics 10, 209–231. Selinker, L. & Lakshmanan, U. 1993. Language transfer and fossilization. The multiple effects principle. In Language Transfer in Language Learning, S. Gass & L. Selinker (eds.), 197– 216. Amsterdam: John Benjamins. Serratrice, L., Sorace, A. & Paoli, S. 2004. Crosslinguistic influence at the syntax-pragmatics interface. Subjects and objects in English-Italian bilingual and monolingual acquisition. Bilingualism: Language and Cognition 7, 183–206. Sharwood Smith, M. 1983. On first language loss in the second language acquirer. In Language Transfer in Language Learning, S. Gass & L. Selinker (eds.), 222–231. Rowley MA: Newbury House. Sharwood Smith, M. 1989. Crosslinguistic influence in language loss. In Bilingualism Across the Lifespan, K. Hyltenstam & L. Obler (ed.), 185–201. Cambridge: CUP. Sharwood Smith, M. 2007. Understanding attrition within a MOGUL framework. In Language Attrition. Theoretical Perspectives, Köpke, B., Schmid, M., Keijzer, M. & S. Dostert (eds.), 39–52. Amsterdam: John Benjamins. Sharwood Smith, M. & Van Buren, P. 1991. First language attrition and the parameter setting model. In First Language Attrition, H. Seliger & R. Vago (eds.), 17–30. Cambridge: CUP. Silva-Corvalán, C. 1991. Spanish language attrition in a contact situation with English. In First Language Attrition, H. Seliger & R. Vago (eds.), 151–171. Cambridge: CUP. Silva-Corvalán, C. 1994. Language Contact and Change. Spanish in Los Angeles. Oxford: OUP. Silva-Corvalán, C. 2000. Sociolingüística y pragmática del español. Washington DC: Georgetown University Press. Silva-Corvalán, C. 2003. Linguistic consequences of reduced input in bilingual first language acquisition. In Linguistic Theory and Language Development in Hispanic Languages, S. Montrul & F. Ordóñez (eds.), 375–397. Somerville MA: Cascadilla. Silva-Corvalán, C. 2007. Subjects in early dual language development. A case study of a Spanish-English bilingual child. In K. Potowski & R. Cameron (eds.), Spanish Contact. Policy, Social and Linguistic Inquiries, 3–22. Amsterdam: John Benjamins. Singleton, D. & Lengyel, Z. (eds.). 1995. The Age Factor in Second Language Acquisition. Clevedon: Multilingual Matters. Singleton, D. & Ryan, L. 2005. Language Acquisition. The Age Factor, 2nd edn. Clevedon: Multilingual Matters. Slabakova, R. 2001. Telicity in the Second Language. Amsterdam: John Benjamins. Slabakova, R. 2006. Is there a critical period for semantics? Second Language Research 22, 1–37. Slabakova, R. 2008. Meaning in the Second Language. Berlin: Mouton de Gruyter.

298 Incomplete Acquisition in Bilingualism

Smit, A., Hand, A., Freilinger, J., Bernthal, J. & Bird, A. 1990. The Iowa articulation norms project and its Nebraska replication. Journal of Speech and Hearing Disorders 55, 779–798. Smith, M. 1935. A study of the speech of eight bilingual children of the same family. Child Development 6, 19–25. Smith, C. & Tager-Flusberg, H. 1982. Metalinguistic awareness and language development. Journal of Experimental Child Psychology 34, 449–468. Smith, N. & Tsimpli, I. 1995. The Mind of a Savant. Language Learning and Modularity. Oxford: Blackwell. Snow, C., Burns, S. & Griffin, P. 1998 (eds.). Preventing Reading Difficulties in Young Children. Washington DC: National Academy Press. Snyder, W. 2007. Child Language: The Parametric Approach. Oxford: OUP. Snyder, W., Senghas, A. & Inman, A. 2001. Agreement morphology and the acquisition of N-drop in Spanish. Language Acquisition 9, 157–173. Song, M., O’Grady, W., Cho, S. & Lee, M. 1997. The learning and teaching of Korean in community schools. In Korean Language in America 2, Y.-H. Kim (ed.), 111–127. American Association of Teachers of Korean. Sorace, A. 1993. Incomplete vs. divergent representations of unaccusativity in native and nonnative Italian grammars of Italian. Second Language Research 9, 22–47. Sorace, A. 1999. Initial states, end-states and residual optionality in L2 acquisition. Proceedings of the 23rd Boston University Conference on Language Development, 666–674. Somerville MA: Cascadilla. Sorace, A. 2000a. Syntactic optionality in L2 acquisition. Second Language Research 16, 93–102. Sorace, A. 2000b. Differential effects of attrition in the L1 syntax of near-native L2 speakers. Proceedings of the 24th Boston University Conference on Language Development, 719–725. Somerville MA: Cascadilla. Sorace, A. 2000c. Gradients in auxiliary selection with intransitive verbs. Language 76, 859– 890. Stromswold, K. 1996. Analyzing children’s spontaneous speech. In Methods for Assessing Children’s Syntax, D. McDaniel, C. McKee & H. Smith-Cairns (eds.), 22–53. Sudhalter, V. & Braine, M. 1985. How does comprehension of passives develop? A comparison of actional and experimental verbs. Journal of Child Language 12, 455–70. Swift, M. 2004. Time in Child Inuktitut: A developmental study of an Eskimo-Aleut language. Berlin: Mouton de Gruyter. Tees, R. & Werker, J. 1984. Perceptual flexibility: Maintenance or recovery of the ability to discriminate non-native speech sounds. Canadian Journal of Psychology 38, 579–590. Thomas, M. 1995. Acquisition of the Japanese reflexive zibun and movement of anaphors in Logical Form. Second Language Research 11, 206–234. Thornton, R. 1990. Adventures in Long-Distance Moving. The Acquisition of Complex WhQuestions. PhD dissertation, University of Connecticut, Storrs. Tomasello, M. & Bates, E. (eds.). 2001. Language Development. Malden MA: Blackwell. Toribio, A. J. 2000. Setting parameter limits on dialectal variation in Spanish. Lingua 10, 315– 341. Torrego, E. 1998. The Dependency of Objects. Cambridge MA: The MIT Press. Torrens, V. 2002. La adquicisión del tiempo y la concordancia. La continuidad en el proceso evolutivo. Madrid: Estudios de la Universidad Nacional de Educación a Distancia (UNED).



References 299

Tsimpli, I. 2007. First language attrition from a minimalist perspective: Interface vulnerability and processing effects. In Language Attrition. Theoretical Perspectives, Köpke, B., Schmid, M., Keijzer, M. & S. Dostert (eds.), 83–98. Amsterdam: John Benjamins. Tsimpli, I. & Dimitrakopoulou, M. 2007. The interpretability hypothesis: Evidence from whinterrogatives in second language acquisition. Second Language Language Research 23, 215–242. Tsimpli, I. & Roussou, A. 1991. Parameter resetting in L2? University College of London Working Papers in Linguistics 3, 149–70. Tsimpli, I. & Smith, N. 1991. Second language learning. Evidence from a polyglot savant. University College of London Working Papers in Linguistics 3, 171–84. Tsimpli, I.-M., Sorace, A., Heycock, C. & Filiaci, F. 2004. First language attrition and syntactic subjects. A study of Greek and Italian near-native speakers of English. International Journal of Bilingualism 8, 157–177. Turian, D. & Altenberg, E. 1991. Compensatory strategies of child first language attrition. In First Language Attrition, H. Seliger & R. Vago (eds.), 207–227. Cambridge: CUP. Ullman, M. 2001. The neural basis of lexicon and grammar in first and second language. The declarative/procedural model. Bilingualism: Language and Cognition 4, 105–122. Unsworth, S. 2005. Child L2, Adult L2, Child L1 Differences and Similarities. A Study on the Acquisition of Object Scrambling in Dutch. PhD dissertation, Utrecht Institute of Linguistics OTS, LOT, Netherlands Graduate School of Linguistics. Unsworth, S. 2007. Age and input in early child bilingualism: The acquisition of grammatical gender in Dutch. In Galana 2 – Proceedings of the Conference on Generative Approaches to Language Acquisition North America 2, A. Belikova, L. Meroni & M. Umeda (eds.), 448– 458. Somerville MA: Cascadilla. Unsworth, S. 2008. Age and input in the acquisition of grammatical gender in Dutch. Second Language Research 24, 365–395. Vago, R. 1991. Paradigmatic regularity in first language attrition. In First Language Attrition, H. Seliger & R. Vago (eds.), 241–252. Cambridge: CUP. Valdés, G. 1995. The teaching of minority languages as ‘foreign’ languages: Pedagogical and theoretical challenges. The Modern Language Journal 79, 299–328. Valdés, G. 2000. Introduction. Spanish for Native Speakers [AATSP Professional Development Series Handbook for teachers K-16, Vol.1]. New York NY: Harcourt College. Valdés, G. & Angelelli, C. 2003. Interpreters, interpreting, and the study of bilingualism. Annual Review of Applied Linguistics 23, 58–78. Valdés, G., Fishman, J., Chávez, R. & Pérez, W. 2006. Developing Minority Language Resources: The Case of Spanish in California. Clevedon: Multilingual Matters. Valenzuela, E. 2006. L2 end state grammars and incomplete acquisition of Spanish CLLD constructions. In Inquiries in Language Development. Studies in Honor of Lydia White, R. S­labakova, S. Montrul & P. Prévost (eds.), 283–304. Amsterdam: John Benjamins. Vargha-Khadem, F., Carr, L., Isaacs, E., Brett, E., Abrams, C. & Mishkin, M. 1997. Onset of speech after left hemispherectomy in a nine-year old boy. Brain 120, 159–82. Ventureyra, V. 2005. A la recherche de la langue perdue. Etude psycholinguistique de l’attrition de la première langue chez des coréens adoptés en France. PhD dissertation. Ecole des Hautes Etudes en Sciences Sociales. Ventureyra, V., Pallier, C. & Yoo, H.-Y. 2004. The loss of first language phonetic perception in adopted Koreans. Journal of Neurolinguistics 17, 79–91.

300 Incomplete Acquisition in Bilingualism

Vihman, M. 1998. A developmental perspective on code switching. Conversations between a pair of bilingual siblings. International Journal of Bilingualism 2, 45–84. de Villiers, J. & de Villiers, P. 1974. Competence and performance: Are children really competent to judge? Journal of Child Language 1, 11–22. de Villiers, P. & de Villiers, J. 1972. Early judgments of semantic and syntactic acceptability by children. Journal of Psycholinguistic Research 1, 229–310. Volterra, V. & Taeschner, T. 1978. The acquisition and development of language by bilingual children. The Journal of Child Language 5, 311–326. Waas, M. 1996. Language Attrition Downunder. Frankfurt: Peter Lang. Werker, J. & Tees, R. 1983. Developmental changes across childhood in the perception of nonnative speech sounds. Canadian Journal of Psychology 37, 278–286. Werker, J. & Tees, R. 1984. Cross-language speech perception, evidence for perceptual reorganization during the first year of life. Infant Behavior and Development 7, 49–63. Werker, J. & Yeung, H. 2005. Infant speech perception bootstraps word learning. TRENDS in Cognitive Sciences 9, 11, 519–527. Wexler, K. 1994. Optional infinitives, head movement and the economy of derivations. In Verb Movement, D. Lightfoot & N. Hornstein (eds.), 305–363. Cambridge: CUP. Wexler, K. 1998. Very early parameter setting and the unique checking constraint: A new explanation of the optional infinitive stage. Lingua 106, 23–79. White, L. 1989. Universal Grammar and Second Language Acquisition. Amsterdam: John Benjamins. White, L. 1996. Clitics in L2 French. In Generative Perspectives on Language Acquisition, H. Clahsen (ed.), 335–366, Amsterdam: John Benjamins. White, L. 2003a. Second Language Acquisition and Universal Grammar. Cambridge: CUP. White, L. 2003b. Fossilization in steady state L2 grammars. Persistent problems with inflectional morphology. Bilingualism: Language and Cognition 6, 129–141. White, L. In press. Grammatical theory: Interfaces and L2 knowledge. In Handbook of Second Language Acquisition, W. Ritchie & T. K. Bhatia (eds.). White, L. & Genesee, F. 1996. How native is near-native? The issue of ultimate attainment in adult second language acquisition. Second Language Research 12, 238–265. White, L., Hirakawa, M. & Kawasaki, T. 1996. Effect of instruction on second language acquisition of the Japanese long-distance reflexive zibun. Canadian Journal of Linguistics/Revue canadienne de linguistique 41, 135–154. White, L., Valenzuela, E., Kozlowska-Macgregor, M. & Leung, Y.-K. I. 2004. Gender agreement in nonnative Spanish. Evidence against failed features. Applied Psycholinguistics 25, 105–133. Wolfe-Quintero, K. 1992. Learnability and the acquisition of extraction in relative clauses and wh-questions. Studies in Second Language Acquisition 14, 39–70. Wong-Fillmore, L. 1991. When learning a second language means losing the first. Early Childhood Research Quarterly 6, 323–346. Wright, S., Taylor, D. & Macarthur, J. 2000. Subtractive bilingualism and the survival of the Inuit language: Heritage-versus second-language education. Journal of Educational Psychology 92, 63–84. Yağmur, K. 1997. First Language Attrition among Turkish Speakers in Sydney. Tilburg: Tilburg University Press.



References 301

Yağmur, K. 2004. Issues in finding the appropriate methodology in language attrition research. In First Language Attrition. Interdisciplinary Perspectives on Methodological Issues, M. Schmid, B. Köpke, M. Keijzer & L. Weilemar (eds.), 133–164. Amsterdam: John Benjamins. Yang, C. 2002. Knowledge and Learning in Natural Language. Oxford: OUP. Yeni-Komshian, G. H., Flege, J. E. & Liu, S. 2000. Pronunciation proficiency in the first and second languages of Korean-English bilinguals. Bilingualism: Language and Cognition 3, 131–149. Yip, V. & Matthews, S. 2000. Syntactic transfer in a Cantonese-English bilingual child. Bilingualism: Language and Cognition 3, 263–280. Yuan, B. 1998. Interpretation of binding and orientation of the Chinese reflexive ziji by English and Japanese speakers. Second Language Research 14, 324–340. Zentella, A. C. 1997. Growing up Bilingual. Oxford: Blackwell.

Index of authors

A Aaronson, D.  50, 134, 152, 155, 156, 157, 158, 195, 265, 271, 286 Abrahamsson, N.  10, 43, 52, 264, 267, 269, 277, 285 Abrams, C.  299 Abutalebi, J.  240, 264, 277 Allen, S.  97, 140, 141, 277, 281 Altenberg, E.  71, 109, 110, 163, 258, 265, 277, 299 Ammerlaan, T.  65, 82, 277 Andersen, R.  69, 277 Angelelli, C.  95, 299 Annahatak, B.  140, 281 Argenti, M.-A.  202, 293 Aronoff, M.  105, 109, 110, 163, 258, 265, 286 Arsenian, S.  95, 277 As, A.  204, 277 Au, T.  211, 223, 224, 235, 277, 287, 293 Aveledo, F.  114, 281 Ax, L.  121, 283 B Bakker, D.  267, 296 Balcom, P.  81, 278 Bard, E.  38, 278 Bartke, S.  71, 281 Barzilay, M.  82, 83, 293 Bates, E.  9, 152, 153, 154, 194, 195, 266, 287, 298 Bauckus, S.  9, 165, 208, 240, 280, 288, 295 Bel, A.  114, 278 Belikova, A.  48, 54, 278, 299 Berko-Gleason, J.  69, 70, 278, 293 Bernardini, P.  105, 278 Bernthal, J.  134, 298 Bhatt, R. 2001.  77, 288

Bialystok, E.  11, 47, 49, 50, 53, 95, 208, 235, 258, 274, 278, 281, 282 Bickerton, D.  15, 278 Bird, A.  134, 203, 298 Birdsong, D.  10, 20, 43, 49, 52, 53, 54, 55, 56, 74, 208, 219, 235, 258, 266, 278, 279, 282 Blake, R.  117, 177, 279 Bley-Vroman, R.  6, 19, 20, 43, 44, 45, 98, 124, 209, 262, 279 Boivin, I.  97, 283 Bolonyai, A.  105, 106, 150, 151, 279 Bongaerts, T.  20, 43, 53, 55, 56, 279 Bornstein, M.  10, 279 Bowles, M.  186, 241, 243, 244, 245, 246, 260, 279, 291 Braine, M.  134, 251, 280, 298 Brauth, S.  10, 280 Brett, E.  299 Brinton, D.  9, 165, 208, 240, 280, 288, 295 Brisk, M. E.  133, 291 Brooks, P.  134, 280 Brown, C.  273, 280 Brown, R.  13, 251, 263, 280 Bruhn de Garavito, J.  213, 214, 233, 235, 280 Bullock, B.  67, 280 Burns, S.  134, 203, 298 Burzio, L.  190, 280 Butler, Y. G.  95, 284 C Cagri, I.  237, 280 Cappa, S.  264, 277 Caramazza, A.  70, 280 Cardinaletti, A.  77, 135, 158, 280, 284

Carr, L.  299 Carroll, S.  232, 237, 269, 273, 280 Chan, C.  43, 44, 54, 90, 209, 234, 262, 273, 284 Chávez, R.  165, 299 Chiu, B.  254, 290 Cho, S.  147, 148, 149, 215, 241, 259, 260, 280, 298 Chomsky, N.  9, 30, 44, 75, 76, 89, 145, 280, 281 Choo, M.  214, 215, 293 Clahsen, H.  43, 71, 114, 124, 281, 285, 300 Clyne, M.  65, 68, 265, 279 Cobo-Lewis, A. B.  95, 146, 293 Colombo, J.  10, 281 Cook, V.  22, 68, 90, 278, 281, 294 Coppieters, R.  48, 49, 52, 55, 56, 262, 281 Coppola, M.  16, 287 Corballis, M.  11, 281 Cornips, L.  58, 271, 281, 285 Corrêa, V.  21, 286 Cowan, N.  134, 280 Crago, M.  97, 140, 141, 277, 281 Craik, F.  274, 281 Crain, S.  4, 9, 44, 95, 281 Curtiss, S.  12, 13, 14, 263, 281 Cyrino, S.  21, 286 D Davies, A.  5, 250, 282 de Bot, K.  64, 65, 68, 70, 81, 265, 279, 280, 286 de Haan, G.  31, 257, 296 de Miguel, E.  191, 291 de Villiers, J.  132, 300 de Villiers, P.  132, 300 Dehaene, S.  202, 293

304 Incomplete Acquisition in Bilingualism

DeKeyser, R.  43, 45, 47, 124, 210, 220, 234, 262, 263, 282 Díaz, R.  138, 295 Dimitrakopoulou, M.  44, 209, 262, 299 Doehring, D.  140, 281 Doherty, M.  132, 282 Dooling, R.  10, 280 Döpke, S.  100, 282 Doughty, C.  29, 282, 285, 289, 291 Dupoux  202, 293 Duran, D.  138, 295 E Eichen, E.  14, 59, 263, 290 Eilers, R.  95, 143, 292, 293 Ellis, N.  210, 221, 282 Ellis, R.  211, 234, 282 Epstein, S.  50, 51, 208, 282 Espinosa, L.  138, 295 Eubank, L.  11, 55, 282 F Filiaci, F.  66, 76, 78, 299 Fish, S.  97, 277 Fishman, J.  165, 299 Flege, J.  22, 50, 52, 58, 67, 74, 81, 152, 194, 195, 208, 258, 265, 266, 273, 279, 282, 294, 301 Flynn, S.  50, 51, 208, 282 Foote, R.  232, 234, 235, 237, 240, 259, 282, 292 Footnick, R.  205, 283 Franceschina, F.  112, 232, 257, 283, 285 Freilinger, J.  134, 298 Fromkin, V.  12, 281 Fromm, E.  204, 283 G Gardner, R.  50, 283 Gass, S.  215, 279, 283, 297 Gavruseva, E.  58, 283 Gawlitzek-Maiwald, I.  96, 283 Genesee, F.  17, 54, 55, 56, 95, 96, 97, 100, 123, 208, 277, 283, 288, 290, 293, 300 Geren, J.  121, 283 Gerfen, C.  67, 280 Giorgi, A.  212, 283

Gleitman, L.  14, 283 Glennen, S.  121, 283 Goad, H.  35, 37, 283 Godson, L.  170, 283 Goebel, R.  267, 296 Gombert, J.  132, 283 Gommans, P.  65, 280 Goral, M.  85, 284 Grabois, H.  121, 122, 202, 265, 293 Green, D.  81, 82, 240, 264, 277, 284 Gregg, K. 1999.  11, 55, 282 Griffin, P.  134, 203, 298 Grimshaw, J.  77, 284 Grinstead, J.  114, 118, 252, 284 Grodzinsky, Y.  70, 284 Grosjean, F.  7, 17, 18, 22, 68, 236, 237, 284 Gross, S.  71, 284 Guasti, M. T.  4, 9, 44, 135, 158, 284 Guillelmon, D.  236, 237, 284 Gürel, A.  65, 69, 81, 82, 85, 87, 88, 89, 200, 284 H Håkansson, G.  104, 123, 124, 163, 226, 227, 228, 284, 296 Hakuta, K.  11, 47, 50, 53, 95, 154, 155, 208, 274, 278, 284 Hall, W.  10, 280 Han, Z.  34, 64, 104, 190, 191, 284 Hand, A.  134, 203, 298 Harley, B.  43, 284 Hattori, H.  44, 90, 209, 262, 285 Hawkins, R.  35, 43, 44, 54, 90, 209, 232, 234, 262, 273, 284, 285 Hayes, R.  77, 288 Haznedar, B.  58, 256, 285 Hernández Pina, F.  112, 285 Hernández, A.  112, 152, 153, 194, 195, 266, 285, 287 Herschensohn, J.  7, 43, 285 Heycock, C.  66, 76, 78, 299 Hirakawa, M.  200, 285, 300 Hong, J.-M.  231, 288 Horgan, D.  251, 285

Hudson Kam, C.  273, 285 Hulk, A.  58, 96, 271, 281, 285, 292 Hulsen, M.  65, 83, 84, 163, 166, 187, 188, 189, 192, 193, 285 Hulstijn, J.  210, 221, 285 Hutz, M.  65, 68, 71, 285 Hyams, N.  253, 254, 285 Hyltenstam, K.  10, 11, 43, 52, 262, 264, 267, 269, 277, 285, 290, 296, 297 I Ionin, T.  58, 124, 175, 176, 286 Isaacs, E.  299 Isurin, L.  151, 152, 202, 286 Itard, J.-M.  13, 286 J Jaeggli, O.  145, 286 Jakobson, R.  69, 70, 286 Jarvis, S.  65, 294 Jaspaert, K.  265, 286 Jegerski, J.  241, 245, 295 Jia, G.  50, 134, 152, 155, 156, 157, 158, 195, 265, 266, 267, 268, 271, 286, 290 Jiang, N.  210, 286 Jo, H.-Y.  150, 286 Johnson, J.  19, 47, 48, 49, 52, 53, 54, 56, 57, 156, 273, 286 Jun, S.  223, 224, 277, 287, 293 K Kagan, O.  9, 165, 208, 240, 280, 288, 295 Kang, B-M.  200, 286 Kasuya, H.  101, 286 Kato, M.  21, 286 Kaufman, D.  105, 109, 110, 163, 258, 265, 286 Kawasaki, T.  200, 300 Kayne, R.  230, 286 Kazmi, H.  58, 290 Kegl, J.  16, 287 Kehoe, M.  96, 287 Keijzer, M.  68, 70, 72, 73, 74, 81, 89, 109, 134, 158, 279, 283, 284, 285, 287, 288, 291, 293, 294, 296, 297, 299, 300 Kim, E. J.  157, 287



Kim, G. E.  147, 287 Kim, H.-S.  155, 157, 231, 288 Kim, J.-H.  187, 199, 200, 201, 215, 259, 260, 288, 298 Knightly, L.  223, 224, 225, 226, 277, 287, 293 Koehn, C.  96, 287 Kohnert, K.  152, 153, 154, 194, 195, 266, 287 Kondo-Brown, K.  9, 147, 165, 287, 294 Kong, H.  231, 288 Köpcke, K.  71, 287 Köpke, B.  65, 67, 68, 71, 82, 136, 265, 268, 279, 283, 284, 285, 287, 288, 291, 293, 294, 296, 297, 299, 300 Kozlowska-Macgregor, M.  233, 300 Krakow, R.  121, 288 Krashen, S.  11, 12, 50, 281, 288 Kravin, H.  102, 288 Kroon, S.  265, 286 L Lafond, L.  77, 288 Lakshmanan, U.  34, 41, 58, 98, 288, 297 Lambert, W.  50, 95, 97, 98, 277, 278, 283, 288, 294 Lanza, E.  97, 101, 288 Lardiere, D.  5, 34, 35, 36, 37, 58, 64, 68, 256, 257, 272, 273, 279, 283, 288 Larson-Hall, J.  43, 45, 282 Lasnik, H.  145, 281 Leadholm, B.  132, 288 LeBihan, D.  202, 293 Lee, D.  45, 288 Lee, M.  147, 148, 149, 214, 215, 241, 259, 260, 293, 298 Lee, Y.-G.  231, 288 Lengyel, Z.  20, 43, 297 Lenneberg, E.  11, 267, 289 Leung, Y.-K. I.  233, 300 Lightbown, P.  29, 289 Lightfoot, D.  66, 67, 270, 289, 300 Lillo-Martin, D.  4, 95, 281 Lin, J.  35, 289 Lipski, J.  163, 172, 289

Index of authors 305

Liszka, S.  35, 285 Litvak, A.  152, 290 Liu, S.  50, 52, 58, 152, 194, 195, 265, 266, 282, 301 Lleó, C.  96, 112, 287, 289 Loban, W.  132, 289 Lock, E.  290 Long, M.  5, 11, 19, 34, 43, 47, 52, 55, 57, 64, 231, 242, 262, 263, 273, 282, 285, 286, 288, 289, 291, 298, 300 López Ornat, S.  112, 114, 289 Luján, M.  186, 289 Luo, S. H.  157, 289 Lynch, A.  179, 180, 289 M Macarthur, J.  140, 141, 300 Mack, M.  5, 68, 282, 289 MacKay, I.  52, 58, 152, 282 Macnamara, J.  95, 289 MacWhinney, B.  273, 289 Major, R.  71, 81, 89, 289 Marcus, G.  71, 281 Marinova-Todd, S.  58, 289 Martin, M.  95, 278, 281 Martohardjono, G.  50, 51, 208, 282 Mason, M.  13, 263, 289 Matsunaga, S.  232, 290 Matthews, S.  301 Maxfield, T.  254, 290 Mayberry, R.  14, 58, 59, 263, 290 Mayer, M.  83, 116, 290 Mc Carthy, C.  174, 290 McDaniel, D.  254, 290, 298 McDonald, J.  58, 290 McElree, B.  152, 290 McWhorter, J.  6, 290 Mehler, J.  202, 293 Meisel, J.  17, 43, 95, 96, 97, 99, 104, 123, 124, 125, 126, 127, 128, 209, 214, 228, 252, 253, 262, 287, 290, 291, 292 Menyuk, P.  133, 291 Merino, B.  141, 142, 143, 144, 179, 259, 291 Miller, B.  50, 132, 278, 288 Miller, J.  50, 132, 278, 288 Mishkin, M.  299

Molis, M.  53, 54, 55, 56, 279 Montrul, S.  31, 55, 56, 77, 114, 128, 144, 145, 146, 164, 172, 180, 181, 183, 184, 185, 186, 191, 192, 193, 195, 196, 197, 199, 200, 201, 211, 212, 213, 215, 228, 229, 230, 231, 232, 233, 234, 235, 237, 238, 239, 240, 241, 243, 244, 245, 246, 259, 260, 265, 268, 273, 279, 287, 291, 292, 297, 299 Morgan-Short, K.  241, 245, 295 Mueller Gathercole, V.  114, 143, 144, 145, 146, 172, 292 Müller, N.  96, 285, 292 Munro, M.  52, 58, 152, 282 Muysken, P.  43, 124, 281 N Navarro, S.  118, 119, 258, 260, 271, 293 Neufeld, G.  52, 292 Neville, H.  14, 292 Newport, E.  14, 15, 16, 19, 47, 48, 49, 52, 53, 54, 56, 57, 59, 156, 262, 263, 273, 274, 275, 283, 285, 286, 292 Nicoladis, E.  96, 97, 100, 121, 122, 202, 265, 283, 293 Nippold, M.  132, 293 O Oh, J.  223, 224, 277, 287, 293 Oller, D. K.  95, 143, 146, 292, 293 Olshtain, E.  82, 83, 293 Oshita, H.  256, 293 Oyama, S.  10, 46, 47, 52, 56, 293 P Paikeday, T. M.  5, 293 Pallier, C.  202, 264, 268, 293, 299 Pan, B.  69, 293 Paoli, S.  119, 260, 297 Paradis, J.  96, 97, 100, 118, 119, 123, 258, 260, 271, 283, 293 Paradis, M.  45, 46, 69, 81, 82, 91, 124, 205, 210, 234, 235, 263, 288, 293, 294 Park, E.  150, 294

306 Incomplete Acquisition in Bilingualism

Parodi, C.  186, 285, 289 Patkowski, M.  47, 52, 294 Pavlenko, A.  64, 65, 66, 294 Peal, E.  95, 294 Pearson, B.  146, 147, 293, 294 Penfield, W.  10, 11, 42, 43, 294 Perani, D.  264, 277 Pérez, W.  117, 165, 280, 289, 294, 299 Pérez-Léroux, A. T.  117, 294 Perner, J.  132, 282 Perpiñán, S.  232, 234, 235, 237, 259, 292 Pfaff, C.  105, 294 Phillips, C.  253, 294 Pianesi, F.  212, 283 Pierce, A.  253, 294 Pinker, S.  9, 45, 74, 75, 294, 295 Pires, A.  21, 108, 135, 158, 218, 286, 294, 295 Poeppel, D.  253, 295 Poline, J.  202, 293 Polinsky, M.  67, 163, 164, 166, 167, 170, 171, 172, 173, 174, 175, 176, 182, 183, 189, 195, 199, 259, 295 Pollock, K.  121, 295 Poplack, S.  96, 295 Potowski, K.  144, 145, 146, 172, 241, 245, 246, 291, 295, 297 Prasada, S.  74, 295 Presco, D.  277 Prévost, P.  58, 74, 208, 256, 285, 295, 299 Price, J.  121, 295 R Radford, A.  105, 253, 295 Rakow, M.  96, 287 Rigler, D.  12, 281 Rigler, M.  12, 281 Rizzi, L.  253, 295 Roberts, J.  121, 288 Roberts, L.  10, 11, 294 Robertson, D.  38, 278 Rocca, I.  114, 281 Rodríguez Louro, C.  77, 228, 229, 292 Rodríguez, J.  77, 138, 139, 228, 229, 292, 295 Rosen, C.  190, 295

Rossing, C.  65, 280 Rothman, J.  21, 108, 135, 158, 218, 286, 294, 295 Rothweiler, M.  71, 281 Roussou, A.  44, 262, 299 Ruiz, A.  95, 296 Russell, J.  29, 296 Ryan, E.  235, 278 Ryan, L.  43, 297 S Sabourin, L.  31, 257, 296 Sameck-Lodovici, V.  77, 284 Samuels, M.  134, 280 Sankoff, G.  15, 296 Santilli, M.  271, 296 Sanz, C.  271, 296 Satz, P.  267, 296 Schachter, J.  19, 20, 37, 43, 45, 124, 209, 210, 255, 262, 267, 279, 288, 296 Schirru, C.  282 Schlyter, S.  102, 103, 104, 105, 123, 124, 125, 126, 128, 208, 220, 249, 250, 258, 259, 278, 296 Schmid, M.  65, 68, 71, 82, 89, 136, 158, 268, 272, 279, 283, 284, 285, 288, 291, 293, 294, 296, 297, 299, 300 Schwartz, B.  30, 50, 58, 98, 124, 125, 208, 256, 273, 285, 296 Scovel, T.  46, 52, 297 Sebastián, E.  11, 114, 292, 297 Sebastián-Gallés, N.  11, 297 Seliger, H.  11, 64, 66, 68, 166, 277, 280, 286, 293, 297, 299 Selinker, L.  5, 34, 41, 297 Senghas, A.  16, 112, 287, 298 Serratrice, L.  119, 260, 297 Shapero, D.  95, 278 Sharwood Smith, M.  64, 67, 75, 283, 297 Silva-Corvalán, C.  109, 111, 115, 116, 117, 119, 126, 163, 166, 176, 177, 178, 179, 180, 183, 186, 195, 265, 297 Singleton, D.  20, 43, 297 Slabakova, R.  32, 33, 55, 56, 211, 292, 297, 299 Smit, A.  134, 203, 298

Smith, C.  132, 298 Smith, M.  43, 44, 64, 90, 95, 298, 299 Smith, N.  298, 299 Snedeker, J.  121, 283 Snow, C.  134, 203, 298 Snyder, W.  4, 112, 251, 254, 298 Song, M.  147, 148, 149, 241, 242, 243, 259, 260, 298 Sorace, A.  20, 37, 38, 39, 49, 51, 66, 69, 75, 76, 77, 78, 80, 90, 117, 119, 182, 183, 190, 191, 192, 253, 260, 278, 283, 285, 297, 298, 299 Soto, P.  114, 292 Spada, N.  29, 296 Sprouse, R.  30, 50, 208, 273, 296 Squires, B.  58, 290 Starke, M.  77, 280 Steele, J.  37, 283 Stowe, L.  31, 257, 296 Stromswold, K.  251, 298 Sudhalter, V.  251, 298 Sung, H.  258, 282 Swift, M.  140, 298 T Taeschner, T.  95, 96, 300 Tager-Flusberg, H.  132, 298 Tamis-LeMonda, S.  134, 280 Tao, S.  121, 288 Taylor, D.  140, 141, 281, 300 Tees, R.  223, 298, 300 Tenunissen, J.  267, 296 Thomas, M.  200, 298 Thornhill, D.  237, 292 Thornton, R.  9, 44, 145, 281, 298 Tomasello, M.  9, 298 Toribio, A.  67, 119, 280, 292, 298 Torrego, E.  184, 298 Torrens, V.  114, 292, 298 Tracy, R.  96, 283 Tsimpli, I.  43, 44, 66, 69, 75, 76, 78, 79, 80, 90, 117, 209, 262, 298, 299 Tsukada, K.  258, 282 Turian, D.  109, 110, 163, 258, 265, 299



U Ullman, M.  45, 74, 91, 95, 210, 234, 263, 268, 294, 299 Unsworth, S.  58, 98, 124, 258, 285, 299 Urbano, R.  95, 293 V Vago, R.  163, 168, 169, 170, 173, 277, 280, 286, 293, 297, 299 Vaid, J.  267, 285 Valdés, G.  95, 163, 165, 223, 299 Valenzuela, E.  39, 40, 41, 233, 299, 300 Van Buren, P.  64, 297 Van der Vlugt, H.  267, 296 Vargha-Khadem, F.  17, 299 Ventureyra, V.  204, 299 Vidal, S.  237, 292 Vihman, M.  97, 299

Index of authors 307

Volterra, V.  95, 96, 300 W Waas, M.  82, 300 Wang, W.  43, 284 Weltens, B.  91, 280 Werker, J.  11, 223, 298, 300 Wexler, K.  58, 124, 253, 286, 295, 300 White, L.  30, 34, 35, 36, 37, 44, 48, 50, 51, 54, 55, 56, 58, 74, 77, 125, 128, 167, 200, 208, 233, 256, 257, 278, 280, 281, 283, 295, 299, 300 Wiese, R.  71, 281 Wiseman, R.  157, 289 Witt, D.  95, 284 Woest, A.  71, 281 Wolfe-Quintero, K.  215, 300

Wong-Fillmore, L.  137, 138, 139, 300 Wright, S.  140, 141, 300 Wu, Y.  152, 286 Y Yağmur, K.  65, 71, 82, 300 Yang, C.  99, 254, 255, 259, 270, 300 Yeni-Komshian, G.  50, 58, 152, 194, 195, 265, 266, 282, 301 Yeung, H.  11, 300 Yip, V.  96, 301 Yoo, H.-Y.  203, 299 Yoon, J.  199, 200, 201, 215, 259, 260, 287 Yuan, B.  200, 301 Z Zentella, A. C.  179, 301 Zurif, E.  70, 280

Index of terms

A A-personal  185, 243–245 Additive bilingualism  97 (cf. Subtractive bilingualism) Adolescents, teenagers  72–75, 132–35, 154, 155, 158, 176 Adoptee(s)  121–123, 127,152, 202–205, 241, 263, 265, 268 Age effect(s)  1, 7 In L1 acquisition of ASL  14 In L2 acquisition  43, 49, 50, 52, 54, 57, 58, 59 In L1 attrition  60, 63, 98, 189, 194, 208, 209, 260, 262–268 Age of arrival (also AoA)  46, 47, 48, 51, 52, 56, 57, 83, 87, 158 Agreement Gender agreement in Dutch  31, 34, 58 in Spanish  111, 112, 144, 172, 225, 226, 232, 233, 234 in Russian  174, 175 in Swedish  228 Verbal agreement in English  34, 35 in Turkish  36 in German  96 in Spanish  113–117, 225, 226, 229, 230 in Russian  182, 183 Adjectival agreement in Dutch  73 Agreement features  77, 128 Possessive agreement in Hungarian  150, 151 American Sign Language (also ASL)  14, 59

Anaphor(s)  187, 199, 200, 201, 206, 215, 242, 259 Arabic  5, 166, 206, 255 ASL  14, 16, 58, 59, 263 Aspect and telicity in Bulgarian  32, 33 grammatical aspect in Spanish (preterite and imperfect)  113–117, 177, 179, 196–199, 212 in Russian  174 B Baseline group (see also control group)  68, 109, 116, 137, 166, 168, 173 Bilingual development  283, 288 Bilingual education  138, 164, 219 Bilingual first language acquisition (also simultaneous bilingualism)  75, 94, 95 Bilingual schools  97, 137, 144, 146, 151, 176 Brazilian Portuguese  5, 21, 67, 135, 218 Bulgarian  32, 33 C Caribbean Spanish  119 Case in Korean  147–149, 214–215, 242 in Russian  171, 172 in Spanish  184–187, 243 Child L2 acquisition  17, 58, 97, 98, 125, 158, 257, 258, 263, 296

Chinese (learners, speakers)  34, 35, 36, 56, 121, 122, 147, 155–157, 187, 202, 227, 232, 254–257 Classroom instruction  28 Classroom (research)  240–242, 245 Clitics in Italian   38–41 in Spanish  184–186, 230, 237–240 Clitic climbing  38, 39, 230 Clitic left dislocations  39, 40, 41, 238 Code switching  96, 97, 101, 105–107 College age (children)  153, 157 Competence (i.e., linguistic competence)  21, 22, 66–69, 75, 76 near-native competence  37, 48, 49, 54, 55 bilingual competence  96, 97, 99, 102, 105–107, 121, 136, 138, 155, 156, 161–164, 205, 216 native competence  210, 211, 249, 250 L2 and heritage language competence  218–222, 235 Complementizer  96, 105, 145, 146, 176 Comprehension  69, 70, 80, 82, 84, 141, 142, 154, 173, 181, 182, 189, 214, 215, 221, 233, 236, 237, 241, 252 Conditional  113, 116, 141, 142, 176, 177, 180 Control group (see also baseline group)  68, 72

310 Incomplete Acquisition in Bilingualism

Count-mass (distinction)  32, 33, 127, 143, 144, 171 Creoles  15, 16, 278 Critical periods  10–12, 59, 249 Crystallization hypothesis  202, 203, 268 Crosslinguistic influence  64, 66, 75, 81, 86, 119 D Deaf  12–16, 58, 59, 263 Declarative/Procedural Model  268 declarative (see also implicit knowledge)  45, 48, 65, 74, 75, 91, 210, 268 procedural (see also explicit knowledge)  45, 65, 74, 91, 95, 210, 220 Definiteness  36, 184, 227 Diachronic change  67, 68 Differential object marking (see also A personal)  184, 186, 206, 244–246, 260 Divergence  34, 49, 51, 68, 170 Dual immersion  52, 58, 63, 97, 136, 137, 143–146, 157 Dutch  31, 32, 53, 56, 58, 72–75, 83–85, 89, 134, 187–189, 192, 257, 271 E Elementary school(s)  133, 136, 143, 147, 162, 163, 224 English  31–40, 47, 48, 52–56, 74, 77, 78, 81, 84, 106, 134, 141–147, 153, 189, 190, 251–255 Explicit knowledge  210, 232, 234, 235 F Failed Functional Features Hypothesis  44, 209, 234 Formal features  34, 35, 39, 75–77, 89, 90, 128, 182, 233, 234 Interpretable features  77, 80, 90, 128, 261

Uninterpretable features  77, 89, 90, 128, 139, 182, 232, 261 Forgetting  67, 81, 152 French  48, 52, 56, 99, 102–105, 135, 141, 202–204, 236, 253 Frequency effects  82, 85, 228 Full Access Hypothesis  30, 32, 51 Full Transfer/Full Access Hypothesis (see also L1 transfer)  30, 32, 51 Functional categories  15, 35, 44, 96 Fundamental Difference Hypothesis  43, 125, 209, 210, 220–222, 248 G Gender agreement (see agreement) Grammaticality judgment task  40, 48, 56, 72, 74, 155, 191, 192, 214, 225, 230, 235, 237, 238, 243, 244 Greek  78–80, 117, 183 H Hebrew  109, 110, 168, 170 Heritage language  120, 136– 139, 147, 149, 150, 157, 161–67, 174, 187, 193, 196, 201, 205, 206, 208, 210, 216–220, 222, 224, 228, 231, 232, 234, 240, 241, 244–248 Heritage language learner  157, 165, 208, 220, 222, 231, 232, 234, 240, 244–248 Heritage language school  147 Heritage speaker(s) (also immigrant children)  21–24, 27, 136, 161–167, 170–77, 199–201, 211–222, 237–244 High School  75, 155 Hypnosis  204, 205 Home language  20, 116, 137, 162

Hungarian  100, 106, 107, 150, 151, 168, 169, 170, 173 I Identity  22, 65, 106, 147, 286 Immersion (see Dual immersion) Immigrant children  20, 97, 120, 136–138, 152, 154, 155, 162, 249 Immigrants (L2 adults)  46, 47, 52, 53, 56, 71, 72, 89 children of immigrants  94, 143, 162, 163, 166–168, 170, 188, 193, 200 (and immigrant children)  157 Impairment  44, 256 Implicit knowledge  91, 220, 235 Incomplete representations  39 Incompleteness Hypothesis  20, 37, 209, 210, 262 Infants  121 Inflectional morphology (see also some entries under Agreement)  35, 37, 66, 71, 72, 102, 103, 109, 111, 123, 126, 141, 146, 150, 162, 167, 171, 177, 206, 216, 241, 247, 251, 258, 272 Initial state (in L2 acquisition)  30, 51, 97, 208 Interlanguage  20, 31–37, 41, 243, 256 Interpretability Hypothesis  44, 209 Inuktitut  140, 141 Irregular morphology  74, 75, 91 Italian  37–39, 46–49, 52, 56, 58, 76–80, 119,120, 135 J Japanese  48, 52, 95, 101, 137, 147, 157, 178, 200, 203, 232, 255, 256 K Korean  147–150, 157, 158, 187, 194, 195, 199–206, 214, 215, 222–224, 231, 241, 242



L L1 acquisition  12–14, 17, 19–22, 27–29, 41–44, 46, 49, 57, 58, 59, 69, 98, 124, 134–136, 163–166, 193, 209, 216, 217, 249–255, 269, 270, 275 L1 transfer  32, 55, 165, 200, 257 (see also Full Transfer/Full Access Hypothesis) prosodic transfer  35, 37 L2 effects  80, 162, 167 Language at home  42, 97, 101, 143, 193 Language change  6, 21, 67 Language choice  100, 155, 217 Language loss  59–61, 64–69, 85, 122, 138, 139, 141–143, 146, 147, 149–159, 163, 262–269 Language maintenance  91, 136, 143, 144, 155, 157, 193 Language processing  60, 202, 263, 269, 273 Language retention  202 Language shift  64, 121, 122, 136, 152–155 Language socialization  147 Length of exposure  193 Length of residence  47, 53, 200 Less is More Hypothesis  273, 275 Lexical access  89, 153, 154, 260 Lexical categories  15 Lexical-semantics  37, 41, 81, 247 Local binding (see also Longdistance binding and Principle B)  200 Logical problem (of language acquisition)  30, 43, 44 Long-distance binding (see also local binding and Principle B)  200, 201, 206 Longitudinal data  93, 116, 164, 166, 251 M Majority language  20, 94, 97, 99, 100, 101, 103, 117, 125, 137,

Index of terms 311

139, 140, 141, 151, 162–164, 201, 216–218 Markedness  70 Memory Implicit/procedural memory system  45, 95, 210, 220 Associative memory  74 and forgetting 65, 67, 75 81, 82, 91, 204, 205 Metalinguistic knowledge/ awareness/skills  75, 95, 132, 133, 134, 217, 219, 221, 226, 234, 237, 240, 244, 247 Minority language  18, 20–22, 93, 94, 98–103, 121, 123, 136, 138–141, 146, 147, 150–152, 158–165 Missing Surface Inflection Hypothesis  74, 234 Modularity  25, 66 N Nicaraguan Sign Language  16 Norwegian  101 Null Subject Parameter  67, 145, 183, 228, 230 Null subject(s)  67, 76–80, 87, 88, 108, 118–120, 145, 162, 182–184, 206, 229, 230, 252, 254 O Onset of bilingualism  22, 24, 63–65, 68, 91, 115, 144, 152, 158, 161, 165, 166, 179, 193, 195, 196, 198, 199, 205, 208, 211, 265 Oral skills  226 P Parameters  15, 17, 18, 30, 44, 76, 89, 96, 99, 124, 125, 167, 186, 204, 209, 214, 220, 228, 230, 250 Partial Access to Universal Grammar  30 Past tense  34, 35, 36, 142, 172, 177, 253, 257 imperfective  113, 116, 175, 176, 177

Performance  5, 14, 19, 34, 35, 36, 48, 49, 52, 53, 54, 55, 57, 59, 64, 65, 69, 73, 76, 90, 93, 134, 141, 142, 143, 146, 155, 161, 182, 205, 213, 215, 216, 221, 226, 232, 234, 235, 240, 241, 242, 243, 245, 251, 252 Phoneme perception task  223 Phonology  25, 29, 35, 37, 41, 43, 46, 49, 56, 57, 66, 70, 71, 77, 81, 89, 96, 134, 167, 168, 182, 187, 205, 218, 220, 223, 224, 225, 247, 248, 255, 256, 260 Picture naming task  84, 85, 233, 234 Picture-matching task  83, 84, 188, 189 Pidgin  15, 16 Plural  32, 71, 72, 82, 112, 113, 134, 168, 175, 190, 192, 230, 231 Positive evidence  29, 278 Postverbal subjects  78, 79, 118, 192, 229, 230, 231 Pragmatic development  133 Pragmatics  4, 19, 66, 70, 80, 117, 150, 158, 159, 167, 182, 183, 225, 241 Pre-school children  101, 115, 139, 140, 218 preterite  33, 55, 113, 114, 116, 177, 179, 196, 197, 198, 199, 211, 212, 213, 260 Preverbal subjects  79 Principle  86, 145 Principle B  86 pro-drop  77, 86, 108, 117, 119, 220, 228, 254, 261 Pronunciation  34, 46, 47, 51–53, 56–58, 66, 67, 70, 71, 89, 163, 167, 194, 195, 225, 232, 241, 250, 258, 265 Puberty  28, 40, 46–50, 57, 60, 88, 91, 98, 164, 165, 193, 207, 209, 214, 217, 219, 233, 236, 255, 263, 267, 271 R Reanalysis  24, 53, 146, 162, 201

312 Incomplete Acquisition in Bilingualism

Relative clauses  71, 108, 133, 135, 141, 142, 146, 152, 178, 180, 214, 215, 218, 220, 255, 256, 259 Representational Deficit Hypothesis  44, 209 Restructuring  32, 38, 44, 66, 74, 80, 85, 170 Russian  3, 110, 151, 152, 161, 163, 171–176, 182, 183, 186, 202, 206, 220, 259, 261 S Simplification  6, 24, 66, 70, 115, 168, 170, 176, 177, 180, 182, 187, 201, 206, 259 South Asian languages  166 Spanish  31–33, 39–41, 54–56, 77, 97, 108, 111–119, 132, 136– 139, 141–155, 163, 166, 171–186, 188, 190–193, 195–197, 199, 206, 208, 211–214, 219, 220, 222, 224, 226, 228–238, 240, 243–246 subjunctive (Spanish)  113, 114, 116, 117, 141, 142, 175–181 Subtractive bilingualism  98

Swedish  102–105, 123–125, 203, 226–228 T Telic vs. atelic (see also aspect)  35 atelic  32, 33, 175 That-t effect  143, 144, 145, 146, 152 Triggers  270 Truth value judgment task(s)  55, 88, 196, 197, 198, 200, 211, 213, 215 Turkish  31, 36, 37, 71, 85–88, 100 U Ultimate attainment  14,18, 27, 34, 37, 43, 44, 49, 51, 52, 55, 57, 63, 64, 98, 161, 165, 207, 211, 241, 273 Unaccusativity  39, 190–192, 230, 231 unaccusative verbs  37, 38, 49, 171, 190–192, 230, 231

unergative verbs  38, 190–192, 230, 231 Unitary Linguistic System Hypothesis  95, 96 Universal Grammar  43–45, 47, 48, 50, 51, 54, 55, 75, 85, 95, 96, 98, 124–128, 167, 208–210, 220, 250, 255, 259 V Verb Second (V2) (see also some entries under word order)  72–74, 99, 104, 105, 125, 134, 227, 228, 230, 261, 270 VOT (voice onset time)  67, 81, 89, 204, 223 W Western Armenian  170, 283 Wh-movement  48, 54–56, 135 Word order  78, 96, 102–105, 108, 118, 125, 147, 183, 227, 237, 238, 240, 252 Wug test  72, 73, 75

In the series Studies in Bilingualism (SiBil) the following titles have been published thus far or are scheduled for publication: 39 Montrul, Silvina A.: Incomplete Acquisition in Bilingualism. Re-examining the Age Factor. 2008. x, 312 pp. 38 Plaza-Pust, Carolina and Esperanza Morales-López (eds.): Sign Bilingualism. Language development, interaction, and maintenance in sign language contact situations. xvi, 379 pp. + index. Expected September 2008 37 Niño-Murcia, Mercedes and Jason Rothman (eds.): Bilingualism and Identity. Spanish at the crossroads with other languages. 2008. vii, 365 pp. 36 Hansen Edwards, Jette G. and Mary L. Zampini (eds.): Phonology and Second Language Acquisition. 2008. vi, 380 pp. 35 Rocca, Sonia: Child Second Language Acquisition. A bi-directional study of English and Italian tenseaspect morphology. 2007. xvi, 240 pp. 34 Koven, Michèle: Selves in Two Languages. Bilinguals' verbal enactments of identity in French and Portuguese. 2007. xi, 327 pp. 33 Köpke, Barbara, Monika S. Schmid, Merel Keijzer and Susan Dostert (eds.): Language Attrition. Theoretical perspectives. 2007. viii, 258 pp. 32 Kondo-Brown, Kimi (ed.): Heritage Language Development. Focus on East Asian Immigrants. 2006. x, 282 pp. 31 Baptista, Barbara O. and Michael Alan Watkins (eds.): English with a Latin Beat. Studies in Portuguese/Spanish – English Interphonology. 2006. vi, 214 pp. 30 Pienemann, Manfred (ed.): Cross-Linguistic Aspects of Processability Theory. 2005. xiv, 303 pp. 29 Ayoun, Dalila and M. Rafael Salaberry (eds.): Tense and Aspect in Romance Languages. Theoretical and applied perspectives. 2005. x, 318 pp. 28 Schmid, Monika S., Barbara Köpke, Merel Keijzer and Lina Weilemar (eds.): First Language Attrition. Interdisciplinary perspectives on methodological issues. 2004. x, 378 pp. 27 Callahan, Laura: Spanish/English Codeswitching in a Written Corpus. 2004. viii, 183 pp. 26 Dimroth, Christine and Marianne Starren (eds.): Information Structure and the Dynamics of Language Acquisition. 2003. vi, 361 pp. 25 Piller, Ingrid: Bilingual Couples Talk. The discursive construction of hybridity. 2002. xii, 315 pp. 24 Schmid, Monika S.: First Language Attrition, Use and Maintenance. The case of German Jews in anglophone countries. 2002. xiv, 259 pp. (incl. CD-rom). 23 Verhoeven, Ludo and Sven Strömqvist (eds.): Narrative Development in a Multilingual Context. 2001. viii, 431 pp. 22 Salaberry, M. Rafael: The Development of Past Tense Morphology in L2 Spanish. 2001. xii, 211 pp. 21 Döpke, Susanne (ed.): Cross-Linguistic Structures in Simultaneous Bilingualism. 2001. x, 258 pp. 20 Poulisse, Nanda: Slips of the Tongue. Speech errors in first and second language production. 1999. xvi, 257 pp. 19 Amara, Muhammad Hasan: Politics and Sociolinguistic Reflexes. Palestinian border villages. 1999. xx, 261 pp. 18 Paradis, Michel: A Neurolinguistic Theory of Bilingualism. 2004. viii, 299 pp. 17 Ellis, Rod: Learning a Second Language through Interaction. 1999. x, 285 pp. 16 Huebner, Thom and Kathryn A. Davis (eds.): Sociopolitical Perspectives on Language Policy and Planning in the USA. With the assistance of Joseph Lo Bianco. 1999. xvi, 365 pp. 15 Pienemann, Manfred: Language Processing and Second Language Development. Processability theory. 1998. xviii, 367 pp. 14 Young, Richard and Agnes Weiyun He (eds.): Talking and Testing. Discourse approaches to the assessment of oral proficiency. 1998. x, 395 pp. 13 Holloway, Charles E.: Dialect Death. The case of Brule Spanish. 1997. x, 220 pp. 12 Halmari, Helena: Government and Codeswitching. Explaining American Finnish. 1997. xvi, 276 pp. 11 Becker, Angelika and Mary Carroll: The Acquisition of Spatial Relations in a Second Language. In cooperation with Jorge Giacobbe, Clive Perdue and Rémi Porquiez. 1997. xii, 212 pp. 10 Bayley, Robert and Dennis R. Preston (eds.): Second Language Acquisition and Linguistic Variation. 1996. xix, 317 pp.

9 8 7

6 5 4 3 2 1

Freed, Barbara F. (ed.): Second Language Acquisition in a Study Abroad Context. 1995. xiv, 345 pp. Davis, Kathryn A.: Language Planning in Multilingual Contexts. Policies, communities, and schools in Luxembourg. 1994. xix, 220 pp. Dietrich, Rainer, Wolfgang Klein and Colette Noyau: The Acquisition of Temporality in a Second Language. In cooperation with Josée Coenen, Beatriz Dorriots, Korrie van Helvert, Henriette Hendriks, Et-Tayeb Houdaïfa, Clive Perdue, Sören Sjöström, Marie-Thérèse Vasseur and Kaarlo Voionmaa. 1995. xii, 288 pp. Schreuder, Robert and Bert Weltens (eds.): The Bilingual Lexicon. 1993. viii, 307 pp. Klein, Wolfgang and Clive Perdue: Utterance Structure. Developing grammars again. In cooperation with Mary Carroll, Josée Coenen, José Deulofeu, Thom Huebner and Anne Trévise. 1992. xvi, 354 pp. Paulston, Christina Bratt: Linguistic Minorities in Multilingual Settings. Implications for language policies. 1994. xi, 136 pp. Döpke, Susanne: One Parent – One Language. An interactional approach. 1992. xviii, 213 pp. Bot, Kees de, Ralph B. Ginsberg and Claire Kramsch (eds.): Foreign Language Research in CrossCultural Perspective. 1991. xii, 275 pp. Fase, Willem, Koen Jaspaert and Sjaak Kroon (eds.): Maintenance and Loss of Minority Languages. 1992. xii, 403 pp.