Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing [1 ed.] 9781617615672, 9781607417101

A bilingual person, in a broad definition, is one who can communicate in more than one language, be it actively (through

221 32 3MB

English Pages 315 Year 2010

Report DMCA / Copyright

DOWNLOAD PDF FILE

Recommend Papers

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing [1 ed.]
 9781617615672, 9781607417101

  • 0 0 0
  • Like this paper and download? You can publish your own PDF file online for free in a few minutes! Sign Up
File loading please wait...
Citation preview

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved. Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved. Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

LANGUAGES AND LINGUISTICS SERIES

BILINGUALS: COGNITION, EDUCATION AND LANGUAGE PROCESSING

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

No part of this digital document may be reproduced, stored in a retrieval system or transmitted in any form or by any means. The publisher has taken reasonable care in the preparation of this digital document, but makes no expressed or implied warranty of any kind and assumes no responsibility for any errors or omissions. No liability is assumed for incidental or consequential damages in connection with or arising out of information contained herein. This digital document is sold with the clear understanding that the publisher is not engaged in rendering legal, medical or any other professional services.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

LANGUAGES AND LINGUISTICS SERIES Critical Discourse Analysis: An Interdisciplinary Perspective Thao Le, Quynh Le and Megan Short (Editors) 2009. ISBN: 978-1-60741-320-2 Critical Discourse Analysis: An Interdisciplinary Perspective Thao Le, Quynh Le and Megan Short (Editors) 2009. ISBN: 978-1-60876-772-4 (Online Book) Building Language Skills and Cultural Competencies in the Military Edgar D. Swain (Editor) 2009. ISBN: 978-1-60741-126-0 Building Language Skills and Cultural Competencies in the Military Edgar D. Swain (Editor) 2009. ISBN: 978-1-60876-597-3 (Online Book)

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Second Languages: Teaching, Learning and Assessment Ryan L. Jikal and Samantha A. Raner (Editors) 2009. ISBN: 978-1-60692-661-1 Aphasia: Symptoms, Diagnosis and Treatment Grigore Ibanescu and Serafim Pescariu (Editors) 2009. ISBN: 978-1-60741-288-5 Building Strategic Language Ability Programs Joshua R. Weston (Editor) 2010. ISBN: 978-1-60741-127-7 Bilinguals: Cognition, Education and Language Processing Earl F. Caldwell (Editor) 2010. ISBN: 978-1-60741-710-1

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

LANGUAGES AND LINGUISTICS SERIES

BILINGUALS: COGNITION, EDUCATION AND LANGUAGE PROCESSING

EARL F. CALDWELL

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

EDITOR

Nova Science Publishers, Inc. New York

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010 by Nova Science Publishers, Inc. All rights reserved. No part of this book may be reproduced, stored in a retrieval system or transmitted in any form or by any means: electronic, electrostatic, magnetic, tape, mechanical photocopying, recording or otherwise without the written permission of the Publisher. For permission to use material from this book please contact us: Telephone 631-231-7269; Fax 631-231-8175 Web Site: http://www.novapublishers.com NOTICE TO THE READER The Publisher has taken reasonable care in the preparation of this book, but makes no expressed or implied warranty of any kind and assumes no responsibility for any errors or omissions. No liability is assumed for incidental or consequential damages in connection with or arising out of information contained in this book. The Publisher shall not be liable for any special, consequential, or exemplary damages resulting, in whole or in part, from the readers’ use of, or reliance upon, this material. Any parts of this book based on government reports are so indicated and copyright is claimed for those parts to the extent applicable to compilations of such works.

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Independent verification should be sought for any data, advice or recommendations contained in this book. In addition, no responsibility is assumed by the publisher for any injury and/or damage to persons or property arising from any methods, products, instructions, ideas or otherwise contained in this publication. This publication is designed to provide accurate and authoritative information with regard to the subject matter covered herein. It is sold with the clear understanding that the Publisher is not engaged in rendering legal or any other professional services. If legal or any other expert assistance is required, the services of a competent person should be sought. FROM A DECLARATION OF PARTICIPANTS JOINTLY ADOPTED BY A COMMITTEE OF THE AMERICAN BAR ASSOCIATION AND A COMMITTEE OF PUBLISHERS. LIBRARY OF CONGRESS CATALOGING-IN-PUBLICATION DATA Bilinguals : cognition, education and language processing / editor, Earl F. Caldwell. p. cm. Includes bibilographical references and index. ISBN 978-1-61761-567-2 (E-Book) 1. Bilingualism. 2. Second language acquisition. I. Caldwell, Earl F. P115.B555 2009 404'.2--dc22 2009032914

Published by Nova Science Publishers, Inc. Ô New York

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

CONTENTS

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Preface

vii

Chapter 1

Cultural and Linguistic Influence on Developmental Neural Basis of Theory of Mind and Self-construal: Whorfian Hypothesis Revisited Chiyoko Kobayashi Frank

Chapter 2

Language Processing in Bimodal Bilinguals Anthony Shook and Viorica Marian

35

Chapter 3

Psycholinguistic Abilities and Phonological Working Memory in Bilingual Children with Specific Language Impairment: A Cross-Cultural Study Dolors Girbau

65

Chapter 4

Psycholinguistic Challenges in Processing the Arabic Language Raphiq Ibrahim

81

Chapter 5

Neurocognitive Aspects of Processing Arabic and Hebrew Raphiq Ibrahim

103

Chapter 6

Visual Word Access in Monolinguals and Bilinguals in English and Spanish John Evar Strid and James R. Booth

123

Chapter 7

Vowels in Semitic Alphabet Languages Raphiq Ibrahim

147

Chapter 8

Methodological Issues in Research on Bilingualism and Multilingualism Lilian Cristine Scherer, Rochele Paz Fonseca and Ana Inés Ansaldo

167

Chapter 9

Bilingualism and Hispanic American Intelligence Test Scores Philip G. Gasquoine, Aracely Cavazos, Juan Cantu and Amy A. Weimer

181

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

1

vi

Contents

Chapter 10

Language Development through a Bilingual Lens Eswen Fava and Rachel Hull

201

Chapter 11

Improving Reading Skills for ESL Learners Using SoundSpel Michael D. Young, Michelle L. Wilson and Alice F. Healy

215

Chapter 12

A Novel Transliteration Approach in an English-Arabic Cross Language Information Retrieval System Ghita Amor-Tijani and Abdelghani Bellaachia

229

Chapter 13

Methods for Cross-Language Information Retrieval Kazuaki Kishida

243

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Index

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

287

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

PREFACE A bilingual person, in a broad definition, is one who can communicate in more than one language, be it actively (through speaking, writing, and/or signing) or passively (through listening, reading, and/or perceiving). More specifically, the terms bilingual and trilingual are used to describe comparable situations in which two or three languages are involved. A generic term for multilingual persons is polyglot. Multilingual speakers have acquired and maintained at least one language during childhood, the so-called first language. The first language (sometimes also referred to as the mother tongue). This new book gathers the latest research from around the globe in this field. Recent cross-cultural research on various cognitive functions found clear cultural influence on how we perceive our world. Likewise, neuroimaging research has found significant influence of culture / language in theory of mind (ToM) – ability to understand mental states of others – and self-construal (which is related to ToM) in the neural level. In Chapter 1, cross-cultural and brain imaging research on ToM and related social cognition are selectively reviewed. I discuss the roles of medial prefrontal cortex (mPFC) and temporoparietal junction (TPJ) that have been consistently implicated in ToM and perspective-taking (or distinguishing “self” from “other”). These structures may be particularly important for culturally unique ways of social cognition related to ToM and self-construal. Next, I briefly review current developmental theories of ToM, and discuss whether or not the recent findings from neuroimaging studies of ToM in children support these theories. Functional relationship between “language regions” and ToM will also be reviewed along with these discussions. Lastly, I present two models of culture / language-dependent and independent ToM development. Recent research suggests differences between bimodal bilinguals, who are fluent in a spoken and a signed language, and unimodal bilinguals, who are fluent in two spoken languages, in regard to the architecture and processing patterns within the bilingual language system. In Chapter 2, we discuss ways in which sign languages are represented and processed and examine recent research on bimodal bilingualism. It is suggested that sign languages display processing characteristics similar to spoken languages, such as the existence of a sign counterpart to phonological priming and the existence of a visual-spatial loop analogous to a phonological loop in working memory. Given the similarities between spoken and signed languages, we consider how they may interact in bimodal bilinguals, whose two languages differ in modality. Specifically, we consider the way in which bimodal bilingual studies may inform current knowledge of the bilingual language processing system, with a particular focus

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

viii

Earl F. Caldwell

on top-down influences, and the fast integration of information from separate modalities. Research from studies looking at both production and perception suggests that bimodal bilinguals, like unimodal bilinguals, process their languages in parallel, with simultaneous access to both lexical and morphosyntactic elements. However, given the lack of overlap at the phonological level (the presumed initial locus of parallel activation in unimodal studies) in bimodal bilinguals’ two languages, we conclude that there are key differences in processing patterns and architecture between unimodal and bimodal language systems. The differences and similarities between unimodal and bimodal bilinguals are placed in the context of current models of bilingual language processing, which are evaluated on the basis of their ability to explain the patterns observed in bimodal bilingual studies. We propose ways in which current models of bilingual language processing may be altered in order to accommodate results from bimodal bilingualism. We conclude that bimodal bilingualism can inform the development of models of bilingual language processing, and provide unique insights into the interactive nature of the bilingual language system in general. A variety of concepts/types of bilingualism and bilingual programs in U.S.A./Europe are presented. The cognitive benefits of bilingual education across several languages are reviewed. Diagnosis and intervention issues in bilingual children with Specific Language Impairment (SLI) / Typical Language Development (TLD) are discussed, including some behavioral and neurophysiology findings concerning language processes. A cross-cultural study was done by comparing children from U.S.A. (with SLI/TLD) and children from Spain (with SLI/TLD), who were involved in a larger project (Girbau & Schwartz, 2007, 2008). Forty-four sequential bilingual children (7;6-10;11 years old), with L1 = Spanish and L2 = English/Catalan, participated. The psycholinguistic abilities in any bilingual group with TLD were significantly higher than in any bilingual group with SLI (Spanish-English/Spanish-Catalan). The similarities of the cross-cultural profiles are discussed. Only children with TLD from Spain produced significantly more correct non-words (in the Spanish Non-word Repetition Task) than children with TLD from U.S.A. (who were exposed to English phonetics). This cross-cultural difference was not found for children with SLI; they all performed poorly in U.S.A. and Spain. The Spanish task was a good marker for SLI in both countries. Our results support the phonological working memory deficit associated with SLI, which appears to be independent of the particular bilingual background. The English Non-word Repetition Task was not sensitive in identifying SLI in these Hispanic unbalanced bilinguals, since English was their L2; their phonotactic representations in L1 seem to determine their performance on the task. The Spanish non-word repetition accuracy correlated significantly with the Auditory Association subtest from the Spanish ITPA, in children with SLI/TLD and in Spain/U.S.A. The task also correlated significantly with the Grammatical Integration subtest for children with SLI and in Spain/U.S.A. Both subtests involve auditory working memory, but the second one has also some visual support through pictures. Implications of the results for the crosscultural identification of SLI in bilinguals are discussed in Chapter 3. The 2006 PISA (Program for International Student Assessment ) report of worldwide scholastic achievements showed that about 50% of Israeli Arabic students were found to exhibit the lowest reading achievement scores in the PISA tests (level 1 and below) as compared to the other participating groups. Also, the MEITZAV national testing program in Israel (2001–2002) showed an achievement gap in language skills (reading and reading

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Preface

ix

comprehension) between Arab students and Jewish students in the school systems. This gap was larger than those found in the other areas tested (mathematics, science and technology, and English). The aim of this chapter is to explore the cognitive basis of these difficulties, specifically the diglossic situation in Arabic. Furthermore, the chapter discusses the unique features of the Arabic language that might contribute to the inhibition and slowness of reading acquisition and might even hinder the acquisition of basic academic skills. Finally, a model with a comprehensive basis (cognitive and neurocognitive) will be built in order to explain the complex linguistic situation of beginning Arabic learners. Chapter 4 is concerned with the cognitive evidence bearing on the nature of the mechanisms of language processing in Arabic which has critical linguistic characteristics and a diglossic factor. Additionally, other aspects, including a neurofunctional perspective, will be discussed. The aim of Chapter 5 is to explore the neurocognitive basis of the difficulties that the Arabic-Hebrew bilingual encounters in processing the Arabic language as a result of the diglossic situation in Arabic (spoken Arabic and Modern Standard, or Literary Arabic). Furthermore, the chapter discusses the unique features of the Arabic language that might contribute to the inhibition and slowness of reading acquisition and might even hinder the acquisition of basic academic skills. In the first section, two case studies of Arabic-Hebrew aphasic patients (M.H. and M.M.) are presented, with different disturbances in the two languages, Arabic (L1) and Hebrew (L2). They exhibited a complementary pattern of severe impairment of either L1 (Arabic) or L2 (Hebrew) constituting a double dissociation. These results suggested that the principles governing the organization of lexical representations in the brain are not similar for the two languages. The second section focuses on the functional architecture of reading in Hebrew and in Arabic. The effects of characteristics of Arabic and Hebrew as Semitic languages on hemispheric functioning were systematically examined. These patterns are compared with the modal findings in the literature, which are usually based on English. Also, the effects of the absence of almost all vowel information, the orthographies of the two languages, and their non-concatenative morphological structure were investigated. It was shown that when languages make different types of demands upon the cognitive system, interhemispheric interaction is dynamic and is suited to these demands. In that regard, both Arabic and Hebrew require a higher level of interhemispheric interaction than does English. Chapter 6 examined if visual word access varies according to language and bilingual status by comparing Spanish and English, priming two syllable CVCV words with bilingual children and monolingual children. The results suggest that lexical access in English is based on a larger phonological sub-lexical unit because of greater report of a unit bigger than the syllable, among both bilingual and monolingual subjects. In contrast, only weak evidence suggested that Spanish lexical access was based on the syllable because of greater report of that unit for bilinguals and monolinguals. Finally, monolingual or bilingual status of the reader did not have influence on English lexical access; however, Spanish bilinguals were influenced by the acquisition of English, suggesting that orthographically opaque languages can have an effect on transparent languages or that immersion, language dominance and literacy experience can influence reading in the other language. While several alphabetic systems are in use, Chapter 7 focuses on the two Semitic alphabet languages—Arabic and Hebrew—especially in the orthography of the two languages. Semitic scripts are unique in that short vowels are represented as diacritics on

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

x

Earl F. Caldwell

consonant letters. The unique characteristics of Arabic and Hebrew orthographies make them unique for investigations among Latin alphabets or even one among the other (Taouk and Coltheart 2004). Hebrew and Arabic are both read from right to left. The aim of Chapter 8 is to discuss important methodological issues regarding research on neurologically preserved bilingual and multilingual populations. It includes the variables to be considered in methodological aspects ranging from individual characteristics to task design and presentation, among others. Hispanic Americans as a group score 0.5SD below White Americans on intelligence test measures (administered in English) that emphasize language processing, but score similarly when visual-perceptual/visual-motor processing is required. The reason for this language decrement is unknown. Chapter 9 considers the possible contribution of bilingualism to this effect, as studies linking bilingualism and cognition (conducted with multiple ethnic groupings) have consistently shown a bilingual disadvantage compared to monolinguals on language processing tasks. Two data sets (older children and adults) of bilingual Hispanic American performance on various intelligence test measures administered in Spanish and English showed evidence of a visual-perceptual/visual-motor over language processing advantage of about 1SD. The size of the visual-perceptual/visual-motor over language advantage was similar in both languages suggesting it is bilingualism-related and not due to low English language proficiency. Bilingualism appears to be a potentially important factor in the Hispanic American language processing decrement seen on intelligence tests, although no direct study on the effect of this variable has yet been conducted. Bilingualism is a fertile resource for studying facets of language development and brain plasticity that may not be apparent in monolinguals. Chapter 10 will summarize historical and contemporary findings in the literature, discuss methodological issues that influence their interpretation, and suggest future directions for examining the neural substrates of language development and the consequences of having two languages in one brain. Chapter 11 examined the effects of using a revised, transparent spelling system SoundSpel, a phonetic reading tool, with learners of English as a Second Language. During 6 training sessions, 12 participants used unaltered material and 12 used SoundSpel texts, in parallel with standard English, when reading American elementary school material. They then answered multiple-choice comprehension questions. Both groups were pre-tested and posttested on comprehension tests of similar elementary school material without SoundSpel. No group differences were found across tests or training (in quiz performance or reading time), suggesting no beneficial or harmful effects from using SoundSpel. A post hoc analysis suggested that SoundSpel would be most beneficial for students who learn to speak English before they learn to read it. One of the main issues facing Cross Language Information Retrieval (CLIR) is untranslatable words, i.e., words not found in dictionaries, which are usually referred to as Out Of Vocabulary (OOV) words. Bilingual dictionaries in general do not cover most proper nouns (e.g., names of places, people, countries, etc.), which constitute a large proportion of OOV words. As they are often primary keys in a query, their correct translation is often necessary to maintain a good retrieval performance. Because they are spelling variants of each other in most languages, an approximate string matching technique against the target database index is usually used to find the target language correspondents of the original query key. The n-gram technique has proven to be the most effective among other approximate string matching techniques. A more complicated issue arises when the languages dealt with

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Preface

xi

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

have different alphabets. The approach usually taken is transliteration. It is applied based on phonetic similarities between the languages involved. However, transliteration by itself cannot guarantee the exact spelling of the transliterated words as found in the document collection. There are a variety of ways that a transliterated word can be spelled despite conventions that might exist. The fact that there is no one correct way of spelling a transliterated word shows the need for a technique that is capable of generating the different spellings found in the document collection. In Chapter 12, we chose to combine both transliteration and the n-gram technique in an English-Arabic CLIR system, in which Arabic documents were searched using English queries. We evaluated the effectiveness of this approach and compared it with other transliteration approaches. Experimental results showed the retrieval improvement gained using our transliteration approach over other existing approaches. Chapter 13 reviews technical methods for enhancing effectiveness of cross-language information retrieval (CLIR), in which target documents are written in different languages from that used for representing a search request. As the Internet has spread since the 1990s, the importance of CLIR has grown, and the research community of information retrieval has been tackling various CLIR problems. The purpose of this article is to overview exhaustively CLIR techniques developed in the research efforts. The following research issues on CLIR are covered: (1) strategies for matching the query and documents written in different languages, e.g., automatic translation or transliteration techniques, (2) techniques for solving the problem of translation ambiguity, (3) formal retrieval models for CLIR such as application of the language modeling, (4) methods for searching a multilingual document collection in which two or more languages are used for writing documents, etc.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved. Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

In: Bilinguals: Cognition, Education and Language Processing ISBN: 978-1-60741-710-1 Editor: Earl F. Caldwell, pp. 1-33 © 2010 Nova Science Publishers, Inc.

Chapter 1

CULTURAL AND LINGUISTIC INFLUENCE ON DEVELOPMENTAL NEURAL BASIS OF THEORY OF MIND AND SELF-CONSTRUAL: WHORFIAN HYPOTHESIS REVISITED Chiyoko Kobayashi Frank* Respecialization in Clinical Psychology, Fielding Graduate University, Santa Barbara, CA, USA

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Abstract Recent cross-cultural research on various cognitive functions found clear cultural influence on how we perceive our world. Likewise, neuroimaging research has found significant influence of culture / language in theory of mind (ToM) – ability to understand mental states of others – and self-construal (which is related to ToM) in the neural level. In this chapter, cross-cultural and brain imaging research on ToM and related social cognition are selectively reviewed. I discuss the roles of medial prefrontal cortex (mPFC) and temporo-parietal junction (TPJ) that have been consistently implicated in ToM and perspective-taking (or distinguishing “self” from “other”). These structures may be particularly important for culturally unique ways of social cognition related to ToM and self-construal. Next, I briefly review current developmental theories of ToM, and discuss whether or not the recent findings from neuroimaging studies of ToM in children support these theories. Functional relationship between “language regions” and ToM will also be reviewed along with these discussions. Lastly, I present two models of culture / language-dependent and independent ToM development.

Introduction Whorf (1956) hypothesized that our language constrains our thoughts and reflects our culturally unique world view. Later, Vygotsky (1967) elaborated this hypothesis, positing that *

E-mail address: [email protected] (Corresponding author)

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

2

Chiyoko Kobayashi Frank

human consciousness (or thoughts) has its basis in linguistic or historical contexts and is enabled only through the internalization of culture-specific symbols (i.e., language). From the late 1960s to mid 1990s, the so-called Sapir-Whorf hypothesis was discredited by cognitive scientists and linguists who emphasized universal and veridical ways of perceiving our world. Throughout these decades a view, which posited that universal linguistic [Chomsky, 1980] and cognitive developmental [Piaget, 1962; Sinclair, 1970] principles determine individuals’ thoughts and cognition, dominated. Despite the long period of obloquy, the Whorfian hypothesis has recently been revived following several new findings from cross-cultural / linguistic studies that have shown some influences of culture / language on people’s representations of conceptual properties [e.g., Boroditsky, 2001; Bowerman and Choi, 2003; Brown and Levinson, 1993; Choi and Bowerman, 1991; Lucy, 1992]. Theory of mind (ToM) is defined as the ability to attribute mental states to oneself or others, and to use such knowledge to make sense of and predict the behavior of agents (Dennett, 1980). Since the first experiment with the chimpanzee [Premack and Woodruff, 1978], various paradigms have been devised to test ToM in humans [Baron-Cohen, 2000]. Among those ToM tasks, a false-belief (FB) task has been the most commonly used for testing normally developing [Wimmer and Perner, 1983] as well as atypical pediatric populations [Baron-Cohen, Leslie, and Frith, 1985; 1986; see also Baron-Cohen, 2000]. In a typical FB task, two characters appear (e.g., Sally and Anne) in a scene. When one character, Sally, is present, Anne, the other character, puts a toy into a basket. Sally then disappears from the scene. While Sally is away, Anne takes the toy out of the basket and puts it into a box. The experimenter then asks the subject the critical false-belief question, “Where will Sally look for the toy?” Nearly universally observed results are that adults and children over 4 years of age correctly answer “basket” whereas younger children (as well as older children and adolescents with autism) fail the task by answering “box” [Baron-Cohen, Leslie, and Frith, 1985; 1986]. These failures reflect their lack of understanding that Sally’s belief about the location of the toy is different from Anne’s [Frith, 2003; Happé, 1993]. A meta-analysis of more than 100 of studies indicated that across cultures children pass various FB style tasks between the first 2.5 and 5 years [Wellman et al., 2001]. However, as I will describe in this chapter, many studies that tested non-English speaking children found significant variability in the passing age of FB tasks (e.g., Chen and Lin, 1994; Goushiki, 1999; Naito, 2003; Naito and Koyama, 2006; Shatz et al., 2003; Vinden, 1996; Wahi and Johri, 1994). These studies have attributed the delays or advancements in the FB task performance in the non-English speaking children to either linguistic and / or cultural difference. For instance, Naito (2003) attributed the poor FB performance in Japanese children to differences in attribution style: Japanese / Asians tend to attribute behaviors / actions to external causes, while Americans / Europeans tend to attribute them to internal or dispositional causes [Nisbett, 2003]. The difference in the attribution style may be related to the difference in the selfconstrual. Social psychological research has found that Americans and Europeans maintain the independent self, yet Asians emphasize the relational or interdependent self [Heine, 2001; Markus and Kitayama, 1991]. Meanwhile, brain imaging research has found increasing evidence that the medial prefrontal cortex (mPFC) and temporo-parietal junction (TPJ) (see Figure 1), which have been consistently implicated in ToM [Frith and Frith, 2003; Saxe et al., 2004], are also related to distinguishing “self” from “other” [Blakemore and Frith, 2003; Jackson and Decety, 2004; see also Decety and Grézes, 2006]; the mPFC being more

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Cultural and Linguistic Influence on Developmental Neural Basis of Theory…

3

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

important for self related judgment [Craik et al., 1999; Kelly et al., 2002; Lieberman et al., 2004; Ochsner et al., 2004; Ochsner et al., 2005] and TPJ being more critical for taking others’ perspectives [D’Argembeau et al., 2007; David et al., 2006]. As I will describe in this chapter, my colleagues and I found significant difference in the ToM specific activity in the TPJ between American and Japanese groups [Kobayashi et al., 2006; 2007b]. When viewing the same ToM cartoons, American children showed greater activity in the TPJ than Japanese children. The difference in ToM-specific brain activity between the two groups may be associated with the cultural difference in self-construal style: i.e., Japanese’ self-other distinction may be more blurred than Americans’ due to Japanese’ interpersonal selfconstrual style. Furthermore, as much as culture is inseparable from language [Vygotsky, 1967], it is reasonable to assume ToM and self-construal style influence (and are influenced by) language not only at the behavioral level but also at the neural level.

Figure 1. Brain diagram showing brain regions implicated in the ToM brain imaging studies. These include medial prefrontal cortex, anterior cingulate cortex, orbitofrontal cortex, and posterior superior temporal sulcus or temporo-parietal junction. Some of the “language regions” of the brain are shown in different colors.

Before I move on to the main body of this chapter, I briefly summarize three prominent theories of ToM development. The three main theories of ToM are “modular”, “theorytheory”, and “simulation”. The modular theory posits that ToM development is a genetically determined innate process [Fodor, 1983; Leslie, 1994; Scholl and Leslie, 1999]. The modular theorists argue that all humans are born with a set of mentalistic concepts that are encapsulated and invulnerable to experience [Fodor, 1983]. In this highly nativistic framework there is no room for cultural variation in ToM. Another theory of ToM, theorytheory hypothesis claims that ToM development is like the development of a scientific theory and it relies on conceptual development. Unlike modularists, theory-theorists accept some role of experience [Gopnik and Wellman, 1992; Wellman et al., 2001]. The third theory of

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

4

Chiyoko Kobayashi Frank

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

ToM is called the simulation hypothesis [Harris, 1991; Harris and Gross, 1996], which, although it admits that ToM development depends upon conceptual development, argues that the concepts are derived from children’s own direct experiences of such states, rather than through some abstract theorizing. These three prominent theories of ToM all predict universality at least in the early years. Even the least nativistic theory-theory assumes innate concepts of ToM which remains relatively resistant to socio-cultural influence throughout development [Wellman et al., 2001]. However, fortunately, the above three are not the only games in town. There is another theory of ToM that emphasizes the influence of the sociocultural effects on ToM [Astington, 1996; Vinden, 1999; Tomasello, 2003; Naito, 2007]. This group of ToM researchers follow Vygotsky’s theory [Vygotsky, 1967] and posit that it is the social-cognitive ability embedded in culture-specific symbolic systems (i.e., languages) that enables children’s ToM. As I discuss below, evidence from both behavioral and neuroimaging studies seems to begin to support this last hypothesis of ToM development. This chapter is divided into three main sections. The first part of the chapter considers results of cross-cultural behavioral studies on ToM and related social cognition and perception. The second part discusses universal or culture-specific neural basis of ToM and related social cognitive / perceptual functions based on the evidence from neuroimaging studies on these functions including ours. Particular emphasis is placed on the comparison between Asians and Anglo-American cultures. Neuroanatomically, my focus is on the comparison between mPFC and TPJ since my colleagues and I found significant difference in the ToM related activity in these regions between Japanese and Americans. The third part discusses main-stream theories of ToM development and whether or not recent findings from neuroimaging research support these theories. Finally, I present two models of ToM development; one representing universal and the other representing culture / languagedependent developmental mechanism of ToM.

Non-universal ToM and Self-construal A meta-analysis over 100 studies found that developmental trajectory of ToM is essentially the same across cultures [Wellman et al., 2001]. Similarly, no difference was found between Canadian, Indian, Peruvian, Thai, and Samoan children in the onset of passing a single FB paradigm [Callaghan et al., 2005]. However, the universal ToM hypothesis has not been uncontested, as several ToM studies conducted outside the Anglo-American cultural or linguistic boundaries have obtained mixed results. Some of these cross-cultural / crosslinguistic studies have supported the universal developmental onset time of ToM [Avis and Harris, 1991; Collaghan et al., 2005; Lee et al., 1999; Naito et al., 1994; Tardiff and Wellman, 2000; Yazdi et al., 2005], whereas others found either delays [Chen and Lin, 1994; Goushiki, 1999; Liu et al., 2008; Louis, 1998; Naito, 2003; Naito and Koyama, 2006; Vinden, 1996] or advancements [Shatz et al., 2003] in ToM for non-English speaking children. For example, onset of FB understanding in Hong Kong children appeared 2 years later than that in Canadian children (Liu et al., 2008). Many of these authors have given linguistic or cultural differences as explanations for the poorer or better performance of the children living in nonAnglo-American countries. For instance, Junin Quechua children’s poor ToM performance has been attributed to their lack of mental state verbs [Vinden, 1996]. In Lee et al.’s (1999) study with Mandarin-speaking children, even though the children’s performance for the FB

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Cultural and Linguistic Influence on Developmental Neural Basis of Theory…

5

task was overall comparable to Western children’s performance, their performance was influenced by the choice of verbs (i.e., three verbs that all mean “think”) used in the FB task. The Mandarin-speaking children performed significantly better when yiwei and dang, which connote that the belief referred to may be false, were used than when xiang (the more neutral verb) was used. The linguistic influence on ToM has also been noted in a few studies that found advanced performance in Turkish and Puerto Rican (PR) Spanish-speaking children [Shatz et al., 2003]. In this study, the Turkish or PR Spanish-speaking children, who have either a specific verb (Turkish) or a case marker (PR Spanish) available to make the falsebelief mental state more explicit, performed better than Brazilian Portuguese or Englishspeaking children who do not have those lexicons. As I mentioned earlier, Naito (2003) has attributed the below-chance ToM performance in 4- and 5-year-old Japanese children to differences between American / European and Asian cultural attribution styles: specifically, people raised in American / European cultures tend to attribute behaviors to internal causes (i.e., traits), while people raised in Asian cultures tend to attribute them to external or situational causes [Masuda and Nisbett, 2001; Nisbett, 2003]. These cultural differences may stem from an even greater difference between Asians and Americans / Europeans in the self-construal. Social psychologists have found that Easterners have a relational self (or inter-connected self with other people in the society) while Westerners have an individual self (or autonomous self separating from others) [Heine, 2001; Markus and Kitayama, 1991; Nisbett, 2003]. The difference in the self-construal presumably affects various human perceptions and cognitions including the causal reasoning. For instance, it has been shown that Westerners remember self-related adjectives better than intimate-other-related adjectives [Lord, 1980, Klein et al., 1989] whereas Chinese remember self-related adjectives no better than intimate-other-related adjectives [Zhu and Zhang, 2002]. According to the culture-dependent attentional hypothesis, a person in an interdependent culture might focus his / her attention on others and away from the self [Markus and Kitayama, 1991]. Cohen and Gunz (2002) tested Americans and Asians with an emotional perspective-taking task and found that when they were in the center of attention, Americans were more likely to take the first-person perspectives by projecting their own emotions onto others, while Asians were more likely to take the third-person perspectives. Similarly, in Wu and Keysar’s (2007) perspective-taking experiment using eye-tracking, Chinese participants were more tuned into their partner’s perspective than American participants were. Since perspective-taking is an important aspect of ToM [Decety and Chaminade, 2003; Samson et al., 2007], the difference between Asians (interdependent culture) and Americans (independent culture) in ToM may be attributed to the cultural difference in how “self” and “others” are construed.

Japanese Self-construal Increasing evidence from socio-psychological studies suggests that Japanese and other Asian cultures encourage the use of “group-agency” more than individualistic “self-agency” to explain various kinds of human behaviors [Ames et al., 2001]. The priority of the groupagency over the self-agency in Japanese culture is reflected in the etymological meaning of self in Japanese; i.e., jibun or “my portion” [Nisbett, 2003]. While Indo-European language speakers may conceive an event based on the “action-agent” model [Werner and Kaplan,

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

6

Chiyoko Kobayashi Frank

1963], Japanese speakers may tend to frame the event as a situation that is beyond the agent’s control [Maynard, 1997]. The sense of demoted self-agency (or the promoted group-agency) is reflected in the fact that Japanese personal references are largely dependent on the person’s socially-conceived roles. For instance, within a Japanese family, parents refer themselves as otosan (“father”) or okasan (“mother”). As a result, Japanese children have been found to acquire pronouns much later than Indo-European language-speaking children [Ide, 1977; see also Clancy, 1985]. Japanese / Asian self concept has been described as the opposite of American / Western self concept [Maynard, 1997; Nisbett, 2003]. For instance, Maynard [1997] describes the Japanese self as a “relational self” and the American self as an “autonomous self”. In Japan, the self-to-society boundaries are much more blurred than in America. In America, each individual is forced to seek autonomous self that may be dependent on the society only psychologically. The diminished sense of self is reflected in the ordinary Japanese discourse, in which an event is summarized as a “thing” (koto) which is beyond the subject’s control [Maynard, 1997]. The demoted self concept pervades the Japanese society across all ages [Clancy, 1985; 1986]. Given these cultural differences in self-construal between Japanese / Asians and Americans / Europeans, the difference found in the FB task performance in Naito (2003) can be attributed more to the strategic difference than to the performance difference. More recently Naito and Koyama (2006) found that Japanese children, when asked to provide the justifications of the protagonist’s behavior, gave more situational justifications (e.g., “The object was there first”) than internal explanations (e.g., “He thinks the toy is still there”) typically given by Western children [Bartsch and Wellman, 1989; Wimmer and Mayringer, 1998]. These results are compatible with findings that, in Japanese and several other nonWestern cultures compared to the West, behaviors / actions are explained in terms of the situations than the internal trait / disposition of the person [Lillard, 1998; Masuda and Kitayama, 2003; Vinden, 1999]. These strategic or qualitative differences in ToM may not be easily detected by the conventional forced-choice FB tasks. Next, I turn our discussion to some of the neuroimaging studies of ToM and related social cognition, and then I discuss recent studies (including ours) that have revealed some qualitative differences in ToM among different cultural / linguistic groups.

Universal Neural Bases of ToM and Self construal Brain imaging studies have examined the neural correlates of ToM using the FB style paradigm in adults [e.g., Brunet et al., 2000; Fletcher et al., 1995; Gallagher et al., 2000; Goel et al., 1995; Happé et al., 1996; Kobayashi et al., 2006; Sabbagh and Taylor, 2000; Saxe and Kanwisher, 2003; Vogeley et al., 2001]. Many of these studies have found significant activity in the mPFC during false-belief conditions [e.g., Flethcer et al., 1995; Gallagher et al., 2000; Goel et al., 1995; Happé et al., 1996] (see Figure 1). In addition, the TPJ has also been suggested to be important for ToM processing. This area was found to become active during both true- and false-belief conditions and not during false representations in a non-social control condition [Saxe and Kanwisher, 2003; Saxe and Wexler, 2005]. Other brain regions implicated in these and other ToM brain imaging studies include anterior cingulate cortex, middle frontal gyrus, precuneus / posterior cingulate cortex, superior temporal sulcus (STS),

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Cultural and Linguistic Influence on Developmental Neural Basis of Theory…

7

orbito-frontal cortex (or ventro-medial prefrontal cortex [vmPFC]) and temporal pole, however, not as consistently as the mPFC and the TPJ [see Frith and Frith, 2003, and Saxe et al., 2004] (see Figure 1). In our group’s study both American and Japanese adults employed the right mPFC [Kobayashi et al., 2006], and children from both cultures employed the vmPFC [Kobayashi et al., 2007b] during the ToM relative to the non-ToM condition. Likewise, the ToM specific mPFC activity was seen in French [Brunet et al., 2000, 2003] and German [Abraham et al., 2008; Ferstl and von Cramon, 2002; Lissek et al., 2008; Vogeley et al., 2001] subjects. Crosscultural neuroimaging research of social cognition is still in its infancy. But besides our study, Zhu et al. (2007) tested Chinese and American / European subjects with an adjective judgment task and found similar activity related to the self in the mPFC in both cultural groups. These results suggest that ToM and related social cognition tasks employ the mPFC universally.

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Which Is More Involved in Universal ToM: Medial Prefrontal Cortex or Temporo-Parietal Junction? As I mentioned earlier, while the majority of the earlier neuroimaging studies of ToM implicated the mPFC as the most important region for ToM [e.g., Brunet et al., 2000; Calarge et al., 2003; den Ouden et al., 2005; Fletcher et al., 1995; Gallagher et al., 2000; Gallagher et al., 2002; Goel et al., 1995; Happé et al., 1996; McCabe et al., 2001; Sabbagh and Taylor, 2000; Vogeley et al., 2001], other later fMRI studies of ToM, in which the experimental condition (i.e., mental state attribution) was closely matched with the control conditions (e.g., asking subjects about any socially relevant information about a person), found more robust brain activity in the TPJ than in the mPFC [Saxe and Kanwisher, 2003; Saxe and Wexler, 2005]. Moreover, the frontal lobe involvement in ToM has been questioned since a few studies on frontal-damaged patients found intact ToM performance [Bird et al., 2004; Fine et al., 2001], while a few other studies found below-chance performance in ToM performance in TPJ-damaged patients [Apperly et al., 2004; Samson et al., 2004]. The precise reason why the earlier brain imaging studies of ToM did not find the TPJ activity is unknown. But one possible explanation is the recent advances in brain imaging methods. Earlier ToM brain imaging studies used positron emission tomography (PET) for scanning. PET data are smoothed with larger filters, resulting in a lower resolution of images than the fMRI data. Thus, it is possible that activity in the TPJ was present in the earlier ToM imaging studies but it was not strong enough to remain significant after the filtering. Another reason may be the improved data analysis methods used for the fMRI studies in general. Many earlier ToM brain imaging studies used fixed-effect analyses for analyzing the imaging data with relatively few participants. However, more recent studies of ToM imaging [e.g., Kobayashi et al., 2006; Saxe and Kanwisher, 2003] included a relatively larger number of subjects (> 10) using random-effect analyses that enable inference to the population [Friston et al., 1999]. Thus, the lack of TPJ activity in earlier ToM imaging results may be explained by the smaller number of subjects used in these studies. Despite these recent seemingly convincing results discussed above, the TPJ’s involvement in ToM may not be universal. In our study, while we found that American adults and children had common brain activity in the bilateral TPJ during both the cartoon and story

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

8

Chiyoko Kobayashi Frank

ToM tasks [Kobayashi et al., 2007a], we did not find the same convergent activity in the TPJ in Japanese adults and children [Kobayashi, 2007; 2008]. Moreover, while viewing the same ToM cartoon, American children activated the TPJ more than Japanese children [Kobayashi et al., 2007b]. Brunet et al. (2000, 2003) tested French adults with comic strips depicting intentional and non-intentional (control) stories, also using PET, and found more brain activity in several regions including the mPFC during the intentional than the control condition. More recently many studies tested German adults with various paradigms using fMRI, and implicated the mPFC or anterior cingulate cortex (ACC) and not the TPJ in ToM [Vogeley et al., 2001; Walter et al., 2004; Sommer et al., 2007; Abraham et al., 2008; Lissek et al., 2008]. In addition, in Zhu et al. (2007) described above, American / European participants activated the left TPJ (close to middle temporal gyrus [MTG]) more than Chinese participants when they processed adjectives related to intimate-others. Similarly, Kim et al. (2005) found activity in the mPFC, and not the TPJ, when Korean participants were engaged in the facial affect judgment (related to ToM). Moreover, in Takahashi et al. (2004), the brain activity specific to the evaluation of guilt and embarrassment in Japanese participants was greater in the mPFC than in the TPJ. These results may indicate that the mPFC is most consistently involved in ToM understanding across cultures / languages. In contrast, the TPJ’s involvement in ToM may be limited to American or British cultures or English language.

Culture Specific Roles of Neural Bases of ToM and Self construal

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Culture Dependent Roles of mPFC and TPJ in Perspective-Taking In addition to its putative role in ToM, as I mentioned earlier, the TPJ has been suggested to be involved in the more general ability of distinguishing self-agency from others [Blakemore and Frith, 2003; Decety and Chamidade, 2003; Jackson and Decety, 2004; see also Decety and Grézes, 2006]. In a PET study of ToM and self, self-related story alone (without ToM) led to increased activity in the right TPJ [Vogeley et al., 2001]. In relation to this, several perspective-taking studies using a variety of tasks implicated the TPJ close to the inferior parietal lobule (IPL) [Grèzes et al., 2004; Lamm et al., 2007; Ruby and Decety, 2001; 2003; 2004; Samson et al., 2007]. The TPJ has been implicated in both emotional [Lamm et al., 2007; Ruby and Decety, 2004] and social cognitive [D’Argembeau et al., 2006; Ruby and Decety, 2003] perspective-taking studies. While the mPFC has also been implicated in the perspective-taking studies [D’Argembeau et al., 2006; David et al., 2006; Ochsner et al., 2005; Ruby and Decety, 2001; 2003; 2004], the mPFC activity has been associated more with the first-person perspective than the third-person perspective [David et al., 2006; D’Argembeau et al., 2007; Ochsner et al., 2004; Ochsner et al., 2005]. For instance, in D’Argembeau et al. (2007) participants judged the extent to which trait adjectives described their own personality or close friends’ personality. When participants judged the adjectives as describing themselves, the mPFC activity was observed. In contrast, when they judged the adjectives as describing close friends the TPJ / IPL activity was found. These results are consistent with the recent findings from the neuroimaging studies of self-referential process [Mitchell et al., 2005; Ochsner et al., 2004; Ochsner et al., 2005]. In these studies, activity in

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Cultural and Linguistic Influence on Developmental Neural Basis of Theory…

9

the mPFC was found when participants judged one’s own, as compared to another 1 individual’s, emotional [Ochsner et al., 2004] or mental [Mitchell et al., 2005] states . Given the cultural difference in the self-construal process (see above), the self-other distinction that these two brain structures mediate may play an important role in contributing to some of the cross-cultural differences in neural activity associated with ToM. As I described earlier, by comparing between Chinese and Americans, Zhu et al. (2007) found overlap in the activity between the self and intimate-other related adjective judgments in the mPFC in Chinese, and not in Americans. In a few ToM brain imaging studies in adults, ToM specific TPJ activity was found in Americans [Kobayashi et al., 2007a], but neither in Japanese [Kobayashi et al., 2008] nor in Koreans [Kim et al., 2005]. For the reasons I discussed above on the dichotomy of independent Western self and interdependent Asian self, the diminished activity in the TPJ during ToM or mentalizing (a term often synonymously used with ToM [Frith and Frith, 2003]) in Japanese [Kobayashi et al., 2007a; 2007b; 2008; Takahashi et al., 2004] and Korean [Kim et al., 2005] adults might reflect the reduced sense of self-other distinction in Asian culture relative to Anglo-American culture. Upon mentalizing, Anglo-Americans may use the TPJ more to detect other-agency in the mentalizing task, whereas Asians may use the mPFC more (or the TPJ less) because they distinguish self from others to a lesser extent than Anglo-Americans based on their interdependent or relational self concept.

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Do Neuroimaging Studies Support Theories of ToM Development? The developmental mechanism of ToM has been investigated with a plethora of paradigms by many researchers. Despite the increasing evidence that supports the developmental relationship between language / culture and various social cognitive capacities (including ToM) [see e.g., de Villiers, 2000, de Villiers and de Villiers, 2000, and Ochs], current main stream developmental theories of ToM discount the cultural / linguistic effects on ToM. In the remaining part of this chapter, I discuss some of those prominent theories of ToM development and whether evidence from the neuroimaging studies of ToM (including ours) supports these theories. Second, I discuss the possible interactions between brain areas associated with language processing and ToM candidate regions, and whether these interactions are occurring universally or in culture / language-dependent ways. Lastly, I present new models for the mechanisms of universal and language / culture-dependent developmental trajectories of ToM.

Neurological Evidence for/against Theories of ToM Development The developmental mechanism of ToM has been investigated with a plethora of paradigms by many researchers. There are three main stream theories of ToM development; “Modular”, “Theory-theory”, and “Simulation”. These theories were developed primarily to 1

Moreover, TPJ’s selective involvement in ToM has recently been questioned because several imaging studies found activity in the right TPJ when the subjects tried to shift their attention to task relevant stimuli [Corbetta et al., 2008]. Therefore, it may be arguable that the TPJ is involved in any tasks (including ToM) that require reorientation of attention [Mitchell, 2008] (but see Scholz et al., 2009 for the counter-evidence of this).

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

10

Chiyoko Kobayashi Frank

account for the results of the FB task performance; failure in normally-developing 3-year-old children and older children with autism and successful performance in 4-year-old children.

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Modular ToM Since Fodor’s highly influential book, “The Modularity of Mind” (1983) was published, the modular view has become influential in cognitive science. According to Fodor’s view, the architecture of each of our sensory (and some of our cognitive) functions can be best described as modular in the sense that it is: (1) informationally encapsulated, (2) unconscious, (3) fast, (4) has shallow outputs, (5) involves obligatory firing, (6) localized, (7) domain specific, and (8) ontogenetically and pathologically universal. Modular ToM theorists argue that the reason why 3-year-olds fail the FB task is not due to any impairments in the core ToM module, but due to immature selection processing (SP) mechanism which is domaingeneral (unlike ToM which is domain-specific) and functions like an inhibitory control [Leslie, 1992; Roth and Leslie, 1998; Leslie et al., 2004]. Leslie and his colleagues (advocates of modular ToM theory) posit that older children with autism are impaired in ToM not because of any damage in the SP part, but because of some dysfunction in the ToM module [Roth and Leslie, 1998] and that this component of ToM comprises the specific innate basis of ToM. Modular hypothesis has recently found some robust support from a series of habituation experiments in infants. These experiments have shown that non-verbal FB tasks are passed, even by 15 month-old infants [Onishi and Baillargeon, 2005, Southgate et al., 2007; Surian et al., 2007]. These studies support the main tenet of modular hypothesis; innate basis of ToM. These studies, however, were criticized because they failed to show why the reality-bias seen in 3 years-olds was not seen in the infants and that infants may be doing three-way association among agent, object, and place without any FB understanding [Perner and Ruffman, 2005]. In terms of neural basis of ToM, the modular hypothesis of ToM predicts that ToM is represented in a highly circumscribed brain region [Gallagher and Frith, 2003]. However, neuroimaging studies of ToM and autism have presented mixed results regarding this prediction. On the one hand, many ToM imaging studies utilizing a variety of ToM paradigms have consistently implicated the mPFC [Brunet et al., 2000; Fletcher et al., 1995; Gallagher et al., 2000; Gallagher et al., 2002; Goel et al., 1995; Happé et al., 1996; Kobayashi et al., 2006; Vogeley et al., 2001; see also Frith and Frith, 2003], indicating some modularity in ToM. On the other hand, structural brain imaging studies on autism alone found abnormalities (usually increases / decreases in volume) in more than 14 brain regions that are widely distributed [Brambilla et al., 2003]. Although the amygdala is most often found to be structurally abnormal in individuals with autism [Schultz et al., 2000], only a few functional brain imaging studies of ToM implicated this region for ToM [e.g., Baron-Cohen et al., 1999]. The increasing evidence suggests that autism is not a modular but a distributed disorder involving functionally connected multiple brain regions [Geschwind and Levitt, 2007; Müller, 2007]. Moreover, it has become increasingly likely that none of the candidate ToM areas are specialized for ToM or mentalizing per se. For instance, it has been shown that the medial prefrontal area is also involved in self-referential activity [Gusnard et al., 2001], social norm transgression [Berthoz et al., 2002], and finding coherence in stories [Ferstl and von Cramon, 2002].

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Cultural and Linguistic Influence on Developmental Neural Basis of Theory…

11

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Table 1. Developmental neuroimaging studies of ToM and related social/ cognitive functions (selected studies in ascending order by year of publication) Authors (Year) Ohnishi, et al. (2004)

Imaging Method fMRI

Liu (2005)

ERP

24 adults and 41 children (57 yrs-old)

Dapretto, et al. (2006)

fMRI

Wang, et al. (2006a)

fMRI

10 TD children and 10 children with ASD (12-13 yearsold) 12 Adults (2333 yrs-old) and 12 children (914 yrs-old)

Wang, et al. (2006b)

fMRI

Kobayashi, et al. (2007a)

fMRI

Kobayashi, et al. (2007b)

fMRI

Moriguchi, et al. (2007)

fMRI

Subject

Task

Main Findings

11 children (713 yrs-old)

Video of hands grasping some objects (imitation). Animation of intentional movement of geometric figures (ToM). Animation based first-order TB and FB task

ToM condition activated right mPFC, bilateral STG, right SMG, right MTG, right TP, right FG, bilateral MOG, and left cerebellum. Both imitation and ToM conditions activated bilateral MTG, right STG, bilateral MOG, and bilateral FG.

18 children/ adolescents with ASD (717 yrs-old) and 18 TD children/ adolescents (816 yrs-old) 16 Adults (1840 yrs-old) and 12 children (812 yrs-old)

12 American and 12 Japanese children (8-12 yrs-old) 16 children/ adolescents (916 yrs old)

Increasing localization in the left frontal region from child-passers to adults. More diffused bilateral frontal activity in children than in adults. Facial imitation TD children activated more right task precentral gyrus, right ACC, bilateral IFG, insula, amygdala, hippocampus, caudate, putamen, and thalamus than children with ASD. Cartoon based Children activated Right STG, Irony task bilateral IFG, right MFG, right STS, and left mPFC more than adults. Adults activated more posterior brain regions (e.g., visual cortex) than children. Story-based irony TD children activated more right IFG (when contextual cues were task available) and bilateral STS (when both types of cues were available) than children with ASD. Children with ASD activated temporal regions more when only prosodic cues were available. Cartoon and story Overall more brain activity in children based second-order than adults. Both Adults and children FB task activated right IPL and TPJ. Adults activated ToM areas (e.g., TPJ) during the story aacondition, but children activated these areas during the cartoon ToM condition. Cartoon and story Both American children and Japanese based second-order bilingual children activated bilateral FB task vmPFC.

Animation of Age related positive correlation in the intentional dorsal mPFC, and negative movement of correlation in the ventral mPFC. geometric figures. Abbreviations: ACC = anterior cingulate cortex, FG = fusiform gyrus, IFG = inferior frontal gyrus, IPL = inferior parietal lobule, MFG = middle frontal gyrus, MOG = middle occipital gyrus, mPFC = medial prefrontal cortex, MTG = middle temporal gyrus, OFC = orbito-frontal cortex, SMG = supramarginal gyrus, STG = superior temporal gyrus, STS = superior temporal sulcus, TP = temporal pole, TPJ = temporo-parietal junction, vmPFC = ventro-medial prefrontal cortex; AS = Asperger’s syndrome, ASD = autism spectrum disorder, HFA = high functioning autism, TD = typically developing

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

12

Chiyoko Kobayashi Frank

Another prediction of the modular ToM hypothesis is that since ToM is relatively unchanging throughout the development (after 18 months), children’s neural basis of ToM would be very similar to adults’ [Fodor, 1983; Scholl and Leslie, 1999]. However, evidence from several recent brain imaging studies of developmental neural basis of ToM argues against this prediction (see Table 1). For instance, Liu (2005) found more diffused bilateral frontal activity in children than adults who showed more localized left frontal activity during the animation-based FB task. Similarly, in our study, children activated many more brain regions than adults during the ToM condition relative to the baseline [Kobayashi et al., 2007a]. Moriguchi et al. (2007) has also found an age related positive correlation in the dorsal mPFC area, but a negative correlation in the ventral mPFC area. Moreover, by examining the neural basis of irony in adults and children, Wang et al. (2006a) found more robust activity in the prefrontal areas in children than in adults, who activated posterior brain regions more during the irony relative to the control conditions. These results seem to argue against the main prediction of the modular hypothesis of ToM: ToM is innate and relatively unchanging throughout development, and so, there should be little variation between adults and children in the neural bases of ToM. Thus, the results from the developmental neuroimaging studies of ToM do not seem to lend support to the strict modular theory of ToM. Nonetheless, one thing that the modular ToM theory (but not others) explains well is the relationship between the executive function (or inhibitory control) and ToM (see Apperly et al., 2009). Although the neuro-functional relationship between the executive function and ToM is still controversial [Saxe et al., 2004, Kain and Perner, 2005], several brain imaging studies of ToM (including ours) have implicated the dorsolateral prefrontal cortex (i.e., a brain region often implicated in the neuroimaging studies of executive function) [BaronCohen et al., 1999; Brunet et al., 2000; Kobayashi et al., 2006; Sanfey et al., 2003]. Thus, neurological evidence may at least support the domain-general SP part of this hypothesis. In addition, the modular hypothesis has recently been revised to focus more on “functional specificity” that does not require the highly circumscribed brain region specifically dedicated to a cognitive function [Barret and Kursban, 2006]. Thus it could still be argued that some core and universal ToM components are present early in human development [Siegal et al., in press]. In fact, our conjunction analysis supports this notion by finding convergent TPJ activity among American and Japanese adults and children [Kobayashi, 2007; 2008]. However, since modular hypothesis remains relatively tacit about the precise developmental mechanisms of ToM, a future task of the proponents of this theory may be to articulate what these mechanisms are.

Theory-Theory ToM The Theory-theory hypothesis of ToM has been developed as an alternative to the modular ToM view, which Wellman and his colleagues describe as being “antidevelopmental” [Wellman et al., 2001]. This hypothesis posits that the adults’ version of ToM, folk psychology, is drastically different from the children’s version because we revise our ToM theory many times throughout life just as a scientist revises his / her theory based on alternative empirical evidence. They further argue that ToM is essentially a dynamic process: it is subject to revision depending upon our individual experiences [Gopnic and Wellman, 1992]. Although the Theory-theorists agree that some innate modules or core structures exist

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Cultural and Linguistic Influence on Developmental Neural Basis of Theory…

13

for ToM, they contend that these structures are subject to extensive revisions throughout one’s life, whereas the Modularists insist that those core modules do not change [Wellman et al., 2001]. According to this theory, the reason why 2-year-olds fail the FB task is that they do not have the concept of “belief”, which develops only sometime after the third birthday when other epistemic concepts (e.g., “knowing” and “thinking”) become available along with lexical expansion [Bartsch and Wellman, 1995]. Evidence from a meta-analysis of more than 100 ToM studies seems to support the theory-theory hypothesis: it failed to show early ToM competence in children younger than 3 years of age, but showed instead a significant covariation between age and ToM performance (but see Yazdi et al. [2006] for counterargument / evidence of this point). Until recently, there has been no neurological evidence to support or reject this hypothesis, as there has been no brain imaging study of ToM development. The results of a few recent developmental brain imaging studies seem to support this hypothesis because these studies have found clear age-related differences in several brain regions involved in ToM and related socio-cognitive functions [Liu, 2005; Liu et al., 2005; Kobayashi et al. 2007a; Moriguchi et al., 2007; Wang et al., 2006a] (see Table 1). These results support the Theorytheorists’ main tenet: some major differences exist between adults’ ToM (a.k.a., folk psychology) and children’s ToM, which develops throughout life. These studies provide some initial evidence for significant changes in the neural basis of ToM and related social cognition between the first 10 years of childhood and adulthood [Liu, 2005; Kobayashi et al., 2007a; Moriguchi et al., 2007; Wang et al., 2006a]. Interestingly, the majority of these studies [Kobayashi et al., 2007a; Moriguchi et al., 2007; Wang et al., 2006a] did not find any behavioral differences between the age groups in the FB / irony task performance. Although Wang et al. (2006a) suggested that the developmental change reflects increasing automatization of ToM / irony understanding as people age, alternatively, this change may reflect changes in strategies that people use for ToM: adults may use different ToM strategies from children. One notable difference between adults and children in the ToM related brain activity found in our study is that even though both adults and children activated the mPFC, children activated more ventral mPFC areas [Kobayashi et al., 2007b]. Similarly, an event relatedpotential (ERP) experiment found more activity in the ventral prefrontal area in 6 years-old children during the animation-based FB than the control condition [Liu et al., 2005]. Moreover, in Moriguchi et al. (2007), a significant positive correlation between ToM related dorsal mPFC activity and age was found. In the vmPFC the relationship was reversed: the activity decreased as age progressed. It has been suggested that the dorsal cingulate area is primarily dedicated to cognitive aspects of behaviors while the ventral cingulate area is more dedicated to emotional aspects of behaviors [see Bush et al., 2000]. In line with these results, an ERP study [Sabbagh, 2004] found vmPFC / orbito-frontal activity while their subjects encoded others’ emotions from eye gazes, and dorsal mPFC activity when they engaged in the cognition-based standard ToM task. These results suggest that ToM may require more emotion-based processing for children but more cognition-based processing for adults. These results also seem to fit well with the Theory-theorists’ prediction that more emotion-laden desire-based ToM understanding precedes more cognition-laden belief-based ToM understanding in development [Bartsch and Wellman, 1995]. One criticism of this theory may be on the universality of ToM. Even through this theory emphasizes influence of individual experiences on ToM development, its view is similar to

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

14

Chiyoko Kobayashi Frank

Modularists in that they argue for the universal developmental of ToM because its underlying assumption is that people’s theory-making ability is the same everywhere [Wellman, 2001; see Lillard, 1998]. As I described earlier, results from several behavioral studies [e.g., Lee et al., 1999; Naito and Koyama, 2006; Shatz et al., 2003] and our neuroimaging studies [Kobayashi, 2007; Kobayashi et al., 2006; 2007b, 2008] argue against the universality of ToM development.

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Simulation ToM A third theory, simulation ToM theory, posits that people do not use any naïve theories of psychology when predicting and explaining the behaviors of others [Goldman, 1989; Harris, 1992]. Simulationists agree that ToM development depends upon conceptual development, but they argue that the concepts are derived from a child’s own direct experience of such states, rather than through some abstract theorizing. When explaining others’ mental states, the child uses his / her own mental states as a model, very much like putting his / her mind into others’ shoes [Harris, 1991, 1992]. Simulation theory has recently been embraced with enthusiasm by neurologists and cognitive scientists following the discovery of the mirror neuron system [Fogassi et al., 2005; Iacoboni, 2005]. Robust activity in the mirror regions has been found in several studies that tested imitation [Decety et al., 1997; Iacoboni, 2005], discrimination of “self” from “other” [Decety and Chaminade, 2004], and reading of others’ intentions [Burgess et al., 2001; Iacoboni et al., 1999]. For example, significant brain activity in these areas was found when monkeys engaged in a task in which they had to infer an experimenter’s intentions [Fogassi et al., 2005]. Several brain imaging studies on human adults also found significant activity in the inferior parietal regions (i.e., a part of the mirror neuron system) while their subjects engaged in imitation tasks [Decety et al., 1997; Chaminade and Decety, 2002; Nakamura et al., 2004]. It has been suggested that these lower-levels of ToM processing (e.g., understanding intentions of others and imitation) form bases for higher-order ToM (e.g., inferring from others’ beliefs) [Malle, 2002; Meltzoff and Brooks, 2001]. However, whether or not the human mirror neuron system is involved in higher-order ToM such as FB reasoning is still controversial. The major reasons are that the mirror neuron regions (the IPL and inferior frontal gyrus [IFG]) are not commonly implicated as ToM candidate areas, and that the FB task requires subjects to identify reasons for the behavior and to predict the next action. “Reasoning” is more than just “decoding” in that it involves prediction of behavior based on the person’s past mental states in addition to the capacity to attribute current mental states to others [Nichols and Stich, 2003; Sabbagh, 2004]. This reasoning might require some abstract theorizing, not only the detection of intentions through simulation. Another criticism of the simulation theory is that simulation theory cannot account for errors that people often make about their own mental state judgment [see Saxe, 2005]. An fMRI study that tested social norm violation within the person’s own culture or others’ culture did not support this theory [Saxe and Wexler, 2005]. Subjects employed the right TPJ more when the stories were about norm-violation within their own cultures than when they were about the same norm-violation in other cultures. These results argue against the main prediction of simulation theory: other minds are represented fundamentally in terms

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Cultural and Linguistic Influence on Developmental Neural Basis of Theory…

15

of their similarity to the person’s own mind (however, see Apperly [2008] for counterargument of this point). Our study with adults and children, however, found some support for this theory by showing convergent activity in the IPL for both cartoon- and story-based ToM tasks [Kobayashi et al., 2007a]. Similarly, Wang et al. (2006a) found a greater activity in the IFG area in children than in adults. One explanation of these results may be that mirror neuron regions are important for children (more than for adults) because these regions are involved in a lower and implicit level of ToM processing. Taken together, these results may indicate that the human mirror neuron system may be involved in a lower-level ToM processing (e.g., detection of intentions) that does not require higher-level abstract theorizing or reasoning behind the action. Another point of contention involves the definition of ToM and empathy. While the simulation process is often thought of as synonymous with the empathy (which is defined as an ability to identify another person’s emotions and thoughts, and to respond to these with an appropriate emotion [Baron-Cohen, 2003; Wheelwright et al., 2006]), ToM and empathy may be different functionally as well as neuro-psychologically [Singer, 2006]. ToM has been conceptualized as the cognitive component of an empathizing system, the residual component being affective empathy [Chakrabarti and Baron-Cohen, 2006; Davis, 1994]. It is clear that affective empathy is not required in the type of ToM tasks used in many neuroimaging studies of ToM. It has been suggested that affective empathy and ToM tasks, especially FB tasks, may tap different psychological capacities. Specifically, affective empathy may be more related to the ability to infer and react to the emotions of others whereas ToM may be more narrowly related to reading the intentions and beliefs of others [Singer, 2006]. However, neuroimaging research using paradigms that tapped empathy has found brain activity in the candidate ToM regions (i.e., mPFC and ACC) [Jackson et al., 2005; Jackson et al., 2006; Lamm et al., 2007; Völlm et al., 2006]. These results suggest that there may not be a clear psychological boundary between empathizing and ToM and that at least in some brain regions these two have functional overlaps [see Oberman and Ramachandran, 2007]. Future research will have to investigate further details of these different levels of processing and neural networks involved in each. A final but important point is that there is no direct evidence of the human mirror neurons that are homologous to the machaque brain. While it is easy to assume that humans have the same mirror neurons as the monkeys, there is no direct evidence of human mirror neurons that respond to action [Agnew et al., 2007]. Thus, the association among human mirror neuron, simulation and ToM remains speculative until further anatomical as well as functional characterizations, are done.

Linguistic Determinism of ToM Despite the observation that the FB reasoning seems to rely upon some verbal ability, all of the above main-stream theories of ToM (except the theory-theory to a limited extent) downplay the contribution of linguistic ability to ToM development. Fortunately, there is another (and less recognized) theory of ToM that emphasizes this component of ToM and is called “linguistic determinism” [de Villiers, 2000; de Villiers and de Villiers, 2000]. These theorists follow the Whorfian theory of language and cognition, positing that ToM develops as language develops in children. A strong form of this hypothesis proposes that linguistic

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

16

Chiyoko Kobayashi Frank

(primarily syntactic and semantic) ability constrains ToM ability. Main advocates of this hypothesis, de Villiers and her colleagues maintain that a sophisticated command of syntax, or more precisely, complement, is necessary for FB task understanding [de Villiers, 2000]. They argue that 3-year-olds fail the FB task because their syntactic commands are not fully developed to handle the propositions embedded in the sentences of the FB task. De Villiers and her colleagues tested 3-to-4 years-old children for FB and syntactic ability, and found correlation between the two [de Villiers, 2000]. Moreover, a longitudinal study [Astington and Jenkins, 1999] has found that earlier language (syntax and semantics) ability predicts later ToM performance and not vice versa, indicating that language competence is a prerequisite for competitive ToM performance. Strong Linguistic determinists (as above) emphasize the contribution of the command in syntax to ToM. However, there is another version of linguistic determinism which emphasizes the contribution of communicative development and the role of socio-cultural experience for ToM. This alternative version stresses non-theoretical, direct interpersonal or social knowledge for ToM understanding [Hobson, 1991; Tomasello, 2003]. Although the proponents of this moderate version of linguistic determinism are in agreement with the idea that language plays a major role in ToM, they argue that any constitutive aspects of language do not constrain ToM development. Their focus is more on culture-specific socio-communicative aspects than constitutive aspects of language and therefore follows the Vygotskyan tradition [see Vygotsky, 1967, and Valsiner, 1989] more closely. A few studies have supported this second version by demonstrating that performance on a FB task is improved by discussing perspectives on the same objects or events with others without any use of the sentential complements [Lohman and Tomasello, 2003; Harris, 2005]. Moreover, a recent developmental study showed that children with high functioning autism or Asperger Syndrome are impaired in pragmatic aspects of language without any syntactic or semantic impairment [Colle et al., 2008]. Neurological studies that examined the relationship between neural correlates of ToM and those of language have obtained mixed results similar to what behavioral studies have found. On the one hand, a severe aphasic patient, who had a wide-range of left hemisphere damage, showed intact performance in some nonverbal ToM tasks, despite failing all other syntax-related tasks [Varley and Siegal, 2000]. On the other hand, evidence suggests that processing of pragmatically coherent sentences also recruits the mPFC area primarily [Ferstl and von Cramon, 2002]. Ours is the first study to compare the effects of language / culture on ToM development, and our results are consistent with this hypothesis. We found clear cultural / linguistic effects on the neural basis of ToM [Kobayashi et al., 2006; 2007b], and at least the cultural effects had little to do with syntax (as both cultural groups saw exactly the same cartoons). Thus, both behavioral and brain imaging results seem to support the notion that language is important for ToM not because of its constitutive aspects (i.e. syntax and semantics) but because of its pragmatic aspects. Consistent with the second version of linguistic determinism of ToM, in several recent developmental studies of ToM and related social / cognitive functions children employed some of the “language regions” for processing the story-based [Kobayashi et al., 2007a; 2007b; Wang et al., 2006b] and / or the cartoon / animation-based tasks [Dapretto et al., 2006; Kobayashi et al., 2007a; 2007b; Moriguchi et al., 2007; Ohnishi et al., 2004; Wang et al., 2006a] (see Table 1). In addition, a study found some interference on the FB reasoning by

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Cultural and Linguistic Influence on Developmental Neural Basis of Theory…

17

verbal shadowing in adults [Newton and de Villiers, 2007]. These results suggest that adults process ToM more verbally than children and ToM develops as language regions develop.

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Language Regions and ToM As described above, increasing neurological as well as psychological evidence suggests that neural correlates of language and ToM interact in development. The classical language regions encompass Broca’s area or Brodmann area (BA) 44 / 45 in the IFG, Wernicke’s area or BA 22 in the superior temporal gyrus (STG), and the angular gyrus and supramarginal gyrus (SMG) [Carter, 1998]. Thanks to new findings from brain imaging studies on various linguistic processing, there is now a broad consensus that syntactic processing is subserved by the left STG and the IFG (specifically, BA 44 and frontal operculum) in adults [Bornkessel et al., 2005; Friederici et al., 2003; Moro et al., 2001]. Semantic processes, in contrast, are supported by the left MTG, the SMG, and BA 45 / 47 in the IFG in adults [Kotz et al., 2002; Poldrack et al., 1999; see also Figure 1 for the language regions of the brain]. Brain imaging studies of language in children or on development are still too scarce to reach a broad agreement on the specialization of those linguistic processes in the different areas of the brain. But increasing evidence suggests that language is less lateralized and involves broader regions in children than in adults [Brauer and Friederici, 2007]. In terms of the involvement of the IFG area in the ToM development, several neuroimaging studies in children found activity in the IFG while the child participants engaged in facial imitation [Dapretto et al., 2006], story- and cartoon-based irony [Wang et al., 2006a; 2006b] and ToM [Kobayashi et al., 2007a; Moriguchi et al., 2007] tasks (see Table 1). We found a three-way interaction in the left IFG (BA 45) for children and adults [Kobayashi et al., 2007a]. Children employed this area more for the cartoon ToM condition, yet the adults used this area more for the story ToM condition. Similarly, in Wang et al. (2006a), children recruited the left IFG (BA 44 and 45) more than adults. Moriguchi et al. (2007) also found activity in the right IFG (BA 45) when children / adolescents processed animation-based ToM task. These results may suggest that the IFG is important for ToM processing during childhood because of its role as a language center. Several recent neuroimaging studies of ToM and related social cognition studies in children implicated the MTG [Kobayashi, 2007; Moriguchi et al., 2007; Onishi et al., 2004; Wang et al., 2006a] (see Table 1). A few studies found bilateral [Onishi et al., 2004] or left MTG activity [Moriguchi et al., 2007] during animation-based ToM conditions in children / adolescents relative to control conditions. In our study, Japanese children showed more activity in the left MTG during the cartoon ToM condition, yet Japanese adults showed more activity in the same area during the story ToM condition [Kobayashi, 2007; 2008]. Similarly, in Wang et al. (2006a), adults used this area more for processing the cartoon-based irony task than children. Thus, children may use the MTG area more for the visual-based ToM and other related social cognition tasks because they try harder to convert the meanings of the cartoons / animations into words and sentences. The STG is not among the ToM regions [see Figure 1; see also Frith and Frith, 2003, and Saxe et al., 2004]. However, a few recent imaging studies have implicated the right STG area for some functions that may be related to ToM: i.e., empathy mapping through facial and hand-gesture imitations [Leslie et al., 2004] and reading eye-gaze directions [Akiyama et al.,

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

18

Chiyoko Kobayashi Frank

2006]. Also, increasing evidence from neurobiological studies suggests that the STG has reciprocal connections with both the mPFC area and the parahippocampal gyrus [see Price, 2005]. Onishi et al. (2004) found activity in the bilateral STG area when 7-13 year-old children processed animation-based ToM tasks. In our study, we found story and cartoon task-specific interaction in the left STG between the American adult and child groups: adults used this area more for the cartoon ToM condition, while children used this area more for processing the story ToM condition [Kobayashi, 2007a]. Similarly, Wang et al. (2006a) found a greater activity in the right STG in children while they processed the cartoon-based irony task than in adults. It has been shown that language processing tasks normally recruit the left hemisphere, but processing of pragmatics employs the right hemisphere [Paradis, 1998]. Also, results of a few studies indicated that processing of story-based FB tasks is like processing pragmatically coherent sentences [Ferstl and von Carmon, 2002; Siegal et al., 1996]. Thus, our speculation is that children use this area more than adults for understanding pragmatics in the ToM (and related socio-cognitive) stories more than adults do. The TPJ area that has been implicated in the more recent brain imaging studies of ToM [Saxe and Kanwisher, 2003; Saxe and Wexler, 2005; Kobayashi et al., 2007a] may include SMG and angular gyrus if we consider it as a region with 10 or more square-centimeter surface as suggested by Saxe (2006). Most of the ToM neuroimaging studies that have implicated the TPJ area have used story-based tasks [Gallagher et al., 2000; Kobayashi et al., 2007a; Saxe and Kanwisher, 2003; Saxe and Wexler, 2005]. Therefore, it is possible that the TPJ-SMG / angular gyrus network is recruited for ToM processing especially when the task requires some sentential semantic / syntactic analyses.

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Universal Developmental Mechanism of ToM As I discussed above, evidence from recent brain imaging of ToM and related social cognition in children seems to support neither the strong modular nor the simulation 2 hypothesis of ToM . The proponents of the other two developmental theories of ToM have had no neurological evidence to support or refute their theories, primarily because of scarcity of neurological studies of ToM in children. A few neuroimaging studies of ToM and related social and cognitive functions in children begin to address some of the predictions these theories might make about the neural basis of ToM. Our results [Kobayashi et al., 2007a] and the result of several others [Liu et al., 2005; Moriguchi et al., 2007; Wang et al., 2006a] support some aspects of the theory-theory hypothesis. The theory-theory (in opposition to the modular theory) would predict major differences between adults’ folk psychology and children’s ToM [Wellman et al., 2001]. Consistent with that prediction, age-related differences in the neural basis of ToM and related social and cognitive functions have been found [Kobayashi et al., 2007a; 2007b; Moriguchi et al., 2007; Wang et al., 2006a] (see Table 1). The theory-theory would also predict that adults’ ToM is based more on a cognition-laden “belief” concept, whereas children’s ToM is based more on an emotion-laden “desire” concept [Bartch and Wellman, 1995]. Results from a few studies have also supported this notion, finding that children recruit the vmPFC more than the dorsal mPFC [Kobayashi et 2

But, as I discussed above, it is still arguable that modular “core” ToM mechanism is present early in development [see Siegal et al., in press].

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Cultural and Linguistic Influence on Developmental Neural Basis of Theory…

19

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

al., 2007a; Liu et al., 2005; Moriguchi et al., 2007]. As described above, the ventral aspect of the ACC and mPFC has been hypothesized to be involved in emotional processing and the dorsal aspect in cognitive processing. Recent findings from brain imaging studies in children have also supported aspects of the second version of the linguistic determinism hypothesis of ToM, showing differential recruiting of language regions during ToM processing depending on the age of the participants and modality of the task (see above). Thus, I have constructed a neural developmental model of ToM incorporating aspects of both of these hypotheses of ToM development (see Figure 2). Although this model has limitations due to the limitations of our study (and of a few other developmental brain imaging studies on ToM and / or related social cognition) and scarcity of developmental data in the ToM neuroimaging literature overall, it attempts to incorporate the findings that have been presented here.

Figure 2. A universal developmental model of ToM neural bases. During childhood, we recruit the vmPFC mainly for understanding ToM more emotionally (through “desire” concept). We acquire language as several language regions of the brain develop. The end result is our folk psychology, which is more cognition or “belief” based and recruits the dorsal mPFC primarily.

I hypothesize that during childhood (before the age of 12), ToM understanding is more closely linked to understanding of emotions, and that young children employ the vmPFC or the ventral part of PFC to process the highly emotion-laden ToM (see Figure 2). At this earlier stage especially, younger children may understand ToM more in terms of the “desire” concept than through the “belief” concept. During late childhood through adolescence, our linguistic ability continues to develop as various language brain regions (e.g., the IFG, MTG, STG, and angular gyrus) mature and language continues to influence ToM development during this time. Finally, in adulthood, ToM is understood in terms of the “belief” concept, and the dorsal mPFC is employed to understand the cognition-laden ToM. However, this model is subject to further empirical testing. It has been suggested that the dorsal mPFC activity is associated with various externally-directed social cognitions, while the ventral

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

20

Chiyoko Kobayashi Frank

mPFC is associated with self-referential process [Mitchell et al., 2005]. Thus, alternatively, the ventral-emotion versus dorsal-cognition dichotomy is a bi-product of externally-directed versus internally-directed social cognition dichotomy. Future neuroimaging studies that will systematically test both children (especially children of ages between 3 and 5 when ToM performance dramatically improves) and adults with either internally- or externally-directed “desire” and “belief” related tasks will be useful for examining how the activity in the different brain regions is associated with the development of different concepts related to ToM and the self-referential process. Further testing of children and adults with various language processing tasks in combination with those tasks will be helpful for examining precisely what aspects of language play the quintessential role for conceptual development and ToM.

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Difference between Anglo-American Culture and Other Cultures in the Developmental Mechanism of ToM The model that I have described above does not reflect some of the cross-cultural (or cross-linguistic) differences in the neural bases of ToM that we observed in our study. Thus, I have modified the model to account for these cultural differences (see Figure 3) A key difference between Anglo-American culture and other cultures in the developmental trajectory of ToM seems to be the extent of TPJ’s involvement in ToM. One main function of the TPJ may be to distinguish self-agency from other-agency [Blakemore and Frith, 2003; Jackson and Decety, 2004; see also Decety and Grézes, 2006]. As I described above, results of recent neuroimaging studies of ToM [Kobayashi et al., 2006; 2007b, 2008) and self [Zhu et al., 2006] indicate that our culturally unique ways of self-construal may affect how we perceive others’ mental states and intentions throughout development. More specifically socio-psychological as well as neuroimaging evidence suggests that Japanese / Asians may make less distinction between self and others than Anglo-Americans because Japanese / Asians embrace the relational self concept (as opposed to the independent self in Anglo-Americans). More overlap of the activation in the mPFC (which has consistently been implicated in self referential processing [see Ochsner et al., 2004; 2005]) was found when Chinese participants judged self-related and intimate other-related adjectives than Americans [Zhu et al., 2006]. My colleagues and I found a greater ToM cartoon specific activation in the TPJ (which has consistently been implicated in processing a third-person perspective or “other” agent [D’Argembeau et al., 2007; David et al., 2006]) in American than Japanese children, despite the fact that both groups of children viewed the same ToM cartoon [Kobayashi et al., 2007b]. Thus, my ToM developmental model for Japanese / Asians differs from that for Anglo-Americans by this one point; less recruitment of the TPJ by the Japanese / Asians throughout development, because of their diminished sense of self and other distinction (see Figure 3). However, for this model, I use the term “other cultures” since, as I mentioned earlier, majority of the ToM brain imaging studies in German and French adults also implicated the mPFC and not the TPJ. Precisely why many other neuroimaging studies of ToM conducted in continental European countries did not find the ToM-specific activity in the TPJ is yet to be examined. It has been demonstrated that, unlike Anglo-American children, French children rarely used subjective “belief” concept to justify the behaviors in

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Cultural and Linguistic Influence on Developmental Neural Basis of Theory…

21

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

FB task (Bradmetz, 1998). Thus, it may be case that the continental European cultures, too, conceptualize ToM in more intersubjective way (therefore, less agent specific way) than Anglo-American culture.

Figure 3. Culture-specific developmental models of ToM neural bases. The ToM developmental trajectory for other cultures (i.e., Asians and some continental Europeans) is the same as that for AngloAmerican cultures except that the bilateral TPJ is not employed due to Asian and some continental European people’s diminished “self” and “other” distinction.

Conclusions Through this chapter, I aimed to examine the Whorfian relativism (i.e., relationship between language / culture and thoughts) in ToM and related social cognition. Especially, I focused on the effects of the relativism at the neural level. In terms of cultural effect, recent studies that compared Asians (Japanese or Chinese) and Westerners in ToM or self construal processing found some evidence to support the Whorfian hypothesis, finding that people understand ToM and judge themselves or intimate-others in culture specific ways. Specifically, the research results indicate that people’s recruitment of the mPFC or TPJ during the ToM and self judgment task is dependent upon their cultural backgrounds. In terms of linguistic effect, a strong form of the Whorfian relativism predicts that a person’s world view and concepts are direct products of the particular language that the person speaks. The studies described above have found evidence of some influence of language on ToM development. This review also considered the extent to which these research findings support or contradict the major developmental theories of ToM; modular, theory, theory, simulation, and linguistic determinism theories. I have demonstrated that these findings are largely in

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

22

Chiyoko Kobayashi Frank

agreement with the theory-theory and a non-extreme version of linguistic determinism of ToM; that non-constitutive aspects of language may constrain ToM development. Thus, I have suggested a new developmental neural model of ToM based upon these two theories. According to this model, children recruit the vmPFC primarily for understanding ToM more emotionally (via. the “desire” concept). As we age, our language matures together with several language regions of the brain, leading to the adults’ folk psychology, which is more cognition-based and recruits the mPFC primarily. Our second developmental model of ToM reflects recent findings from cross-cultural and / or developmental neuroimaging studies of ToM and related social cognition; that language and culture affect ToM throughout development. The main difference between AngloAmericans and other cultures in the developmental mechanism of ToM may be in how “self” and “other” agencies are perceived for understanding ToM in each culture. The diminished self and other distinction in Japanese / Asians (and some other non-Anglo-American cultural groups) may be reflected in their lesser recruitment of the TPJ regions for processing ToM compared to Anglo-Americans, who tend to make more clear-cut distinction between “self” and “other”. A clear limitation of current cross-cultural ToM research is that, as Vinden (1999) and Lillard (1998) pointed out, the standard FB task may not be valid for all cultures because many non-Anglo-American cultures do not construe behaviors as internal / intentional. Results of Naito and Koyama (2007) have shown that many Japanese children rarely give desire-based explanation to account for the false-belief of the protagonist. These results (together with Vinden’s [1999] results) call into question the applicability of the developmental order of ToM concepts – from “desire” based understanding to “belief” based understanding (Wellman and Liu, 2004) – to all cultures. Taken together, the results from the developmental and / or cross-cultural neuroimaging studies of ToM and related social cognition support the mild version of the Whorfian hypothesis: one’s language and culture influence one’s thoughts. They also highlight that the development of ToM is a dynamic, socio-interactive and, perhaps, domain-general process involving linguistic, affective and cognitive development. However, this is not to deny that universal (and perhaps non-linguistic) “core” ToM component is present early in human development. In fact it may be more plausible to assume such domain-specific component exists. Our study was an initial effort to dissociate universal from culture / languagedependent component of ToM. More ToM neuro-developmental studies on younger children, infants, and non-human primates are needed to advance the theoretical models of ToM development.

Acknowledgments I thank Dr. Elise Temple, Dr. Simon Baron-Cohen, Dr. Gary H. Glover, Dr. Barbara C. Lust, Dr. Michael Siegal, and Dr. Michael J. Spivey for discussion. I also thank Randall Frank for assistance. Parts of this chapter have been published in a different form in “Language and thought: Cultural and linguistic neural bases of theory of mind”, VDM Verlag: Saarbrücken, Germany.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Cultural and Linguistic Influence on Developmental Neural Basis of Theory…

23

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

References Abraham, A., Werning, M., Rakoczy, H., von Cramon, D. Y., & Schubotz, R. I. (in press) Minds, persons, and space: An fMRI investigation into the relational complexity of higher-order intentionality. Conscious & Cognition. Agnew, Z. K., Bhakoo, K. K., & Puri, B. K. (2007). The human mirror system: A motor resonance theory of mind-reading. Brain Research Reviews, 54, 286-293. Akiyama, T., Kato, M., Muramatsu, T., Saito, F., Nakachi, R., & Kashima, H. (2006). A deficit in discriminating gaze direction in a case with right superior temporal gyrus lesion. Neuropsychologia, 44(2),161-170. Ames, D. R., Knowles, E. D., Morris, M. W., Kalish, C. W., Rosati, A. D., & Gopnik, A. (2001). The social folk theorist: Insights from social and cultural psychology on the contents and contexts of folk theorizing. In Malle, B. F., Moses, L. J., & Baldwin, D. A. (Eds.), Intentions and intentionality (pp. 208-329). Cambridge, MA: MIT Press. Apperly, I.A. (2008). Beyond Simulation-Thoery-theory: Why social cognitive neuroscience should use its own concepts to study “theory of mind”. Cognition, 107, 266-283. Apperly, I. A., Samson, D., & Humphreys, G. W. (2009). Studies of adults can inform accounts of theory of mind development. Developmental Psychology, 45(1), 190-201. Apperly, I. A., Williams, E., & Williams, J. (2004). Three to four-year-olds’ recognition that symbols have a stable meaning: pictures are understood before written words. Child Development, 75(5), 1510-1522. Astington, J.W., & Jenkins, J.M. (1999). A longitudinal study of the relation between language and theory-of-mind development. Developmental Psychology, 35(5), 13111320. Avis, J., & Harris, P.L. (1991). Belief-desire reasoning among Baka children: Evidence for a universal conception of mind. Child Development, 62, 460-467. Baron-Cohen, S. (2000). Theory of mind and autism: A fifteen years. In: Baron-Cohen, S., Tager-Flusberg, H., & Cohen, D. (Eds.), Understanding other minds: Perspective from autism (2nd edition, pp. 3-20). Oxford: Oxford University Press. Baron-Cohen, S. (2003). The essential difference: men, women and the extreme male brain. London: Penguin. Baron-Cohen, S., Leslie, A.M., & Frith, U. (1985). Does the autistic child have a “theory of mind”? Cognition, 21, 37-46. Baron-Cohen, S., Leslie, A.M., & Frith, U. (1986). Mechanical, behavioral and intentional understanding of picture stories in autistic children. British Journal of Developmental Psychology, 4, 113-125. Baron-Cohen, S., Ring, H. A., Wheelwright, S., Bullmore, E. T., Brammer, M. J., Simmons, A., & Williams, S. C. (1999). Social intelligence in the normal and autistic brain: an fMRI study. European Journal of Neuroscience, 11, 1891-1898. Barrett, H. C., & Kurzban, R. (2006). Modularity in cognition: Framing the debate. Psychological Review, 113, 628-647. Bartsch, K., & Wellman, H. (1995). Children talk about the mind. Oxford University Press, New York. Berthoz, S., Armony, J.L., Blair, R.J.R., & Dolan, R.J. (2002). An fMRI study of intentional and unintentional (embarrassing) violation of social norms. Brain, 125, 1696-1708.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

24

Chiyoko Kobayashi Frank

Bird, C. M., Castelli, F., Malik, O., Frith, U., & Husain, M. (2004). The impact of extensive medial fontal lobe damage on ‘theory of mind’ and cognition. Brain, 127, 914-928. Blakemore, S.-J., & Frith, C.D. (2003). Self-awareness and action. Current Opinion in Neurobiology, 13, 219-224. Bornkessel, I., Zysset, S., Friederici, A. D., von Cramon, D. Y., & Schlesewsky, M. (2005). Who did what to whom? The neural basis of argument hierarchies during language comprehension. Neuroimage, 26, 221-233. Boroditsky, L. (2001). Does language shape thought? Mandarin and English speakers’ conception of time. Cognitive Psychology, 43, 1-22. Bowerman, M., & Choi, S. (2001). Shaping meanings for language: Universal and languagespecific in the acquisition of spatial semantic categories. In Bowerman, M., & Levinson, S. C. (Eds.), Language acquisition and conceptual development, (pp. 475-511). Cambridge: Cambridge University Press. Bradmetz, J. (1998) Acquisition of a theory of mind and intellectual development in the child aged 3 to 5. Current Psychology & Cognition, 17, 95-113. Brambilla, P., Harden, A. Y., di Nemi, S. U., Caverzasi, E., Soares, J. C., Perez, J., & Barale, F. (2004). The functional neuroanatomy of autism. Functional Neurology, 19(1), 9-17. Brauer, J., & Friederici, A. D. (2007). Functional neural networks of semantic and syntactic processes in the developing brain. Journal of Cognitive Neuroscience, 19(10), 16091623. Brown, P., & Levinson, S. C. (1993). Linguistic and nonlinguistic coding of spatial arrays: Exploration in Mayan cognition. Working Paper 24. Nijmegen: Max Planck Institute for Psycholinguistics, Cognitive Anthropology Research Group. Brunet, E., Sarfati, Y., Hardy-Bayle, M-C., & Decety, J. (2000). A PET investigation of the attribution of intentions with a nonverbal task. Neuroimage, 11, 157-166. Brunet, E., Sarfati, Y., Hardy-Baylé, M.-C., & Decety, J. (2003) Abnormalities of brain function during a nonverbal theory of mind task in schizophrenia. Neuropsychologia, 41, 1574-1582. Burgess, P. W., Quayle, A., & Frith, C. D. (2001). Brain regions involved in prospective memory as determined by position emission tomography. Neuropsychologia, 39(6), 545555. Bush, G., Luu, P., & Postner, M. J. (2000). Cognitive and emotional influences in anterior cingulate cortex. Trends in Cognitive Science, 4(6), 215-222. Calarge, C., Andreasen, N.C., & O’Leary, D.S. (2003). Visualizing how one brain understands another: A PET study of theory of mind. American Journal of Psychiatry, 160(11), 1954-1964. Callaghan, T., Rochat, P., Lillard, A., Claux, L., Odden, H., Itakura, S., Tapanya, S., & Singh, S. (2005). Synchrony in the onset of mental-state reasoning. Psychological Science, 16(5), 378-384. Carter, R. (1998). Mapping the mind. University of California Press, Berkeley. Chakrabarti, B., & Baron-Cohen, S. (2006). Empathizing: neurocognitive developmental mechanisms and individual differences. Progress in Brain Research, 156, 403-417. Chaminade, T., & Decety, J. (2002). Leader or follower? Involvement of the inferior parietal lobule in agency. Neuroreport, 13(15), 1975-1978. Chen, M.J., & Lin, Z.X. (1994). Chinese preschoolers’ difficulty with theory-of-mind tests. Bulletin of the Hong Kong Psychological Society, 32/33, 34-46.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Cultural and Linguistic Influence on Developmental Neural Basis of Theory…

25

Choi, S., & Bowerman, M. (1991). Learning to express motion events in English and Korean: The influence of language-specific lexicalization patterns. Cognition, 41, 83-121. Chomsky, N. (1980). Rules and representations. Oxford: Basil Blackwell. Clancy, P. M. (1985). The acquisition of Japanese. In Slobin, D. I. (Ed.), The crosslinguistic study of language acquisition (Vol. 2, pp. 373-524) Hillsdale, NJ: Lawrence Erlbaum Associates. Clancy, P. M. (1986). The acquisition of communicative style in Japanese. In Schieffelin, B. B., & Ochs, E. (Eds.), Language socialization across cultures (pp. 213-250). Cambridge: Cambridge University Press. Cohen, D, & Gunz, A. (2002). As seen by the other… : Perspectives on the self in the memories and emotional perceptions of Easterners and Westerners. Psychological Science, 13, 55-59. Colle, L., Baron-Cohen, S., Wheelwright, S., ven der Lely, H. K. J. (2008). Narrative discourse in adults with high-functioning autism or Asperger syndrome. Journal of Autism and Developmental Disorders, 38, 28-40. Corbetta, M., Patel, G., Shulman, G. L. (2008). The reorienting system of the human brain: From environment to theory of mind. Neuron, 58, 306-324. Craik, F. I. M., Moroz, T. M., Moscovitch, M., Stuss, D. T., Winocur, G., Tulving, E., & Kapur, S. (1999). In search of the self: A position emission tomography study. Psychological Science, 10, 26-34. Dapretto, M., Davies, M. S., Pfeifer, J. H., Scott, A. A., Sigman, M., Bookheimer, S. Y., & Iacoboni, M. (2006). Understanding emotions in others: mirror neuron dysfunction in children with autism spectrun disorders. Nature Neuroscience, 9(1), 28-30. D’Argembeau, A., Ruby, P., Collette, F., Degueldre, C., Balteau, E., Luxen, A., Maquet, P., & Salmon, E. (2007). Distinct regions of the medial prefrontal cortex and perspective taking. Journal of Cognitive Neuroscience, 19(6), 935-944. David, N., Bewernick, B. H., Cohen, M. X., Newen, A., Lux, S., Fink, G. R., Shah, N. J., & Vogeley, K. (2006). Neural representation of self versus other: Visual-spatial perspective taking and agency in a virtual ball-tossing game. Journal of Cognitive Neuroscience, 18(6), 898-910. Davis, M. H. (1994). Empathy: a social psychological approach. Boulder, CO: Westview Press. Decety, J., & Chaminade, J. (2004). When self represents the other: A new cognitive neuroscience view of psychological intention. Conscious and Cognition, 12, 577-596. Decety, J., & Grézes, J. (2006). The power of simulation: imaging one’s own and other’s behavior. Brain Research, 1079(1), 4-14. Decety, J., Grézes, J., Costes, N., Perani, D., Jeannerod, M., Procyk, E. et al., (1997). Brain activity during observation of actions. Influence of action content and subject’s strategy. Brain, 120(10), 1763-1777. Dennett, D.C. (1980). The milk of human intentionality (commentary on Searle). Behavioral and Brain Sciences, 3, 428-30. Den Ouden, H. E. M., Frith, U., Frith, C., & Blakemore, S.-J. (2005). Thinking about intentions. Neuroimage, 28, 787-796. De Villiers, J.G. (2000). Language and theory of mind: What are the developmental relationships? In Baron-Cohen, S., Tager-Flusberg, H., & Cohen, D. (Eds.),

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

26

Chiyoko Kobayashi Frank

Understanding other minds: Perspective from autism (2nd edition, pp. 83-123). Oxford: Oxford University Press. De Villiers, J.G., & de Villiers, P.A. (2000). Linguistic determinism and the understanding of false beliefs. In Mitchell, P., & Riggs, K. J. (Eds.) Children’s reasoning and the mind (pp.191-228). East Sussex, UK: Psychology Press. Ferstl, E.C., & von Cramon, D.Y. (2002). What does the frontomedian cortex contribute to language processing: coherence or theory of mind? Neuroimage, 17, 1599-1612. Fine, C., Lumsden, J., & Blair, R. J., 2001. Dissociation between ‘theory of mind’ and executive function in a patient with early left amygdala damage. Brain, 124, 287-298. Fletcher, P.C., Happé, F., Frith, U., Baker, S. C., Dolan, R. J., Frackowiak, R. S. J., & Frith, C. D. (1995). Other minds in the brain: A functional imaging study of ‘theory of mind’ in story comprehension. Cognition, 57, 109-128. Fodor, J. A. (1983). The modularity of mind. Cambridge, MA: MIT Press. Fogassi, L., Ferrari, P. F., Gesierich, B., Rozzi, S., Chersi, F., & Rizzolatti, G. (2005). Parietal lobe: From action organization to intention understanding. Science, 308, 662-667. Friederici, A. D., Rüschemeyer, S. A., Hahne, A., & Fiebach, C. J. (2003). The role of left inferior frontal and superior temporal cortex in sentence comprehension: Localizing syntactic and semantic process. Cerebral Cortex, 13, 170-177. Friston, K. J., Holmes, A. P., Price, C. J., Buchel, C., & Worsley, K.J. (1999). Multisubject fMRI studies and conjunction analyses. Neuroimage, 10(4), 385-396. Frith, U. (2003). Autism: Explaining the enigma. Malden, MA: Blackwell Publishing. Frith, U., & Frith, C. D. (2003). Development of neurophysiology of mentalizing. Philosophical Transactions of the. Royal Society of. London, Series B, 358(1431), 459473. Gallagher, H.L., & Frith, C.D. (2003). Functional imaging of ‘theory of mind.’ Trends in Cognitive Science, 7(2), 77-83. Gallagher, H.L., Happé, F., Frunswick, N., Fletcher, P.C., Frith, U., & Frith, C.D. (2000). Reading the mind in cartoons and stories: an fMRI study of ‘theory of mind’ in verbal and nonverbal tasks. Neuropsychologia, 38, 11-21. Gallagher, H.L., Jack, A.I., Roepstorff, A., & Frith, C.D. (2002). Imaging the intentional stance in a competitive game. Neuroimage, 16, 814-821. Geschwind, D. H., & Levitt, P. (2007). Autism spectrum disorders developmental disconnection syndromes. Current Opinion in Neurobiology, 17, 103-111. Goel, V., Grafman, J., Sadato, N., & Hallet, M. (1995). Modeling other minds. Neuroreport, 6(13), 1741-1746. Goldman, A. (1989). Interpretation psychologized. Mind and Language,. 4, 161-185. Gopnik, A., & Wellman, H.M. (1992). Why the child’s theory of mind really is a theory. Mind and Language, 7(1-2), 145-171. Goushiki, T. (1999) Preschool children’s understanding of self-mind and other-mind using “theory of mind” tasks. Japanese Journal of Educational Psychology, 47, 354-363. Grèzes, J., Frith, C. D., & Passingham, R. E. (2004). Inferring false beliefs from the actions of oneself and others: an fMRI study. Neuroimage, 21, 744-750. Gusnard, D. A., Akbudak, E., Shulman, G. L., & Raichle, M. E. (2001). Medial prefrontal cortex and self-referential mental activity: relation to a default mode of brain function. Proceedings of the National Academy of Sciences, USA, 98(7), 4259-4264.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Cultural and Linguistic Influence on Developmental Neural Basis of Theory…

27

Gusnard, D. A., & Raichle, M. E. (2001). Searching for a baseline: functional imaging and the resting human brain. Nature Reviews Neuroscience, 2(10), 685-694. Happé, F.G. (1993). Communicative competence and theory of mind in autism: a test of relevance theory. Cognition, 48, 101-119. Happé, F., Ehler, S., Fletcher, P., Frith, U., Johansson, M., Gillberg, C., Dolan, R., Frackowiak, R., & Frith, C. (1996). ‘Theory of mind’ in the brain: evidence from a PET scan study of Asperger syndrome. Neuroreport, 8, 197-201. Harris, P. L. (1991). The work of the imagination. In A. Whitman (Ed.), Natural theories of mind: The evolution, development and simulation of everyday mindreading (pp. 283304). Oxford: Oxford University Press. Harris, P. L. (1992). From simulation to folk psychology: The case for development. Mind and Language,. 7, 120-144. Harris, P. L. (2005). Conversation, pretense and theory of mind. In Astington, J. W., & Baird, J. A. (Eds.), Why language matters for theory of mind (pp. 70-83). Oxford: Oxford University Press. Harris, P.L., & Gross, D. (1996) Children’s understanding of real and apparent emotion. In: Astington, J.W., Harris, P.L. & Olson, D.R. (Eds.), Developing theories of mind. (pp. 295-314). New York: Cambridge University Press. Hayes, C.M. (1998). Theory of mind in nonhuman primates. Behavioral and Brain Sciences, 21, 101-134. Heine, S. J. (2001). Self as cultural product: an examination of East Asian and North American elves. Journal of Personality, 69, 881-906. Hobson, R. (1991). Against the theory of “Theory of Mind.” British Journal of Developmental Psychology, 9, 33-51. Iacoboni, M. (2005). Neural mechanisms of imitation. Current Opinion in Neurobiology, 15, 632-637. Iacoboni, M., Woods, R.P., Brass, M., Bekkering, H., Mazziotta, J.C., & Rizzolatti, G. (1999). Cortical mechanisms of human imitation. Science, 286, 2526-2528. Ide, S. (1977). Sex and personal referents of Japanese children. Proceedings of the Symposium on Japanese Sociolinguistics, Summer Institute of the Linguistic Society of America. University of Hawaii, Honolulu. Jackson, P. L., Brunet, E., Meltzoff, A. N., & Decety, J. (2006). Empathy examined through the neural mechanisms involved in imagining how I feel versus how you feel pain. Neuropsychologia, 44(5), 752-761. Jackson, P.L., & Decety, J. (2004). Motor cognition: a new paradigm to study self other interactions. Current Opinion in Neurobiology, 14, 259-263. Jackson, P. L., Meltzoff, A. N., & Decety, J. (2005). How do we perceive the pain of others? A window into the neural processes involved in empathy. Neuroimage, 24(3), 771-779. Kain, W., & Perner, J. (2005). What fMRI can tell us about the ToM-EF connection: falsebeliefs, working memory, and inhibition. In Schneider, W., Schumann-Hengsteler, R., & Sodian, B. (Eds.), Young children’s cognitive development (pp. 189-217). Erlbaum, Mahwah, NJ: Lawrence Erlbaum. Kelly, W. M., Macrae, C. N., Wyland, C. L., Caglar, S., Inati, S., & Heatherton, T. F. (2002). Finding the self? An event-related fMRI study. Journal of Cognitive Neuroscience,. 14, 785-794.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

28

Chiyoko Kobayashi Frank

Kim, J-W., Kim, J-J., Jeong, B. S., Ki, S. W., Im, D-M., Lee, S. J., & Lee, H. S. (2005). Neural mechanism for judging the appropriateness of facial affect. Cognitive Brain Research, 25, 659-667. Klein, S. B., Loftus, J., & Burton, H. A. (1989). Two self-reference effects: The importance of distinguishing between self-descriptiveness judgments and autobiographical retrieval in self-referent encoding. Journal of Personality and Social Psychology, 56, 853-865. Kobayashi, C. (2007). Language and thought: Cultural and linguistic influence on developmental neural bases of theory of mind. Ph.D Thesis. Cornell University, NY. Kobayashi, C. (2008). Language and thought: Cultural and linguistic influence on developmental neural bases of theory of mind. Saarbruecken, Germany: VDM Verlag. Kobayashi, C., Glover, G. H., & Temple, E. (2006). Cultural and linguistic influence on neural bases of ‘Theory of Mind’: An fMRI study with Japanese bilinguals. Bran and Language, 98(2), 210-220. Kobayashi, C., Glover, G. H., & Temple, E. (2007a). Children’s and adults’ neural bases of verbal and nonverbal ‘Theory of Mind’. Neuropsychologia, 45, 1522-1532. Kobayashi, C., Glover, G. H., & Temple, E. (2007b). Cultural and linguistic effects on neural bases of ‘Theory of Mind’ in American and Japanese children. Brain Research, 1164, 95107. Kobayashi, C., Glover, G. H., & Temple, E. (2008). Switching language switches mind: linguistic effects on developmental neural bases of ‘Theory of Mind’. Social, Cognitive and Affective Neuroscience, 3, 62-70. Kotz, S. A., Cappa, S. F., von Cramon, D. Y., & Friederici, A. D. (2002). Modulation of the lexical-semantic network by auditory semantic priming: An event-related functional MRI study. Neuroimage, 17, 1761-1772. Lamm, C., Batson, C. D., & Decety, J. (2007). The neural substrate of human empathy: effects of perspective-taking and cognitive appraisal. Journal of Cognitive Neuroscience, 19(1), 42-58. Lee, K., Olson, D.R., & Torrance, N. (1999). Chinese children’s understanding of falsebeliefs: The role of language. Journal of Child Language, 26, 1-21. Leslie, A.M. (1992). Autism and the ‘theory of mind’ module. Current Directions in Psychological Science, 1, 18-21. Leslie, A.M. (1994). Pretending and believing: issues in the theory of ToMM. Cognition, 50, 211-238. Leslie, A. M., Friedman, O., & German, T. P. (2004). Core mechanisms in 'theory of mind'. Trends in Cognitive Science, 8(12), 529-533. Leslie, K. R., Johnson-Frey, S. H., & Grafton, S. T. (2004). Functional imaging of face and hand imitation: Towards a motor theory of empathy. Neuroimage, 21(2), 601-607. Lieberman, M. D., Jarcho, J. M., & Satpute, A. B. (2004). Evidence-based and intuition-based self-knowledge: an fMRI study. Journal of Personality and Social Psychology, 87, 421435. Lillard, A. (1998). Ethnospychologies: Cultural variations in theories of mind. Psychological Bulletin, 123, 3-32. Lissek, S., Peters, S., Fuchs, N., Witthaus, H., Nicolas, V., Tegenthoff, M., Juckel, G., & Brüne, M. (2008) Cooperation and deception recruit different subsets of the theory-ofmind network. PLoS One, 3(4), 1-11.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Cultural and Linguistic Influence on Developmental Neural Basis of Theory…

29

Liu, D. (2005). Neural correlates of children’s theory of mind development. Ph.D Thesis. University of Michigan, MI. Liu, D., Sabbagh, M. A., Gehring, W. J., & Wellman, H. M. (2005), April. An ERP study of 6-year-olds’ theory of mind. Poster presented at the 2005 annual meeting of the Cognitive Neuroscience Society, New York City, NY. Liu, D., Wellman, H. M., & Sabbagh, M.A. (2008). Theory of mind development in Chinese children: A meta-analysis of false-belief understanding across cultures and language. Developmental Psychology, 44(2), 523-531. Lohman, H., & Tomasello, M. (2003). The role of language in the development of false belief understanding: A training study. Child Development, 74, 1130-1144. Lord, C. G. (1980). Schemas and images as memory aids: Two modes of processing social information. Journal of Personality and Social Psychology, 38, 257-269. Louis, B.J. (1998). Does Childhood Bilingualism facilitate the Development of Theory of Mind? Psychology Ph.D. Thesis. Yale University, New Haven, CT. Lucy, J. A. (1992). Grammatical categories and cognition: A case study of the linguistic relativity hypothesis. Cambridge: Cambridge University Press. Malle, B. F. (2002). The relation between language and theory of mind in development and evolution. In: Govon, T., & Malle, B.F. (Eds.), The evolution of language our of prelanguage (pp. 265-284). Amsterdam: Benjamins. Markus, H., & Kitayama, S. (1991). Culture and the self: Implications for cognition, emotion, and motivation. Psychological Review, 98, 224-253. Masuda, T., & Nisbett, R.E. (2001). Attending holistically vs. analytically: Comparing the context sensitivity of Japanese and Americans. Journal of Personality and Social Psychology, 81, 922-934. Maynard, S. K. (1997). Japanese communication: Language and thought in context. University of Hawai’i Press: Honolulu. McCabe, K., Houser, D., Ryan, L., Smith, V., & Trouard, T. (2001). A functional imaging study of cooperation in two-person reciprocal exchange. Proceedings of the National Academy of Sciences, USA, 98, 11832-11835. Meltzoff, A. N., & Brooks, R. (2001). “Like Me” as a building block for understanding other minds: Bodily acts, attention, and intention. In Malle, B. F., Moses, L J., & Baldwin, D. A. (Eds.), Intentions and intentionality (pp. 171-191). Cambridge, MA: MIT Press. Mitchell, J. P. (2008). Activity in right temporo-parietal junction is not selective for theoryof-mind. Cerebral Cortex, 18, 262-271. Mitchell, J. P., Banaji, M. R., & Macrae, C. N. (2005). The link between social cognition and self-referential thought in the medial prefrontal cortex. Journal of Cognitive Neuroscience, 17, 1306-1315. Moriguchi, Y., Ohnishi, T., Mori, T., Matsuda, H., & Komaki, G. (2007). Changes of brain activity in the neural substrates for theory of mind during childhood and adolescence. Psychiatry and Clinical Neurosciences, 61, 355-363. Moro, A., Tettamanti, M., Perani, D., Donati, C., Cappa, S. F., & Fazio, F. (2001). Syntax and the brain: Disentangling grammar by selective anomalies. Neuroimage, 13, 110-118. Müller, R. A. (2007). The study of autism as a distributed disorder. Mental Retardation and Developmental Disabilities Research Reviews, 13(1), 85-95.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

30

Chiyoko Kobayashi Frank

Naito, M. (2003). The relationship between theory of mind and episodic memory: Evidence for the development of autonoetic consciousness. Journal of Experimental Child Psychology, 85, 312-336. Naito, M. (2007) Current status of theory of mind research and future direction (Kokoro no shinri no genjyo to kongo no kadai). Annual Review of Japanese Child Psychology, Kaneko Shobou, 47, 1-37, (in Japanese). Naito, M., Komatsu, S., & Fuke, T. (1994). Normal and autistic children’s understanding of their own and others’ false belief: A study from Japan. British Journal of Developmental Psychology, 12, 403-416. Naito, M., & Koyama, K. (2006). The development of false-belief understanding in Japanese children: Delay and difference? International Journal of Behavioral Development, 30(4), 290-304. Nakamura, A., Maess, B., Knösche, T. R., Gunter, T. C., Bach, P, & Friederici, A. D. (2004). Cooperation of different neuronal systems during hand sign recognition. Neuroimage, 23, 25-34. Newton, A. M., & de Villiers, J. G. (2007). Thinking while talking. Psychological Science, 18(7), 574-579. Nichols, S., & Stich, S. P. (2003). Mind reading: an integrated account of pretence, selfawareness, and understanding other minds. Oxford: Oxford University Press. Nisbett, R.E. (2003). The geography of thought. New York: The Free Press. Oberman, L. M., & Ramachandran, V. S. (2007). The simulating social mind: The role of the mirror neuron system and simulation in the social and communicative deficits of Autism Spectrum Disorders. Psychological Bulletin, 133, 310-327. Ochs, E. (1985). Introduction. In Schieffelin, B. B., Ochs, E. (Eds.), Language socialization across cultures (pp. 1-13). Cambridge: Cambridge University Press. Ochsner, K. N., Knierim, K., Ludlow, D. H., Hanelin, J., Ramachandran, T., Glover, G., & Mackey, S. C. (2004). Reflecting upon feelings: an fMRI study of neural systems supporting the attribution of emotion to self and other. Journal of Cognitive Neuroscience, 16(10), 1746-1772. Ochsner, K. N., Beer, J. S., Robertson, E. R., Cooper, J. C., Gabrieli, J. D. E., Kihsltrom, J. F., & D’Esposite, M. (2005). The neural correlates of direct and reflected self-knowledge. Neuroimage, 28, 797-814. Ohnish, T., Moriguchi, Y., Matsuda, H., Mori, T., Hirakata, M., Imabayashi, E., Hirao, K., Nemoto, K., Kaga, M., Inagaki, M., Yamada, M., & Uno, A. (2004). The neural network for the mirror system and mentalizing in normally developed children: an fMRI study. Neuroreport, 15(9), 1483-1487. Onishi, K. H., & Baillargeon, R. (2005). Do 15-month-old infants understand false beliefs? Science, 308, 255-258. Paradis, M. (1998). The other side of language: Pragmatic competence. Journal of Neurolinguistics, 11, 1-10. Perner, J., & Ruffman, T. (2005). Infants’ insight into the mind: How deep? Science, 308, 214-216. Piaget, J. (1962). Play, dreams, and imitation in childhood. New York: Norton. Poldrack, R. A., Wagner, A. D., Prull, M. W. Desmond, J. E., Glover, G. H., & Gabrieli, J. D. (1999). Functional specialization for semantic and phonological processing in the left inferior frontal cortex. Neuroimage, 10, 15-35.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Cultural and Linguistic Influence on Developmental Neural Basis of Theory…

31

Premack, D., & Woodruff, G. (1978). Does the chimpanzee have a theory of mind? Behavioral and Brain Sciences, 1, 515-526. Price, J. L. (2005). Free will versus survival: Brain systems that underlie intrinsic constrains on behavior. Journal of Comparative Neurology, 493, 132-139. Roth, D., & Leslie, A. M. (1998). Solving belief problems: towards a task analysis. Cognition, 66, 1-31. Ruby, P., & Decety, J. (2001). Effect of subjective perspective taking during simulation of action: a PET investigation of agency. Nature Neuroscience, 4(5), 546-550. Ruby, P., & Decety, J. (2003). What you believe versus what you think they believe: a neuroimaging study of conceptual perspective-taking. European Journal of Neuroscience, 17, 2475-2480. Ruby, P., & Decety, J. (2004). How would you feel versus how do you think she would feel? A neuroimaging study of perspective-taking with social emotion. Journal of Cognitive Neuroscience, 16(6), 988-999. Sabbagh, M. A. (2004). Understanding orbitofrontal contributions to theory-of-mind reasoning: Implications for autism. Brain and Cognition, 55, 209-219. Sabbagh, M.A., & Taylor, M. (2000). Neural correlates of theory-of-mind reasoning: an event-related potential study. Psychological Science, 11(1), 46-50. Samson, D., Apperly, I.A., Chiavarino, C., & Humphreys, G. W. (2004). Left temporoparietal junction is necessary for representing someone else’s belief. Nature Neuroscience, 7(5), 499-500. Samson, D., Apperly, I. A., & Humphreys, G. W. (2007). Error analyses reveal contrasting deficits in “theory of mind”: Neuropsychological evidence from a 3-option false belief task. Neuropsychologia, 45, 2561-2569. Sanfey, A. G., Rilling, J. K., Aronson, J. A., Nystrom, L. E., & Cohen, J. D. (2003). The neural basis of economic decision-making in the ultimatum game. Science, 300, 17551758. Saxe, R. (2005). Against simulation: the argument from error. Trends in Cognitive Sciences, 9(4), 174-179. Saxe, R., Carey, S., & Kanwisher, N. (2004). Understanding other minds: linking developmental psychology and functional neuroimaging. Annual Review of Psychology, 55, 87-124. Saxe, R., & Kanwisher, N. (2003). People thinking about thinking people: the role of the temporo-parietal junction in “theory of mind.” Neuroimage, 19, 1835-1842. Saxe, R., & Wexler, A. (2005). Making sense of another mind: the role of the right temporoparietal junction. Neuropsychologia, 43, 1391-1399. Scholl, B. J., & Leslie, A. M. (1999). Modularity, development and ‘theory of mind.’ Mind and Language, 14(1), 131-153. Scholz, J., Triantafyllou, C., Whitfield-Gabrieli, S., Brown, E. N., & Saxe, R. (2009). Distinct regions of right temporo-parietal junction are selective for theory of mind and exogenous attention. PLoS One, 4(3), e4869. Schultz, R.T., Romanski, L.M., & Tsatsanis, K.D. (2000). Neurofunctional models of autistic disorder and Asperger syndrome: Clues from neuroimaing. In Klin, A., Volkmar, F. R., & Sparrow, S. S. (Eds.), Asperger syndrome (pp. 172-209). New York: Guilford Press.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

32

Chiyoko Kobayashi Frank

Shatz, M., Diesendruck, G., Martinez-Beck, I., & Akar, D. (2003). The influence of language and socioeconomic status on children’s understanding of false belief. Developmental Psychology, 39(4), 717-729. Siegal, M., Carrington, J., & Radel, M. (1996). Theory of mind and pragmatic understanding following right hemisphere damage. Brain and Language, 53, 40-50. Siegal, M., Frank, C. K., Surian, L., Hjelmquist, E. (in press). Theory of Mind and Bilingual Cognition. Sinclair, M. (1970). The transition from sensory-motor behavior to symbolic activity. Interchange, 1, 119-126. Singer, T. (2006). The neuronal basis and ontogeny of empathy and mind reading: Review of literature and implications for future research. Neuroscience and Biobehavioral Reviews, 30, 855-863. Sommer, M., Döhnel, K., Sodian, B., Meinhardt, J., Thoemer, C., & Hajak, G. (2007). Neural correlates of true and false belief reasoning. Neuroimage, 35, 1378-1384. Southgate, V., Senju, A., & Csibra, G. (2007). Action anticipation through attribution of false belief by 2-year-olds. Psychological Science, 18(7), 587-592. Surian, L., Caldi, S, & Sperber, D. (2007). Attribution of beliefs by 13-month-old infants. Psychological Science, 18(7), 580-586. Takahashi, H., Yahata, N., Koeda, M., Matsuda, T., Asai, K., & Okubo, Y. (2004). Brain activation associated with evaluative processes of guilt and embarrassment: an fMRI study. Neuroimage, 23, 967-974. Tardiff, T., & Wellman, H.M. (2000). Acquisition of mental state language in Mandarin- and Cantonese-speaking children. Developmental Psychology, 36 (1), 25-43. Tomasello, M. (2003). The key is social cognition. In D. Gentner, & S. Goldin-Meadow (Eds.), Language in mind: Advances in the study of language and thought (pp. 48-57). Cambridge, MA: MIT Press. Valsiner, J. (1989). Human development and culture: The social nature of personality and its study. Lexington, MA: Lexington Books. Varley, R., & Siegal, M. (2000). Evidence for cognition without grammar from causal reasoning and ‘theory of mind’ in an agrammatic aphasic patient. Current Biology, 10, 723-726. Venkatraman, V., Siong, S. C., Chee, M. W., & Ansari, D. (2006). Effect of language switching on arithmetic: a bilingual FMRI study. Journal of Cognitive Neuroscience, 18(1), 64-74. Vinden, P. (1996). Junin Quechua children’s understanding of mind. Child Development, 67, 1707-1716. Vinden, P. G. (1999). Children’s understanding of mind and emotion: A multi-culture study. Cognition & Emotion, 13, 19-48. Völlm, B. A., Taylor, A. N. W., Richardson, P., Corcoran, R., Stirling, J., McKie, S., Deakin, J. F. W., & Elliot, R. (2006). Neural correlates of theory of mind and empathy: a functional magnetic resonance imaging study in a nonverbal task. Neuroimage, 29(1), 9098. Vogeley, K., Bussfeld, P., Newen, A., Herrmann, S., Happé, F., Falkai, P., Maier, W., Shah, N.J., Fink, G.R., & Zilles, K. (2001). Mind reading: Neural mechanisms of theory of mind and self-perspective. Neuroimage, 14, 170-181.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Cultural and Linguistic Influence on Developmental Neural Basis of Theory…

33

Vygotsky, L. S. (1967). Thought and language. In Hanfmann, E., & Vakar, G. (Eds. and Trans.). Cambridge, MA: MIT Press. Walter, H., Adenzato, M., Ciaramidaro, A., Enrici, I., Pia, L., & Bara, B. G. (2004) Understanding intentions in social interaction: The role of the anterior paracingulate cortex. Journal of Cognitive Neuroscience, 16(10), 1854-1863. Wang, A. T., Lee, S. S., Sigman, M., & Dapretto, M. (2006a). Developmental changes in the neural basis of interpreting communicative intent. Social, Cognitive and Affective Neuroscience, 1, 107-121. Wang, A. T., Lee, S. S., Sigman, M., & Dapretto, M. (2006b). Neural basis of irony comprehension in hcildren with autism: the role of prosody and context. Brain, 129(4), 932-943. Wellman, H.M., Cross, D.C., & Watson, J. (2001). Meta-analysis of theory of mind development: The truth about false belief. Child Development, 72(3), 655-684. Wellman, H. M., & Liu, D. (2004). Scaling of theory-of-mind tasks. Child Development, 75, 523-541. Werner, H., & Kaplan, B. (1963). Symbol formation. New York: John Wiley and Sons. Wheelwright, S., Baron-Cohen, S., Goldenfeld, N., Delaney, J., Fine, D., Smith, R., Weil, L., & Wakabayashi, A. (2006). Predicting autism spectrum quotient (AQ) from the systemizing quotient-revised (SQ-R) and empathy quotient (EQ). Brain Research, 1079, 47-56. Whorf, B.L. (1956). Language, thought, and reality: Selected writing of Benjamin Lee Whorf. (J.B. Carroll, ed.). Cambridge, MA: MIT Press. Wimmer, H., & Mayringer, V. (1998). False belief understanding in young children: Explanations do not develop before predictions. International Journal of Behavioral Development, 22, 403-422. Wimmer, H., & Perner, J. (1983). Belief about belief: representation and constraining function of wrong beliefs in young children’s understanding of deception. Cognition, 13, 103-128. Wu, S., & Keysar, B. (2007). The effect of culture on perspective taking. Psychological Science, 18(7), 600-606. Yazdi, A. A., German, T. P., Defeyter, M. A., & Siegal, M. (2006). Competence and performance in belief-desire reasoning across two cultures: The truth, the whole truth and nothing but the truth about false belief? Cognition, 100(2), 343-368. Zhu, Y., & Zhang, L. (2002). An experimental study on the self-reference effect. Science in China, Series C 45, 120-128. Zhu, Y., Zhang, Li., Fan, J., & Han, S. (2007). Neural basis of cultural influence on selfrepresentation. Neuroimage, 34, 1310-1316.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved. Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

In: Bilinguals: Cognition, Education and Language Processing ISBN: 978-1-60741-710-1 Editor: Earl F. Caldwell, pp. 35-64 © 2010 Nova Science Publishers, Inc.

Chapter 2

LANGUAGE PROCESSING IN BIMODAL BILINGUALS Anthony Shook and Viorica Marian Northwestern University, Illinois, USA

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Abstract Recent research suggests differences between bimodal bilinguals, who are fluent in a spoken and a signed language, and unimodal bilinguals, who are fluent in two spoken languages, in regard to the architecture and processing patterns within the bilingual language system. In this chapter, we discuss ways in which sign languages are represented and processed and examine recent research on bimodal bilingualism. It is suggested that sign languages display processing characteristics similar to spoken languages, such as the existence of a sign counterpart to phonological priming and the existence of a visual-spatial loop analogous to a phonological loop in working memory. Given the similarities between spoken and signed languages, we consider how they may interact in bimodal bilinguals, whose two languages differ in modality. Specifically, we consider the way in which bimodal bilingual studies may inform current knowledge of the bilingual language processing system, with a particular focus on topdown influences, and the fast integration of information from separate modalities. Research from studies looking at both production and perception suggests that bimodal bilinguals, like unimodal bilinguals, process their languages in parallel, with simultaneous access to both lexical and morphosyntactic elements. However, given the lack of overlap at the phonological level (the presumed initial locus of parallel activation in unimodal studies) in bimodal bilinguals’ two languages, we conclude that there are key differences in processing patterns and architecture between unimodal and bimodal language systems. The differences and similarities between unimodal and bimodal bilinguals are placed in the context of current models of bilingual language processing, which are evaluated on the basis of their ability to explain the patterns observed in bimodal bilingual studies. We propose ways in which current models of bilingual language processing may be altered in order to accommodate results from bimodal bilingualism. We conclude that bimodal bilingualism can inform the development of models of bilingual language processing, and provide unique insights into the interactive nature of the bilingual language system in general. “The analytic mechanisms of the language faculty seem to be triggered in much the same ways, whether the input is auditory, visual, even tactual…” -Noam Chomsky (2000, p. 100-101) Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

36

Anthony Shook and Viorica Marian

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

1. Introduction to the Chapter One of the most striking features of bimodal bilingualism (which refers to fluency in both a signed and a spoken language) is the total lack of phonological overlap between the two languages. Bimodal bilinguals are able to create distinct, meaningful utterances with two separate sets of articulators, and have two output channels, vocal and manual. In contrast, unimodal bilinguals, whose two languages are spoken, utilize only one modality for both input and output. Moreover, bimodal bilinguals are able to perceive distinct linguistic information in two domains, via listening to speech and via visually perceiving signs. Although unimodal bilinguals utilize visual information as well, it acts primarily as a cue to facilitate understanding of auditory input, rather than providing a source of visual linguistic input independent from the auditory signal. Research on bimodal bilingualism carries implications for understanding general language processing. For example, one issue at the heart of conceptual modeling of language is the level of influence of bottom-up versus top-down processing. Specifically, when we process language, how much information do we gain from the signal itself (e.g. bottom-up input from phonological or orthographic features) and how much do we gain from higherorder knowledge (e.g., top-down input from background information, context, etc.)? While the existence of both top-down and bottom-up influences is universally acknowledged, the degree to which each holds sway over the language processing system is not entirely clear. Another important question is how bilinguals integrate auditory and visual information when processing language and whether that process differs between unimodal and bimodal bilinguals. To what extent does the ability to retrieve information from two separate modalities facilitate language comprehension? Is information from separate modalities accessed simultaneously or serially? Given the alternate input/output structure found in bimodal bilinguals, and lack of linguistic overlap between signed and spoken languages, it is important to consider what studies about bimodal bilingualism can tell us about bilingual language processing in general. Since the vast majority of bilingual research is performed with unimodal bilinguals, it is somewhat unclear what similarities and differences exist between the two groups. Furthermore, comparing both the structural-linguistic and cognitive-processing aspects of unimodal and bimodal bilingual groups can illuminate the effects of modality on language processing. In the present chapter, we will review recent research on bimodal bilingualism and contrast it with results from unimodal studies in order to expand understanding of bilingual language processing. To study the influence that research on bimodal bilingualism has on both the mechanisms of bilingual processing and the architecture of the underlying language system, we will outline several models of bilingual language processing and examine how well they account for the results seen in recent bimodal bilingual research. The present chapter consists of two main parts. The first part focuses on linguistic and cognitive aspects of sign languages in native signers and in bimodal bilinguals. Specifically, we will (a) compare and contrast how sign languages are represented linguistically, by examining previous work on the structural characteristics of sign languages, (b) discuss the cognitive patterns of sign language, in contrast to spoken language, by examining the similarities between phenomena found in spoken language research with those found in sign language research, and (c) examine results

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Language Processing in Bimodal Bilinguals

37

from studies looking first at language production and then at language perception in bimodal bilinguals, which will be directly contrasted with previous results from unimodal bilingual studies. In the second part of the chapter, we will introduce several models of bilingual language processing, focusing on models of both (a) language production and (b) language perception, and discuss them in light of the results from bimodal bilingual studies. We will conclude by suggesting that spoken and signed languages are represented similarly at the cognitive level and interact in bimodal bilinguals much the same way two spoken languages interact in unimodal bilinguals, and that bimodal bilingual research can highlight both the strengths and weaknesses of current models of bilingual language processing.

2. Representation and Processing of Sign Languages

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

2.a. Structure of Sign Languages Current models that explain the structure of sign languages are primarily based on the study of American Sign Language, and its contrast with spoken English. It is important to note that just as spoken languages differ in phonology, morphology and syntax, so do sign languages. For example, not only do American Sign Language (ASL) and British Sign Language (BSL) utilize separate lexicons, they display morphological distinctions as well, such as BSL’s use of a two-handed finger-spelling system compared to ASL’s one-handed finger-spelling system (Sutton-Spence and Woll, 1999). There are also phonological distinctions, in that phonological segments in one language do not necessarily exist in the other, much like in spoken languages (sign languages use handshape, location of the sign in space, and motion of the sign as phonological parameters). We will focus mainly on the phonological aspects of sign language structure, while briefly discussing certain morphosyntactic traits. Given that much of the current body of knowledge about sign language structure is based on American Sign Language, we will focus specifically on the relationship between the phonologies of ASL and spoken English, in order to highlight some of the fundamental differences between spoken and signed languages in general. The most salient difference between signed and spoken languages is that signed languages exist in a spatial environment, and are expressed grammatically through the manipulation of the body, notably the hands and face, within a linguistic sign-space, which is a physical area centered around the front of the speaker’s body. Like actors on a stage, the mechanics of grammar occur within this sign-space. This variance in articulatory location and modality results in interesting syntactic differences. While English uses prepositional information to determine the location and relation of object, ASL creates schematic layouts of objects within the sign-space to determine their relationship in space, as well as to show motion. For example, rather than describe the movement of some object from point A to point B, the lexical item is presented and then physically moved. Syntactically, movement of verbs within the sign-space can determine the objects and subjects of sentences. Consider, for instance, the sign “give.” Whereas in English the difference between “I give you X” and “You give me X” is determined by word order and thematic role (subject “I” versus object “me”), ASL differentiates the two through variation in direction of movement of the verb. The two nominal entries are placed in the sign space, and the sentence structure is determined by whether the speaker moves the “give” sign from himself to his interlocutor, or vice versa

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

38

Anthony Shook and Viorica Marian

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

(see Figure 1, top and middle panels). This system also allows for ASL to have a less strict word order than English, given that nominal entries in space can be manipulated freely (Bellugi, Poizner, and Klima, 1989). Secondly, ASL marks tense lexically using temporal adverbs, rather than morphologically marking verbs as in English (i.e., whereas forming the past tense of “walk” in English involves adding a past tense morpheme to create “walked,” in ASL “walk” becomes “walked” by combining the sign for “walk” with a temporal sign like “before” or “yesterday”). One could argue that this, too, is related to the nature of sign-language articulation. If we consider the sign-space as a stage on which lexical items can be placed and physically referenced, then we may consider the addition of temporal adverbs as settingmarkers. This way, rather than consistently marking signs throughout an entire utterance, signers merely need to indicate the time to the listener once.

Figure 1. Signs for the ASL phrases “I give you” and “you give me,” and the word “bite”. All images © 2006, www.Lifeprint.com. Used by permission.

Another interesting difference between ASL and English is the use of facial expressions to mark certain morphosyntactic elements, like relative clauses, topicalization and conditionals (Liddell, 1980). For example, when producing conditional statements in ASL, the signs are accompanied by the raising of the eyebrows. The eyebrows are furrowed downward to accompany wh-questions. Often, combinations of manual and non-manual constructions are used in creating signs (Liddell, 1980) – for example, the sign for “bite” involves both a manual motion and a biting motion of the mouth (see Figure 1, lower panel).

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Language Processing in Bimodal Bilinguals

39

They can also act as suprasegmental features. A head shake can negate all or part of a sentence, even though a manual sign for negation exists as well. Much like English, ASL contains a sublexical phonological system that combines segments according to combinatorial rules, but which uses manual rather than oral features. At the phonological level, current models of ASL recognize three main parameters, which are handshape, location of the sign in space, and movement of the sign (Brentari, 1998; Stokoe, 1960). Each of these parameters can be further broken down into a finite set of phonemes, but each sign contains at least one feature from all three parameters. In other words, a sign consists of a specific handshape that is held in a particular point in the sign space, and is then moved in a particular way. Each parameter can be varied independently of the others, which can result in pairs of signs that match in two parameters and vary only in the third. This differs from spoken language phonology not only in modality of structure, but also in temporal relationships of the features. Since spoken languages unfold sequentially, the phonemes do not overlap within words. In ASL, and other sign languages, the features selected to create a sign do overlap temporally. The handshape used to form a sign occurs simultaneously with the location of the sign in space. As we will discuss later, this temporal overlap can provide a unique avenue into lexical and sublexical selection mechanisms during language processing.

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

2.b. Cognitive Representation of Sign Languages While comparisons of spoken and sign language often focus on the structural differences between the two languages, a number of studies have examined similarities in patterns of cognitive processing across both languages. Here, we discuss studies that compare psycholinguistic phenomena found in spoken language with those found in signed languages. In doing so, we show that there are similarities in the way that linguistic information is processed across signed and spoken languages, which allows us to view them as equally represented on a cognitive level, and therefore examine their interactions in bimodal bilinguals. In order to inspect the way in which phonological information is handled in users of signed languages, Dye and Shih (2006) performed a study that examined the role of phonological priming in British Sign Language (BSL). The authors asked whether a sign could facilitate the activation of a phonologically similar sign. This notion was based on results that suggest spoken words can facilitate the activation of phonologically similar words (Goldinger, Luce, Pisoni and Marcario, 1992; but see Marslen-Wilson, 1990; Praamsta, Meyer and Levelt 1994). Dye and Shih tested the reaction times of monolingual native signers of BSL in a lexical decision task where sign targets were preceded by primes that shared none, one, two, or all three parameters of sign language phonology. They found that native signers were significantly faster at correctly naming lexical items when the preceding prime overlapped in location, as well as in location and movement (but not in other dimensions). Interestingly, the authors used both signs and non-signs to test the priming effect and found that native BSL signers only showed priming in response to real signs. This implies that the locus of lexical priming is actually at the level of the lexicon in native signers – while in English, non-words are capable of priming words, this is not the case with native signers of

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

40

Anthony Shook and Viorica Marian

BSL. Were the features themselves producing the lexical priming effect, one would expect to see priming due to heavily overlapping non-signs. However, the same paradigm used with non-native signers showed non-sign-to-sign priming, suggesting that the effect cannot be attributed exclusively to the lexicon. This suggests qualitatively different processes for handling input when considering the model of human speech perception. It is unclear whether the difference arises due to changes in processing patterns while maintaining standard architecture, or if the structure of the language processing system as a whole is different when developed around visual/gestural languages. Given models of lexical access like MarlsenWilson and Welsh’s Cohort Theory (1978) that conceive of the build-up of phonological information to activate phonologically similar lexical items, an issue arises with the results yielded by native-signers. Specifically, auditory information is analyzed temporally, such that the input activates all lexical items that match, and over time as the system gets more information, the number of activated potential targets decreases. If this theory held true, native signers should show the same non-sign-to-sign priming effect as non-native signers, due to the fact that features that overlap with items in the signer’s lexicon are presented. However, it appears that the featural information alone is not enough to promote the activation of overlapping lexical items. It seems then that the nature of language acquisition influences the development of the language processing system. This raises several questions. First, to what extent is the system able to alternate between one lexical access mechanism versus the other – in other words, do non-native signers ever access lexical items the same way native signers do, or vice versa? Dye and Shih’s non-native group was comprised of subjects who had learned BSL later in life, but were born profoundly deaf. It is, therefore, not the case that their processing system was influenced by spoken language processing. Still, the non-native signers’ processing patterns matched those predicted by the Cohort theory better than the native signers. The second question that arises is how bimodal bilinguals might access their lexicons. Age of acquisition of the two languages obviously plays a role, where native signers who learned to speak later in life might process more like Dye and Shih’s native signers, and native speakers who learned a sign language later in life might process more like the nonnative group. However, if a person learns both languages simultaneously, the predictions become less clear. One possibility is that lexical access could become task based, utilizing both processes in different circumstances. Slowiaczek and Pisoni (1986) tested monolingual English speakers in a lexical decision priming task and their results contradicted the predictions of the Cohort model – initial phonemes between prime and target actually caused inhibition of lexical access. However, initial phoneme overlap showed facilitation of response in identification-in-noise tasks. Phonological overlap affected the system differently across tasks. The same concept could be generalized to bimodal bilinguals, which could suggest that simultaneous bimodal bilinguals access their lexicon differently dependent on the nature of the input. Another possibility is that the system simply uses both mechanisms for lexical access simultaneously. It is possible to think of the features of sign language phonology as analogous to certain features of spoken language phonology. Voice-onset time (VOT) can act as a cue to certain phonemes, and studies have shown that listeners are able to use fine-grained VOT information during lexical access to determine the word being spoken (McMurray, Tanenhaus, and Aslin, 2002). Dye and Shih (2006) showed that location and movement act as more salient cues to priming than handshape in non-native signers. So, perhaps the bimodal

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Language Processing in Bimodal Bilinguals

41

system is able to simultaneously process the featural information of signs, while simultaneously utilizing the lexical-level access shown by native signers in the study. To examine this, one could replicate the priming paradigm utilized by Dye and Shih and compare groups of Native-ASL speakers, Native-English speakers and ASL-English bilinguals. Previous research has provided evidence for the existence of a visuospatial articulatory loop in working memory for sign languages. In users of a spoken language, the phonological loop in verbal working memory consists of a phonological storage buffer, which holds phonological information in working memory, and a rehearsal process, which refreshes the items in the storage buffer, preventing them from fading quickly (Gathercole and Baddeley, 1993). The evidence for the existence and subsequent separation of these two components comes from experimental effects such as the phonological similarity effect, the word-length effect, and articulatory suppression. If the same effects could be seen in users of signlanguage, this would suggest that the cognitive system is capable of treating spatial sensorimotor information as it would language information from an auditory modality. In other words, specific linguistic experience shapes the kind of input that the phonological loop deems relevant, but does not necessarily change the way the system functions. So do the phonological similarity, word-length and articulatory suppression effects occur in signlanguage users? The phonological similarity effect refers to the phenomenon that words in a list that share phonological information are more difficult to recall than words that are phonologically diverse. Research has shown that lists of signs that contain the same handshape show worse recall than lists with diverse handshapes (Krakow and Hanson, 1985; Wilson and Emmorey, 1997). This provides evidence for a phonological similarity effect in ASL. The word-length effect shows that lists of long words are harder to recall than lists of short words. Wilson and Emmorey (1998) tested ASL signers with lists consisting of signs with long movement, and lists of signs with short, local movement. Since movement is a physical process, it requires more time to make large movements within a sign than it does to make small movements, thus increasing the temporal load on the listener. The results showed that lists of temporally long signs were recalled worse than those with short signs. Lastly, articulatory suppression is the effect in which repetition of phonemes or syllables that use relevant articulators disrupts the rehearsal mechanism of the phonological loop. The result is worse performance with suppression than without. This has also been shown in speakers of ASL (Wilson and Emmorey, 1997). When subjects were asked to make motoric hand movements (alternating fist and open hand) during list memorization, they showed worse recall than when they kept their hands still. In the same study, Wilson and Emmorey showed no interaction between articulatory suppression and phonological similarity, which is in accord with results from spoken language studies and suggests that the two tasks affect different aspects of the phonological loop. There is also compelling evidence to suggest that children learning ASL as a native language develop much like children learning a spoken language. Children learning ASL reach the milestones of language development at about the same rate as those learning spoken languages (Bonvillian and Folven, 1993; Pettito and Marentette, 1991). One well-documented developmental phenomenon originates in Werker and Tees’ (1984) study suggesting that infants under one-year of age were capable of discriminating sounds from their non-native language, but lost that ability as they grew older. This is an example of categorical perception, the phenomenon by which sounds are placed in phonemic categories and listeners are unable

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

42

Anthony Shook and Viorica Marian

to distinguish between sounds that fall within the same category. The categories available to any given listener are based upon the relevant phonemes of the listener’s native language. In regards to ASL signers, this raises two questions. First, does ASL display categorical perception? Emmorey, McCullough and Brentari (2003) examined whether native deaf signers displayed categorical perception based on differences in handshape. Much like in auditory studies, the experimenters developed continua of signs using still images, where the endpoints represented prototypical productions of some handshape. They found that native signers demonstrated categorical perception for the handshape stimuli, but a group of hearing non-signers did not. This result was corroborated by Baker, Idsardi, Golinkoff, and Petitto (2005), who also found that native ASL signers categorized handshapes linguistically, rather than on a purely perceptual basis. The second question is whether or not this ability develops similarly in signers and speakers. Baker, Golinkoff, and Petitto (2006) found that 4-month old hearing infants, who were not learning ASL, displayed categorical perception of handshape stimuli based on linguistic properties of the input, while 14-month infants failed to do so. This suggests a pattern of perceptual shift for ASL that is nearly identical to spoken languages, where infants initially have the capacity to linguistically categorize perceptual input, but lose this ability as their perceptual system becomes specialized to their own language. Taken together, these studies suggest strong similarities between users of spoken languages and signed languages on a cognitive level. Sign languages show many of the same phenomena as spoken languages, such as lexical priming, categorical perception, and the presence of an articulatory loop, suggesting that spoken and signed languages may be processed by a similar language mechanism. In the next section, we examine the way two languages that differ in modality interact within bilinguals who are fluent in both.

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

2.c. Processing Patterns in Bimodal Bilinguals It has become commonly accepted in the field of bilingual study that unimodal bilinguals activate their two languages in parallel (Blumenfeld and Marian, 2007; Canseco-Gonzalez, et al., 2005; Marian and Spivey, 2003; Weber and Cutler, 2004). However, little work has been done to suggest that bimodal bilinguals activate their languages in parallel as well. Emmorey and colleagues (Casey and Emmorey, 2008; Emmorey, Borinstein, Thompson, and Gollan, 2008) have performed a series of experiments looking at how ASL might be active during production of English in bilingual users of both languages. In several studies, bimodal bilinguals were asked to tell a story to listeners, in English. Listeners were either known to be bilinguals as well (Emmorey et al., 2008), or their language background was unknown to the speaker (Casey and Emmorey, 2008). During English production, the experimenters recorded the hand gestures that were spontaneously created by the bilingual speakers. The results showed that bimodal bilinguals produced a significant number of what Emmorey and colleagues refer to as code-blends, which are semantically related signs inserted simultaneously with the related lexical item in speech. While bimodal bilinguals produced code-blends with both groups of listeners, they produced more code-blends when the listener was known to be a bimodal bilingual. The findings from these studies suggest that bimodal bilinguals access their non-target language during production, even up to the point where semantically related signs are produced concomitantly with speech. This is in contrast with some previous work that

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Language Processing in Bimodal Bilinguals

43

suggests unimodal bilingual lexical access during speech is language-specific (e.g., Costa and Caramazza, 1999, but see Colomé, 2001). If lexical access during production is languagespecific in unimodal bilinguals, but language-independent in bimodal bilinguals, does that imply that the bimodal bilingual processing system is, in some way, different from that of the unimodal bilingual, beyond the surface distinction of modality of input? Emmorey et al. (2008) discuss one possible reason for the potential difference between unimodal and bimodal bilinguals in language production. While unimodal bilinguals are physiologically limited to using one language at a time (it is impossible to express a concept in both French and English simultaneously, for example), bimodal bilinguals face no such limitation. It is possible that the processing architecture is the same, but bimodal bilinguals are able to exploit the cross-modal nature of their languages to make use of a skill that unimodal bilinguals possess, but cannot access. Determining whether bimodal bilinguals process language differently than unimodal bilinguals or whether unimodal bilinguals and bimodal bilinguals process language similarly (i.e., the two groups employ the same mechanisms, but unimodal bilinguals are unable to produce both of their languages simultaneously due to biological constraints) has important implications for modeling the bilingual system. There is also evidence to suggest that in addition to simultaneous production of lexical items, higher-order, ASL-specific morphosyntactic features can be found when bimodal bilinguals produce English. Pyers and Emmorey (2008) examined the interaction of ASL facial expressions with English grammatical constructions. ASL uses facial expressions to mark certain grammatical features of sentences (e.g., furrowed brows accompany whquestions, raised eyebrows occur with conditionals, etc). The authors found that when bimodal bilinguals are speaking English, they produce grammatically relevant facial expressions simultaneously. Furthermore, the authors recorded the timing of facial productions and found that the facial expressions were very closely time-locked with the English grammatical constructions. This implies that bimodal bilinguals utilize a language system that integrates grammatical information from both languages at the same time, rather than separating the syntactical systems of the two languages (Hartsuiker, Pickering, and Veltkamp, 2004). Another group of bimodal bilinguals, which Dufour (1997) refers to as sign-text bilinguals, can also provide a unique window into bilingual language processing. Sign-text bilinguals are those who are fluent in a sign language, as well as the written form of a spoken language. While sign-text bilinguals do not display the salient characteristic of both speaking and signing a language, they nevertheless process two languages with very different grammatical structures. Unfortunately, little work has been done to directly contrast the two language forms in sign-text bilinguals. However, as Dufour explains, there are studies that use text-based input as the stimuli or materials for studies with deaf signers. While these studies fail to control for the wide range of relevant variables found to influence bilingual status (such as age of acquisition, proficiency, etc.), they nevertheless provide an insight into the interaction between the two modalities. Hanson (1982) examined the interaction between the signed and written modality. ASLEnglish sign-text bilinguals were presented with lists of words that were structurally similar (i.e., all the signs were similar), phonologically similar (the English translations had similar sounds) or orthographically similar (the English words were spelled similarly). The deaf participants were separated into two groups, where one received the word lists as signs, and

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

44

Anthony Shook and Viorica Marian

the other received the lists as English written forms. Subjects performed a probe recall task, where an item from one of the lists was given, and subjects had to provide the item that followed it in the list. Hanson found that subjects who received the sign presentation showed worse recall to the phonologically and structurally similar lists, compared to control lists matched for frequency and item length. Signers who received the English lists showed significantly worse recall only for the phonologically similar list. Hanson’s study, from a bilingual standpoint, consists of two conditions – L1 recall of signs, and L2 recall of English word forms. If the L1 and L2 lexicons were separate (or at least privy to separate input), then we would expect to see signers who received the signed lists perform worse only with the structurally similar lists (akin to the phonological similarity effect), since the phonologically and orthographically related lists differed based on non-L1 features. However, this was not the case, which suggests that when native signers were presented with signs, they encoded the information in both L1 and L2 phonological structure. In other words, the perception and subsequent encoding of a sign activated structures in the L2, even though the two languages did not share modality. However, the results from the signers who received English word lists suggest something different – here, the input form only seems to disrupt recall in lists with relevant similarities, namely phonology of English. Since Hanson was not intending to study the interaction between languages, she did not document characteristics of her subjects that could have affected the results, such as whether they were taught via total communication or oralist methodology1. The reason this is relevant is because it is unclear how the signers had access to the kind of distinctive phonological information found in English if they were unable to hear those distinctions. Previous work has focused on the functional equivalence of sign and speech in learning phonological information (Hanson, 1982; Leybaert, 2000; Miller, 2004), suggesting that underlying phonological representations can be equally tapped by sign or speech input. However, more recent work by McQuarrie and Parrila (2008) seems to suggest this is not the case; when rhyme judgments based on phonological information are contrasted with those based on visual or motor/tactile information, “phonological” facilitation is not seen. There are, then, two possible explanations for Hanson’s result. First, perhaps the participants in her study were trained in an oralist tradition, and did in fact have some motoric knowledge of the way in which to produce certain phonemes, and that phono-motor information was enough to map onto some underlying phonological structure. Another possibility is that the effects were caused by stimulus design – the list involving phonologically similar items was simply more difficult by default. If we accept the latter explanation, we’re forced to ignore the phonologically similar lists, and are left with the result that when the words are presented as signs, they are encoded as sign structures, and perhaps more troubling, that words presented orthographically result in no specific encoding at all (it is unclear why the orthographically related lists did not also show interference, given that they would appear to be the most salient). Regardless of which explanation is correct, we are left with a result that suggests single-language encoding in certain conditions, and therefore, separate lexicons. 1

Total Communication, commonly referred to as Simultaneous Communication, refers to the practice of using various methods of communication (e.g., signing, writing, oral, etc.) to educate children. Oralism refers to the practice of teaching deaf children to pronounce spoken languages, and to understand spoken languages via lipreading. Oralism, by default, involves teaching the child to produce spoken language, and Total Communication often involves an oral component.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Language Processing in Bimodal Bilinguals

45

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Along with sign-text bilinguals, Dufour also mentions another understudied group – speech-sign bilinguals, who are more commonly referred to simply as bimodal bilinguals. Unfortunately, little work has been done to look at the processing of perceptual information in speech-sign bimodal bilinguals. In our lab, we are currently examining bimodal bilinguals who are fluent in ASL and spoken English to determine whether both languages are activated in parallel (Shook and Marian, in progress). More specifically, bimodal bilinguals are given a recognition task in which they are shown four images and asked to click on one of them while their eye movements are recorded. In experimental trials, two pictures (a target and a competitor) share three of the four phonological parameters in ASL (see Figure 2).

Figure 2. Example of eye-tracking display from Shook and Marian (in progress). The target item is “chair” which shares three of four parameters (handshape, location, orientation) with the sign for “train,” differing only in motion.

Pilot data suggest that bimodal bilinguals look at competitor items that share phonological similarity to the target more than at semantically and phonologically unrelated control items. This implies that bimodal bilinguals activate both languages simultaneously during language comprehension. One interesting aspect of this result is the prediction that it makes regarding the mechanism with which the non-target language (ASL) is activated. Rather than relying on bottom-up information, the lack of phonological overlap between the two languages necessitates the inclusion of a top-down pathway in order to explain how ASL may be activated during a purely English task (see Figure 3). The finding that bimodal bilinguals activate both languages simultaneously during comprehension is not surprising if we consider how audio and visual information is integrated generally, including in monolinguals and unimodal bilinguals. Likely the most famous

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

46

Anthony Shook and Viorica Marian

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

account of audio-visual integration in monolingual speech recognition involves the McGurk effect (McGurk and MacDonald, 1976) in which an auditory signal, /ba/, is presented with a video of a speaker articulating the syllable /ga/, resulting in subjects reporting perception of the syllable /da/. Marian (2009) discusses how this integration affects bilingual language processing, suggesting that “as a word unfolds, incoming auditory input is combined online with incoming visual input and the two sources mutually interact to exclude options that are not plausible in at least one modality.” It could be argued that bimodal bilinguals are even better attuned than monolinguals or unimodal bilinguals at integrating information from separate modalities, because their processing systems are trained to do just that. In addition, research suggests that users of sign languages show greater visual attention in the periphery, and increased peripheral motion processing (Bavelier, Dye, and Hauser, 2006; Bavelier, Tomann, Hutton, Mitchell, Corina, Liu and Neville, 2000). If bimodal bilinguals are more attentive to the visual field due to their experience with a signed language, we might expect to see a greater influence of visual information in language processing.

Figure 3. A graphical representation of top-down activation patterns in bimodal bilinguals activating both English and American Sign Language during a comprehension task.

A key difference between bimodal and unimodal audio-visual integration arises when one considers the mechanism by which lexical items are activated. In the unimodal case, the net activation of a lexical item is the sum of activation garnered by the phonological (or auditory information) as well as the extra, top-down activation provided by visual information. Conversely, bimodal bilinguals viewing congruous code-blends will activate lexical items in both languages that correspond to the same referent in the semantic system. Activation from both lexical items feeds to the semantic level and results in high activation of the semantic node, which can feed back to both lexical items, causing their activation levels to increase in turn. Note that this implies a significant top-down contribution to increased lexical activation. Not only that, it also implies that simultaneous parallel activation of lexical items between languages that do not share modality in bimodal bilinguals, when considering code-blend contexts, is due to bottom-up information.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Language Processing in Bimodal Bilinguals

47

Conversely, the audio-visual integration concept could work in one of two ways for bimodal bilinguals using only one of their languages. As previously mentioned, eye-tracking studies with unimodal bilinguals have shown that bilingual subjects performing an experiment in their L1 are more likely to look at competitor objects in a visual scene when the L2 lexical item for the competitor overlaps phonologically with the L1 target (e.g., Marian and Spivey, 2003). This is explained via bottom-up accounts where phonological information activates items from both lexicons, which is similar to the mechanism that drives parallel activation in bimodal bilinguals viewing code-blends. However, if bimodal bilinguals using only one language (e.g., spoken English) in these eye-tracking tasks also show parallel activation patterns, it must be due to a top-down mechanism. This can occur in two possible ways. First, as in the case of code-blends, activation from the phonological level feeds upward to the semantic level, which then feeds back down to corresponding lexical entries for both languages. Second, it could occur in the same fashion that Marian (2009) suggests audiovisual integration occurs. While the auditory signal is sending information up the processing chain, the visual information has supplied the conceptual information, which once again feeds back down to the corresponding lexical items in both languages, and to the cross-modal phonological information for both entries. This dichotomy is testable by virtue of the fact that each account makes different predictions. Should the information need to be fed up to the semantic level before feeding back to activate lexical entries in the non-target language, we should expect subjects to take more time to disambiguate between the target and the competitor relative to unimodal bilinguals, since the competition occurs later in the processing stream. If the second account is true, and the visual information is integrated at the conceptual level, which then feeds back down, we might expect the rate of disambiguation for bimodals and unimodals to be about equal. With the increase of research on bimodal bilingualism in recent years, we are beginning to see comparisons between unimodal bilinguals and their bimodal peers. While there seem to be similarities in function between the two groups, there are also some differences in the mechanisms that underlie their language processing. The research outlined in the previous sections has established certain similarities between the processing patterns of bimodal and unimodal bilinguals, enabling us to consider how phenomena found in studies with unimodal bilinguals might manifest in bimodal bilinguals. In other words, how might the architecture of the language processing system vary? Do bimodal and unimodal bilinguals utilize the same information (e.g., bottom-up segmental information or top-down linguistic context) to the same degree? To begin to answer these questions, it may be useful to consider current models of bilingual language processing in light of bimodal bilingual research. In doing so, we may gain a deeper understanding of how and why bimodal bilinguals process language the way that they do, as well as how bilinguals in general utilize linguistic input.

3. Modeling Language Processing in Bimodal Bilinguals 3.a. Modeling Language Production in Bimodal Bilinguals One of the most immediate challenges of developing a model of language production in bimodal bilinguals is accommodating the ability to produce both languages simultaneously. This phenomenon greatly increases the complexity of the processing system. Consider the

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

48

Anthony Shook and Viorica Marian

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

notion of lexical access. Rather than designing a system in which one target is chosen from many activated lexical items, the Emmorey et al. (2008) results suggest the need for a mechanism that is able to choose separate (but semantically related) lexical items from languages of variant modality for simultaneous production. Furthermore, the notion that the majority of the simultaneously-produced items are semantically related implies that prior to the lexical selection process, a shared conceptual store provides information to the different languages. Also, one needs to consider the way in which syntactic information interacts between the two languages, given how morpho-syntactically meaningful ASL facial expressions occur with English speech (Pyers and Emmorey, 2008) and how bimodal bilinguals’ English constructions sometimes show influence from ASL syntax (Bishop and Hicks, 2005). To our knowledge, only one model of bilingual language processing has been specifically proposed for bimodal bilinguals (see Figure 4). Emmorey et al. (2008) adapted Levelt’s model of Speech Production (1989), and integrated it with a model of speech and gesture production proposed by Kita and Özyrük (2003). In the model proposed by Emmorey et al., the grammatical, phonological and lexical aspects of production of both ASL and English are separate but connected, and are all activated by a Message Generator, which relays conceptual information to the two languages. In other words, propositional or semantic information is sent to both languages simultaneously, and that conceptual unit is determined before lexical selection occurs. Each formulator then encodes the input based on the matrix language (which is the language that submits the syntactic frame). The input is encoded with English grammatical constructs when English is the matrix language and with ASL grammatical constructs when ASL is the matrix language. The result is that code-switched or code-blended productions which are of the non-matrix language are produced in accordance with the matrix language’s grammar.

Figure 4. Model of Code-blend production in bimodal bilinguals as proposed in Emmorey, Borinstein, Thompson, and Gollan (2008). This model integrates Levelt’s (1989) model of speech production with Kita and Özyrük’s (2003) model of co-speech gesture. Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Language Processing in Bimodal Bilinguals

49

Kita and Özyrük’s (2003) addition is the Action Generator, which is a general mechanism used for creating an “action plan.” In this sense, it is not inherently linguistically biased. Rather, it generates movements in both real and imagined space, guided by knowledge of spatial features and information. Though independent of the Message Generator, the two interact; this can result in the gestures generated by the Action Generator being influenced by linguistic information. Casey and Emmorey (2008) suggested that bimodal bilinguals produced more iconic gestures and gestures from a character viewpoint than non-signers. They hypothesized that this was due to the interaction between the Action Generator and Message Generator. Feedback from the Formulator levels of the model (e.g., the ASL Formulator, which includes phonological and lexical information) to the Message Generator likely imbues the latter with information about how to encode the spatial properties of ASL. Furthermore, the authors suggested that this can prime the Action Generator to produce more iconic gestures even when the speaker is simply using English. As a whole, the ASL-English code-blend model proposed by Emmorey et al. (2008) sufficiently explains the results shown in the code-blend literature. It also provides an insight into the timing of bilingual language production systems. While there is some debate about the level at which lexical selection is made in bilingual production (e.g., Finkbeiner, Gollan, and Caramazza, 2006), Emmorey et al.’s (2008) results suggest a late-selection mechanism. The majority of code-blended signs produced by bilingual participants in their study were semantic equivalents to the simultaneously produced English word (approximately 81%). If lexical selection occurred at an earlier stage, say with concept formation, one would expect either no code-blends (rather, code-switches), or more code-blends that produced unrelated, or non time-locked, signs (also see Casey and Emmorey, 2008). However, it is unclear exactly how well this model can generalize to unimodal bilingual data, or conversely, whether that function is necessary. On the one hand, we can simply view the model as specific to bimodal production. In this scenario, we assume variant architectures for unimodal and bimodal bilinguals. This has the negative side-effect of rendering the support of late lexical selection via bimodal data useless when talking about unimodal bilinguals. However, it seems that bilingual production models for unimodal bilinguals are not dissimilar from the basic architecture of the ASL-English code-blend model, and as Emmorey et al. claim, differences seen between the outputs of unimodal versus bimodal productions are not due to systematic or architectural differences, but rather to the biological constraint of not being able to articulate two words at the same time that bimodal bilinguals are not subject to. In light of this notion of biological constraint perhaps masking the architectural or systematic nature of the bilingual language production system, one should examine models created to accommodate unimodal bilingual production patterns, and consider how well they are able to account for bimodal data. There are at least four such models of bilingual production, each of which provides a slightly different explanation for how bilinguals access their lexicons, specifically in regards to how they correctly choose items from the intended target language. One such model was proposed by Costa, Miozzo and Caramazza (1999) and contends that lexical selection during bilingual speech production is language specific. In other words, when choosing a lexical item to produce, the production system has inherent knowledge of both that lexical item’s language category, as well as the intended target language of the speaker, and the lexical selection mechanism will only choose members of the intended language. While this account suggests a language-specific lexical-level selection mechanism,

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

50

Anthony Shook and Viorica Marian

the bimodal data seem to suggest otherwise. Selection likely occurs later in the stream, given that speakers produce simultaneous ASL and English fairly often (35.71% of all utterances in Emmorey et al., 1998, occurred with a code-blended sign). Also, if the lexical selection mechanism were language specific, we would not expect to see code-blends at all. Another model of bilingual language production is Green’s Inhibitory Control (IC) model (1998). Green proposed that rather than having a language specific selection mechanism that activates only one language at a time, conceptual information actually activates all candidates, regardless of language, and the non-target language is then suppressed through some inhibition mechanism (support for this notion often comes from studies showing increased inhibitory control in bilinguals even in non-linguistic tasks, see Bialystok, 1999; and Bialystok, Craik, Klein, and Viswanathan, 2004). In the IC model, since both lexicons are activated, the notion of language non-specific selection suggested by the bimodal data is supported. Furthermore, the IC model posits specific asymmetries in language activation in that a speaker’s L1 is likely to be more strongly activated by the semantic system than a speaker’s L2. Also, since the amount of suppression put forth by the inhibitory mechanism is directly proportional to the amount of activation, L1 words should be more strongly suppressed during L2 use than L2 words during L1 use. Emmorey et al. (2008) found that while producing English, bimodal bilinguals were highly likely to produce concurrent singleword ASL signs. However, when signing ASL, no examples of English single-word intrusions were found. In Emmorey et al.’s subjects, ASL was the L1, and English was the L2. If ASL is more suppressed during production of English, and English less suppressed during ASL production, we would expect to see the opposite result. Specifically, if L1 (ASL) is more strongly suppressed during L2 (English) use, and L2 less strongly suppressed during L1 use, then more English intrusions should be found during ASL productions than ASL intrusions during English production, which was not the case in Emmorey’s results. A third model of lexical selection in bilingual language production comes from La Heij (2005) who suggests that concept selection is more important than lexical selection. In this model, preverbal messages carry information about the speaker’s intended language, as well as information about things like register, and the conceptual notion itself. La Heij’s model presupposes language-specific selection (like Costa et al., 1999), while as previously mentioned, the bimodal data suggest language-non-specific selection. However, La Heij’s model also supposes that the preverbal message at the semantic level is capable of activating related semantic concepts at the lexical level. One possible point of support comes from differences found in the rates of code-blending between Emmorey et al. (2008) and Casey and Emmorey (2008). In the former, bimodal bilingual subjects relayed stories to other bimodal bilingual subjects. In the latter, bilinguals relayed stories to monolinguals, and showed fewer code-blends than in the Emmorey et al. study. According to La Heij, this would be due to the fact that the preverbal message was more strongly balanced towards English in the Casey and Emmorey study, based on the subjects’ knowledge that English was the only shared language. However, this model also seems to predict symmetrical interference across languages. The preverbal message contains the conceptual idea and the intended target-language, which then activates the target language lexical representation, as well as the semantic translationequivalent, to a lesser degree. It’s intuitive to claim, then, the activation of the non-target language translation equivalent should be the same, regardless of which of the two languages is the intended language (at least in equally proficient bilinguals). This prediction should

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Language Processing in Bimodal Bilinguals

51

result in symmetrical interference between languages, which was not seen in the Emmorey et al. study. More recently, Finkbeiner et al. (2006) proposed an account in which lexical selection occurs at the verbal output stage, rather than at the lexical stage, called the response-selection account. According to this account, well-formed phonological information about activated lexical items is held within an output buffer, and from those low-level phonological entries, some selection mechanism chooses the correct target. Within each entry in the output buffer is an item’s phonological and lexical information (e.g., language identity and grammatical class). Intuitively, this is appealing in that it allows for language-non-specific activation, and posits a late-stage selection mechanism that can explain the parallel activation of ASL and English during production as shown in the studies by Emmorey and colleagues (which suggest that lexical items are chosen late during the language production process). The proposal presumes that items in the output buffer are examined serially, beginning with the response that comes first. So, for translation equivalents, semantic priming causes the target and the translation equivalent to be highly active and occur early within the list – the nontarget language item is then quickly rejected based on the target-language mismatch. If one considers only semantic priming, the mechanism by which bimodal bilinguals might code-blend is fairly straightforward. If the two items are both highly active, and the biological “one thing at a time” constraint is not present, both items may be produced. However, the response-output account suggests that the fact that the two items occur in different languages means that the non-target item (in this case, the ASL sign) should be rejected quickly, and not have time to maintain enough activation to reach the production stage. The response-output account does not explain what happens to an item when it is rejected. If activation decays slowly, high levels of activation may still cause the language system to produce a rejected item. If non-target lexical items are actively suppressed, then the account fails to explain the bimodal data. Even if we accept that the selection-mechanism can choose two possible referents for simultaneous production, we still run into the issue of how to deal with non-translation equivalent, non-target language distractors. Empirical findings suggest that these types of distractors cause slower response times in unimodal bilinguals - in other words, they take longer to be rejected. If they take longer to be rejected, then they should continue to gain activation over time, and be more likely to be produced simultaneously in bimodal bilinguals. This does not seem to be the case however, as the vast majority of code-blend productions are translation-equivalents (Emmorey et al. 2008). In support of the serial-nature of the response-output model, there is some recent research to suggest that bimodal bilinguals may actually produce code-blends serially. Emmorey, Petrich and Gollan (2008) performed a picture-naming study where bimodal bilinguals were asked to produce an English word, an ASL sign, or both simultaneously. They found that for bimodal bilinguals, the time required to produce an English word during a code-blend was significantly longer than during the production of an English word or an ASL sign alone. This suggests that during a code-blend, the ASL structure is being constructed first, followed by the English structure. This seems to contrast with Emmorey et al.’s (2008) results that showed a time-lock between production of speech and signs during code-blends. One possible explanation is that the motor planning phase for each language production is different, and the production system is able to coordinate the signal so that they both occur temporally locked. However, if this were the case, we’d expect to see differences in timing for sign-alone and speech-alone productions, which we do not. Another possibility is that the delay of English

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

52

Anthony Shook and Viorica Marian

production during code-blends is simply due to an over-taxed motor coordination system. Further research is required to tease apart these possibilities. While the response-selection proposal seems to cover a sizable portion of empirical incongruities found in other bilingual production models, neither it, nor the other models discussed in this chapter, are currently equipped to explain the bimodal data. Though the majority of bilingual models, in both production and perception, tend to focus on phonological, orthographic and lexical levels (so called lower-order levels), there is also some evidence to indicate syntactic transfer across languages. Research suggests that Spanish-English (Hartsuiker, Pickering, and Veltkamp, 2004) and Dutch-English (Schoonbaert, Pickering, and Hartsuiker, 2007) bilinguals show priming of sentence structures across their two languages, such that the use of a syntactic structure in L1 can prime the use of that same structure in L2. This implies a cross-linguistic, integrated syntactic system where sentence structures that overlap in both languages can utilize lexical items from both languages. Currently, no bilingual models have been developed beyond the level of the lexicon and though many include links to the semantic system, the morphosyntactic system is not usually incorporated in these models. There is also evidence to suggest similar syntactic integration in bimodal bilinguals. Bishop and Hicks (2005) provided samples of written English by children of deaf adults (CODAs) and found that their English constructions showed a good deal of influence from ASL grammatical structure (e.g. missing determiners, dropped subjects, etc.). Emmorey et al. (2008) found similar English constructions in some of their subjects as well. In addition, Pyers and Emmorey (2008) showed that when producing English sentences, bimodal bilinguals tended to produce grammatically relevant ASL facial expressions. These studies may suggest the possibility that the syntactic system is somewhat overlapping even when the modalities of a bilingual’s two languages do not match. Future work will need to determine the nature of syntactic interaction in both unimodal and bimodal bilingual populations, in order to capture a more comprehensive picture of bilingual language processing.

3.b. Modeling Language Perception in Bimodal Bilinguals Arguably the most well developed model of bilingual language processing is the Bilingual Interactive Activation+ model (BIA+, see Figure 5). Initially a bilingual version of the Interactive Activation model (proposed by McClelland and Rumelhart, 1981), the BIA was created to explain how orthographic information is processed in bilinguals (Djikstra and van Heuven, 1998). The model consisted of an orthographic feature level that fed to a letter level. The letter level then provided weighted activation to lexical entries that shared orthographic information with the activated features, according to position (so, graphemes in initial position activated all words with that grapheme in initial position, and inhibited all entries that did not). At the word level, all words, regardless of language, were stored in one lexicon, and had lateral inhibition. This supported the notion of non-selective access, where words in both languages could be equally activated by bottom-up information. The final level consisted of language nodes, which represented each language in the bilingual. These nodes had two functions – first, they categorized words in the single lexicon as belonging to one language or the other. Secondly, summed activation

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Language Processing in Bimodal Bilinguals

53

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

levels from words of one lexicon could activate the language node they corresponded to, which they could further facilitate, and inhibit the words of the other language.

Figure 5. The Bilingual Interactive Activation+ model (BIA+) adapted from Dijkstra and Van Heuven (2002).

One major issue with the BIA, in its initial implementation, was its focus on purely orthographic stimuli. Dijkstra and Van Heuven (2002) proposed an extension to the BIA, labeled the BIA+, which included extensions for phonological and semantic representations via a separate model called SOPHIA (the Semantic, Orthographic, PHonological Interactive Activation model; see Figure 6). Language nodes are still present but no longer inhibit the non-target language - They are simply used to supply lexical items with a category. However, though SOPHIA adds the ability for BIA+ to process phonological information, the model still uses orthographic information as input. While it is capable of modeling the interactions between orthographic and phonological information, it is not very good at making predictions regarding solely phonological input. Due to this reliance on written language, BIA+ is limited in its ability to explain results found in bimodal bilingual studies using hearing signers. At first glance, the BIA+ may be well equipped to explain the results found within the sign-text bilingual subgroup of bimodal bilinguals (see Hanson, 1982). Since one of a sign-

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

54

Anthony Shook and Viorica Marian

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

text bilingual’s languages relies entirely on orthographic information, the BIA+ model may be able to capture the interactions between the written language and the signed language.

Figure 6. The Semantic, Orthographic, Phonological Interactive Activation model (SOPHIA) adapted from Dijkstra and Van Heuven (2002). Connections between orthographic and phonological pathways allow for interaction between the two domains.

However, further examination of results from sign-text bilingual studies suggests a problem for BIA+, as the data imply that the two lexicons of English written form versus ASL are separate, but capable of being encoded simultaneously when accessed via L1. The BIA+/SOPHIA combination actually posits a temporal delay between the processing of orthographic information and phonological information within L2 items relative to L1 items. In certain tasks, where understanding is guided by L1 orthographic codes, the L2 semantic or phonological information may not have time to influence the perception of L1 – in other words, we should see no L1-L2 interference when the task demands deem orthographic L1 input as most significant. Furthermore, the data from sign-text studies seem to advocate the notion of separate lexicons, which is in direct contrast to the integrated lexicon of the BIA+.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Language Processing in Bimodal Bilinguals

55

Yet, we must consider whether to view sign-text bilingualism in the same category as unimodal (speech-speech) bilinguals, or bimodal (sign-speech) bilinguals. The two languages in a sign-text bilingual are uniquely separated, in that one is housed entirely within the phonological (as it applies to the phonological properties of sign languages) realm while the other is exclusively orthographic. They also make use of entirely different grammars. This could result in weakened links between the orthographic and phonological pathways, affecting their interaction. In unimodal bilinguals, the systems and mechanisms utilized by the two languages are shared. In bimodal sign-speech bilinguals (e.g., ASL-English), lexical items in English could map to translation-equivalent lexical items in ASL via associative learning. This mapping could be generated initially by semantic links, but could also result in the development of lateral links between the two lexical systems (or modalities within the same lexical system). Since the orthography of language in the SOPHIA model is connected laterally to the phonology of the same language at every level of structure, one could then design a system where English orthography is able to connect to ASL phonology through an exclusively lateral chain. This suggests a greater similarity between unimodal and bimodal sign-speech bilingual processing systems than either unimodal or bimodal sign-speech systems with sign-text. Perhaps the simplest way to circumvent the issues of the BIA+ is to change the input structure. Most models of language processing are based on single-modality input – often this is done to simplify the enormous task of modeling the entire language system. In order to accurately capture the pattern of bimodal bilingual processing, however, models of bilingual language processing require multiple input structures and need to re-envision the connections between them. Also, the input structures themselves will have to be altered. The BIA+ utilizes a positional system –orthographic input at the letter level activates lexical items whose orthography matches not only in character, but also in position. It isn’t enough for two lexical items to share a “d” in their orthography, but that “d” has to occur in the same position in order for two words to coactivate. For sign, it is much more difficult to recognize a serialized structure, and so the sign input structure must likely do away with the positional form. Still, the bimodal sign-speech bilingual system consists of distinctive language pathways. It would be difficult to imagine, due to the cross-modal nature of the bimodal bilingual, that the phonologic, lexical and feature levels represented in the BIA+ could be appropriately shared in a bimodal bilingual. The total lack of overlap at any level seems to preclude this notion. Instead, were we to attempt to adapt the BIA+ to bimodal results, one way to design it could be the inclusion of a third pathway that includes lateral links – phonology of L1, phonology of L2 and orthography of L2. The lateral links, as previously mentioned, could be borne from associative links via a semantic pathway (like the Hebbian notion of cells that fire together, wire together; Hebb, 1949). In fact, Emmorey et al. (2008) showed that bimodal bilinguals who produced code-blends (simultaneous production of sign and speech) did not produce signs that were propositionally different from the speech output – the signs were often closely related semantically, suggesting that the same conceptual notion from a shared semantic level activates items in both lexicons. In order to better account for results from studies of sign-speech bimodal bilinguals, it may be more appropriate to look at models that consider auditory information as the primary source of input. One such model is the Bilingual Model of Lexical Access (BIMOLA, Grosjean, 2008; see Figure 7), a model of bilingual speech perception based on the TRACE model of speech perception (McClelland and Elman, 1986). The BIMOLA posits separate

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

56

Anthony Shook and Viorica Marian

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

clusters of phonemes and words for each language, though they are housed within a single set at each level. This means that L1 words do not compete with L2 words during auditory recognition at the lexical level. This does not mean that the two languages cannot be activated in parallel; rather, the separation of language sets acts as a categorization tool like the language nodes found in the BIA+. The BIMOLA also has a “global language information” node that informs the system of contextual information.

Figure 7. The Bilingual Model of Lexical Access (BIMOLA) as proposed in Thomas and Van Heuven (2005).

The most salient issue with the BIMOLA regarding its ability to cope with bimodal bilingual data is that its input is restricted to the auditory modality. However, it should be possible to change the modality of input features. Let us suppose then, that the BIMOLA were equipped with featural information for signed languages as well as spoken languages. Since the features of signed phonology and spoken phonology do not overlap, then we must immediately change the overlapping feature level to two separate feature levels that feed independently upward to the phonological level. The fact that BIMOLA already posits separate lexicons and phonological levels is congruent with bimodal processing; however, there may be an issue in that the languagespecific lexical and phonological sets are housed within the same larger set. BIMOLA does

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Language Processing in Bimodal Bilinguals

57

not clearly define how the set is categorized. If the set is based on shared featural information that guides the larger set of phonological entities (which would allow for shared phonemes across languages) then separate language clusters aren’t enough – the model would require separate sets at each level. This restructuring of the model has important implications for bimodal bilingual processing patterns during language comprehension. Since previous studies show parallel activation in unimodal bilinguals (Blumenfeld and Marian, 2007; Canseco-Gonzalez, et al., 2005; Marian and Spivey, 2003; Weber and Cutler, 2004;), it is possible that bimodal bilinguals also activate their two languages in parallel during language comprehension (e.g., Shook and Marian, in progress). If this is the case, then the featural level, which is the presumed primary locus of parallel activation in unimodal bilinguals, cannot be the cause of parallel activation in bimodal bilinguals. Instead, it must be due to feedback from the semantic system, or connections between languages at a lexical level (see Figure 3). BIMOLA does not include a semantic level in its architecture, nor does it allow for lateral activation or inhibition. Notably, the BIA+ also does not specifically include the semantic system in its architecture either, but both models posit a shared conceptual system (for support, see Finkbeiner, Nicol, Nakamura and Greth, 2002 and Li and Gleitman, 2002). This is, however, a fairly straightforward fix. If we assume a shared conceptual store, then it is simply another level above the lexicon which, importantly, must include both feed-forward and feed-back connections in order to explain parallel processing in bimodal bilinguals. Note that one model that was not included in the present discussion is the SelfOrganizing Model of Bilingual Processing (SOMBIP, Li and Farkas, 2002). While the SOMBIP is uniquely qualified to look at the influence of developmental patterns on the structure of the bilingual lexicon, it currently makes no predictions regarding how a bilingual’s two languages might interact (regardless of modality) or concerning the mechanisms that underlie language processing in a fully developed user. In summary, no perceptual model of bilingualism currently explains bimodal bilingual data. In order to account for these data, the models must be altered to include visual-linguistic input structures beyond orthography, and they require feedback systems to allow for top-down influence on the activation of lexical items, primarily for the unused language. Further research needs to be done, however, before we understand more fully the level of interaction between the two languages in a bimodal bilingual.

Conclusions This chapter summarizes existing knowledge about language processing in bimodal bilinguals and compares phenomena found in bimodal research to those found in unimodal research in order to create a more complete model of bilingual language processing. The question of whether languages that vary in modality, such as ASL and English, are represented the same way in the brain is not a trivial one. It requires that the two languages be placed on an even playing field. In other words, one could argue that similar representation requires similar function. While on the surface, ASL seems vastly different from English, there are structural similarities. For example, the phonological systems in both languages are

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

58

Anthony Shook and Viorica Marian

based on the combination of finite, meaningful parts. Perhaps more importantly, sign languages display many of the same features of spoken languages, such as phonological priming, an articulatory loop in working memory, and categorical perception. It is intuitive to believe that the functional similarities between the two languages may lead to representational similarities as well, where signed and spoken languages exist in the same language processing system. Furthermore, this overlap of two languages in one system should look much like that found in a unimodal system. However, the processing patterns of bimodal and unimodal bilingual groups are not always congruous. For example, bimodal bilinguals code-blend (produce a sign and speech simultaneously) while unimodal bilinguals code-switch (produce lexical items from both languages at different times in the same sentence). It is possible that this difference between groups is due solely to the unimodal bilinguals’ biological constraint of simply being unable to produce two spoken languages at once. If so, it is unclear whether the relationship between two languages in unimodal and bimodal bilinguals is similar. Results by Emmorey et al. (2008) suggest a semantic overlap between lexical items in code-blends, which implies that a single concept from the semantic system activates items from two lexicons. Bimodal bilinguals also seem to show increased occurrence of interfering morphosyntactic markers across languages (Pyers and Emmorey, 2008). These studies indicate that languages of different modalities seem to strongly interact in bimodal bilinguals. The question of how exactly two languages of different modalities interact remains. One possibility is that language, regardless of modality, is represented similarly in the brain. Work by Emmorey and colleagues seems to suggest that although the modalities aren’t shared, there is overlap between languages in bimodal bilinguals, much like we might expect to see in unimodal bilinguals. While spoken and signed languages vary in surface and structural aspects, research suggests that they are processed very similarly. In light of this, one could argue that the distinction between spoken and signed languages on a representational level is seamless – they utilize the same processes within the same architecture. In this account, the differences seen between the two groups may be based on two things – non-linguistic differences or constraints, and degree of processing. The first refers to surface level differences based on the nature of the language itself. For example, as we have discussed, bimodal bilinguals’ code-blending is not necessarily based on processing differences, but the lack of a biological constraint found in unimodal bilinguals. The second refers to the degree to which certain mechanisms are utilized. As previously discussed, users of signed languages have been shown to have greater visual field perception (Bavelier et al. 2000), which could result in greater attention to visual detail during language processing. This suggests the possibility that visual and linguistic information are more strongly linked in users of signed languages, and subsequently, bimodal bilinguals. In this scenario, the structure of the processing system is not inherently different between unimodal and bimodal groups, but the degree to which certain mechanisms influence speech is distinct. However, we must still entertain two different possibilities for modeling the bilingual language system. The first is the notion that perhaps bimodal bilinguals utilize an entirely different system than unimodal bilinguals. The differences found between bimodal bilinguals and unimodal bilinguals can suggest that the reason current models of bilingual language processing are unable to accommodate bimodal data is due to differences in the two systems. Indeed, arguments that explain differences between groups by positing separate processing patterns or architectures are used to compare modality differences within unimodal bilingual

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Language Processing in Bimodal Bilinguals

59

models as well. Thomas and Van Heuven (2005) explain the differences between the BIA+ and the BIMOLA by saying “the modelers implicitly assume that the different demands of recognition in each modality have led to different functional architectures.” This is an immediately appealing argument, in that it allows researchers to parse up the different domains of language processing. Note that Thomas and Van Heuven are simply talking about the functional differences between the visual and auditory domain in perception. This says nothing about production, and the way in which the production system in bilinguals (as well as monolinguals) maps onto the receptive system. Note that while this path of separate domain/separate task modeling is tempting, there are issues with implementation. It would be difficult to argue that separate domains within language processing do not interact in some way, whether it be through shared semantic representations or influence at lower levels of processing. Consider the process of audiovisual integration – even in monolinguals, there is an obvious combination of the auditory and visual modalities that influences speech perception. What this means is that, even if we model each domain separately, the different domains still need to be integrated in some way. For the sake of parsimony, it seems simpler to assume shared architecture from the first stages of development, and build upward. While creating a fully implemented model is an enormous task, it seems like the most cost-effective method of developing models that can handle data from bimodal bilingual studies, considering that these require a system that can integrate information from different domains or modalities. This idea leads directly to the second possibility, that unimodal and bimodal bilinguals utilize the same architecture to process language. Indeed, studies of sign language suggest similar functional capacity between signed languages and spoken languages at a cognitive level. Furthermore, the bimodal studies suggest that bimodal bilinguals show similar language function as unimodal bilinguals, such as parallel processing of their two languages. While Emmorey et al. (2008) point out differences in the production patterns of bimodal bilinguals compared to unimodal bilinguals, the authors also provide evidence to suggest the differences are due to output ability (bimodal bilinguals can produce two lexical items simultaneously) rather than the structure of the system itself (since code-blends tended to be semantically linked). With the similarities in processing between unimodal and bimodal groups in mind, rather than creating separate models for unimodal and bimodal bilingual groups, it may be most useful to develop models that are capable of adjudicating unimodal and bimodal bilingual data without forcing variant architectures. Perhaps the best method would be to consider localist models in which weights between nodes can be readily shifted not just between phonology and orthography, but also between separate phonological pathways for languages of different modalities. Furthermore, bilinguals show an influence of audio-visual integration, and models of language processing should be capable of implementing that integration. With language processing architecture that includes connections between modalities, both in terms of input (visual or auditory) and language (L1 or L2), one may be able to make systematic predictions about the way bilinguals process their two languages, regardless of modality, simply by varying the interactions between domains instead of the structure of the system itself. By looking at bimodal bilinguals, it becomes clear that computational models require the ability to accommodate and integrate information from a variety of modalities and domains. This necessitates the development of more harmonious models which take into account both

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

60

Anthony Shook and Viorica Marian

the functional convergence and structural divergence found in bimodal bilingual processing patterns, and consider language processing not to be domain-specific, but experience-based, in the sense that the underlying mechanisms that govern language are the same, but how they are used varies dependent on the type of input the system receives. Furthermore, by studying bimodal bilinguals, we may begin to gain a clearer picture of how language processing is affected by cross-modal input. If we are to expand our understanding of bilingual language processing, we must take into account those groups (like bimodal bilinguals) who are on the periphery of current research, in order to widen the boundaries of our knowledge about what the bilingual language system is capable of. In doing so, we may more fully understand bilingual language processing, and be better equipped to describe the mechanisms that govern language.

Author Note The authors would like to thank James Bartolotti, Caroline Engstler, Jenna Luque and Scott Schroeder for comments and feedback on earlier drafts of this manuscript. Correspondence should be directed to Anthony Shook at [email protected] or at the Department of Communication Sciences and Disorders, Northwestern University, 2240 Campus Drive, Evanston IL 60208.

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

References Baker, S.A., Idsardi, W.J., Golinkoff, R.M., and Pettito, L. (2005). The perception of handshapes in American Sign Language. Memory and Cognition, 33(5), 887-904. Baker, S.A., Golinkoff, R.M., and Pettito, L. (2006) New insights into old puzzles from infants’ categorical discrimination of soundless phonetic units. Language Learning and Development, 2(3), 147-162. Bavelier, D., Tomann, A., Hutton, C., Mitchell, T., Corina, D., Liu, G., and Neville, H. (2000). Visual attention to the periphery is enhanced in congenitally deaf individuals. Journal of Neuroscience, 20, 1-6. Bavelier, D., Dye, M.W.G., and Hauser, P.C. (2006). Do deaf individuals see better? Trends in Cognitive Sciences, 10(11), 512-518. Bialystok, E. (1999). Cognitive complexity and attentional control in the bilingual mind. Child Development, 70, 636-644. Bialystok, E., Craik, F.I.M., Klein, R., and Viswanathan, M. (2004). Bilingualism, aging, and cognitive control: Evidence from the Simon task. Psychology and Aging, 19, 290-303. Bishop, M., and Hicks, S. (2005). Orange eyes: Bimodal bilingualism in hearing adults from deaf families. Sign Language Studies, 5(2), 188-230. Blumenfeld, H., and Marian, V. (2007). Constraints on parallel activation in bilingual spoken language processing: Examining proficiency and lexical status using eye-tracking. Language and Cognitive Processes, 22(5), 633-660. Bonvillian, J.D., and Folven, R.J. (1993). Sign Language Acquisition: Developmental Aspects. In M. Marschark and M.D. Clark. (Eds.), Psychological Perspectives on Deafness, pp. 229-265. Hillsdale, NJ: Lawrence Erlbaum Associates.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Language Processing in Bimodal Bilinguals

61

Brentari, D. (1998). A Prosodic Model of Sign Language Phonology. MIT Press. Canseco-Gonzalez, E. (2005). “Carpet or Carcel” effects of speaker type, fluency, and language mode on bilingual lexical access. Abstracts of the 5th International Symposium on Bilingualism, pp. 156-157. Casey, S., and Emmorey, K. (2008). Co-speech gesture in bimodal bilinguals. Language and Cognitive Processes, 1-23. Chomsky, N. (2000). New Horizons in the Study of Language and Mind. Cambridge University Press. Colomé, A. (2001). Lexical activation in bilinguals’ speech production: Language-specific or language-independent? Journal of Memory and Language, 45(4), 721-736. Costa, A. and Caramazza, A. (1999). Is lexical selection in bilingual speech production language-specific? Further evidence from Spanish-English and English-Spanish bilinguals. Bilingualism: Language and Cognition, 2, 231-244. Costa, A., Miozza, M. and Caramazza, A. (1999). Lexical selection in bilinguals: Do words in the bilingual’s two lexicons compete for selection? Journal of Memory and Language, 41, 365-397. Dijkstra, T. and van Heuven, W. (1998). The BIA model and bilingual word recognition. In J. Grainger and A.M. Jacobs (Eds.), Localist Connectionist Approaches to Human Cognition (pp. 189-225). Mahwah, NJ: Erlbaum. Dijkstra, A., and Van Heuven, W. J. B. (2002). The architecture of the bilingual word recognition system: From identification to decision. Bilingualism: Language and Cognition, 5,175-197. Dufour, R. (1997). Sign language and bilingualism: Modality implications for bilingual language representation. In J.F. Kroll and A. M. B de Groot (eds.), Tutorials in Bilingualism: Psycholinguistic Perspectives. pp. 301-330. Mahwah, NJ: Lawrence Erlbaum Associates. Dye, M. W. G., and Shih, Shui-I. (2006). Phonological priming in British Sign Language. In L. M. Goldstein, D. H. Whalen and C. T. Best (Eds.), Laboratory Phonology 8: Varieties of Phonological Competence. Berlin: Mouton de Gruyter. Emmorey, K., McCullough, S., and Brentari, D. (2003). Categorical Perception in American Sign Language. Language and Cognitive Processes, 18(1), 21-45. Emmorey, K., Borinstein, H.B., Thompson, R., and Gollan, T.H. (2008). Bimodal bilingualism. Bilingualism: Language and Cognition, 11(1), 43-61. Emmorey, K., Petrich, J., and Gollan, T.H. (2008). Simultaneous production of ASL and English implies serial access in production. Presented at the 49th Annual Meeting of the Psychonomic Society, November 16th, 2008, Chicago IL. Finkbeiner, M., Nicol, J., Nakamura, K., and Greth, D. (2002). The role of language in memory for actions. Journal of Psycholinguistic Research, 31(5), 447-457. Finkbeiner, M., Gollan, T.H., and Caramazza, A. (2006). Lexical access in bilingual speakers: What’s the (hard) problem? Bilingualism: Language and Cognition, 9(2), 153-166. Gathercole, S. E., and Baddeley, A. D. (1993). Working Memory and Language . Hillsdale, NJ: Erlbaum. Goldinger, S.D., Luce, P.A., Pisoni, D.B., and Marcario, J.K. (1992). Form-based priming in spoken word recognition: The roles of competition and bias. Journal of Experimental Psychology: Learning, Memory and Cognition, 18, 1210–1238.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

62

Anthony Shook and Viorica Marian

Grainger, J. (1993). Visual word recognition in bilinguals. In R. Schreuder and B. Weltens (Eds.), The Bilingual Lexicon (pp. 11–26). Amsterdam: John Benjamins. Green, D. W. (1998). Mental control of the bilingual lexico-semantic system. Bilingualism: Language and Cognition, 1, 67-82. Grosjean, F. (1997). Processing mixed languages: Issues, findings and models. In A.M.B. de Groot and J.F. Kroll (Eds.), Tutorials in Bilingualism: Psycholinguistic Perspectives (pp. 225-254). Mahwah, NJ: Erlbaum. Grosjean, F. (2008). Studying bilinguals. Oxford: Oxford University Press. Hartsuiker, R.J., Pickering, M.J., and Veltkamp, E. (2004). Is syntax separate or shared between languages? Cross-linguistic syntactic priming in Spanish-English bilinguals. Psychological Science, 15(6), 409-414. Hanson, V. (1982). Short-term recall by deaf signers of American Sign Language: Implications of encoding strategy for order recall. Journal of Experimental Psychology: Learning, Memory, and Cognition, 8, 572-583. Hebb, D.O. (1949) The Organization of Behavior. New York: Wiley. Hildebrandt, U., and Corina, D. (2002). Phonological similarity in American Sign Language. Language and Cognitive Processes, 17(6), 593-612. Kita, S. and Özyrük, A. (2003). What does cross-linguistic variation in semantic coordination of speech and gesture reveal?: Evidence for an interface representation of spatial thinking and speaking. Journal of Memory and Language, 48, 16-32. Krakow, R., and Hanson, V. (1985). Deaf signers and serial recall in the visual modality: Memory for signs, fingerspelling and print. Memory and Cognition, 13, 265-272. La Heij, W. (2005). Selection processes in monolingual and bilingual lexical access. In J.F. Kroll and A.M.B de Groot (eds.), Handbook of Bilingualism: Psycholinguistics Approaches, pp. 289-307. New York: Oxford University Press. Levelt, W.J.M. (1989). Speaking: From Intention to Articulation. Cambridge, MA: MIT Press. Leybaert, J. (2000). Phonology acquired through the eyes and spelling of deaf children. Journal of Experimental Child Psychology, 75(4), 291-318. Li, P. and Farkas, I. (2002). A self-organizing connectionist model of bilingual processing. In R. Heredia and J. Altarriba (Eds.), Bilingual Sentence Processing, pp. 59-85. North Holland: Elsevier. Li, P., and Gleitman, L. (2002). Turning the tables: Language and spatial reasoning. Cognition, 83, 265-294. Liddell, S. (1980). American Sign Language Syntax. The Hague: Mouton. Marian, V. (2009). Audio-visual integration during bilingual language processing. In A. Pavlenko (Ed.), The Bilingual Mental Lexicon. Clevedon, UK: Multilingual Matters. Marian, V., and Spivey, M. (2003). Bilingual and monolingual processing of competing lexical items. Applied Psycholinguistics, 24, 173-193. Marslen-Wilson, W.D. (1990). Activation, competition, and frequency in lexical access. In G. T. M. Altmann (Ed.), Cognitive Models of Speech Processing: Psycholinguistic and Computational Perspectives. Cambridge, MA: MIT Press. Marslen-Wilson, W. D., and Welsh, A. (1978). Processing interactions and lexical access during word recognition in continuous speech. Cognitive Psychology, 10, 29-63.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Language Processing in Bimodal Bilinguals

63

McClelland, J. L., and Rumelhart, D. E. (1981). An interactive activation model of context effects in letter perception: Part 1. An account of Basic Findings. Psychological Review, 88, 375-407. McClelland, J. L., and Elman, J. L. (1986). The TRACE Model of Speech Perception. Cognitive Psychology, 18, 1-86. McGurk, H. and MacDonald, J. (1976). Hearing lips and seeing voices: A new illusion. Nature, 264, 746-748. McMurray, B., Tanenhaus, M.K., and Aslin, R.N. (2002). Gradient effects of within-category phonetic variation on lexical access. Cognition, 86(2), 33-42. McQuarrie, L., and Parrila, R. K. (2008). Deaf children’s awareness of phonological structure: Rethinking the ‘functional-equivalence’ hypothesis. Journal of Deaf Studies and Deaf Education, doi:10.1093/deafed/enn025. Meuter, R.F.I, and Allport, A. (1999). Bilingual language switching in naming: Asymmetrical costs in language selection. Journal of Memory and Language, 40, 25-40. Miller, P. (2004). Processing of written words by individuals with prelingual deafness. Journal of Speech, Language, and Hearing Research, 47, 979-989. Petitto, L.A., and Marentette, P.F. (1991). Babbling in the manuel mode: Evidence for the ontogeny of language. Science, 251, 1493-1496. Praamsta, P., Meyer, A.S., and Levelt, W.J.M. (1994). Neuropsychological manifestations of phonological processing: Latency variation of a negative ERP component timelocked to phonological mismatch. Journal of Cognitive Neuroscience, 6(3), 204-219. Pyers, J.E., and Emmorey, K. (2008). The face of bimodal bilingualism: Grammatical markers in American Sign Language are produced when bilinguals speak English to monolinguals. Psychological Science, 19(6), 531-536. Schoonbaert, S., Hartsuiker, R.J., and Pickering, M.J. (2007). The representation of lexical and syntactic information in bilinguals: Evidence from syntactic priming. Journal of Memory and Language, 56(2), 153-171. Shook, A., and Marian, V. (in progress). Parallel processing in bimodal bilinguals. Manuscript in Preparation. Slowiaczek, L.M., and Pisoni, D.B. (1986). Effects of phonological similarity on priming in auditory lexical decision. Memory and Cognition, 14(3), 230-237. Spivey, M., and Marian, V. (1999). Crosstalk between native and second languages: Partial activation of an irrelevant lexicon. Psychological Science, 10, 281–284. Stokoe, W. (1960). Sign language structure: An outline of the visual communication systems of the American Deaf, Studies in Linguistics, Occasional Papers 8. Available from Silver Spring, MD: Linstok Press. Supalla, T., and Webb, R. (1995). The grammar of International sign: A new look at pidgin languages. In K. Emmorey and J. Reilly (eds.), Sign, Language and Space, pp. 333-352. Hillsdale, NJ: Lawrence Erlbaum Associates. Sutton-Spence, R. and Woll, B. (1999). The Linguistics of British Sign Language. Cambridge, U.K.: Cambridge University Press. Thomas, M.S.C., and van Heuven, W.J.B. (2005). Computational models of bilingual comprehension. In J. F. Kroll and A. M. B. de Groot (Eds.), Handbook of Bilingualism: Psycholinguistic Approaches. Oxford: Oxford University Press. Weber, A., and Cutler, A. (2004). Lexical competition in non-native spoken- word recognition. Journal of Memory and Language, 50, 1-25.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

64

Anthony Shook and Viorica Marian

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Werker, J.F., and Tees, R.C. (1984). Phonemic and phonetic factors in adult cross-language speech perception. Journal of the Acoustical Society of America, 75(6), 1866-1877. Wilson, M., and Emmorey, K. (1997). A visuospatial “phonological loop” in working memory: evidence from American Sign Language. Memory and Cognition, 25, 313-320. Wilson, M., and Emmorey, K. (1998). A “word length effect” for sign language: Further evidence on the role of language in structuring working memory. Memory and Cognition, 26, 584-590.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

In: Bilinguals: Cognition, Education and Language Processing ISBN: 978-1-60741-710-1 Editor: Earl F. Caldwell, pp. 65-80 © 2010 Nova Science Publishers, Inc.

Chapter 3

PSYCHOLINGUISTIC ABILITIES AND PHONOLOGICAL WORKING MEMORY IN BILINGUAL CHILDREN WITH SPECIFIC LANGUAGE IMPAIRMENT: A CROSS-CULTURAL STUDY Dolors Girbau* University Jaume I, Castelló, Spain

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Abstract A variety of concepts/types of bilingualism and bilingual programs in U.S.A./Europe are presented. The cognitive benefits of bilingual education across several languages are reviewed. Diagnosis and intervention issues in bilingual children with Specific Language Impairment (SLI) / Typical Language Development (TLD) are discussed, including some behavioral and neurophysiology findings concerning language processes. A cross-cultural study was done by comparing children from U.S.A. (with SLI/TLD) and children from Spain (with SLI/TLD), who were involved in a larger project (Girbau & Schwartz, 2007, 2008). Forty-four sequential bilingual children (7;6-10;11 years old), with L1 = Spanish and L2 = English/Catalan, participated. The psycholinguistic abilities in any bilingual group with TLD were significantly higher than in any bilingual group with SLI (Spanish-English/Spanish-Catalan). The similarities of the cross-cultural profiles are discussed. Only children with TLD from Spain produced significantly more correct non-words (in the Spanish Non-word Repetition Task) than children with TLD from U.S.A. (who were exposed to English phonetics). This cross-cultural difference was not found for children with SLI; they all performed poorly in U.S.A. and Spain. The Spanish task was a good marker for SLI in both countries. Our results support the phonological working memory deficit associated with SLI, which appears to be independent of the particular bilingual background. The English Non-word Repetition Task was not sensitive in identifying SLI in these Hispanic unbalanced bilinguals, since English was their L2; their phonotactic representations in L1 seem to determine their performance on the task.

* Correspondence: Dolors Girbau, Department of Basic, Clinical & Biological Psychology, University Jaume I, Campus Riu Sec, 12071 Castelló, Spain. E-mail address: [email protected]. Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

66

Dolors Girbau The Spanish non-word repetition accuracy correlated significantly with the Auditory Association subtest from the Spanish ITPA, in children with SLI/TLD and in Spain/U.S.A. The task also correlated significantly with the Grammatical Integration subtest for children with SLI and in Spain/U.S.A. Both subtests involve auditory working memory, but the second one has also some visual support through pictures. Implications of the results for the crosscultural identification of SLI in bilinguals are discussed.

1. Introduction Measuring the individual proficiency in two particular languages is a challenge for professionals of bilingualism, especially when the features of these languages are very different or the availability of tests is limited. It becomes more complex when we need to identify children with communication disorders, particularly Specific Language Impairment (SLI). The present book chapter is part of a larger cross-cultural research in sequential bilingual (native Spanish-speaking) children with either SLI or Typical Language Development (TLD), from U.S.A. and Spain.

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

1.1. Basic Concepts and Types of Bilingualism There are many approaches to the definition of bilingualism, which are usually linked to some particular types of bilingualism (see Girbau, 2002, for a review). A bilingual person is considered as an individual who is usually high proficient in two languages and is able to use any of them with similar ability in any communicative context. This definition corresponds to a type of balanced and active bilingualism, (opposite to unbalanced/passive bilingualism). Usually they are early bilinguals (rarely late bilinguals), who learn both languages during their childhood. The most representative situation of this early exposure to a bilingual context is the so-called family bilingual, whose parents speak respectively two different languages to their child since he/she is born. Thus, family bilinguals acquire both languages simultaneously (i.e., they are simultaneous bilinguals). However, some individuals are significantly more proficient in one of the two languages to which they are exposed; in this case they use to be more proficient in their native language that is spoken at home. This unbalanced (or incipient) bilingualism is usually related with sequential bilingualism (i.e., their second language is acquired after their first one). Sometimes this type of bilingual may become passive bilingual, a person who is native speaker in one language and is capable of understanding another one despite not speaking it for several possible circumstances (not having enough proficiency, considering it unnecessary, etc.). This is more frequent in communities with dyglossic bilingualism, in which speakers use one language for formal contexts and another language for informal settings. Finally, each language includes many linguistic variants (phonological, semantic, etc.) that are found not only across the different countries/regions in which it is spoken, but also within smaller geographic areas.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Psycholinguistic Abilities and Phonological Working Memory…

67

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

1.2. Cognitive Benefits of Bilingual Education Bilingual education is developed throughout multiple contexts across life-span, but home and school are settings of particular relevance for language development. Bilingual programs at schools in the United States are somewhat different from those found across several European countries (including Spain). In U.S.A., there are several approaches to bilingual instruction, which are usually transitory (i.e., applied only to a limited number of grades), as opposed to the continuous bilingual education in several European bilingual regions (aiming that children will become proficient bilinguals). For example, the dual language program in U.S.A. (also known as "two-way bilingual program"), groups native speakers of English with native speakers of another language (e.g., Spanish). Instruction is provided in both languages usually on alternate days, or sometimes according to academic subjects, etc. The main goal is often that students will be proficient in English (L2) at the end of primary school, but not necessarily proficient in their native language (L1). Thus, if at some point parents also begin to speak English with their children at home (instead of their L1), the attrition of L1 is speeded up in favor of English so that they will probably become monolinguals of English soon. However, when their native language is the only one that is spoken at home, children can more easily become balanced bilinguals (especially if the quality of language input is good). Finally, some children can finish the school with a very poor proficiency in any language, as we will discuss in this book chapter. Obviously, the age of beginning bilingual education (and the quality/amount of language input) will be crucial to develop their bilingual skills; we all know the sponge-like learning capacity of young typically developing children. According to the U.S. Census Bureau (2008), in 2007 more than 45 million people nationwide were estimated to be of Hispanic heritage (with a projection of 75 million by 2025). How many of them will develop a balanced bilingualism is an open question. Spanish was the language spoken at home by more than 34 million people in 2006. However, this Hispanic population come mostly from low Socioeconomic Status (SES) which can affect negatively language acquisition (Schuele, 2001; Suro, 2005), and also increase the deficits associated with language impairment. A low SES is usually associated to poor language input at home. In Spain (a multilingual European country with more than 46 million inhabitants), the linguistic immersion program in Catalan is the most popular one in the region of Catalonia. Catalan is a romance minority language that is spoken by around 11 million people (in Catalonia, Valencia, Balearic Islands, etc.). It has more lexical, phonological and morphosyntactic similarities with Spanish than English. That program is applied across all grades by using their own minority language (Catalan) for the instruction of all subject matter; Spanish is often only taught as a separate subject. Its goal is that students will become balanced bilinguals, since many of them are more exposed to Spanish (a majority co-official language) in their daily life, with an approximate estimation of half of them having Catalan (the other half having Spanish) as home-speaking language (Generalitat de Catalunya, 2008). In other areas (e.g., Valencian region), it is more popular having Spanish as the main teaching language and Catalan only taught as a separate subject. This type of program is usually more associated to passive bilinguals (except for those who speak the minority language at home). More cross-cultural research is needed to find out which is the best instruction approach for each particular individual to become balanced bilingual. Probably the election of the best

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

68

Dolors Girbau

bilingual program for a particular child will depend mostly on their bilingual baseline and context (i.e., degree of proficiency/exposure to the two languages). The cognitive advantage of bilingualism has been demonstrated across many studies including one of the first works from Montréal (Peal & Lambert, 1962). The authors already concluded about the bilingual superiority in cognitive flexibility, divergent thought, metalinguistic skills, general intelligence development, etc. More recently, the benefits of bilingualism have been found across several language processes. For example, early biliterate bilinguals (L1 = Russian / L2 = Hebrew) showed superior word/pseudoword accuracy and phonological awareness in Hebrew than earlyliterate monolinguals of Hebrew (Schwartz, Share, Leikin, & Kozminsky, 2008). This bilingual advantage was interpreted as coming from a transfer of their Russian skills, a language with a complex syllabic structure and a fully fledged alphabet. Other studies with more than 100 bilinguals/multilinguals and 100 monolinguals concluded that bilinguals (and even more multilinguals) are more able to suppress irrelevant information in dual-task language conditions, involving long-term working memory that benefits from this bilingual education, (cf. Ransdell, Arecco, & Levy, 2001). In a study on a lexical retrieval task, bilinguals with matched vocabulary scores were better in a letter fluency test than monolinguals (whose performance was similar to bilinguals with lower vocabulary), (Bialystok, Craik, & Luk, 2008). However, the age at which bilingual education begins seems to modulate this bilingual advantage (e.g., in certain contexts requiring high-level processing). In a research with Spanish/English listeners, the maximum level of noise at which English speech was intelligible was significantly lower for late bilinguals (who learned it after age 14) than for early bilinguals (learning it before age 6) and English monolinguals; all bilinguals were native Mexican-Spanish speakers, (Mayo, Florentine, & Buus, 1997). Another study found similar English academic achievements (including reading and word knowledge in grades 6, 8 and 12) for Mexican-Americans who were either in a monolingual English or maintenance bilingual program at 1st-5th grade (Medina, Saldate, Mishra, 1985). The cognitive advantage of bilingual education has also been studied in relation to certain types of pathology with some interesting findings. Particularly, research supports that bilinguals show an advantage in cognitive control and executive processes, which are boosted by their experience in controlling attention to two languages without interferences between them since their childhood (Bialystok, 2007). All these cognitive benefits across life-span may help to explain that bilingualism has been found to protect against the onset of dementia, which began 4 years later in bilinguals than monolinguals according to a report with almost 200 patients (Bialystok, Craik, & Freedman, 2007).

1.3. Bilingualism, Language Processes and Specific Language Impairment Bilingualism has also been an approach of interest to researchers studying Specific Language Impairment (SLI). SLI is characterized by a deficit in language production and/or understanding in the absence of other significant developmental or hearing deficits (Bishop & Leonard, 2000; Leonard, 1998; Schwartz, 2008). Children with SLI have expressive and/or receptive language scores that are significantly below those of their age-matched children with Typical Language Development (TLD), but their standardized scores of nonverbal

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Psycholinguistic Abilities and Phonological Working Memory…

69

intelligence are within normal limits (DSM-IV-TR, 2000; ICD-10, 1993). The DSM-IV-TR differentiates between either expressive or mixed receptive-expressive language disorders (315.31/315.32), whereas ICD-10 distinguishes between either expressive or receptive language disorders (F80.1/F80.2). However, experts in SLI are still far from an agreement about possible subtypes of SLI. Cross-cultural research of SLI could be a helpful approach to shed more light on this issue and other features of SLI. In fact, identifying Specific Language Impairment is still a challenge for professionals even in school-aged children. The diagnosis is even harder in bilingual populations, which need to be assessed in both languages (to confirm that their psycholinguistic abilities for each language are significantly below the age norms). Particularly, bilingual assessment usually needs to face the limited availability of: (a) similar reliable standardized tests across languages (which are more limited for minority languages); (b) norms that are based on bilingual population (not on monolinguals); (c) bilingual professionals that are qualified to assess the two languages (especially for languages from another country); etc. On the other hand, the increasing mobility of individuals and information has driven an unstoppable trend toward cultural uniformity, including the extinction of some minority languages (since language and culture are closely related). However, a need remains for further development of cross-cultural measures that can reliably identify SLI with worldwide uniformity, especially considering its high incidence. SLI affects approximately 7% of monolingual English-speaking kindergarten children (8% of boys and 6% of girls) in U.S.A. (Tomblin et al., 1997). According to the authors, parental education also modulates this prevalence. More epidemiologic studies are needed including bilingual populations. There are several empirical studies about language skills and cognitive processes in bilingual children with SLI. One of these works compared two groups of monolingual children with SLI (French-/English-speakers) with a group of age-matched French-English simultaneous bilingual children with SLI of around 7 years of age (Paradis, Crago, Genesee, & Rice, 2003). The three groups showed similar deficit patterns with respect to tense-marking morphology in spontaneous language samples (from play sessions). According to their results, the authors suggested that bilingual education might not interfere with the overall course of language acquisition (at least in morphology) under conditions of impairment. However, they also suggested that adding some groups of children with TLD could clarify more the influence of bilingual education (excluding SLI) in the domain of grammatical morphology. From a clinician neurophysiology approach, changes in the left/right auditory cortices (including strengthening of the mismatch response) were found in children with SLI mostly bilingual of Swedish/Finnish languages (6-7 years), after receiving a phonological intervention program, (Pihko et al., 2007). The 8-week program focused mostly upon phonological discrimination, awareness and production, with a special emphasis on sounds that occur in Swedish (the dominant language for most of them) but not in Finnish; MEG recordings were obtained before and after the intervention. In fact, phonological working memory deficits have been found in monolingual and bilingual children with SLI. Particularly, the Non-word Repetition Task in English (Dollaghan & Campbell, 1998; Gathercole & Baddeley, 1996) and Spanish (Girbau & Schwartz, 2007, 2008) seem to work as a possible good screener for identifying SLI in native speaking children of one of these respective languages. The English task appears to lose sensitivity when it is administered in the child's L2; at least this was the case for Spanish-

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

70

Dolors Girbau

English sequential bilingual children performing the Non-word Repetition Task in English as L2 (Girbau, 2008).

1.4. Purposes of the Research Our main goal is to present a comparative cross-cultural analysis between two groups of children from U.S.A. (with SLI/TLD) and two groups of children from Spain (with SLI/TLD), who were involved in some previous studies (Girbau, 2008; Girbau & Schwartz, 2007, 2008). All children were sequential bilingual with L1 = Spanish, but with a different L2 (either English or Catalan). We are interested in comparing the psycholinguistic abilities of the four groups of bilingual children with SLI/TLD. We also want to compare their performance in the Spanish Non-word Repetition Task in order to know the possible cross-cultural variants between groups. Finally, we will analyze the relationship between language skills, Phonological Working Memory and age by comparing the bilingual groups.

2. Method

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

2.1. Participant Recruitment in U.S.A. and Spain More than fifty native Spanish-speaking children with language impairment or typical development (with no articulation deficit) were referred by school psychologists, speech pathologists and/or teachers. All children with SLI were receiving language therapy (except for two participants who were scheduled to begin receiving intervention). Their SLI status was also confirmed by the judgments of their parents and teachers, which appears to be a highly reliable indicator of language status (Restrepo 1998). Their parents signed a consent form to participate in the study.

2.1.1. Recruitment of Spanish-English Bilingual Children Most of the Spanish-English sequential bilingual children were recruited from a dual language public school in New York City (NYC, Bronx). This school provides instruction to native speakers of either English or Spanish (50% in each classroom), both in English and Spanish languages on alternate days. All participants came from Spanish-speaking homes in NYC, with low/low-middle Socio-Economic Status (SES). According to the SES questionnaire (Hollingshead, 1975) that was administered to 15 participants, their SES average was middle-low, (M = 28.87 [Range = 14-58], SD = 12.28); we could not get enough SES information from 7 participants. The native language for both parents was Spanish. For all children, Spanish was their first language and English was their second language, according to their testing scores, parent questionnaire, teachers, etc.

2.1.2. Recruitment of Spanish-Catalan Bilingual Children The Spanish-Catalan bilingual children were recruited from two middle/middle-low SES public schools in Spain (Castelló). The main teaching language at both schools was Spanish;

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Psycholinguistic Abilities and Phonological Working Memory…

71

Catalan was also taught as an obligatory language course. In the Valencia region, the diglossic bilingualism is frequent: Spanish is usually considered a formal (academic) language and Catalan a non-formal language. In that region, there are many passive bilinguals who are Spanish speakers that understand Catalan (but they do not usually speak it). Our participants came from Spanish-speaking homes, but they also understood Catalan. According to the SES questionnaire (Hollingshead, 1975) that was administered to all participants, their SES average was middle (M = 31.48 [Range = 11-66], SD = 13.53).

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

2.2. Bilingual Testing of Children in U.S.A. and Spain Each participant passed a hearing screening at 20 dB for at least 5 frequencies (4000, 3000, 2000, 1000, and 500 Hz) at our university facility, following the American National Standards Institute (2004a, 2004b). Their parents completed a questionnaire to determine the extent to which Spanish was the child's primary language (and the exposure to other languages), their SES, the family and child's history of language deficits, their place of birth, and that the child did not have any history of neurological disorders, or behavior features of autism. According to parents' information, all children had Spanish as a first language and either English or Catalan as a second one. The primary Spanish language measure was the Spanish version of the Illinois Test of Psycholinguistic Abilities with norms from Spain (ITPA; Kirk, McCarthy, & Kirk, 2001). It was administered to compare children from U.S.A. and Spain, and because of the scarce availability of reliable Spanish tests with norms from U.S.A. for this age range. The ITPA consists of 11 subtests from which we selected the 4 subtests of language production/comprehension that most closely correspond to those measured by more contemporary language tests. In the Grammatical Integration subtest, the child completes sentences spoken by the psychologist according to related pictures (e.g., "Here there is a dog, here there are two … (dogs)"). In the Auditory Comprehension subtest, the child listens to brief stories and answers to questions about the stories by pointing to pictures. In the Auditory Association subtest, the participant completes sentences spoken by the psychologist (e.g., "The father is big, the child is… small."). The Verbal Expression subtest is a lexical fluency task; the child is given a category (e.g., body parts, fruits) and is asked to say as many items in that particular category as possible in one minute. The ITPA was individually administered to all participants. Each subtest raw score was converted into a z-score through a formula [(raw score – M) / SD)]; the M and SD were from the particular age norms. For each children's group, we calculated a mean of the z-scores for each subtest of the ITPA, and for the average z-score of the four subtests (overall mean).

2.2.1. Bilingual Testing of Spanish-English Speaking Children Spanish-English bilinguals included eleven children with SLI and 11 age-matched children with TLD (7;6 to 10;10). Their parents were born in Latin American Spanishspeaking countries and they were native monolingual speakers of Spanish with some exposure to English. The mean age for the group of children with SLI was 8;10 (SD = 13.42 months, Range = 90 to 126 months) and the mean age for the group with TLD was 9;1 (SD = 14.83 months, Range = 90 to 130 months). There were five boys and six girls in each group.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

72

Dolors Girbau

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Each participant with SLI was matched on age (in months) and gender with a child with TLD, as closely as possible. The average age difference [TLD – SLI] was M = 3.55 months (SD = 3.42). All participants performed within normal limits on the TONI-3 – Test of Nonverbal Intelligence (TONI-3; Brown, Sherbenou, & Johnsen, 1997). For children with SLI, the IQ average was M = 95.09 with SD = 6.82 (Range = 88 to 107), and for children with TLD, M = 103.64 and SD = 12.92 (Range = 84 to 122). The children with SLI had a significant low proficiency in both Spanish and English languages (Figure 1). In the Spanish ITPA test, the participants with SLI who were included all scored: (a) below -1.82 SD from the mean z-score on at least two of the four selected subtests; and (b) below -1.94 SD on the average z-score of these four language-relevant subtests. The children with TLD who were included all scored: (a) no lower than -1.28 SD from the mean on at least two of the four selected subtests; and (b) no lower than -1.35 SD from the mean z-scores of these four language-relevant subtests.

Figure 1. Mean standard deviations in relation to mean scores from the norms, (and Standard Deviations), for the Spanish ITPA subtests scores: bilingual children with TLD/SLI from U.S.A./Spain

The primary English language measure was the English CELF-3 Screening test (Semel, Wiig, & Secord, 1996). The raw score was converted into a z-score by using the formula mentioned above for the ITPA. As expected, the proficiency in English (L2) was significantly below the age norms in the group of children with TLD (M = -1.19, SD = 1.42) and children with SLI (M = -2.42, SD = 0.67).

2.2.2. Bilingual Testing of Spanish-Catalan Speaking Children Eleven children with SLI and 11 children with TLD, aged 8;3–10;11, participated in the study that was done in Spain. There were four girls and seven boys in each group. Each child

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Psycholinguistic Abilities and Phonological Working Memory…

73

with SLI was matched for age and gender with a child with TLD, as closely as possible. The average age difference [TLD – SLI] was M = 1.55 months with SD = 3.53. The mean age for the group of children with SLI was M = 9;5, SD = 8.54 months (ranging from 99 to 131). For the group of children with TLD, M = 9;6, SD = 7.88 months (Range=103 to 127 months). The children with TLD had performance intelligence IQs ranging from 94 to 128 (M = 111.08, SD = 9.19), on the basis of the "Batería de Aptitudes Diferenciales y Generales" test (BADYG E2; Yuste, 2002). The children with SLI had IQs based on the TONI-2 (Brown et al., 2000) ranging from 83 to 132 (M = 101.36, SD = 17.72). As for the Spanish ITPA, the children with SLI who were included all scored: (1) below zero on the average z-score of the four language-relevant subtests; and (2) at least -1.0 SD below the means on at least two of the four selected subtests. The children with TLD who were included all scored: (1) above zero on the average z-score of these four subtests; and (2) more than -1.0 SD from the mean on any of the four selected subtests; (see Figure 1). The availability of tests for Catalan language at this age range is scarce, but Catalan and Spanish are romance languages that are very close to each other. Their knowledge of Catalan language (L2) was assessed through the parent questionnaire (including a detailed section on child's language background and use), the professionals working at the school, and our talk with children. They were passive bilinguals of Catalan.

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

2.3. Spanish/English Non-word Repetition Tasks and Bilingualism A Non-word Repetition Task following Spanish phonotactic patterns was administered to all participants (Girbau & Schwartz, 2007, 2008). It includes 20 audio-recorded pseudowords pseudowords (e.g., /mún.tir/), four at each of five syllable lengths. Moreover, an English Nonword Repetition Task (Dollaghan & Campbell, 1998) was administered only to SpanishEnglish Speaking Children (counterbalancing the order of administration of both tasks); (see Girbau, 2008, for additional details). There were 16 recorded non-words (e.g., /teivak/), which were phonotactically English in their segmental content, four at each of four syllable lengths. The tasks were presented by the author to a particular child individually through a computer with headphones. Immediately after listening to each pseudoword, the participant had to repeat it. We analyzed the number of correct non-words for each of the two tasks.

3. Results We compared the age (in months) of the two bilingual groups (Spanish-English/SpanishCatalan) for either children with TLD or children with SLI. The two bilingual groups did not differ significantly in age for children with TLD [F(1, 20) = 1.05, p = .32] or children with SLI [F(1, 20) = 2.13, p = .16], on the basis of the one-way ANOVAs. In our previous studies, the two language status groups (SLI/TLD) did not differ significantly in age either for Spanish-English bilinguals (Girbau & Schwartz, 2008) or Spanish-Catalan bilinguals (Girbau & Schwartz, 2007). Therefore, the age variable was not considered in the next analyses about their psycholinguistic abilities, phonological working memory, and the correlation between them.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

74

Dolors Girbau

3.1. Psycholinguistic Abilities in Bilingual Children The psycholinguistic profiles of these four subgroups, on the basis of the Spanish ITPA test, are shown in Figure 1. Both bilingual groups of children with TLD (from Spain/U.S.A.) scored notoriously above any of the two bilingual groups of children with SLI, in all ITPA subtests scores. However, Spanish-English bilingual children with SLI performed somewhat poorer than Spanish-Catalan bilingual children with SLI in the mentioned ITPA, especially in the Grammatical Integration subtest. The average difference (in the same direction) between the two bilingual groups was rather small for children with TLD.

3.2. Phonological Working Memory in Bilingual Children

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

The focus of the analyses for phonological working memory was on the percentage of correct non-words. We analyzed both the total scores in the Non-word Repetition Task and a combined score for three-, four-, and five-syllable non-words (3-4-5 composite), since most errors occurred on these items. We transformed all percentages using arcsine prior to statistical analysis.

Figure 2. Mean overall/3-4-5 syllable composite percentages of correct non-words (and Standard Deviations) in bilingual children with TLD/SLI from U.S.A./Spain

Several one-way analyses of variance (ANOVAs) revealed that only the two bilingual groups with TLD differed in the percentage of total number of non-words correct [F(1, 20) = 32.22, p < .00005] and in the 3-4-5 composite [F(1, 20) = 37.35, p < .00001]. The two bilingual groups with SLI did not differ significantly in the percentage of total number of non-words correct [F(1, 20) = 1.11, p = .30] or the 3-4-5 composite [F(1, 20) = .28, p = .60].

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Psycholinguistic Abilities and Phonological Working Memory…

75

Thus, results in the Spanish Non-word Repetition Task showed some cross-cultural differences only for children with TLD; all children with SLI performed poorly. The accuracy in this task for children with TLD from Spain was higher than for children from U.S.A. (see Figure 2). Particularly, for the subset of items from 3 to 5 syllables in length, both groups differed almost 30%.

3.3. Relation of Bilingual Children's Psycholinguistic Abilities to Phonological Working Memory

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

We were also interested in analyzing the relation between the non-word repetition accuracy and the Spanish ITPA test for either children with SLI or children with TLD. In our previous studies, we found significant correlations for Auditory Association and Grammatical Integration subtests in both groups of children together (with SLI and TLD), for each of the two countries (Girbau & Schwartz, 2007, 2008). However, the correlation with the Auditory Comprehension subtest was only significant in the participants from U.S.A. (not in those from Spain).

Note. * p < .05 ** p < .01 *** p < .005. For all correlation scores, according to the Bonferroni correction, the significant p-value starts at p = .0125 (i.e., the probability of .05 divided by the 4 correlations we performed for each hypothesis). According to this, the correlations between the Auditory Association and the total percent correct non-words (p = .013) for children with TLD did not reach significance. The correlation of Grammatical Integration scores with the composite 3-4-5-syllable percent of correct nonwords (p = .025) for children with TLD was not considered significant either.

Figure 3. Pearson correlations between the four subtests of the Spanish ITPA test, and the overall/3-4-5 syllable composite percentages of correct non-words in the Spanish Non-word Repetition task.

In the present study, we focused in each of the two groups of children (with either SLI or TLD) from both countries together. We calculated Pearson Product Moment Correlations Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

76

Dolors Girbau

between the raw scores of the four ITPA subtests, and the overall non-word repetition percentages or the 3-4-5 syllable composite percentages (Figure 3). Significant correlations for the Auditory Association subtest were found in the two groups of children with SLI/TLD (for both countries together). The correlation between the Grammatical Integration subtest and the Spanish Non-word Repetition Task was significant in children with SLI, but it did not reach significance in children with TLD (after applying the Bonferroni correction).

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

4. Discussion The psycholinguistic abilities in any of the two bilingual groups of children with TLD are significantly higher than in any bilingual group of children with SLI, independently of their particular second language (Catalan/English). Thus, the ITPA test was sensitive in identifying SLI for both bilingual groups with Spanish as a first language (and L2 = Catalan/English), despite the norms of the test were based on monolingual speakers from Spain. However, Spanish-English bilingual children with SLI from U.S.A. performed somewhat poorer than Spanish-Catalan bilingual children with SLI from Spain in the ITPA, especially in the Grammatical Integration subtest. This may suggest that some psycholinguistic abilities of those children with SLI could be a bit more affected by a lower SES, poorer language input and/or a more challenging bilingual context (since Spanish/English languages are more different than Spanish/Catalan). Another possible explanation may be related with the grammatical content and language of the speech-language therapy that almost all children with SLI were receiving. For example, in Spain it was mostly focused in their L1 (Spanish), and in U.S.A. mostly in their L2 (English). Several authors have outlined the importance of assessing and also treating bilingual children in their mother tongue (e.g., Pert & Stow, 2001; Stow & Dodd, 2003). Pihko et al. (2007) found some significant brain's changes in bilingual children with SLI, after they received intervention with a special emphasis in the dominant language. On the other hand, we analyzed the performance on the Spanish Non-word Repetition Task by the four groups of unbalanced bilingual children. The phonological working memory task was sensitive in identifying SLI both in Spain (Girbau & Schwartz, 2007) and U.S.A. (Girbau & Schwartz, 2008). According to these works, both groups of bilingual children with SLI (from Spain/U.S.A.) showed a significant poorer performance than the two respective groups of bilingual children with TLD (from Spain/U.S.A.). However, in the present study we found some significant cross-cultural differences between the two groups of children with TLD, but not for the two groups of children with SLI. Children with TLD from Spain produced a significant higher number of correct non-words than children with TLD from U.S.A. This effect was somewhat larger for the subset of non-words from 3 to 5 syllables in length, since most errors occurred on these items. This significant difference may be explained by their different social/bilingual backgrounds, including the exposure to English phonetics in children from NYC; English and Spanish phonetics are more different than Catalan/Spanish phonetics. Previous research (Schwartz et al., 2008) found a bilingual advantage, associated to the transfer from Russian (L1) to Hebrew (L2), for some Hebrew phonological tasks different from our non-word repetition task. However, the levels of transfer between two languages depend on their particular characteristics/similarities.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Psycholinguistic Abilities and Phonological Working Memory…

77

Conversely, both groups of children with SLI showed a similar phonological working memory deficit. This deficit was also found to be a strong marker for SLI in monolingual children (e.g., Dollaghan & Campbell, 1998; Gathercole & Baddeley, 1996). The usual difficulty that children with SLI exhibit to perform this phonological task may explain the similar poor performance for both groups with SLI (despite their different bilingual backgrounds). It is also important to consider that the Non-word Repetition Task in child's L1 (Spanish) was a good marker for SLI. However, this was not found when it was administered in the child's second language (English), since both groups of children with SLI/TLD performed with similar low accuracy (Girbau, 2008). This seems to reflect the extent to which children’s phonotactic representations in their dominant language influence their performance on non-word repetition. It also emphasizes the importance of assessing children's language skills in their first language (besides assessing their L2). A previous work found similar deficit patterns in tense-marking morphology for French-/English-monolingual children with SLI and French-English simultaneous bilingual children with SLI of around 7 years of age (Paradis et al., 2003). Finally, the non-word repetition accuracy was found to correlate significantly with the Auditory Association subtest from the Spanish ITPA, in either children with SLI or children with TLD (for both countries together). This subtest also correlated significantly in both groups of children together (with SLI and TLD), in either Spain (Girbau & Schwartz, 2007) or U.S.A. (Girbau & Schwartz, 2008). Thus, our results support a strong relation of Phonological Working Memory to Auditory Association, independently of the presence of language impairment and the type of bilingual contexts (i.e., the particular second language, SES, etc.). In fact, this subtest involves auditory working memory demands similar to the non-word repetition task. It requires child’s oral completion of a sentence presented orally by the psychologist without any visual support. This would explain the strong correlation that was found across all our studies. The Grammatical Integration subtest correlated significantly for children with SLI, but the correlation did not reach significance for children with TLD (after applying the Bonferroni correction). This subtest also correlated significantly in both groups of children together (with SLI and TLD), in either Spain (Girbau & Schwartz, 2007) or U.S.A. (Girbau & Schwartz, 2008). Thus, our results support a relation of Phonological Working Memory to Grammatical Integration, mostly in children with SLI, and independently of the type of bilingual contexts. This subtest requires child’s oral completion of a sentence presented orally by the examiner with visual support (related pictures). Thus, it involves some auditory working memory demands similar to the non-word repetition task, but thanks to the visual support it seems to be a bit less auditory/memory demanding than the Auditory Association subtest. Maybe children with TLD benefitted somewhat more from the visual support, so they required less auditory working memory resources. The general high strengths of all these correlations are not usual in publications that examine the relationship between this phonological working memory task and any measures of language (see Girbau & Schwartz, 2007, 2008, for a review). More research is needed to better know the universal language/cognitive deficits in bilingual children with SLI. It would be interesting to compare these bilinguals with Spanish monolinguals without any exposure to a second language (i.e., from a monolingual region). A further development of standardized tests and instruments across languages, especially for minority languages, would also be helpful to identify the growing population of bilingual

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

78

Dolors Girbau

children with SLI. In the near future, neurophysiology and brain imaging techniques will probably shed more light to our understanding of the best approach to assessment, education, and intervention for bilingual children with SLI.

Acknowledgments This research was funded by a grant from the "Ministerio de Educación, Cultura y Deporte" of Spain (Secretaría de Estado de Universidades e Investigación) PR2003-0061 to the first autor; and two national grants from the "Instituto de Salud Carlos III-Ministerio de Sanidad y Consumo", FIS-PI041733, and "Ministerio de Educación y Ciencia", SEJ200760325/PSIC, (D. Girbau, P.I.). I would like to thank Richard G. Schwartz for his advice and collaboration in planning this research.

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

5. References American National Standards Institute (2004a). Methods for manual pure-tone threshold audiometry (ANSI S3.21-2004). New York: American National Standards Institute. American National Standards Institute (2004b). Specifications for audiometers (ANSI S3.62004). New York: American National Standards Institute. American Psychiatric Association (2000). Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition, Text Revision (DSM-IV-TR). Bialystok, E. (2007). Cognitive effects of bilingualism: How linguistic experience leads to cognitive change. International Journal of Bilingual Education and Bilingualism, 10(3), 210-223. Bialystok, E., Craik, F. I. M., & Freedman, M. (2007). Bilingualism as a protection against the onset of symptoms of dementia. Neuropsychologia, 45(2), 459-464. Bialystok, E., Craik, F.I.M., & Luk, G. (2008). Lexical access in bilinguals: Effects of vocabulary size and executive control. Journal of Neurolinguistics, 21(6), 522-538. Bishop, D. V. M., & Leonard, L. B. (Eds.) (2000). Speech and language impairments in children: causes, characteristics, intervention and outcome. Hove: Psychology Press. Brown, L., Sherbenou, R. J., & Johnsen, S. K. (1997). TONI-3: Test of Nonverbal Intelligence (3rd ed.). Austin, TX: PRO-ED. Brown, L., Sherbenou, R. J., & Johnsen, S. K. (2000). TONI-2. Test de inteligencia no verbal: apreciación de la habilidad cognitiva sin influencia del lenguaje [Test of Nonverbal Intelligence]. Madrid: TEA. Dollaghan, C. A., & Campbell, T. (1998). Nonword repetition and child language impairment. Journal of Speech, Language, and Hearing Research, 41, 1136–1146. Gathercole, S. E., & Baddeley, A. D. (1996). The Children's Test of Nonword Repetition. London: Psychological Corporation. Generalitat de Catalunya (2008). Estudi sociodemogràfic i lingüístic de l’alumnat de 4t d’ESO de Catalunya: Avaluació de l’educació secundària obligatòria 2006 [Sociodemographic and linguistic study of 4 grade secondary school students in Catalonia: Assessment of obligatory secondary education 2006]. Col·lecció Informes d’avaluació 11. Barcelona: Generalitat de Catalunya.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Psycholinguistic Abilities and Phonological Working Memory…

79

Girbau, D. (2002). Psicología de la Comunicación [Psychology of communication]. Barcelona: Ariel. Girbau, D. (2008). A Bilingual Approach to the Study of Phonological Short-Term Memory: Spanish-English Speaking Children with and without Specific Language Impairment. In F. Columbus (Ed.), New Research on Short-Term Memory (pp. 223-240). New York: Nova Science Publishers. Girbau, D., & Schwartz, R. G. (2007). Non-word repetition in Spanish-speaking children with Specific Language Impairment (SLI). International Journal of Language & Communication Disorders, 42(1), 59-75. Girbau, D., & Schwartz, R. G. (2008). Phonological Working Memory in Spanish-English bilingual children with and without Specific Language Impairment. Journal of Communication Disorders, 41(2), 124-145. Hollingshead, A. B. (1975). Four-factor Index of Social Status. Unpublished manuscript. New Haven, CT: Yale University. Kirk, S. A., McCarthy, J. J., & Kirk, W. D. (2001). ITPA: Test Illinois de Aptitudes Psicolingüísticas [Illinois Test of Psycholinguistic Abilities]. Madrid: TEA. Leonard, L. B. (1998). Children with Specific Language Impairment. Cambridge, MA: MIT Press. Mayo, L. H., Florentine, M., & Buus, S. (1997) Age of second-language acquisition and perception of speech in noise. Journal of Speech, Language and Hearing Research, 40, 686-693. Medina, M., Saldate, M., & Mishra, S. (1985).The sustaining effects of bilingual education: A follow-up study. Journal of Instructional Psychology, 12(3), 132-139. Paradis, J., Crago, M., Genesee, F., & Rice, M. (2003). French-English bilingual children with SLI: How do they compare with their monolingual peers? Journal of Speech, Language and Hearing Research, 46, 1-15. Peal, E., & Lambert, W. E. (1962). The relation of bilingualism to intelligence. Psychological Monographs, 76(27), 1-23. Pert, S., & Stow, C. (2001). Language remediation in mother tongue: A paediatric multilingual picture resource. International Journal of Language and Communication Disorders, 36, 303-308. Pihko, E., Mickos, A., Kujala, T., Pihlgren, A., Westman, M., Alku, P., Byring, R., & Korkman, M. (2007). Group intervention changes brain activity in bilingual languageimpaired children. Cerebral Cortex, 17, 849-858. Ransdell, S., Arecco, M. R., & Levy, C. M. (2001). Bilingual long-term working memory: The effects of working memory loads on writing quality and fluency. Applied Psycholinguistics, 22(1), 113-128. Restrepo, M. A. (1998). Identifiers of predominantly Spanish-speaking children with language impairment. Journal of Speech, Language, and Hearing Research, 41(6), 1398– 1411. Schuele, C. M. (2001). Socioeconomic influences on children’s language acquisition. Journal of Speech-Language Pathology and Audiology, 25(2), 77–88. Schwartz, R. G. (Ed.) (2009). The handbook of child language disorders. New York, NY: Psychology Press.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

80

Dolors Girbau

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Schwartz, M., Share, D. L., Leikin, M., & Kozminsky, E. (2008). On the benefits of biliteracy: just a head start in reading or specific orthographic insights? Reading and Writing, 21(9), 905-927. Semel, E., Wiig, E. H., & Secord, W. A. (1996). CELF-3 Screening Test (Clinical Evaluation of Language Fundamentals— Third Edition). San Antonio, TX: Harcourt. Stow, C., & Dodd, B. (2003). Providing an equitable service to bilingual children in the UK: a review. International Journal of Language and Communication Disorders, 38(4), 351377. Suro, R. (2005). Special Section: Hispanic Americans. A growing minority. The World Almanac and Book of Facts 2005, 7. Tomblin, J. B., Records, N. L., Buckwalter, P., Zhang, X., Smith, E., & O'Brian, M. (1997). The prevalence of specific language impairment in kindergarten children. Journal of Speech, Language, and Hearing Research, 40, 1245-1260. U.S. Census Bureau (2008). Statistical Abstract of the United States: 2009 (128th Edition). Washington, DC: U.S. Census Bureau. World Health Organization (1993). The ICD-10 Classification of Mental and Behavioural Disorders: Diagnostic Criteria for Research. Geneva: World Health Organization. Yuste, C. (2002). BADYG E2. Baterıía de Aptitudes Diferenciales y Generales [BADYG E2. Battery of Differential and General Abilities]. Madrid: Cepe.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

In: Bilinguals: Cognition, Education and Language Processing ISBN: 978-1-60741-710-1 Editor: Earl F. Caldwell, pp. 81-102 © 2010 Nova Science Publishers, Inc.

Chapter 4

PSYCHOLINGUISTIC CHALLENGES IN PROCESSING THE ARABIC LANGUAGE Raphiq Ibrahim Learning Disabilities Department, and The Edmond J. Safra Brain Research Center for the Study of Learning Disabilities, University of Haifa, Haifa, Israel

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Abstract The 2006 PISA (Program for International Student Assessment ) report of worldwide scholastic achievements showed that about 50% of Israeli Arabic students were found to exhibit the lowest reading achievement scores in the PISA tests (level 1 and below) as compared to the other participating groups. Also, the MEITZAV national testing program in Israel (2001–2002) showed an achievement gap in language skills (reading and reading comprehension) between Arab students and Jewish students in the school systems. This gap was larger than those found in the other areas tested (mathematics, science and technology, and English). The aim of this chapter is to explore the cognitive basis of these difficulties, specifically the diglossic situation in Arabic. Furthermore, the chapter discusses the unique features of the Arabic language that might contribute to the inhibition and slowness of reading acquisition and might even hinder the acquisition of basic academic skills. Finally, a model with a comprehensive basis (cognitive and neurocognitive) will be built in order to explain the complex linguistic situation of beginning Arabic learners. This chapter is concerned with the cognitive evidence bearing on the nature of the mechanisms of language processing in Arabic which has critical linguistic characteristics and a diglossic factor. Additionally, other aspects, including a neurofunctional perspective, will be discussed.

1. Literacy and the Arabic Language It is clear that reading and writing are crucial functioning factors in a literate society. Illiteracy statistics (UNESCO- 2001) reported by the International Bureau of Education (IBE) within and across the Arab States showed that illiteracy rates vary widely in our region, ranging from 10.2% in Jordan to 59% in Mauritania. These high rates of illiteracy that

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

82

Raphiq Ibrahim

characterize the Arab region, first of all, seem to indicate that the educational system is failing and that there is a growing inadequacy and deterioration of education in the Arab states. UNICEF statistics for the PISA (Program for International Student Assessment) report of worldwide scholastic achievements (2006) placed Israel in 40th place out of 57 countries in reading achievement. This marked a drop from 30th place in 2002. Data indicated that about 50% of Israeli Arabic students were found to exhibit the lowest achievement scores in the PISA tests (level 1 and below) as compared to the other participating countries. The Ministry of Education’s MEITZAV (national testing program) in 2001–2002 showed an achievement gap in language skills (reading and reading comprehension) between Arab students and Jewish students in the school systems. This gap was larger than those found in the other areas tested (mathematics, science and technology, and English). Hebrew-speaking students achieved an average of 79%, while the Arab students’ average was 60.9%. Education officials agree that the poor reading skills in Arabic have a broad impact on a student’s achievement potential. At the same time, they admit that the Arabic language faces a real crisis in the education system, and pedagogical programs needed to change this situation. Sociolinguists looked at these data from a social and societal viewpoint on literacy. They see in the “diglossic” situation in Arabic a marked differentiation between two related varieties of Arabic: on one hand, fuṣḥa, which is mostly used for “high” functions such as formal prayers, speeches, or lectures; and, on the other hand, ammia, Arabic dialects usually used for “low” functions, defined as home and family discourse or trade and market conversations within Arab societies. According to Maamouri (1998), the gap between the language of orality (spoken dialect at home) and the language of literacy, more commonly referred to as Modern Standard Arabic (MSA), seems to be a major cause of low learning achievement in schools and low adult literacy levels everywhere in the Arab region. Maamouri described this nicely. He claimed that fuṣḥa, which is at the same time “formal Arabic” is difficult to learn and use because it is nobody’s native language. He added that the mixture of language patterns in classrooms (fuṣḥa and dialectal Arabic code-switching) is a cause of serious pedagogical problems, sometimes leading to a lack of adequate language competence, to low linguistic selfconfidence, and usually to consequent social problems. Researchers from a psycholinguistics perspective showed that Arab speakers recognize single words in Arabic more slowly and less accurately in comparison to those for whom Hebrew and English are the first language (Frost, Katz and Bentin, 1997; Bentin and Ibrahim, 1996). Also, in text reading, other researchers showed poor reading achievement (speed and accuracy) in Arabic by Arab students (Saiegh-Haddad, 2003; Abu-Rabia, 2002). This might be related to some of the hardships that native Arabic beginning readers encounter and might even hinder their basic acquisition of basic academic skills. In this chapter I will elaborate this issue.

1.1. Review of Arabic Reading and Comprehension Research The question that many researchers tried to answer is whether a literate Arabic speaker is a bilingual de facto. Ayari (1996) and Maamouri (1998) thought that the diglossic situation in Arabic might be related to some of the hardships that native Arabic beginning readers encounter and might even hinder their basic acquisition of basic academic skills. To elaborate

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Psycholinguistic Challenges in Processing the Arabic Language

83

this issue, an important longitudinal research study on the acquisition of literacy and Arabic reading skills was done by Wagner in Morocco (1993; see also, Wagner, 2004). Wagner’s objective was to provide a profile of the variability that exists in Arabic literacy acquisition in Morocco. Wagner disscussed how Arabic speakers in Morocco become literate. He showed that knowledge in year one of Arabic letters, their graphemic variability and pronunciation predicted more than 30% of the variance in reading achievement five years later. Early decoding skills at the single-word level explain an additional 14% of the same variance. Wagner’s conclusion showed that there is “substantial reason to believe” that learning to read in Arabic necessitates an even greater reliance on decoding skills than in other languages. Wagner highlights “the absence of vocalization diacritics as the main reason behind the growing difficulty of decoding for word recognition and paragraph comprehension, a difficulty which mars advanced Arabic reading stages and requires knowledge of appropriately correct inflectional endings and the ability to place full and correct diacritical marking”. In an empirical research study undertaken on primary school reading errors and the role of diacritics for beginning readers, Azzam (1990) examines the misreadings and misspellings that Arab primary school children make and identifies vocalization and its use of diacritical markings as the main culprit. Her research seems to suggest that diacritical markings are significantly important in the process of reading and comprehending written language at all levels of Arabic reading. Salim Abu-Rabia (1998) investigated the effect of vowels on reading accuracy in Arabic orthography. Four kinds of written fuṣḥa Arabic texts (narrative, informative, poetic, and Koranic) were administered to sixty-four native Arabic speakers. Three texts of each kind were presented in three reading conditions: correctly vocalized, unvocalized, and wrongly vocalized. The most important finding of this study is that vowels were found to significantly influence the reading of both poor and skilled readers in the four fuṣḥa writing styles in all three conditions. It was also found that both skilled and poor readers improved their reading accuracy in all writing styles when they read with vowels. This last study reinforces and supports similar previous findings obtained by Abu-Rabia (1997), where it was demonstrated that the vowels and the sentence context were significant factors for word recognition for both skilled and poor fuṣḥa readers. Abu-Rabia (2000) investigated the contention that reading difficulties in Arabic in elementary school result from the diglossic situation of fuṣḥa, the language of books and school instructions, and its opposition to the spoken dialect of the home. Starting from the belief shared by educators, teachers, and parents that the exposure of young Arabic speakers to fuṣḥa in the preschool period is not useful and a burden to all, AbuRabia compared the reading comprehension performance of first and second grade children who had been experimentally exposed to literary Arabic throughout their preschooling period with the reading performance of a parallel control group only exposed to spoken Arabic during that period. He found, contrary to the commonly held belief, that the early exposure of Arab preschool children to fuṣḥa text (stories) enhances their reading comprehension abilities and improves their performance in reading comprehension tests two years later. Finally, more than his scientific findings, Abu-Rabia’s conclusions (2000: 155) are worth noting: (a) policymakers may incorporate this pedagogy in all preschool years, (b) educating elementary-school teachers and kindergarten teachers in diglossic issues, and (c) the recommendation that “teachers at all levels use literary Arabic as the language of instruction.” Elinor SaieghHaddad (2003) examined phonemic awareness and pseudo-word decoding in kindergarten

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

84

Raphiq Ibrahim

and first grade Arabic native children. She hypothesized that because native speakers of Arabic first learn to read in fuṣḥa, a language structurally different from the local dialect they grow up speaking, the linguistic differences between the two Arabic language varieties would interfere with the acquisition of basic reading processes in fuṣḥa. Saiegh-Haddad studied the role of oral language in the acquisition of basic fuṣḥa reading processes with purpose of researching the interface between exposure to fuṣḥa and top-level comprehension skill development, a vital issue for a theory of initial reading acquisition in diglossic or bi-dialectal settings. Going beyond just establishing a possible causal link between exposure to fuṣḥa and achieving toplevel comprehension reading skill development, Saiegh-Haddad (2003) addressed many aspects of such questions. First, do diglossic variables or linguistic distance parameters interfere with the acquisition of basic reading processes in fuṣḥa? Second, which diglossic structures interfere with the acquisition of basic reading skills—the phonological, syntactic, morphosyntactic or lexical? Third, which reading skills (phonemic awareness, word decoding, reading fluency, or reading comprehension) are sensitive to diglossic variables? That study focused on phonemic awareness and pseudo-word decoding because both are prerequisites to the acquisition of word reading. Its findings showed that although the first grade children seemed to have benefited from the increased exposure to fuṣḥa structures that formal literacy instruction allowed, they still found the task of isolating standard phonological structures quite difficult. The study showed that diglossia and the phonological distance between the two varieties of Arabic were related to the native decoding ability of the young Arab children. The findings on the relationships between the two forms of Arabic to the relations existing between LA and Hebrew using semantic and repetition priming techniques will be represented. In further study, we discuss if Arab children evince the metalinguistic abilities that have been found to characterize bilingual children and how these abilities affect reading acquisition. In related issue, we discuss the perceptual processes involved in the recognition of Arabic letters compared to Hebrew letters. In this discussion we tried to highlight the role of the additional visual complexity that characterizes Arabic orthography in reading acquisition.

1.2. Research on Bilingualism and Diglossia Research on bilingualism over the past three decades has focused on three central issues. The first is the nature of representation of words in the bilinguals’ mental lexicon, the second is the anatomical and functional organization of language abilities in the brains of monolinguals and bilinguals, and the third is the effect of bilingualism on cognitive and paralinguistic development. The consensus on these issues today may be summarized thus: although bilinguals have a more complex and possibly multistructured mental lexicon which is influenced by the idiosyncratic context in which the languages have been learned and by the structural relations between the languages (e.g., de Groot, 1992), this complexity need not result in differential cortical organization of linguistic abilities between bilinguals and monolinguals (Paradis, 1998). The central view was focused on the lexical organization of the two languages in bilinguals (Kirsner, 1986; for recent reviews see Kroll de and Groot, 1997). Of particular interest were questions addressing the relationship between semantically related words and translation equivalents across languages and the manner in which words in each

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Psycholinguistic Challenges in Processing the Arabic Language

85

language are connected to their meanings (e.g., de Groot, 1995; Grainger and Frenck-Mestre, 1998). Concerning the issue of cognitive representation of the two languages in bilinguals, several models have been proposed. Conceptual mediation models advocate a basic symmetry and independence of the two languages in the bilingual cognitive-linguistic system, at least for fluent bilinguals (Gerard and Scarborough, 1989; Grainger and Dijkstra, 1992). According to such models the words of each language are represented separately at the lexical level and connected indirectly via a common semantic system, which is accessed independently from each lexicon (Chen and Leung, 1989; Schwanenfeugel and Ray, 1986). Alternatively, lexical association views concur that each language is represented in a different lexicon, but propose that words in the two languages are interconnected and interact at the lexical level (Chen and Ho, 1986; Tzelgov, Henik, Sneg and Baruch, 1996). Assuming that the connection between words and their meanings is stronger in the first (L1) than in the second language (L2), a hierarchical model has been proposed by which the access of a word in the second language to its meaning is often mediated by its translation equivalent in the first language (Kroll, 1993; Kroll and Stewart, 1994). The connections between translation words at the lexical level are asymmetric but it is agreed by researchers that the strength of these connections is determined by the bilingual’s proficiency in the second language ( see, the revised hierarchical model, Kroll and Stewart, 1994; Chen and Leung, 1989; Tzelgov, Henik, and Leiser, 1990). Also, factors such as concreteness, morphological or phonological similarity between L1 and L2 words, seemed to affect translation between languages (de Groot, Dannenburg, and van Hell, 1994). Without committing ourselves to any of the models above, a safe conclusion is that the bilingual structural and functional organization is influenced by numerous factors, some related to the speaker characteristics and others related to words characteristics (for a review of such factors see de Groot, 1995). Assuming that the organization of languages in the cognitive-linguistic system is influenced by functional as well as by linguistic factors, Ibrahim and his colleagues (Ibrahim and Aharon-Peretz, 2005; Ibrahim, submitted) examined if diglossia in Arabic that presents a rare case in which speakers of the language actually use two languages (LA and SA) concurrently and intensively as a matter of course and not as an exception, constitute a private case of bilingualism. In Israel, Arab students who attend the Arab school system learn Hebrew and, at the high-school level, most students are as proficient in Hebrew as they are in LA. Therefore, if LA constitutes a second language for the Arabic speaker, then the results obtained regarding the lexical status of LA words and their connections to meaning should concur with patterns found for Hebrew. On the other hand, if the psychological reality has led to the combination of both forms of Arabic in a single lexicon, the results of linguistic manipulations between the two forms of Arabic should resemble those known to exist when the same linguistic manipulations are performed within a language. Two experiments, were designed to examine whether the intensive, daily interactive use of SA and LA , along with a psychological reality in which the two languages may be considered two forms of one language, may affect the lexical organization of these two languages in the cognitivelinguistic system of adult native speakers of SA. The relations between the two forms of Arabic were compared to the relations existing between Hebrew and SA. We compared semantic priming effects within Spoken Arabic, with the effects found across languages with written Arabic or in Hebrew being the other language. When both the primes and the targets were presented in SA, the semantic priming effect was significantly greater than when the prime and target words were from different languages. Further, the cross-language priming

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

86

Raphiq Ibrahim

effects on RTs were virtually the same, regardless of whether the second language was Hebrew or LA. Although the pattern of the priming effects on accuracy was not as clear as that on RTs, it was not radically different. Therefore, the overall results of the two experiments suggest that the representation of LA is that of L2, similar to Hebrew and that SA and LA have the status of two separate languages in the cognitive systems of Arabic native adults. Hence, from the linguistic (rather than social) perspective, literate Arabic speakers could be considered, de facto, bilinguals and contribute tothe debate exists as to whether the two forms of Arabic represent different languages, or whether this is a diglossic situation (Eid, 1990). However, a possible reason for the results in this study might be that the similar morphological structure of all three Semitic languages enabled the formation of strong lexical links between the words in SA, LA, and Hebrew, links that may have mediated semantic priming from L2 primes to L1 targets. It seems that second language primes facilitate performance with first language targets by attributing the effect to lexical connections between words in the second language and their translations in the first language, rather than attributing it to conceptual connections in the semantic network (Kroll, Sholl, Altarriba, Luppino, Moynihan and Sandres, 1992). In further study, Ibrahim (2006) examine to what extent the other linguistic effects (e.g., morpho-phonological similarity) are modulated by the psychological status of the second language. In that study, fast passing from one linguistic code to another in the auditory modality was experimentally studied. The mechanism of language switching, mentioned in the introduction called by other researchers “code switching” (Moreno, Federmeier and Kutas, 2002). Hence, with similar proficiency in Hebrew and LA, larger repetition effects between SA and LA than between Hebrew and LA translation equivalents could at least partly be attributed to a difference in subjective perception of Hebrew but not LA as a second language. The study of Ibrahim, (Ibrahim and Aharon-Perez, 2005) was designed to examine this hypothesis. Similar cross-language semantic priming effects from LA to SA and from Hebrew to SA was found, both about half the magnitude of the within language (SA) priming effects. These findings align nicely with the previously reported asymmetry in cross-lingual semantic priming (Altarriba, 1990; Keatly and deGelder, 1992; Keatly, Spinks and deGelder, 1994). The interpretation of the difference between the patterns of priming effect shed light on the lexical organization of the native speaker. The observed asymmetry of priming efficiency is usually attributed to the fact that words in a second language have looser connections with their meanings than do words in the first language. Therefore, the above described pattern of semantic priming effects suggests that, at least in regard to their connections with the semantic network, LA, as well as Hebrew, constitute second languages for the bilingual native speaker of SA. This directs us with a second account, which is priming between translation equivalents in Hebrew and LA reflected primarily semantic relationship and morpho-phonemic similarity (de Groot and Nas, 1991), Based on a distributed conceptual representations model (see for example, McClelland, Rumelhart and Hinton, 1986) de Groot (e.g., 1993, 1995) accounted for the morpho-phonemic similarity effect assuming that, having a common etymology, cognate translation equivalents share more meaning features than non-cognate translation equivalents and therefore priming between cognate translations is more effective than between noncognate translations. In essence, according to her view, repetition priming between translation equivalents, like semantic priming, originates from activation of common semantic features in the conceptual system. Because, the morpho-phonemic similarity for both priming languages was based on a shared root, the semantic overlap between cognate translation equivalents in

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Psycholinguistic Challenges in Processing the Arabic Language

87

SA and LA and Hebrew and LA should have been very similar. In further study, Ibrahim (2006) examined if the difference in semantic overlap of Hebrew-LA and SA-LA translation equivalants remained constant across lags. The results showed larger priming effects for cognate Hebrew–LA than SA–LA pairs at lag 0, on the one hand, and smaller priming effects at lag 4, on the other hand. Ibrahim concluded, that non-linguistic factors qualified the influence of the linguistic factors in determining the magnitude of the morpho-phonemic similarity effects. Specifically, he proposed that among these factors are lexical-episodic associations, which are apparently stronger between translation equivalents in two languages that are interactively and concomitantly used on every-day basis (such as SA and LA), than between translation equivalents in languages that are not concomitantly used (such as Hebrew and LA). In concert with previous findings, this study indicated that despite extremely intensive and concomitant use of Spoken and Literary Arabic does not bring about a change in their status as first and second languages, respectively. However, under such circumstances, associative links are formed between translation equivalents at the lexical level. The strength of these associations is determined by the frequency of concomitant use and by the psychological perception of the two languages as being forms of a single language.

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

1.2.1. Metalinguistic Ability, Bilingualism and Reading Metalinguistic awareness refers to the ability to think about the linguistic nature of the message and to be aware of certain properties of language such as its arbitrariness and phonological structure (Ben Zeev, 1977; Dash and Mishra, 1992; Edwards and Christophersen, 1988). To be metalinguistically aware is to solve linguistic problems such as the detection of ambiguity and grammaticality (Galambos and Hakuta, 1988; Galambos and Goldin-Meadow, 1990), to create new words (Titone, 1994), and to be able to segment words into their constituents (Campbell and Sais, 1995). All of these require an awareness of language as a system and the ability to access and manipulate knowledge about that system (Bialystok and Ryan, 1985). The received view is that the challenges posed by two linguistic systems promote the development of cognitive strategies that result in heightened metalinguistic abilities A second view in this area of research has been focused on the degree of bilingualism necessary for the emergence of metalinguistic advantage. The concept of “degree” of bilingualism can be seen in two ways. The view that has received attention is the definition of degree as level of facility or exposure to two languages. Here the results are equivocal, with some researchers reporting positive effects of bilingualism with minimal exposure to a foreign language (Yelland, Pollard, and Mercuri, 1993), while others have suggested that these effects are only seen when the child has acheived high facility in both languages (Cummins, 1987). Exposure to more than one language at an early age results in heightened awareness of the arbitrary and phonological aspects of language (Bialystok, 1991). The research reported here focuses on the last finding. Studies of the effect of bilingualism on cognitive and linguistic development have shown that the consistent and most conspicuous effect of bilingualism at an early age is on the development of metalinguistic awareness, where bilingual children achieve higher levels of performance than age matched monolinguals. In previous research in our lab (Eviatar and Ibrahim, 2001), we asked how early exposure to two languages affects the cognitive system. Our approach was as follows: Given that exposure to two languages results in a specific pattern of performance on three types of tests

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

88

Raphiq Ibrahim

(arbitrariness, phonological awareness, and vocabulary size) and given that this pattern is interpreted as reflecting the functioning of the cognitive system, will children exposed to the two forms of Arabic show this pattern? That is, are the two forms of Arabic different enough from each other to result in the pattern typical of bilingualism? To acieve this goal, Arabicspeaking children, who are exposed to both SA and LA were compared to Russian-Hebrew bilinguals and Hebrew-speaking monolinguals in tests included language arbitrariness, phonological awareness, and vocabulary. All of the children were in kindergarten or in first grade. The results showed the classic pattern resulting from exposure to two languages: higher performance levels in metalinguistic tests, and lower performance levels in the vocabulary measure as compared to monolinguals. The Arab children’s performance levels mimicked those of the bilingual children for the most part, and suggested that exposure to Literary Arabic in early childhood affects metalinguistic skills in the same manner as that reported for children exposed to two different languages. In addition, the comparison of kindergartners with the first graders revealed that exposure to literacy enhanced phonological awareness across all three groups. From the whole findings we concluded, that Arabic speaking children who are exposed to Literary Arabic behave as bilinguals. In a recent study (Ibrahim, Eviatar and Aharon-Peretz, 2007), we tried to explore the possibility that bilingualism, which has been shown to affect metalinguistic abilities, influences reading performance via these abilities. The consensus in the field is that learning a second language permits children to view their language as one system among others, and thereby enhances their reading ability. It is believed that the systematic separation of form and meaning that is experienced in an early bilingualism gives children added control of language processing. Focus on reading performance and the variables that influence it have revealed strong correlative relations with metalinguistic skills. The majority of previous investigations of the relationship between bilingualism and reading ability were conducted in English and other Indu-Europian languages. The general pattern of the effects of bilingualism is the following: bilinguals achieve higher scores than monolinguals on tests of arbitrariness (Edwards and Christofersen, 1988) and phonological awareness (Dash and Mishra, 1992), and lower scores than monolinguals on tests of vocabulary size (Doyle, Champagne and Segalwitz, 1978). Concerning phonological awareness many studies have demonstrated that children’s performance in various phonological awareness tasks are strongly related to the acquisition of reading skills in English (Bradly and Bryant, 1985), Italian (Cossu, G., Shankweiler, Liberman, Katz, and Tola, 1988), French (Bertlson, Morais, Alegria, and Content, 1985), Spanish (deManrique and Gramigna,1984). These researches hypothesized that the ability to reflect on phonemes presupposes the ability to reflect on words, but not vice verse. For example, to segment the word “cat” into its constituent phonemic elements, children must be able to dissociate the phonological realization of the word from its referent. However, these correlative studies tell us very little about the nature of this relationship. Specifically, we tried in our recent study (Ibrahim, Eviatar and Aharon-Peretz, 2007) to explore how the advantages in phonological awareness revealed by the Russian-Hebrew bilinguals and the Arab children are related to reading performance in first grade. Single word reading as well as connected text measures were compared among two groups of first graders of monolingual Hebrew speakers, bilingual Russian-Hebrew speakers. The two groups performed all tasks in Hebrew. An interesting result concerning language group results was that, the Arab children had lower scores than monolinguals on the tests of vocabulary and

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Psycholinguistic Challenges in Processing the Arabic Language

89

their performance on the text reading measures is significantly poorer. This support to the hypothesis that the bilingual children reading Hebrew were paying more attention to the task, as they made less errors than the monolingual children reading Hebrew. However, when the children were reading text, the large attentional demand made by the Russian-Hebrew in the letter and word identification stage resulted in less attentional resources available to higher processing of syntax and comprehension. This hypothesis is further supported by examination of the types of errors made by the bilingual children reading the text in Hebrew, which were mostly inaccuracies related to using false affixes (diacritics or letters) that generally represent the syntactic roles in the sentence and not false identification of the word itself. The whole results showed that language experience affected reading as Russian-Hebrew bilinguals were faster and more accurate in reading text than monolingual Hebrew children and both better than Arab children. A closer look at the results revealed that exposure to second language in early childhood affect reading skill among children in the first grade. In that regard this finding converges with other’s reports showed that bilingualism is powerful predictor of the speed and effieciency of reading acquisition (Da Fontoura and Siegel, 1995).

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

1.3. Linguistic Considerations Although the Arab children scores on tests of phonological awareness were higher than those of monolingual Hebrew speakers, their scores on tests of reading achievement were lower. We suggested that this is due to the complexity of Arabic orthography as compared with Hebrew orthography. Indeed, comparing Arabic and Hebrew orthographic systems with English, additional complexity is found in both orthographies but to a much larger extent in Arabic than in Hebrew. The two orthographies differ in two central aspects. The first aspect is related to shape of some letters that differ depending on their placement in the word. This phenomenon is much less extensive in Hebrew than in Arabic. In Hebrew there are five letters that change shape when they are word final: ( ‫ך‬-‫ כ‬,‫ם‬-‫מ‬, ‫ן‬-‫נ‬,‫ץ‬-‫צ‬,‫ף‬-‫) פ‬. The Arabic writing system is an alphabetic system with twenty eight basic consonant letters. Most of these consonants show a very close resemblance in form, with only additional dots or strokes to distinguish them from each other. They are usually composed of one base form and most of them have up to three or four distinct variant shapes. Graphemic variants differ depending on whether they occur independently (non-connectors) or in word initial, mid- or final position(for example, ). The second characteristic the phoneme /h/ is represented as: has to do with diacritics and dots. In Hebrew, dots occur only to mark vowels and as a stressmarking device (dagesh). In the case of three letters, this stress-marking device (which does not appear in unvowelized scripts) changes the phonemic representation of the letters from fricatives (v, x, f) to stops (b, k, p for the letters ‫ פ ק ב‬respectively). In the unvowelized form of the script, these letters can be disambiguated by their place in the word, as only word or syllable initial placement indicates the stop consonant. In Arabic, the use of dots is more extensive: many letters have a similar or even identical structure and are distinguished only on the basis of the existence, location and number of dots (e.g., the Arabic letters representing ) become the graphemes representing /th/ and /b/ ( /t/ and /n ( changing the number or location of dots.

) by adding or

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

90

Raphiq Ibrahim

1.3.1. Effects of Language Structure

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Orthography Recently, studies examining diverse orthographies including Serbo-Croatian, Chinease, Japanese, Hebrew and Arabic, have aimed to understand how differences among these orthographies influence processing and how word recognition processes are influenced by the characteristics of the writing systems (Frost, 1994). Orthography-related challenges, which usually result from centuries of use and misuse of the script, aggravate the linguistic problems described above. In order to deal with these linguistic issues in Arabic, we focused first on the spatial aspects of Arabic orthography and its influence on letter identification (Ibrahim, AharonPeretz and Eviatar, 2007). We tested the hypothesis that the graphemic complexity of Arabic is larger than in Hebrew, and that this results in additional perceptual load. The subjects tested were all adolescent healthy native Arabic speakers who had mastered Hebrew as a second language. We used oral and visual variants of the trail making test (Reitan and Wolfson, 1985) in both languages. Both versions have two levels of complexity: Level A requires connecting visually numbers or letters in order. Level B in the two modalities requires alternation between letters and numbers. Level B in the oral version requires declamation of the alternation. Performance time was the dependent variable. At the low level of complexity (Level A) there were no differences between performance in Hebrew and in Arabic in either the oral and the visual versions. In the more complex version (Level B), language (Hebrew or Arabic) did not affect speed in the oral version, but in the visual version, Arabic was performed significantly slower than Hebrew. We presented three versions of the Arabic visual test: only connected letters, only nonconnected letters, and a version using both types of letters. The main result showed that performance time was significantly faster in the nonconnected letters only version and both (connected and nonconnected) are slower than Hebrew letters. These results indicate that native Arabic speakers process Arabic orthography (first language) slower than Hebrew orthography (second language), and suggest that this is due to the additional complexity of the Arabic orthography. These findings supported our major conclusion that Arabic letters are harder to identify than Hebrew letters as a result of their greater visual complexity. We suggested that to these results could be implications regarding didactic methods of the learning Arabic orthography in early childhood and about the validity of tests that involve Arabic script in native Arabic speaking populations. In a further study (Eviatar and Ibrahim, 2004), we examined directly the effects of grapheme–phoneme conversion in English, Hebrew and Arabic, using a lateralized nonsense syllable identification task. The syllables were constructed as consonant–vowel– consonant (CVC) trigrams (the vowels in Hebrew and Arabic were letters that double as consonants or vowels), and the task of the participants was to identify the three letters. The stimuli were presented vertically in three conditions: left visual field (LVF), right visual field (RVF), and bilaterally (BVF). The participants were university students, native readers of each of the languages. In this study we were interested in errors, so we titrated exposure duration independently for each participant, in order to achieve a duration that resulted in 50% errors. This paradigm allowed us to measure three dependent measures that indexed different aspects of the task. The first measure was the mean exposure duration that was reached in each of the three groups of participants in order to achieve a 50% error rate. This is an index of the speed

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Psycholinguistic Challenges in Processing the Arabic Language

91

at which native readers can identify letters in each of the languages. This measure revealed that the Arabic readers required significantly longer exposure durations that the readers of Hebrew and English, and that Hebrew readers required significantly longer exposure durations that the readers of English. Thus, these results suggest that English letters are easier to identify, that Hebrew letters, and that Arabic letters are the hardest to identify. The second measure in this study was the total number of errors in each presentation condition (LVF, RVF, and BVF). This measure revealed that all of the participants showed a right visual field advantage (RVFA) that reflects specialization of the LH for this linguistic task. This advantage, the difference between performance levels in the LVF and in the RVF, was significantly greater in the Arabic speakers than in the other groups, as a result of poorer performance in the LVF in Arabic than in the other languages, while performance in the other presentation conditions (RVF, BVF) was equivalent among the groups. The third measure in this study was the difference between errors on the first letter and errors on the last letter of the trigram, a qualitative measure of sequential processing (Levy, Heller, Banich, and Burton, 1983; Eviatar, Hellige, and Zaidel, 1997). This qualitative measure revealed that Arabic and Hebrew speakers evinced a similar pattern that was different from the one shown by the English speakers. We interpreted this as indicating a different division of labor between the hemispheres while reading English or the Semitic languages, and attributed the difference to the demands made by concatenative versus nonconcatenative word morphology. We discuss this issue in more detail below. Of importance to us here is that it took longer for Arabic speakers to identify Arabic letters than it did for Hebrew speakers to identify Hebrew letters or English speakers to identify English letters. In addition, the large difference between performance levels in the two visual fields of Arabic speakers suggested that there are large differences in the abilities of the hemispheres in letter identification in Arabic, but not in the other languages. Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Morphology Arabic and Hebrew, as Semitic languages, are characterized by a nonconcatenative, highly productive derivational morphology (Berman, 1978). In Semitic these two languages words are constructed by combining a consonantal root (that carries most of the semantic information) and a word pattern that includes vowels as well as consonants, and provides information about the word class and its morphological status, as well as the complete unequivocal structure of the word. Hence, each word in Hebrew or Arabic is, at the very least bi-morphemic, but none of the composing morphemes are words by themselves. In most words, the core meaning is conveyed by the root, while the phonological pattern conveys word class information. For example, Hebrew, the word (TARSHEEM) consists of the root (R,SH,M) and the phonological pattern TA- -I- and the word (SIFRA) consists of the root (SFR) and the phonological pattern –I- -A in which every line represents a consonant. In Arabic the word (TAKREEM) consists of the root (KRM, whose semantic space includes things having to do with respect) and the phonological pattern TA- -I-. The combination results in the word “honor”. Unlike the Latin orthography in which vowels are represented by letters, in Arabic and Hebrew vowels are not part of the alphabet letters. The letters that make up the root may be dispersed across the word, interdigitated with letters that can double as vowels and other consonants that belong to the morphological pattern. Also, in Hebrew and Arabic, there are four letters which also specify long vowels, in addition to their role in

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

92

Raphiq Ibrahim

signifying specific consonants (in Hebrew: “alif , vav, yud”; in Arabic: “alif” or“imaala, waaw, yaa”). However, in some cases it is difficult for the reader to determine whether these dual-function letters represent a vowel or a consonant. When vowels do appear (in poetry, children's books and liturgical texts), they are signified by diacritical marks above, below or within the body of the word. The three Arabic diacritics are: a, i, u. Additional diacritical marking, the shadda, is used for lexical differentiation. Most of the grammatical functions at both the morphological and syntactic levels are represented by the short vowels, which also represent mood and case endings in the Verb-Subject-Object literary (fuṣḥa) syntax. From psycholinguistic view, inclusion of these marks specifies the phonological form of the orthographic string, making it completely transparent in terms of orthography/phonology relations. As the majority of written materials do not include the diacritical marks, a single printed word is often not only ambiguous between different lexical items (this ambiguity is normally solved by semantic and syntactic processes in text comprehension), but also does not specify the phonological form of the letter string. Thus in their unpointed form, Hebrew and Arabic orthographies contain a limited amount of vowel information and include a large number of homographs. For example, the bare unvowelized fuṣḥa form SH-R-B-T has five readings and five corresponding semantic interpretations: (a) sharebtu “I drank”; (b) sharebta “You (singular/masculine) drank”; (c) sharebti “You (singular/feminine) drank”; (d) sharebat “She drank”; and (e) shuribat “It (singular/feminine) was drunk.” Despite the similarity between Arabic and Hebrew, there are interesting differences. First, Arabic has special case of diglossia that is not exists in Hebrew. The state of affairs in Arabic is rare, since speakers of the language actually use two languages concurrently and intensively as a matter of course and not as an exception. In Arabic, the spoken form, which is ammia (local dialect) used by speakers of the language in a specified geographic area for daily verbal communication, is differentiated from the fuṣḥa (literary form), which is the language all speakers of Arabic, from all over the world, read and write in. Also, literary Arabic is universally used in the Arab world for formal communication and is known as written Arabic or Modern Standard Arabic (MSA) and Spoken Arabic appears partly or entirely in colloquial dialect and it is the language of day-to-day communication and has no written form. Hence, from the ecological point of view, SA and MSA could be considered as an instance of diglossia, that is, a social environment in which a community uses two forms of the same language concomitantly (Ferguson, 1959). Although sharing a limited subgroup of words, the two languages differ on several levels: phonetic, phonologic, morpho-syntactic and semantic.

1.3.2. Psycholinguistic Consideration Looking at the literature, much of the empirical evidence presenting different models of languages processing came from the Indo-Europian languages. For example, different models that were suggested to explain identification of printed words (by their whole-word visualorthographic or by phonological mediation process) were based on investigation in English language. English language have dominated experimental research because, it was implicitly assumed that reading processes (as well as the cognitive processes) are universal. Only in the past twenty years researchers tried to generalize these data to other languages like Chinese, Serbo-Cruatian, Hebrew and Arabic (see Frost, R., Katz, L., and Bentin, S. [1987]; Chen, H.C., and Ho, C. [1986]; Bentin, S., and Ibrahim, R. [1996]). Given that all models of word recognition for example are in principle possible, the focus of most contemporary studies has

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Psycholinguistic Challenges in Processing the Arabic Language

93

shifted from attempting to determine which of these models are better supported by empirical evidence, to understand how the different kinds of information provided by different orthographic systems in different languages interact during word recognition (Frost, 1994). For example, in Hebrew (and also in Arabic), letters represent mostly consonants, whereas vowels may be represented in print by a set of diacritical marks (points). These points are frequently not printed, and under these circumstances, isolated words are phonologically and semantically ambiguous. Nevertheless, it has been found that in both Hebrew (Bentin, Bargai, and Katz, 1984) and Arabic (Roman and Pavard, 1987) the addition of phonological disambiguating vowel points inhibits (rather than facilitates) lexical decision. On the basis of such results, it has been suggested that, at least in Hebrew, correct lexical decisions may be initiated on the basis of orthographic codes, before a particular phonological unit has been accessed (Bentin and Frost, 1987). These findings joins the list of studies mentioned previously that have all found evidence for cumbersome processing of the Arabic orthography and are congruent with the studies of McCusker, Hillinger, and Bias (1981). These researchers suggested that three factors influence the relative time course of orthographical and phonological code activation in word recognition: the participant’s reading ability, the task demands, and the complexity of the stimuli. In the case of Hebrew and Arabic, the third factor could be related to that the complexity of orthography and morphology that increases the perceptual load, thus affecting word identification in reading. In the case of Arabic, an important part of the literacy problem is posed by the Arabic orthographic system and its failure to support easy and efficient reading. Evidently, previous research on reading acquisition in the Arabic language has revealed that this process is slower even than in Hebrew (Azzam, 1989, 1993; Ibrahim and Eviatar, 2001). In skilled readers, it has been found that reaction times for visual recognition of Arabic words by Arabic speakers are longer than reaction times for Hebrew words by Hebrew speakers (Bentin and Ibrahim, 1996), English words by English speakers, and Serbo-Croatian words by Serbo-Croatians (Frost, Katz, and Bentin, 1987). When visual Arabic-word recognition was compared with visual Hebrew-word recognition in native Arabic speakers, Arabic words took longer to be recognized, although the Arabic words were recognized faster than the Hebrew words when the words were presented in the auditory modality (Ibrahim, 1998). Comparing reading processes in Arabic and French bilingual individuals, Roman and Pavard (1987) used oculomotor recording techniques to evaluate visual scanning strategies. They found that although mean reading time did not differ between Arabic and French texts (note that for conveying identical content, the number of words needed in Arabic is less than in French because Arabic morphology is nonconcatenative), gaze duration per word was significantly longer in Arabic (342 ms) than in French (215 ms). This phenomenon also has been found in comparisons of Hebrew and English text reading, in which the morphology of Hebrew is dense and similar to that of Arabic, and English morphology is concatenative and more similar to French (Shimron and Sivan, 1994). Comparisons of reading Arabic and French suggest that word-recognition processes may be slightly different in these two languages, possibly because of the additional morphologic complexity of Arabic relative to French (Courrieu and Do, 1987; Farid and Grainger, in press). Orthographic semantic view can also be found in studies of normal word recognition. Based on the cumulating data in Arabic we hypothesized: first that the complexity of Arabic orthography increases the perceptual load, thus slowing word identification in Arabic. Second, the diglossia in Arabic creates the possibility that to read in literary Arabic is to read

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

94

Raphiq Ibrahim

in second language. Indeed, and in accordance with our hypothesis, Ibrahim, Aharon-Peretz and Eviatar (2002), in their study on highly proficient Arabic-Hebrew bilinguals revealed that Arabic letters are processed slower than are Hebrew letters. Theoretically, some of the slowness in processing may stem from a conflict at the access to phonological codes (reflected by difficulties at the production stage) caused by the diglotic situation and the interference between spoken and literary Arabic (Bentin and Ibrahim, 1996; Ibrahim and Aharon-Peretz, 2005). Saiegh-Haddad (2003) investigated this factor directly by addressing aspects such as diglossic variables or linguistic distance parameters interfere with the acquisition of basic reading processes in Literary Arabic. The psycholinguistic and neurolinguistic aspects of this issue will be discussed in details after this informative review.

2. Neurolinguistic Basis of Processing Arabic In the last two decades, there has been a great concern to determine the neural basis of language processing in native (L1) and second (L2) language speakers beside a wide investigation in exploring the abstract representation of languages in their cognitive system (see Bookheimer, 2002 for a review). In this part I tried to elaborate this issue, taking advantage of a some unique features found in the Semitic languages (Hebrew and Arabic) in which they differ in various characteristics (orthography, phonology, morphology and syntax) from English as described before. Specifically, I tried to explore the interaction of these characteristics with hemispheric abilities.

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

2.1. Arabic Language and Hemispheric Effects Examination of the differences in orthography/phonology relations among the languages, together with the language experience of the participants, revealed that strategies of phonological encoding that are specific to an orthography seem to be used also while reading a second language (Eviatar, 1999), and that the processing of Arabic orthography seems to make different demands on the cognitive system both in beginning (Ibrahim, Aharon-Perez and Eviatar, 2002) and in skilled readers (Ibrahim, Eviatar, and Aharon-Perez, 2007). In a divided visual field study we (Eviatar, Ibrahim, and Ganayim, 2004) explored the locus of this difference and showed that adult native Arabic speakers who can read both Arabic and Hebrew are better at identifying letters in Hebrew than in Arabic, and that the main disadvantage for Arabic letters is in the left visual field when they are exposed to the right hemisphere (RH). We asked native Arabic speakers and native Hebrew speakers to perform a lateralized letter matching task in both Arabic and Hebrew, using a physical identity criterion (the Arabic speakers were literate in both languages, but the Hebrew speakers could not read Arabic, and thus performed the task as a pattern matching task). The pattern of results in response times was as we expected, revealing a RVFA in all of the conditions in which the participants could read (all except Hebrew speakers in Arabic, who showed a slight LVFA in this condition). The results of the accuracy measure were quite dramatic. In Hebrew, both groups revealed low error rates and equivalent performance in the two visual fields (both hemispheres are able to match letters quite well). In Arabic, the

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Psycholinguistic Challenges in Processing the Arabic Language

95

Hebrew speakers made many errors, with equivalent performance in the two visual fields. Arabic speakers revealed good performance in the RVF (their LH was able to match letters in Arabic as well as in Hebrew). In the LVF, however, Arabic speakers made as many errors as Hebrew speakers (who cannot read the language)! We hypothesized that the reason for this RH disadvantage in letter recognition is the complexity of grapheme-phoneme relations in Arabic. In order to examine our hypothesis that the RH cannot differentiate between very similar different letters in Arabic, whereas the LH can do so, we created a global-local task with two types of incongruent stimuli: one where the two letters on the two levels of the hierarchical stimulus were physically very different from each other: ‫ ت‬and ‫ ;م‬and another where the two letters were very similar to each other: ‫ ت‬and ‫ ب‬. Participants were required to attend to the local or the global levels of these hierarchical stimuli in different blocks. We measured the difference between congruent (where the same letter was used on both levels of the stimuli) versus incongruent conditions (where the letter in the global level was made out of small versions of the other letter). This difference indicated the degree of interference between the levels. In the first condition, where the letters were very different from each other, we replicated the results of other studies in other languages (e.g., Lamb, Robertson, and Knight, 1990). We found a global precedence (responses to the global level were always faster than to the local level (Navon 1977), and an asymmetry in the degree of interference between the two visual fields: stronger interference from the global level to the local level in the LVF, and the opposite pattern in the RVF. This pattern has been used to support the hypothesis that the LH is relatively more sensitive to the local aspects of visual stimuli and that the RH is relatively more sensitive to the global aspects of these stimuli. Most interestingly, however, in the second condition, where the two letters differed only in the number and placement of dots, there was no incongruence effect in the LVF at all, while congruent stimuli were faster than incongruent stimuli in the RVF in the local conditions. These results show that the RH cannot discriminate between letters that differ only in the placement or number of dots (e.g., /t/- ‫ ت‬and /b ‫ ب‬- / ), but that the LH can do so. Thus, in the case of Arabic, an important part of the literacy problem is posed by the Arabic orthographic system and its failure to support easy and efficient reading. Previous research on reading acquisition in the Arabic language has revealed that this process is slower than in Hebrew. We have shown that in both beginning and skilled readers, letter discrimination in Arabic is quite difficult. Concerning morphology, first I will examine the effects of word morphology in Arabic on the process of reading, and on the division of labor between the cerebral hemispheres in the early stages of visual word recognition. A number of psycholinguistic studies (Frost and Bentin, 1992; Frost, Katz and Bentin, 1997; Deutsch, Frost, and Forster, 1998; Berent, 2002) have explored the effect of the Hebrew morphology on lexical access and the structure of the mental lexicon. One of the conclusions from these studies is that the nonconcatenative and agglutinative morphological structure of Hebrew, together with the distributional properties of abstract word forms, results in the inclusion of subword morphological units in the mental lexicon of Hebrew speakers. Similarly, Prunet, Beland, and Idrissi (2000) report a case study of an Arabic-French agrammatic patient, who showed identical deficits in the two languages, except for a specific type of error, metathesis, in which he modified the order of the root consonants, with the vowel patterns remaining intact, only in Arabic, not in French. They interpret this finding as reflecting the manner in which words are stored in the mental lexicon

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

96

Raphiq Ibrahim

in the two languages: whole words plus affixes in French, and roots plus word patterns in Arabic. These findings converge with the conclusions of Eviatar (1999, Experiment 4) and Eviatar and Ibrahim (2004), who showed that nonsense CVC trigrams are processed sequentially in both visual fields in English, but in neither visual field in Hebrew and in Arabic, and hypothesized that this is because Hebrew and Arabic nonwords cannot be read sequentially. A similar conclusion for words was reached by Farid and Grainger (1996), who showed that initial fixation position in a word results in somewhat different response patterns in French and in Arabic. In French, fixation slightly to the left of the word's center results in best recognition for both prefixed and suffixed words, while in Arabic, prefixed words result in best recognition from leftward fixations and suffixed words result in best recognition from rightward initial fixations. They suggest that this is due to the greater importance of morphological structure in Arabic, because “...much of the phonological representation of the word can be recovered only after successfully matching the consonant cluster to a lexical representation” (p.364), that is, after extraction of the root. Berent (2002) has also concluded that in Hebrew, “Speakers decompose the root from the word pattern in on-line word identification…” (p. 335). Most recently we reported that the different manner in which words are constructed in English and in Hebrew and Arabic has an effect on the division of labor between the cerebral hemispheres in a lateralized lexical decision task (Eviatar and Ibrahim, 2007). We presented native speakers of Arabic, Hebrew, and English with morphologically simple and complex words and nonwords in their native language, and measured indexes of hemispheric integration. Morphological complexity was operationalized differently in English than in the Semitic languages. In English we defined monomorphemic words as morphologically simple, and derivations (e.g., farmer=farm+er) as morphologically complex. Morphologically complex nonwords were made up of legal morphemes in illegal combinations (e.g., logly). In Arabic and in Hebrew we defined a word as morphologically simple if the root+wordform structure was not transparent (e.g., the word is not easily divisible into these morphemes or the root is not generative, and appears only in that form), and as morphologically complex if it was easily and transparently divisible into these elements. Morphologically complex nonwords were created by inserting nonexistent roots into legal word forms. In English, we replicated the findings of previous studies: similarly to Iacoboni and Zaidel (1996), we showed that while the RH is able to independently recognize nonwords; it draws upon resources of the LH when encountering words. Similarly to Burgess and Skodis (1993) in English, and to Koenig, Wetzel, and Carramazza (1992) in French, we showed that for the English speakers, only the LH was sensitive to the morphological complexity of the stimuli. Morphological complexity affected words and nonwords in the same manner, with complex stimuli requiring longer latencies to be identified either as a word or as a nonword only in the RVF. As opposed to the English speakers, both groups of speakers of the Semitic languages showed bilateral sensitivity to morphological complexity. In addition, the Arabic and Hebrew readers showed higher values on our indexes of interhemispheric integration, suggesting more intensive hemispheric cooperation during the reading of Hebrew and Arabic than of English. Interestingly, in both languages, morphological complexity had opposing effects for words and for nonwords. Morphological complexity, or transparency of the root+wordform structure, facilitated the recognition of words and decelerated the rejection of nonwords.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Psycholinguistic Challenges in Processing the Arabic Language

97

We suggested that the nonconcatenative morphology of the Semitic languages, in which words are analyzed into their root and word-form constituents, requires that both hemispheres be sensitive to morphological structure. The automatic analysis of a character string into a recognizable word-form and a root resulted in faster recognition of complex words than of the simple words, which are not divisible in this way. This analysis also resulted in slower responses to complex nonwords than to simple nonwords, which did not contain a recognizable word form. Thus, the word form made complex nonwords more “wordlike”, requiring a more intensive search before they could be correctly rejected in the lexical decision task. In sum, these findings reveal the dynamic properties of the hemispheric relations, reflecting the flexibility of the system when it has to deal with different types of stimuli. The morphological structures of the Semitic languages make it necessary of the RH to be sensitive to morphology (either on its own or by using LH facilities via interhemispheric channels). This pattern in discernable in Hebrew, which the RH can read, but not in Arabic, in which it has a specific difficulty, or in English, which does not require morphological decomposition.

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Conclusions All of the findings of our group’s studies related to Arabic language, in addition to the findings of other researchers (see Abu-Rabia, 1997, 1998, 2000; Saiegh-Haddad, 2003), support the notion that Arabic has unique features that contribute to the inhibition and slowness of the reading process. Furthermore, I argue for inclusion of the neurofunctional perspective as a comprehensive basis for the discussion of teaching a second language (L2). After all, teachers deal every day with the ability of students to learn and they necessitate the knowledge of the structural relationship between languages and of relevant pedagogical methods to allow them to monitor the learning process, checking if it is optimally effective, and to intervene to shape it toward effectiveness. The findings do not allow us to ignore the fact that the normal Arab child (and to a further extent, the dyslexic child) who encounters special difficulties in reading acquisition needs special pedagogical methods and systematic professional intervention to overcome these difficulties that the Arabic language imposes. In that regard, although this chapter attempts to shed light on the dynamics of processing diglossic languages, it could be considered applicable to other bilingual populations, since verbal information mechanisms are universal.

References Abu-Rabia, S. (1997). Reading in Arabic orthography: The effect of vowels and context on reading accuracy of poor and skilled native Arabic readers. Reading and Writing: An Interdisciplinary Journal, 9, 65-78. Abu-Rabia, S. (1998). “Reading Arabic texts: Effects of text type, reader type and vowelization”. Reading and Writing: An Interdisciplinary Journal 10: 105-119. Abu-Rabia, S. (2000). Effects of exposure to literary Arabic on reading comprehension in a diglossic situation. Reading and Writing: An Interdisciplinary Journal 13: 147-157.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

98

Raphiq Ibrahim

Altarriba, J. (1992). The representation of translation equivalents in bilingual memory. In R. J. Harris (Ed.), Cognitive processing in bilinguals. (pp.157-174). Amesterdam: Elsevier. Azzam, R. (1989). Orthography and reading of the Arabic language. In P. G. Aaron and R. M. Joshi (Eds.), Reading and writing disorders in different orthographic systems (pp. 1–29). New York: Kluwer Academic. Azzam, R. 1990. The nature of Arabic reading and spelling errors of young children: a descriptive sudy. Unpublished doctoral dissertation, Teachers College, Columbia University, N.Y. USA. Azzam, R. (1993). The nature of Arabic reading and spelling errors of young children. Reading and Writing, 5, 355–385. Ayari, S. 1996. “ diglossia and illiteracy in the Arab world.” Language, Culture and Curriculum 9:243-252. Bentin, S., and Ibrahim, R. (1996). New evidence for phonological processing during visual word recognition: The case of Arabic. Journal of Experimental Psychology: Learning, Memory, and Cognition, 22, 309-323. Berman, R.A. (1978). Modern Hebrew structure. Tel Aviv, Israel: University Publishing. Bentin, S., and Ibrahim, R. (1996). New evidence for phonological processing during visual word recognition: The case of Arabic. Journal of Experimental Psychology: Learning, Memory, and Cognition, 22, 309-323. Ben Zeev, S. (1977). The influence of bilingualism on cognitive strategy and cognitive development. Child Development, 48, 1009 1018. Bentin, S., and Frost, R. (1987). Procsessing lexical ambiguity and visual word recognition in a deep orthography. Memory and Cognition, 15, 13-23. Bentin, S., Bargai, N., and Katz, L. (1984). Graphemic and phonemic coding for lexical access: Evidence from Hebrew. Journal of Experimental Psychology: Learning, Memory and Cognition, 10, 353-368. Berent, I. (2002). Identity avoidance in the Hebrew lexicon: implications for symbolic accounts of word formation. Brain and language, 81(1-3), 326-41. Bertelson, P., Morais, J., Alegria, J., and Content, A. (1985). Phonetic analysis capacity and learning to read. Nature, 313, 73–74. Bertelson, P., Morais, J., Alegria, J., and Content, A. (1985). Phonetic analysis capacity and learning to read. Nature, 313, 73–74. Bookheimer.S. (2002). “Functional MRI of Language: New Approaches to Understanding the Cortical Organization of Semantic Processing.” Annual Review of Neuroscience, 25:15188. Bialystok, E.. (1991). Language processing in bilingual children. Cambridge, England, UK: Cambridge University Press. Bialistok, E. and Ryan, E.B. (1985). Toward a definition of metalinguistic skill. MerrillPalmer Quarterly, 31, 229-251. Bradley, L. and Bryant, P.E. (1983). Categorizing sounds and learning to read: A causal connection. Nature, 301(5899),419 421. Campbell, R., and Sais, E., (1995). Accelerated metalinguistic (phonological) awareness in bilingual children. British Journal of Developmental Psychology, 13(1), 61 68. Cummins,J. (1987). Bilingualism, language proficiency, and metalinguistic development. In P. Homel and M. Palij, (Eds) Childhood bilingualism: Aspects of linguistic, cognitive, and social development. (pp. 57-73). Hillsdale, NJ, USA: Lawrence Erlbaum Associates.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Psycholinguistic Challenges in Processing the Arabic Language

99

Chen, H.-C., and Ho, C. (1986). Development of stroop interference in Chinese-English bilinguals. Journal of Experimental Psychology: Learning, Memory, and Cognition, 12, 397-401. Chen, H. C., and Leung, Y. S. (1989). Patterns of lexical processing in a nonnative language. Journal of Experimental Psychology: Learning, Memory and Cognition, 15, 316-325. Courrieu, P., and do, P.(1987). Perceptual analysis of word in Arabic. In J.K. O’Regan andA. Levy-Schoen (Eds.) Eye Movement: From Physiology to Cognition.(pp.451-458), NorthHolland: Elsevier. Dash, U. N. and Mishra, H. C. (1992). Bilingualism and metalinguistic development: Evidence from Kond tribal culture. Psychological Studies. 37(2 3), 81 87. Da Fontoura, H.L., and Siegel L.S. (1995). Reading, syntactic, and working memory skills of bilingual PortugueseEnglish Canadian children. Reading and Writing: An Interdisciplinary. De Groot, A. M. B., Dannenbur, J. G. and van aell (1994). Forward and backward word translation by bilinguals. Journal of Memory and Language, 33, 600-629. De Manrique, A. M. B., Gramigna, S. (1984). La segmentación fonológica y silábica en ni nos de preescolar y primer grado. Lectura y Vida, 5/1, 4–14. Deutsch, A., Frost, R. and Forster, K. I. (1998). Verbs and Nouns are Organized and Accessed Differently in the Mental Lexicon: Evidence from Hebrew. Journal of Experimental Psychology: Learning, Memory and Cognition, 24(5), 1238-1255. Doyle, A., Champagne, M., and Segalowitz, N. (1978).Some issues on the assessment of linguistic consequences of early bilingualism. In M. Paradis (Ed.) Aspects of bilingualism (pp. 13 20). Colombia, SC: Hornbeam Press. Eviatar, Z. (1999). Cross-language tests of hemispheric strategies in reading nonwords. Neuropsychology, 13 (4), 1-18. Eviatar, Z., Hellige, J. B., and Zaidel, E. (1997). Individual differences in hemispheric specialization: Effects of gender and handedness. Neuropsychology, 11, 562-576. Eviatar, Z. and Ibrahim, R. (2001). Bilingual is as bilingual does: Metalinguistic abilities of Arabic-speaking children. Applied Psycholinguistics, 21(4), 451-471. Eviatar. Z., and Ibrahim, R. (2007). Morphological structure and hemispheric functioning: The contribution of the right hemisphere to reading in different languages. Neuropsychology, 21(4), 470-484. Eviatar. Z., Leiken. M. and Ibrahim, R. (1999). Phonological processing of second language phonemes: A selective deficit in a bilingual aphasic. Language Learning, 49.1, 121-141. Edwards, D., and Christophersen, H. (1988). Bilingualism, literacy and meta linguistic awareness in preschool children. British Journal of Developmental Psychology, 6(3), 235 -244. Eviatar, Z., Ibrahim, R., and Ganayim, D. (2004). Orthography and the hemispheres: Visual and linguistic aspects of letter processing. Neuropsychology, 18(1), 174-184. Eviatar, Z. and Ibrahim, R. (2004). Morphological and Orthographic Effects on Hemispheric Processing of Nonwords: A Cross-Linguistic Comparison. Reading and Writing:17, 691705. Eid, M. (1990). Arabic linguistrics: The current scene. In M. Eid (Ed) Perspectives on Arabic linguistics I. (pp 3 37). Amsterdam: John Benjamins Publishing Company. Farid, M., and Grainger, J., (1996). How initial fixation position influences visual word recognition: A comparison of French and Arabic. Brain and Language, 53, 351-361.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

100

Raphiq Ibrahim

French. In J. K. O’Regan and A. Levy-Schoen (Eds.), Eye movement: From physiology to cognition (pp. 431-440). Amsterdam, The Netherlands: North Holland Elsevier. Frost, R. (1994). Prelexical and postlexical strategies in reading: Evidence from a deep and a shallow orthography. Journal of Experimental Psychology: Learning,Memory, and Cognition, 20, 1-16. Frost, R., Bentin, S. (1992). Processing phonological and semantic ambiguity: Evidence from semantic priming at different SOAs. Journal of Experimental Psychology: Learning Memory and Cognition, 18 (1), 58-68. Frost, R., Katz, L., and Bentin, S. (1987). Strategies for visual word recognitionand orthographical depth: A multilingual comparsion. Journal of Experimental Psychology: Human Perception and Performance, 13, 104-114. Galambos, S. J., and Hakuta, K. (1988). Subject specific and task specific characteristics of metalinguistic awareness in bilingual children. Applied Psycholinguistics, 9(2), 141 162. Galambos, S. J., and Goldin Meadow, S. (1990). The effects of learning two languages on levels of metalinguistic awareness. Cognition, 34(1), 1 56. Gerard, L. D., and Scarbourough, D. L. (1989). Language specific lexical access of homographs by bilinguals. Journal of Experimental Psychology: Learning, Memory and Cognition, 15.2, 305-315. Grainger, J. and T. Dijkstra. (1992) On the Representation and Use of Language Information in Bilinguals." In Cognitive Processing in Bilinguals (ed.) R. J. Harris, 207-220. Amsterdam: North Holland. Groot, de. M. B. A., (1992). Determinants of word translation. Journal of Experimental Psychology: Learning, Memory, and Cognition, 18, 1000-1018. Groot, de. M. B. A., Dannenburg, L,. and Van Hell, J. G. (Submitted). Forward and backward word translation by bilinguals. Journal of Memory and Language. Groot, de. M. B. A. and Nas, L. J. G. (1991). Lexical representation of cognates and noncognates in compound bilinguals. Journal of Memory and Language, 30, 90- 123. Groot de, A.M.B. (1995). Determinants of bilingual lexicosemantic organization. Computer Assisted Language Learning, 8, 151-180. Grainger, J., and Frenck-Mestre, C. (1998). Masked priming by translation equivalents in proficient bilinguals. Language and Cognitive Processes, 13, 901-623. Iacoboni, M., and Zaidel, E. (1996). Hemispheric independence in word recognition: Evidence from unilateral and bilateral presentations. Brain and Language, 53, 121-140. Ibrahim, R., Eviatar, Z., and Aharon Peretz, J. (2002). The characteristics of the Arabic orthography slow it’s cognitive processing. Neuropsycholgy, 16(3), 322-326. Ibrahim, R.and Aharon-Peretz, J. (2002). Text processing and perceptual characteristics of Arabic orthograpy. Neurology update, 9, 22-24. (Hebrew). Ibrahim, R. and Aharon-Peretz, J. (2005). Is literary Arabic a second language for native Arab speakers?: Evidence from a semantic priming study. The Journal of Psycholinguistic Research, 34(1),51-70. Ibrahim, R., Eviatar. Z., and Aharon Peretz, J. (2007). Metalinguistic awareness and reading performance: A cross language comparison. The Journal of Psycholinguistic Research, 36(4), 297-317. Ibrahim, R (submitted) Is a literate Arabic speaker bilingual? New evidence from a repetition priming study.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Psycholinguistic Challenges in Processing the Arabic Language

101

Ibrahim, R. (2006). Morpho-Phonemic similarity within and between languages: A factor to be considered in processing Arabic and Hebre. Reading and Writing, 19, (6), 563-586. International Bureau of Education (IBE)—UNICEF statistics for Literacy UNESCO–Beirut in 2001. Keatley, C., and DeGelder, B. (1992). The bilingual primed lexical decision task: cross language priming disappears with speeded responses. European Journal of Cognitive Psychology, 4(4), 273-292. Keatley, C., Spinks, J., and DeGelder, B. (1990). Asymmetrical semantic facilitation between language: Evidence for separate representational system in bilingual memory Unpublished Manuscript, University of Tilberg, the Netherlands. Kirsner, K. (1986). “Lexical representation: is a bilingual account necessary”. Language Processing in Bilinguals: Psycholinguistic and Neuropsychological Perspectives ed. by J. Vaid, 21-45. Hillsdale, NJ: Lawrence Erlbaum. Koenig, O., Wetzel, C., and Caramazza, A. (1992). Evidence for different types of lexical representations in the cerebral hemispheres. Cognitive Neuropsychology, 9, 33–45. Kroll, J. F.,Sholl, A., Altarriba, J., Luppino, C., Moynihan, L., and Sanders, C. (1992, November). Cross-language semantic priming: Evidence for independent lexical and conceptual contributions. Paper presented at the Thirty-Third Annual Meeting of the Psychonomic Society , St. Louis, MO. Kroll, J. F., and de Groot, A. M. B. (1997). Lexical And Conceptual Memory in the Bilingual: Mapping Form to Meaning in Two Languages. In de Groot, A. M. B. and Kroll, J. F. (Eds.), Tutorials In Bilingualism: Psycholinguistic Perspectives. (pp. 201-224). Mahwah, NJ: Lawrence Erlbaum Publishers. Kroll, J. F., and Stewart, E. (1994). Category interference in translation and picture naming: Evidence for asymmetric connections between bilingual memory representations. Journal of Memory and Language. 33, 149-174. Lamb, M.R., Robertson, L.C., and Knight, R.T., (1990). Component mechanisms underlying the processing of hierarchically organized patterns: Inferences from patients with unilateral cortical lesions. Journal of Experimental Psychology: Learning, Memory, and Cognition, 16, 471-483. Levy, J., Heller, W., Banich, M.T., and Burton, L. (1983). Are variations among right handed individuals in perceptual asymmetries caused by characteristic arousal differences between hemispheres? Journal of Experimental Psychology: Human Perception and Performance, 9, 329-359. Navon, D. (1977). Forest before trees: The precedence of global features in visual perception. Cognitive Psychology, 9, 353-383. Maamouri, M. (1998). Language education and human development: Arabic diglossia and its impact on the quality of education in the Arab region”. The Mediterranean Development Forum. Washington, DC: The World Bank. McClelland, J, L., Rumelhart,D, E., and Hinton, G, E. (1986). The appeal of parallel distributed processing. In D. E. Rumelhart and J.L. McClelland (Eds.) parallel distributed processing. Cambridge, Mass .: M.I.T. press. Meitzav - national testing program of 5th Grade. (2003). Ministry of Education, Jerusalem, October.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

102

Raphiq Ibrahim

Moreno, E. M., Federmeier, K. D., and Kutas, M. (2002). Switching languages, switching palabras (words): An electrophysiological study of code-switching. Brain and Language, 80, 188-207. Paradis, M, (1998). Aphasia in bilinguals: How atypical is it? In P. Coppens and Y. Lebrun (Eds), Aphasia in atypical populations. (pp. 35 66). Mahwah, NJ: Lawrence Erlbaum Associates. PISA- Program for International Student Assessment(2006). report of worldwide scholastic achievements. Prunet, J., Beland, R., and Idrissi, A. (2000). The mental representation of Semitic words. Linguistic Inquiry 31, 4, p. 609-648. Roman, G., and Pavard, B. (1987). A comparative study: How we read Arabic and French. In J. K. O’Regan and A. Levy-Schoen (Eds.), Eye movement: From physiology to cognition (pp. 431-440). Amsterdam, The Netherlands: North Holland Elsevier. Reitan, R. M., and wolfson, D. (1993). The Halsted reitan neuropsychological test battery: Theory and clinical interpretation (2nded.). Tucson, AZ: Neuropsychology Press. Saiegh- haddad, Elinor. 2003. “Linguistic distance and initial reading acquisition: The case of Arabic. Diglossia” Applied Psycholinguistics 24: 431-451. Schwanenflugel, P. J., and Rey, M. (1986). Interlingual semantic facilitation: Evidence for a common representational system in the bilingual lexicon. Journal of Memory and Language, 25, 605-618. Shimron, J. and Sivan, T. (1994). Reading profeciency and orthography: Evidence from Hebrew and English. Language learnining. 44: 5-27. Titone, R. (1994). Bilingual education and the development of metalingusitic abilities: A research project. International Journal of Psycholinguistics 10(1), 5 14. Tzelgov, J., Henik, A., and Leiser, D. (1990). Controlling Stroop interference: Evidence from bilingual task. Journal of Experimental Psychology: Learning, Memory and Cognition, 16, 760-771. Tzelgov J.,Henik,A.,Sneg,R.,and baruch, O.(1996 ) Unintentional Word Reading Via the Phonological Route: The Stroop Effect With Cross-Script Homophones. Journal of Experimental Psychology: Learning, Memory, and Cognition, 22, 2, 336-349. Wagner, Daniel. 2004. “Literacy(ies), culture(s), and development(s): The ethnographic challenge”. Reading Research Quaterly 39: 234-241. Wagner, Daniel. A. 1993 Literacy, Culture, and Development: Becoming literate in Morocco USA: Cambridge University Press.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

In: Bilinguals: Cognition, Education and Language Processing ISBN: 978-1-60741-710-1 Editor: Earl F. Caldwell, pp. 103-122 © 2010 Nova Science Publishers, Inc.

Chapter 5

NEUROCOGNITIVE ASPECTS OF PROCESSING ARABIC AND HEBREW Raphiq Ibrahim Learning Disabilities Department, and The Edmond J. Safra Brain Research Center for the Study of Learning Disabilities, University of Haifa, Haifa, Israel

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Abstract The aim of this chapter is to explore the neurocognitive basis of the difficulties that the Arabic-Hebrew bilingual encounters in processing the Arabic language as a result of the diglossic situation in Arabic (spoken Arabic and Modern Standard, or Literary Arabic). Furthermore, the chapter discusses the unique features of the Arabic language that might contribute to the inhibition and slowness of reading acquisition and might even hinder the acquisition of basic academic skills. In the first section, two case studies of Arabic-Hebrew aphasic patients (M.H. and M.M.) are presented, with different disturbances in the two languages, Arabic (L1) and Hebrew (L2). They exhibited a complementary pattern of severe impairment of either L1 (Arabic) or L2 (Hebrew) constituting a double dissociation. These results suggested that the principles governing the organization of lexical representations in the brain are not similar for the two languages. The second section focuses on the functional architecture of reading in Hebrew and in Arabic. The effects of characteristics of Arabic and Hebrew as Semitic languages on hemispheric functioning were systematically examined. These patterns are compared with the modal findings in the literature, which are usually based on English. Also, the effects of the absence of almost all vowel information, the orthographies of the two languages, and their non-concatenative morphological structure were investigated. It was shown that when languages make different types of demands upon the cognitive system, interhemispheric interaction is dynamic and is suited to these demands. In that regard, both Arabic and Hebrew require a higher level of interhemispheric interaction than does English.

1. The Neural Basis of Bilingualism The question of how multiple languages are represented in the brain remains unresolved. On one hand, data have shown that there is one neural representation of multiple languages

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

104

Raphiq Ibrahim

(Moretti, Bava, Torre, Antonello, Zorzon, and Zivandinov, 2001; Paradis, 1990), called “the linguistic domain” approach. On the other hand, data have indicated that bilingual persons could have distinct cortical language areas (Ojemann, Ojemann and Lettich, 1989; Dehaene, Dupoux, and Mehler 1997) and called by some researchers the ‘‘language-membership principle’’ approach. According to this approach, first language (L1) and second language (L2) representations would be, to some extent, sustained by different brain areas, since they take different language membership values. Cases of selective aphasia and other results from neuro-imaging techniques demonstrate a dissociation between multiple language representations in the cognitive system of the brain (Dehaene et al., 1997). In their study of French-English bilinguals using the fMRI technique, Dehaene and his colleagues (1997) found a dissociation between cortical areas involved in the French (L1) and English (L2) languages. The regions of the left superior temporal sulcus, superior and middle temporal gyri showed consistent activation across subjects during presentation of L1. According to this hypothesis, a bilingual aphasic is found to recover selectively in one language while the other is lost (see Green, 2005; Green and Price, 2001). The classical model assigns language functions to two regions in the left hemisphere: the inferior frontal region and the temporo-parietal region of the brain. Injuries in the general boundaries of these cortical areas have resulted in clinically and linguistically different aphasic syndromes, referred to as Broca’s aphasia (agrammatic) and Wernicke’s aphasia (paragrammatic). In recent years, most localization theory of language lost its categorical power. Some researchers provided evidence for people whose language’s components (like semantics) are localized in the right hemisphere (Fabbro, 2001; Fabbro, Pesenti, Facoetti, Bonanomi, Libera, and Lorusso, 2001). Additional evidence has come from epileptic bilingual patients who performed a picture-naming task in L1 and L2 while having different brain areas stimulated (e.g., Lucas, McKhann, and Ojemann, 2004). Lucas and his colleagues suggested that both overlapping and distinct brain regions are involved in the representations of multiple languages of a bilingual. Furthermore, some studies of the activity of the brain have revealed that the same brain regions are responsible for the representation of multiple languages of a bilingual regardless of the grade of similarity between these languages (Klein, Zatorre, Milner, and Meyer, 1994). In this view, the organization of the lexical representations of the two languages (L1 and L2) would be governed by variables such as grammatical class and semantic category regardless of language membership. Klein, Milner, Zatorre, Zhao, and Nikelski (1999) compared cerebral organization of two typologically distant languages: English and Mandarin Chinese. The subjects, proficient in both languages, had learned their L2s during adolescence. Mandarin was chosen as it differs from English in its specified use of pitch and tone. The study examined the influence of linguistic structure on cerebral blood flow (CBF) patterns in subjects when they performed a noun-verb generation task. The task conditions consisted of repeating nouns in Mandarin, repeating nouns in English, generating a verb for a noun in Mandarin, and generating a verb for a noun in English. Overall, the pattern of CBF increase seen in response to the L1 was strikingly similar to that seen for the L2. This finding led to the conclusion that in fluent bilinguals who use both languages in daily life lexical search utilizes common cortical areas. More recently, based on the results of an eventrelated functional magnetic resonance imaging (ER-fMRI), there is a shared neural mechanism for the processing of native and second languages (Pu et al., 2001). Moreover, Illes and his colleagues (1999) examined brain activation in bilingual participants who sequentially learned English and Spanish (or vice versa). The participants became fluent in

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Neurocognitive Aspects of Processing Arabic and Hebrew

105

their L2s a decade after L1 acquisition but were proficient in both languages. Subjects were presented with 480 concrete and abstract English nouns and their Spanish translations. Participants performed tasks that required semantic and non-semantic decisions about those words. The semantic activation for both languages occurred in the same cortical locations. Further, no activation difference was observed in a direct comparison of semantic judgments in English and Spanish. The researchers suggested that, according to the resolution provided by functional MRI, a common neural system mediates semantic processes for the two languages in the bilingual brain. They concluded that learning a new language after puberty does not require the addition of a new semantic processing system or the recruitment of new cortical regions. A third hypothesis claimed that both “language membership“ and “linguistic domain” would affect the way L2 information is represented in the brain. Concerning the issue of cognitive representation of the two languages in bilinguals, several psycholinguistic models have been proposed. Current models of lexical access in bilingual speakers typically assume that the semantic system is shared by the two languages of a bilingual (De Bot, 1992; Green, 1986; 1998; Kroll and Stewart, 1994). In other words, each semantic/conceptual representation is connected to its corresponding lexical nodes in the two languages. Although, some researchers (e.g., Van Hell, and De Groot, 1998) have claimed that conceptual representations are language dependent, recent proposals widely favor the idea that, at least for common words, bilingual subjects have a unique conceptual store shared by both languages. If the semantic system is shared by the two languages of a bilingual, the question that arose was whether there is a spreading activation between the semantic system and the lexical system regardless of the language programmed for response. Another question that researchers tried to answer was whether the activation of the semantic system spread to the two languages of a bilingual. Some researchers claimed that there is parallel activation of the two languages of a bilingual regardless of the language chosen for production (De Bot, 1992; Green, 1986). In other words, current models follow the general spreading activation principle and assume that there is parallel activation of the two lexicons of a bilingual. Levelt (1989) assumed that concepts are represented as indivisible nodes and the nodes corresponding to a concept are linked to the nodes of semantically related concepts. For example, the activation of the conceptual node corresponding to the picture (e.g., bird) “spreads” some activation to other semantic representations that are associated with it (such as tree, plane). Other models (Caramazza, 1997; Dell, 1986) assumed that concepts (e.g., canary) are represented as a bundle of semantic features (bird, can fly, two legs) and the activation of a given concept (e.g., bird) would activate part of the semantic representation of other related concepts (e.g., penguin) because some of their semantic features are shared. Regardless of the specific mechanisms, these two proposals share the assumption that in the course of naming a picture, several semantic representations are activated to some degree. This is either because semantic representations are interconnected or because they share several semantic features. In that regard, psycholinguistic models assumed that words of each language are represented separately at the lexical level and connected indirectly via a common semantic system, which is accessed independently from each lexicon (Chen and Leung, 1989; Schwanenfeugel and Ray, 1986). In neuro-linguistic terms, these models suggest that separate but overlapping regions are involved in the processing of more than one language.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

106

Raphiq Ibrahim

Cases of bilingual aphasia afford an excellent opportunity to study language processes. The pattern of aphasia following brain injury to a bilingual is very diverse and therefore results obtained should be wearily approached. Previous studies showed that, a brain lesion could selectively disrupt one language but not the other (Ojemann, 1983) and bilingual persons could have distinct cortical language areas (Dehaene et al., 1997). On the other hand, there are numerous reports of aphasia simultaneously affecting both of a bilingual patient’s languages following lesions of the left hemisphere suggesting that, both overlapping and distinct brain regions are involved in the representations of multiple languages (Fabbro, 2001). A further complication to the resolution of this issue comes from the fact that the cortical organization of L2 in relation to L1 seems to depend on various factors such as level of proficiency, age of acquisition and exposure (e.g., Kim et al., 1997). This conflicting data can be resolved with case studies selected bilingual aphasic patients indicating dissociation and a double dissociation between first language (L2) and second language (L1). However, researchers must keeps in mind that lesions in the brain are often widespread. In the first section of this chapter, I report the performance of two Arab-Hebrew bilinguals who had suffered a brain lesions tumor and were undergone a surgery. As a result of these brain lesions, the linguistic abilities were impaired and exhibited different symptomologies in the both languages. In one case, the patient (M.H.) is an Arabic-Hebrew bilingual who suffered from a lesion to the brain following a brain tumor and evinced dissociation between his ability to perceive and produce his second language (Hebrew). In the second case, another patient (M.M.) who suffered a brain damage following hemorrhage evinced complementary picture with a double dissociation of the M.H. case report.

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

1.1. Arabic and Hebrew: Background and Characteristics As Semitic languages, words in Arabic and Hebrew have similar morphological structures. Regardless if these words based on inflectional or Derivational forms, the morpheme-based lexicon of these families implies the existence of roots and templates (word patterns?). Roots are recognized as autonomous morphemes expressing the basic meaning of the word. Roots are abstract entities that are separated by vowels adding morphological information (e.g., in Arabic, the perfective /a-a/ in daraba ‘hit’, or the passive /u-i/ in duriba “was hit” and in Hebrew , the perfective /a-a/ in lakah “took”, or the passive /ni-a/ in nilkah “was taken”). Other researchers defined both Semitic languages as non-concatenative, highly productive derivational morphology (Berman, 1978). According to this approach, most words are derived by embedding a root (generally trilateral) into a morpho-phonological word pattern when various derivatives are formed by the addition of affixes and vowels. Also, in Arabic and Hebrew, there are four letters which also specify long vowels, in addition to their role in signifying specific consonants (in Arabic there are only three – a, u, y ‫)ا و ي‬. However, in some cases it is difficult for the reader to determine whether these dual-function letters represent a vowel or a consonant. When vowels do appear (in poetry, children’s books and liturgical texts), they are signified by diacritical marks above, below or within the body of the word. Inclusion of these marks specifies the phonological form of the orthographic string, making it completely transparent in terms of orthography/phonology relations.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Neurocognitive Aspects of Processing Arabic and Hebrew

107

In regard to semantics, the core meaning is conveyed by the root, while the phonological pattern conveys word class information. For example, in Arabic the word (TAKREEM) consists of the root (KRM, whose semantic space includes things having to do with respect) and the phonological pattern TA—I. The combination results in the word ‘honor’. In Hebrew, the word (SIFRA) consists of the root (SFR- whose semantic space includes things having to do with counting) and the phonological pattern –I—A, which tends to occur in words denoting singular feminine nouns, resulting in the word “numeral”. As the majority of written materials do not include the diacritical marks, a single printed word is often not only ambiguous between different lexical items (this ambiguity is normally solved by semantic and syntactic processes in text comprehension), but also does not specify the phonological form of the letter string. Thus, in their unpointed form, Hebrew and Arabic orthographies contain a limited amount of vowel information and include a large number of homographs. Comparing to Hebrew, Arabic includes much larger number of homographs thus, it is much more complicated than Hebrew. Despite the similarity between these languages, there are major differences between Arabic and Hebrew. First, Arabic has special case of diglossia that does not exist in Hebrew. Literary Arabic is universally used in the Arab world for formal communication and is known as “written Arabic”, also called “Modern Standard Arabic” (MSA), and spoken Arabic appears partly or entirely in colloquial dialect and is the language of everyday communication and has no written form. Although sharing a limited subgroup of words, the two forms of Arabic are phonologically, morphologically, and syntactically different. This added complexity is found in several characteristics that occur in both orthographies, but to a much larger extent in Arabic than in Hebrew. The later has to do with orthography that includes letters, diacritics and dots. In the two orthographies some letters are represented by different shapes, depending on their placement in the word. Again, this is much less extensive in Hebrew than in Arabic. In Hebrew there are five letters that change shape when they are word final: ( ‫ך‬-‫ כ‬,‫ם‬-‫מ‬, ‫ן‬-‫נ‬,‫ץ‬-‫צ‬,‫ף‬-‫) פ‬. In Arabic, 22 of the 28 letters in the alphabet have four shapes each (for example, the phoneme /h/ is represented as: ‫ ـﻬـ‬، ‫ ـ ﻪ‬، ‫ هـ‬، ‫)ﻩ‬. Thus, the graphemephoneme relations are quite complex in Arabic, with similar graphemes representing quite different phonemes, and different graphemes representing the same phoneme. Concerning dots in Hebrew, they occur only as diacritics to mark vowels and as a stress-marking device (dagesh). In the case of three letters, this stress-marking device (which does not appear in unvowelized scripts) changes the phonemic representation of the letters from fricatives (v, x, f) to stops (b, k, p for the letters ‫ פ ק ב‬respectively). In the unvowelized form of the script, these letters can be disambiguated by their place in the word, as only word or syllable initial placement indicates the stop consonant. In Arabic, the use of dots is more extensive: many letters have a similar or even identical structure and are distinguished only on the basis of the existence, location and number of dots (e.g., the Arabic letters representing /t/ and /n ،‫ ن‬, ‫) ت‬ become the graphemes representing /th/ and /b/ (‫ ث‬, ‫ )ب‬by adding or changing the number or location of dots. Many studies have demonstrated Bilinguals do not recognize written words exactly the same as monolinguals. For example, it was proven that visual word identification in L2 is affected by the native language of the reader (e.g., Wang, Koda, and Perfetti, 2003). However, the opposite is true as well: knowledge of L2 may have impact on the identification of printed L1 words was published by Bijeljac-Babic, Biardeau, and Grainger (1997). In comparative studies of different languages, there are two points of comparison: The speech

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

108

Raphiq Ibrahim

and the writing system. Thus, in comparing Arabic and Hebrew reading, we compare examples of two related language families (semetic languages) that are similar in their morphological structure but radically differ in their orthographic and phonetic systems. Recent studies on Hebrew (Frost, Deutsch and Forster, 1997) and Arabic (Mahadin, 1982.) support the assumption that roots can be accessed as independent morphological units. Also, in the area of speech perception, differences in the phonetic perception of L1 and L2 was found between native (for reviews see Flege, 1992) and nonnative speakers (Eviatar, Leikin, and Ibrahim, 1999).

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

1.2. Case Report: M.H. M.H. was a 41-year-old right-handed male native speaker of Arabic (Ibrahim, 2008). He was born in Israel, acquired the Hebrew language at fourth grade and his Hebrew competence was very high (a high school teacher). He was brought to the hospital with sudden onset of fever and confusion and diagnosed as suffering from herpes simplex virus. On the fifth day in the hospital he exhibited a sudden high-grade headache, vomiting and disturbance of consciousness. Radiological findings showed acute, massive intracranial hemorrhage in the left temporal lobe, compressing the central line of the brain contra-laterally and moderate hemorrhage and encephalomyelitis in left temporal lobe and right frontal subdural hemorrhage. After surgery he was sent to rehabilitation and for two months he was hospitalized. During this period he developed an acute onset of a neurological deficit, epileptic status with left temporal focus and exhibited amnestic aphasia and his spontaneous language production was non-fluent, with grammatical disruptions and common anomic states. M.H. was administered neuropsychological tests after the rehabilitation period and dissociation between the performance in the two languages was obtained. In Arabic M.H. exhibited almost fluent speech with exhibited word-finding pauses and paraphasic errors with limited disturbances in auditory comprehension. In contrast, more disturbances appeared in Hebrew, which constitute a second language. In the written language, M.H. has encountered problems in reading and writing compared to Arabic. M.H. exhibited non-fluent speech with anomia, disturbances in auditory comprehension without difficulties in repetition. The results revealed different patterns emerged in both languages, though they were more severe in Hebrew. In addition, some preserved abilities were observed in single-word reading and some writing to dictation in Hebrew. M.H. received intensive language therapy in Arabic and in Hebrew for three months and showed significant improvement in both languages, more in Arabic. The improvement in Arabic was in all linguistic abilities but in Hebrew a mild improvement was noticed in his spontaneous speech and auditory comprehension, whereas naming ability remained without changes. His speech in Arabic is fluent and grammatically correct but with occasional paraphasias and prominent word-finding difficulties. His reading and writing abilities improved significantly only in Arabic. M.H.’s naming abilities were impaired in all modalities and in all types of naming tasks. However, these deficits were not equivalent in the two languages, where Arabic was more productive. Phonemic priming was effective and M.H.’s performance improved if he received more than one syllable. With treatment, a significant improvement of auditory comprehension (including single-word comprehension) was gradually appeared. To rule out symptoms due to right frontal hemorrhage, tasks assessing visuospatial and frontal difficulties were conducted. The patient

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Neurocognitive Aspects of Processing Arabic and Hebrew

109

demonstrated good visual ability. M.H. demonstrated good copying and construction abilities and the non-verbal abstraction was near to his age norms, consistent with intact visual perception and reasoning skills. In regard to phonologica abilities, M.H.’s performance was dependent on word length, with better performance on short words (three to five letters). Both Arabic and Hebrew are languages with deep orthography where there is no one-to-one correspondence between letters and sounds because most Arabic and Hebrew vowels are not instantiated as letters. This is probably reflected in his relatively similar performance in both languages. In reading and writing, M.H.’s read aloud single and short words using a direct visual strategy but in some cases this strategy was not successful and he turned to letter-byletter reading, resulting in literal. His strategy for reading in Hebrew was similar, but resulted in poor performance. Spontaneous writing (in Arabic) was in good level single words and word combinations without literal paragraphias. In Hebrew, writing to dictation was possible only at the level of words with literal paragraphias (for example, the word mapa, “map,” was written as maba, which does not constitute a word. Overall, the patient M.H. displayed somewhat different symptomatologies in his two languages. The results of the standard examination showed that M.H. suffered from different language impairment in Arabic and Hebrew, with a significantly more prominent disorder in Hebrew. Moreover, he displayed different progress in both languages in consequence of language therapy, though progress in Arabic was greater. This clinical picture is of interest because Arabic is structurally not very distant from Hebrew (especially in terms of morphology and syntax). It is important to remind that, although Arabic is the native language, the prior level of language competence in the two languages was almost equivalent.

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

1.3. Case Report: M.M. This case report verified the existence of a double first language (L1)/second language (L2) dissociation (Ibrahim, 2008). An Arabic-Hebrew bilingual (M.M.), with similar cultural background which who acquired Hebrew at age eight and considered to be balanced ArabicHebrew bilingual (A retired Israeli army soldier) suffered from brain damage following a left hemisphere tumor (oligodendroglioma) and craniotomy. M.M. underwent surgery and a left frontal craniotomy was carried out. After surgery, the patient was sent to a rehabilitation period and was hospitalized for two months. During this period he developed epilepsy and was treated with anti-convulsions drugs. In addition, because of the motor Aphasia, he was undergone speech therapy for a long period. After his recovery, the same material were used and a series of linguistic tasks taken from Western Aphasia Battery (WAB; Kertesz, 1982) and the Boston Naming Test (BNT; Kaplan, Goodglass, and Weintraub, 1983) was administered in Arabic and Hebrew to evaluate M.M.’s efficiency of different components of his linguistic processing system including: fluency, repetition, naming, spelling, category and letter generation and other visuospatial tasks. The results revealed dissociation between the performance in the two languages was obtained. In Hebrew, M.M. exhibited mild disturbances. The speech is grammatically correct but with occasional literal and semantic paraphasias and slight word-finding difficulties without disturbances in auditory comprehension and without difficulties in repetition. In contrast, more disturbances were appeared in Arabic, his native language. As mentioned before, M.M. exhibited non fluent speech in Arabic with prominent word finding difficulties,

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

110

Raphiq Ibrahim

disturbances in auditory comprehension and with mild difficulties in repetition. In naming, literal and semantic paragraphias were exhibited in Arabic (for example, in the literal paraphasia, the word noor (“flower”) was replaced by nowara, which is not word, and in semantic paraphasia in Hebrew, for example, the word mihoga (“lead compasses”) was replaced by igol, which means “circle”. In written language, M.M. encountered problems in reading and writing more in Arabic. Different patterns emerged in both languages, though they were more severe in Arabic, the native language. However, in Arabic, some preserved abilities were observed in single-word reading and some writing to dictation. M.M.’s most evident initial as well as residual aphasic symptom was a marked difficulty in confrontation naming in both languages. Initially (at least two years after surgery), M.M. demonstrated an almost typical pattern of severe motor aphasia in both languages (Benson, 1979; Luria, 1975). M.M.’s naming abilities were impaired in all modalities and in all types of naming tasks. However, these deficits were not equivalent in the two languages, where Hebrew was more productive. In regard to visual abilities M.M. demonstrated good visual ability. However, he demonstrated moderate copying difficulties and construction abilities (Clock drawing) and his score was consistent with his intact visual perception. The non verbal abstraction on the Wisconsin Sorting Cards (WCST) was below his age norms (reached on category) and he exhibited preservative reactions leading to disorders in reasoning skills (Milner, 1963). In regard to Phonological abilities, M.M. was presented with three auditory tasks following Luria (1970): (a) counting the number of letters in individual words (i.e., saying how many letters there are in a spoken word), (b) counting the numbers of syllables in an individual word, and (c) synthesizing words from individually pronounced letters (i.e., recognizing an auditorally spelled word). The performance of M.M. was dependent on word length, with better performance on short words (three to five letters). Concerning Reading and Writing, reading aloud in Arabic revealed a letter-by-letter strategy and exhibited poor performance in this language compared to Hebrew. The spontaneous writing in Hebrew was in better level than Arabic in all types of words (single words and word combinations). In Arabic, writing to dictation was possible only at the level of single words. The whole results of the standard examination showed that M.M. suffered from different language impairment in Arabic and Hebrew, with a significantly more prominent disorder in Hebrew. The initial diagnosis was that M.M. suffering from amnestic aphasia. During the period of the language treatment, M.M. was administered various tests to investigate further the nature of his impairments in the two languages. Moreover, he displayed parallel progress in both languages in consequence of language therapy, though progress in Arabic was greater. This clinical picture is of interest because Arabic is structurally not very distant from Hebrew (especially in terms of morphology and syntax). It is important to remind that, although Arabic is the native language, the prior level of language competence in the two languages was almost equivalent. The pattern of the results is complementary to the recent case study of M.H. (Ibrahim, submitted) that exhibited dissociation between languages. Given that M.M. residual left brain damage, evinced more deficits in L1 perception and production than L2 (which differs in from native language), and given that in recent case report, M.H. provided a dissociation between processing L1 and L2, the data support the position that distinct brain regions are involved in the representations of multiple languages of a bilingual. This support the conclusion that patient with a more prominent L1 impairment usually have lesion centered on the left hemisphere areas, while a more prominent L2 impairment are observed in patient with damage limited to right hemispheres areas. Also, the

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Neurocognitive Aspects of Processing Arabic and Hebrew

111

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

cases of M.H. and M.M., both native Arab speakers who acquired Hebrew (both Semitic languages), join experimental data in neurolinguistics and shed light on the relationship between language and mechanisms of neurobiology, and offer new psycholinguistic evidence to understand the dynamics of processing two languages in bilingual. In that regard, these both findings are also compatible with a cognitive study gained in our lab (see, Ibrahim, Aharon-Peretz, 2005). In that study, the response times (RTs) to target words in Arabic was not influenced by a previous appearance of its translation equivalent in Hebrew. These findings suggest that the lexical representations of Arabic and Hebrew words are equivalent, both reflecting the typical organization of L2 in a separate lexicon (Gerard and Scarborough, 1989). This finding converges with other reports established in cross-lingual semantic priming (de Groot and Nas, 1991), and repetition priming (de Groot and Nas, 1991). These findings indicated a possible relationship between the two L1 and L2 via the semantic level. This formulation fits the Hybrid Model of lexical representation in the bilingual brain (de Bot, 1992; de Groot, 1992). According to this model, a common semantic system is connected to two independent lexical systems corresponding to each of the two languages known by the bilingual. The ease of access to each lexicon from semantic memory depends on such factors as the age at which the lexical item was acquired and the frequency and recency of access (Snodgrass and Tsivkin, 1995). This will create a preference for choosing the native lexical item, particularly in the presence of aphasic disturbances. M.M. demonstrated such preference for Arabic in all the naming tasks. M.M.’s perception deficits suggest that bilinguals may possess two separate switching mechanisms: a lexical/semantic mechanism. M.M. provides evidence that Hebrew as a second language has a subsystem that is independent from Arabic and that this subsystem was more fragile and, therefore, more sensitive to brain damage. In the following chapter I will present the relationship between language structures (like orthography and morphological structures) and the performance asymmetries of hemispheric functioning.

2. Higher Cognitive Functions in Processing Arabic and Hebrew An important part of the literacy problem is posed by the Arabic orthographic system and its failure to support easy and efficient reading. Previous research on reading acquisition in the Arabic language has revealed that this process is slower than in Hebrew (Abu-Rabia, 1997; Saiegh-Haddad, 2003). The next section focuses on the hardship of letter and word discrimination in both beginning and skilled readers.

2.1. Orthographic Complexity Although Arab childrens’ scores on tests of phonological awareness were higher than those of monolingual Hebrew speakers, their scores on tests of reading achievement were lower (Eviatar and Ibrahim, 2001; Ibrahim, Eviatar, and Aharon-Perez, 2007). We suggested that this is due to the complexity of Arabic orthography as compared with Hebrew orthography. We tested the hypothesis that the graphemic complexity of Arabic is larger than in Hebrew, and that this results in additional perceptual load (Ibrahim, Aharon-Peretz and Eviatar, 2002). The subjects tested were all adolescent healthy native Arabic speakers who

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

112

Raphiq Ibrahim

had mastered Hebrew as a second language. We used oral and visual variants of the trail making test (Reitan and Wolfson, 1985) in both languages. Both versions have two levels of complexity: Level A requires connecting visually numbers or letters in order. Level B in the two modalities requires alternation between letters and numbers. Level B in the oral version requires declamation of the alternation. Performance time was the dependent variable. At the low level of complexity (Level A) there were no differences between performance in Hebrew and in Arabic in either the oral and the visual versions. In the more complex version (Level B), language (Hebrew or Arabic) did not affect speed in the oral version, but in the visual version, Arabic was performed significantly slower than Hebrew. These findings supported our major conclusion that Arabic letters are harder to identify than Hebrew letters as a result of their greater visual complexity. In two divided visual field studies we have shown that Arabic letters are harder to identify that English and Hebrew letters and have suggested that the locus of this difficulty is in the right hemisphere. In the first study (Eviatar and Ibrahim, 2004) we examined directly the effects of grapheme–phoneme conversion in English, Hebrew and Arabic, using a lateralized nonsense syllable identification task. The syllables were constructed as consonant– vowel– consonant (CVC) trigrams (the vowels in Hebrew and Arabic were letters that double as consonants or vowels), and the task of the participants was to identify the three letters. The stimuli were presented vertically in three conditions: left visual field (LVF), right visual field (RVF), and bilaterally (BVF). The participants were university students, native readers of each of the languages. In this study we were interested in errors, so we titrated exposure duration independently for each participant, in order to achieve a duration that resulted in 50% errors. This paradigm allowed us to measure three dependent measures that indexed different aspects of the task. The first measure was the mean exposure duration that was reached in each of the three groups of participants in order to achieve a 50% error rate. This is an index of the speed at which native readers can identify letters in each of the languages. This measure revealed that the Arabic readers required significantly longer exposure durations that the readers of Hebrew and English, and that Hebrew readers required significantly longer exposure durations that the readers of English. Thus, these results suggest that English letters are easier to identify that Hebrew letters, and that Arabic letters are the hardest to identify. The second measure in this study was the total number of errors in each presentation condition (LVF, RVF, and BVF). This measure revealed that all of the participants showed a right visual field advantage (RVFA) that reflects specialization of the LH for this linguistic task. This advantage, the difference between performance levels in the LVF and in the RVF was significantly larger in the Arabic speakers than in the other groups, as a result of poorer performance in the LVF in Arabic than in the other languages, while performance in the other presentation conditions (RVF, BVF) was equivalent among the groups. The third measure in this study was the difference between errors on the first letter and errors on the last letter of the trigram, a qualitative measure of sequential processing (Levy, Heller, Banich, and Burton, 1983; Eviatar, Hellige, and Zaidel, 1997). This qualitative measure revealed that Arabic and Hebrew speakers evinced a similar pattern that was different from the one shown by the English speakers. Of importance to us here is that it took longer for Arabic speakers to identify Arabic letters than it did for Hebrew speakers to identify Hebrew letters or English speakers to identify English letters. In addition, the large difference between performance levels in the two visual fields of Arabic speakers suggested

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Neurocognitive Aspects of Processing Arabic and Hebrew

113

that there are large differences in the abilities of the hemispheres in letter identification in Arabic, but not in the other languages. In divided visual field study we (Eviatar, Ibrahim, and Ganayim, 2004) explored the locus of this difference, and showed that adult native Arabic speakers who can read both Arabic and Hebrew, are better at identifying letters in Hebrew than in Arabic, and that the main disadvantage for Arabic letters is in the left visual field, when they are exposed to the right hemisphere (RH). We asked native Arabic speakers and native Hebrew speakers to perform a lateralized letter matching task in both Arabic and Hebrew, using a physical identity criterion (the Arabic speakers were literate in both languages, but the Hebrew speakers could not read Arabic, and thus performed the task as a pattern matching task). The pattern of results in response times was as we expected, revealing a RVFA in all of the conditions in which the participants could read (all except Hebrew speakers in Arabic, who showed a slight LVFA in this condition). The results of the accuracy measure were quite dramatic. In Hebrew, both groups revealed low error rates and equivalent performance in the two visual fields (both hemispheres are able to match letters quite well). In Arabic, the Hebrew speakers made many errors, with equivalent performance in the two visual fields. Arabic speakers revealed good performance in the RVF (their LH was able to match letters in Arabic as well as in Hebrew). In the LVF, however, Arabic speakers made as many errors as Hebrew speakers (who cannot read the language)! We hypothesized that the reason for this RH disadvantage in letter recognition is the complexity of grapheme-phoneme relations in Arabic. In order to examine our hypothesis that the RH cannot differentiate between very similar different letters in Arabic, whereas the LH can do so, we created a global-local task with two types of incongruent stimuli: one where the two letters on the two levels of the hierarchical stimulus were physically very different from each other: ‫ ت‬and ‫ ;م‬and another where the two letters were very similar to each other: ‫ ت‬and ‫ ب‬. Participants were required to attend to the local or the global levels of these hierarchical stimuli in different blocks. We measured the difference between congruent (where the same letter was used on both levels of the stimuli) versus incongruent conditions (where the letter in the global level was made out of small versions of the other letter). This difference indicated the degree of interference between the levels. In the first condition, where the letters were very different from each other, we replicated the results of other studies in other languages (e.g., Lamb, Robertson, and Knight, 1990). We found a global precedence (responses to the global level were always faster than to the local level (Navon 1977), and an asymmetry in the degree of interference between the two visual fields: stronger interference from the global level to the local level in the LVF, and the opposite pattern in the RVF. This pattern has been used to support the hypothesis that the LH is relatively more sensitive to the local aspects of visual stimuli and that the RH is relatively more sensitive to the global aspects of these stimuli. Most interestingly, however, in the second condition, where the two letters differed only in the number and placement of dots, there was no incongruence effect in the LVF at all, while congruent stimuli were faster than incongruent stimuli in the RVF in the local conditions. These results show that the RH cannot discriminate between letters that differ only in the placement or number of dots (e.g. /t/- ‫ ت‬and /b ‫ ب‬- / ), but that the LH can do so.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

114

Raphiq Ibrahim

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

2.2. Word Morphology Arabic and Hebrew, as Semitic languages, are constructed by combining a consonantal root (that carries most of the semantic information) and a word pattern that includes vowels as well as consonants, and provides information about the word class and its morphological status. Roman and Pavard (1987) used oculomotor recording techniques to evaluate visual scanning strategies. They found that although mean reading time did not differ between Arabic and French texts (note that for conveying identical content, the number of words needed in Arabic is less than in French because Arabic morphology is more dense), gaze duration per word was significantly longer in Arabic (342 ms) than in French (215 ms). This phenomenon also has been found in comparisons of Hebrew and English text reading, in which the morphology of Hebrew is dense and similar to that of Arabic, and English morphology is concatenative and more similar to French (Shimron and Sivan, 1994). A number of psycholinguistic studies (Feldman et al., 1995; Frost et al., 1997; Deutsch, Frost, and Forster, 1998; Berent, 2002) have explored the effects of the morphology and orthography of Hebrew on lexical access and the structure of the mental lexicon. One of the conclusions from these studies is that the nonconcatenative and agglutinative morphological structure of Hebrew, together with the distributional properties of abstract word forms, results in the inclusion of subword morphological units in the mental lexicon of Hebrew speakers. Similarly, Prunet, Beland, and Idrissi (2000) report a case study of an Arabic-French agrammatic patient, who showed identical deficits in the two languages, except for a specific type of error, metathesis, in which he modified the order of the root consonants, with the vowel patterns remaining intact, only in Arabic, not in French. They interpret this finding as reflecting the manner in which words are stored in the mental lexicon in the two languages: whole words plus affixes in French, and roots plus word patterns in Arabic. These findings converge with the conclusions of Eviatar (1999, Experiment 4) and Eviatar and Ibrahim (2004), who showed that nonsense CVC trigrams are processed sequentially in both visual fields in English, but in neither visual field in Hebrew and in Arabic, and hypothesized that this is because Hebrew and Arabic nonwords cannot be read sequentially. A similar conclusion for words was reached by Farid and Grainger (1996), who showed that initial fixation position in a word results in somewhat different response patterns in French and in Arabic. In French, fixation slightly to the left of the word's center results in best recognition for both prefixed and suffixed words, while in Arabic, prefixed words result in best recognition from leftward fixations and suffixed words result in best recognition from rightward initial fixations. They suggest that this is due to the greater importance of morphological structure in Arabic, because “...much of the phonological representation of the word can be recovered only after successfully matching the consonant cluster to a lexical representation” (p.364), that is, after extraction of the root. Berent (2002) has also concluded that in Hebrew, “Speakers decompose the root from the word pattern in on-line word identification…” (p. 335). Most recently we reported that the different manner in which words are constructed in English and in Hebrew and Arabic has an effect on the division of labor between the cerebral hemispheres in a lateralized lexical decision task (Eviatar and Ibrahim, 2007). We presented native speakers of Arabic, Hebrew, and English with morphologically simple and complex words and nonwords in their native language, and measured indexes of hemispheric integration. Morphological complexity was operationalized differently in English than in the

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Neurocognitive Aspects of Processing Arabic and Hebrew

115

Semitic languages. In English we defined monomorphemic words as morphologically simple, and derivations (e.g., farmer=farm+er) as morphologically complex. Morphologically complex nonwords were made up of legal morphemes in illegal combinations (e.g., logly). In Arabic and in Hebrew we defined a word as morphologically simple if the root+wordform structure was not transparent (e.g., the word is not easily divisible into these morphemes or the root is not generative, and appears only in that form), and as morphologically complex if it was easily and transparently divisible into these elements. Morphologically complex nonwords were created by inserting nonexistent roots into legal wordforms. In English, we replicated the findings of previous studies: similarly to Iacoboni and Zaidel (1996), we showed that while the RH is able to independently recognize nonwords; it draws upon resources of the LH when encountering words. Similarly to Burgess and Skodis (1993) in English, and to Koenig, Wetzel, and Carramazza (1992) in French, we showed that for the English speakers, only the LH was sensitive to the morphological complexity of the stimuli. Morphological complexity affected words and nonwords in the same manner, with complex stimuli requiring longer latencies to be identified either as a word or as a nonwords only in the RVF. As opposed to the English speakers, both groups of speakers of the Semitic languages showed bilateral sensitivity to morphological complexity. In addition, the Arabic and Hebrew readers showed higher values on our indexes of interhemispheric integration, suggesting more intensive hemispheric cooperation during the reading of Hebrew and Arabic than of English. Interestingly, in both languages, morphological complexity had opposing effects for words and for nonwords. Morphological complexity, or transparency of the root+wordform structure, facilitated the recognition of words and decelerated the rejection of nonwords. We suggested that the nonconcatenative morphology of the Semitic languages, in which words are analyzed into their root and word-form constituents, requires that both hemispheres be sensitive to morphological structure. The automatic analysis of a character string into a recognizable word-form and a root resulted in faster recognition of complex words than of the simple words, which are not divisible in this way. This analysis also resulted in slower responses to complex nonwords than to simple nonwords, which did not contain a recognizable word form. Thus, the word form made complex nonwords more “wordlike”, requiring a more intensive search before they could be correctly rejected in the lexical decision task. In general, we found that the manner in which words are formed in these different languages resulted in different types of interhemispheric division of labor in the lexical decision task. Specifically, we showed that when languages make different types of demands upon the cognitive system, interhemispheric interaction is dynamic and is suited to these demands. Arabic and Hebrew require a higher level of interhemispheric interaction than does English.

2.3. Reading in the Nonnative Language Recently, we examined the interaction of the effects of reading Arabic and other languages (Ibrahim and Eviatar, 2008). We took advantage of the high facility of Arab university students in Hebrew and English, in order to examine the manner in which such multilingual brains deal with word morphology. We used the same paradigm described above,

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

116

Raphiq Ibrahim

with trilingual Arab participants making lexical decisions on morphologically simple and complex words and nonwords, in Arabic, in Hebrew, and in English. One of the interesting results from this study is that in the RVF/LH there was a significant difference between performance levels in the three languages, reflecting significantly better performance in Arabic, which these participants learned to read first, than in Hebrew and English, which these participants consider their nonnative languages. However, in the LVF/RH, there was no difference between performance levels in the three languages. We interpret this as reflecting the specific RH deficit in reading Arabic, which lowers performance in the LVF for this language, such that it is not better than the second and third languages, in which these participants have lower facility. Another interesting result from this study is that the participants showed the same patterns of interhemispheric cooperation in the three languages, suggesting that they used the same reading strategies in all of the languages. For Hebrew, the patterns are similar to the ones shown by native Hebrew speakers, suggesting that morphological processes are similar in these similar languages. However, the patterns shown by native Arabic speakers in English are different from the patterns shown by native English speakers. Thus, our participants were reading a second (or third) language with the same mechanisms as the first learned language. This type of pattern was also reported by Eviater (1999) for native Hebrew readers recognizing nonsense syllables in English. Eviater (1999) suggested that this is due to the demand for morphological decomposition in Hebrew that determines reading strategies for other languages as well. The results reported here suggest that these same demands occur for Arabic readers.

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Conclusion All of the findings of our group’s studies related to Arabic language, in addition to findings of other researchers (see Abu-Rabia, 1997; Saiegh-Haddad, 2003), support the notion that Arabic has unique features that contribute to the inhibition and slowness of the reading process. Furthermore, the chapter argued for inclusion of the neurofunctional perspective as a comprehensive basis for the discussion of the organization of two languages in the cognitive system of the Arabic-Hebrew bilingual and how we should treat teaching Hebrew as second language (L2). It was also suggested that bilinguals may possess two separate switching mechanisms: a lexical/semantic mechanism. In that regard, the data provided evidence that Hebrew as a second language has a subsystem that is independent from Arabic and that this subsystem is more fragile and, therefore, more sensitive to brain damage.

References Abu-Rabia, S. (1997). Reading in Arabic orthography: The effect of vowels and context on reading accuracy of poor and skilled native Arabic readers. Reading and Writing: An Interdisciplinary Journal, 9, 65-78. Abu-Rabia, S. (1998). “Reading Arabic texts: Effects of text type, reader type and vowelization”. Reading and Writing: An Interdisciplinary Journal 10: 105-119.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Neurocognitive Aspects of Processing Arabic and Hebrew

117

Abu-Rabia, S. (2000). Effects of exposure to literary Arabic on reading comprehension in a diglossic situation. Reading and Writing: An Interdisciplinary Journal 13: 147-157. Aglioti, S., Smania, N., Manfredi, M., Berlucchi, G. (1996). Disownership of left hand and objects related to it in a patient with right brain damage. Neuroreport, 8(1), 293-296. Altarriba, J. (1992). The representation of translation equivalents in bilingual memory. In R. J. Harris (Ed.), Cognitive processing in bilinguals. (pp.157-174). Amesterdam: Elsevier. Azzam, R. (1989). Orthography and reading of the Arabic language. In P. G. Aaron and R. M. Joshi (Eds.), Reading and writing disordersin different orthographic systems (pp. 1–29). New York: Kluwer Academic. Azzam, R. 1990. The nature of Arabic reading and spelling errors of young children: a descriptive sudy. Unpublished doctoral dissertation, Teachers College, Columbia University, N.Y. USA. Azzam, R. (1993). The nature of Arabic reading and spelling errors ofyoung children. Reading and Writing, 5, 355–385. Ayari, S. 1996. “ diglossia and illiteracy in the Arab world.” Language, Culture and Curriculum 9:243-252. Bentin, S., and Ibrahim, R. (1996). New evidence for phonological processing during visual word recognition: The case of Arabic. Journal of Experimental Psychology: Learning, Memory, and Cognition, 22, 309-323. Benson, D. F. (1979). Neurologic correlates of anomia. In H.Whitaker and H. A.Whitaker (Eds.), Studies in neurolinguistics: Vol. 4. New York: Academic Press. Berman, R.A. (1978). Modern Hebrew structure. Tel Aviv, Israel: University Publishing. Bentin, S., and Ibrahim, R. (1996). New evidence for phonological processing during visual word recognition: The case of Arabic. Journal of Experimental Psychology: Learning, Memory, and Cognition, 22, 309-323. Ben Zeev, S. (1977). The influence of bilingualism on cognitive strategy and cognitive development. Child Development, 48, 1009 1018. Bialystok, E.. (1991). Language processing in bilingual children. Cambridge, England, UK: Cambridge University Press. Bialistok, E. and Ryan, E.B. (1985). Toward a definition of metalinguistic skill. MerrillPalmer Quarterly, 31, 229-251. Bentin, S., and Frost, R. (1987). Procsessing lexical ambiguity and visual word recognition in a deep orthography. Memory and Cognition, 15, 13-23. Bentin, S., Bargai, N., and Katz, L. (1984). Graphemic and phonemic coding for lexical access: Evidence from Hebrew. Journal of Experimental Psychology: Learning, Memory and Cognition, 10, 353-368. Berent, I. (2002). Identity avoidance in the Hebrew lexicon: implications for symbolic accounts of word formation. Brain and language, 81(1-3), 326-41. Bertelson, P., Morais, J., Alegria, J., and Content, A. (1985). Phonetic analysis capacity and learning to read. Nature, 313, 73–74. Bertelson, P., Morais, J., Alegria, J., and Content, A. (1985). Phonetic analysis capacity and learning to read. Nature, 313, 73–74. Bookheimer.S. (2002). “Functional MRI of Language: New Approaches to Understanding the Cortical Organization of Semantic Processing.” Annual Review of Neuroscience, 25:15188.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

118

Raphiq Ibrahim

Bradley, L. and Bryant, P.E. (1983). Categorizing sounds and learning to read: A causal connection. Nature, 301(5899),419 421. Campbell, R., and Sais, E., (1995). Accelerated metalinguistic (phonological) awareness in bilingual children. British Journal of Developmental Psychology, 13(1), 61 68. Caramazza, A. (1997). How many levels of processing are there in lexical access? Cognitive Neuropsychology, 14, 177-208. Cummins,J. (1987). Bilingualism, language proficiency, and metalinguistic development. In P. Homel and M. Palij, (Eds) Childhood bilingualism: Aspects of linguistic, cognitive, and social development. (pp. 57-73). Hillsdale, NJ, USA: Lawrence Erlbaum Associates. Chen, H. C., and Leung, Y. S. (1989). Patterns of lexical processing in a nonnative language. Journal of Experimental Psychology: Learning, Memory and Cognition, 15, 316-325. Courrieu, P., and do, P.(1987). Perceptual analysis of word in Arabic. In J.K. O’Regan andA. Levy-Schoen (Eds.) Eye Movement: From Physiology to Cognition.(pp.451-458), NorthHolland: Elsevier. Dash, U. N. and Mishra, H. C. (1992). Bilingualism and metalinguistic development: Evidence from Kond tribal culture. Psychological Studies. 37(2 3), 81 87. de Bot, K. (1992). A bilingual production model: Levelt’s “speaking” model adapted. Applied Linguistics, 13, 1–24. Da Fontoura, H.L., and Siegel L.S. (1995). Reading, syntactic, and working memory skills of bilingual PortugueseEnglish Canadian children. Reading and Writing: AnInterdisciplinary. Dehaene. S., Dupoux. E., and Mehler. J. (1997). Anatomical variability in the cortical representation of first and second language. Neuroreport, 8, 3809–3815. Dell, G. S. (1986). A spreading activation theory of retrieval in sentence production. Psychological Review, 93, 283-321. Dell, G. S., Schwartz, M. F., Martin, N. M., Saffran, E. M., and Gagnon, D. A. (1997). Lexical access in aphasic and nonaphasic speakers. Psychological Review, 104, 801-838. Deutsch, A., Frost, R. and Forster, K. I. (1998). Verbs and Nouns are Organized and Accessed Differently in the Mental Lexicon: Evidence from Hebrew. Journal of Experimental Psychology: Learning, Memory and Cognition, 24(5), 1238-1255. Doyle, A., Champagne, M., and Segalowitz, N. (1978).Some issues on the assessment of linguistic consequences of early bilingualism. In M. Paradis (Ed.) Aspects of bilingualism (pp. 13 20). Colombia, SC: Hornbeam Press. De Manrique, A. M. B., Gramigna, S. (1984). La segmentación fonológica y silábica en ni nos de preescolar y primer grado. Lectura y Vida, 5/1, 4–14. Eviatar, Z. (1995) Eye movements and attention: Effects on lateralized ignoring. Brain and Cognition, 29, 137-150. Eviatar, Z. (1996) Orthography/phonology relations and hemispheric functioning. Brain and Cognition, 32 (2), 151-155. Eviatar, Z. (1997) Language experience and right hemisphere tasks: The effects of scanning habits and multilingualism. Brain and Language, 58, 157-173. Eviatar, Z. (1999). Cross-language tests of hemispheric strategies in reading nonwords. Neuropsychology, 13 (4), 1-18. Eviatar. Z., Leiken. M. and Ibrahim, R. (1999). Phonological processing of second language phonemes: A selective deficit in a bilingual aphasic. Language Learning, 49.1, 121-141.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Neurocognitive Aspects of Processing Arabic and Hebrew

119

Eviatar, Z. and Ibrahim, R. (2001). Bilingual is as bilingual does: Metalinguistic abilities of Arabic-speaking children. Applied Psycholinguistics, 21(4), 451-471. Eviatar, Z., and Ibrahim, R. (2007). Morphological structure and hemispheric functioning: The contribution of the right hemisphere to reading in different languages. Neuropsychology, 21(4), 470-484. Edwards, D., and Christophersen, H. (1988). Bilingualism, literacy and meta linguistic awareness in preschool children. British Journal of Developmental Psychology, 6(3), 235 -244. Eid, M. (1990). Arabic linguistrics: The current scene. In M. Eid (Ed) Perspectives on Arabic linguistics I. (pp 3 37). Amsterdam: John Benjamins Publishing Company. Eviatar, Z., Hellige, J. B., and Zaidel, E. (1997). Individual differences in hemispheric specialization: Effects of gender and handedness. Neuropsychology, 11, 562-576. Eviatar, Z., Ibrahim, R., and Ganayim, D. (2004). Orthography and the hemispheres: Visual and linguistic aspects of letter processing. Neuropsychology, 18(1), 174-184. Eviatar, Z. and Ibrahim, R. (2004). Morphological and Orthographic Effects on Hemispheric Processing of Nonwords: A Cross-Linguistic Comparison. Reading and Writing:17, 691705. Fabbro, F. (2001).The bilingual brain: cerebral representation of languages. Brain and Language, 79, 211-222. Farid, M., and Grainger, J., (1996). How initial fixation position influences visual word recognition: A comparison of French and Arabic. Brain and Language, 53, 351-361. French. In J. K. O’Regan and A. Levy-Schoen (Eds.), Eye movement: From physiology to cognition (pp. 431-440). Amsterdam, The Netherlands: North Holland Elsevier. Frost, R. (1994). Prelexical and postlexical strategies in reading: Evidence from a deep and a shallow orthography. Journal of Experimental Psychology: Learning,Memory, and Cognition, 20, 1-16. Frost, R., Forster, I. K., and Deutsch, A. (1997). What can we learn from the morphology of Hebrew? A masked-priming investigation of morphological representation. Journal of Experimental Psychology : Learning, Memory, and Cognition. 4, 1-28. Frost, R., Katz, L., and Bentin, S. (1987). Strategies for visual word recognitionand orthographical depth: A multilingual comparsion. Journal of Experimental Psychology: Human Perception and Performance, 13, 104-114. Galambos, S. J., and Hakuta, K. (1988). Subject specific and task specific characteristics of metalinguistic awareness in bilingual children. Applied Psycholinguistics, 9(2), 141 162. Galambos, S. J., and Goldin Meadow, S. (1990). The effects of learning two languages on levels of metalinguistic awareness. Cognition, 34(1), 1 56. Gerard, L. D., and Scarbourough, D. L. (1989). Language specific lexical access of homographs by bilinguals. Journal of Experimental Psychology: Learning, Memory and Cognition, 15.2, 305-315. Groot, de. M. B. A., (1992). Determinants of word translation. Journal of Experimental Psychology: Learning, Memory, and Cognition, 18, 1000-1018. Groot, de. A. M. B., Dannenbur, J. G. and van aell (1994). Forward and backward word translation by bilinguals. Journal of Memory and Language, 33, 600-629. Groot, de. M. B. A. and Nas, L. J. G. (1991). Lexical representation of cognates and noncognates in compound bilinguals. Journal of Memory and Language, 30, 90- 123.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

120

Raphiq Ibrahim

Green, D. W. (1986). Control, activation and resource: A framework and a model for the control of speech in bilinguals. Brain and Language, 27, 210-223. Green, D. W. (1998). Mental control of the bilingual lexico-semantic system. Bilingualism: Language and Cognition, 1, 67-81. Ibrahim, R. (2008). Performance in L1 and L2 observed in Arabic-Hebrew bilingual aphasic following brain tumor: A case of double dissociation. Psychology Research and Behavior Management. 8, 11-19. Ibrahim, R. (submitted). Selective deficit of second language: A case of Arabic-Hebrew bilingual brain-damaged patient. Behavioral and Brain Functions. Ibrahim, R., Eviatar, Z., and Aharon Peretz, J. (2002). The characteristics of the Arabic orthography slow it’s cognitive processing. Neuropsycholgy, 16(3), 322-326. Ibrahim, R.and Aharon-Peretz, J. (2002). Text processing and perceptual characteristics of Arabic orthograpy. Neurology update, 9, 22-24. (Hebrew) Ibrahim, R. and Aharon-Peretz, J. (2005). Is literary Arabic a second language for native Arab speakers?: Evidence from a semantic priming study. The Journal of Psycholinguistic Research, 34(1),51-70. Ibrahim, R. (2006). Morpho-Phonemic similarity within and between languages: A factor to be considered in processing Arabic and Hebrew. Reading and Writing, 19, (6), 563-586. Ibrahim, R., Eviatar. Z., and Aharon Peretz, J. (2007). Metalinguistic awareness and reading performance: A cross language comparison. The Journal of Psycholinguistic Research, 36(4), 297-317 Lawrence Erlbaum Associates. Iacoboni, M., and Zaidel, E. (1996). Hemispheric independence in word recognition: Evidence from unilateral and bilateral presentations. Brain and Language, 53, 121-140. Illes. J., Francis. W. S., and Desmond. J. E, (1999). Convergent cortical representation of semantic processing in bilinguals. Brain and Language,70, 347–363. Kaplan, E., Goodglass, H., and Weintraub, S. (1983). Boston Naming Test. Philadelphia: Lea and Febiger. Keatley, C., and DeGelder, B. (1992). The bilingual primed lexical decision task: cross language priming disappears with speeded responses. European Journal of Cognitive Psychology, 4(4), 273-292. Keatley, C., Spinks, J., and DeGelder, B. (1990). Asymmetrical semantic facilitation between language: Evidence for separate representational system in bilingual memory Unpublished Manuscript, University of Tilberg. The Netherlands. Kertesz, A. (1982). Western aphasia battery. New York: Grune and Stratton. Kim. K. H., Relkin. N. R., and Lee. K. M. (1997). Distinct cortical areas associated with native and second languages. Nature, 388, 171–174. Klein, D., Zatorre, R.J.,Milner,B., and Meyer, E.(1994). Left putamenactivation when speaking a second language: Evidence from PET. Neuroreport, 5, 2295–2297. Klein, D., Milner, B., Zatorre, R.J., Zhao, V., and Nikelski, J. (1999). Cerebral organization of bilinguals: A PET study of Chinese-English verb generation. NeuroReport, 10, 28412846. Kirsner, K. (1986). “Lexical representation: is a bilingual account necessary”. Language Processing in Bilinguals: Psycholinguistic and Neuropsychological Perspectives ed. by J. Vaid, 21-45. Hillsdale, NJ: Lawrence Erlbaum. Kroll, J. F.,Sholl, A., Altarriba, J., Luppino, C., Moynihan, L., and Sanders, C. (1992, November). Cross-language semantic priming: Evidence for independent lexical and

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Neurocognitive Aspects of Processing Arabic and Hebrew

121

conceptual contributions. Paper presented at the Thirty-Third Annual Meeting of the Psychonomic Society , St. Louis, MO. Kroll, J. F., and de Groot, A. M. B. (1997). Lexical And Conceptual Memory in the Bilingual: Mapping Form to Meaning in Two Languages. In de Groot, A. M. B. and Kroll, J. F. (Eds.), Tutorials In Bilingualism: Psycholinguistic Perspectives. (pp. 201-224). Mahwah, NJ: Lawrence Erlbaum Publishers. Kroll, J. F., and Stewart, E. (1994). Category interference in translation and picture naming: Evidence for asymmetric connections between bilingual memory representations. Journal of Memory and Language. 33, 149-174. Lamb, M.R., Robertson, L.C., and Knight, R.T., (1990). Component mechanisms underlying the processing of hierarchically organized patterns: Inferences from patients with unilateral cortical lesions. Journal of Experimental Psychology: Learning, Memory, and Cognition, 16, 471-483. Luria, A. R. (1970). Traumatic aphasia: Its syndromes, psychology, and treatment. The Hague: Mouton. Lucas, T. H., McKhann, G. M., and Ojemann, G. A. (2004). Functional separation of languages in the bilingual brain: A comparison of electrical stimulation language mapping in 25 bilingual patients and 117 monolingual control patients. Journal of neurosurgery, 101, 449–457. Maamouri, M. (1998). Language education and human development: Arabic diglossia and its impact on the quality of education in the Arab region”. The Mediterranean Development Forum. Washington, DC: The World Bank. Major, R.C. (1999). Chronological and stylistic aspects of second language acquisition of consonant clusters. Language Learning 49 Supplement 1, 123-150. Mahadin. R. S. (1982). The morphophonemics of the Standard Arabic triconsonantal verbs. Doctoral dissertation, University of Pennsylvania, Philadelphia. McClelland, J, L., Rumelhart,D, E., and Hinton, G, E. (1986). The appeal of parallel distributed processing. In D. E. Rumelhart and J.L. McClelland (Eds.) parallel distributed processing. Cambridge, Mass .: M.I.T. press. Meyers. J.E., Meyers. K. R.(1995). Rey complex figure test and recognition trial: professional manual. Odessa. FL: Psychological Assessment Resources. Milner. B., (1963). Effects of brain lesions on card sorting. Arch Neurol, 9, 90–100. Moreno, E. M., Federmeier, K. D., and Kutas, M. (2002). Switching languages, switching palabras (words): An electrophysiological study of code-switching. Brain and Language, 80, 188-207. Moretti, R., Bava, A., Torre, O., Antonello, R. M., Zorzon, M., Zivandinov, R. (2001). Bilingual aphasia and subcortical–cortical lesions. Perceptual and Motor Skills, 92, 803–814. Navon, D.,(1977). Forest before trees: The precedence of global features in visual perception. Cognitive Psychology, 9, 353-383. Ojemann, G.A. (1983). Brain organization for language from the perspective of electrical stimulation mapping. Behavioral and Brain Sciences, 6, 189-206. Ojemann. G., Ojemann. J., and Lettich. E. (1989). Cortical language localization in left, dominant hemisphere. An electrical stimulation mapping investigation in 117 patients. J. Neurosurg., 71, 316–326,

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

122

Raphiq Ibrahim

Paradis, M, (1998). Aphasia in bilinguals: How atypical is it? In P. Coppens and Y. Lebrun (Eds), Aphasia in atypical populations. (pp. 35 66). Mahwah, NJ: Lawrence Erlbaum Associates. Pouratian. N., Bookheimer. S. Y., and O’Farrell. A. M. (2000). Optical imaging of bilingual cortical representations. Case report. J. Neurosurg. 93, 676–681. Prunet, J., Beland, R., and Idrissi, A. (2000). The mental representation of Semitic words. Linguistic Inquiry 31, 4, p. 609-648. Pu, Y., Liu, H. L., Spinks, J. A., Mahankali, S., Xiong, J., and Feng, C. M. (2001). Cerebral hemodynamic response in Chinese (first) and English (second) language processing revealed by event-related functional MRI. Magnetic Resonance Imaging, 19, 643–647. Roman, G., and Pavard, B. (1987). A comparative study: How we read Arabic and French. In J. K. O’Regan and A. Levy-Schoen (Eds.), Eye movement: From physiology to cognition (pp. 431-440). Amsterdam, The Netherlands: North Holland Elsevier. Reitan, R. M., and wolfson, D. (1993). The Halsted reitan neuropsychological test battery: Theory and clinical interpretation (2nded.). Tucson, AZ: Neuropsychology Press. Saffran, E. M., and Schwartz, .M. F. (1994). Of Cabbages and Things: Semantic Memory From a Neuropsychological Perspective – A Tutorial Review. In C. Umiltà and M. Moscovitch (Eds.), Attention and Performance: XV. Conscious and Nonconscious Information Processing (pp. 507–536). Cambridge: MIT Press. Saiegh- Haddad, Elinor. 2003. “Linguistic distance and initial reading acquisition: The case of Arabic. Diglossia” Applied Psycholinguistics 24: 431-451. Schwanenflugel, P. J., and Rey, M. (1986). Interlingual semantic facilitation: Evidence for a common representational system in the bilingual lexicon. Journal of Memory and Language, 25, 605-618. Shimron, J. and Sivan, T. (1994). Reading profeciency and orthography: Evidence from Hebrew and English. Language learnining. 44: 5-27. Simos. P. G., Castillo. E. M., and Fletcher. J. M. (2001). Mapping of receptive language cortex in by using magnetic source imaging. J. Neurosurg., 95, 76–81. Snodgrass, J. G., and Tsivkin, S. (1995). Organization of the bilingual lexicon: Categorical versus alphabetic cuing in Russian-English bilinguals. Journal of Psycholinguistic Research, 24, 145–162. Van Hell, J. G., and de Groot, A. M. B. (1998). Conceptual representation in bilinguals memory: Effects of concreteness and cognate status in word association. Bilingualism: Language and Cognition,1 (3), 193-211. Wang, M., Koda, K., and Perfetti, C. A. (2003). Alphabetic and nonalphabetic L1 effects in English word identification: A comparison of Korean and Chinese L2 learners. Cognition, 87, 129-149.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

In: Bilinguals: Cognition, Education and Language Processing ISBN: 978-1-60741-710-1 Editor: Earl F. Caldwell, pp. 123-145 © 2010 Nova Science Publishers, Inc.

Chapter 6

VISUAL WORD ACCESS IN MONOLINGUALS AND BILINGUALS IN ENGLISH AND SPANISH John Evar Strid∗,1 and James R. Booth2 1

Northeastern Illinois University, IL, USA 2 Northwestern University, IL, USA

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Abstract This research examined if visual word access varies according to language and bilingual status by comparing Spanish and English, priming two syllable CVCV words with bilingual children and monolingual children. The results suggest that lexical access in English is based on a larger phonological sub-lexical unit because of greater report of a unit bigger than the syllable, among both bilingual and monolingual subjects. In contrast, only weak evidence suggested that Spanish lexical access was based on the syllable because of greater report of that unit for bilinguals and monolinguals. Finally, monolingual or bilingual status of the reader did not have influence on English lexical access; however, Spanish bilinguals were influenced by the acquisition of English, suggesting that orthographically opaque languages can have an effect on transparent languages or that immersion, language dominance and literacy experience can influence reading in the other language.

Keywords: Bilingualism –Spanish/English –Reading –Visual word activation –Priming. What kind of adaptation do bilingual readers of alphabetic languages make to the different languages that they are reading? Do they get the same phonological information out of letters regardless of which language they are reading? Do they follow the same pathway to visual lexical access as monolingual readers in both of the languages? These are primary questions that the present research sets out to address by directly comparing young bilingual and monolingual readers of English and Spanish.



E-mail address: [email protected]. Correspondence concerning this article should be addressed to John Evar Strid, Department of Linguistics, Northeastern Illinois University, 5500 N. St. Louis Ave. Chicago, Il 60625.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

124

John Evar Strid and James R. Booth

Relatively little research has examined bilingual literacy –with most research on this subject focusing on whether the accomplishment of knowing two languages confers some sort advantage in terms of metalinguistic awareness –generally finding an advantage in terms of word awareness and syntactic awareness (e.g., Ben Zeev, 1977; Bialystok, 1988; Galambos and Goldin-Meadow, 1990). The few studies that have studied phonological awareness in bilinguals have found an early advantage over monolinguals that diminishes with the onset of literary instruction (Bruck and Genessee, 1995; Yelland, Pollard, and Mercuri, 1993). This bilingual phonological awareness advantage is conditioned by the relationship between the two languages and the nature of their writing systems, with a study of first and second graders showing that English-Spanish bilinguals outperform English monolinguals who in turn outperformed Chinese-English bilinguals in a phoneme segmentation task (Bialystok, et al., 2003). This notion of the nature of the script affecting lexical access is central to the present study and one that has been approached in two distinct manners. First, some have argued that the consistency of the script may affect lexical access, particularly the use of phonology in lexical access when reading in a writing system that is more opaque (Katz and Frost, 1992).  Some  experimental  evidence  comparing  monolinguals  in  languages  with  scripts  of  varying  opacity  suggests  that  monolingual  readers  differ  in  their  access  of  phonology  and that readers of opaque scripts are less reliant on phonology (Frost, Katz, and Bentin, 1987; Kang and Simpson, 1996). Ziegler and Goswami (2005, 2006) proposed a grain size theory of reading, according to which readers of transparent scripts such as Spanish are much more reliant on small grain size units, such as the phoneme, using the highly consistent grapheme-phoneme correspondences to efficiently convert the letters into phonemes and to gain lexical access. In comparison, readers of scripts with relatively poor sound-letter correspondence, such as English, are forced to be flexible in the unit or grain size of lexical access, at times relying on small grain size grapheme/phoneme conversion, while in others relying on larger units such as onset/rimes and words. In support of the theory, they point to evidence that monolingual readers of transparent scripts reach ceiling in word recognition tasks within the first year of literacy instruction, while English accuracy is much lower (Bruck, Genesee, and Caravolas, 1997; Ellis and Hooper, 2001: Frith, Wimmer, and Landerl, 1998; Goswami, Gombert and De Barerra, 1998; Seymour, Aro, and Erskine, 2003) and research findings that suggest that English readers have greater sensitivity to units that are larger in grain size in comparison to readers of transparent scripts (Brown and Deavers, 1999; Goswami, Ziegler, Dalton, and Schneider, 2001; Goswami, Ziegler, Dalton, and Schneider, 2003). Second, some research has directly examined how bilinguals adapt to reading in different languages. Most of these studies have tested if script differences between alphabetic, syllabic, or logographic writing systems affect lexical access in a second language –with results that suggest that the nature of an L1 script may affect phonological access in L2 reading (e.g., Koda, 1989, 2007; Saito, Inoue, and Nomura, 1979; Wang, Koda, and Perfetti, 2003; Wang, Perfetti, and Liu, 2003). Relatively little research has examined bilingual readers in both languages to see if their reading varies in the different languages or has compared them directly to monolinguals to see if lexical access varies according to bilingual status. Some have examined the speed and ease of bilingual readers in both their languages, finding that their reading was generally superior in L1 (e.g., Favreau, Komoda, and Segalowitz, 1980; Favreau and Segalowitz, 1983),

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Visual Word Access in Monolinguals and Bilinguals in English and Spanish

125

but that this difference in speed can be affected by the nature of the scripts (Shimron and Sivan, 1994). Recent research examining bilingual adults in English and Spanish suggests that readers access words via different units of lexical access, with readers in English using larger units and in Spanish segmenting words into syllables (Strid and Booth, 2007), a finding that supports the grain size theory (Ziegler and Goswami, 2005, 2006). However, fMRI results reported by Tan et al. (2003) showed bilingual Chinese-English readers activating similar neural networks for a rhyming task in both Chinese and English (languages written in scripts quite distinct in their relation to phonology), but the activated areas were quite distinct from those activated by monolingual English readers. In summary, some results comparing reading in monolinguals and bilinguals suggest that readers may not completely adapt to the second language (Tan et al., 2003), while others suggest that they eventually do (Strid and Booth, 2007). Additional research is needed to clearly test bilingual adaptation during reading. In general, previous research argues that monolingual readers of English and Spanish access words in differing manners. In particular, Spanish and English readers differ in their use of syllables in lexical access during reading, perhaps following from Spanish being a syllable timed language, with relatively equal syllable weight and clearly defined syllable structure (Carreires, Alvarez, and de Vega, 1993; Harris, 1983; Sánchez-Casas, 1996). In contrast, English is a stressed timed language, with alternation of strong and weak syllables and less well defined syllables (Jensen, 2000; Kahn, 1976; Rubach, 1996; Selkirk, 1982). Thus leading readers during visual word recognition in the two languages to exploit different levels of the prosodic hierarchy –a layered representation of the internal phonological structure of words (Blevins, 1995; Jensen, 2000; Nespor and Vogel, 1986; Selkirk, 1982). Much previous research has consistently shown that the syllable is a key unit of lexical access in Spanish (Álvarez, de Vega, and Carreiras, 1998; Álvarez, Carreiras, and de Vega, 2000; Álvarez, Carreiras, and Taft, 2001; Álvarez, Carreiras, and Perea, 2004; Carreiras et al., 1993; Carreiras and Perea, 2002; Domínguez, de Vega, and Cuetos, 1993; Perea and Carreiras, 1998). Recent research directly comparing Spanish monolingual to English monolingual readers found that Spanish readers showed more orientation to syllables regardless of reading level, while better readers of English showed greater orientation to units larger than the syllable in comparison to weaker readers (Taft, Álvarez, and Carreiras, 2007). In general, research on the use of the syllable in visual word access in English suggests that activation of the syllabic level of the prosodic hierarchy may come later during phonological assembly (e.g., Ashby and Rayner, 2004; Ferrand, Seguí, and Humphreys, 1997; Jared and Seidenberg, 1990; Schiller, 2000). In fact, the precise unit of visual word access in English is a matter of some contention, with many contradictory proposals (e.g., Butterworth, 1983; Chialant and Caramazza, 1995; Cole, Beauvillain, and Seguí, 2000, Lukatela, Gligorijevic, Kostic, and Turvey, 1980; Taft, 1979, 1994; Taft and Forster, 1975, 1976; Ziegler and Goswani, 2005). Overall, the unit of reading used in lexical access in the two languages appears to vary, with Spanish readers clearly exploiting the syllabic level of the prosodic hierarchy and English readers exploiting a larger or variable unit. Children who are developing language learners seem to make a universal, sequential progression from sensitivity to larger phonological units to increased sensitivity to smaller phonological units even before beginning to learn literacy skills. In tasks designed to test knowledge of different phonological levels, children master word-level skills before syllablelevel skills, syllable-level skills before onset-rime level skills, and onset-rime level skills

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

126

John Evar Strid and James R. Booth

before phoneme-level skills (Anthony and Lonigan, 2004; Anthony, et al., 2002; Stanovich, 1992; Treiman and Zukowski, 1996). The finest grained level, phoneme awareness, only appears to develop when children are explicitly taught to break words into phonemes during literacy instruction. Thus children are only successful at tasks that involve phoneme segmentation when they are beginning to learn to read and write, perhaps because individual phonemes are less salient in speech (Goswami and Bryant, 1990; Zeigler and Goswami, 2005). This progression of sensitivity from larger to smaller phonological units appears to be generalized in the process of literacy acquisition (Goswami, 1988; Zeigler and Goswami, 2005). In English, evidence points to developing readers needing to develop sensitivity to multiple levels of grain size, since the phoneme level provides inconsistent information due to poor correspondence between graphemes and phonemes in written English. In contrast, more reliable information regarding the pronunciation of English words can be obtained at the level of rimes, with evidence showing that children begin to develop a special sensitivity to this larger grained level (Treiman, Mullenix, Bijeljac-Babic, and Richmond-Welty, 1995). For this reason, skilled English readers develop sensitivity to multiple grain levels, with even young developing readers aged 7, 8, and 9, showing signs of using both larger units and smaller units simultaneously in phonological access in comparison to readers of a language such as German, where fine-grained units provide accurate information and the stimuli were controlled structurally to allow German readers access to higher level grain sizes if they needed it (Goswami, Ziegler, Dalton, and Schneider, 2003). Because German is comparable to Spanish in its degree of correspondence between graphemes and phonemes and beginning readers in both languages show high accuracy levels in word reading after the first year of literacy instruction (Seymour et al., 2003), arguably Spanish and German pattern similarly in contrast to English with respect to the grain size used by developing readers. In general, readers of English and Spanish appear to develop phonological representations along somewhat different trajectories due to the differing natures of the two languages and the varying correspondence between the languages and the scripts with which they are written. English readers develop phonological access through larger units such as rimes, as well as working on acquiring sensitivity to fine grained units such as phonemes. In contrast, Spanish readers are more sensitive to the fine-grained level of phonemes, since the script has good sound/letter correspondence. In addition, they develop greater sensitivity to syllables, due to the clear quality of syllables in Spanish and the fact that the traditional teaching of literacy in Spanish emphasizes syllables (the syllable segmentation and blending method) (Denton, Hasbrouck, Weaver, and Riccio, 2000). This emphasis on syllables in teaching seemingly calls attention to this level of the prosodic hierarchy early on in reading development, with some research evidence suggesting that Spanish readers aged 8 to 13 years may already have developed some sensitivity to syllables (Jiménez and Rodrigo, 1994). The goal of the present research is to compare visual word access in developing bilingual readers to monolinguals to see if the two subject populations differ in their lexical access, using English and Spanish as the comparison languages. The differences in phonological structure between the two languages should cause readers to lexically activate words in different manners. The unit of lexical access should differ between Spanish and English, with readers in Spanish showing a preference for the syllable and readers in English accessing words through a larger unit in the prosodic hierarchy. In addition, bilingual readers should

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Visual Word Access in Monolinguals and Bilinguals in English and Spanish

127

demonstrate adaptation to the languages they are reading in by showing a similar pattern of word access to monolinguals.

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Method Subjects: Eighty-five 11 to 15 year old subjects participated in this experiment in 3 different categories: English-Spanish bilingual, English monolingual and Spanish monolingual. The twenty-six bilingual English-Spanish readers were recruited at a dual language immersion elementary school and a regular elementary school with a large bilingual population both in Chicago. They were paid $10 for the hour-long experiment. Two participants were excluded from analysis because their standardized reading comprehension scores fell two standard deviations below the norm in either English or Spanish, leaving twenty-four participants. The sample included 10 boys and 14 girls. The participants ranged in age from 146 months to 180 with an average of 160.4. Twenty-nine monolingual English readers recruited at public and private elementary schools in the metropolitan Chicago area participated in the experiment. They were paid $5 for the half-hour long experiment. The standardized measures to determine reading comprehension level in English and Spanish (where appropriate) resulted in the exclusion of three participants from analysis because of standardized scores two standard deviations above the norm –meaning they had reading comprehension in English beyond the norm -leaving twenty-six participants. Two additional participants were unable to perform the experimental task, by not providing complete words in their answers, leaving 24 participants. The sample included 10 boys and 14 girls. The age of the participants ranged from 155 to 176 months, with an average of 166.1. Thirty monolingual Spanish readers recruited at a private middle school in Puebla, Mexico and at a dual language immersion public school in Chicago participated in the experiment. They were paid $5 for the half-hour long experiment. The standardized measures to determine reading comprehension level in Spanish and English resulted in the exclusion of two participants from analysis who tested as having too much reading knowledge of English –one within two standard deviations of the norm and the other testing greater than two standard deviations above the norm, leaving 28 participants. Four additional participants were unable to perform the experimental task, by not providing complete words in their answers, leaving 24 participants. The sample included 15 boys and 9 girls. The age of the participants ranged from 142 to 174 months, with an average of 159.3. Design and Materials: This study used a partial identity masked priming task, since this technique is designed to tap into early phonological assembly processes (Berent and Perfetti, 1995; Lukatela and Turvey, 2000; Perry and Ziegler, 2002). To the largest extent possible, equivalent materials in both languages for all test conditions were developed. All stimuli in both languages consisted of two syllable words phonologically fitting a CVCV schema, with stress on the first syllable –the most common pattern in Spanish and one that exists in English (albeit more rarely). Orthographically, 30 words in English fit the pattern CVCV and 30 fit the pattern CVCVC, due to exhausting all word candidates that fit the first orthographic pattern. All 60 Spanish stimuli fit the orthographic and phonological pattern CVCV. The orthographic difference between English and Spanish arose from the limited number of

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

128

John Evar Strid and James R. Booth

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

orthographic CVCV words in English with stress on the first syllable. Obvious cognates were excluded from the stimuli in both languages. (See Appendix 1 for a list of the stimuli used.) The stimuli were equated for their word frequency and bigram frequency using the CELEX database for English (Baayen, Piepenbrock, and Gulikers, 1995) and the Diccionario de frecuencias de las unidades lingüísticas del castellano [Dictionary of frequencies of the linguistic units of Spanish] (Alameda and Cuetos, 1995) for Spanish. The word sets showed no significant difference between the two languages for word frequency, and for token frequency (how often two letters appear in the same position in words in a corpus of the language) for the first two bigrams. The difference between the two languages for the last bigram’s token frequency follows from the fact that relatively few letters can appear in word final position in Spanish, with most words ending in either ‘a’ or ‘o’. For this reason, we will report analyses excluding the last bigram, examining only units of which it is not a part. Essentially this statistical difference between languages tells us that given the second consonant, that Spanish readers would be more likely to be able to guess the vowel that follows the second consonant. Additionally, all stimuli were controlled for the token transitional probabilities of letters (the probability that one letter appears after another in words in a corpus), with no significant difference between languages. Within both languages, there was a significant difference between the transitional probability of the first two letters and the third/fourth letters in both word sets, but no significant difference existed between languages in either word set for this difference. Procedures: For bilingual participants, testing took place either in two separate sessions of a half hour each during recess in the middle of the school day (on days as close to each other as possible), or during a session when participants made a special trip into school on a non-attendance day or after school, completing both languages at the same time. For monolingual participants, the half hour testing took place either in a session during recess in the middle of the school day or when participants were pulled out of class. The investigators worked with school authorities to ensure that all testing was held in all locales in a dim room with shades, with similar equipment placement. The subjects at the different schools did not differ statistically in age. For bilingual subjects, whether the English and Spanish tests were given on one day or two resulted in no statistical difference in overall accuracy rates. For bilingual subjects, the total time of testing lasted about an hour for each subject. Participants were randomly assigned to whether they completed the English or the Spanish part of the testing first. When they were in either the Spanish or the English condition, the bilingual experimenter spoke to them in only that language. Before beginning the experiment in each language, the participants took standardized tests in both languages to determine their reading level. In English, the Passage Comprehension (PC) sub-test from Woodcock Reading Mastery Test was used, while in Spanish the equivalent sub-test from the Batería WoodcockMuñoz was selected (Woodcock, 1987; Woodcock and Muñoz, 1995). Monolingual subjects only took the standardized test in the appropriate language to determine their reading level and completed the experiment in the appropriate language. If they had had some exposure to the other language, they took the standardized test in the other language after completing the experimental procedure. Participants viewed all stimuli on a SVGA monitor in a darkened room, with shaded windows providing background ambient lighting. The Apple computer used Psyscope

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Visual Word Access in Monolinguals and Bilinguals in English and Spanish

129

software designed to control the priming experiment (Cohen, MacWhinney, Flatt, and Provost, 1993). Participants recorded their responses by writing on a test sheet. Participants saw different primes that corresponded to the first three letters of the word (CVC#), the first two letters (CV##), the first vowel and second consonant (#VC#), or, in the control condition, a series of number signs. These primes were chosen to examine the effect of priming the syllable, priming across the syllable boundary, and priming a unit larger than the syllable. The primes and the targets matched up in the manner demonstrated in Table 1. Four different counterbalancing conditions varied the primes that were shown with different words. Balanced presentation of stimuli ensured that each participant saw each condition equally often, and that each item was preceded by each prime type equally often. Participants were assigned randomly to counterbalancing conditions and viewed stimuli within each counterbalancing condition in a randomized order of presentation. Participants viewed primes and targets on a standard black background with white letters. The participants first saw a fixation cross on the computer screen before each trial, and were asked to press an external button box to begin each trial. When they triggered the trial, they then saw the prime in upper case letters and the target in lower case letters flash on the computer screen for 32 ms each, followed by a series of number signs for an additional 500ms. The presentation of lower and upper case letters helped ensure that the task was not simply visual priming, but instead accessing abstract letter representations. The rate of presentation ensured interpretable accuracy levels (Berent and Perfetti, 1995; Lukatela and Turvey, 2000; Perry and Ziegler, 2002). Previous findings have held that less than 40% accuracy for word identification precludes strategic processing with this technique (Xu and Perfetti, 1999). Before beginning each part of the experiment, participants completed a practice set of four examples. The subject read instructions presented on the computer screen and the experimenter discussed them with the subject to ensure understanding before beginning the practice set. Participants were instructed to give their best guess even if they were uncertain of what they had seen. They were reminded of the importance of writing words with the appropriate number of letters for all trials. After completing the experiment, the researcher asked them questions about their language background, recording their answers on a Language Background Questionnaire. Table 1. Experimental Conditions Status Test Conditions

Control

Prime SOF# SO## #OF# ####

Target Sofa Sofa Sofa sofa

Results Subjects: The standardized reading tests show that the bilingual participants who participated in this experiment were overall more proficient readers in English. The mean Standardized Score (SS) of the PC subtest in English was 108.1 (SD = 13.2), while in Spanish

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

130

John Evar Strid and James R. Booth

the mean was 98.5, (SD = 13.9). The SS on this test is based around 100, with half the participants in a particular age range scoring either higher or lower during norming. The SD of the sample population during norming was 15 points. To ensure that subjects were truly bilingual in terms of literacy - i.e. that they were proficient readers in both languages -we defined language literary dominance according to specific criteria. Using the SD value from the normed test, English or Spanish dominants had scores greater than 2/3 of a SD on PC in their dominant language, while balanced bilinguals had scores that differed by less than 2/3 of a SD between the two languages. According to this analysis, 11 participants were more proficient readers in English, 10 had balanced literacy skills and 3 comprehended what they read in Spanish better. The participants’ English dominance in terms of literacy was confirmed by matched pair t-tests comparing their scores in the two languages, with a significant difference in scores for the PC subtest (t(23) = 3.137, p = .005). For participants with balanced literacy skills, their English scores were still significantly greater than their Spanish, although not as marked (t(9) =2.36, p= .043). Some results from the Language Background Questionnaire also suggested that bilingual subjects had superior English literacy skills. From the questionnaire, 19 participants’ first language was Spanish, while 3 learned English first and 2 reported learning both languages from their earliest memory. The average age when participants began to learn their second language was 5.3 years old. Most participants had spent far more time in an English speaking country than in a Spanish speaking one (10.6 vs. 2.5 years). When asked to rate their overall proficiency in the two languages on a scale from 1 to 6, participants’ ranked themselves slightly higher in Spanish (5.0) than in English (4.9) –the only self rating question focusing on oral speech. When asked about their reading experience in the two languages on a 1-6 scale, participants stated that they had read more in English than in Spanish (4.9 vs. 3.9). Finally, participants stated that they spoke more English during a typical day than Spanish (57.5% of the day vs. 42.5%). Monolingual participants in their respective languages had similar reading scores as the bilingual subjects on the standardized measure. The mean SS on the PC subtest for the English monolingual participants was 107.7, (SD = 8.9). Only five of the twenty-four participants had had any exposure to Spanish and completed the PC subtest in Spanish; all of the Spanish scores were more than one and a half standard deviations below their English score and more than one and a half standard deviations below the normal score of 100. The mean score in Spanish for these participants was 72 (SD = 4.4). The mean SS on the PC subtest for the Spanish monolingual participants was 103.2 (SD = 8.3). All of the twenty-four participants had had some exposure to English; for this reason all completed the PC subtest in English; all of their English scores were more than one and a half standard deviations below their Spanish score and more than one and a half standard deviations below the normal score of 100. The mean score in English was 54.5 (SD = 13.7). The English PC SS of the English monolingual participants was not significantly greater than the Spanish score of the Spanish monolingual participants, (t(23) = 1.81, p = .078). The Language Background Questionnaire yielded similar results. Fourteen of the English monolingual participants’ reported that their first language was English, while 3 learned Spanish first, 2 reported learning both languages at the same time, 2 learned Gujarati, 1 each learned Malayalan, Korean, and Punjabi. The average age when English participants learned their second language was 6.2 years old, while 4 reported never learning another language. All English monolinguals participants had spent far more time in an English speaking country

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Visual Word Access in Monolinguals and Bilinguals in English and Spanish

131

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

than in a Spanish speaking one (13.0 vs. 0.4 years). Also, English participants’ self-rating of their language proficiency in English on a six point scale was higher than in Spanish (5.1 vs. 2.4). Participants reported reading far more in English than in Spanish on a six point scale (5.6 vs. 1.7). Finally, English participants reported using English 84.9% of the time in comparison to only 9.6% for Spanish. For the Spanish monolingual participants, all 24 reported that their first language was Spanish. The mean age when the Spanish participants started learning English was 6.5 years. The Spanish participants reported that they had spent an average of 12.5 years in a Spanish speaking country, while they (3 subjects) only spent 0.2 years in an English speaking country. They rated their overall Spanish proficiency at 4.9 on a six point scale and their English proficiency at 2.8. When asked to rate how much they had read in Spanish on a six point scale, they gave themselves 4.8, while in English 2.2. They reported spending 94.5% of a typical day using Spanish, and 5.4% of the day using English. The standardized test results for the monolingual subjects did not differ statistically from those of the bilingual subjects in either language. In English, the mean PC standardized score for monolingual subjects was 107.7 (SD = 8.9), in comparison to an average score of 108.1 (SD = 13.2) for bilinguals (t(46) = .13, p = .90), a statistically non-significant difference (t(46) =-.13, p = .9).Spanish monolingual readers had a mean PC standardized score of 103.2 (SD = 8.3), while bilinguals had a mean of 98.5 (SD = 13.9), a difference that was also not statistically different (t(46) =1.43, p = .16). Priming Task: For each language and prime type accuracy in identifying first consonant/first vowel (CV), and the first consonant/first vowel/second consonant (CVC), was calculated and averaged by subjects and items. We chose to examine recognition of the CV because it corresponds to the syllable; if word activation proceeds via syllables in Spanish, significant proportions of subjects should identify only this unit. Similarly, English readers should identify larger units, like the CVC.1 We analyzed independent identification of both units. Independent identification means that we are defining identification of CV, for example, as when only those letters together were correctly recognized, without being a part of any larger unit such as CVC or word. While we report subject data for bilingual subjects in Table 2 and for monolingual subjects in Table 3, we report statistically significant findings for both subject and item analyses, using 2 x 2 x 2 ANOVAs to examine bilinguals and monolinguals (First consonant/first vowel (CV) primed vs. unprimed x First vowel/second consonant (VC) primed vs. unprimed x English vs. Spanish) examined accuracy of identification of CV, and CVC units, with prime type as a within subject and item variable and with language as a within subject factor and a between item factor for bilinguals and as a between subject and item factor for monolinguals. For bilingual status, 2 x 2 x 2 ANOVAs for each language (First consonant/first vowel (CV) primed vs. unprimed x First vowel/second consonant (VC) primed vs. unprimed x Monolingual vs. bilingual) examined accuracy of identification of CV, and CVC units, with bilingual status as a between subjects and items factor. 1

While the CVC in the English stimuli corresponds to the BOSS proposed by Taft (1979), our purpose is not to examine or to argue theoretically for word activation specifically through the BOSS for English. While word activation through the BOSS unit is certainly possible, we are instead simply arguing for activation through a larger unit in English, and the CVC is the only larger unit in our stimuli short of the word.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

132

John Evar Strid and James R. Booth Table 2. Accurate responses according to prime types for Bilingual Children (with standard error)

Language English

Spanish

Prime CVC# CV## #VC# #### CVC# CV## #VC# ####

CV 15.14(2.09) 56.03(3.98) 1.64(0.59) 5.10(1.56) 12.61(1.40) 55.28(2.92) 2.49(0.88) 6.25(1.48)

CVC 31.51(3.53) 5.69(1.40) 5.65(1.35) 0.78(0.78) 20.46(2.42) 5.02(1.10) 4.87(1.09) 1.04(0.61)

Table 3. Accurate responses according to prime type for Monolingual Children (with standard error) Language English

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Spanish

Prime CVC# CV## #VC# #### CVC# CV## #VC# ####

CV 11.16(2.22) 51.38(3.82) 1.93(0.85) 3.72(1.10) 11.20(1.93) 46.09(2.91) 2.94(0.96) 7.78(1.53)

CVC 31.29(2.86) 8.67(1.49) 2.27(0.80) .60(0.41) 12.35(1.83) 5.80(1.14) 3.53(0.94) .89(0.65)

For bilingual subjects, the priming results are demonstrated in Figure 1, which presents identification rates of CVs and CVCs based on different prime types. For CV identification, the prime type main effects were significant. The main effect of priming CV was significant according to subjects (F(1,23) = 296.83, p < .001) and items (F(1,118) = 471.36, p < .001). The main effect of priming VC was also significant according to subjects (F(1,23) = 172.48, p < .001) and items (F(1,118) = 323.93, p < .001). The interaction between CV primes and VC primes was significant according to subjects (F(1,23) = 120.77, p < .001) and items (F(1,118) = 281.06, p < .001), with CV## primes (and to a much lesser extent CVC# primes) resulting in greater recognition of CV in comparison to the other primes. However, no interactions between language and prime type were significant. Turning to identification of CVC, the main effect of CV primes was significant according to subjects (F(1,23) = 96.64, p < .001) and items (F(1,118) = 98.4, p < .001). According to both subjects (F(1,23) = 106.61, p < .001) and items (F(1,118) = 88.35, p < .001), CVC identification was significantly aided by VC primes. The interaction between CV Primes and VC Primes was significant according to subjects (F(1,23) = 60.58, p < .001) and items (F(1,118) = 60.08, p < .001), with CVC# resulting in greater recognition of CVC than the other primes.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Visual Word Access in Monolinguals and Bilinguals in English and Spanish

133

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Figure 1. Accuracy of identification of CV and CVC (+SE) according to prime type in English and Spanish for Bilingual Children.

Next, the interaction between CV Primes and Language was significant for both subjects (F(1,23) = 5.22, p = .032) and items (F(1,118) = 4.97, p = .028), with greater priming effect of CV primes in English. Also the interaction between VC Primes and Language was significant according to both subjects (F(1,23) = 7.16, p = .013) and items (F(1,118) = 4.46, p = .037), with greater priming effect of CV primes in English. Finally, the three way interaction between CV primes, VC primes and Language approaches significance according to subjects (F(1,23) = 4.89, p = .073) and is significant according to items (F(1,118) = 5.16, p = .025. Paired sample t-tests for subjects and independent sample t-tests for items further tested the significant interactions between the main effects and looked at the effect of the different prime types on CVC identification. According to the t-tests, CVC# primes resulted in significantly greater CVC recognition in English according to subjects (t(23) = 2.69, p = .007) and items (t(118) = 2.39, p = .009). To test if dominance affected performance, we contrasted the balanced bilinguals and the English dominant subjects, entering the results into a 2 x 2 x 2 x 2 ANOVA (CV-prime x VCprime x Language x Dominance). The three Spanish dominant participants were excluded from this analysis, to create two clear categories. No main effects of dominance were significant, nor were any interactions between dominance and any other main effects. Turning to the monolingual subjects’ priming results shown in Figure 2, the prime type main effects for CV identification were significant. The main effect of priming CV was significant according to subjects (F(1,46) = 225.88, p < .001) and items (F(1,118) = 401.3, p < .001). The main effect of VC primes was also significant according to subjects (F(1,46) = 197.05, p < .001) and items (F(1,118) = 225.72, p < .001). The interaction between CV Primes and VC Primes was significant according to subjects (F(1,46) = 171.44, p < .001) and items (F(1,118) = 179.15, p < .001), with CV## primes resulting in greater recognition of CV in comparison to the other primes.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

134

John Evar Strid and James R. Booth

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Figure 2. Accuracy of identification of CV and CVC (+SE) according to prime type in English and Spanish for Monolingual Children.

Finally, the interaction between CV primes and language was significant according to items (F(1,118) = 4.39, p = .038) but not according to subjects (F(1,46) = 2.24, p = .141) – with CV Primes leading to slightly higher identification rates of syllables in English in comparison to Spanish. In order to better understand the interaction between CV primes and language and to examine the effect of individual primes, we ran paired sample t-tests for subjects and independent sample t-tests for items. The t-tests show that #### primes result in significantly greater accuracy of identification of syllables in Spanish according to both subjects (t(46) = 2.15, p = .019) and items (t(118) = 2.08, p = .02). In contrast, CVC# primes, CV## primes and #VC# primes did not result in significantly different CV identification rates in the two languages. Turning to identification of CVC, all main effects and all interactions were significant. The main effect of CV primes significantly affected accuracy in identifying CVC according to subjects (F(1,46) = 139.39, p < .001) and items (F(1,118) = 80.99, p < .001). The main effect of VC primes was also significant according to subjects (F(1,46) = 93.07, p < .001) and items ((F(1,118) = 63.18, p < .001). The main effect of language was also significant according to subjects (F(1,46) = 18.88, p < .001) and items (F(1,118) = 10.35, p = .002). The interaction between CV primes and language was significant according to both subjects (F(1,46) = 29.58, p < .001) and items (F(1,118) = 16.88, p < .001). Additionally, the interaction between VC primes and language was significant according to subjects (F(1,46) = 18,94, p < .001 and items (F(1,118) = 12.38, p = .001). Also, the interaction between CV primes and VC primes was significant according to subjects (F(1,46) = 38.16, p < .001) and items (F(1,118 = 41.38, p < .001), with greater identification of CVC with CVC# primes. Finally, the three way interaction between CV primes, VC primes, and language was significant according to subjects (F(1,46) = 17.94, p < .001) and items (F(1,118) = 19.64, p < .001).

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Visual Word Access in Monolinguals and Bilinguals in English and Spanish

135

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

In order to better understand the interactions between prime types and language and to examine the effect of individual primes, we ran independent sample t-tests for subjects and for items. The t-tests show that CVC# primes result in significantly greater accuracy of identification of CVC in English according to both subjects (t(46) = 5.58, p < .001) and items (t(118) = 4.43, p = .001). Once again, this finding supports the hypothesis that English readers activate a larger unit. In contrast, #VC# primes, CV## primes and #### primes did not result in statistically different CVC identification rates between the languages. Turning now to the effect of bilingual status on the priming results shown in Figure 3, in English none of the interactions between prime types and bilingual status were significant for the different recognition units under consideration: CV and CVC. In contrast, in Spanish, many of the bilingual status interactions were significant for the different units. For example, for CV identification, the interaction between CV primes and bilingual status was significant according to subjects (F(1,46) = 4.98, p = .031) and items (F(1,118) = 4.94, p = .028. This interaction between CV primes and bilingual status was further examined with independent sample t-tests comparing CV recognition. Only CV## primes resulted in a significant difference according to subjects (t(46) = 2.23, p = .016) and items (t(118) = 1.80, p = .03), with bilingual readers identifying more CV units than monolingual readers. Turning to Spanish readers’ identification of CVC, the interaction between VC primes and bilingual status was significant according to subjects (F(1,46) =11.31, p = .002 and items (F(1,118) = 4.27, p = .041. Also, the three way interaction between CV primes, VC primes and bilingual status was significant according to subjects (F(1,46) = 4.72, p = .035) and items (F(1,118) = 4.79, p = .031). Further examining the interactions with independent sample ttests showed that CVC# primes resulted in significantly greater CVC identification for bilingual readers according to subjects (t(46) = 2.67, p = .005) and items (t(118) = 2.06, p = .021).

Figure 3. Accuracy of identification of CV and CVC (+SE) according to prime type in Spanish for Monolingual and Bilingual Children. Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

136

John Evar Strid and James R. Booth

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Discussion This experiment furthers the case that readers of Spanish and English depend on different levels of the prosodic hierarchy in visual recognition of words. These results provide evidence that visual word access in Spanish and English makes use of different units, suggesting that readers in English activate some larger unit in early word recognition during reading while providing weak evidence that Spanish readers may activate words via syllables. The present experiments’ results support previous suggestions that English readers use some unit larger than the syllable, consistent with previous claims that English readers access words via larger grained units (Butterworth, 1983; Chialant and Caramazza, 1995; Cole, et al., 2000, Lukatela, et al., 1980; Taft, et al., 2007; Taft and Forster, 1975, 1976; Taft, 1979, 1994; Ziegler and Goswami, 2005, 2006). In our study, both bilingual and monolingual readers recognized significantly more CVC units in English (as compared to Spanish), given CVC# primes. In fact, monolingual English readers and bilingual English readers did not differ from one another in CVC recognition given CVC# primes. In addition, bilinguals reading Spanish recognized significantly more CVC units in Spanish than monolinguals, suggesting that they were accessing words in a similar fashion in both languages and in a manner comparable to English monolinguals –a claim that will be discussed in further depth below. In comparison, few results suggested greater use of the syllabic level of the prosodic hierarchy in Spanish. When comparing languages, Spanish monolingual readers only differed in CV recognition from English monolinguals given #### primes, recognizing more CV units. While this finding provides scant evidence that Spanish readers rely more on the syllabic level of the prosodic hierarchy, being in the baseline condition, it nonetheless suggests that this level of the prosodic hierarchy may have some importance in Spanish for developing readers. When examining bilingual status, Spanish bilingual readers recognized more CV units than Spanish monolinguals given CV## primes. This finding comparing bilingual to monolingual readers runs contrary to expectation, because one would predict more fully developed syllabic representations and activation for Spanish monolinguals in comparison to bilinguals. This lack of evidence of syllabic activation is surprising (especially for monolinguals) in light of the large body of previously mentioned research that has shown that the syllable is a key unit of lexical access in Spanish for adult monolingual readers (Álvarez, et al., 1998; Álvarez, et al., 2000; Álvarez, et al., 2001; Álvarez, et al., 2004; Carreiras, et al., 1993; Carreiras and Perea, 2002; Domínguez, et al., 1993; Perea and Carreiras, 1998). However, most of the research demonstrating the importance of the syllabic level of the prosodic hierarchy in word activation in Spanish was done with adults with fully mature reading skills. Young Spanish monolinguals may have yet to develop fully the strong syllabic representations found in adult readers of Spanish. Relatively little research has been done that clearly demonstrates syllabic sensitivity in developing Spanish readers. While Jiménez and Rodrigo (1994)’s finding that developing 8-13 year old Spanish readers react to words with higher frequency syllables more slowly than those with lower frequency syllables in a lexical decision task suggests early syllable sensitivity, this was not the primary focus of their study and little additional research has been done with developing readers of Spanish to fully understand the time course, evolution and sequence of the growth in sensitivity to this level of the prosodic hierarchy.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Visual Word Access in Monolinguals and Bilinguals in English and Spanish

137

That children’s sensitivity to language appropriate phonological units is still developing is well documented in the literature, generally moving from larger to smaller units (Anthony and Lonigan, 2004; Anthony et al., 2002; Stanovich, 1992; Treiman and Zukowski, 1996). Because of this developing phonological sensitivity, younger readers may have less developed representation of the appropriate unit of visual word access and they may have not yet acquired the robust syllable activation found in adult Spanish word reading. While the present experiments did not completely control for the effect of teaching techniques, both experimenter observation and teacher reports suggested that both the monolingual and bilingual Spanish developing readers had been taught literacy using the traditional syllable segmentation and blending method.2 At this stage, young readers may still be working on acquiring activating words through this level of the prosodic hierarchy. Additionally, bilingual readers may have been affected in their mode of word activation by exposure to English, particularly since they were more experienced readers of English. The weak evidence for syllable activation in Spanish in the present experiment contrasts with the results of our previous experiments with adult subjects, in which adult bilinguals, reading in both English and Spanish, activated significantly more CV units in Spanish compared to English in response to both C### and CV## primes in an experiment that primed within the syllable boundary, as well as when primed with #V## in an experiment that primed across the syllable boundary (Strid and Booth, 2007). Therefore, the logical interpretation of the varying use of syllables in adult and children readers is the difference in their experience and expertise in reading. Specifically, the difference between adults and children argues that the children are still developing their language representations, whereas the adult bilinguals are more fully adapted to the appropriate phonological level of activation for both of their languages. Turning to the effect of bilingualism on reading, the  bilinguals  reading in Spanish showed evidence of being influenced by exposure to English. Comparing bilinguals’ reading in Spanish to monolinguals, shows that the bilinguals activated more CVC units than monolinguals given CVC# primes, the same as what they recognized in English and what English monolinguals identified given the same primes. This finding suggests that the bilinguals must have been affected by their exposure to a language with an opaque script. In contrast, in the present study, bilinguals reading in English did not differ statistically in activation units from monolinguals. Bilingual exposure to Spanish did not appear to affect their English reading. So while bilinguals may adapt to the language that they are reading (e.g., Shimron and Sivan, 1994; Oren and Breznitz, 2005; Strid and Booth, 2007), their direct access to phonology may be disrupted by exposure to a script that has poor sound-letter correspondence. The suggestion that bilingual readers are affected by exposure to a language with poor sound/letter correspondence or by language dominance, agrees with previous research that demonstrated that bilingual readers from a first language background where the script contains less phonological information are less dependent on phonology while reading in a

2

In an observation of literacy training in Spanish at the dual language immersion school that provided over 60% of the bilingual subjects, the teacher was using the syllable segmentation and blending method. Also, when asked about the literacy training of the Spanish monolingual subjects, the lead teacher of the Spanish monolingual subjects indicated that this method had also been used with them (J.D.D. Peña-Ortega, personal communication, July 24, 2006).

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

138

John Evar Strid and James R. Booth

second language in which the script offers greater phonological information (Koda, 1989; Saito, et al., 1979; Wang, et al., 2003: Wang, et al., 2003). Finally, our finding suggesting that bilinguals activated words through the processes of their dominant language agrees with research showing bilinguals activated words in their L2 using the same neural networks as in their L1 (Tan et al., 2003). In conclusion, these experiments investigated the effect on being bilingual on lexical access in two languages, testing whether young Spanish and English monolingual and bilingual readers differed in the units of activation in visual word access. The results suggest that lexical access in English is based on a larger phonological sub-lexical unit because of greater report of a unit bigger than the syllable, whereas there was little evidence that Spanish lexical access was based on the syllable in children. Finally, monolingual or bilingual status of the reader did not have influence on English lexical access; however, Spanish bilinguals were influenced by the acquisition of English, suggesting that orthographically opaque languages can have an effect on transparent languages or that language dominance may be an important factor.

Acknowledgments

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

The authors would especially like to thank Janet Pierrehumbert and Matt Goldrick for their many valuable suggestions and contributions. Also this research would not have been possible without the support of staff and students at UPAEP (Puebla, Mexico), InterAmerican Magnet School (Chicago, Il.), Sawyer School (Chicago, Il.), Waldorf School (Chicago, Il.), and CCSD 93 (Carol Stream, Il.). This research was partially supported by  a  Graduate  Research  Grant  from  the  University  Research  Grants  Committee  at  Northwestern  University. 

Appendix: Test Stimuli English Test Words Group 1: bevy gory nosy tuna wavy yoga memo hazy lily duly pony posy wary

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Visual Word Access in Monolinguals and Bilinguals in English and Spanish

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

rosy tidy lazy zero navy sofa vary pity copy holy busy duty tiny lady baby body many Group 2: savor cedar diner giver lager mover donor meter rotor rover sever maker poker finer lever razor ruler sober safer liver baker vicar lover fever sugar cover river

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

139

140

John Evar Strid and James R. Booth paper later water

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Spanish Test Words Group 1: bono gafe leño jota doma bobo beca foro cano vilo coco loro reja mona bala dale rota paja roca tomo vela hoja raro loco dedo luna baja boca cara cada Group 2: jaco gula faja lego faro feto

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Visual Word Access in Monolinguals and Bilinguals in English and Spanish

141

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

duna nata rima foca pato mole pala lodo mago mora foco pera lata sano codo lobo mono raza rico rato mala sola bajo modo

References Alameda, J. R., and Cuetos, F. (Eds.). (1995). Diccionario de frecuencias de las unidades linguísticas del castellano [Dictionary of frequencies of the linguistic units of Spanish]. Oviedo: Servicio de Publicaciones de la Universidad de Oviedo. Álvarez, C., Carreiras, M., and de Vega, M. (2000). Syllable-frequency effect in visual word recognition: Evidence of sequential-type processing. Psicologica, 21(3), 341-374. Álvarez, C. Carreiras, M., and Perea, M. (2004). Are syllables phonological units in visual word recognition? Language and Cognitive Processes, 19(3), 427-452. Álvarez, C. J., Carreiras, M., and Taft, M. (2001). Syllables and morphemes: Contrasting frequency effects in Spanish. Journal of Experimental Psychology: Learning, Memory, and Cognition, 27(2), 545-555. Álvarez, C., de Vega, M., and Carreiras, M. (1998). The syllable as an activational unit in reading trisyllabic words. Psicothema, 10(2), 371-386. Anthony, J. L., and Lonigan, C. J. (2004). The nature of phonological awareness: Converging evidence from four studies of preschool and early grade school children. Journal of Educational Psychology, 96, 43–55. Anthony, J. L., Lonigan, C. J., Burgess, S. R., Driscoll, K., Phillips, B. M., and Cantor, B. G. (2002). Structure of preschool phonological sensitivity: Overlapping sensitivity to rhyme, words, syllables, and phonemes. Journal of Experimental Child Psychology, 82, 65–92.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

142

John Evar Strid and James R. Booth

Ashby, J., and Rayner, K. (2004). Representing syllable information during silent reading: Evidence from eye movements. Language and Cognitive Processes, 19(3), 391-426. Baayen, R. H., Piepenbrock, R., and Gulikers, L. (1995). The CELEX lexical database (version release 2) [CD-ROM]. Philadelphia, PA: Linguistic Data Consortium, University of Pennsylvania. Ben-Zeev, S. (1977). The influence of bilingualism on cognitive strategy and cognitive development. Child Development. 48(3), 1009-1018. Berent, I., and Perfetti, C. (1995). A rose is a reez: The two-cycles model of phonology assembly in reading English. Psychological Review, 102(1), 146-184. Bialystok, E. (1988). Levels of bilingualism and levels of linguistic awareness. Developmental Psychology. 24(4), 560-567. Bialystok, E., Majumder, S., and Martin, M. (2003). Developing phonological awareness: Is there a bilingual advantage? Applied Psycholinguistics. 24(1), 27-44. Blevins, J. (1995). The syllable in phonological theory. In J. Goldsmith (Ed.), The handbook of phonological theory. Cambridge, Ma: Blackwell, pp. 206-244. Branum-Martin, L., Mehta, P. D., Fletcher, J. M., Carlson, C. D., Ortiz, A., Carlo, M., et al. (2006). Bilingual phonological awareness: Multilevel construct validation among Spanish-speaking kindergarteners in transitional bilingual education classrooms. Journal of Educational Psychology, 98(1), 170-181. Brown, G. D. A., and Deavers, R. P. (1999). Units of analysis in nonword reading: Evidence from children and adults. Journal of Experimental Child Psychology, 73, 208–242. Bruck, M., and F. Genesee (1995). Phonological awareness in young second language learners. Journal of Child Language. 22(2), 307-324. Bruck, M., Genesee, F. and Caravolas, M. (1997). A cross-linguistic study of early literacy acquisition. In B. A. Blachman (Ed.), Foundations of reading acquisition and dyslexia: Implications for early intervention, Mahwah, NJ: Erlbaum, pp. 145–162. Butterworth, B., (1983). “Lexical representation.” In B. Butterworth (Ed.), Language Production, Vol. 1, London: Academic Press. Carreiras, M., Álvarez, C., and de Vega, M. (1993) Syllable frequency and visual word recognition in Spanish. Journal of Memory and Language, 32(6), 766-780. Carreiras, M., and Perea, M. (2002). Masked priming effects with syllabic neighbors in a lexical decision task. Journal of Experimental Psychology: Human Perception and Performance, 28(5), 1228-1242. Chialant, D., and Caramazza, A. (1995). Where is morphology and how is it processed? The case of written word recognition. In L. Feldman, (Ed.), Morphological aspects of language processing. Hillsdale, NJ: Lawrence Erlbaum. Cohen, J. D., MacWhinney, B., Flatt, M., and Provost, J. (1993). “PsyScope: An interactive graphic system for designing and controlling experiments in the psychology laboratory using Macintosh computers.” Behavior Research Methods, Instruments, and Computers, 25, 257-271. Cole, P., Beauvillain, C., and Seguí, J. (1989). On the representation and processing of prefixed and suffixed derived words: A differential frequency effect. Journal of Memory and Language, 28(1), 1-13. Cummins, J. (1979). Linguistic interdependence and the educational development of bilingual children. Review of Educational Research, 49, 225-251.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Visual Word Access in Monolinguals and Bilinguals in English and Spanish

143

Cummins, J. (1984). Language Proficiency, Bilingualism, and Achievement. In J. Cummins, Bilingualism and Special Education: Issues in Assessment and Pedagogy. San Diego, Ca: College Hill. Denton, C., Hasbrouck, J., Weaver, L., and Riccio, C. (2000). What do we know about phonological awareness in Spanish? Reading Psychology, 21, 335–352. D'Angiulli, A., Siegel, L. S. and Serra, E. (2001). The development of reading in English and Italian in bilingual children. Applied Psycholinguistics, 22(4), 479-507. Domínguez, A., de Vega, M., and Cuetos, F. (1997). Lexical inhibition from syllabic units in Spanish visual word recognition. Language and Cognitive Processes, 12(4), 401-422. Ellis, N. C., and Hooper, A. M., (2001). Why learning to read is easier in Welsh than in English: Orthographic transparency effects evinced with frequency-matched tests. Applied Psycholinguistics, 22, 571–599. Favreau, M., Komoda, M., and Segalowitz, N. (1980). Second language reading: Implications of the word superiority effect in skilled bilinguals. Canadian Journal of Psychology, 34(4), 370-380. Favreau, M., and Segalowitz, N. (1983). Automatic and controlled processes in the first- and second-language reading of fluent bilinguals. Memory and Cognition, 11(6), 565-574. Ferrand, L. Seguí, J., and Humphreys, G. (1997). The syllable's role in word naming. Memory and Cognition, 25(4), 458-470. Frith, U., Wimmer, H., and Landerl, K., (1998). Differences in phonological recoding in German- and English-speaking children. Scientific Studies of Reading, 2, 31–54. Frost, R., Katz, L., and Bentin, S., (1987). Strategies for visual word recognition and orthographic depth: a Multilingual Comparison. Journal of Experimental Psychology: Human Perception and Performance, 13, 104-115. Galambos, S. J., and S. Goldin-Meadow (1990). The effects of learning two languages on levels of metalinguistic awareness. Cognition. 34(1), 1-56. Goswami, U. (1988). Orthographic analogies and reading development. Quarterly Journal of Experimental Psychology, 40A, 239-268. Goswami, U., and Bryant, P.E. (1990). Phonological skills and learning to read. Hillsdale, NJ: Erlbaum. Goswami, U., Gombert, J. E., and de Barrera, L. F., (1998). Children’s orthographic representations and linguistic transparency: Nonsense word reading in English, French, and Spanish. Applied Psycholinguistics, 19, 19–52. Goswami, U., Ziegler, J. C., Dalton, L., and Schneider, W. (2001). Pseudohomophone effects and phonological recoding procedures in reading development in English and German. Journal of Memory and Language, 45, 648–664. Goswami, U., Ziegler, J. C., Dalton, L., and Schneider, W. (2003). Nonword reading across orthographies: How flexible is the choice of reading units? Applied Psycholinguistics, 24, 235–247. Harris, J. (1983). Syllable Structure and Stress in Spanish: a Nonlinear Analysis. Cambridge, Ma: MIT Press. Jared, D., and Seidenberg M. (1990). Naming multisyllabic words. Journal of Experimental Psychology: Human Perception and Performance, 16, 92-105. Jensen, J. (2000). Against Ambisyllabicity. Phonology, 17, 187-235.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

144

John Evar Strid and James R. Booth

Jiménez, J. E., and Rodrigo, M. (1994). Is it true that the differences in reading performance between students with and without LD cannot be explained by IQ? Journal of Learning Disabilities, 27, 155-163. Kahn, D. (1976). Syllable-based generalization in English phonology. Ph.D. dissertation, M.I.T. Kang, H., and Simpson, G., (1996). Development of semantic and phonological priming in a shallow orthography. Developmental Psychology, 32, 860-866. Katz,  L.,  and  Frost,  R.  (1992).  The  reading  process  is  different  for  different  orthographies:  the  Orthographic  Depth  Hypothesis.  In  R.  Frost  and  L.  Katz  (Eds.), Orthography, phonology, morphology and meaning. Amsterdam: Elsevier, 67-84. Koda, K. (1989). Effects of L1 orthographic representation on L2 phonological coding strategies. Journal of Psycholinguistic Research, 18(2), 201-222. Koda, K. (2007). Reading and Language Learning: Crosslinguistic Constraints on Second Language Reading Development. Language Learning 57:Suppl. 1, 1–44. Lindgren, S. D., de Renzi, E. and Richman, L. (1985). Cross-national comparisons of developmental dyslexia in Italy and the United States. Child Development, 56(6): 14041417. Lukatela, G., Gligorijevic, B., Kostic, A., and Turvey, M. (1980), Representation of inflected nouns in the internal lexicon. Memory and Cognition, 8, 415-423. Lukatela, G., and Turvey, M. (2000). An evaluation of the two-cycles model of phonology assembly. Journal of Memory and Language, 42(2), 186-207. Oren, R., and Breznitz, Z. (2005). Reading processes in L1 and L2 among dyslexic as compared to regular bilingual readers: Behavioral and electrophysiological evidence. Journal of Neurolinguistics, 18(2), 127-151. Nespor, M., and Vogel, I. (1986). Prosodic Phonology. Dordrecht: Foris. Paulesu, E., Demonet, J., Fazio, F., McCrory, E., Chanoine, V., Brunswick, N., et al. (2001). “Dyslexia: Cultural diversity and biological unity.” Science, 291, 2165–2167. Perea, M., and Carreiras, M. (1998). Effects of syllable frequency and syllable neighborhood frequency in visual word recognition. Journal of Experimental Psychology: Human Perception and Performance, 24(1), 134-144. Perry, C., and Zieglar, J. (2002). On the nature of phonological assembly: Evidence from backward masking. Language and Cognitive Processes, 17(1), 31-59. Rubach, J. (1996). Shortening and ambisyllabicity in English. Phonology, 13, 197-237. Sánchez-Casas, R. (1996). Lexical access in visual word recognition: The contribution of word form. In M. Carreiras, J. García-Albea and N. Sebastián-Gallés (Eds.), Language processing in Spanish (pp. 21-59). Mahwah, NJ: Lawrence Erlbaum Associates. Saito, H., Inoue, M., and Nomura, Y. (1979). Information processing of Kanji and Kana. Psychologia, 22, 195-206. Schiller, N. (2000). Single word production in English: The role of the subsyllabic units during phonological encoding. Journal of Experimental Psychology: Learning, Memory, and Cognition, 26(2). 512-528. Selkirk, E. (1982). The syllable. In H. van der Hulst and N. Smith (Eds.), The structure of phonological representations, Part 2 (pp. 337-383). Dordrecht: Foris. Seymour, P. H. K., Aro, M., and Erskine, J. M. (2003). Foundation literacy acquisition in European orthographies. British Journal of Psychology, 94, 143–174.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Visual Word Access in Monolinguals and Bilinguals in English and Spanish

145

Shimron, J., and Sivan, T. (1994). Reading proficiency and orthography: Evidence from Hebrew and English. Language Learning, 44(1), 5-27. Stanovich, K.E., (1992). Speculations on the causes and consequences of individual differences in early reading acquisition. In P.B. Gough, L. C. Ehri, and R. Treiman, (Eds.), Reading Acquisition (pp. 307-342). Hillsdale, NJ: Lawrence Erlbaum Associates, Ltd. Strid, J. E., and Booth, J. (2007). The Effect of Phonological Structure on Visual Word Access in Bilinguals. Journal of Psycholinguistic Research 36(5), 383-409. Taft, M. (1979). Lexical access via an orthographic code: The basic orthographic syllabic structure (BOSS). Journal of Verbal Learning and Verbal Behavior, 18(1), 21-39. Taft, M. (1994). Interactive-activation as a framework for understanding morphological processing. Language and Cognitive Processes, 9(3), 271-294. Taft, M., Álvarez, C.J., and Carreiras, M. (2007). Cross-language differences in the use of internal orthographic structure when reading polysyllabic words. The Mental Lexicon, 2 (1), 46-63. Taft, M., and Forster, K. (1975). Lexical storage and retrieval of prefixed words. Journal of Verbal Learning and Verbal Behavior, 14(6), 638-647. Taft, M., and Forster, K. (1976). Lexical storage and retrieval of polymorphemic and polysyllabic words. Journal of Verbal Learning and Verbal Behavior, 15(6), 607-620. Tan, L., Spinks, J., Feng, C., Siok, W., Perfetti, C., Xiong, J., et al. (2003). Neural systems of second language reading are shaped by native language. Human Brain Mapping 18, 158-166. Treiman, R., Mullenix, J., Bijeljac-Babic, R., and Richmond-Welty, E. D. (1995). The special role of rimes in the description, use, and acquisition of English orthography. Journal of Experimental Psychology General, 124(2), 107-136. Treiman, R., and Zukowski, A., (1996). Children's sensitivity to syllables, onsets, rimes, and phonemes. Journal of Experimental Child Psychology, 61, p. 193-215. Wang, M., Koda, K., and Perfetti, C. (2003). Alphabetic and nonalphabetic L1 effects in English word identification: A comparison of Korean and Chinese English L2 learners. Cognition, 87(2), 129-149. Wang, M., Perfetti, C. and Liu, Y. (2003). Alphabetic readers quickly acquire orthographic structure in learning to read Chinese. Scientific Studies of Reading, 7(2), 183-207. Woodcock, R. W. (1987). Woodcock Reading Mastery Tests - Revised. Circle Pines, MN: American Guidance Service. Woodcock, R. W. and Muñoz-Sandoval, A. (1995). La Batería Woodcock-Muñoz [The Woodcock-Muñoz Tests]. Chicago: Riverside Publishing. Xu, B., and Perfetti, C. (1999). Nonstrategic subjective threshold effects in phonemic masking. Memory and Cognition, 27(1), 26-36. Yelland, G. W., Pollard, J., and Mercuri, A. (1993). The metalinguistic benefits of limited contact with a second language. Applied Psycholinguistics, 14(4), 423-444. Ziegler, J. C., and Goswami, U. (2005). Reading acquisition, developmental dyslexia, and skilled reading across languages: A psycholinguistic grain size theory. Psychological Bulletin, 131(1), 3–29. Ziegler, J. C., and Goswami, U. (2006). Becoming literate in different languages: similar problems, different solutions. Developmental Science, 9, 429-436. Ziegler, J. C., Perry, C., Ma-Wyatt, A., Ladner, D., and Schulte-Körne, G., (2003). Developmental dyslexia in different languages: Language specific or universal? Journal of Experimental Child Psychology, 86, 169–193.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved. Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

In: Bilinguals: Cognition, Education and Language Processing ISBN: 978-1-60741-710-1 Editor: Earl F. Caldwell, pp. 147-165 © 2010 Nova Science Publishers, Inc.

Chapter 7

VOWELS IN SEMITIC ALPHABET LANGUAGES Raphiq Ibrahim Learning Disabilities Department, and The Edmond J. Safra Brain Research Center for the Study of Learning Disabilities, University of Haifa, Haifa, Israel

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Abstract While several alphabetic systems are in use, this chapter focuses on the two Semitic alphabet languages—Arabic and Hebrew—especially in the orthography of the two languages. Semitic scripts are unique in that short vowels are represented as diacritics on consonant letters. The unique characteristics of Arabic and Hebrew orthographies make them unique for investigations among Latin alphabets or even one among the other (Taouk and Coltheart 2004). Hebrew and Arabic are both read from right to left.

Theoretical Background Before focusing on the vowelization system in Arabic and Hebrew, a theoretical background will be presented about vowels in alphabetic writing systems in general. When the first alphabet was created a few thousand years ago, it contained only consonants—no vowels at all (Shimron 1993). Vowels were added 1,000 years later by the Greeks. The main consideration behind this addition was to represent spoken words in a more comprehensive manner in consonantal homographs, and also to represent phonemes of the spoken words more completely (Shimron 1993). Vowels can strengthen the graphemes-phonemes relationship but can never make it perfect, because the relation between the latter and the sound is not always one-on-one. A letter by itself may represent more than one sound just as one sound can represent more than one specific letter (e.g. Taylor 1980).

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

148

Raphiq Ibrahim

The Role of Vowels in English Reading The common English alphabet contains 5 vowels (a, e, i, o, u) which join the consonants but do not function as consonants. Many researches demonstrate the important roles vowels can play. Some of the vowelization roles are as follows: Venezky (1970) and Massaro, Venezky and Taylor (1979) suggested that vowels played a “marker” function. That is, some vowels in a specific location in a word mark the pronunciation of the other consonants and vowels of this word. Venezky gave an example for a word that ends with a silent “e” in the word ”race”. The silent "e" determines the pronunciation of the “a” as well as the “c” which must be pronounced differently in the context of other words. To disambiguate homophones is another critical role of vowels in other alphabetic languages such as Hebrew and Arabic, which will be discussed latter, e.g. “plain” and “plane”— two words pronounced in the same way while the vowels mark the difference in meaning.

Adams (1981, 1990) tried to identify the role of vowels in the process of reading. Beside her general perception concerning the uninformative function of vowelization in the reading process, she also identified a positive role of the vowels in reading English texts, depending basically on the structure of the syllable. Why particularly syllables? Words, especially long ones, are composed of syllables which must include vowels, mostly at the center of the syllable. Assuming reading involves perception and conserving syllables as perceptual blocks—disconnectedly to the function of phonological awareness of the reader—vowels must assume a basic role in accelerating the process of reading (Adams 1981,1990).

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Characteristics of Arabic and Hebrew Arabic has been the main language of communication for more than hundreds of millions of individuals. In addition, the written form of Arabic is experienced by many more peoples around the world in the form of religious texts (particularly the Quran). Indeed, the written form of Arabic has a special function in representing a standard form of Arabic (Modern Standard Arabic or MSA), which provides a common language across the Arab world. The diversity and geographical separation of countries where Arabic is spoken has led to a variety of versions (local forms or dialects) being used, which means that the local language used by an individual may not be understood that well by another individual from another part of the Arab world. Hence MSA is the common language of communication across the Arab world and the written form represents this common language. Since MSA is not a mother-tongue of anyone in the Arab world, acquiring good skills in MSA occurs primarily during schooling (Maamouri, 1998). Given that lexical items will vary to differing degrees (depending on the local language) between MSA and mother-tongue, and syntactic and morphological rules in MSA will vary from the child’s spoken experience, the learning of the written form, therefore, can be argued to provide a route to support understanding across the Arab speaking countries, as well as its role in its cultural expression. However, despite the potential significance of written text in cultural and religious aspects of Arabic life, there are still a large number of individuals who cannot read formally (potentially some 9% of the world’s

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Vowels in Semitic Alphabet Languages

149

illiterate individuals live in the Arab world). Hence, it is an important language to consider in terms of literacy acquisition and factors that may be related to literacy learning problems. In Hebrew and Arabic, all verbs and most nouns are written primarily as consonantal roots that are differently affixed and vowelled to form the words of the lexicon (Berman, 1978). The simple letter-sound correspondence in vowelled Arabic script makes it transparent or shallow (Shimron 1999). Vowelled texts are appropriate for beginning and unskilled readers, and most written materials do not include vowels. In Arabic, of the vowel signs of the three short vowels, two are positioned above the letter “fatHa” (َ )=a, “damme” (ُ )= u the consonants, and one below “kasra”(ِ )=i. Although these vowels are not letters, the combination with consonants forms CV syllables (Taouk and Coltheart, 2004). For example, the consonant t, when combined with these vowel-signs would form different syllables ending in short vowel sounds: ta, tu and ti. Furthermore, there are “double fatHa” (ً ) sounds “an,” “double damme” (ٌ ) sounds “on,” and “double kasra” (ٍ ) sounds “in.” These vowel signs are used in cases of undefined subjects. The corresponding long vowels for “fatHa” =a, “damme” = u, and “kasra” =i and for the diphthongs ai and au are alif (for a), waw (for u, au) and ia (for i, ai) possibly with the addition of the signs for short vowels (Jensen, 1970). In addition to the diacritics for short vowels, the long vowels and the diphthongs, there are four other reading signs: the “skoon” (ْ ) which signals absence of vowel, “shada” (ّ ) which signals doubling of consonant, “maddah” ( ~ ) which signals doubling of letter alif and “hamzeh” ( ‫ ) أ‬which signals the glottal-stop sound. In Hebrew, the diacritical marks are dots and strokes which are usually posited below the letters, and sometimes above or in the middle of the letters. Letters are presented as syllables when combined with vowel-signs, for example, the letter b can be presented with: (ba), (be), (bi), (bo) or (bu) when combined with the kamatz (ָ ) and patach (ַ ) for the phoneme (a), tsere (ֵ ) for the phoneme (e), cherek (ִ) for the phoneme i, and segol (ֶ ) for the phoneme e—besides the two letters “yod” and “vav” because sometimes the vowels are presented by letters: “yod = ‫ ”י‬for the phoneme i and “vav = ‫“ו‬for the phonemes (o) or (u). There are three letters in Arabic (‫ ي‬,‫ و‬,‫ ) ا‬and four in Hebrew (‫ י‬,‫ ו‬,‫ ה‬,‫ )א‬which in addition to their role in signifying specific consonants, also specify long vowels. In some cases, it is difficult for the reader to determine whether these dual-function letters represent a vowel or a consonant. When vowels do appear (in poetry, children's books and liturgical texts), they are signified by diacritical marks above, below or within the body of the word. Inclusion of these marks completely specifies the phonological form of the orthographic string, making it completely transparent in terms of orthography/ phonology relations. Due to the fact that Arabic and Hebrew words are almost based on roots compounded of three letters, and many words could be formed by using different vowels in each letter, even by using different affixes, which makes different words look identical when written without vowels (e.g., Abu-Rabia and Siegal, 1995). Thus in their unpointed form, the Hebrew and Arabic orthographies contain a limited amount of vowel information and include a large number of homographs. Recent studies on reading in Arabic and Arabic orthography indicate different conclusions from other studies conducted in Latin orthographies (e.g. Abu-Rabia, 1997a) or even differs from the Hebrew orthography and reading process despite sharing common characteristics with this Semitic language, the homograph phenomenon is very distinct in that

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

150

Raphiq Ibrahim

almost one third of the words in Arabic unvowelized texts might be homographs, namely, typifying several meanings (Abu-Rabia, 1997a). A distinct feature of the Arabic language compared to other languages is the vowels at the end of the words that determine the structural function of the word in the sentence and hence changes the whole meaning (Abu-Rabia, 1997a). For example, the meaning of the word water, represented with the word “‫ ”ﻣﺎء‬is dependent on the vowel at the end of the word, which determines whether it is the defined or undefined subject or object. Hence, considerable knowledge such as literary Arabic syntax, vocabulary, and contextual interpretations differentiate skilled readers from poor readers (Abu-Rabia, 1997a), especially while reading unvowelized texts. The differences between Arabic and Hebrew writing systems may be strengthened through studies which investigated the influence of vowels on reading accuracy in Hebrew among good and poor readers (e.g. Frost, Katz and Bentin 1987; Navon and Shimron, 1981, 1984). Although the authors did not intend to compare the two orthographies, addressing the comparison is possible. Frost and his colleagues (1987) used words that carry one meaning (not homographs), and they concluded that vowels do not facilitate naming in Hebrew. On the other hand, Navon and Shimron (1981, 1984) concluded that although skilled Hebrew readers can manage fluent reading without vowels, those readers do not ignore vowels even if they are instructed to do that. However, they concluded that even skilled readers were not sensitive to different vowel marks that represented the same phoneme (e.g. /a/). Frost and his colleagues (1987) concluded that in spite of the ambiguity of the unpointed Hebrew, good readers did not need vowels for reading. Abu-Rabia (1997a) has shown that both skilled and poor high-school level readers of Arabic utilize both context and diacritical vowels extensively. In fact, Abu-Rabia and Siegel (1995) have claimed that unvoweled isolated words in Arabic are almost impossible to read correctly, due to the large number of homographs in the language. Their conclusion was that in word recognition, poor Arabic readers rely on context more than skilled readers (as in English). The findings of Abu-Rabia (1997a) contradict findings obtained from Hebrew. Participants from both groups (poor and skilled readers) were presented with isolated unvowelized words but showed a floor effect; a finding that contradicts what was found in Latin orthographies (e.g. Bruck 1990; Perfetti 1985) or even findings obtained from Hebrew orthographies (as mentioned above). This finding consolidates the importance of vowels for reading in Arabic. It was also concluded that Arabic could facilitate word accuracy in reading context either for poor or skilled readers. Interestingly, a close examination of Abu-Rabia's (1997a) data reveals that although both context and vowelization help both poor and good readers, there seems to be an interaction between these variables. Vowelization facilitates correct reading stronger in good than in poor readers, while context shows the opposite effect: it facilitates correct reading stronger in poor than in good readers. This is an interesting finding, suggesting that phonological information is more facilitative for good readers, while semantic information is more facilitative for poor readers, converging with general models of reading and reading acquisition (Frost, 1995; Stanovich and Feeman, 1981). Finally, it can be concluded that reading Arabic texts could be qualified as interactive and dynamic process (Abu-Rabia 1997). Abu Rabia (1999) investigated second and sixth grade native Arabic speakers on silent reading comprehension of vowelized and unvowelized texts. It was found that vowels facilitated reading comprehension, and it was concluded that the process involved vowels

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Vowels in Semitic Alphabet Languages

151

beside the context. Abu Rabia (2001) found that even highly proficient adult native Arabic speakers benefited from vowels for comprehending texts. The researcher concluded that this was seemingly due to the nature of the Arabic orthography. The process of reading unvowelized texts in Arabic necessitates focusing on the morphological structure of words (root identification), a process that needs additional cognition effort for lexical access and compensates, in addition to the context and prior knowledge, for the lack of vowels (Stanovich 1980). A research conducted by Badry (1983) confirmed this conclusion. It was found that identifying word roots is an essential process for reading comprehension in Arabic. Two additional domains add to the complexity of both orthographies, but to a much larger extent in Arabic than in Hebrew. The first has to do with the role of dots. In Hebrew, dots are a diacritic used as a stress marking device (dagesh). For example, this stress marking device (which does not appear in unvowelized scripts) changes the phonemic representation of the letters from fricatives (v, x, f) to stops (b, k, p for the letters ‫ פ ק ב‬respectively). However, ‫ בּ‬and ‫ ב‬represent the same letter, which has two phonemic representations. In the unvowelized form of Hebrew script, these different phonemes can be disambiguated by their place in the word, as only word or syllable initial placement indicates the stop consonant. In Arabic, dots are not diacritics, but an integral part of the grapheme, where many different letters have a similar or even identical base structure and are distinguished only on the basis of the existence, location and number of dots (for example, the phonemes t, b, and n are respectively represented by the letters: ‫ب ت‬, and ‫ن‬, and the phonemes r and z by ‫ ر‬and ‫)ز‬. The second difference between Hebrew and Arabic orthography that we believe is relevant here is that in Hebrew, diacritics/dots usually appear under the letters and represents vowels. In Arabic, dots in themselves do not have phonetic value but are part of the consonant letter (integral parts of the letters). For example, letters have a similar or even identical structure are distinguished on the basis of the existence, location and number of dots (e.g. the Arabic letters representing /t/ and /n (‫ )ن ت‬will present /th/ and /b/ (‫ ) ب ث‬according to the number or location of dots. An additional characteristic that contributes to the complexity of the two orthographies is that some letters are represented by different shapes, depending on their placement in the word. Again, this is much less extensive in Hebrew than in Arabic. In Hebrew there are five letters that change shape when they are word final: ( ‫ך‬-‫ כ‬,,‫ם‬-‫מ‬, ‫נ‬,‫ץ‬-‫צ‬,‫ף‬-‫פ‬-‫) ן‬. In Arabic, 22 of the 28 letters in the alphabet have four shapes each (word initial, medial, final, and when they follow a non-connecting letter, for example, the phoneme /h/ is represented by the graphemes: ‫))هـ ـﻬـ ﻩ ـﻪ‬, and six have two shapes each, final and separate. Thus, the grapheme phoneme relations are quite complex in Arabic, with similar graphemes representing quite different phonemes, and different graphemes representing the same phoneme.

Phonological Processing Arabic has a highly regular/transparent orthography when presented in the marked or vowelized form of the writing system (in this respect, it is similar to many other Semitic languages; for example, Hebrew). A regular/transparent orthography is one where there is a relatively simple relationship between the written form and the language sounds that the written form represents: i.e., there is close to a one-to-one correspondence between

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

152

Raphiq Ibrahim

graphemes and phonemes. In other orthographies (English is the best example), this correspondence is less transparent, meaning that a letter may represent several sounds, and a particular sound may be represented by several different letters. However, the Arabic language is based on a highly derivational morphological system. Once learning of the basic association between written and verbal form has taken place, the emphasis of the written form is on meaning, which is primarily conveyed by morphological components. Hence, despite languages such as Arabic having a highly regular orthography when fully marked (or fully vowelized), this form of the orthography is rarely used in most literary texts read by the more experienced reader (the exception, most likely, being religious texts). Once beyond initial schooling grades, the Arabic child is likely to experience text in which short vowel markers are removed, leading to an orthography that is opaque in its relationship between letters and sounds, and to texts that contain a large number of homographic words (i.e., words that look alike but which represent different concepts and are pronounced differently). Such nonvowelized text needs to be read ‘in context’. This means that an adult or child experiencing such writings will have to decipher the context within which a word is written, such as the meaning of words around the homograph or the general theme of the passage, or the grammatical structure of a phrase, to be able to understand the meaning of the word and even pronounce that word correctly. Hence, Arabic word processing may rely on phrase/sentence processing or text comprehension to a larger extent than found in some other languages (even English). As discussed above, word decoding may be mediated by phonological processing. Such processes may be critical for the ability to translate a written letter string into an appropriate pronunciation, and this role has been investigated in Arabic. For example, Abu-Rabia, Share and Mansour (2003) compared the performance of reading disabled Arabic learners with both age matched and reading level matched peers in tests of phonological and orthographic processing as well as measures of syntax, morphological awareness, working memory and visual memory. Results of this study indicated that the most severe deficiencies amongst the reading disabled group were found for measures of phonological awareness, in contrast to their relative strengths in orthographic processing. Abu-Rabia and Taha (2006) interpreted phonemic errors that were fairly stable across Arabic school grades as evidence that the Arabic learner tends to rely on their phonological/decoding/alphabetic skills for longer than would be predicted based on current literacy developmental models derived from Latin-based scripts (eg, Frith, 1985). Abu-Rabia and Taha explained that this might be a specific effect within Arabic, due to the complex phonology and orthography of Arabic. The importance of phonological processing measures as predictors of Arabic literacy levels was also noted by Al-Mannai and Everatt (2005) who examined predictors of literacy development among grade 1 through grade 3 Arabic speaking learners in Bahrain. Regression analysis indicated that pseudo-word reading was the best predictor of variability, with a rhyme awareness task also strongly predicting variation in word level literacy – again consistent with the need to use phonological and letter-sound decoding processing skills. The importance of phonological measures as a potential predictor of reading ability levels amongst fourth and fifth graders in Egypt led Elbeheri and Everatt (2007) to conclude that reading Arabic, like English, depends to a large extent on phonological processing skills. Such evidence was used by Elbeheri, Everatt, Reid and Al-Mannai (2006) to propose that models of English literacy acquisition, and literacy learning difficulties, could be applicable to understanding the same processes in Arabic.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Vowels in Semitic Alphabet Languages

153

However, these authors also argue that further research is necessary to allow firm conclusions to be made given that variations from predictions based on English language models were identified. These were reasoned as potentially related to specific orthographic and/or morphological features of Arabic.

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Morphological Processing The Arabic written form may be described as complex for the beginning reader/writer. For example, the positioning of dots or marks either above, below or, sometimes, within letter shapes distinguishes different letters or grammatical rules. These letter shapes vary depending on their positioning at the beginning, middle or end of a word, meaning that the Arabic child will experience cursive text with letter shapes varying dependent on position within the word, with marks being included or not depending on the level of the text (vowelized or nonvowelised as described above). In addition, despite its cursive form, some letters in Arabic (one-way connectors) do not join to both letters around them, meaning that the size of a space needs to be used to distinguish a word boundary. Therefore, recognizing an individual feature, such as a letter or a letter combination, within Arabic text may be a more complex process than doing the same thing in English. Potentially consistent with this complexity argument, Ibrahim, Eviatar and Aharon-Peretz (2002) found that when biliterate children were given a trail making task, in which participants had to serially order letters while matching them with numbers, an Arabic orthography condition was significantly slower than a Hebrew orthography condition, even though Arabic was the first language of the individuals tested. These findings led the authors to argue that the complexity of the Arabic orthography makes it difficult to process. Additionally, Elbeheri and Everatt (2007) found that a word chains task (in which participants had to indicate word boundaries in a random series of Arabic written words from which the spaces between words had been removed) was highly related to reading levels amongst Egyptian primary school children, and this relationship was larger than the analogous correlations for the phonological and decoding measures in the study. Therefore, orthographic complexity may be an additional hurdle for the Arabic child when learning letter-sound decoding. This may be particularly salient when the transition between vowelized and non-vowelized forms is encountered. For example, Abu-Rabia (1999, 2001) investigated the influence of using vowelized and non-vowelized variations of the Arabic script among 2nd and 6th grade children and adults, and found that the vowelized form of the Arabic script tended to increase the levels of reading comprehension shown by these readers. Hence, although the vowelized form is more visually/orthographically complex than the non-vowelized form, making the link between letters and sounds simple may improve reading skills, even for skilled/adult readers. Although, in a later study, Abu-Rabia (2007) examined the reading skills of typical and dyslexic Arabic native readers (grades 3, 6, 9 and 12) and found that vowelization (either within words or at the end of words as a measure of syntactic knowledge) was not a predictor of reading accuracy or reading comprehension. Interestingly, whereas Abu-Rabia (2007) did not find vowelization predictive of Arabic reading amongst dyslexic and control children, there was an effect of morphology. Along with spelling ability, the identification and/or production of morphological units was generally predictive of reading (both accuracy and

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

154

Raphiq Ibrahim

comprehension) in both groups across the grade range studied. Therefore, morphology, rather than orthographic effects, may be the reason why studies have found differing effects from that predicted by English language data. In addition, Boudelaa and Marslen-Wilson (2001, 2005) have found that priming by morphological units was different from that obtained from orthographic/phonological controls suggesting that morphology and orthographic influences on word processing need to be treated somewhat independently in models of Arabic reading ability (see also Mahfoudhi, 2007).

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Statistical Data in Israel Analysis of data regarding reading abilities conducted and published by the Israeli Ministry of Education revealed that the acquisition process of reading Arabic (in the Arab sector) is slower than in Hebrew (Jewish sector) with the final achievement level somewhat lower. Since 1975, there has been a serious confusion as a result of these results of reading estimation tests among Arabic speakers (Bashi, Sorel and Daniel, 1981) that led to continuing assessments for 17 years, with the results being published in 1992. The results of these analyses were identical. Furthermore, the intervention period did not contribute to any fundamental progress among Arabic speakers (Azaiza 1997). It was found that approximately 23% of third graders in Arab school were not able to solve basic questions such as differentiating letters within words (e.g. to find the letter “‫ ”س‬in simple words, corresponding to the very low percentage among Hebrew speakers in the same grade (about 4%). Despite the attribution of several causes for this phenomenon, such as the learner himself (Kais, 1978), readability of the text (Adar, 1987), the pragmatic aspect (Vipond 1980), and others, the present research will focus on the unique characteristics of the Arabic language as an important and influential source for this phenomenon. Furthermore, previous studies on reading acquisition in the Arabic language have revealed identical results (Azzam, 1984, 1993; Ibrahim, Eviatar and Aharon-Peretz, 2007). In this regard, it has been found that reaction times for visual recognition of Arabic words by Arabic speakers are longer than reaction times for Hebrew words by Hebrew speakers (Bentin and Ibrahim, 1996). Moreover, when visual Arabic-word recognition was compared with visual Hebrew-word recognition among native Arabic speakers, Arabic words took longer to be recognized, although the Arabic words were recognized faster than the Hebrew words when the words were presented in the auditory modality. In comparison to other languages, lexical decisions and naming were inhibited for Arabic words compared with Hebrew, English and Serbo-Croatian (Frost, Katz and Bentin, 1987). Also, large word-frequency effect was found in naming as well as in making positive lexical Decisions. This finding is very interesting, because, Arabic (like Hebrew) considered as ‘deep’1 orthography.

1

.Voweled Arabic orthography is considered “shallow” for the unambiguous grapheme-phoneme relation, for example the voweled word: ‫ﺐ‬ َ ‫( َآ َﺘ‬kataba) had one reading option, while unvowelized orthography is considered “deep” for the grapheme-phoneme relation that is ambiguous because unvoweled word in Arabic has a number of reading options, for example: the unvoweled word (ktb) ‫آﺘ ﺐ‬in Arabic can be read in several ways: ‫آَُﺘُـــــﺐ‬- kutub,- َ‫ َُآﺘِـــــــﺐ‬kutiba, َ‫آَﺘَـــــــﺐ‬- kataba.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Vowels in Semitic Alphabet Languages

155

A number of studies were conducted in an attempt to examine possible sources of slowness in reading acquisition of Arabic, in relation to some other languages.The first source is related to the psycholinguistic basis for being the literate Arabic speaker a bilingual de facto. The second source focus on the relationship between the Arabic orthographic system and cognitive processes that might be involved during word recognition. In particular, these focused on the specific characteristics of Arabic orthography and the ways these characteristics may influence the acquisition of reading (e.g Ibrahim and Aharon-Peretz, 2005; Feitelson et al, 1993; Ibrahim, Eviatar and Aharon-Peretz, 2002; Eviatar, Ibrahim and Ganayim, 2004; Eviatar and Ibrahim, 2004). Israeli Arab adults, can be minimally considered multilingual, with ammia-SA (Spoken Arabic) as L1, and fuṣḥa-Literary more commonly referred to as ‘Modern Standard Arabic’ (MSA) ), Hebrew and English as additional languages. SA is a local dialect which accompanies children from birth and throughout their whole lives, including their study period. This situation in Arabic termed “diglossia,” a phenomenon referring to the existence of two forms/systems of the same language (Ferguson, 1959). The spoken dialect has no written form, while the written form of the MSA is taught in schools parallel to learning how to read and write. Literary Arabic or Modern Standard Arabic, the formal Arabic, universally used in the Arab world for formal communication and writing and has become part of everyday life, as it is the language in which news is reported (both written and oral) and it is the language of prayer and of formal public occasions. Although sharing a limited subgroup of words, the two forms of Arabic are sematically, phonologically and syntactically different. Related to this issue, the following background is pertinent: There is a semantic gap between SA and MSA: For example: the word balcony in English is ‫" ﺑﺮﻧ ﺪة‬baranda" in SA while it is ‫" ﺷ ﺮﻓﺔ‬shorfa" in MSA. Therefore, a lexico/semantic representation of a spoken word might differ from its representation in the standard version although it is related to the same concept. There is also phonological gap between SA and MSA. For example: the word dog (‫)آﻠ ﺐ‬ is pronounced as /kalb/ in the classical Arabic and /kalib/ in the spoken one (adding the phoneme /i/), or the word officer is pronounced as /dabet/ in the classical Arabic and /zabet/ in the spoken one (substituting the phoneme /d/ with /z/). Therefore, a specific phonological representation of a spoken word might differ from its phonological representation in the standard version although it is phonetically related to it (Saiegh-Haddad, 2004). Recent psycholinguistic studies which have addressed this issue directly (e.g. Ibrahim et al. 2005) have revealed that the two forms of Arabic function as two separate language systems, such that a literate Arabic speaker is essentially a bilingual. Another study (Eviatar and Ibrahim, 2001) has shown that young Arab children who have been exposed to Literary Arabic function as bilinguals on tests of metalinguistic awareness. A recent study (Ibrahim et al. 2007) compared reading measures in Arab, Hebrew monolingual, and Hebrew-Russian bilingual first grader children groups. Two main questions were discussed in their study: (1) does greater phonological awareness of bilinguals affect reading performance; (2) how could orthographic characteristics of a language influence reading performance and how does this interact with the effects of phonological awareness. The questions were tested with phonological awareness tests (initial/final phoneme detection, phoneme/syllable deletion) and through reading measures (reading single real words, nonwords, pointed text) in three groups of first graders. The results have shown better functioning

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

156

Raphiq Ibrahim

on meta-linguistic measures among Russian–Hebrew bilinguals and Arabic speakers than Hebrew speakers. On the other hand, text reading revealed a significant effect of the language on reading performance (accuracy and speed) when Arab children exhibited the slowest speed and the highest mean number of errors. Russian- Hebrew children exceeded the two groups, with Hebrew monolinguals in between (but with no significant difference between Hebrew monolinguals and Russian- Hebrew bilinguals). However, Arabic speakers have shown better performance in single word and non-word reading over the other two groups, but they exhibited very weak relationships between phonological abilities and text reading – a result that did not fit the strong correlative relations between phonological awareness and reading skills found in many researches on languages such as: English (e.g. Stanovich, Cunningham and Cramer, 1984). Planned computations revealed that phonological ability predicted reading performance differently: over 60% of the variance in text reading (speed and accuracy) was predicted by phonological tests for monolingual Hebrew readers, a similar percentage for Hebrew-Russian bilingual readers, and only 30% for Arabic readers. Even though the Arab children had higher scores than monolinguals on tests of phonological awareness, those abilities particularly did not facilitate text reading performance for Arab native speakers. Hence, the researchers concluded that the Arabic native speakers experienced more difficulty in relation to Hebrew monolinguals and bilinguals in language processing, which might be related to the visual complexity of Arabic orthography (the characters of the Arabic alphabet system will be elaborated on in the following sections). It was also concluded that the characteristics of the language that children learned to read must have much greater impact than the status of the language in the children’s linguistic history.

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Research in Vowel Perception in Arabic and Hebrew The processing of orthography has two aspects: orthography is a representation of linguistic units (e.g., phonemes in alphabetic scripts, syllables in logographic scripts), and it also has purely visual aspects. This dual characteristic is evident in research on reading in general, with differing weights given to the linguistic and the visual aspects of the task (Stein and Walsh, 1997). On the ground of the described literary review, and given the complex Arabic orthography and the considerable similarities such as differences between Arabic and the Hebrew orthography, In recent study, Shalabna, Ibrahim and Eviatar (submitted), examined the degree to which orthographic complexity and lexical status can affect vowels detection in Arabic and Hebrew, among 6th and 3rd graders who are native Arabic speaking, and considered normal readers in Hebrew and Arabic. For obtaining this purpose, we used a series of detection tasks. The targets was a specific vowel (“fatHa” in Arabic stimuli and “patah” in Hebrew stimuli). A voweled stimulus was presented on a computer screen, and the participant was asked to decide if the specific vowel exists or not by pressing the proper button on the keyboard (yes/no). We hypothesized that lexical information will facilitate perception of short vowels, apparent in better performance in real words as compared with unmeaning stimuli; hence, children will show word superiority effect. In addition, we phonological information would not facilitate perception of short vowels; hence, reading the stimulus will not help the children in facilitating their performance. Finally, we hypothesized that children will process orthography differently, as a

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Vowels in Semitic Alphabet Languages

157

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

function of the complexity level of the orthography, so the result will be two-fold: on the one hand, no differences would be found in performance between the Hebrew stimuli and Arabic stimuli which are constructed of 3 separate letters without dots as an integral part of the letters. But on the other hand, significant difference would be noticeable between stimulus constructed of separate letters and stimulus constructed of connected letters of Arabic with or without dots as integrative part of the letter (apparent that connection of letters in Arabic indicates complexity of the Arabic orthography). All participants were taken from 3rd graders and 6th graders who are native Arabic speakers. All children were recruited from an elementary school from the lower Galilee in Israel in which Arabic is the official language. The stimuli were of three levels of lexicality: real words, pseudo-words and strings of letter-like stimuli (semi word pattern of matched lines and curves). All stimuli were strings of three letters or semi-letters (patterns that look similar to letters) and diacritical marks (short vowels) that were part of the consonants as well as some of the vowels were presented on them. Most of short vowels from the two languages were included. The diacritics were presented beneath and/or above the letters according to the acceptable rules in Arabic and beneath the letters according to the acceptable rules in Hebrew orthography. The three categories of each kind of words in Arabic, differing in their complexity, as follows: a) stimuli constructed of non-dotted separate letters, letters without dots comprised as an integral part of them. This category is defined as having simple orthography. b) stimuli constructed of non-dotted connected letters. c) stimuli constructed of dotted and connected letters, connected letters with dots comprised as an integral part of them. This category is defined as having complex orthography. Such categorization does not exist in Hebrew because the Hebrew orthography is not complex and words are always constructed of separate letters (simple orthography). The fatHa ”َ “ which was designated as the target diacritic, existed in half of the stimuli (real words/ pseudo-words/ semi words) in each category in Arabic. The patah “ַ” that was designated as the target diacritic in Hebrew existed in half of the stimuli (real words/ pseudowords/ semi words) in each category in Hebrew. Examples of the stimuli in Arabic are shown in Table 1. In Hebrew only lexicality was manipulated, as these levels of complexity do not exist in Hebrew orthography. Examples of the stimuli in Arabic are shown in Table 2. The task of the participants was to detect the FatHa “َ “ by pressing the right square parentheses “]” if they see FatHa in the stimulus in the center of the screen, and by pressing the left square parentheses “[” if the FatHa does not appear in the presented stimulus. In the second set all the Hebrew stimuli appeared one by one on a computer screen randomly from the 3 groups. The participants were asked to detect the patah “ַ” by pressing the right square parentheses “]” if they see patah in the stimulus presented on the screen, and pressing the left square parentheses “[” if the patah does not appear in the presented stimulus.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

158

Raphiq Ibrahim

Table 1. Examples of stimuli in the Arabic conditions with transliterations for words and pseudo-words Arabic stimuli: Lexicality levels Real words

pseudo-words

Semi- words

Orthography groups

With ”َ “

Without ”َ “

non-dotted separate letters

‫( ُدرَ ٌر‬dorarun)

ِ ‫( إرْم‬ermi)

non-dotted connected letters dotted connected letters non-dotted separate letters

‫( ﻣَﻄَ ٌﺮ‬matarun)

‫( ُﻣ ْﻌ ٍﺪ‬moaden)

‫ﺞ‬ َ

‫ﺧ ﺒ ٌﺰ‬ ُ (khobzun)

‫( َﻧ َﺘ‬nataja)

‫ح‬ ٌ ‫( َو َر‬warahun)

‫( وِ ْد ٌم‬wedmun)

non-dotted connected letters

‫ﺴ ٌﻢ‬ ْ َ‫( ﻋ‬asmon)

‫( ﻟُﻜ ٌﺪ‬lokdon)

dotted connected letters non-dotted separate letters

‫ﻲ‬ َ ‫ﺑَﺸ‬ (basheya)

‫ﺨ ﺾ‬ ِ ‫( ُﻓ‬fokhed)

non-dotted connected letters

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

dotted connected letters

Table 2. Examples of stimuli from the Hebrew condtions with transliterations for words and pseudo-words stimuli in Hebrew Real words

With “ַ” ‫( ִשׂי ַח‬seiah)

Without “ַ” ‫( ֶע רֶב‬erev)

pseudo-words

‫( ַצ דֵב‬tsadev)

‫( ֶד רֶר‬derer)

semi words

Comparing Hebrew Orthography with Arabic Orthographies Comparison between grades revealed differences in performance between the two grades (a main grade effect was found). For the stimuli presented in Arabic, 6th graders responded faster and made fewer mistakes and were more sensitive to differences between stimuli. These differences may be accounted for by natural developmental differences, and by differences in the extent of exposure to the language between third and six graders. In the Hebrew stimuli, 6th graders exceeded third graders only in response times; however, in

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Vowels in Semitic Alphabet Languages

159

mistake percentages and sensitivity measures no differences were found between the two grades. The difference in response time may be accounted for by a developmental factor which in part may be due to greater exposure to the Hebrew writing for six graders. Though this finding is not applicable to all dependent measures in Hebrew, it gives a different perspective for the exposure to Hebrew. This, in turn, accounts for the difference between findings for Hebrew and Arabic. Thus, despite the three year gap between the two groups of children, the gap in exposure to Hebrew writing system is not equal to that of the Arabic writing system, and is less intensive. A three year gap between the two groups in exposure to the Arabic writing means a gap in exposure to the Arabic in most classes during all the days the student is at school, alongside an almost daily exposure to the Arabic writing system at home (stories, newspapers, etc.). However, a three year gap in exposure to the Hebrew means a gap in exposure to the Hebrew writing system in a very limited number of weekly Hebrew classes, alongside a very limited exposure to the Hebrew writing in daily life. Thus, these findings may account for the effect of the great exposure to the Arabic in relation to the Hebrew orthographies, and its contribution to the increased sensitivity to Arabic script, which was reflected in the best accomplishments in the three dependent measures among six graders. The less intensive exposure to the Hebrew script contributed only partially to increasing sensitivity to the Hebrew writing system. In regard to lexical and phonological effect on perception, Our data does not support the second hypothesis about the “ word superiority effect” because no differences in performance between words and non words were found, either among 6th graders or 3rd graders in both measures: RTs and errors percentage (except of dotted and connected letters among 3rd graders). These findings were obtained for the Hebrew stimuli as well, as no differences between words and pseudo-words were found in all three measures: RTs and errors percentage, as well as in the sensitivity measure (except for a marginal difference between non-words and pseudo-words and between words among third graders). These findings indicate that words and pseudo words are somewhat processed in the same way. According to our hypothesis, this finding indicates that the lexical information was not necessary in the process of the vowels detection. Phonological information also did not facilitate perception of short vowels appended to the letters/semi letters, because children in both grades mostly did not perceive the real words faster and more accurately than pseudo or semi stimuli. In other words, they did not show words superiority effect (except the case of the reaction time of non-dotted and connected letters and marginally in the case of the errors percentage of dotted and connected letters among 6th graders). In an attempt to examine the way children perceive vowels in more detail, a comparison of the difference lexicality levels of words, pseudo and semi words was performed in each of the different complex levels. The comparison was performed between the groups and among the groups, and the obtained results indicated the same pattern in response times and a partial pattern in error percentages, but the main common finding indicated no differences between words and non words in all levels. these findings further strengthening the previous findings regarding the phonological and lexical effect on the performed task. Compared to previous research, the evidence from the present research regarding the reduced phonological and lexical effect on visual perception of vowelization in the Arabic script are in line with some studies in the field, while they contradict others. For example, the

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

160

Raphiq Ibrahim

results abtained by Bentin, Bargai and Kats (1984) were found consistent with our results, and indicated that the phonological code does not constitute an important component in the process of word identification as does the orthographic code (an elaboration of the orthographic processing will be presented in later sections). Bentin and his colleagues (1984) examined the degree of phonological mediation in the access to a lexicon of single words in Hebrew in a series of three trials. In the first task, the subjects had to implement the presented stimuli, while the second task entailed a lexical decision. Unvowelized words were used in both tasks, whereby some of them had one reading option while others had a number of reading options (phonological ambiguous). In the first trial, it was found that subjects implemented the ambiguous words slower than the words with one reading option, and the researchers concluded that phonology was of importance in this task. However, in the second task this difference between the two types of stimuli was not found. It appears that the findings indicate that phonology is involved in the naming process but not in the lexical decision. However, the researchers concluded that the ambiguousness effect is subsequent to the lexical decision. Because the finding is contradictory to what is found in research literature on other languages considered to be shallow in relation to Hebrew (e.g. English), where naming was faster than the lexical decision, in the present research it was concluded that the naming was slow apparently due to the fact that naming in Hebrew is a post-lexical process. In the third trial, a lexical decision priming paradigm was used, in which real words were presented as prime, along with pseudo words orthographically or phonologically similar to the real words presented. All stimuli were vowelized. The results support the researchers’ hypothesis in that it was found that the decision of pseudo words orthographically similar to the prime was slow in relation to pseudo words of the other kind. Thereby, the conclusion was that the orthographic code was the dominant one, and not the phonological. In contrast, the research results in the field are not consistent with other studies in this field. Abu Rabia and Taha (2006) found that phonology presents the basic factor for spelling errors in Arabic across grades, as phonological errors were the most dominant among Arabic speakers between 1st and 9th grades. Identical results were recieved in a study by Abu Rabia, Share and Mansour (2003), as they found that the most severe deficit among dyslexic Arabic readers was obsereved to be in phonological awareness. Furthermore, a study conducted by Bentin and Ibrahim (1996), found evidence of phonological processing during visual identification of words in Arabic in a research examining lexical decision and naming of real words, pseudo-words, along with transliterations of words from spoken Arabic (as mentioned in the introduction, have no written form and are expected to be novel orthographic stimuli for the participants), and attempted to examine whether the reader ignores the phonologic information of the presented stimuli when this information helps the subject to perform the task more efficiently. In their study, it was found that both naming as well as the lexical decision were delayed when stimuli with transliterations were presented, whereby the functioning was poorer than when pseudo-words were presented. The researchers attributed this finding to the mediating factor phonology plays, and this proves that the subjects did not ignore the phonology of these words, although the reading that relied on phonological decoding and not the word pattern (because they were not in fact exposed to the patterns of these words before the trial) delayed them. The research findings in the same scope in some studies in Hebrew language are also not consistent with our findings. (e.g. Liberman 1992, Perfetti and Bell, 1991) as they found that

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Vowels in Semitic Alphabet Languages

161

phonology is a mediator for semantic memory. Some even maintain that this component is activated automatically during word identification (e.g. Van Orden 1991). Similar findings were also obtained in studies on the Hebrew language (Navon and Shimron, 1981, Navon and Shimron, 1985). Their results showed that wrong phonological representation may interfere in correct stimuli naming, and thus the chance for automatic phonological processing in the process of words identification. The findings of the role of orthography in the present research surprisingly indicate that it is not the connection in letters in the Arabic writing which constitutes a cognitive load, but apparently the dots which constitute an integral part of the letter (there are approximately 15 such letters in Arabic). It is this that the present research proved unequivocally, and therefore vowels constituting a component of the letter cannot be considered while ignoring other features of the Arabic letters. As mentioned in the introduction, the grapheme-phoneme relation in Arabic is especially difficult due to the multiple forms of each letter in relation to its location (beginning, middle, and end) within a word. In conclusion, The result gained by our studies showed very clearly that detection of the target was faster and more accurate in Hebrew always, in both age groups. This reflects the perceptual complexity of Arabic orthography. In addition, the differences between the grades revealed that although both connectivity and the existence of dots affected both groups, it seems that connectivity of the letters has a larger effect in 3 grade and the existence of dots a larger effect in 6 grade. The absence of a word superiority effect suggests that reders of Arabic may not parse connected words into separate letters.

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Educational Recommendations Reading acquisition in the Arabic language is unique due to the orthography and diglossia characterizing this language. The present research adds to the accumulating body of research that brings further evidence to the cognitive weight of the Arabic language orthography, which leads to to importance of increased, directed exposure to reading at an early age, and the importance to fostering various learning environments. School managements have the responsibility to consider the following recommendations: first, schools have the responsibility to encourage reading in every possible way, such as designing a library in every classroom, giving reading homework, holding reading contests, and holding discussions with students on books they have read. Second, increasing the weekly hours of teaching the Arabic language is recommended, as the hours provided today are not sufficient, especially as the Arab pupils begin to study a third language, English, which comes at the expense of the hours allocated to teaching the Arabic language (Azaizeh, 1997). It is recommended that teachers utilize various methods in instilling the basics of writing, such as integration between cartoon television programs for writing acquisition, computer games, various activities to draw children’s attention, etc. Undoubtedly, parents have an active role in fostering the process from an early age in their children, exposing them to stories and creating an environment that encourages reading. Finally, in light of the fact that Arab children learning how to write in Arabic encounter not only the difficulty of the letter, but also the fact that s/he needs to learn about 80 letters until the end of the first grade, as most of the letters (22 out of 28) have at least 3 different

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

162

Raphiq Ibrahim

forms, depending on their position within a word. Therefore, the question as to why do children have to learn the alphabetical code within one year, and what if they learn the letters in more than one year arises.

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

References Abu Rabia (1999). The effects of Arabic vowels on the reading comprehension of second-and sixth-grade native Arab children. Journal of psycholinguistic research, 28(1), 93-101. Abu-Rabia, S. (1997a). Reading in Arabic orthography: The effects of vowels and context on reading accuracy of poor and skilled native Arabic readers. Reading and Writing : An Interdisciplinary Journal 9: 65-78. Abu-Rabia, S. (2001). The role of vowels in reading Semitic scripts: Data from Arabic and Hebrew. Reading and Writing: An Interdisciplinary Journal, 14, 39–59. Abu-Rabia, S. (2007). The role of morphology and short vowelization in reading Arabic among normal and dyslexic readers in grades 3, 6, 9, and 12. Journal of Psycholinguistic Research. Vol 36(2), 89-106. Abu-Rabia, Share and Mansour, (2003). Word recognition and basic cognitive processing among reading disabled and normal readers in Arabic. Reading and Writing: An Interdisciplinary Journal, 16, 423-442. Abu-Rabia, S., and Siegel, L. S. (1995). Different orthographies, different context effects: The effects of Arabic sentence context on skilled and poor readers. Reading Psychology 16, 1–19. Abu-Rabia, S. and Taha H. (2006). Phonological Errors in Arabic Spelling Across Grades 19. Journal of Psycholinguistic Research, 35, 167-188. Adams, M. J. (1981). What good is orthographic redundancy? In O. J. L. Tzeng and H. Singer (Eds.), Perception of print (pp. 197-221). Hillsdale, NJ: Erlbaum. Adams, M.J. (1990). Beginning to read: thinking and learning about print, Cambridge, MA: Mit Press. Adar, L., Bloom-Kulka, S., Nir, R. Meisler, R. and Weil, T. (1987). Detection of difficulties in reading comprehenstion. Jerusalem: The institute of educational furthering research, School of education, Hebrew university. Al-Mannai, H. and Everatt, J. (2005). Phonological processing skills as predictors of literacy amongst Arabic speaking Bahraini school children. Dyslexia, 11, 269-291. Azaizeh, K., (1997). Reading fluency and comprehension in the Arab elementary schools in Israel: mediating factors. Submitted. Azzam, R. (1984). Orthography and reading of the Arabic languge. In J. Aaron and R. M. Joshi (Eds.), Reading and writing disorders in different orthographic systems (pp. 1–29). Kluwer Academic. Azzam, R. (1993). The nature of Arabic reading and spelling errors of young children. Reading and Writing, 5, 355–385. Badry, F. (1983). Acquisition of lexical derivational rules in Moroccan Arabic: Implications for the development of standard Arabic as a second language through literacy. Ph.D. Dissertation, university of California, Barkeley. Bashi, I. , Sorel, R., Daniel D. (1981). The educational achievement of the Arab elementary school in Israel. Jerusalem: The school of education, Hebrew university, 8-138.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Vowels in Semitic Alphabet Languages

163

Bentin and Ibrahim, 1996. New Evidence for Phonological Processing During Visual Word Recognition: The Case of Arabic. Journal of Experimental Psychology, 22 (2), 309-323. Bentin S, Bargai N, Katz L. (1984), Orthographic and phonemic coding for lexical access: evidence from Hebrew: Journal of Experimental Psychology: Learning, Memory, and Cognition, 10(3), 353-368. Berman, R. A. (1978). Modern Hebrew structure. Tel Aviv, Israel: University Publishing. Boudelaa, S. and Marslen-Wilson, W.D. (2001). Morphological units in the Arabic mental lexicon. Cognition, 81, 65-92. Boudelaa, S. and Marslen-Wilson, W.D. (2005). Discontinuous morphology in time: Incremental masked priming in Arabic. Language and Cognitive Processes. 20, 207-260. Bruck, M. (1990). Word-recognition skills of adults with childhood diagnoses of dyslexia. Developmental Psychology, 26, 439–454. Elbeheri, G. and Everatt, J. (2007). Literacy ability and phonological processing skills amongst dyslexic and non-dyslexic speakers of Arabic. Reading and Writing, 20, 273294. Elbeheri, G., Everatt, J., Reid, G. and Al-Mannai, H. (2006). Dyslexia assessment in Arabic. Journal of Research in Special Educational Needs, 6, 143-152. Eviatar, Z. and Ibrahim, R. (2004). Morphological and orthographic effects on hemispheric Processing of nonwords: A Cross-linguistic comparison. Reading and Writing:17, 691705. Eviatar, Z., and Ibrahim, R. (2000). Bilingual is as bilingual does: Metalinguistic abilities of Arabic-speaking children. Applied Psycholinguistics, 21, 451–471. Eviatar, Z., Ibrahim, R., and Ganayim, D. (2004). Orthography and the hemispheres: Visual and linguistic aspects of letter processing. Neuropsychology, 16 (3) 322-326. Feitelson D., Goldstein, Z., Iraqi, G. and Share, D. (1993). Effects of listening to story reading on aspects of literacy acquisition in a diglossia situation, Reading research quarterly, 28, 71-79. Ferguson C.A. (1959). “Diglossia”. World, 15, 325-340. Frith, U. (1985). Beneath the surface of developmental dyslexia. In K.E. Patterson, J.C. Marashall and M. Coltheart (Eds.), Surface dyslexia. London: Lawrence Erlbaum Associates. Frost, R. (1995). Phonological computation and missing vowels: Mapping lexical involvement in reading. Journal of Experimental Psychology: Learning, Memory, and Cognition, 21, 398–408. Frost, R., Katz, L., and Bentin, S. (1987). Strategies for visual word recognition and orthographical depth: A multilingual comparison. Journal of Experimental Psychology: Human Perception and Performance, 13, 104–114. Ibrahim, R., and Aharon-Peretz, J (2005). Is literary Arabic a second language for native Arab speakers?:Evidence from a semantic priming study. The Journal of Psycholinguistic Research, 34(1), 51–70. Ibrahim, R., and Eviatar, Z. (2001). The effects of meta linguistic awareness on the acquisition of reading: A cross-language comparison. Manuscript submitted for publication. Ibrahim, R., Eviatar, Z. and Aharon-Peretz, (2007). Metalinguistic awareness and reading performance: A cross language comparison. Journal of Psycholinguistic Research, 36(4), 297-317.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

164

Raphiq Ibrahim

Ibrahim, R., Eviatar, Z., and Aharon-Perez, J. (2002). Do the characteristics of Arabic orthography slow it’s cognitive processing? Neuropsychology 16 (3) 322-326. Jensen, H. (1970). Signs, sympol and script. London: Goerge Allen and Unwin. Kais, I., (1978), Verbal behavior and socialization, Eyunim behinuch, 19. Haifa, pp. 131-152. Liberman, A. M. (1992). The relation of speech to reading and writing. In R. Frost and L Katz (Eds.), Orthography, phonology, morphology, and meaning (pp. 193-210). Amsterdam, The Netherlands: North Holland Elsevier. Maamouri, M. (1998). Language education and human development: Arabic diglossia and its impact on the quality of education in the Arab region”. The Mediterranean Development Forum. Washington, DC: The World Bank. Mahfoudhi, A. (2007). Roots and patterns in Arabic lexical processing. In E. Benmamoun (Ed.), Perspectives on Arabic linguistics XIX. Amsterdam: Benjamins. Massaro, D. W., Venezky, R. L., Taylor, G. A. (1979). Orthographic regularity, positional frequency, and visual processing of letter strings: A reply to Krueger's comments. Journal of Experimental Psychology, 108(1), 131-132. Navon, D., and Shimron, J. (1981). Does word naming involve grapheme-to-phoneme translation? Evidence from Hebrew. Journal of Verbal Learning and Verbal Behavior, 20, 97–109. Navon, D., and Shimron, J. (1985). Source and function of phonological codes in processing written Hebrew (Final report to the US–Israeli Binational Science Foundation). Haifa, Israel: University of Haifa, Institute of Information Processing and Decision Making. Perfetti, C. A., and Bell, L. (1991). Phonemic activation during the first 40 ms of word identification: Evidence from backward masking and masked priming. Journal of Memory and Language, 30, 473-485. Perfetti, C.A. (1985). Reading ability. New York: Oxford University Press. Pollatsek , Bolozky , Well and Rayner (1981). Asymmetries in the perceptual span for Israeli readers. Brain and Language, 14(1), 174-180. Saiegh-Haddad, E. (2004). The impact of phonemic and lexical distance on the phonological analysis of words and pseudowords in a diglossic context. Applied Psycholinguistics, 25, 495-512. Shalabna S., Ibrahim, R. and Eviatar. Z. (Submitted). What affects vowels perception among skilled native Arabic readers?. Shimron, J. (1993). The role of vowels in reading: A review of studies of English and Hebrew, Psychological Bulletin, 114, 52–67. Shimron, J. (1999). The role of vowel signs in Hebrew: Beyond word recognition. Reading and Writing: An Interdisciplinary Journal. 11, 301–319. Smolka, E. and Eviatar, Z. (2006). Phonological and orthographic visual word recognition in the two cerebral hemispheres: Evidence from Hebrew . Cognitive Neuropsychology, 23(6), 972 – 989. Stanovich , K.E., Cunningham, A. andCramer, B. (1984). Assessing phonological awareness in kindergarten children: Issues of task comparability. Journal Experimental Child Psychology, 38, 175-190. Stanovich, K. E. and Feeman, D. G. (1981). A longitudinal study of sentence context effects in second-grade children: Tests of an interactive-compensatory model. Journal of Experimental Child Psychology, 32(2), 185-199.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Vowels in Semitic Alphabet Languages

165

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Stanovich, K. E. (1980). Toward an interactive-compensatory model of individual differences in the development of reading fluency. Reading research quarterly, 16, 32-71. Stein, J. and Walsh, V. (1997). To see but not to read; the mango-cellular theory of dyslexia. Trends in Neurosciences, 20(4): 147-152. Taouk, M., and Coltheart, M. (2004). The cognitive processes involved in learning to read in Arabic. Reading and Writing: An Interdisciplinary Journal, 17, 27–57. Taylor, I. (1980).The Korean writing system: an alphabet? A syllabary? A logography? In: P.A. Kolers, M.E. Wrolstad, and H. Bouma (Eds.), Processing of Visual Language (vol. 2, pp. 67-82). New York: Plenum Press. Van Orden, G. C. (1991). Phonological mediation is fundamental to reading. In D. Besner and G. Humphreys (Eds.), Basic Processes in Reading: Irtsual Word Recognition (pp. 77103). Hillsdale, N J: Edbaum. Venezky, R. I. (1970). The structure of English orthography. The Hague, The Netherlands: Mouton. Vipond, D. (1980). Micro and macroprocesses in text comprehension: Journal of Verbal Learning and Verbal Behaviour, 19, 276-296.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved. Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

In: Bilinguals: Cognition, Education and Language Processing ISBN: 978-1-60741-710-1 Editor: Earl F. Caldwell, pp. 167-179 © 2010 Nova Science Publishers, Inc.

Chapter 8

METHODOLOGICAL ISSUES IN RESEARCH ON BILINGUALISM AND MULTILINGUALISM Lilian Cristine Scherer1, Rochele Paz Fonseca2 and Ana Inés Ansaldo3 1

2

Universidade de Santa Cruz do Sul (UNISC), Brazil Pontifícia Universidade Católica do Rio Grande do Sul (PUCRS), Brazil 3 University of Montreal, Centre de Recherche Institut Universitaire de Gériatrie de Montréal (CRIUGM), Canada

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Introduction Research on language processing in the brain has increased and has become prominent in the past decades, fostered mainly by the adoption of neuroimaging techniques. For instance, language studies on the lesion paradigm, with which neuropsychology was born, have considerably improved by the use of refined neuroimaging techniques (Price, Noppeney & Friston, 2006). Some of the techniques mainly used in research on language processing are functional magnetic resonance imaging (fMRI), positron emission topography (PET), magnetoencephalography (MEG), electroencephalography (EEG), and functional nearinfrared spectroscopy (fNIRS), among others, which have allowed researchers to acquire in vivo brain images of the order of seconds or even milliseconds (for a review please see Démonet, Thierry and Cardebat, 2005). The availability of these increasingly refined techniques, allied to the research interest in understanding language processing and its related cognitive components has led to the emergence of a new research field, named Neuropsycholinguistics. In this field, a growing topic of study is bi/multilingualism and its neural correlates. Neuroimaging research on bi/multilingualism has investigated both comprehension and production, by adopting a variety of experimental designs and criteria for participants’ recruitment. This diversity, although informative, has represented a challenge for drawing general conclusions from the data.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

168

Lilian Cristine Scherer, Rochele Paz Fonseca and Ana Inés Ansaldo

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

The term “bilingual” has been defined in distinct ways, depending on the researcher´s concept on the degree of language mastering considered to be necessary for a person to be characterized as such. The definition here to be adopted is the one proposed by Grosjean (1994), according to which the term ‘bilingual’ refers to an individual who uses two or more languages or dialects in his or her everyday life, regardless of the context of use. Whenever this definition is taken into account, it is possible to conceive that more than half of the world´s population can be considered bilingual (Giussani, Roux, Lubrano, Gaini & Bello, 2007). Nowadays, it is common also to choose the word “multilingual” to characterize individuals who speak more than two languages. This word will be used in this sense throughout the chapter, although the word “bilingual” is preferably used by some authors. In this scenario of multilingualism, understanding how languages coexist in the brain is of relevant importance, since this type of knowledge may support studies on first and second language acquisition and neuropsychological rehabilitation of language and communication impairment, and language teaching, among others. Moreover, it may contribute to enhance our understanding about the processing of executive functions in cognitive neuropsychology, such as the switching component, which has been investigated in bilingual participants (Von Studnitz & Green, 2002). Concerning research on bi/multilingualism, a rigorous methodological control needs to be taken by researchers in order to guarantee data validity and reliability, since several cooccurring factors may influence results and their interpretation. If not controlled, these interferences may confound data analyses and theoretical discussions. The aim of this chapter is to discuss important methodological issues regarding research on neurologically preserved bilingual and multilingual populations. It includes the variables to be considered in methodological aspects ranging from individual characteristics to task design and presentation, among others.

1. Why Care about Methodological Issues in Research on Bilingual and Multilingual Language Processing? From the most recent neuroimaging studies on language processing in the brain, it has become evident that a large circuitry involving several brain areas is responsible for language processing, not only the classical areas of Broca and Wernicke regions as it has been thought in the past (Démonet, Thierry & Cardebat, 2005; Joanette et al., 2008). Moreover, these studies have shown that it is a hard enterprise trying to dissociate language processing from other cognitive abilities, such as memory, attention and inhibitory control, since language is controlled by and several times controls and organizes these cognitive abilities in an interdependent way. This is one of the most robust reasons to justify the importance of language studies—language sustains and conveys most of our reasoning and several communicative processes, which represent integrating and distinguishing characteristics of human beings. If the study of language processing in the brain is so relevant, the investigation of concomitant languages´ processing can be even more illustrative of how language systems are organized and function in the brain. This is the case because each language can be studied as an entire entity, or still, it can be analyzed in its constitutive parts, namely phonetics and

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Methodological Issues in Research on Bilingualism and Multilingualism

169

phonology, syntax, semantics, discourse and pragmatic components. Furthermore, it may be analyzed at the word, sentence or discourse level, each one with its specific degree of complexity. Considering the complexity of this arena and the recency of these studies, it is plausible to assume that further investigations should widely be implemented in order to reach a thorough knowledge of how languages and their components interconnect, interact, are stored and retrieved in the most varied situations of language(s) use. As seen from above, language studies are intrinsically complex, due to the complexity of language itself. However, besides linguistic considerations, which will not be fully described in this chapter, other aspects are as well relevant in research in bi- and multilingualism. One of them relates to the participants´ profile. Several individual characteristics may affect data collection and analyses. Similarly, choices taken by the researcher regarding task design and its administration may as well interfere in the results. These aspects will be discussed in the section that follows.

2. Fundamental Methodological Issues to Be Considered in Research on Bi- and Multilingualism

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

2.1. Issues Related to the Bi/Multilinguals’ Individual Characteristics Concerning the role of individual characteristics in language(s) processing by bilingual and multilingual populations, some of the most relevant aspects to be taken into consideration in order to look for sample homogeneity are: 1) age and form of L2/L3 acquisition, 2) chronological age, 3) language mastering in the four abilities (reading, writing, speaking and listening comprehension), 4) language use, 5) socio-cultural aspects, and 6) general cognitive abilities. All of these issues may interact with the behavioral and neuroimaging results to be obtained. Information about the aspects listed above, which characterize the research population, may be gathered by the application of questionnaires. Important items assessed by questionnaires usually are 1) a) the number of months/years using the languages, b) the learning method (informal or formal), and the number of years of formal education received in the languages, c) the degree of language exposure (at home, school, work, through media), and d) the degree of proficiency in oral and written comprehension and expression, within each language (as an illustration, please see the questionnaire proposed by Marian, Blumenfeld and Kaushanskaya, 2007). Questionnaires may as well test the participant’s performance in grammaticality judgment tasks, reading fluency, productive vocabulary and sound awareness. Each of the items listed above, whose control is essential for seeking participants’ homogeneity, will be now explained.

2.1.1. Age and Form of L2 Acquisition Age and form (whether in a formal or informal context) of acquisition are very much related, since early L2 learning tends to occur in a natural environment, whereas late L2 learning (after infancy) generally occurs in formal, academic settings. This distinction has led to the debate on learning versus acquiring a language. Acquiring a language would be

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

170

Lilian Cristine Scherer, Rochele Paz Fonseca and Ana Inés Ansaldo

internalizing it unconsciously, effortlessly, as in the case of immersion settings, whereas learning would be consciously internalizing the L2 (or L3, L4, and others), by means of learning language rules and vocabulary. However, this distinction is not clear, since in several situations the learner is not able to distinguish whether s/he acquired or learned a specific item. Another way of demonstrating how vague and subtle this distinction is comes from the debate on the role of declarative versus procedural memory systems. Some research has suggested that incidental L2 learning would rely on procedural, implicit memory, while a formally learned L2 would rely more heavily upon declarative, explicit memory (for a debate on this topic, including the impact of memory systems in language recovery from aphasia, please see Paradis, 2004). This hypothesis has been challenged by studies which suggest that L2 use and practice may allow the interaction between these two memory systems (Ullman, 2004). Much research has tried to analyze to what extent a more incidental versus a more formal L2 acquisition would rely on implicit or explicit memory systems, in a sense that explicit memories may turn into implicit ones by the achievement of an automatic use (McClelland, McNaughton & O´Reilly, 1995). Specifically considering the role of age of acquisition in the organization and functioning of the neural substrates for language processing, evidence has not been convergent. For instance, studies at the single-word level, using word completion (Chee, Tan & Thiel, 1999), semantic judgment (Chee, Hon, Lee & Soon, 2001), naming (Hernandez, Martinez & Kohnert, 2000), and noun generation (Briellmann et al., 2004) tasks, have reported overlapping activations for L1 and L2, whatever the age of acquisition is. Conversely, studies developed by Kim and colleagues (1997) and Wartenburger and colleagues (2003) report an impact of age of L2 acquisition on the neural substrate for L2 processing. To illustrate, Wartenburger and colleagues (2003) examined early and late bilinguals’ performance in a syntactic and semantic judgment task; the authors report that late bilinguals show more extended activations in Broca’s area, the inferior frontal gyrus (BA44/6) and the right hemisphere homologous region than early bilinguals in the syntactic task, whereas no difference in the activation patterns was observed across groups in the semantic judgment task. Thus, the authors suggest that late L2 acquisition will have an impact on the neural substrate of morphosyntax but not on the circuitry sustaining semantic processing. Finally, Bialystok (1997) alerts to the importance of a careful analysis of different factors affecting bilinguals’ language representation before attributing research results to age of acquisition. She emphasizes the necessity of considering three determinant issues, namely 1) the correspondence between structures in the L1 and in the L2 (the more similar, the easier the assimilation of the L2 will be, independently of the age factor); 2) the length of residence (the amount of time spent using the L2), and 3) the amount of formal education in the L2. These factors have barely been considered in studies on the age of acquisition factor. Fledge et al. (1999) also emphasize the role of formal education in L2 and of the exposure to and use of the second language in language organization in the bilingual brain. Their research challenges the notion of the existence of a maturational critical period. From the discussion above, it is clear that no conclusive evidence has been found yet in terms of the role of age and form of acquisition in languages’ organization and functioning in the bi/multilingual brain. Therefore, it is important to control group sampling considering these two factors in order to analyze data in the light of the different results and discussions brought by the literature so far.

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Methodological Issues in Research on Bilingualism and Multilingualism

171

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

2.1.2. Chronological Age Neurofunctional studies on brain circuitry for cognitive processing have shown different patterns of brain activation for younger adults and elderly adults when executing the same task. Several theories have aimed to explain these differences, based on varying assumptions according to the changes registered in the aging brain. Thus, some assumptions to explain the changing neurofunctional organization in aging are: 1. a growing involvement of frontal areas, possibly due to the fact that frontal regions seem to be the first areas affected by agerelated changes (Tisserand, 2002); 2. a posterior-anterior shift in aging (PASA), a theory which postulates that there is a migration from posterior, including parietal and occipital regions, to frontal areas (Davis et al., 2007), and 3) a de-differentiation process, observable by the sharing of labor by the hemispheres to solve tasks which were previously developed mainly in one hemisphere; in other words, the hemispheric asymmetry reduction in older adults – the HAROLD model (Cabeza, 2002). This pattern has been specifically observed with elderly adults with high performance in cognitive tasks. The discussion reported above relates to brain organization in a monolingual elderly population. Very little research has addressed the neurobiological and neurofunctional bases of elderly bilinguals’ language processing so far. Most data come from behavioral studies, such as those recently developed by Bialystok and Craik (2007) and Bialystok, Craik and Ryan (2006). These studies have shown the impact of bilingualism mainly on executive functions, representing an advantage in terms of the control of executive functions by bilingual elderly samples in the comparison to monolingual elderly controls. Thus, the complexity of dealing with two languages seems to improve the processing of executive components. Moreover, Bialystok, Craik and Freedman (2007) report a four-year delay in the appearance of the first symptoms indicating Alzheimer’s disease in elderly bilinguals, in comparison to monolingual Alzheimer’s elderly patients. These data are in accordance with the hypothesis of the establishment and reinforcement of a greater number of connections in the bilinguals’ brain (Giussani et al., 2007), and suggest that bilinguals would have access to a cognitive reserve which could compensate for the early signs of healthy and unhealthy aging. The literature on cognitive reserves in monolinguals has pointed to the importance of years of formal education, among other factors, in enlarging these reserves and, thus, in aiding the prevention of neurological signs of aging (Stern, 2009). It may be possible to postulate, therefore, that formal education may as well have its impact on building cognitive reserves in elderly bilinguals. Considering the evidence presented above, the necessity of taking into consideration the changing patterns of brain activation as revealed by neuroimaging studies and the specificities in behavioral results obtained by an elderly population becomes clear. All assumptions made by the studies need to be drawn in the light of theories on the aging brain, and have to be carefully drawn specially in an aging bilingual population, a research field which needs to be further developed.

2.1.3. Proficiency in the Four Linguistic Abilities (Reading, Writing, Speaking and Listening) Evidence has undoubtedly shown that the proficiency attained in the second and in other languages is a relevant factor to determine languages distribution and functioning in the

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

172

Lilian Cristine Scherer, Rochele Paz Fonseca and Ana Inés Ansaldo

bi/multilingual brain (Muñoz & Marquardt, 2008; Perani & Abutalebi, 2005). Not only the overall proficiency in the languages is crucial, but also the proficiency across language levels in particular (i.e., in speaking, reading, writing and oral comprehension). Brain regions participate differently in the several stages a language learner passes towards languages’ mastering. For instance, in early acquisition stages, parahipocampal and right hemisphere regions are activated in L2 processing (Paradis, 2004). Wider and more distributed right hemisphere activation has been reported, particularly in frontal areas, in participants with lower levels of L2 processing (Dehaene et al, 1997; Perani et al, 1996); conversely, highly proficient bilinguals have shown an overlap of L1 and L2 networks in the left hemisphere (Perani et al., 1998). In the review article organized by Abutalebi and Green (2007), they report higher L2 related activations with low proficiency participants, not only in regions traditionally involved in L1 processing, but also in regions responsible for the ‘cognitive control’, such as the prefrontal cortex (BA 9, 46, 47), the anterior cingulate cortex, and the inferior parietal cortex. According to the authors, these activation patterns reflect monitoring processes aiming at inhibiting incorrect responses, and filtering out unnecessary information available in the environment. Thus, in terms of methodological issues, the observance of the proficiency level attained by the research participant in each of the languages whose processing is to be analyzed should be taken into account, since the degree of proficiency determines specific recruitment of brain circuitry, as suggested by empirical evidence. Conversely, the type of linguistic ability to be measured (whether reading, writing, speaking or oral comprehension ability) needs as well to be considered. This is the case because bilinguals and multilinguals may not be equally proficient in each one of these levels, in the same way as monolinguals are not equally proficient in them. This variability may result from the varying levels of exposure and use of each of the abilities in a daily basis, as well as from varying levels of motivation to improve one or other ability in special. Therefore, an accurate assessment of the research participants’ ability in the field to be tested (reading, writing, speaking or listening comprehension ability) cannot be neglected, in order to assure homogeneity of the experimental sample. What has been the case in several experiments is that researchers adopt self-report questionnaires requiring the participant to self-evaluate his/her competence in each of these abilities. This is probably not a reliable measure if taken in isolation, without the concomitant administration of a valid language proficiency test, since self-evaluations may be subjective and not portrait the real proficiency level, as in the case of highly proficient participants whose high demands on their own competence may lead them to attribute to themselves lower scores than those attributed by lower level proficiency participants whose sense of self ability/achievement is higher.

2.1.4. Language Exposure and Use Neglected in the majority of the research on bi- and multilingualism implemented in the past decades, the role of language(s)’ exposure and effective use in the neural substrate for languages processing seems to be relevant, together with motivational factors (Abutalebi, Cappa & Perani, 2001; Bialystok et al., 2005, Hellermann & Vergun, 2007, among others). A language which is frequently used in daily activities turns to be more easily and automatically retrieved and produced, with a lowered threshold level (Green, 1998).

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Methodological Issues in Research on Bilingualism and Multilingualism

173

Another important methodological aspect to be taken into account regarding the participants’ profile is the amount of years of formal education received in the target language(s) (Parente, Fonseca, & Scherer, 2008). Bi/multilinguals who attended courses addressed in the language(s) to be investigated tend to have higher ability in tasks which are normally developed in formal educational settings, such as writing and reading abilities, as well as a higher facility in answering tasks which measure these abilities (Fledge, 1999).

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

2.1.5. Socio-cultural Aspects Related to the Bi/Multilingual Sample In order to achieve a high sample homogeneity, along with the control of the factors described above, other issues to be observed are bi/multilinguals’ socio-cultural background and socioeconomic status (Bialystok, 2006). Therefore, when analyzing, for example, the bilingual advantage in executive control, possible differences in socioeconomic status and cultural background must as well be taken into account. Claiming that these factors have not always been observed in bi/multilingual research, Morton and Harper (2007) developed a study in which they tried to show that the control of ethnicity and socioeconomic status would attenuate the bilingual advantage. To do so, they administered the Simon task to a sample of 17 mono- and 17 bilingual children aged 6–7. Bilingual and monolingual children performed similarly, but children with higher socioeconomic status outperformed their homologous groups of lower socioeconomic status. However, Bialystok (2009) claims that their study did not exactly replicate previous research showing a bilingual advantage, since the number and the ages of their sample were not comparable to previous studies; in addition, the tasks parameters were not the same, as stated by the author. To sum up, it becomes evident that bi/multilingual participants’ samples need to be recruited with a concern to their socioeconomic and cultural conditions, since these aspects directly relate to the amount of exposure to and experience with linguistic material. Moreover, caution needs to be taken when pointing to ethnic aspects in this type of research, so that data do not serve as a tool to lead to prejudice and to misleading conclusions. This has already occurred in the past, when researchers were not always concerned with samples’ nature and their proficiency and language use, variables straightly linked to cultural aspects. As a consequence, for many times bilingual immigrants were segregated and treated with prejudice, since their performance in cognitive tasks, sometimes measured in their nonproficient language, was taken as an indicative of their performance in IQ scales.

2.1.6. General Cognitive Abilities Associated to Language Processing One of the most remarkable human cognitive abilities, language processing relies upon an array of cognitive functions, including memory, attention and inhibitory control (for instance, the control for inhibiting erroneous inferences) and cognitive flexibility (the revision and correction of a given hypothesis). Individual differences may modulate the participants’ ability in recruiting each one of these capacities. Therefore, an important methodological concern should be to make sure that discrepancies in performance in language tasks are not motivated by differential abilities in other cognitive functions involved in the processing of the experimental task. Ideally, correlations between neuroimaging and behavioral data should be used to ponder the influence of a given cognitive ability on a particular language task, for

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

174

Lilian Cristine Scherer, Rochele Paz Fonseca and Ana Inés Ansaldo

instance, the impact of working memory capacity on the participant’s performance in syntactic processing, depending upon the level of syntactic complexity (Suh et al., 2007). Thus, a manner of verifying a possible influence of these and other cognitive abilities on language tasks would be the administration of neuropsychological batteries, whose results could be correlated with the experimental data of mono and bi/multilingual populations.

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

2.2. Practical Issues in Implementing Data Collection Beyond the influence exerted by participants’ individual aspects in data drawn from a bi/multilingual population, several issues related to research implementation concerning data collection may as well influence the data to be obtained. More specifically, some fundamental methodological aspects to be observed are: 1) language typology (considering types of language groups); 2) the level of linguistic complexity (word, sentence or discourse level); 3) psycholinguistic aspects in task design (considering psycholinguistic criteria); 4) the mode of stimuli presentation (whether visual or aural); 5) the bi/multilingual condition of the examiner; 6) session organization for task application and its role in language mixing/shifting, and 7) the adoption of monolingual controls. Each one of these issues will now be focused. Considering types of languages in terms of their original family groups and, thus, the degree of overlap between their structural properties, it is important to consider that for research participants with lower L2 proficiency certain experimental tasks may be more easily solved whenever the target languages belong to the same linguistic family field, as it is the case with languages such as Italian, Spanish, Portuguese and French. Contrarily, native speakers of ideographic languages such as Japanese and Chinese may not benefit from this advantage when reading, let us say, in English, for example. In terms of neuroimaging, a very reduced number of studies on bi/multilingualism has focused on neuroimaging patterns of activation considering language typology. One study that illustrates this type of investigation is the one developed by Tan and collaborators (2003). They investigated the neural mechanisms of reading in Chinese (as L1) and in English (as L2), two languages with different phonological and orthographic systems, in order to analyze the effect of L1 typology on the acquisition of an L2. Differently from orthographic languages, written Chinese is based on logograms, single characters that represent a word. Their fMRI data have shown the recruitment by bilinguals of similar areas for reading Chinese and English words – left middle frontal and posterior parietal gyri, two cortical regions involved in spatial information representation, spatial working memory and coordination of cognitive resources, such as the central executive system, while different regions were recruited by English monolinguals. The conclusion drawn from the study was that the bilinguals were adopting their L1 strategies to L2 reading, and thus, that language experience tunes the cortex. Other research, this time with PET but with a similar aim, conducted by Klein, Milner, Zatorre, Zhao and Nikelski (1999), compared verb generation and word repetition in Mandarin (L1) and English (late acquired L2). The researchers observed a shared circuitry for both languages, in both types of tasks. Thus, their results do not corroborate those brought by Tan and colleagues (2003), who reported the occurrence of different activation patterns for word processing in both languages. The divergent results brought by these two illustrative studies demonstrate the necessity of further research.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Methodological Issues in Research on Bilingualism and Multilingualism

175

The complexity of the linguistic task is also something to be considered when analyzing behavioral and imaging data, since the types of cognitive demands and of linguistic abilities vary as a function of the task (Démonet, Thierry & Cardebat, 2005; De Groot, Borgwaldt, Bos & Van den Eijnden, 2002). Thus, the type of processing demanded for word processing is different from the one required for the sentence level, or even for discourse processing. In this final level, for instance, higher degrees of inferences are active in the comparison to word processing, when it comes, for instance, to pragmatic abilities for interpreting the context in an online way. Another way of manipulating task complexity is varying the instruction demands. For instance, the passive reading of a text recruits brain regions in a different way than tasks demanding the research participant to answer true or false questions or to correct wrong statements about the text (Tzourio, Nkanga-Ngila & Mazoyer, 1998). The same applies to overt or covert syntactic or semantic judgments of sentences. Thus, whenever designing an experiment, researchers need to be aware of the fact that the results to be obtained vary as a function of the tasks demands and of the linguistic level to be analyzed (whether word, sentence or discursive-pragmatic level). Psycholinguistic aspects in task design are as well a matter of concern for task designers of studies on bi/multilingual populations. A rigorous psycholinguistic control of variables while designing the linguistic task is fundamental. In the case of the sentence level, for example, these variables include sentence structure (single, compound or complex sentences, direct or indirect order, passive or active voice), semantic and syntactic processing demands (literal or connotative meanings, metaphors, the number and types of verbal complements), among others; at the word level, word frequency, imageability, word length and prototypicality are some of the aspects known to modulate processing (Friederici, Fiebach, Schlesewsky, Bornkessel & von Cramon, 2006). All of these variables certainly affect monolingual research as well; however, in the case of bi/multilingual research they are even more crucial to be balanced, since more than one language is in general investigated, with their data compared. The fourth aspect pointed out above as one of the aspects to be observed while designing linguistic tasks for research with a bi/multilingual population is the mode of stimuli presentation. Neuroimaging evidence has already demonstrated that brain activity varies as a function of the tasks’ presentation mode (Démonet, Thierry & Cardebat, 2005). Thus, occipital areas tend to be recruited in visually presented tasks, due to the nature of these tasks, while orally presented tasks may activate temporal regions, since these areas are involved in listening activities. This seems to be too simple, but it cannot confound researchers while drawing conclusions from the data, since some brain regions may typically be involved in oral or visual stimuli processing, independently of the linguistic demands imposed by the activity. The fifth aspect listed above is the bi/multilingual condition of the examiner. Although very much neglected, the language status of the tester, the person in charge of conducting the experiment, who directly interacts with the participant, may be taken into account. More specifically, according to Grosjean (1999), data brought by experimental studies (including neuroimaging ones) may be affected by the fact of the tester himself/herself being or not a bi/multilingual. This may interfere in the results because as soon as the participant notices the possibility of communicating with the tester in a determined language, the likeliness of language switching (changing from one language to the other) and mixing (borrowing structural items from one language to the other(s)) increases. In a monolingual situation, when the tester masters only one of the languages whose processing is going to be examined,

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

176

Lilian Cristine Scherer, Rochele Paz Fonseca and Ana Inés Ansaldo

or omits the fact of being a speaker of those languages, mixing and switching are less likely to occur. Examiners’ features are deeply explored by Ardila (2005), who assumes that the examiner´s culture can as well influence on the results obtained by cognitive instruments. The sixth aspect mentioned above relates to session organization for task application. Before starting data acquisition, the researcher needs to deeply think on the consequences of testing more than one language during the same session. Sometimes, depending on the aim of the study, this may be important, when, for instance, s/he intends to investigate differences in the threshold level for semantic activation as a function of word frequency and language proficiency. However, in some cases, setting two acquisitions apart, with an interval of some weeks, may be important if, for instance, the same stimuli are used in different languages, to reduce memory effects on the task, since the participant may remember the stimuli and this may influence in the speed and accuracy in his/her performance in the other language(s). Finally, the choice of whether adopting or not monolingual control groups should be considered. Since several studies developed with bi/multilingual populations do not have comparable studies in each of the monolingual native languages' counterparts, almost no L1 baseline data are available. Thus, in order to allow the researcher to establish a comparison between native speakers’ performance in the L1 and non-native speakers’ performance in the target language, it may be advisable to have monolingual controls, as suggested by Vaid and Hull (2002). There are still other important aspects to consider when doing research with bi/multilingual populations, such as the case of special language systems, like Braille and sign language, which can as well be considered to be a second (or first) language of a part of the world’s population. Similarly, the study of bi/multilingualism within an illiterate population should be carefully analyzed, since illiteracy studies on monolingual populations have demonstrated the occurrence of specific patterns of brain activation in these groups, representing a factor that would certainly modulate brain circuitry in a bi/multilingual population as well.

Conclusions Bilingualism and multilingualism are special arenas to study linguistic and cognitive related aspects underlying the brain circuitry in human communication, studies fostered lately by the application of ever refining neuroimaging techniques. Although much research has already been developed, it is still difficult to draw conclusive findings even on fundamental issues related to language organization and functioning in the brain. One of the reasons for certain discrepancies in the data found so far is the adoption of several types of methodological orientations, including the chosen language aspects, the participants’ profiles and the task designs, aspects which, in their turn, affect several cognitive processes hardly dissociable from language processing. Consequently, this methodological heterogeneity influences behavioral and neuroimaging data. These divergences turn the postulation of theoretical explanations into a hard task. Thus, further studies are still needed in order to achieve a deeper understanding of the interaction between languages and their correlated cognitive functions in the architecture of the bi/multilingual brain. Moreover, straight methodological measurements have to be adopted, in order to obtain consistent, reliable, replicable data, since several variables which characterize these

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Methodological Issues in Research on Bilingualism and Multilingualism

177

populations have to be taken into account. A straightforward developed research may bring data which will, in turn, provide evidence for neuropsychology and rehabilitation sciences on the specificities of healthy (and impaired) bi/multilingual language processing, improving assessment and training methods specifically designed to these populations. In this chapter, we aimed to focus on the most relevant aspects to be considered in terms of methodological concerns in bi/multilingual research, but we certainly did not cover all of them, since further research may bring other contributions to the field. However, an assumption can be postulated: behavioral and neuroimaging studies will be essential for more homogenous and precise results regarding language processing in the healthy and injured brain, as well as regarding the relation between linguistic and other cognitive components.

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

References Abutalebi, J., Cappa, S. F., & Perani, D. (2001). The bilingual brain as revealed by functional neuroimaging. Bilingualism, 4, 179-190. Abutalebi, J., & Green, D. (2007). Bilingual language production: The neurocognition of language representation and control. Journal of Neurolinguistics, 20, 3, 242-275. Ardilla, A. (2005). Cultural values underlying psychometric cognitive testing. Briellmann, R. S., Saling, M. M., Connell, A. B., Waites, A. B., Abbott, D. F., & Jackson, G. D. (2004). A highfield functional MRI study of quadri-lingual subjects. Brain and Language, 89(3), 531-542. Bialystok , E. (1997). The structure of age: In search of barriers to second-language acquisition. Second Language Research, 13, 116-137. Bialystok, E., Craik, F. I., Grady, C., Chau, W., Ishii, R., Gunji, A., et al. (2005). Effect of bilingualism on cognitive control in the Simon task: Evidence from MEG. Neuroimage, 24, 1, 40-49. Bialystok, E. (2006). The impact of bilingualism on language and literacy development. In T. Bhatia, & W. Ritchie (Eds.). The handbook of bilingualism (pp: 577-601). New York: Blackwell. Bialystok, E., Craik, F. I., & Ryan, J. (2006). Executive control in a modified anti-saccade task: effects of aging and bilingualism. Journal of Experimental Psychology: Learning Memory, and Cognition, 32, 1341-1354. Bialystok, E., & Craik, F. I. (2007). Bilingualism and naming: implications for cognitive assessment. Journal of the International Neuropsychological Society, 13, 209-211. Bialystok, E., Craik, F. I., & Freedman, M. (2007). Bilingualism as a protection against the onset of symptoms of dementia. Neuropsychologia, 45, 459-464. Bialystok, E. (2009). Claiming Evidence from Non-evidence: A Reply to Morton and Harper. Developmental Science, 12, 4, 499-503. Cabeza. R. (2002). Hemispheric Asymmetry Reduction in Older Adults: The HAROLD Model. Psychology and Aging, 17, 1, 85-100. Chee, M. W., Hon, N., Lee, H. L., & Soon, C. S. (2001). Relative language proficiency modulates BOLD signal change when bilinguals perform semantic judgments. Blood oxygen level dependent. Neuroimage, 13, 1155–1163.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

178

Lilian Cristine Scherer, Rochele Paz Fonseca and Ana Inés Ansaldo

Chee, M. W., Tan, E. W. L., & Thiel, T. (1999). Mandarin and English single word processing studied with functional magnetic resonance imaging. The Journal of Neuroscience, 19, 3050-3056. Davis, S. W., Dennis, N. A., Daselaar, S. M., Fleck, M. S., & Cabeza, R. (2007). Qué PASA? The posterior anterior shift in aging. Cerebral Cortex, published online on Oct. 8, 2007. De Groot, A.M.B., Borgwaldt, S., Bos, M., & Van den Eijnden, E. (2002). Lexical decision and word naming in bilinguals: Language effects and task effects. Journal of Memory and Language, 47, 91-124. Dehaene, S., Dupoux, E., Mehler, J., Cohen, L., Paulesu, E., Perani, D., van de Moortel, P., Lehericy, S., & Le Bihan, D. (1997). Anatomical variability in the cortical representation of first and second language. Neuroreport, 8, 3809-3815. Démonet, J. F., Thierry, G. T., & Cardebat, D. (2005). Renewal of the Neurophysiology of Language: Functional Neuroimaging. Physiol. Review, 85, 49-95. Flege, J.E, Komshian, Y., & Liu, S. (1999). Age Constraints on Second –Language Acquisition, Journal of Memory and Language, 41, 78-104. Friederici, A. D., Fiebach, C. J., Schlesewsky, M., Bornkessel, I. D., & von Cramon, D. Y. (2006). Processing linguistic complexity and grammaticality in the left frontal cortex. Cerebral Cortex, 16, 1709-1717. Giussani, C., Roux, F. E., Lubrano, V., Gaini, S. M. & Bello, L. (2007). Review of language organisation in bilingual patients: what can we learn from direct brain mapping? Acta Neurochirurgica, 149, 1109-1116. Green, D. W. (1998). Mental control of the bilingual lexicosemantic system. Bilingualism. Language and Cognition, 1, 67-81. Grosjean, F. (1994). Individual bilingualism. In The encyclopedia of language and linguistics (pp. 1656-1660). Oxford: Pergamon Press. Grosjean, F. (1999). The bilingual’s language modes. In J. L. Nicol (ed.). One Mind, Two Languages: Bilingual Language Processing. Oxford: Blackwell. Hellermann, J., & Vergun, A. (2007). Language which is not taught: the discourse marker use of beginning adult learnars of English. Journal of Pragmatics, 39, 157-179. Hernandez, A. E., Martinez, A., & Kohnert, K. (2000). In search of the language switch: An fMRI study of picture naming in Spanish-English bilinguals. Brain and Language, 73, 421-431. Joanette, Y., Ansaldo, A.I., Carbonnel, S., Ska, B., Kahlaoui, K., Nespoulous, J.L. (2008). Communication, language and the brain: from past anterior to near future. Rev. Neurol. (Paris), 164 Suppl. 3, S83-90. Kim, K. S., Relkin, N. R., Lee, K. M., & Hirsch, J. (1997). Distinct cortical areas associated with native and second languages. Nature, 388, 171-174. Klein, D., Milner, B., Zatorre, R.J., Zhao, V., & Nikelski, J. (1999). Cerebral organization in bilinguals: A PET study of Chinese-English verb generation. NeuroReport, 10, 13, 28412845. Marian, V., Blumenfeld, H. K., & Kaushanskaya, M. (2007). The Language Experience and Proficiency Questionnaire (LEAPQ): Assessing Language. Journal of Speech, Language, and Hearing Research, 50, 4, ProQuest Psychology Journals. McClelland, J. L, McNaughton, B. L., & O’Reilly, R. C. (1995). Why There are Complementary Learning Systems in the Hippocampus and Neocortex: Insights from the Successes and Failures of Connectionist Models of Learning and Memory. Psychological Review, 102, 419-457.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Methodological Issues in Research on Bilingualism and Multilingualism

179

Morton, J. B., & Harper, S. N. (2007). What did the Simon say? Revisiting the bilingual advantage. Developmental Science, 10, 6, 719-726. Munõz, M. L., & Marquardt, T. P. (2008). The performance of neurologically normal bilingual speakers of Spanish and English of the short version of the bilingual aphasia test. Aphasiology, 22, 1, 3-19. Paradis, M. (2004). A Neurolinguistic Theory of Bilingualism (Studies on Bilingualism, 18). Amsterdam: John Benjamins Publishing Co. Parente, M.A.P.M., Fonseca, R.P., & Scherer, L.C. Literacy as a determining factor for brain organization: from Lecours´contribution to the present day. Dementia & Neuropsychologia,2(3), 165-172. Perani, D., & Abutalebi, J. (2005). The neural basis of first and second language processing. Current Opinion in Neurobiology, 15, 2, 202-206. Perani, D., Dehaene, S., Grassi, F., Cohen, L., Cappa, S. F., Dupoux, E., Fazio, F., & Mehler, J. (1996). Brain processing of native and foreign languages. NeuroReport, 7, 2439-2444. Perani, D., Paulesu, E., Sebastian-Gallés, N., Dupoux, E., Dehaene, D., Bettinaedi, V., Cappa, S., Fazio, F., & Mehler, J. (1998). The bilingual brain: Proficiency and age of acquisition of the second language. Brain, 121, 1841-1852. Price, C., Noppeney, U., & Friston, K.L. (2006). Functional Neuroimaging and Neuropsychologically Impaired Patients. In R. Cabeza and A. Kingstone (Eds.), Handbook of Functional Neuroimaging and Cognition, MIT: Cambridge, MA. Stern, Y. (2009). Cognitive reserve. Neuropsychologia, 47, 2015-2028. Suh, S., Yoon, H. W., Lee, S., Chung, J-Y., Cho, Z-H., & Park, H. W. (2007). Effects of syntactic complexity in L1 and L2: An fMRI study of Korean-English bilinguals. Brain Research, 1136, 178-189. Tan, L. H., Spinks, J. A., Feng, C. M., Siok, W. T., Perfetti, C. A., Xiong, J., Fox, P. T., Gao, J. H., (2003). Neural systems of second language reading are shaped by native language. Human Brain Mapping, 18, 158-166. Tisserand, D. J., Pruessner, J. C., Sanz Arigita, E. J., van Boxtel, M. P., Evans, A. C., Jolles, J., et al. (2002). Regional frontal cortical volumes decrease differentially in aging: an MRI study to compare volumetric approaches and voxel based morphometry. NeuroImage, 17, 2, 657-669. Tzourio, N., Nkanga-Ngila, B., & Mazoyer, B. (1998). Left planum temporale surface correlates with functional dominance during story listening. NeuroReport, 9, 829-33. Ullman, M. T. (2004). Contributions of neural memory circuits to language: The declarative/procedural model. Cognition, 92, 1- 2, 231-270. Vaid, J., & Hull, R. (2002). Re-envisioning the bilingual brain using functional neuroimaging: Methodological and interpretative issues. In F. Fabbro (Ed.), Advances in the neurolinguistics of bilingualism: Essays in Honor of Michel Paradis (315-355). Udine Forum: Udine University Press. Von Studnitz, R.E., & Green, D.W. (2002). Interlingual homograph interference in GermanEnglish bilinguals: Its modulation and locus of control. Bilingualism: Language and Cognition, 5, 1-23. Wartenburger, I., Heekeren, H.R., Abutalebi, J., Cappa, S.F., Villringer, A, & Perani, D. (2003). Early setting of grammatical processing in the bilingual brain. (2003). Neuron, 37, 1, 150-170.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved. Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

In: Bilinguals: Cognition, Education and Language Processing ISBN: 978-1-60741-710-1 Editor: Earl F. Caldwell, pp. 181-199 © 2010 Nova Science Publishers, Inc.

Chapter 9

BILINGUALISM AND HISPANIC AMERICAN INTELLIGENCE TEST SCORES Philip G. Gasquoine, Aracely Cavazos, Juan Cantu and Amy A. Weimer University of Texas – Pan American, TX, USA

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Abstract Hispanic Americans as a group score 0.5SD below White Americans on intelligence test measures (administered in English) that emphasize language processing, but score similarly when visual-perceptual/visual-motor processing is required. The reason for this language decrement is unknown. This chapter considers the possible contribution of bilingualism to this effect, as studies linking bilingualism and cognition (conducted with multiple ethnic groupings) have consistently shown a bilingual disadvantage compared to monolinguals on language processing tasks. Two data sets (older children and adults) of bilingual Hispanic American performance on various intelligence test measures administered in Spanish and English showed evidence of a visual-perceptual/visual-motor over language processing advantage of about 1SD. The size of the visual-perceptual/visual-motor over language advantage was similar in both languages suggesting it is bilingualism-related and not due to low English language proficiency. Bilingualism appears to be a potentially important factor in the Hispanic American language processing decrement seen on intelligence tests, although no direct study on the effect of this variable has yet been conducted.

Introduction There are differences in the mean scores of groupings of minority and White Americans on intelligence, achievement, and neuropsychological tests. Probably the most well-known is a 10 to 15 point (.67 to 1SD) decrement in Full-Scale IQ by groupings of African vs. White Americans (Dickens and Flynn, 2006; Jensen, 1969; Neisser et al, 1996; Prifitera, Weiss, and Saklofske, 1998; Rushton and Jensen, 2006; Williams and Ceci, 1997). The cause of this and other minority vs. White American group score differences on intelligence tests has been much debated. Popular explanations have included test bias in all its many possible

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

182

Philip G. Gasquoine, Aracely Cavazos, Juan Cantu et al.

manifestations (e.g., Flaugher, 1978; Schmidt and Hunter, 1974; Van de Vijver and Phalet, 2004), group demographic differences (e.g., in socioeconomic status or level of parental education; Prifitera et al., 1998), and genetics (Jensen, 1969; Rushton and Jensen, 2006). None have survived scientific scrutiny. Even a task force convened by the Board of Scientific Affairs of the American Psychological Association to review the issue failed to identify a clear explanation (Neisser et al., 1996). Hispanic Americans as a group also score below White Americans on intelligence tests (administered in English). The discrepancy is of the order of 7 to 8 points (.5SD) on language measures (e.g., Verbal IQ) with no discrepancy typically being recorded on visualperceptual/visual-motor measures (e.g., Performance IQ: Neisser et al., 1996; Puente and Salazar, 1998). In the field of intelligence testing, the distinction between language and visual-perceptual/visual-motor measures predates World War I. Henry Goddard proposed use of the language processing heavy Binet-Simon scale for the screening of immigrants at Ellis Island in New York Harbor using translators. After about 80% of European adult immigrants, many of whom had minimal education, scored as “feeble-minded” (i.e., mental age below 12) during pilot testing (Gregory, 2007; Strickland, 2000), Ellis Island physicians began the search for an intelligence test that minimized the language component by using visualperceptual/visual-motor tasks (Boake, 2002). Conceptually, the distinction between language and visual-perceptual/visual-motor processing corresponds to the differential specialization of the dominant (language) and non-dominant (visual-perceptual) hemispheres of the brain. Superior performance on visual-perceptual/visual-motor over language measures has led to a common contemporary practice of assessing the intelligence of Hispanic Americans via the use of exclusively visual-perceptual/visual-motor (often referred to as “culture-fair”) intelligence tests (Harris, Wagner, and Cullum, 2007). Examples of such tests include the Leiter International Performance Scale – Revised (Roid and Miller, 1995) and the Test of Nonverbal Intelligence - 2 (Brown, Sherbenou, and Johnson, 1990). This practice changes the fundamental definition of intelligence from that of general mental ability to a more specific definition that emphasizes visual-perceptual/visual-motor skills while ignoring the assessment of language abilities. The change may reduce predictive validity as the modern Western schooling system emphasizes the use of language processing much more than visualperceptual/visual-motor processing. One factor that potentially moderates the Hispanic American language vs. visualperceptual/visual-motor processing difference is bilingualism. As bilingualism is much more frequent among Hispanics than White Americans it is puzzling that this factor has not generated more interest. This is likely due to intelligence test developers using a census category definition for Hispanic Americans. Identified through self-report, they are included in test standardization samples in proportion to their representation within the general population. This focuses attention away from potentially important within group differences in acculturation, educational level, English language proficiency, and bilingualism, all of which are factors that may influence intelligence test scores. In a few isolated studies researchers have investigated such within group differences post hoc with interesting results. For example, Prifitera et al. (1998) analyzed data from the Wechsler Intelligence Scale for Children – III (WISC-III; Wechsler, 1991) standardization sample and found the Performance – Verbal IQ decrement among Hispanic American children decreased as parental education increased, a trend not evident for African Americans.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Bilingualism and Hispanic American Intelligence Test Scores

183

Although the relationship between bilingualism and intelligence test scores in Hispanic Americans has yet to be subjected to direct experimental test, there have been studies of indirect relevance, most especially investigations on the effect of bilingualism on cognition that have been conducted with multiple ethnic groupings. Intelligence and cognition are similar concepts with different definitions.

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Defining Intelligence and Cognition Even experts disagree on the exact definition of intelligence (Neisser et al., 1996). It is operationally defined here as what is measured by standardized, individually administered, intelligence tests. There are many different such tests but the most widely used in the United States are the Wechsler scales (Prifitera and Saklofske, 1998) that consist of a number of subtests sampling different subsystems of cognition. All subtests require multiple subsystems of cognition for successful completion, but each subtest emphasizes a specific subsystem, such as working memory. Based on factor analytical studies, modern versions of the Wechsler scales, like the WISC-IV (Wechsler, 2003), have replaced the traditional split between Verbal and Performance IQs with four IQ indices: (a) Verbal Comprehension; (b) Perceptual Reasoning; (c) Working Memory; and (d) Processing Speed, that are all combined in the Full-Scale IQ. Scores from standardized, individually administered, intelligence tests like the WISC-IV have considerable practical utility in predicting scholastic attainment and, in conjunction with achievement and other measures, have been widely used within school systems to identify children with mental retardation, learning disability, and the gifted and talented. American minorities have historically been over-represented in the first two categories and underrepresented in the third. In 1979 such statistics led to a ban in the state of California on the use of intelligence tests for the placement of African American children in special education classes for children with mental retardation (Larry P. v. Wilson Riles, 1979). This decision was later overturned by the same judge (Crawford et al. v. Honig et al., 1994), as no satisfactory alternative measure to determine special education placement had been found. In contrast to African American children, Hispanic Americans are not over-represented nationally among children with mental retardation but they are over-represented among those diagnosed with learning disability by 7% and under-represented in gifted and talented programs by 50% (National Research Council, 2002). The term cognition is considered here as a general or all-encompassing term, similar to intelligence, with a number of overlapping subsystems such as attention, memory, language, visual-perceptual processing, motor skills, and executive functions. These subsystems may be considered distinct in that certain parts of the brain have been found to be specialized for their operation, like the hippocampus for memory, Broca’s and Wernicke’s Areas for language, or the frontal lobes for executive functions. A change to any one of these subsystems can potentially have an impact on any other subsystem as they are highly interactive in producing behavior. Research has shown that bilingualism can have a positive, negative, or no effect on different subsystems of cognition.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

184

Philip G. Gasquoine, Aracely Cavazos, Juan Cantu et al.

Effect of Bilingualism on Cognition

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Positive Effect There has been much reasoned argument for the potential positive effect of bilingualism on cognition (e.g., Bialystok, 2007; Kovacs, 2007), but the primary experimental support comes from a series of developmental studies using tasks that required suppression of a previously correct response. Bialystok (1999) features a typical experimental paradigm demonstrating this type of effect. Groupings of bilingual (English/Chinese-speaking) and monolingual (English-speaking) children aged 3 to 6 years were matched for English proficiency on a receptive language task (Peabody Picture Vocabulary Test – Revised) and administered the Dimensional Change Card Sort Task. This is a simplified version of the Wisconsin Card Sorting Test (which has norms for ages 6 years 6 months through 89 years 11 months; Heaton, Chelune, Talley, Kay, and Curtiss, 1993), widely used in clinical neuropsychological practice. In the Dimensional Change Card Sort the child sorts cards featuring two dimensions (color and form), first on one dimension and then on the second dimension. Sorting on the second dimension requires suppression of the previously correct response. Bilingual and monolingual children scored equivalently while sorting on the first dimension (there was a ceiling effect), but the bilingual children were significantly better at sorting on the second dimension. Related response inhibition tasks that have shown a bilingual over monolingual superiority have been found across the lifespan. Tests on which the effect has been demonstrated include: (a) theory of mind tasks for young children (Bialystok and Senman, 2004; Goetz, 2003). For example, Goetz used a False Belief Task in which children were asked what was inside a candy box in which a car had been secretly placed. The correct response should be candy. After the children were shown that there was in fact a car inside the box they were asked what a friend who has not seen inside the box would think it contained. The correct answer (candy) required inhibition of the child’s own knowledge that there was a car inside the box. (b) The Simon task where participants must respond to the dimension of a stimulus (e.g., color) despite misleading cue information on half the trials (i.e., a red stimulus appearing above a blue response key; Bialystok, Martin, and Viswanathan, 2005). (c) The Posner Attentional Network Task where participants must indicate the direction of an arrow accompanied by either congruent or incongruent cue information (Costa, Hernandez, and Sebastian-Galles, 2008); and (d) the Stroop test when the measure is time reading color names printed in an incongruent color (i.e., “RED” printed in green ink) minus time to read the same color names printed in the congruent color (Bialystok, Craik, and Luk, 2008). Note that Rosselli et al. (2002) found bilingual adults were slower than monolinguals in reading the congruent color names on the Stroop test. Response inhibition tasks like the Stroop and Wisconsin Card Sorting Test have long been associated in the practice of clinical neuropsychology with executive functioning and the frontal lobes. Deficient performance on such tests is assumed to indicate executive dysfunction from such neurological conditions as severe traumatic brain injury. Response inhibition is a subcomponent of executive function, a term that covers a vast array of behaviors often referred to as high level cognitive functions. There is no consensus classification scheme but this term is typically used to refer to concepts like social behavior,

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Bilingualism and Hispanic American Intelligence Test Scores

185

personality, personal memories, and self-awareness (Stuss, Gallup, and Alexander, 2001), volition, planning, purposive action, and effective performance (Lezak, Howieson, and Loring, 2004), and working memory, cognitive control, set shifting, theory of mind, and delay of gratification (Farah et al., 2006). There is no evidence that bilingualism has a positive effect on any other subcomponent of executive function aside from response inhibition. There is evidence (see Costa et al., 2008) that suggests bilingual persons are actively using an inhibitory control mechanism during all language processing as both languages become active concurrently. It is hypothesized that the bilingual advantage in response inhibition stems from having to constantly suppress interference from the language not currently in use (Bialystok et al., 2008).

Negative Effect

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Negative effects of bilingualism on cognition have been reported for many language processing tasks. Young bilingual children score lower than monolinguals in receptive vocabulary (Bialystok, 2007). Adult bilingual participants, in comparison to monolinguals, have: more tip-of-the-tongue retrieval failures (Gollan and Acenas, 2004); naming deficits in both speed and accuracy (Bialystok et al., 2008; Gollan, Montoya, Fennema-Notestine, and Morris, 2005; Roberts, Garcia, Desrochers, and Hernandez, 2002); speed deficits in reading color names (Rosselli et al., 2002); and reduced scores on letter and category fluency tasks (Bialystok et al., 2008; Gollan, Montoya, and Werner, 2002). These linguistic deficits have been recorded in testing situations where there is time pressure to retrieve words and are not apparent in general social situations. The reason for the bilingual linguistic deficit is still a matter of conjecture, although it is thought that the parallel activation of both languages causes interlanguage interference thereby slowing processing time and increasing the possibility of errors (Bialystok et al., 2008).

No Effect Cognitive subsystems on which bilingual and monolingual participants perform equivalently include the development of phonological awareness skills necessary for reading (Bialystok, 2002) and visual-perceptual working memory tasks such as Corsi Blocks (Bialystok et al., 2008).

Methodological Factors in Bilingualism/Cognition Studies Although the results of bilingualism/cognition studies have been consistent in showing a positive effect of bilingualism on response inhibition and a negative effect on language processing tasks, several methodological factors may limit generalization of this result. Most of these studies have compared language proficiency matched groupings of bilingual and monolingual participants on the cognitive task under review (e.g., Bialystok, 1999; Bialystok and Senman, 2004; Bialystok et al., 2005; 2008; Costa et al., 2008; Goetz, 2003; Harris et al., 2007). As bilingual participants score below monolinguals on language processing tasks this practice may actually result in a comparison between a bilingual grouping from a distribution

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

186

Philip G. Gasquoine, Aracely Cavazos, Juan Cantu et al.

at the higher end of the normal distribution than that of the monolinguals. In other words, language proficiency score distributions for monolingual and bilingual participants are unlikely to be equally matched. An alternative design compares groupings of balanced and unbalanced bilingual participants (e.g., Gasquoine, Croyle, Cavazos-Gonzalez, and Sandoval, 2007; Gollan, Fennema-Notestine, Montoya, and Jernigan, 2007). In this alternative design there are no monolingual controls and bilingualism is viewed as a continuum from dominance in one language, through balanced bilingualism, to dominance in the second language. It is preferable if a study on bilingualism and cognition uses two languages of administration, as it avoids procedural confounds from either testing the bilingual group twice and the monolingual group only once (e.g., Rosselli et al., 2002), or testing the monolingual group in a language they do not understand (e.g., Harris, Cullum, and Puente, 1995). One disadvantage of using two languages of administration is that practice effects may differentially apply to certain cognitive tasks and can cause ceiling effects. It may also be easier to match balanced and unbalanced bilingual groupings than monolingual and bilingual groupings on relevant demographic variables. Group mean cognitive test score differences can be difficult to interpret as bilingualism-related if there are naturally occurring demographic differences between groupings in acculturational level, educational level, or socioeconomic status (SES). This is a frequent problem whenever White and minority samples are compared on cognitive tests (Gasquoine, 1999; 2001; Hosp and Reschly, 2004). For example in children, SES correlates highly with scores from intelligence, achievement, and neuropsychological tests, especially when language processing is involved (e.g., Noble, McCandliss, and Farah, 2007). As SES is a broad, multidimensional, psychological construct with imprecise operational definition it is difficult to determine exactly what aspect(s) of SES are responsible for the effect. When group demographic differences do occur (e.g., in Bialystok and Senman, 2004 the monolingual and bilingual groupings differed on SES and English language proficiency) analysis of covariance (ANCOVA) should not be used to provide “correction”, as it requires the assumption that there is no meaningful relationship between the covariate and bilingual group membership. ANCOVA was designed to correct for covariate differences between groupings formed by random assignment (Miller and Chapman, 2001). Bilingualism/cognition studies tend to use selected participants with high language proficiency. As example, Bialystok et al. (2005), who studied participants from across the lifespan, wrote: “Five-year-old children with enough linguistic experience to be considered bilingual by the rigid criteria necessary to conduct controlled research are rare” (p. 107). Young adult participants were university undergraduates and older participants were “selected according to rigorous criteria for language experience and education” (p. 113). Participants with limited language proficiency were excluded. As example, “Eleven additional bilingual children were not included because their English was too limited” (Bialystok, 1999, p. 638). Findings from such selected participants may not generalize to a naturally occurring bilingual population that will include individuals with varying degrees of language proficiency across the full spectrum. Bilingualism is a multidimensional construct with many possible classification schemes (Grosjean, 1998; Rhodes, Ochoa, and Ortiz, 2005). One such classification that appears especially pertinent to the generalization of bilingualism/cognition research findings is the elective vs. circumstantial distinction of Valdes and Figueroa (1996). Elective bilinguals

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Bilingualism and Hispanic American Intelligence Test Scores

187

chose to learn a second language, whereas circumstantial bilinguals are forced to learn a second language due to life circumstances. It can be surmised that elective bilinguals have an inborn talent that enables them to more easily achieve proficiency in two languages while sustaining the motivation to do so. Such participants appear more likely to be selected for inclusion in bilingualism/cognition studies as a consequence of high scores on language proficiency tests. In contrast, a naturally occurring bilingual population of Hispanic Americans, whose bilingualism is related to migration, will contain a mix of elective and circumstantial bilinguals with language proficiency skills that include those in the lower values. Another crucial element in bilingualism/cognition research is the assessment of language proficiency and dominance. The distinction between these two bilingualism dimensions is that proficiency involves an ability rating or total score in a language, whereas dominance involves difference scores between proficiency measures in two languages. Self-report (e.g., Gasquoine et al., 2007; Harris et al., 2007), likert scale self-ratings (e.g., Ardila, Rosselli, Ostrosky-Solis, Marcos, Granda, and Soto, 2000; Bialystok et al., 2008; Gasquoine et al., 2007; Gollan et al., 2002; Harris et al., 2007; Rosselli et al., 2002), and standardized tests of language ability (e.g., Gasquoine et al., 2007; Gollan et al., 2007; Harris et al., 1995; 2007; Rosselli et al., 2002) have all been used to assess proficiency and dominance. Gasquoine et al. (2007) compared all three in the measurement of language dominance and found self-report likert difference scores provided no improvement over participant self-report of dominant language. The latter correlated highly (r = .77) with standardized test generated language proficiency difference scores in a Hispanic American adult sample. Several standardized tests of language ability have been employed to assess proficiency in bilingualism/cognition research. Most measure only a single dimension of language, such as oral expression (Harris et al., 1995; 2007), picture naming (Boston Naming Test: Gollan et al., 2007; Rosselli et al., 2002), or receptive vocabulary (Peabody Picture Vocabulary Test – Revised: Bialystok, 1999; Goetz, 2003). This might be problematic when determining dominance, as it can vary across different language dimensions like speaking and comprehending (Grosjean, 1998; Rhodes et al., 2005). To provide a more comprehensive assessment of language dominance with Hispanic Americans, Gasquoine et al. (2007) used the difference in total scores from the Spanish and English language versions of the Woodcock-Munoz Language Survey – Revised (WMLS-R; Woodcock, Munoz-Sandoval, Ruef, and Alvarado, 2005). These total scores are derived from subtests that assess multiple language dimensions, including oral expression, language comprehension, reading, and writing. When measuring dominance it is frequently necessary to establish some criteria to distinguish balanced from unbalanced bilingual participants. Perfect balance is rare and there is no accepted standard cut-off (Gollan et al., 2007). Some studies provide vague information as to how balance was determined. For example, Bialystok et al. (2005) used “a strict set of questions regarding the participants’ language experiences” (p. 110). Similarly, Costa et al. (2008) wrote: “Bilingual participants used both…(languages)…at a native speaker level, i.e., they have a high proficiency level in speaking, comprehending, reading, pronouncing and writing both of the languages” (p. 68). These definitions are difficult to replicate. More operational definitions have involved splitting the range of difference scores obtained on a proficiency measure post-hoc, so that balanced and unbalanced groupings have roughly equal numbers (Gasquoine et al., 2007; Gollan et al., 2007). In Gollan et al. the classification

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

188

Philip G. Gasquoine, Aracely Cavazos, Juan Cantu et al.

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

measure (Boston Naming Test) was unfortunately also the primary outcome measure, a situation that should be avoided as the authors themselves noted. With these methodological concerns in mind, two data sets showing intelligence test performance of bilingual Hispanic Americans in both Spanish and English are reviewed with the goal of estimating the effect of bilingualism on these scores. If the Hispanic American visual-perceptual/visual-motor over language processing advantage is bilingualism-related it should be present in both languages. If the advantage is a consequence of limited English language proficiency it should be present only for the English administration. The first data set comes from Gasquoine et al. (2007), a study on the effect of language of administration on various neuropsychological measures among a grouping of consecutive, neurologically intact, bilingual, Hispanic American adults. The second data set consists of WISC-IV profiles of a selected, language proficiency balanced, grouping of bilingual, Hispanic American, older children. The two data sets have the following design features in common: 1. All participants were residents of the Rio Grande Valley region of Texas. Over 90% of this area’s residents are of Hispanic heritage. Many are subjectively fluent in both Spanish and English, switch easily between languages, have little accent in either language, and speak Spanish (either exclusively, or in combination with English) at home. Overton, Fielding, and Simonsson (2004) reported that area public schools have large numbers of students in: (a) bilingual programs (i.e., the curriculum is taught in both Spanish and English in varying proportions - about 25% of students); (b) English as a second language programs (i.e., English only instruction - about 10%); and (c) students in migrant programs (i.e., parents are farm workers who move to Northern states at harvest time - about 10%). 2. Participants were volunteers who were considered for inclusion if they were of Hispanic descent and subjectively bilingual (i.e., able to converse in both Spanish and English). 3. Language dominance was determined by subtracting the WMLS-R age-corrected total cluster scale score (M = 100, SD = 15) in English (Form A) from that in Spanish for each participant. Each language version of the WMLS-R has seven subtests (Picture Vocabulary, Verbal Analogies, Letter-Word Identification, Dictation, Understanding Directions, Story Recall, and Passage Comprehension). Spanish subtests are all adaptations of the parallel tests in English. For the English version, norms were derived from over 4500 K through 12th grade Americans, census matched on various demographics including race/ethnicity. For the Spanish version, norms were from over 1100 native monolingual Spanish speakers mostly from Mexico, Argentina, Panama, and Costa Rica. 4. Language proficiency was measured as WMLS-R cognitive-academic language proficiency (CALP) scores. These reflect “language proficiency in academic situations, or those aspects of language proficiency that emerge and become distinctive with formal schooling” (Woodcock et al., 2005, p. 61). CALP scores can range from 1 (Negligible) to 6 (Very Advanced). 5. A repeated measures design was employed whereby each participant was tested in two sessions by the same bilingual examiner. At each session only Spanish or English was used and participants were instructed that they could only respond in

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Bilingualism and Hispanic American Intelligence Test Scores

189

that same language. Half the participants (randomly chosen) in each data set were tested in Spanish at the first session and English at the second.

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Data Set 1: Language Proficiency and WAIS-III Digit Span and Matrix Reasoning Subtest Scores in Bilingual, Hispanic American Adults Participants in Gasquoine et al. (2007) were 36 consecutive adult volunteers aged 20 through 65 years. There were 13 males and 23 females with a mean age of 31.7 (SD = 11.8) years. Education ranged from 9 to 18 (M = 14.4, SD = 2) years. All participants were of Mexican heritage, as determined by ancestry, and were either born in the United States (22) or Mexico (14). Those born in Mexico had lived in the United States from 3 to 25 (M = 13.5, SD = 7) years. Most were educated entirely in the United States (28), with four being educated in Mexico, and four in both countries. Most (32) were sequential bilinguals with Spanish as their first language. Second language acquisition usually began at school. Most (28) still spoke Spanish exclusively at home. Participants were tested over two sessions, 1461 (M = 25, SD = 8.9) days apart. CALP scores ranged from 2 to 5 in Spanish and 3 to 5 in English. Mean CALP scores were 3.5 (SD = .8) in Spanish and 3.3 (SD = .6) in English. Both these means are in the Limited to Fluent range of language proficiency. They did not differ significantly t(35) = 1, ns. Among the many neuropsychological instruments administered in this study were the Digit Span and Matrix Reasoning subtests from the Wechsler Adult Intelligence Scale –III (WAIS-III: Wechsler, 1997). The Spanish adaptation of the Digit Span subtest was taken from the Bateria Neuropsicologica (Artiola i Fortuny, Hermosillo, Heaton, and Pardee, 1999). Digit Span is a measure of verbal working memory that requires the immediate repetition of increasingly longer sequences of digits both forwards and backwards. Instructions for the Matrix Reasoning subtest were translated into Spanish using standard back-translation techniques. The Matrix Reasoning subtest requires solving visual-perceptual sameness, symmetry, and analogy problems. Both Digit Span and Matrix Reasoning have a national mean of 10 (SD = 3). Table 1 shows the mean WMLS-R total cluster scale scores together with mean WAIS-III Digit Span and Matrix Reasoning scale scores for the Spanish and English language administrations. These scores were analyzed by ANOVA in the original study as part of a larger data set and the Digit Span scores were found to differ significantly at p < .05, being higher in English. Longer digit spans in English than Spanish have been attributed to differences in the phonological length of digits (e.g., Ardila et al., 2000). The Matrix Reasoning combined mean scale score was at the 63rd percentile of national norms and was more than 1SD higher than the WMLS-R total cluster combined scale score mean at the 23rd percentile. The Digit Span combined mean scale score was intermediate at the 37th percentile of national norms. This sample was well-educated (as is typical of studies that involve the psychological testing of volunteers, Kaplan and Saccuzzo, 2005, p. 363) with an average of more than two years of college. The mean Matrix Reasoning subtest scale score at the 63rd percentile of national norms was closer to expected ability levels, as judged from years of education completed, than the WMLS-R total cluster scale score at the 23rd percentile. This visualperceptual over language processing advantage was more than double the size of the

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

190

Philip G. Gasquoine, Aracely Cavazos, Juan Cantu et al.

Performance minus Verbal IQ difference (.5SD) previously reported for Hispanic Americans on intelligence tests administered in English (e.g., Neisser et al., 1996; Prifitera et al., 1998; Puente and Salazar, 1998). The size of the advantage was similar in both Spanish and English administrations, suggesting it is bilingualism-related and not due to limited English language proficiency. Table 1. Means (and Standard Deviations) for 36 Bilingual, Hispanic American, Adults on the WMLS-R and WAIS-III Digit Span and Matrix Reasoning subtests in Spanish and English

Spanish

English

WMLS-R cluster scale score

87.9 (19.7)

85.8 (14.5)

SpanishEnglish Difference +2.1

Matrix Reasoning scale score

10.7 (2.8)

11.3 (2.3)

-0.6

11.0

Digit Span scale score

8.7* (1.9)

9.5* (2.4)

-0.8

9.1

Combined Mean 86.9

* p < .05. Key: WMLS-R = Woodcock Munoz Language Survey – Revised; WAIS-III = Wechsler Adult Intelligence Scale, Third Edition.

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Data Set 2: WISC-IV Subtest Scores and IQ Indices from a Language-Balanced Grouping of Bilingual Hispanic American Older Children Spanish is the only foreign language for which a range of standardized cognitive assessment techniques has been developed for use within the United States. This has been dictated by legal challenges and population density demands. Legal rulings from the consent decree of Diana v. State Board of Education (1973) to the Individuals with Disabilities Education Act (2004) have held that the intellectual assessment of Hispanic American school children should minimize language bias by conducting administration in the dominant language of the child. At the last census 78% of Hispanic Americans over the age of five reported speaking Spanish at home (U.S. Bureau of the Census, 2000). Spanish-speaking, Hispanic Americans comprise 77% of limited English proficient children in public schools in the United States, far outnumbering the next largest foreign language-speaking group, the Vietnamese, at just 2% (Rhodes et al., 2005). Spanish language assessment instruments developed for use with Hispanic Americans provide an ideal opportunity to study the effects of bilingualism on cognition, if the equivalence of the Spanish and English versions has been established (Fernandez, Boccaccini, and Noland, 2007). One recent Spanish language adaptation that claims equivalence to the English version is the WISC-IV Spanish (Wechsler, 2005), an intelligence test designed primarily for use with immigrant Hispanic American children who are Spanish dominant and learning English as a second language while acculturating to the United States. The WISC-IV Spanish standardization sample consisted of 851 children with five or less consecutive years of education in the continental United States and self-reported Spanish language dominance.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Bilingualism and Hispanic American Intelligence Test Scores

191

The equivalence of the Spanish and English versions of the WISC- IV was determined within the standardization samples by comparing respective IQ and subtest scores. Generally, mean scores were lower in Spanish than English but the two samples also differed in parental education levels. To circumvent this confound, a group of children from the Spanish standardization sample was matched (gender, age, and parental educational level) with a group of Hispanic American children from the English standardization sample. Small differences in scores were found (e.g., Full-Scale IQs were 2 points higher in English), supporting the equivalence of the two versions at the IQ and subtest score level. In our analysis WISC-IV Spanish and English subtest and IQ index scores from a language-balanced grouping of local born, Hispanic American, older children were compared. We were primarily interested in confirming the presence of a visual-perceptual/visual-motor over language processing advantage and determining if the advantage was present in both languages or only in English. If the advantage is present in both languages it suggests it is bilingualism-related and not due to limited English language proficiency.

Method

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Participants Participants were 13 girls and 11 boys from three public school districts in Hidalgo County in the Rio Grande Valley region of Texas. Inclusion criteria were: (a) Hispanic descent; (b) born and educated in the United States; (c) aged 10 through 16 years; (d) subjectively bilingual (i.e., able to converse in both Spanish and English); and (e) in regular classes with no history of evaluation for special education, or mental health diagnosis. Participants were consecutive volunteers who gave personal assent and whose parents agreed to their inclusion. The 24 children had a mean age of 13.6 (SD = 1.7) years and were in the 4th through 11th grades. All were of Mexican heritage, except for one whose parents were from Colombia. There were 12 simultaneous bilinguals who learned both Spanish and English from infancy and 12 sequential bilinguals who learned Spanish first and then English at school. All spoke Spanish either exclusively (7 cases), or in combination with English, at home.

Measures The 10 core subtests (Block Design [BD], Similarities [SI], Digit Span [DS], Picture Concepts [PC], Coding [CD], Vocabulary [VO], Letter-Number Sequencing [LNS], Matrix Reasoning [MR], Comprehension [CO], and Symbol Search [SS]) of the WISC-IV Spanish and English versions were administered in a standardized fashion with the sole exception that answers in English were neither prompted nor scored as correct on the Spanish version. These subtests generate scale scores with national M = 10 (SD = 3). Administration of these subtests allowed calculation of all IQ indices, with national age-corrected M = 100 (SD = 15): Verbal Comprehension Index (VCI), Perceptual Reasoning Index (PRI), Working Memory Index (WMI), and Processing Speed Index (PSI), in addition to Full-Scale IQ (FSIQ). Norms for subtests comprising the PRI (BD, PC, and MR) and PSI (CD and SS) are identical in Spanish

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

192

Philip G. Gasquoine, Aracely Cavazos, Juan Cantu et al.

and English, as changes in administration were restricted to instruction translation. Norms for subtests comprising the VCI (SI, VO, and CO) and WMI (DS and LNS) were calibrated to obtain equivalence across languages, as these subtests had changes to item content and scoring rules.

Procedure and Data Analysis Participants were tested in two sessions 6-63 (M = 19, SD = 17) days apart. The length of each session was approximately three hours. Three bilingual Mexican American examiners were used, all with Masters level training in the assessment of intelligence and language proficiency in children. A language-balanced grouping was obtained by: (a) subtracting the WMLS-R total cluster score in English from that in Spanish for each participant. The resulting difference scores (positive scores indicate greater Spanish proficiency) ranged from +13 to -82; and (b) excluding participants with difference scores > 20. This cutoff was determined a posteriori to exclude the least number of participants while matching Spanish and English language skills as much as possible. The resulting language-balanced grouping was comprised of 18 participants with difference scores ranging from +13 to -18 and equal numbers either side of zero. Language skills were still not quite equally balanced, as indicated by a summed group WMLS-R difference score of -27 points (i.e., 1.5 points per participant in favor of English). Language of administration at the first session was counterbalanced within the grouping.

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Results WMLS-R CALP scores ranged from 2 to 5 in Spanish and 3 to 5 in English. Mean CALP scores were 3.7 (SD = .6) in Spanish and 3.8 (SD = .5) in English. Both these means are in the Limited to Fluent range. They did not differ significantly t(17) = .9, ns. Table 2 shows the mean Spanish and English WISC-IV subtest scores. They were analyzed in a 2 (language) x 10 (subtest score) repeated measures ANOVA. The main effect of language, F(1, 17) < 1, ns, η² = .03 and the language x subtest score interaction, F(9, 153) = 1.7, ns, η² = .1 were not significant. The main effect of subtest score was significant F(9, 153) = 5.9, p < .01, η2 = .3. Post hoc comparisons using the Bonferroni procedure showed the combined Spanish and English MR mean (11.0) was significantly higher than the corresponding means for VO (7.4), BD (7.6), and CO (8.3) at p < .05. No other comparison among the subtest scores was significant. Table 3 shows the mean Spanish and English WISC-IV IQ indices. They were analyzed in a 2 (language) x 4 (IQ index) repeated measures ANOVA. The main effect of language, F(1, 17) < 1, ns, η² = .01 and the language x IQ index interaction, F(3, 51) = 2.5, ns, η² = .13 were not significant. The main effect of IQ index was significant F(3, 51) = 4.8, p < .05, η2 = .22. Post hoc comparisons using the Bonferroni procedure showed the combined Spanish and English VCI mean (89.3) was significantly below the corresponding PSI mean (104.6) at p < .05. No other comparison among the IQ indices was significant.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Bilingualism and Hispanic American Intelligence Test Scores

193

Table 2. Mean (Standard Deviation) Spanish and English WISC-IV Subtest Scores Obtained by a Language-Balanced, Bilingual Grouping (N = 18). The National Mean is 10 (3) for Every Subtest

Spanish

English

Similarities

8.0 (2.0)

9.5 (2.9)

SpanishEnglish Difference -1.5

Vocabulary

7.0 (2.2)

7.9 (2.7)

-0.9

7.4 *

Comprehension

8.1 (3.2)

8.6 (3.2)

-0.5

8.3 ˆ

Block Design

7.7 (2.3)

7.5 (2.3)

+0.2

7.6˚

Picture Concepts

9.7 (3.6)

9.6 (3.9)

+0.1

9.7

Matrix Reasoning

10.8 (3.9)

11.2 (3.8)

-0.4

11.0 * ˆ ˚

Digit Span

9.6 (2.1)

9.0 (2.4)

+0.6

9.3

Letter-Number-Sequencing

9.8 (3.2)

9.2 (4.5)

+0.6

9.5

Coding

11.6 (4.4)

11.3 (4.6)

+0.2

11.5

Symbol Search

9.9 (3.2)

10.2 (3.3)

-0.3

10.1

Combined Mean 8.7

*, ˆ, ˚ denotes pairs of means that differ at p < .05. Key: WISC-IV = Wechsler Intelligence Scale for Children, Fourth Edition.

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Table 3. Mean (Standard Deviation) IQ Indices for the Spanish and English Versions of the WISC-IV Obtained by a Language Balanced Bilingual Grouping (N = 18). The National Mean is 100 (15) for Every IQ Index

Spanish

English

Spanish – English Difference

Combined Mean

Verbal Comprehension Index

86.8 (12.5)

91.9 (15.3)

-5.1

89.3*

Perceptual Reasoning Index

96.4 (17.5)

96.7 (18.8)

-0.2

96.6

Working Memory Index

97.4 (13.4)

94.0 (17.9)

+3.4

95.8

Processing Speed Index

104.6 (19.6)

104.7 (19.9)

-0.1

104.6*

Full-Scale IQ

94.1 (14.5)

95.2 (19.7)

-1.1

94.7

IQ Index

* p < .05. Key: WISC-IV = Wechsler Intelligence Scale for Children, Fourth Edition.

Discussion Studies on adults (e.g., Gasquoine et al., 2007; Gollan et al., 2002; Harris et al., 1995; Rosselli et al., 2002) and children (e.g., Naglieri, Otero, DeLauder, and Matto, 2007) have shown that groupings of balanced bilingual participants perform equally well on cognitive tests in both languages. This finding was supported here as there were no significant differences between any pair of corresponding Spanish and English WISC-IV subtest or IQ

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

194

Philip G. Gasquoine, Aracely Cavazos, Juan Cantu et al.

index scores. When Spanish and English scores were combined, Vocabulary, Comprehension, and Block Design subtest scores were significantly below Matrix Reasoning and the Verbal Comprehension Index was significantly below the Processing Speed Index. Effect sizes were small in, but these results support the concept that Hispanic Americans perform better when tests emphasize visual-perceptual/visual-motor compared to language processing. Support was not unequivocal as participants scored relatively poorly on the Block Design subtest, a visual-constructional task. The visual-perceptual/visual-motor advantage occurred in both languages, suggesting it is a consequence of some aspect of the bilingual experience and is not due to limited English language proficiency. The size of the language vs. visual-perceptual/visual-motor difference was of the order of 1SD. Specifically, the largest significant differences for subtest and index scores respectively were: (a) the Matrix Reasoning subtest score (11.0) was 1.2 SD units higher than the Vocabulary subtest score (7.4); and (b) the Processing Speed Index (104.6) was 1 SD unit higher than the Verbal Comprehension Index (89.3). This is about double the size of the previously reported visual-perceptual/visual-motor advantage for Hispanic Americans in English language intelligence test standardization samples (e.g., Neisser et al., 1996; Prifitera et al., 1998; Puente and Salazar, 1998).

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Conclusion Bialystok (2007) posed the question: “Are bilinguals more intelligent than monolinguals?” (p. 220). The answer appears to be “no”, at least if intelligence is defined as what is measured on popular standardized intelligence tests, a definition that is not without its critics (e.g., Sternberg, 2004). Studies investigating the effect of bilingualism on cognition have consistently shown a bilingual over monolingual advantage in response inhibition tasks, but there is a monolingual over bilingual advantage on tasks that emphasize language processing. Popular standardized intelligence tests like the Wechsler scales do not have items that require response inhibition, whereas approximately half the total number of items emphasize language processing skills. Despite low scores on intelligence tests, there are major social and economic advantages to bilingualism, especially in borderland regions where individuals speaking two different languages intermix. Analysis of two data sets of Wechsler intelligence test scores from bilingual, Hispanic American residents of the Rio Grande Valley region of Texas suggested a visualperceptual/visual-motor over language processing advantage of about 1SD. Scores on tests emphasizing visual-perceptual/visual-motor processing skills were in the average range of intellectual ability and were consistent with educational levels, but scores on tests emphasizing language processing were low. The low scores occurred with both English and Spanish language administrations and so were more likely bilingualism-related than a consequence of limited English language proficiency. The size of the visualperceptual/visual-motor over language processing advantage was about double that typically reported for Hispanic American groupings in intelligence test standardization samples (in which the extent of bilingualism is unknown). Analysis of data sets, as was done here, provides an indirect link between bilingualism and intelligence test scores of Hispanic Americans. The analysis has suggested bilingualism is a factor, but does not specifically implicate it. Other potential explanations for the visual-

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Bilingualism and Hispanic American Intelligence Test Scores

195

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

perceptual/visual-motor over language processing advantage include relevant demographic characteristic (like SES or quality of education) unique to the Rio Grande Valley region of Texas where the participants were resident. This is a community that is evolving culturally secondary to large numbers of recent immigrants. There is no reason to assume that immigrants, as a grouping, formed a normal distribution of cognitive skills in their country of origin. In fact it is more likely that the opposite is the case. A more direct test of the relationship between bilingualism and intelligence test scores would come from studies that compared the intelligence test scores of demographically matched groupings of monolingual and bilingual Hispanic Americans. Intelligence test developers could also facilitate study of this issue by reporting the frequency of bilingualism in standardization samples as assessed by self-report. The practical implications of low scores by Hispanic Americans on intelligence test measures that emphasize language processing are not limited to special education placement in K through 12. Intelligence test scores tend to correlate highly with achievement test scores such as the Scholastic Assessment Test (SAT) and the Graduate Record Examination (GRE) that are used to determine admission to college. Bilingual vs. monolingual group differences in these achievement test scores could result in disproportionate numbers of Hispanic Americans being excluded from college, an admission policy that would violate the 14th Amendment. Solutions to this problem are complex, involving variants of affirmative action (Gasquoine, 2008). Why bilingualism might lead to low scores on language based IQ/achievement test measures remains a matter of conjecture. The most plausible explanation developed to date is that parallel activation of both languages causes interlanguage interference thereby slowing processing time and increasing the possibility of errors under certain stressful environmental conditions such as those involved with IQ/achievement testing (Bialystok et al., 2008).

Author Note This research was supported by a Faculty Research Grant from the University of Texas – Pan American to the first author. Correspondence should be addressed to Philip G. Gasquoine, Ph.D., Graduate Psychology Program Director, Department of Psychology and Anthropology, University of Texas – Pan American, 1201 W. University Drive, Edinburg, Texas 78541. Electronic mail may be sent to [email protected].

References Ardila, A., Rosselli, M., Ostrosky-Solis, F., Marcos, J., Granda, G., and Soto, M. (2000). Syntactic comprehension, verbal memory, and calculation abilities in Spanish-English bilinguals. Applied Neuropsychology, 7, 3-16. Artiola i Fortuny, L., Hermosillo, D., Heaton, R.K., and Pardee, R.E. (1999). Manual de normas y procedimientos para la bateria neuropsicologica en Espanol. Tucson, AZ: m Press.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

196

Philip G. Gasquoine, Aracely Cavazos, Juan Cantu et al.

Bialystok, E. (1999). Cognitive complexity and attentional control in the bilingual mind. Child Development, 70, 636-644. Bialystok, E. (2002). Acquisition of literacy in bilingual children: A framework for research. Language Learning, 52, 159-199. Bialystok, E. (2007). Cognitive effects of bilingualism: How linguistic experience leads to cognitive change. International Journal of Bilingual Education and Bilingualism, 10, 210-223. Bialystok, E., Craik, F., and Luk, G. (2008). Cognitive control and lexical access in younger and older bilinguals. Journal of Experimental Psychology: Learning, Memory, and Cognition, 34, 859-873. Bialystok, E., Martin, M. M., and Viswanathan, M. (2005). Bilingualism across the lifespan: The rise and fall of inhibitory control. International Journal of Bilingualism, 9, 103-119. Bialystok, E., and Senman, L. (2004). Executive processes in Appearance-Reality tasks: The role of inhibition of attention and symbolic representation. Child Development, 75, 562579. Boake, C. (2002). From the Binet-Simon to the Wechsler-Bellevue: Tracing the history of intelligence testing. Journal of Clinical and Experimental Neuropsychology, 24, 383-405. Brown, L., Sherbenou, R. J., and Johnson, S. K. (1990). TONI-2. Test of Nonverbal Intelligence. Examiner’s manual. Austin, TX: Pro-edn. Costa, A., Hernandez, M., and Sebastian-Galles, N. (2008). Bilingualism aids conflict resolution: Evidence from the ANT task. Cognition, 106, 59-86. Crawford et al. v. Honig et al. (1994). 37 F. 3d 485, 487. Diana v. State Board of Education. (1973, June 18). C-70-37 RFP (Northern District of California) (Stipulated settlement). Dickens, W. T., and Flynn, J. R. (2006). Black Americans reduce the racial IQ gap: Evidence from standardization samples. Psychological Science, 17, 913-920. Farah, M. J., Shera, D. M., Savage, J. H., Betancourt, L., Giannetta, J. M., Brodsky, N. L., et al. (2006). Childhood poverty: Specific associations with neurocognitive development. Brain Research, 1110, 166-174. Fernandez, K., Boccaccini, M. T., and Noland, R. M. (2007). Professionally responsible test selection for Spanish-speaking clients: A four-step approach for identifying and selecting translated tests. Professional Psychology: Research and Practice, 38, 363-374. Flaugher, R. L. (1978). The many definitions of test bias. American Psychologist, 33, 671679. Gasquoine, P. G. (1999). Variables moderating cultural and ethnic differences in neuropsychological assessment: The case of Hispanic-Americans. Clinical Neuropsychologist, 13, 376-383. Gasquoine, P. G. (2001). Research in clinical neuropsychology with Hispanic American participants: A review. Clinical Neuropsychologist, 15, 2-12. Gasquoine, P. G. (2008). Psychology of multicultural competence: Critical issues. New York: Linus Publishing. Gasquoine, P. G., Croyle K. L., Cavazos-Gonzalez, C., and Sandoval, O. (2007). Language of administration and neuropsychological test performance in neurologically intact Hispanic American bilingual adults. Archives of Clinical Neuropsychology, 22, 991-1001. Goetz, P. J. (2003). The effects of bilingualism on theory of mind development. Bilingualism: Language and Cognition, 6, 1-15.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Bilingualism and Hispanic American Intelligence Test Scores

197

Gollan, T. H., and Acenas, L. A. (2004). What is a TOT? Cognate and translation effects on tip-of-the-tongue states in Spanish-English and tagalong-English bilinguals. Journal of Experimental Psychology: Learning, Memory, and Cognition, 30, 246-269. Gollan, T. H., Fennema-Notestine, C., Montoya, R. I., and Jernigan, T. L. (2007). The bilingual effect on Boston Naming Test performance. Journal of the International Neuropsychological Society, 13, 197-208. Gollan, T. H., Montoya, R. I., Fennema-Notestine, C., and Morris, S. K. (2005). Bilingualism affects picture naming but not picture classification. Memory and Cognition, 33, 12201234. Gollan, T. H., Montoya, R. I., and Werner, G. (2002). Semantic and letter fluency in SpanishEnglish bilinguals. Neuropsychology, 16, 562-576. Gregory, R. J. (2007). Psychological testing: History, principles, and applications (5th ed.). Boston: Allyn and Bacon. Grosjean, F. (1998). Studying bilinguals: Methodological and conceptual issues. Bilingualism: Language and Cognition, 1, 131-149. Harris, J. G., Cullum, C. M., and Puente, A. E. (1995). Effects of bilingualism on verbal learning and memory in Hispanic adults. Journal of the International Neuropsychological Society, 1, 10–16. Harris, J. G., Wagner, B., and Cullum, C. M. (2007). Symbol vs. digit substitution task performance in diverse cultural and linguistic groups. Clinical Neuropsychologist, 21, 800-810. Heaton, R. K., Chelune, G. J., Talley, J. L., Kay, G. G., and Curtiss, G. (1993). Wisconsin Card Sorting Test manual: Revised and expanded. Odessa, FL: Psychological Assessment Resources. Hosp, J. L., and Reschly, D. J. (2004). Disproportionate representation of minority students in special education: Academic, demographic, and economic predictors. Exceptional Children, 70, 185-199. Individuals with Disabilities Education Act. (2004). 20 U.S.C., 1400. Jensen, A. R. (1969). How much can we boost IQ and scholastic achievement? Harvard Educational Review, 39, 1-123. Kaplan, R. M., and Saccuzzo, D. P. (2005). Psychological testing: Principles, applications, and issues, (6th ed.). Belmont, CA: Thompson. Kovacs, A. M. (2007). Beyond language: Childhood bilingualism enhances high-level cognitive functions. In I. Keckes and L. Albertazzi (Eds.), Cognitive aspects of bilingualism. New York: Springer. Larry P. v. Wilson Riles. (1979). 343 F. Supp. 1306 (Northern District California 1972), affd 502 F. 2d 963 (9th Circuit). Lezak, M. D., Howieson, D. B., and Loring, D. W. (2004). Neuropsychological assessment (4th ed.). New York: Oxford University Press. Miller, G. A., and Chapman, J. P. (2001). Misunderstanding analysis of covariance. Journal of Abnormal Psychology, 110, 40-48. Naglieri, J.A., Otero, T., DeLauder, B., and Matto, H. (2007). Bilingual Hispanic children’s performance on the English and Spanish versions of the Cognitive Assessment System. School Psychology Quarterly, 22, 432-488. National Research Council. (2002). Minority students in special and gifted education. Washington, DC: National Academy Press.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

198

Philip G. Gasquoine, Aracely Cavazos, Juan Cantu et al.

Neisser, U., Boodoo, G., Bouchard, T. J., Boykin, A. W., Brody, N., Ceci, S. J., et al. (1996). Intelligence: Knowns and unknowns. American Psychologist, 51, 77-100. Noble, K. G., McCandliss, B. D., and Farah, M. J. (2007). Socioeconomic gradients predict individual differences in neurocognitive abilities. Developmental Science, 10, 464-480. Overton, T., Fielding, C., and Simonsson, M. (2004). Decision making in determining eligibility of culturally and linguistically diverse learners: Reasons given by assessment personnel. Journal of Learning Disabilities, 37, 319-330. Prifitera, A., and Saklofske, D. (1998). WISC-III clinical use and interpretation: Scientistpractitioner perspectives. San Diego, CA: Academic Press. Prifitera, A., Weiss, L. G., and Saklofske, D. H. (1998). The WISC-III in context. In A. Prifitera and D. H. Saklofske (Eds.) WISC-III: Clinical use and interpretation: Scientistpractitioner perspectives. San Diego, CA: Academic Press, pp. 1-38. Puente, A. E., and Salazar, G. D. (1998). Assessment of minority and culturally diverse children. In A. Prifitera and D. H. Saklofske (Eds.) WISC-III clinical use and interpretation: Scientist-practitioner perspectives. San Diego, CA: Academic Press, pp. 227-248. Rhodes, R. L., Ochoa, S. H., and Ortiz, S. O. (2005). Assessing culturally and linguistically diverse students: A practical guide. New York: Guilford Press. Roberts, P. M., Garcia, L. J., Desrochers, A., and Hernandez, D. (2002). English performance of proficient bilingual adults on the Boston Naming Test. Aphasiology, 16, 635-645. Roid, G. H., and Miller, L. J. (1995). Leiter International Performance Scale – Revised. Wood Dale, IL: Stoelting. Rosselli, M. A., Ardila, A., Santisi, M. N., Del Rosario Arecco, M., Salvatierra, J., Conde, A., et al. (2002). Stroop effect in Spanish-English bilinguals. Journal of the International Neuropsychological Society, 8, 819–827. Rushton, J. P., and Jensen, A. R. (2006). The totality of available evidence shows that the race IQ gap still remains. Psychological Science, 17, 921-922. Schmidt, F. L., and Hunter, J. E. (1974). Racial and ethnic bias in psychological tests: Divergent implications of two definitions of test bias. American Psychologist, 29, 1-13. Sternberg, R. J. (2004). Culture and intelligence. American Psychologist, 59, 325-338. Strickland, B. R. (2000). Misassumptions, misadventures, and the misuse of psychology. American Psychologist, 55, 331-338. Stuss, D. T., Gallup, G. G., Jr., and Alexander, M. P. (2001). The frontal lobes are necessary for ‘theory of mind’. Brain, 124, 279-286. U.S. Bureau of the Census. (2000). Population reports: 2000. Washington, DC: Author. Valdes, G., and Figueroa, R. A. (1996). Bilingualism and testing: A special case of bias. Norwood, NJ: Ablex. van de Vijver, F. J. R., and Phalet, K. (2004). Assessment in multicultural groups: The role of acculturation. Applied Psychology: An International Review, 53, 215-236. Wechsler, D. (1991). Wechsler Intelligence Scale for Children - Third Edition. San Antonio, TX: Psychological Corporation. Wechsler, D. (1997). Wechsler Adult Intelligence Scale – Third Edition. San Antonio, TX: Psychological Corporation. Wechsler, D. (2003). Wechsler Intelligence Scale for Children – Fourth Edition. San Antonio, TX: Psychological Corporation.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Bilingualism and Hispanic American Intelligence Test Scores

199

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Wechsler, D. (2005). Wechsler Intelligence Scale for Children – Fourth Edition, Spanish. San Antonio, TX: Psychological Corporation. Williams, W. M., and Ceci, S. J. (1997). Are Americans becoming more or less alike? Trends in race, class, and ability differences in intelligence. American Psychologist, 52, 11261235. Woodcock, R. W., Munoz-Sandoval, A. F., Ruef, M. L., and Alvarado, C. G. (2005). Woodcock-Munoz Language Survey – Revised. Itasca, IL: Riverside Publishing.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved. Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

In: Bilinguals: Cognition, Education and Language Processing ISBN: 978-1-60741-710-1 Editor: Earl F. Caldwell, pp. 201-214 © 2010 Nova Science Publishers, Inc.

Chapter 10

LANGUAGE DEVELOPMENT THROUGH A BILINGUAL LENS Eswen Fava and Rachel Hull∗ Department of Psychology, Texas A and M University College Station, TX, USA

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Abstract Bilingualism is a fertile resource for studying facets of language development and brain plasticity that may not be apparent in monolinguals. This chapter will summarize historical and contemporary findings in the literature, discuss methodological issues that influence their interpretation, and suggest future directions for examining the neural substrates of language development and the consequences of having two languages in one brain.

Introduction Historical Context: Where Have We Come From? The notion that language “lives” in the left hemisphere (LH) emerged from the classic observations of Broca (1861) and Wernicke (1911), who found that patients with (LH) focal lesions showed consistent and systematic behavioral deficits for language. Over time, additional lesion deficit data have largely supported the early findings of Broca and Wernicke, and the assumption that language is typically a left-lateralized processing event has become a generally accepted tenet. However, there has been a noteworthy subset of cases that do not follow the predicted pattern. For instance, numerous studies have found that LH lesions sustained in early childhood do not typically result in language deficits or pronounced speech problems (e.g., Thal, 1997; Vargha-Khadem, O'Gorman, and Watters, 1985; Woods and Carey, 1979, 2000; Woods and Teuber, 1978). Other studies have found the same pattern in young (but not older) patients and have suggested that reorganization of language ∗

E-mail address: [email protected]. 979-845-7146 (phone), 979-845-4727 (fax),

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

202

Eswen Fava and Rachel Hull

representation from the LH to the right hemisphere (RH) is more likely if injury occurs before the age of five (Duncan et al., 1997; Muller et al., 1999; Muller, Rothermel, Behen et al., 1998; Muller, Rothermel, Muzik et al., 1998). Although it was initially assumed that early LH injury caused reorganization only when lesions were extensive or impinged upon language areas, some studies have indicated that a reorganization of left lateralized language processing to the RH can occur in children even when lesions are remote from classic language areas. For example, children with motor deficits (stemming from LH motor cortex damage) showed an atypical left ear advantage during a dichotic listening language task, indicating that the RH was more involved in their language processing (Isaacs, Christie, Vargha-Khadem, and Mishkin, 1996). There has also been some neuroimaging evidence that the brain may bring about a change in lateralization for language even when damage does not involve areas traditionally thought to be critical for language processing. For instance, an fMRI study by DeVos et al., (1995) found an association between RH localization for language processing and childhood onset tumors in the LH (see also Liégeois et al., 2004; Springer et al., 1999). At present, the reason for the association between remote lesions and profound changes in language localization remains unclear, although the involvement of frontal areas in redistribution of language processing has been suggested (Dehaene et al., 1997; Hahne and Friederici, 2001; Hernandez, Martinez, and Kohnert, 2000; Liégeois et al., 2004; Weber-Fox and Neville, 2001; Weber-Fox and Neville, 1999). What is clear, however, is that the brain has the capacity to respond to different kinds of LH damage by redistributing language function to healthy RH areas, particularly when the damage occurs during early development (Booth et al., 2000; Staudt et al., 2002). Thus, even if one accepts that the LH is dominant for language processing in most brain-intact individuals, there is certainly evidence to suggest that neural plasticity can operate to reorganize language representation in response to brain damage. Moreover, quite apart from pathology, it is known that the degree of hemispheric dominance varies with differences in genetic predispositions, developmental experiences, gender, and environment (Springer et al., 1999). Given that early brain damage can alter brain organization for language, it is important to consider what other substantial early experiences could have a similar effect. An obvious experience to examine would be early exposure to more than one language system. That is, what might the neural repercussions be of fitting two language systems into one developing brain? It is critical to note that the lesion-deficit outcomes discussed thus far have almost exclusively involved monolingual patients; indeed, the vast majority of patients whose data contributed to the established belief that language is constrained to the left hemisphere were monolingual (see review by Martin, 2000). Given that bilinguals can be expected to share all of the variations found among monolinguals, as well as additional variations arising from circumstances unique to bilingualism (e.g., the age of second language learning, proficiency in the second language, similarity of the first and second language), it seems reasonable that we might expect more variation in brain organization for language in bilinguals, particularly when the onset of bilingualism occurs during early development. Indeed, hints of a difference between monolinguals and bilinguals in terms of the neural architecture of language were suggested when lesion deficit studies began to appear with bilingual aphasics. Both early (Borod, Koff, Lorch, and Nicholas, 1985; Geschwind and Galaburda, 1985; Rasmussen and Milner, 1977) and more recent studies of bilinguals with brain damage (Booth et al., 2000; Liégeois et al., 2004; Naeser et al., 2004) seem to indicate

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Language Development through a Bilingual Lens

203

evidence of brain plasticity for language beyond that suggested in monolingual lesion studies. For instance, numerous case studies of bilingual aphasics have shown differential post-morbid recovery of the two languages (Abutalebi, Miozzo, and Cappa, 2000; Fabbro, 2001; Vaid, 2002; Vaid and Hull, 2002). In addition, some bilingual aphasics can access one language (but not the other) on a given day, then show the reverse pattern later (Paradis, 1998). There is also some limited evidence to suggest that bilinguals show a higher frequency of crossed aphasia (language deficits following damage to the RH) relative to monolinguals (Karanth and Rangamani, 1988; Vaid, 2002). Although lesion deficit studies can point toward the operation of brain plasticity for language, they leave unclear whether specific language deficits are the result of trauma to a specialized brain component at the lesion site, or if the damaged area is simply part of a larger neural network that mediates a given component of language (Abutalebi, Cappa, and Perani, 2001; Szaflarski, Holland, Schmithorst, and Byars, 2006; Vaid and Hull, 2002). Moreover, although lesion deficit studies can provide time-slice glimpses of static circumstances that are valuable in refining theories and models of language representation, the literature is poorly positioned to analyze language development as a process (Brown et al., 2005; Holland et al., 2001; Schlaggar et al., 2002; Springer et al., 1999; Szaflarski et al., 2002). Nevertheless, these studies taken together suggest specific directions for investigating how brain plasticity may give rise to differential development of language representation in the face of profoundly different language experiences, and identify bilingualism as a unique circumstance to explore the brain’s full potential for developing a neural architecture for language.

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Contemporary Context: Where Are We Now? In response to the compelling evidence from the lesion deficit literature, studies of the bilingual brain have escalated dramatically in the last three decades. The theoretical approaches that have dominated bilingualism research have centered on either the age of second language acquisition or proficiency in the second language, and several hypotheses have been introduced to explain findings in the literature. One of the first was the Age Hypothesis (Genesee, Hamers, Lambert, Mononen, Seitz, and Starck, 1978; Vaid and Genesee, 1980), which advocates cognitive and developmental states as the principle mechanisms that drive functional brain organization for multiple languages. In general, the Age Hypothesis states that the more similar the cognitive and brain maturational states are when the first and second languages are acquired, the more similar will be their functional organization. Thus, the Age Hypothesis posits that the age of second language (L2) acquisition will predict brain functional organization for language. Another of the earliest hypotheses, the Stage Hypothesis (Albert and Obler, 1978; Galloway and Krashen, 1980; Obler, 1981; Schneiderman, 1986), focused instead on L2 proficiency as the principle mechanism driving bilingual language organization. In this view, low L2 proficiency is thought to be associated with a heavier reliance on RH strategies that involve using contextual cues to meaning, whereas increased proficiency yields a shift to LH dominance as syntactic and phonological processing become more automated (see also Ullman, 2004). Thus, the Stage Hypothesis says that L2 proficiency will predict brain functional organization for language.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

204

Eswen Fava and Rachel Hull

More recent hypotheses have attempted to reconcile the Age and Stage approaches. Based on outcomes from two comprehensive meta-analyses of the behavioral bilingual laterality literature, Hull and Vaid (2007) have introduced the Anchoring Hypothesis, which states that the neural architecture of language is established early in development, during a period when the brain is especially plastic (i.e., before the age of six), that this pattern persists into adulthood, and that it anchors later-learned languages to the same pattern. More specifically, for people who learn only one language during early childhood (i.e., monolinguals and late bilinguals who learn only one language before the age of 6), the prevailing pattern is LH dominance, and consequently any languages learned after age 6, regardless of proficiency, will also be LH dominant. However, the degree of LH dominance in late bilinguals is predicted to vary as a function of L2 proficiency; that is, low-proficiency late bilinguals are expected to show more LH involvement relative to their high-proficiency counterparts. In contrast, those who learn two (or more) languages during early childhood (i.e., before age 6) are expected to demonstrate bilateral involvement during the processing of both languages, and also for any languages learned later in life. Hull and Vaid argue that the increased involvement of the RH in early bilinguals may be a consequence of the precocious development of RH-specialized pragmatic strategies to support dual-language processing (e.g., monitoring social usage of the two languages; see Beeman and Chiarello, 1998; Boatman, 2004; Genesee, 1980; Hull and Vaid, 2006; Obler, 1981); thus, the relatively early establishment of and reliance upon these RH pathways is thought to yield a pattern of bilateral involvement for language processing that persists through the lifespan. Two additional hypotheses that predict language organization differences in monolinguals and early bilinguals have recently emerged. The Signature Hypothesis (Kovelman, Baker, and Petitto, 2008) suggests that early, proficient bilinguals will recruit different neural tissue than monolinguals across all language processing contexts. This hypothesis is based largely on behavioral language mastery differences in brain-intact bilinguals and takes the theoretical approach that the languages of early, proficient bilinguals are subserved by neural substrates distinct from those of monolinguals. In contrast, the Functional Switching Hypothesis (Kovelman et al., 2008), which is based on differences in language impairment or recovery patterns among bilingual aphasics, assumes that early, proficient bilinguals will demonstrate recruitment of neural tissue similar to that of monolinguals, but only when bilinguals are processing language in a monolingual context (i.e., while suppressing the unused language). In a bilingual context (e.g., switching between languages), bilinguals are predicted to recruit additional and distinct neural tissue. This hypothesis is consistent with theoretical accounts that suggest bilinguals preferentially rely on an attentional, language switching mechanism that exists outside the language systems per se (Green, 1998; Paradis, 2000). In a test of the Signature and Functional Switching hypotheses, Kovelman et al. (2008) used Near Infrared Spectroscopy (NIRS) to examine whether sustained, early exposure to two languages altered the brain’s organization of processing language. They found that in a monolingual context, classic language areas such as Broca’s Area showed similar recruitment in monolinguals and early bilinguals, but only the bilinguals also showed bilateral recruitment of frontal areas associated with working memory, namely, dorsolateral prefrontal cortex (DLPFC) and inferior frontal cortex (IFC). In a bilingual context, early bilinguals demonstrated a similar pattern as during the monolingual context, but they also showed relatively increased recruitment of the RH working memory areas. The authors concluded that

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Language Development through a Bilingual Lens

205

the outcomes of their studies were consistent with their Signature Hypothesis (but not their Functional Switching Hypothesis). The results of these studies were also consistent with the Anchoring Hypothesis. In what follows, we summarize other findings from the contemporary bilingualism literature that support the Anchoring and Signature hypotheses. Behavioral evidence. Data from behavioral bilingualism studies, as a general rule, have the benefit of well-established paradigms and tasks that demonstrate validity in their measurement of subtle constructs such as pragmatics, lexico-semantics, and syntax. The three most commonly used behavioral paradigms for inferring the lateralization of languages have been dichotic listening, tachistoscopic viewing and dual task, where each method typically relies on a different response measure (accuracy, reaction time, and interference size, respectively). In dichotic listening studies, participants typically hear words in one or the other ear, and accuracy of recall by ear is the dependent variable. In visual half-field studies, words are presented to one or the other visual field, and the dependent variable is either recognition speed or speeded word judgments of different sorts for stimuli varying on visual, phonetic, semantic or syntactic dimensions. Finally, dual task studies typically measure how much interference finger tapping with the left- or right-hand finger (hence engaging the right or left hemisphere, respectively) produces on a concurrent language task, such as reading for comprehension. Behavioral studies and reviews of hemispheric involvement for language in brain-intact individuals have consistently suggested differences in laterality between monolinguals and at least some types of bilinguals (Bialystock and Ryan, 1985; Birdsong and Molis, 2001; Goral, Levy, and Obler, 2002; Hull and Vaid, 2005; Johnson and Newport, 1989; Long, 1990; Newport, Bavelier, and Neville, 2001; Vaid, 1984, 1987; Vaid and Frenck-Mestre, 1991). Especially in light of outcomes from the bilingual aphasia literature, the behavioral results suggest that language representation differences might arise from developmental sources (consistent with the Anchoring and Signature hypotheses). In order to identify and quantify any systematic differences in underlying patterns of language representation in the brain, two recent meta-analyses examined the behavioral language laterality literature and identified several consistent patterns across experimental studies. Hull and Vaid (2006) specifically compared behavioral studies of monolinguals and bilinguals, coding also for variables such as testing paradigm (dichotic listening, visual hemifield, dual task), age of L2 acquisition (early, late), and level of L2 proficiency (proficient, non-proficient). They observed that late bilinguals and monolinguals reliably showed LH language dominance, regardless of proficiency level, whereas early bilinguals reliably showed bilateral involvement for language. A subsequent meta-analysis by Hull and Vaid (2007) focused specifically on disentangling laterality differences among bilinguals. Consistent with findings for bilinguals in their earlier meta-analysis ( Hull and Vaid, 2006), the outcomes demonstrated reliably different patterns of language lateralization for bilinguals with early (bilaterally organized) versus late bilingualism onset (LH dominant). In addition, lateralization within each bilingual subgroup overlapped across first and second languages; that is, both the L1 and L2 of early bilinguals were organized bilaterally (consistent with the Anchoring and Signature hypotheses), whereas both languages were LH dominant in late bilinguals (consistent with the Anchoring hypothesis). Moreover, the 2007 meta-analysis showed increased LH dominance for non-proficient late bilinguals relative to proficient ones (consistent with the Anchoring Hypothesis).

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

206

Eswen Fava and Rachel Hull

Taken together, outcomes from the behavioral bilingual laterality literature clearly point to differences in brain organization for language associated with systematic differences in both the developmental stage during which the L2 is acquired and the amount of experience (i.e., proficiency) with the L2. However, while partial answers to questions about language representation in the bilingual brain have emerged from behavioral sources, such measures are by their nature limited in what they can reveal about neural pathways. Neuroimaging evidence. Several alternative (i.e., non-behavioral) techniques for inferring hemispheric organization for language have more recently emerged, namely electroencephalographic (EEG) recording, event related brain potentials (ERPs), positron emission tomography (PET), functional magnetic resonance imaging (fMRI), and nearinfrared spectroscopy (NIRS). These techniques typically utilize an entirely different form of participant responses relative to behavioral paradigms. Generally speaking, measures of the onset and duration of electrical activity in response to a language stimulus (EEG, ERP) have been used to identify the timing and character of neural activity for language processing, whereas differences in the amount of regional cerebral blood flow during a language task relative to some baseline, non-language task (fMRI, PET, NIRS) have been used to infer the location of neural structures and pathways that support language processing. The remainder of this chapter will focus primarily on neuroimaging evidence about language from NIRS and fMRI (but see Hull and Vaid, 2005, for a review of the relevant EEG/ERP literature). Numerous neuroimaging studies have lent support to the long-standing notion that language in monolinguals is LH dominant (Dehaene-Lambertz, 2000; Hickok and Poeppel, 2000; Holowka and Petitto, 2002; Knecht et al., 2001; Paus et al., 1996; Pujol, Deus, Losilla, and Capdevila, 1999; Scott and Wise, 2004). For instance, a specific role for phonological processing has been indicated in the left IFG (Devlin, Matthews, and Rushworth, 2003; Gabrieli, Poldrack, and Desmond, 1998; Klein, Milner, Zatorre, Meyer, and Evans, 1995), and left superior temporal gyrus (STG) (Hickok, 2001; Hughdal et al., 1999; Karbe, 1995; Parker et al., 2005). In addition, activity in the most posterior and medial part of the STG at the temporo-parietal junction has been associated with speech production rather than perception (Wise et al., 2001), whereas both overlapping and unique areas of neural activation have been identified in frontal and posterior areas during speech production and perception, (Hickok, Buchsbaum, Humphries, and Muftuler, 2003; Hickok and Poeppel, 2000; Parker et al., 2005). Although neuroimaging evidence with late bilinguals has typically demonstrated the traditional LH dominance for language seen in monolinguals (e.g., Klein et al., 1994; Vingerhoets et al., 2003; for reviews, see Indefrey, 2006 and Vaid and Hull, 2002), the outcomes with early bilinguals, in contrast, have been markedly less consistent with the notion of a unilateral LH dominance for language. For instance, one fMRI study of multilingual speakers with varying ages of non-native language acquisition and different levels of non-native language proficiency (Briellman et al., 2004) demonstrated that late multilingual speakers consistently showed LH dominance for processing in all languages, with increased LH activation for less proficient languages. However, the lone early multilingual in this study appeared to show overlapping, bilateral activation patterns for all languages, including those for which the early bilingual was not proficient. These outcomes are consistent with the Anchoring Hypothesis. Another fMRI study used a covert picture naming task to reveal that early bilinguals showed equal activation bilaterally in the left and right STG while switching between L1 and

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Language Development through a Bilingual Lens

207

L2 (Hernandez et al., 2000). In addition, an fMRI study involving deaf and hearing signlanguage users showed that both hearing and deaf American Sign Language (ASL)-English early bilinguals demonstrated bilateral activation during sign language processing, whereas late ASL-English bilinguals and ASL or English monolinguals displayed primarily LH activation (Neville and Bavelier, 1998). Finally, a structural MRI study found increased gray matter density in bilinguals relative to monolinguals in the inferior parietal lobe near the temporo-parietal junction, and significantly more so for early than late bilinguals, suggesting that the onset age of bilingualism and/or the length of experience as a bilingual might alter the very structure of the brain (Mechelli et al., 2004). This outcome suggests that variations in hemispheric dominance for language might be explained in terms of different amounts of gray matter volume and density. Another possibility is that the mechanism of increased dominance or density might be length of exposure and the associated automatization of language processing (Lucas, McKhann, and Ojemann, 2004), although the intrinsic link between exposure and proficiency may be difficult to ultimately disentangle. Taken together, data from behavioral and neuroimaging studies provide converging support for the Anchoring Hypothesis and the Signature Hypothesis. However, innovative approaches are needed to produce new evidence to refine and connect these hypotheses.

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Where Are We Moving? Like behavioral measures, neuroimaging measures have some notable limits on what they can reveal about language and the brain. PET studies involve a radioactive tracer that makes it impractical (even unethical) to test individuals on multiple tasks and/or trials, or even to test certain populations at all. fMRI is temporally sluggish, expensive, unavailable to most researchers, and limited in the paradigms it can test (e.g., overt speech production). That is, extreme sensitivity to motion artifacts often requires modifying tasks for fMRI (e.g., covertly “thinking of” words instead of physically responding), making it impossible to behaviorally validate participants’ responses and quite possibly not capturing the full range of activation associated with actual speech production. NIRS represents an alternative to fMRI in such cases, because it is more robust to motion artifacts and can therefore be used with the same kinds of tasks used in behavioral paradigms (e.g., overt speech production, handwriting, etc.). In addition, NIRS can be safely used with protected populations (e.g., healthy infants) previously restricted in traditional neuroimaging paradigms. However, NIRS is limited to tracking activation in the cortex and thus is not appropriate for measuring subcortical activity, for which fMRI remains the most informative paradigm at present. We suggest that the future of language research lies in the marriage of behavioral and neuroimaging techniques. In studies with adult bilinguals, this strategy of combining techniques has already begun to produce promising outcomes. One recent study used NIRS with an established behavioral measure (i.e., picture-naming) to demonstrate that cortical activity associated with overt speech production was localized to the left (but not right) STG in monolingual adults (Hull, Bortfeld, and Koons, 2009). A study with late bilinguals combined methodologies from a behavioral picture-naming task with a neuroimaging paradigm (language mapping using electronic stimulation of the cortex) and found that left temporal regions were shared for L1 and L2 processing (Lucas, McKhann and Ojemann, 2004). However, the authors further reported that L1 processing was also related to activation

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

208

Eswen Fava and Rachel Hull

in the frontal cortex, whereas L2 processing was related to additional sites in mid- to posterior temporal areas, thereby suggesting a functional separation of languages. Importantly, this study suggests that while a classic left lateralization is present in late bilinguals (consistent with the Anchoring Hypothesis), there may still be potential for the two languages to be separated within the LH. Other evidence for a functional LH separation for L1 and L2 in late bilinguals includes an innovative study by Marian, Spivey, and Hirsch (2003), who used both eye-tracking (a behavioral measure) and fMRI to test late Russian-English bilinguals on a picture-recognition task using both phonetic and visual within- and between-language distractors. The results pointed toward common activation across languages for lower levels of language processing (e.g. phonetic), whereas higher-level processing (e.g. lexical) was supported by distinct areas that appeared to be preferentially associated with one language or the other. In studies with infants, Dehaene-Lambertz and colleagues (Dehaene-Lambertz, Dehaene, and Hertz-Pannier, 2002; Dehaene-Lambertz et al., 2006) used fMRI to demonstrate left lateralized hemodynamic responses to speech in 2-4 month old infants. In a NIRS study with neonates from monolingual environments, newborns showed LH dominance for their native language relative to non-language stimuli (Peña et al., 2003). A separate NIRS study with older infants (aged 6-9 months) also demonstrated preferential activation in the left temporal region in response to native language stimuli by Bortfeld, et al., (2009). In contrast, neonates in an EEG study showed RH dominance for processing the native language, and the same infants showed bilateral activation during subsequent testing at 2.5 months (Radicevic, Vujovic, Jelicic, and Sovilj, 2008). Zaramella et al., (2001) also found bilateral activation when using NIRS to measure newborn infants processing of tonal frequency sweeps (although this stimulus may not capture exactly the same activity as true linguistic input). Thus, although there is growing evidence for a left lateralized preference for native language processing in infants, at least those from monolingual environments, the data are limited and there is no clear consensus on how or when hemispheric specializations may develop in either monolingual or bilingual babies. We suggest that this debate could be addressed using NIRS and other imaging and behavioral techniques concurrently, or at least on the same infants in temporal proximity as well as longitudinally. In addition, future studies combining NIRS and overt speech production paradigms in older children and adults could shed new light on the notion a frontal-posterior network of auditory-motor integration, which Hickok et al. (2003) have suggested guides speech development and supports speech production in monolinguals. It would be interesting and valuable to test the hypothesis using NIRS and behavioral measures of overt speech production, and to extend it to include bilingual participants.

Conclusion The language development literature is at a critical crossroads. Forward progress in understanding exactly how, when, and why the critical junctures emerge hinges on refining how researchers frame and test experimental questions. Ample evidence from behavioral and neuroimaging methods suggests the brain’s developmental sensitivity to the acquisition of multiple language systems, and therefore warrants this as a promising frontier to be explored. We suggest the most effective studies will combine existing techniques to further illuminate

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Language Development through a Bilingual Lens

209

the ways in which early language experience influences brain structural and functional organization. Moreover, the emergence of the NIRS technique will allow us to design unprecedented studies that access previously underexplored populations (namely infants and children). Such studies are anticipated to provide an improved means to address the complexity and range of language organization in the intact, developing brain, and to facilitate necessary connections between existing knowledge and theoretical explanations across research domains.

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

References Abutalebi, J., Cappa, S. F., and Perani, D. (2001). The bilingual brain as revealed by functional neuroimaging. Bilingualism: Language and Cognition, 4, pp 179-190. Abutalebi, J., Miozzo, A., and Cappa, S. F. (2000). Do subcortical structures control language selection‚ in polyglots? Evidence from pathological language mixing. Neurocase, 6(1), 51 - 56. Albert, M., and Obler, L. K. (1978). The bilingual brain: Neuropsychological and neurolinguistic aspects of bilingualism. New York: Academic Press. Beeman, M., and Chiarello, C. (1998). Complementary left and right hemisphere language comprehension. Current Directions in Psychological Science, 7, 2-8. Bialystock, E., and Ryan, E. B. (1985). Toward a Definition of Metalinguistic Skill. MerrillPalmer Quarterly, 31(3), 229-251. Birdsong, D., and Molis, M. (2001). On the evidence for maturational constraints in secondlanguage acquisition. Journal of Memory and Language, 44, 235-249. Boatman, D. (2004). Cortical bases of speech perception: Evidence from functional lesion studies. . Cognition, 92, 47-65. Booth, J., MacWhinney B., Thulborn K.R., Sacco, K., Voyvodic, J.T., and HM, F. (2000). Developmental and lesion effects in brain activation during sentence comprehension and mental rotation. Developmental Neuropsychology, 18(2), 139-169. Borod, J. C., Koff, E., Lorch, M. P., and Nicholas, M. (1985). Channels of Emotional Expression in Patients With Unilateral Brain Damage. Arch Neurol, 42(4), 345-348. Bortfeld, H., Fava, E., and Boas, D. A. (2009). Identifying cortical lateralization of speech processing in infants using near-infrared spectroscopy. Developmental Neuropsychology, 34(1), 52-65. Briellmann, R. S., Saling, M. M., Connell, A. B., Waites, A. B., Abbott, D. F., and Jackson, G. D. (2004). A high-field functional MRI study of quadri-lingual subjects. Brain and Language, 89, 531-542. Broca, P. (1861). Remarques sur le siege de la faculte du langage articule; suivies d’une observation d’aphemie. Paper presented at the Bulletin de la Société Anatomique de Paris, Paris, France. Brown, T. T., Lugar, H. M., Coalson, R. S., Miezin, F. M., Petersen, S. E., and Schlaggar, B. L. (2005). Developmental Changes in Human Cerebral Functional Organization for Word Generation. Cereb. Cortex, 15(3), 275-290. Dehaene, S., Dupoux, E., Mehler, J., Cohen, L. G., Paulesu, E., and Perani, D. (1997). Anatomical variability in the cortical representation of first and second language. NeuroReport, 8, 3809-3815.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

210

Eswen Fava and Rachel Hull

Dehaene-Lambertz, G. (2000). Cerebral Specialization for Speech and Non-Speech Stimuli in Infants. Journal of Cognitive Neuroscience, 12(3), 449-460. Dehaene-Lambertz, G., Dehaene, S., and Hertz-Pannier, L. (2002). Functional Neuroimaging of Speech Perception in Infants. Science, 298. Dehaene-Lambertz, G., Hertz-Pannier, L., Dubois, J., Meriaux, S., Roche, A., Sigman, M., et al. (2006). Functional organization of perisylvian activation during presentation of sentences in perverbal infants. Proceedings of the National Academy of Sciences, 103, 14240-14245. Devlin, J. T., Matthews, P. M., and Rushworth, M. F. S. (2003). Semantic Processing in the Left Inferior Prefrontal Cortex: A Combined Functional Magnetic Resonance Imaging and Transcranial Magnetic Stimulation Study Journal of Cognitive Neuroscience, 15(1), 71-84. DeVos, K., Wyllie, E., Geckler, C., Kotagal, P., and Comair, Y. (1995). Language dominance in patients with early childhood tumors near left hemisphere language areas. Neurology, 45(2), 349-356. Duncan, J., Moss, S., Bandy, D., Manwaring, K., Kaplan, A., and Reinian, E. (1997). Use of positron emission tomography for presurgical localization of eloquent brain areas in children with seizures. Pediatric Neurosurgery, 26, 144-156. Fabbro, F. (2001). The Bilingual Brain: Cerebral Representation of Language. Brain and Language, 79(2), 211-222. Gabrieli, J. D. E., Poldrack, R. A., and Desmond, J. E. (1998). The role of left prefrontal cortex in language and‚Äâmemory. Proceedings of the National Academy of Sciences of the United States of America, 95(3), 906-913. Galloway, L. (1980). Cerebral Organization in Bilingualism and Second Language. Paper presented at the Research in Second Language Acquisition: Selected Papers of the Los Angeles Second Language Acquisition Research Forum. Genesee, F., Hamers, J., Lambert, W. E., Mononen, L., Seitz, M., and Starck, R. (1978). Language processing strategies in bilinguals: A neuropsychological study. Brain and Language, 5, 1-12. Geschwind, N., and Galaburda, A. M. (1985). Cerebral lateralization: Biological mechanisms, associations, and pathology: III. A hypothesis and a program for research. Archives of Neurology, 42(7), 634-654. Goral, M., Levy, E., and Obler, L. (2002). Neurolinguistic aspects of bilingualism. International Journal of Bilingualism, 6(4), 411-440. Green, D. W. (1998). Mental control of the bilingual lexico-semantic system. Bilingualism: Language and Cognition, 1, 67-81. Hahne, A., and Friederici, A. (2001). Processing a second language: late learners' comprehension mechanisms as revealed by event-related brain potentials. Bilingualism: Language and Cognition, 4(2), 123-141. Hernandez, A. E., Martinez, A., and Kohnert, K. (2000). In Search of the Language Switch: An fMRI Study of Picture Naming in Spanish-English Bilinguals. Brain and Language, 73(3), 421-431. Hickok, G. (2001). Functional anatomy of speech perception and speech production: Psycholinguistic implications. Journal of Psycholinguistic Research, 30, 225-234.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Language Development through a Bilingual Lens

211

Hickok, G., Buchsbaum, B., Humphries, C., and Muftuler, T. (2003). Auditory-Motor Interaction Revealed by fMRI: Speech, Music, and Working Memory in Area Spt. Journal of Cognitive Neuroscience, 15(5), 673-682. Hickok, G., and Poeppel, D. (2000). Towards a functional neuroanatomy of speech perception. TRENDS in Cognitive Sciences, 4(4), 131-138. Holland, S. K., Plante, E., Weber Byars, A., Strawsburg, R. H., Schmithorst, V. J., and Ball, W. S. (2001). Normal fMRI Brain Activation Patterns in Children Performing a Verb Generation Task. NeuroImage, 14(4), 837-843. Holowka, S., and Petitto, L.-A. (2002). Left Hemisphere Cerebral Specialization for Babies While Babbling. Science, 297(5586), 1515. Hughdal, K., Bronnick, K., Kyllingsbrk, S., Law, I., Gade, A., and Paulson, O. B. (1999). Brain activation during dichotic presentations of consonant-vowel and musical instrument stimuli: a 15O-PET studyfn2. Neuropsychologia, 37(4), 431-440. Hull, R., Bortfeld, H., and Koons, S. (2009). Near-infrared spectroscopy and cortical responses to speech production. The Open Neuroimaging Journal, 3, 26–30. Hull, R., and Vaid, J. (2005). Clearing the cobwebs from the study of the bilingual brain: Toward converging evidence laterality and ERP. In J. Kroll and A. M. B. deGroot (Eds.), Handbook of bilingualism: Psycholinguistic approaches (pp. 480-496). London: Oxford University Press. Hull, R., and Vaid, J. (2006). Laterality and language experience. Laterality: Asymmetries of Body, Brain and Cognition, 11(5), 436 - 464. Hull, R., and Vaid, J. (2007). Bilingual language lateralization: A meta-analytic tale of two hemispheres. Neuropsychologia, 45, 1987-2008. Indefrey, P. (2006). A Meta-analysis of Hemodynamic Studies on First and Second Language Processing: Which Suggested Differences Can We Trust and What Do They Mean? In M. Gullberg and P. Indefrey (Eds.), The cognitive neuroscience of second language acquisition. (pp. 279-304). Malden, MA: Blackwell. Isaacs, E. B., Christie, D., Vargha-Khadem, F., and Mishkin., M. (1996). Effects of hemispheric side of injury, age at injury, and presence of seizur disorder on functional ear and hand asymmetries in hemiplegic children. Neuropsychologia, 34, 127-137. Johnson, J. S., and Newport, E. L. (1989). Critical Period Effects in Second Language Learning: The Inluence of Maturational State on the Acquisition of English as a Second Language. Cognitive Psychology, 21, 60-99. Karanth, P., and Rangamani, G. N. (1988). Crossed aphasia in multilinguals. Brain and Language, 34(1), 169-180. Karbe, H. (1995). Long-term prognosis of poststroke aphasia studied with positron emission tomography. . Archives of Neurology, 52(2), 186-190. Klein, D., Milner, B., Zatorre, R., Meyer, E., and Evans, A. C. (1995). The neural substrates underlying word generation: a abilingual functional-imaging sutdy. PNAS, 92(7), 28992903. Knecht, S., Dräger, B., Flöel, A., Lohmann, H., Breitenstein, C., Deppe, M., et al. (2001). Behavioural relevance of atypical language lateralization in healthy subjects. Brain:, 124(8), 1657-1665. Kovelman, I., Baker, S. A., and Petitto, L.-A. (2008). Bilingual and Monolingual Brains Compared: A Functional Magnetic Resonance Imaging Investigation of Syntactic

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

212

Eswen Fava and Rachel Hull

Processing and a Possible and#x201C;Neural Signatureand#x201D; of Bilingualism. Journal of Cognitive Neuroscience, 20(1), 153-169. Liégeois, F., Connelly, A., Cross, J. H., Boyd, S. G., Gadian, D. G., Vargha-Khadem, F., et al. (2004). Language reorganization in children with early- onset lesions of the left hemisphere: an fMRI study Brain, 127, 1229-1236. Long, M. (1990). Maturational constraints on language development. Studies in Second Language Acquisition, 12, 251-285. Lucas, T. H., McKhann, G. M., and Ojemann, G. A. (2004). Functional sep- aration of languages in the bilingual brain: A comparison of electrical stimulation language mapping in 25 bilingual patients and 117 monolin- gual control patients. Journal of Neurosurgery, 101, 449–445. Marian, V., Spivey, M., and Hirsch, J. (2003). Shared and separate systems in bilingual language processing: Converging evidence from eyetracking and brain imaging. Brain and Language, 86(1), 70-82. Martin, R. C. (2000). Contributions from the neuropsychology of language and memory to the development of cognitive theory. Journal of Memory and Language, 43, 149-156. Mechelli, A., Crinion, J. T., Noppeney, U., O'Doherty, J., Ashburner, J., Frackowiak, R. S., et al. (2004). Neurolinguistics: Structural plasticity in the bilingual brain. Nature, 431(7010), 757-757. Muller, R.-A., Behen, M., Rothermel, R., Muzik O, Chakraborty, P., and Chugani, H. (1999). Brain organization for language in children, adolescents, and adults with left hemisphere lesion: a PET study. Progressive Neuropsychopharmacology and Biological Psychiatry, 23, 657-668. Muller, R.-A., Rothermel, R., Behen, M., Muzik, O., Mangner, T., and Chakraborty, P. (1998). Brain organization of language after early unilateral lesion: a PET study. Brain and Language, 62, 422-451. Muller, R.-A., Rothermel, R., Muzik, O., Becker, C., Fuerst, D., and Behen, M. (1998). Determination of language dominance by [15O]-water PET in children and adolescents: a comparison with the Wada Test. Journal of Epilepsy, 11, 152-161. Naeser, M. A., Martina, P. I., Bakera, E. H., Hodgea, S. M., Sczerzeniea, S. E., Nicholasa, M., et al. (2004). Overt propositional speech in chronic nonfluent aphasia studied with the dynamic susceptibility contrast fMRI method NeuroImage, 22(1), 29-41. Neville, H. J., and Bavelier, D. (1998). Neural organization and plasticity of language. Current Opinion in Neurobiology, 8(2), 254-258. Newport, E. L., Bavelier, D., and Neville, H. J. (2001). Critical thinking about critical periods: Perspectives on a critical period for language acquisition. In E. Dupoux (Ed.), Language, brain, and cognitive development: Essays in honor of Jacques Mehler (pp. 481-502). Cambridge, MA: MIT Press. Obler, L. K. (1981). Right Hemisphere Participation in Second Language Acquisition. In K. Diller (Ed.), Individual Differences and Universals in Language Learning Aptitude (pp. 53-64). Rowley, MA Newbury Press. Paradis, M. (1998). The other side of language: Pragmatic competence Journal of Linguistics, 11(1-2), 1-10. Paradis, M. (2000). Language lateralization in bilinguals: enough already! In L. Wei (Ed.), The Bilingualism Reader (pp. 365-372): Routledge.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Language Development through a Bilingual Lens

213

Parker, G. J. M., Luzzi, S., Alexander, D. C., Wheeler-Kingshott, C. A. M., Ciccerelli, O., and Lambon Ralph, M. A. (2005). Lateralization of ventral and dorsal auditory-language pathways in the human brain NeuroImage, 24(3), 656-666. Paus, T., Tomaiuolo, F., Otaky, N., MacDonald, D., Petrides, M., Atlas, J., et al. (1996). Human Cingulate and Paracingulate Sulci: Pattern, Variability, Asymmetry, and Probabilistic Map Cerebral Cortex, 6, 207-214. Peña, M., Maki, A., Kovacić, D., Dehaene-Lambertz, G., Koizumi, H., Bouquet, F., et al. (2003). Sounds and silence: an optical topography study of language recognition at birth. Proceedings of the National Academy of Sciences of the United States of America., 100(20), 11702-11705. Pujol, J., Deus, J., Losilla, J. L., and Capdevila, A. (1999). Cerebral lateralization of language in normal left-handed people studied by functional MRI Neurology, 52, 1038. Radicevic, Z., Vujovic, M., Jelicic, L., and Sovilj, M. (2008). Comparative findings of voice and speech: language processing at an early ontogenetic age in quantitative EEG mapping. Experimental Brain Research, 184(4), 529-532. Rasmussen, T., and Milner, B. (1977). The role of early left-brain injury in determining lateralization of cerebral speech functions. Annals of the New York Academy of Science, 299(September), 355-369. Schlaggar, B. L., Brown, T. T., Lugar, H. M., Visscher, K. M., Miezin, F. M., and Petersen, S. E. (2002). Functional Neuroanatomical Differences Between Adults and School-Age Children in the Processing of Single Words. Science, 296(5572), 1476-1479. Schneiderman, E. I. (1986). Right Hemisphere Processing and Foreign Language Aptitude. ITL: Review of Applied Linguistics, 71, 43-64. Scott, S., and Wise, R. (2004). The functional neuroanatomy of prelexical processing in speech perception. Cognition, 92(1-2 ), 13-45. Springer, J. A., Binder, J. R., Hammeke, T. A., Swanson, S. J., Frost, J. A., Bellgowan, P. S. F., et al. (1999). Language dominance in neurologically normal and epilepsy subjects: A functional MRI study. Brain, 122(11), 2033-2046. Staudt, M., Lidzba, K., Grodd, W., Wildgruber, D., Erb, M., and Kr‰geloh-Mann, I. (2002). Right-Hemispheric Organization of Language Following Early Left-Sided Brain Lesions: Functional MRI Topography. NeuroImage, 16(4), 954-967. Szaflarski, J. P., Binder, J. R., Possing, E. T., McKiernan, K. A., Ward, B. D., and Hammeke, T. A. (2002). Language lateralization in left-handed and ambidextrous people: fMRI data. 59(2), pp 238-244. Szaflarski, J. P., Holland, S. K., Schmithorst, V. J., and Byars, A. W. (2006). An fMRI study of language lateralization in children and adults. Human Brain Mapping, 27(3), 202-212. Thal, D. J. (1997). Origins of language disorders. Developmental Neuropsychology, 13(Special Issue: Origins of language disorders), 233-237. Ullman, M. T. (2004). Contributions of memory circuits to language: the declarative/procedural model. Cognition, 92(1-2), 231-270. Vaid, J. (1984). On Readings on Aphasia in Bilinguals and Polyglots. Journal of the History of the Behavioral Sciences, 20, 372-374. Vaid, J. (1987). Visual field asymmetries for rhyme and syntactic category judgements in monolinguals and fluent early and late bilinguals. Brain and Language, 30, 263-277. Vaid, J. (2002). Bilingualism. In V. S. Ramachandran (Ed.), Encyclopedia of the human brain (Vol. 1, pp. 417-434). San Diego: Elsevier.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

214

Eswen Fava and Rachel Hull

Vaid, J., and Frenck-Mestre, C. A. (1991). Incidental memory for format of presentation of number stimuli: Evidence from monolinguals and bilinguals. Brain and Cognition, 17(2), 272-284. Vaid, J., and Genesee, F. (1980). Neuropsychological approaches to bilingualism: A critical review. . Canadian Journal of Psychology(34), 417-445. Vaid, J., and Hull, R. (2002). Re-envisioning the bilingual brain using functional neuroimaging: Methodological and interpretive issues. In F. Fabbro (Ed.), In the Neurolinguistics of Bilingualism: Essays in Honor of Michel Paradis. (pp. 315-355). Udine Forum: Udine University Press. Vargha-Khadem, F., O'Gorman, A. M., and Watters, G. V. (1985). Aphasia and Handedness in Relation to Hemisphere Side, Age at Injury and Severity of Cerebral Lesion during Childhood. Brain, 108(3), 677-696. Vingerhoets, G., Van Borsel, J., Tesink, C., van den Noort, M., Deblaere, K., Seurinck, R., et al. (2003). Multilingualism: an fMRI study. NeuroImage, 20, 2181-2196. Weber-Fox, C., and Neville, H. J. (2001). Sensitive Periods Differentiate Processing of Openand Closed-Class Words: An ERP Study of Bilinguals. J Speech Lang Hear Res, 44(6), 1338-1353. Weber-Fox, C.-M., and Neville, H.-J. (1999). Functional neural subsystems are differentially affected by delays in second language immersion: ERP and behavioral evidence in bilinguals. In D. E. Birdsong (Ed.), Second language acquisition and the Critical Period Hypothesis. Second language acquisition research. (pp. 23-38). (pp. 191 pp). Mahwah, NJ, US: Lawrence Erlbaum Associates, Inc. Wernicke, C. (1911). The symptom of complex aphasia. In C. AE (Ed.), Diseases of the nervous system (pp. 265–324). New York: Appleton. Wise, R. K., Scott, S., Blank, C. S., Mummery, C. J., Murphy, K., and Warburton, E. A. (2001). Separate neural subsystems within `Wernicke's area' Brain, 124(1), 83-95. Woods, B. T., and Carey, S. (1979). Language deficits after apparent clinical recovery from childhood aphasia. Annals of Neurology, 6(5), 405-409. Woods, B. T., and Teuber, H.-L. (1978). Changing patterns of childhood aphasia. Annals of Neurology, 3(3), 273-280. Zaramella, P., Freato, F., Amigoni, A., Salvadori, S., Marangoni, P., Suppjei, A., et al. (2001). Brain auditory activation measured by near-infrared spectroscopy (NIRS) in neonates. Pediatric Research, 49, 213-219.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

In: Bilinguals: Cognition, Education and Language Processing ISBN: 978-1-60741-710-1 Editor: Earl F. Caldwell, pp. 215-227 © 2010 Nova Science Publishers, Inc.

Chapter 11

IMPROVING READING SKILLS FOR ESL LEARNERS USING SOUNDSPEL

Michael D. Young, Michelle L. Wilson and Alice F. Healy University of Colorado, USA

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Abstract This study examined the effects of using a revised, transparent spelling system SoundSpel, a phonetic reading tool, with learners of English as a Second Language. During 6 training sessions, 12 participants used unaltered material and 12 used SoundSpel texts, in parallel with standard English, when reading American elementary school material. They then answered multiple-choice comprehension questions. Both groups were pre-tested and post-tested on comprehension tests of similar elementary school material without SoundSpel. No group differences were found across tests or training (in quiz performance or reading time), suggesting no beneficial or harmful effects from using SoundSpel. A post hoc analysis suggested that SoundSpel would be most beneficial for students who learn to speak English before they learn to read it.

Introduction Learning to read is an important goal of elementary education. Specifically students need to move from sounding out words to sight reading, except, perhaps, for low-frequency words. Children typically have a larger speaking vocabulary than reading vocabulary, so a challenge for them is to match their speaking vocabulary to printed words. Towards this end, young readers often make use of dictionaries with phonetic notation. Mastering the spelling system, which includes 26 letters and approximately 42 sounds, can be a difficult task, as evident in the following example: “Though the rough cough and the hiccough plough me through, I ought to cross the lough” (Mole, 2003, p. 4). To read this sentence successfully, the pronunciation of each of these instances of the letter string -ough would need to be memorized. In this way, English differs from other, more transparent languages that have better letter-to-sound correspondence, and hence would require less memorization.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

216

Michael D. Young, Michelle L. Wilson and Alice F. Healy

Illustrating this fact, Patel, Snowing, and de Jong (2004) showed that Dutch children (ages 611) reading Dutch, which is a phonetically transparent language, were more accurate and faster at pronouncing visually presented words and nonwords than were comparable American children reading English, especially at younger ages. Reading skill development has been shown to be highly associated with phonological processing skills. Evidence that phonological awareness is related to reading comes from longitudinal studies that have found that differing degrees of phonological awareness among children (before they even learn to read) can predict reading skill through at least age 6 (e.g., Bryant, MacLean, Bradley, and Crossland, 1990). Some phonological awareness develops early, although awareness of phonemes as units develops later than an awareness of syllables, onsets, and rimes. Phonological awareness can develop without reading instruction, but in the absence of reading an alphabetic system, this capacity seems to be retarded (Mann, 1986). However, it has been shown that specific training in phonological awareness can help with remediation of reading problems, particularly for poor adult readers learning an alphabetic system (Rayner, Foorman, Perfetti, Pesetsky, and Seidenberg, 2001). For example, training with “talking computers” that pronounce words in text has been effective at improving word recognition, phonological decoding, and phonological awareness in children with reading difficulties (Olson, Wise, Ring, and Johnson, 1997). In the present study, phonological processing is investigated by using a program called SoundSpel, developed by the American Literacy Council (Mole, 2003), which converts English text into a more predictable spelling system. Specifically, SoundSpel is a phonetic spelling system with unambiguous, easy to learn, letter-to-sound correspondences, including improved vowel representations and—unlike standard dictionaries—no additional characters or diacritical marks. English text can be written using a format called DoubleLine, in which every line of text appears in parallel with the same line converted directly below into SoundSpel, with the SoundSpel version of a given standard English word placed immediately under it. If a reader encounters an unfamiliar word in English, using DoubleLine, he or she can refer to the line below, which in effect contains a transcription (i.e., a respelling) of the word according to its pronunciation. For instance, the word “glad” is pronounced as it is spelled, relying on no special rule or exception for proper pronunciation. In SoundSpel, it would be written “glad” as in English. Yet the word “give” is an instance when the phonological representation of a word lacks congruency with its graphemic representation because of the terminal silent letter “e”. That word in SoundSpel would be written “giv” (without the superfluous final letter “e”). Thus, DoubleLine would be useful because it could function as a point of reference, representing words according to their phonological features, thereby providing a systematic link between written and spoken language. In the present study, we explore the possibility of using DoubleLine to improve reading comprehension with adults learning English as a second language (ESL). Two groups of nonnative English speakers were compared: The SoundSpel group was instructed in SoundSpel and read English text written in DoubleLine during training, and the control group was not exposed to SoundSpel or DoubleLine but read the same English text written so that every line of text appeared in parallel with the same line directly below it (i.e., duplicated). Before and after training, both groups of participants were given tests. The pretest was equivalent to the posttest, and neither test included DoubleLine or duplicated lines; instead the text was printed in ordinary English. During training and testing, each English passage that was read was followed by multiple-choice comprehension questions. Two measures of performance were

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Improving Reading Skills for ESL Learners Using SoundSpel

217

collected: number correct on the reading comprehension test of a given passage and total time to read the passage and answer the comprehension questions. The two passages used in testing were both at the fourth grade level, whereas the six passages used in training included three at the third grade level followed by three at the fourth grade level. These reading levels were selected to match the current reading levels of the ESL participants as closely as possible. If DoubleLine effectively promotes phonological awareness in adults, then we should find that performance on both measures is better during training for the SoundSpel group than for the control group. Most importantly, if the phonological training derived from SoundSpel transfers to reading normal English, then any improvement in performance from the pretest to the posttest should be larger for the SoundSpel group than for the control group. Thus, a comparison of SoundSpel and control groups during training would reveal any immediate benefits of DoubleLine, whereas a comparison of the two groups on the posttest would reveal any long-term benefits of prior exposure to DoubleLine.

Method

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Participants Participants for this study were recruited from the International English Center ESL program at the University of Colorado and the InterCambio de Comunidades ESL program in the Boulder community. This population was limited to non-native English speaking adults with a reading level ranging from approximately third to fourth grade. A total of 24 participants were included in the analyses. An additional participant was tested, but the data from that participant were excluded because that participant did not follow directions in one of the training sessions. The 24 participants could be divided by native language background into the following groups: Arabic (2), French (2), Korean (6), Japanese (4), Nepali (1), Spanish (8), and Ukrainian (1). Recruitment methods included informational fliers and wordof-mouth. Upon completion of the entire eight-part study, participants were compensated $100 for their time.

Design For the tests, one between-subjects independent variable was the experimental condition (control, SoundSpel). An additional independent between-subjects variable was the order of the pretest and posttest (A then B, B then A). Test type (pretest, posttest) was a withinsubjects independent variable. Other within-subjects variables that pertained to training but not the tests were grade level of material (third or fourth) and story number (first, second, or third) within a given grade level. The primary dependent variable of this study was reading comprehension, measured by the number of correct responses on the quizzes. Time taken to read each text and respond to the multiple-choice comprehension questions constituted a second dependent variable.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

218

Michael D. Young, Michelle L. Wilson and Alice F. Healy

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Materials and Apparatus Materials used for this study were from American third and fourth grade reading passages taken from Collections: In Good Company and Collections: Sometimes I Wonder (Allington, Cortez, Cunningham, Sebesta, and Tierney, 1989b, 1989c). In the test portion of the study, occurring at the beginning and end of the experiment, participants read the unmodified fourth grade Texts A and B in a counterbalanced order. Text A was 12 pages long (1838 words) with a picture on each page. Text B was 20 pages long (5766 words) with 9 small illustrations. The texts used in the training portion of this study varied according to condition. The control condition training texts were double-spaced paragraphs with an identical line written below each line of the paragraph. Texts for the SoundSpel group were also in the format of doublespaced paragraphs; however, below each line that appeared in standard English spelling was that same line converted to SoundSpel DoubleLine (see Appendix A). All documents for the SoundSpel group had been converted using a Microsoft Word macro, provided by the American Literacy Council, and printed in Courier font. The first three texts were at the third grade level, and the last three texts were at the fourth grade level. Training Text 3-1 was 42 pages long (4217 words) with 21 illustrations. Text 3-2 was 22 pages (2192 words) with 14 pictures, and Text 3-3 was 28 pages (3154 words) with 9 pictures. Training Text 4-1 was 58 pages (6965 words) and had 9 illustrations, Text 4-2 was 54 pages (6631 words) with 8 pictures, and Text 4-3 was 52 pages (6714 words) and had 14 illustrations. Multiple-choice questions (four questions per text) were used to quiz reading comprehension. These multiple choice questions were developed by the experimenters and tested for clarity using native speaking undergraduates in a preliminary experiment (see Appendix B). Each multiple-choice question had three different answers to choose from. Most of the participants in the SoundSpel condition (except the first four tested) used a standard ruler to minimize any distraction of the double lines in reading these passages. (They were told to hide the SoundSpel line when reading the standard English text unless it was needed to decipher the standard English text.) Participants in the control condition were not given the ruler. Time to complete reading each passage and answering the comprehension questions was recorded in minutes by the experimenter using a watch or wall clock.

Procedure This study consisted of eight sessions: pretest, six training sessions, and posttest. With the exception of 4 participants (whose sessions were all on 8 separate days), sessions were paired over the course of 4 days—both were administered on a given day with a short break in between sessions (e.g., 5 min). Participants were assigned to one of two conditions based on a fixed rotation depending on their time of arrival for the first session: the control condition or the SoundSpel condition. Because there was no systematic relation between time of arrival and language background, assignment of participants to conditions was essentially random. All participants were then handed an unmodified pretest text (A or B), followed by a comprehension quiz including four multiple-choice questions. Participants in the experimental condition then received approximately 5 min of training on the SoundSpel spelling system from the experimenter.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Improving Reading Skills for ESL Learners Using SoundSpel

219

Instructing participants in SoundSpel involved introducing every sound-to-letter(s) representation for the consonants (which remained relatively unchanged). Instruction proceeded by then introducing the less familiar, but systematic, SoundSpel representation of short and long vowels. Participants in the SoundSpel group saw all the consonants and then all the vowels on PowerPoint slides or printouts of the slides (two for the consonants followed by two for the vowels), with a sample word or words containing a given letter to illustrate that letter’s pronunciation. The experimenter pronounced aloud for the participant each letter and many of the sample words (all of those for the vowels). After participants in the SoundSpel condition were tutored on SoundSpel, they were instructed to read aloud a short passage transcribed using SoundSpel (see Appendix C). The task of reading aloud the transcribed passage equipped the participant with (a) some familiarity in using the SoundSpel system, (b) an opportunity to receive feedback from the experimenter, and (c) a chance to demonstrate that he or she was, in fact, able to comprehend and use this new phonological tool. Training sessions followed the initial SoundSpel instruction and testing session. There were six training sessions, which were broken up by grade level. The first three training session texts involved third grade stories. The subsequent three training session texts involved fourth grade stories. After reading each text, the experimenter administered a multiple-choice quiz consisting of four multiple-choice questions regarding its corresponding story. Participants completed the quiz without the aid of the text. Generally, the experimenter started timing upon distribution of the passage and ended upon completion of the passage’s corresponding comprehension questions, although for some or all sessions of 4 of the participants (2 in each condition) only the reading of the passage was timed. Following these six training sessions was the final test session, in which the posttest was given. If participants were initially administered Text A, they received Text B following the six training sessions (or vice-versa). Time needed to complete this experiment was estimated to be a total of four 2-hour sessions (or eight 1-hour sessions for 4 participants), although a given session often took somewhat less time than estimated.

Results Tests A mixed factorial analysis of variance was conducted on the number of correct responses, out of four total questions per test, for comprehension questions on tests before and after training, both of which were based on fourth grade materials (see Figure 1). The analysis included the between-subjects factor of condition (control, SoundSpel) and the withinsubjects factor of test type (pretest, posttest). There were no significant main effects or interaction, F(1, 22) < 1 in each case. Specifically, there was no significant improvement from the pretest to the posttest for either the control condition (pretest = 3.000, posttest = 3.000) or the SoundSpel condition (pretest = 3.167, posttest = 3.000). A second mixed factorial analysis of variance, including the same factors, was conducted on the response time (in minutes) to complete reading each passage and responding to the four multiple-choice questions. Again, there were no significant main effects or interaction; F(1, 22) < 1 both for the main effect of condition and for the interaction of condition and test type; F(1,22) = 2.331, MSE = 151.042, p = .1411 for the main effect of test type. The means,

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

220

Michael D. Young, Michelle L. Wilson and Alice F. Healy

which are provided in Figure 2, show a small decrease in response time across tests of comparable magnitude for the control condition (pretest = 33.750 min, posttest = 28.333 min) and the SoundSpel condition (pretest = 33.750 min, posttest = 28.333 min).

Mean Number Correct (out of 4)

4

3

Pretest Posttest

2

1

0

Control

SoundSpel

Subject Group

Figure 1. Mean number correct (out of 4 questions total per test) on pretest and posttest for two conditions. Error bars show standard errors of the mean.

Mean Reading Time (in minutes)

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

40

30

Pretest Posttest

20

10

0

Control

SoundSpel

Subject Group

Figure 2. Mean reading time (in minutes) on pretest and posttest for two conditions. Error bars show standard errors of the mean.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Improving Reading Skills for ESL Learners Using SoundSpel

221

Thus, participants in the SoundSpel condition did not differ significantly from those in the control condition in terms of their performance on the tests before and after the training phase with respect to both accuracy and speed.

Training With respect to training, a mixed factorial analysis of variance was conducted on the number of correct responses (out of four total questions per story) on the comprehension questions given after each passage (see Figure 3). The analysis included the between-subjects factor of condition (control, SoundSpel) and the within-subjects factors of passage grade level (third, fourth) and story number within grade level (1, 2, 3). The analysis revealed a significant main effect of story number, F(2, 44) = 11.932, MSE = 0.514, p < .0001, reflecting a decline in the number of correct responses from Story 1 (3.667) to Story 2 (2.958), with Story 3 (3.229) in between. Also, there was a significant main effect of grade level, F(1, 22) = 6.459, MSE = 0.474, p = .0186, reflecting a lower mean number of correct responses for the fourth grade stories (3.139) than for the third grade stories (3.431). In addition, there was a significant interaction of story number and grade level, F(2, 44) = 4.421, MSE = 0.457, p = .0178. For the third grade stories, the mean number of correct responses was highest for the first story (3.875), lowest for the second story (2.875), and intermediate for the third story (3.542). For the fourth grade stories, the mean number of correct responses decreased across stories (first = 3.458; second = 3.042; third = 2.917). Control SoundSpel

Mean Number Correct (out of 4)

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

4

3

2

1

0

First

Second

Third Grade

Third

First

Second

Third

Fourth Grade

Figure 3. Mean number correct (out of 4 questions total per story) during training session for two conditions as a function of grade level and story number. Error bars show standard errors of the mean.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

222

Michael D. Young, Michelle L. Wilson and Alice F. Healy 60 Control SoundSpel

Mean Reading Time (in minutes)

50

40

30

20

10

0

First

Second

Third Grade

Third

First

Second

Third

Fourth Grade

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Figure 4. Mean reading time (in minutes) during training session for two conditions as a function of grade level and story number. Error bars show standard errors of the mean.

These differences across stories may be due to the content, length, and nature of the stories themselves, which were not counterbalanced across participants, because the stories were distributed in a fixed order within each grade level. There were no other significant effects; in particular, there was not a main effect of condition, F(1, 22) < 1, an interaction of condition and grade level, F(1, 22) < 1, an interaction of condition and story number, F(2, 44) = 1.095, MSE = 0.514, p = .3436, or a three-way interaction of condition, grade level, and story number, F(2, 44) < 1. Thus, there were essentially no differences between the control and SoundSpel conditions in accuracy during training. A mixed factorial analysis of variance was also conducted on response time (in minutes) needed to complete reading a given passage and responding to the four multiple choice questions during the training phase. As seen in Figure 4, there were three significant effects in this analysis: the main effect of grade level, F(1, 22) = 64.019, MSE = 125.124, p < .0001, the main effect of story number, F(2, 44) = 3.537, MSE = 50.484, p = .0376, and the interaction of grade level and story number, F(2, 44) = 6.419, MSE = 70.919, p = .0036. Third grade stories took less time to complete (34.250 min) than fourth grade stories (49.167 min). This result can be directly attributed to the overall longer length of the fourth grade passages relative to the third grade passages. Also, the first passage of the third grade stories had a longer response time (39.167 min) than the subsequent third grade stories (second = 29.208 min, third = 34.375 min). However, there was less change across the three fourth grade stories (first = 48.333 min, second = 50.625 min, third = 48.542 min). This pattern might reflect the fact that participants were particularly slow at reading the very first story they encountered in training, which was always the initial third grade story. There were no other significant effects; in particular, there was not a main effect of condition, F(1, 22) < 1, an interaction of condition and grade level, F(1, 22) < 1, an interaction of condition and story number, F(2, 44)

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Improving Reading Skills for ESL Learners Using SoundSpel

223

< 1, or a three-way interaction of condition, grade level, and story number, F(2, 44) < 1. Thus, there were no differences at all between the control and SoundSpel conditions in reading time during training, despite the differences in the material read.

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Discussion With respect to both accuracy and speed, participants in the SoundSpel condition did not differ significantly in their performance from those in the control condition on the tests before and after training. These results do not support the hypothesis that SoundSpel participants would demonstrate greater improvement from pretest to posttest in their reading comprehension. Although SoundSpel did not enhance performance at the posttest, it also did not depress performance, even though the testing conditions did not match the training conditions. This observation is noteworthy because of former studies showing that test performance is depressed whenever testing and training conditions do not correspond, in accordance with both the principle of transfer appropriate processing (e.g., Morris, Bransford, and Franks, 1977; Roediger, Weldon, and Challis, 1989) and the principle of procedural reinstatement (e.g., Healy, Wohldmann, and Bourne, 2005). Hence, there were no facilitative or detrimental effects of training with SoundSpel. One reason why we might not have found a significant advantage of training with SoundSpel concerns the language background of the participants. It seems most likely that participants would benefit from SoundSpel to the extent that they can understand spoken English better than they can understand written English. For participants with a Far East Asian language background (viz., Korean and Japanese), reading English typically precedes listening to English, so it seems unlikely for them that comprehension of spoken English would be superior to that of written English. In contrast, for participants with other language backgrounds, it is more likely that they encountered spoken English before written English. Thus, to determine whether SoundSpel would have a training advantage for individuals who speak English better than they read it, we conducted a post hoc analysis on the training data in which we divided the participants into two groups: Far East Asian and other. This analysis yielded the expected pattern of results; the interaction between native language category and condition was significant by a one-tailed test, F(1, 20) = 3.297, MSE = 0.938, p = .0422. As shown in Figure 5, the participants with a Far East Asian native language showed better comprehension accuracy during training in the control condition than in the SoundSpel condition. In contrast, the other participants showed better comprehension accuracy during training using SoundSpel than without using SoundSpel. This result suggests that future studies investigating SoundSpel be restricted to individuals who learn to speak English before they learn to read it, such as children and ESL students with a language background other than Far East Asian. Past studies have demonstrated that the phonology of the first language exerts a very strong influence on the acquisition of second language vocabulary (Feldman and Healy, 1998). Problems then occur for ESL students whose native language has different phonological rules (i.e., different sound-to-spelling rules) than English. SoundSpel can help such ESL students better understand written English by simplifying the sound-to-spelling mapping rules.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

224

Michael D. Young, Michelle L. Wilson and Alice F. Healy

Mean Number Correct (out of 4)

4

Control SoundSpel

3

2

1

0

Far East

Other

Subject Group

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Figure 5. Mean number correct (out of 4 questions total per story) during training session for two conditions as a function of native language background. Error bars show standard errors of the mean.

Thus, the ESL students can use SoundSpel to identify those words that are in their spoken vocabulary but have an orthography with less familiar sound-to-spelling correspondences. However, if the particular words or syllables encountered during training with SoundSpel do not include those that appear on the test, the ESL students will not show any benefit from SoundSpel training. A subsequent study should, therefore, ensure that the words encountered in the tests are also included among the training passages. We found that story grade level had a significant impact on comprehension accuracy during training, with third grade stories more accurate overall than fourth grade stories. This finding, although not unexpected, is important for two reasons. First, it provides a manipulation verification because the third grade materials were assumed to be easier to read than the fourth grade materials. Second, it verifies that the material selected was appropriate for the reading abilities of the participants (i.e., the material was not so simple that comprehension was at the ceiling, as it was for the college undergraduates in the preliminary experiment; see Appendix B). There was also a significant impact of story grade level on response time. However, this finding is difficult to interpret because the fourth grade stories were longer (i.e., included more pages) than the third grade stories, and the fourth grade stories had a smaller number of pictures per page than the third grade stories. Also, story grade level was confounded with position in the sequence, with all third grade stories appearing before fourth grade stories. However, this confounding only works against the finding that performance was worse on fourth grade stories than on third grade stories because the additional practice should probably improve performance. We also found significant effects of story number for both comprehension and reading time during training. The effect of story number is difficult to interpret because the stories were not counterbalanced across sessions. Also, the effect of story number on comprehension

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Improving Reading Skills for ESL Learners Using SoundSpel

225

is difficult to understand because there was neither a consistent increase nor a consistent decrease in performance across stories; of the three stories performance was best overall on Story 1 and worst on Story 2. For reading time there was little difference for the fourth grade stories, but for the third grade stories participants were disproportionately slow in reading the first story. That finding could be due either to specific aspects of the story itself or to the fact that participants need some minimal practice with the unusual format in which lines are repeated before they can read the stories efficiently. A suggestion for improving performance on SoundSpel would be to make the SoundSpel instruction more congruent with its actual use. In the present experiment, SoundSpel instruction did not include DoubleLine although DoubleLine was used during the training session. Practice with DoubleLine during SoundSpel instruction may lead to better performance during training for the SoundSpel group. Although we used SoundSpel in this experiment to promote the skill of silent reading, it may prove to be a more effective tool for promoting the skill of reading aloud. The advantage of SoundSpel is that by simplifying the spelling-to-sound mapping it allows readers to go from the orthographic representation of a word to its phonological (i.e., spoken) representation. Thus, reading aloud, which requires the phonological representation, should be facilitated by the use of SoundSpel, whereas reading silently might not require access to the phonological representation, so should not show as large a benefit from SoundSpel. Therefore a future study might compare the SoundSpel and control conditions in a situation where both training and testing require the participants to read passages aloud, followed by multiple-choice comprehension quizzes, as in the present experiment. In such a study, any advantages of SoundSpel should be magnified.

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Author Notes This research was supported in part by a grant from the American Literacy Council, contract DASW01-03-K-0002 from the Army Research Institute, and grant W911NF-05-10153 from the Army Research Office, all to the University of Colorado (Alice Healy, Principal Investigator). Michelle Wilson’s work on this project was also supported in part by a summer undergraduate research fellowship (SURF) award from the University of Colorado. We are indebted to Kathleen Shea for her help in constructing the experimental materials and in developing the multiple-choice comprehension questions. Correspondence concerning this report should be sent to Alice F. Healy, Department of Psychology and Neuroscience, 345 UCB, University of Colorado, Boulder, CO 80309-0345.

Appendix A. Excerpt of SoundSpel Version of a Passage Used in the Experiment A Raccoon’s Life Julia Cunningham’s story about a talking raccoon is a A Raccoon’s Lief Julia Cunningham’s story about a tauking raccoon is a fantasy, a make-believe story. Even though the talking raccoon in Macaroon fantasy, a maek-beleev story. Eeven tho th tauking raccoon in Macaroon is not true to life, the environment the author describes could be real. In the

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

226

Michael D. Young, Michelle L. Wilson and Alice F. Healy is not troo to life, th envieronment th author descriebs cuud be reel. In th spring and summer a leafy, green forest is a comfortable place for raccoons. spring and sumer a leefy, green forest is a cumfortabl plaes for raccoons.

Appendix B. Preliminary Experiment

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Prior to testing ESL students, 48 native English-speaking University of Colorado undergraduates participated in a preliminary experiment to ensure the adequacy of the materials. These students, who were tested individually, received credit in an introductory psychology course for their participation, which lasted approximately 1 hour. Eighteen passages were selected from third, fourth, and fifth grade readers (Allington, Cortez, Cunningham, Sebesta, and Tierney, 1989a, 1989b, 1989c), with six passages from each grade level. Each passage was accompanied by four three-alternative multiple-choice comprehension questions written by the experimenters. Each student was given three passages to read, one from each grade level, and each passage was read by 8 students. The texts were written in the same format as the pretest and posttest; that is, they did not contain DoubleLine but were written double-spaced with no repeated lines or SoundSpel text. Immediately after reading a given passage, students answered the multiple-choice questions for that passage, which occurred on a separate page given to them by the experimenter. The overall accuracy was very high for all of the passages; the mean proportion correct was over .95. Nevertheless, the passages used in the experiment were selected from the third and fourth grade passages using the criterion that all selected passages had the highest proportion of correct responses by the 8 students who read them.

Appendix C. Passage Read Aloud by Participants Learning SoundSpel System It was about th midl of a laet spring morning when th Hors caem inside frum th far sied of th bair rij and stuud for a whiel on th hieest part. He apeerd to be foer or five years oeld, compactly bilt and with a ruf coet sumwherr between bloo-grae and mous culor. He wor no owner’s brand, and th oenly distinktiv mark on him was th straengj mask-like pach of darker culor covering his foerhed and upper muzl. Th untiedy tangles and mats of cockleburs and mud in his long tael and maen markt him as a raenj hors, and not wun that had simply straed off frum sum ranch or farm.

References Allington, R. L., Cortez, J., Cunningham, P. M., Sebesta, S. L., and Tierney, R. J. (1989a). Collections: Between times. Glenview, IL: Scott Foresman and Company. Allington, R. L., Cortez, J., Cunningham, P. M., Sebesta, S. L., and Tierney, R. J. (1989b). Collections: In good company. Glenview, IL: Scott Foresman and Company.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Improving Reading Skills for ESL Learners Using SoundSpel

227

Allington, R. L., Cortez, J., Cunningham, P. M., Sebesta, S. L., and Tierney, R. J. (1989c). Collections: Sometimes I wonder. Glenview, IL: Scott Foresman and Company. Bryant, P. E., MacLean, M., Bradley, L. L., and Crossland, J. (1990). Rhyme and alliteration, phoneme detection, and learning to read. Developmental Psychology, 26, 429-438. Feldman, A. and Healy, A. F. (1998). Effect of first language phonological configuration on lexical acquisition in a second language. In A. F. Healy and L. E. Bourne, Jr. (Eds.), Foreign language learning: Psycholinguistic studies on training and retention (pp. 5576). Mahwah, NJ: Erlbaum. Healy, A. F., Wohldmann, E. L., and Bourne, L. E., Jr. (2005). The procedural reinstatement principle: Studies on training, retention, and transfer. In A. F. Healy (Ed.), Experimental cognitive psychology and its applications (pp. 59-71). Washington, DC: American Psychological Association. Mann, V. A. (1986). Phonological awareness: The role of reading experience. Cognition, 24, 65-92. Mole, A. (Ed.) (2003). Stories in reformed spelling. Boulder, CO: American Literacy Council. Morris, C. D., Bransford, J. D., and Franks, J. J. (1977). Levels of processing versus transfer appropriate processing. Journal of Verbal Learning and Verbal Behavior, 16, 519-533. Olson, R. K., Wise, B., Ring, J., and Johnson, M. (1997). Computer-based remedial training in phoneme awareness and phonological decoding: Effects on the posttraining development of word recognition. Scientific Studies of Reading, 1, 235-253. Patel, T. K., Snowling, M. J., and de Jong, P. F. (2004). A cross-linguistic comparison of children learning to read in English and Dutch. Journal of Educational Psychology, 96, 785-797. Rayner, K., Foorman, B. R., Perfetti, C. A., Pesetsky, D., and Seidenberg, M. S. (2001). How psychological science informs the teaching of reading. Psychological Science in the Public Interest, 2, 31-74. Roediger, H. L., III, Weldon, M. S., and Challis, B. H. (1989). Explaining dissociations between implicit and explicit measures of retention: A processing account. In H. L. Roediger, III, and F. I. M. Craik (Eds.), Varieties of memory and consciousness: Essays in honour of Endel Tulving (pp. 3-41). Hillsdale, NJ: Erlbaum.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved. Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

In: Bilinguals: Cognition, Education and Language Processing ISBN: 978-1-60741-710-1 Editor: Earl F. Caldwell, pp. 229-242 © 2010 Nova Science Publishers, Inc.

Chapter 12

A NOVEL TRANSLITERATION APPROACH IN AN ENGLISH-ARABIC CROSS LANGUAGE INFORMATION RETRIEVAL SYSTEM Ghita Amor-Tijani∗ and Abdelghani Bellaachia Department of Computer Science, The George Washington University Washington DC, USA

Abstract

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

One of the main issues facing Cross Language Information Retrieval (CLIR) is untranslatable words, i.e., words not found in dictionaries, which are usually referred to as Out Of Vocabulary (OOV) words. Bilingual dictionaries in general do not cover most proper nouns (e.g., names of places, people, countries, etc.), which constitute a large proportion of OOV words. As they are often primary keys in a query, their correct translation is often necessary to maintain a good retrieval performance. Because they are spelling variants of each other in most languages, an approximate string matching technique against the target database index is usually used to find the target language correspondents of the original query key. The n-gram technique has proven to be the most effective among other approximate string matching techniques. A more complicated issue arises when the languages dealt with have different alphabets. The approach usually taken is transliteration. It is applied based on phonetic similarities between the languages involved. However, transliteration by itself cannot guarantee the exact spelling of the transliterated words as found in the document collection. There are a variety of ways that a transliterated word can be spelled despite conventions that might exist. The fact that there is no one correct way of spelling a transliterated word shows the need for a technique that is capable of generating the different spellings found in the document collection. In this study, we chose to combine both transliteration and the n-gram technique in an English-Arabic CLIR system, in which Arabic documents were searched using English queries. We evaluated the effectiveness of this approach and compared it with other transliteration approaches. Experimental results showed the retrieval improvement gained using our transliteration approach over other existing approaches.



E-mail address: [email protected], [email protected]

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

230

Ghita Amor-Tijani and Abdelghani Bellaachia

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Introduction Issues related to ambiguity and Out Of Vocabulary (OOV) words in Cross Language Information Retrieval (CLIR) have been addressed by several studies. Using a dictionarybased approach in information retrieval, queries are translated by replacing source language query keys with their target language equivalents provided by online dictionaries. This approach, however, presents several problems: previous studies have demonstrated that automatic Word by Word (WBW) translation of queries using Machine Readable Dictionaries (MRD) results in a 40-60% drop in effectiveness compared to monolingual retrieval (Hull and Grefenstette, 1996). The main factors that cause this loss in effectiveness are 1) OOV query keys, most of which are not covered by general dictionaries; 2) the processing of inflected words; 3) phrase identification and translation; and 4) lexical ambiguity in source and target languages. The problems associated with OOV words include compound words and cross lingual spelling variants, specifically proper nouns and technical terms (Pirkola, Hedlund, Keskuslato, and Järvelin, 2000; Pirkola, Hedlund, Keskuslato, and Järvelin, 2001). Around 50% of OOV words were observed to be proper nouns (Davis and Ogden, 1998). Studies presented at the 2002 Text REtrieval Conference (TREC2002) on English-Arabic cross language track noted that performance drops by more than 50% when proper nouns are not covered by the dictionaries used (Larkey, AbdulJaleel, and Connell, 2003). Their correct translation is therefore crucial to the overall retrieval performance of a CLIR system. General dictionaries only include the most commonly used proper nouns and technical terms. Those proper nouns and technical terms not found in dictionaries are subsequently considered untranslatable by the CLIR system processing them. As they are usually primary keys in the query, their correct translation might be necessary to maintain a good retrieval effectiveness. A common method for handling OOV words is to carry them over unchanged to the target query. This can, however, lead to deterioration in retrieval effectiveness. As OOV words have spelling variants in most languages, an approximate string matching technique against the target database index could be used to find equivalents of the original query key in the target language. Transliteration could also be determined based on the phonetic similarities of the languages in question. The n-gram technique is a language-independent technique which has been reported to be even more effective than other string matching techniques such as Soundex or Phonix, which are based on phonetic similarity (Zobel and Dart, 1995). In n-gram matching, query keys and terms in the index of the document set are split into sub-strings of length n. The n-gram sets of the query key and the document's indexed terms are compared, and the best matching words are used as the key’s correspondents. The number of matching words retrieved is based on the threshold used. The n-gram technique is effective because proper nouns as well as technical terms are usually spelling variants of each other in such languages (Pirkola, Keskustalo, Leppänen, Känsälä, and Järvelin, 2002). Sometimes, an approximate string matching technique misses the correct target word as the best match; a transformation rule based translation (TRT) was applied to generate intermediate forms (Pirkola, Toivonen, Keskustalo, Visala, and Järvelin, 2003). However, the TRT approach may be useless if it just generates a set of translation equivalent candidates but is not able to identify the one correct equivalent for a source word. Using regular frequency patterns of generated word forms for a source word was the alternative method to recognize the correct target word (Pirkola,

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

A Novel Transliteration Approach in an English-Arabic Cross Language…

231

Toivonen, Keskustalo, and Järvelin, 2006). Those were obtained using the TRT (Pirkola et al., 2003). This approach was demonstrated to effectively identify the target language equivalent of an OOV word. As we are dealing with a CLIR system involving English and Arabic, which are two languages with different alphabets, a string matching technique cannot be applied. Different versions of transliteration were used as the alternative methods to render the query term orthography into that of the target language (Darwish, Doermann, Jones, Oard, and Rautiainen, 2002; Larkey, Connell, and AbdulJaleel, 2003). Experiments showed that generating multiple spelling variants for Arabic transliterated words improves effectiveness (Larkey, AbdulJaleel, and Connell, 2003). One of the transliteration models that was shown to improve retrieval effectiveness is the n-gram based statistical transliteration technique studied by the University of Massachusetts (UMass) (AbdulJaleel and Larkey, 2003). This model generates multiple Arabic transliterations based on a mapping derived from a character-level alignment of a list of Arabic/English word pairs. There are other standard transliterations like the library of congress scheme (Library of Congress, 1998); which mainly follows the transliterations used by the best known Arabic-English dictionary, the Hans Wehr dictionary of Modern Written Arabic edited by J.M. Cowan. Our approach was to extract the spelling variants of the transliterated words from the document collection, rather than to generate them automatically. An approximate string matching technique seemed to be the best technique to be used for that purpose. In our transliteration approach, which we refer to as Transliteration N-Gram (TNG), we combine transliteration with an approximate string matching technique. We first generate one transliteration to have the OOV word in the same alphabet as our document collection. Then, we use the n-gram as it was shown to be the most effective among other approximate string matching techniques, such as Soundex and Phonix (Zobel and Dart, 1995). Unlike fuzzy translation (Pirkola et al., 2003), our technique focuses on two languages of different orthographies and the two steps of data transformation are used for different purposes. This paper is organized as follows. In Section 2, our system is illustrated and the TNG technique, which is our transliteration approach, is defined. Section 3 shows the experimental results and the effectiveness of our transliteration approach on the TREC 2002 (English-Arabic) cross language track. A comparison of our work with UMass statistical transliteration and the Arabic transliteration model used by the library of congress is also presented. Finally, Section 4 concludes the paper.

TNG Framework In this section, we describe our approach of transliteration and the main components of our system. We first present the architectural model that illustrates the steps of query processing and then give a detailed description of those steps.

Architectural Model The flow chart illustrated in Figure 1 shows the steps taken to process the query file. We used the English query file from TREC2002. The English query file used consisted of fifty queries.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

232

Ghita Amor-Tijani and Abdelghani Bellaachia

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Figure 1. TNG Framework.

Each query is processed individually by first running it through a tokenizer to extract the different tokens that are checked against an English stop word list. After stop words are removed from the query, all terms are stemmed using the Porter stemmer (Porter, 1980). Using an English-to-Arabic dictionary, stemmed terms are translated. In case there is an entry for each term in the query, the Arabic query is formed and fed to the search engine. Otherwise, terms without an entry in the dictionary are transliterated using a modified version of UMass statistical transliteration (AbdulJaleel and Larkey, 2003). Either one transliteration is generated for each OOV query term to form the final query, or spelling variants of that transliteration are extracted from the list of terms in the index of our document collection using the n-gram technique. The main difference using TNG is the new approach to generate multiple transliterations in a CLIR system involving two languages of different alphabets; more specifically, the Arabic language. TNG is a combination of transliteration and n-gram string matching technique. Once the OOV words are processed using either transliteration or TNG, the final Arabic query is formed and fed to the search engine. Examples of the data transformation that occurs in the transliteration and the ngram processing steps of OOV words are given later in Table 1. Table 1. Examples of Spelling Variants Extracted Using Ngram0 OOV word Clinton

Transliteration ‫( نﻮﺗﻧﻳﻟﻜ‬Clinton)

Spelling variants extracted using Ngram0 ‫( نﻮﺗﻧﻟﻜ‬Clnton) ،‫(نﻮﺗﻨﻳﻳﻟﻜ‬Cliinton) ،‫(نﻮﺗﻨﻟﻜ‬Clnton) ‫(نﺗﻧﻳﻟﻜ‬Clintn) ،‫( نﻮﺗﻳﻧﻴﻟﻜ‬Cliniton) ،‫( نﻮﺗﻴﻨﻟﻜ‬Clniton)

Query Processing Using the TNG technique, both transliteration and an n-gram mapping are applied to get a set of possible transliterations. One transliteration is first generated and the n-gram string matching technique is then applied on the transliterated term and the list of terms in the index

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

A Novel Transliteration Approach in an English-Arabic Cross Language…

233

of the document collection. The transliteration model followed is similar to the one used by UMass, known as the statistical transliteration. Their transliteration model is based on a character-level alignment performed on a list of Arabic/English word pairs. Segments of English characters that mapped to a single Arabic letter were extracted and given a probability based on how many Arabic segments an English segment corresponded to and how often the mapping occurred. Instead of generating multiple transliterations, we only produce one transliteration using the equivalent Arabic letters with the highest probability. As for the n-gram technique, it is used on the transliterated term and the list of stemmed words in the index of the document collection. Five different approaches were used and compared. The basic idea of the n-gram technique is to consider the set of pairs of adjacent characters of words to be compared and calculate the degree of similarity between them. Similarity is based on the number of similar n-grams and the total number of unique n-grams. Digrams (sets of two characters) were used as they proved to give better performance than trigrams (sets of three characters) for this specific data collection. Five types of character combinations were considered in the calculation of the similarity value: −

Ngram0: only adjacent characters (combination of two letters) are considered. Word abcd



Ngram1: adjacent and non-adjacent characters separated by one character are considered.

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Word abcd





Digram set for ngram1 {ab,ac,bc,bd,cd}

Ngram2: adjacent and non-adjacent characters separated by one character and by two characters are considered. Word abcd



Digram set for ngram0 {ab,bc,cd}

Digram set for ngram2 {ab,ac,ad,bc,bd,cd}

Ngram0-1: adjacent (first set) and non-adjacent (second set) characters separated by one character are considered. Only digrams belonging to the same category set are compared. Word Digram sets for ngram0-1 abcd {ab,bc,cd}{ac,bd} Ngram0-2: adjacent (first set) and non-adjacent (second set) characters separated by one character and by two characters are considered. Only digrams belonging to the same category set are compared. Word abcd

Digram sets for ngram0-2 {ab,bc,cd}{ac,ad,bd}

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

234

Ghita Amor-Tijani and Abdelghani Bellaachia

The five different approaches of the n-gram technique were indeed used to extract the different spellings of OOV words using a transliterated form, which can be one of the different ways the word is spelled. Examples of spelling variants of transliterated words extracted using ngram0 are illustrated in Table 1. Those approaches were compared and experimental results showed that using the TNG technique that combines transliteration with the ngram0 function showed the best retrieval performance (Bellaachia and Amor-Tijani, 2008). This is the TNG approach that will be used in the next comparisons.

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Experimental Results The Indri search engine from the Lemur1 project was used as a tool for information retrieval. It is a cooperative effort between the University of Massachusetts and Carnegie Mellon University. The document set used is known as the "Arabic Newswire Part 1". It was obtained from LDC. It contains approximately 383,872 news articles taken from the “Agence France Presse” (AFP) Arabic newswire dated from May 13, 1994 through December 20, 2000. The collection is about 896MB. The documents are represented in Unicode and encoded in utf-8. The category of the Arabic language used is the Modern Standard Arabic (MSA). This collection consists of 76 Million tokens; among them are 666,094 unique words. This number is reduced to 240,823 after the collection is stemmed and stop-words are removed. This corpus was indexed using the application “buildindex” available in Indri. TREC 2002 topics2 in both English and Arabic were used to search the Arabic document set. Relevance judgments3 corresponding to the topics and the document set were used to evaluate the effectiveness of our retrieval system since they give information about which documents are relevant to what query. Fifty queries were used in our experimental approach. Only the topic queries which are typical user’s queries were considered in order to evaluate our techniques. Below is an example of an English query:

Number:AR26 Kurdistan Independence Description: How does the National Council of Resistance relate to the potential independence of Kurdistan? Narrative: Articles reporting activities of the National Council of Resistance are considered on topic. Articles discussing Ocalan's leadership within the context of the Kurdish efforts toward independence are also considered on topic.

The English topics were processed before they were fed to the search engine. The TREC2002 query file consists of a title, description and a narration field for each of the 50 queries. Only the topic field, which corresponds to short queries, was considered in our study as it represents typical user queries. The queries were first tokenized and stop-words were 1

http://www.lemurproject.org/indri/ http://trec.nist.gov/data/topics_noneng/index.html 3 http://trec.nist.gov/data/qrels_noneng/index html 2

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

A Novel Transliteration Approach in an English-Arabic Cross Language…

235

removed. These are common words with no content-specific information. The resulting query terms were then stemmed by the Porter stemmer. The stemmed query terms were subsequently translated using a corpus generated from a set of parallel UN documents provided by the LDC. It provides a varying set of translation probabilities for the plausible translations of each entry term. Its English entries are processed with the Porter stemmer and the Arabic translations are stemmed using Al-Stem provided by Kareem Darwish from the University of Maryland. For each query term, the translation evaluated at the highest probability was considered to be the best translation. The final translated query was formed using these translations. Previous research showed that considering translations with probability higher than 0.15 gave the best performance for this dictionary, referred to as UNdict (Larkey and Connell, 2003). We decided to use only the best translation as we are not testing the efficiency of this dictionary, but rather other issues more specifically related to word-by-word translation. In case no entry was found in the dictionary, the query term was passed unchanged; unless further processing was considered. Once the translated query was ready, the application “runquery” from Indri was used to search the Indri repositories built from the TREC documents. The indexed documents were searched and a list of ranked documents was retrieved reflecting relevant documents to different queries. The following results are for 50 topic queries. All proper nouns used in topic queries were detected manually and their translation was disconnected from the dictionary for two reasons. The effectiveness of the TNG technique can be better reflected in the results. Also, typical dictionaries usually do not include translations for proper nouns. Thirty one queries included untranslatable words, among which 29 included about 25 proper nouns (e.g. Sadeq AlSadr, Lebanon) and 9 adjectives (e.g. Iranian). Variants of the OOV term were chosen based on a Similarity Value (SIM) calculated using Equation 1. A SIM threshold was chosen after comparing the performance of different runs. All terms with a SIM higher than or equal to 0.4 were considered in the translated query. Those terms were combined with the synonym operator “#wsyn” from Indri using the SIM value as the weight of the term. Equation 1: Similarity Value SIM (N1, N2)= | N1∩N2 | / | N1UN2 | where, N1 and N2 are digram sets of two words. | N1∩N2 |: number of similar digrams | N1UN2 |: number of unique digrams Precision and recall are two metrics used to evaluate retrieval performance in an IR system: Precision is the measure of the system’s capacity to retrieve relevant documents at the top ranking. It is the ratio of relevant documents retrieved over the total number of all documents retrieved. A high precision is achieved when most of the retrieved documents are relevant. Equation 2: Precision

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

236

Ghita Amor-Tijani and Abdelghani Bellaachia

Recall is the measure of the system’s capacity to retrieve all relevant documents. It is the ratio of relevant documents that are retrieved over the total number of all relevant documents available. A perfect recall ratio of 1.0 is achieved when the system retrieves all relevant documents.

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Equation 3: Recall

To compute both precision and recall, the documents obtained in the search results using the IR system are compared against the relevance judgment file which specifies the relevant documents for each query. The latter is usually predefined for a specific set of queries and document set. A group of people assess the collection according to specific criteria and determine which documents are relevant to which queries and which ones are not. This data is stored in the relevance judgment file. When the queries are run using the IR system, the variables needed to compute both precision and recall are identified by comparing the results obtained using Indri with the ones in the relevance judgment file. The relevant documents are those in the relevance judgment file, and the retrieved documents are those obtained with the IR system. Mean average precision (MAP) is used to measure how relevant the retrieved documents are. Precision at any recall level is calculated as the average precision of all queries at that recall value. A standard tool used by the TREC community known as “trec_eval” is used in order to get the system’s performance data. It takes both the results file from the application “runquery” from Indri and the standard set of judged results from TREC and utilizes them as parameters to generate tables reflecting performance as a measure of precision and recall. For the evaluation of each run, the number of retrieved documents, the number of relevant documents, and the number of retrieved documents that are relevant is generated. A table of precision averages at different recall values (0.0, 0.1,…,1.0) is generated for each run. The average precision values given are the percent of retrieved documents that are relevant after a certain percentage of all relevant documents is retrieved. In other words, precision at recall 0.1 is precision after 10% of relevant documents are retrieved. Values are averaged over all queries to determine the MAP of the run. The statistical significance of performance improvement is evaluated using the Wilcoxon signed ranks test with the p-level set at 0.5. Different runs were carried out and compared for all 50 queries to compare the performance of the different transliteration approaches: −

Mono: In the monolingual run, Arabic queries were used to search the document set. Al-Stem stemmer was used to stem both the Arabic queries and the document set. Stemming was included in the monolingual run as well, because our interests do not lie in evaluating the stemmer.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

A Novel Transliteration Approach in an English-Arabic Cross Language… −

− − −

237

Transliterated: English queries were used to search the document set. First, stopwords were removed. The queries were then stemmed with the Porter stemmer. Translations for query terms were looked up in the UNdict. OOV words were then transliterated generating one transliteration corresponding to the one with the highest probability using UMass transliteration model. Congress: English queries were translated and OOV words were transliterated using the congress transliteration model. UMass: English queries were translated and OOV words were transliterated using UMass statistical transliteration model. TNG: English queries were translated and OOV words in translated English queries were first transliterated, then the ngram0 function was applied on the transliterated term and the indexed stemmed terms. Different possible spelling variations of the transliterated word were derived using the SIM value as the lower bound of possible correct spellings. Those were grouped a weighted synonym “#wsyn” operator in the final query.

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Retrieval Performance A comparison was made between our TNG approach, the Library of Congress transliteration scheme (Library of Congress, 1998), and the statistical transliteration used by the University of Massachusetts (UMass) (AbdulJaleel and Larkey, 2003). The difference between the three transliteration approaches is explained earlier in Section 1 and Section 2.1. The main difference between the models is that the Library of Congress scheme is based on a one-to-one mapping generating one transliteration for each word, whereas the other two models generate multiple transliterations based on two different approaches. As for the TNG technique, the ngram0 function was used in the comparison as it is explained earlier in Section 2.2. Table 2 and Table 3 illustrate the effectiveness of all three approaches. Table 2. Interpolated Average Precision, Averaged over 50 Queries Recall 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Mono 0.5968 0.3845 0.3192 0.2831 0.2494 0.2306 0.1839 0.1567 0.1298 0.0651 0.0134

Translated 0.273 0.1621 0.1391 0.1276 0.1028 0.0965 0.0897 0.0787 0.065 0.0532 0.0028

TNG 0.4379 0.2793 0.243 0.2055 0.1744 0.1509 0.1264 0.1039 0.0722 0.0555 0.0054

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

238

Ghita Amor-Tijani and Abdelghani Bellaachia Table 3. Interpolated Average Precision, Averaged over 50 Queries Transliterated 0.3211 0.203 0.1802 0.1601 0.135 0.1228 0.1112 0.0951 0.0755 0.0535 0.0074

Recall 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

UMass 0.3544 0.2221 0.1971 0.1721 0.1458 0.1256 0.1128 0.0978 0.0794 0.0567 0.0074

Congress 0.323 0.2016 0.1801 0.1524 0.1228 0.1031 0.0914 0.0792 0.0646 0.0488 0.0032

TNG 0.4379 0.2793 0.243 0.2055 0.1744 0.1509 0.1264 0.1039 0.0722 0.0555 0.0054

It displays the 11-point interpolated average precision values for both runs, averaged over the 50 TREC 2002 queries used. The MAP values were also obtained and compared. Precision of the monolingual run is first calculated by using the Arabic translated TREC queries. Precision using transliterated queries is then calculated using both the UMass transliteration with the highest probability and the congress transliteration model. Finally, precision using the TNG technique is calculated and compared to precision using transliteration only. Table 2 and Figure 2 show the benefit of handling untranslatable words, which are mostly proper nouns in this case. Using the TNG technique significantly improves average precision.

Mono Translated TNG

0.6 0.5 Precision

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

0.7

0.4 0.3 0.2 0.1 0 1

2

3

4

5

6 Recall

7

8

9

10

Figure 2. Precision-Recall Graph of the Mono, Translated, and TNG Runs.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

11

A Novel Transliteration Approach in an English-Arabic Cross Language…

239

0.5 Transliterated Congress UMass TNG

0.45 0.4 Precision

0.35 0.3 0.25 0.2 0.15 0.1 0.05 0 0

0.1

0.2

0.3

0.4

0.5 0.6 Recall

0.7

0.8

0.9

1

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Figure 3. Precision-Recall Graph of the Transliterated, Congress, UMass, and TNG Runs.

As explained in Section 2, adjacent characters of the transliterated word are compared to the set of stemmed words in the index of the document collection. Those words with similarity value higher or equal to 0.4 are considered in the translated query. The improvement gained using the TNG technique on OOV words shows the importance of correctly translating query terms, specifically proper nouns that are usually primary keys in a query. The significant improvement also emphasizes the need of having multiple spellings of those words in the translated query. Table 3 and Figure 3 show the difference in performance between generating one and multiple transliterations for proper nouns. We notice that using TNG, which combines transliteration and the n-gram technique, gives a better performance than the statistical transliteration used by UMass.The difference in performance can be explained by the fact that transliteration by itself, although it can generate different spellings of a word, cannot cover all the variations of a word spelling used in the document collection. The n-gram technique compensates this shortcoming by allowing a wider range of existing spelling variations to be used in the search query. Both those approaches give a better performance than the transliterated run, the modified version of the statistical transliteration approach, and the library of congress transliteration model. These observations emphasize the importance of generating multiple transliterations to maintain a good retrieval performance.

Summary of Results Table 4 and Table 5 summarize the results obtained at different runs. These results were obtained when processing 50 topic queries. The percent improvement in effectiveness is given as a percentage of transliteration alone in which one transliteration is produced for each OOV word (%Transliterated). The different approaches used are also compared to the

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

240

Ghita Amor-Tijani and Abdelghani Bellaachia

monolingual run (%Mono). This comparison represents how effective each technique is, as the monolingual run reflects a CLIR system without the effect of translation ambiguity. Table 4 shows the improvement in effectiveness of the TNG and UMass techniques over the transliterated run. A significant improvement of 28% and 7% was obtained respectively. The difference in performance between the transliteration scheme used in the transliterated run and by the Library of Congress is insignificant. Both approaches generate one possible transliteration, which limits the retrieval performance of the system. Table 5 depicts the performance of all runs compared to the monolingual run. The Library of Congress scheme yields about the same performance as the transliterated run. Both approaches generate one transliteration for each OOV word. The transliteration generated might not necessarily be used in the document collection. The statistical transliteration used by UMass gives a better performance than the transliterated run as multiple transliterations as generated. An effectiveness of 72% of that of the monolingual run is achieved when TNG is applied. Table 4. Percent Improvement of the Congress, UMass, and TNG Runs over the Transliterated Run Transliterated Congress UMass TNG

MAP 0.1209 0.1123 0.1294 0.1553

%Transliterated ---7% 7% 28%

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Table 5. Percent Effectiveness of the Translated, Transliterated, Congress, UMass, and TNG Runs Compared to the Monolingual Run Mono Translated Transliterated Congress UMass TNG

MAP 0.2168 0.0989 0.1209 0.1123 0.1294 0.1553

%Mono --46% 56% 52% 60% 72%

Table 6. Percent Improvement of the TNG and UMass Techniques over The Library of Congress Transliteration Scheme Congress UMass TNG0

MAP 0.1123 0.1294 0.1553

%UMass --15% 38%

This improvement is significant as effectiveness is only 46% of that of the monolingual run when OOV words are not processed. These results demonstrate the significant improvement in effectiveness gained in our CLIR system when OOV words are processed,

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

A Novel Transliteration Approach in an English-Arabic Cross Language…

241

and specifically when variations of transliterations are extracted from the document set rather than generated automatically. The results in Table 6 show that TNG and UMass statistical transliteration give a significant improvement in MAP of 38% and 15%, respectively, over the Library of Congress transliteration model. This is because using the Library of Congress transliteration scheme only generates one transliteration, while it has been shown in this study as well as previous studies that generating multiple transliteration spellings is necessary for a good retrieval performance.

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Conclusion In this study we have compared retrieval performance using three transliteration schemes: the Library of Congress, UMass, and TNG. Both the UMass and TNG techniques gave better results than the standard transliteration model used by the Library of Congress. The difference in retrieval effectiveness is due to the fact that the Library of Congress scheme uses a one-toone mapping to produce the corresponding transliteration, thus generating one spelling possibility. It has been shown in previous studies as well as in this study that generating multiple transliteration spellings enhances retrieval performance. Indeed, the TNG technique and UMass statistical transliteration model gave good results. They yield a retrieval performance better than the Library of Congress model by 15% and 38%, respectively. Experimental results also show that the TNG based approach significantly improves retrieval effectiveness and gives a MAP that is 20% better than the UMass statistical transliteration, which also produces different transliterations. The improvement in effectiveness is due to the use of different spelling variants of transliterated terms used in the existing document collection, in the translated Arabic query. We can conclude that generating multiple transliterations indeed enhances performance and correctly choosing the possible spelling variants is important to further improve it.

References AbdulJaleel, N., and Larkey, L.S. (2003). Statistical transliteration for English-Arabic cross language information retrieval. Proceedings of the twelfth international conference on information and knowledge management (CIKM 2003) (pp. 139-146). New Orleans, LA: ACM Press. Bellaachia, A. and Amor-Tijani, G. (2008). Proper nouns in English-Arabic cross language information retrieval. Journal of the American Society for Information Science and Technology, 59(12), 1925-1932. Darwish, K., Doermann, D., Jones, R., Oard, D., and Rautiainen, M. (2002). TREC-10 experiments at University of Maryland CLIR and video. Proceedings of Text Retrieval Conference TREC10 (TREC2001) (pp. 549-562). Gaithersburg, MD. Davis, M.W. and Ogden, W.C. (1998). Free resources and advanced alignment for crosslanguage text retrieval. Proceedings of the sixth text retrieval conference (TREC-6) (pp. 385-394). Gaithersburg: NIST Special Publication.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

242

Ghita Amor-Tijani and Abdelghani Bellaachia

Diab, M. (2004). An unsupervised approach for bootstrapping Arabic word sense tagging. Proceedings of the Workshop on Arabic Based Script Languages at the International Conference on Computational Linguistics (COLING 2004) (pp. 43-50). Stroudsburg, PA: ACL. Diab, M., Hacioglu, K., and Jurafsky, D. (2004). Automatic tagging of Arabic text: From raw text to base phrase chunks. Proceedings of Human Language Technology and the North American Chapter of the Association for Computational Linguistics (HLT-NAACL 2004) (pp.149-152). Stroudsburg, PA: ACL. Hull, D.A., and Grefenstette, G. (1996). Querying across languages: a dictionary-based approach to multilingual information retrieval. Proceedings of the 19th Annual International ACM SIGIR Conference on Research and Development in Information Retrieval (pp. 49-57). New York City: ACM Press. Larkey, L.S., AbdulJaleel, N. and Connell, M. (2003). What's in a name?: Proper names in Arabic cross language information retrieval. CIIR Technical Report, IR-278. [Available at http://ciir.cs.umass.edu/pubfiles/ir-278.pdf]. Larkey, L.S., and Connell, M. (2003). Structured Queries, Language Modeling, and Relevance Modeling in Cross-Language Information Retrieval. Information Processing and Management Special Issue on Cross Language Information Retrieval. 41, 457-473. Larkey, L.S., Connell, M., and AbdulJaleel, N. (2003). Hindi CLIR in thirty days. ACM Transactions on Asian Language Information Processing (TALIP), 2(2), 130-142. New York City: ACM Press. Library of Congress. (1998). ALA-LC Romanization tables: transliteration schemes for non_roman scripts. [Available at http://www.loc.gov/catdir/cpso/roman.html]. Pirkola A, Hedlund T, Keskuslato H, and Järvelin K. (2000). Cross-Lingual Information Retrieval Problems: Methods and findings for three language pairs. ProLISSa Progress in Library and Information Science in Southern Africa. First biannual DISSAnet Conference. Pretoria, 26-27. Pirkola, A., Hedlund, T., Keskustalo, H., and Järvelin, K. (2001). Dictionary-based cross language information retrieval: Problems, methods, and research findings. Information Retrieval, 4(3/4), 209-230. Hingham, MA: KAP. Pirkola, A., Keskustalo, H., Leppänen, E., Känsälä, A.P., and Järvelin, K. (2002). Targeted sgram matching: A novel n-gram matching technique for cross- and monolingual word form variants. Information Research, 7(2) [Available at http://InformationR.net/ir/72/paper126.html]. Pirkola, A., Toivonen, J., Keskustalo, H., and Järvelin, K. (2006). FITE-TRT: A high quality translation technique for OOV words. Proceedings of the 21st Annual ACM Symposium on Applied Computing (SAC 2006) (pp. 1043-1049). New York City: ACM Press. Pirkola, A., Toivonen, J., Keskustalo, H., Visala, K., and Järvelin, K. (2003). Fuzzy translation of cross-lingual spelling variants. Proceedings of the 26th annual international ACM SIGIR conference on Research and development in information retrieval (pp. 345352). New York City: ACM Press. Porter, M.F. (1980). An algorithm for suffix stripping, Program 14(3), 130-137. Zobel, J., and Dart, P. (1995). Finding approximate matches in large lexicons. SoftwarePractice and Experience, 25(3), 331-345.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

In: Bilinguals: Cognition, Education and Language Processing ISBN 978-1-60741-710-1 c 2010 Nova Science Publishers, Inc. Editor: Earl F. Caldwell, pp. 243-286

Chapter 13

M ETHODS FOR C ROSS -L ANGUAGE I NFORMATION R ETRIEVAL Kazuaki Kishida School of Library Information Science, Keio University, Japan

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Abstract This article reviews technical methods for enhancing effectiveness of cross-language information retrieval (CLIR), in which target documents are written in different languages from that used for representing a search request. As the Internet has spread since the 1990s, the importance of CLIR has grown, and the research community of information retrieval has been tackling various CLIR problems. The purpose of this article is to overview exhaustively CLIR techniques developed in the research efforts. The following research issues on CLIR are covered: (1) strategies for matching the query and documents written in different languages, e.g., automatic translation or transliteration techniques, (2) techniques for solving the problem of translation ambiguity, (3) formal retrieval models for CLIR such as application of the language modeling, (4) methods for searching a multilingual document collection in which two or more languages are used for writing documents, etc.

1.

Introduction

Cross-language information retrieval (CLIR) is a special type of information retrieval (IR) that enables a set of documents written in one language to be searched using queries made in another language. For example, it would be convenient for Japanese people to be able to find documents in English by entering a search request in Japanese into the retrieval system. Especially, as the Internet has spread since the 1990s, the importance of CLIR for allowing users to access information resources written in various languages on the Internet has grown. Some search engines already allow users to use a translation function for searching and viewing web sites in foreign languages. Another area in which CLIR plays an important role is patent retrieval, by enabling searchers to locate easily related patents registered in foreign countries. The IR research community has therefore been tackling a wide range of CLIR problems. The Workshop of Cross-Linguistic Information Retrieval held in August 1996 during the

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

244

Kazuaki Kishida

ACM SIGIR Conference marked a turning point for research on CLIR. Since then, many retrieval experiments have been attempted, and have yielded unexpected enhancements in CLIR techniques. Intuitively, it would seem that machine translation (MT) could easily solve CLIR problems, because CLIR tasks can be reduced to the execution of standard monolingual IR if a given search query or its target documents are automatically translated into the other language. However, as yet there is no perfect MT system that always returns the correct results of translation, and it is suspected that such a system will be developed in the near future. Thus, in the IR field, various unique methods of language processing or document ranking have been explored for improving CLIR performance. Such research has also yielded deeper insights on some previously unconsidered aspects of IR. More specifically, CLIR includes two types of retrieval: 1. bilingual information retrieval (BLIR) 2. multilingual information retrieval (MLIR) In the case of BLIR, the documents are written in a single language different from the query language as in the above example (i.e., a Japanese-English bilingual search). Meanwhile, the target in MLIR is a heterogeneous set of documents written in two or more languages. For example, when Japanese, French and English documents are concurrently searched for a Japanese query, it is MLIR (see Figure 1). In general, MLIR is more complicated than BLIR as discussed later.

BLIR Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

query: Japanese

document collection English

MLIR query: Japanese

document collection French English Japanese

Figure 1. BLIR and MLIR. This article explains in detail the principal techniques of CLIR by extending a previous review [Kishida, 2005]. Specifically, it discusses matching strategies, translation techniques, term disambiguation in the process of translation, formal CLIR models, and MLIR methods.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Methods for Cross-Language Information Retrieval

2.

245

Matching Strategies

2.1.

Matching Operation in CLIR

The most basic operation in IR is to compare the subject representation of a given search request (i.e., ‘query’) with that of each target document, and to measure topical similarity between them. If the degree of similarity (or relatedness) vi between the query q and a document di (i = 1, ..., N ) were calculated by a method, the value vi would enable us to generate a search output in which documents are ranked in descending order of their similarity with the given query. This similarity is often called the document score or retrieval status value (RSV), and is operationally defined in a retrieval model such as a vector space model or a probabilistic model. For example, in the framework of language modeling (LM), vi is computed as the probability that a set of query terms is generated from a given document di , i.e., vi = P (Ωq |di ) where Ωq is a set of terms included in q. Several formulae for computing this probability have been proposed, and a simple one of which is as follows [Hiemstra, 1998a]. P (Ωq |di ) =

Y

αP (t|di ) + (1 − α)P (t),

(1)

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

t∈Ωq

where t is a term in the query and α is a mixing parameter (0 ≤ α ≤ 1). Actually, P (t|di ) in the formula is easily estimated as the relative frequency of term occurrence in the document di and P (t) as the proportion of documents in which the term t appears. Inevitably, when the query and documents are written in different languages, it is not possible to estimate P (t|di ) without matching correctly a query term t with the corresponding term s in each document. Such kind of matching operation is important for CLIR. There are four types of strategies for matching a query with a set of documents in the context of CLIR [Oard and Diekema, 1998]: • Translation 1. Query translation 2. Document translation 3. Interlingual techniques • No translation – Cognate matching

2.2.

Query Translation and Document Translation

Query translation is the most widely used matching strategy for CLIR due to its tractability, i.e., the retrieval system does not have to change inherent components (e.g., index files) at all in response to queries in any language if an external translation module that can convert the text of the query into the document language is incorporated. This is a remarkable advantage of query translation in practice.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

246

Kazuaki Kishida

However, it is difficult to resolve term ambiguity arising in the process of query translation because “queries are often short and short queries provide little context for disambiguation” [Oard and Diekema, 1998]. Suppose that a given query is “post service”. The word “post” has some different senses related to “mail”, “position” and so on, and this short query does not provide sufficient information for selecting a correct translation. Many empirical analyses of transaction log data in search engines on the Internet have shown that actual queries usually consist of about two terms as in this example (e.g., see [Markey, 2007]). In such a situation, the effectiveness of query translation is often limited. Therefore, a few researchers have attempted to translate target documents into the query language (i.e., document translation) due to the fact that sentences included in documents tend to be more complete and to be translated correctly [Oard and Hackett, 1998, Franz et al., 2000, Braschler and Sch¨auble, 2001]. In fact, an experiment [Oard and Hackett, 1998] showed that document translation using commercial MT software outperforms query translation in BLIR from German to English. This result suggests the potential of document translation while the short query problem remains unsolved even if documents were perfectly translated. Furthermore, it is possible to combine the results from query translation and document translation to form a hybrid approach (see Figure 2), e.g., we can use the average (or weighted average) of two document scores which were computed from query translation and document translation respectively in order to rank documents for final output [Scott McCarley, 1999, Braschler, 2004, Kang et al., 2005, Kishida and Kando, 2006]. An advantage of the hybrid approach is that it increases the possibility of correctly identifying documents having the same subject content with the query. Suppose that a term t is included in a given query and its corresponding term in the language of documents is s. If a tool for translating from the query language to the document language can not translate t into s correctly, the system will fail to find documents containing term s by this query translation. However, if another tool for translation in the reverse direction, i.e., the document language into the query language, can identify term t from term s, matching between the query and documents including term s becomes successful. search Japanese query

Japanese docs

query translation

document translation

doc list merge final list

search English query

English docs

doc list

Figure 2. Hybrid of query and document translation. For implementing document translation or the hybrid approach, it is important to solve the problem that document translation is a very cost-intensive task, i.e., it would take too Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Methods for Cross-Language Information Retrieval

247

long to translate all documents by commercial MT software. One possible solution is to employ a high-speed algorithm with low complexity at the cost of translation quality such as a ‘fast document translation’ algorithm [Franz et al., 2000], which is based on a statistical approach developed by IBM research group [Brown et al., 1993]. Also, simple replacement of each term in documents with its translation using a bilingual dictionary is often used as a convenient technique for document translation. Of course, such dictionary-based method is also used for query translation when appropriate MT software is not available as discussed later.

2.3.

Interlingual Techniques

In interlingual techniques, an intermediate space of subject representation into which both the query and the documents are converted is used to compare them. There are two main categories of this technique [Oard and Diekema, 1998]: 1. Matching within a space generated by latent semantic indexing (LSI) 2. Matching via multilingual thesauri

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

2.3.1.

LSI-based Technique

Suppose that there is a document-aligned parallel corpus in which N documents are included and each document has a pair of equivalent texts in two different languages. If the two languages correspond to the query and document languages respectively, this corpus can be used to produce an intermediate space for BLIR. We denote an M1 × N term-document frequency matrix in a language by X1 , where M1 is the distinct number of terms of this language contained in the corpus and each component xji of this matrix (j = 1, ..., M1 ; i = 1, ..., N ) is the frequency (or normalized frequency) of j-th term tj within the text of a document di . Similarly, X2 is assumed to be an M2 × N term-document frequency matrix for another language. Then we constitute an M × N matrix such that   X1 X= , (2) X2 where M = M1 + M2 . According to the theory of linear algebra, the matrix X can be broken down such that X = UΛVT where U is an M × r orthogonal matrix, Λ is an r × r diagonal matrix, V is an N × r orthogonal matrix and r is rank of X. This decomposition is called singular value decomposition (SVD). Latent semantic indexing (LSI) theory [Deerwester et al., 1990, Landauer and Littman, 1990] extracts b principal diagonal elements from Λ (b < r) and interprets that they represent ‘latent’ meanings included in the original X. As a result, a new indexing space Ub Λb VTb is constructed for conceptual retrieval where Ub and Vb are matrices in which only b columns are extracted from original ones, respectively, and Λb is a diagonal matrix including only b principal elements. Since X is derived from a parallel corpus, Ub Λb VTb can be interpreted as a multilingual indexing space in which each meaning is represented independent of its language expressions. Therefore, if we can project a query and a document in different languages into this

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

248

Kazuaki Kishida

indexing space, it is possible to compare them within the space. One of the methods is to −1 T T compute Λ−1 b Ub di and Λb Ub q where di and q are subject representations (e.g., termfrequency vectors) of di and q, respectively [Rehder et al., 1998, Landauer et al., 2007]. Note that di is an M -dimensional vector and its (M1 + 1)-th to M -th elements are zero if di is written in the first language. Similarly, the first to M1 -th elements are zero in the vector of q in the second language. Therefore, an inner product of the two vectors dTi q T T −1 T is always zero, but (Λ−1 b Ub di ) Λb Ub q > 0 if di has semantic similarity with q in the multilingual indexing space. This means that we can use a similarity measure based on T T −1 T the inner product (Λ−1 b Ub di ) Λb Ub q as vi for ranking documents. Similar approaches were employed by some researchers (e.g, see [Berry and Young, 1995, Dumais et al., 1996, Littman et al., 1998]).

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

2.3.2.

Thesaurus-Based Techniques

Another type of interlingual approach is to map a term in a language into the corresponding term in another language via linguistically neutral labels of concepts registered in bilingual thesauri such as WordNet [Diekema et al., 1999, Bian and Lin, 2001], ULMS (Unified Medical Language System) [Eichmann et al., 1998] and so on. For example, we can use sense labels of ‘synsets’ (sets of synonymous words) provided in WordNet. The English word “train” has the sense label “train/1” implicating a line of railway cars, and it is mapped to an Italian synset including “convoglio” and “treno” in MultiWordNet1 . The label “train/1” can be employed as an interlingua for English to Italian BLIR in the case that “train” implicating a line of railway cars is included in an English query (logically, the matching operation with this interlingua is equivalent to using directly such translations as “convoglio” and “treno” listed in the multilingual thesaurus). It is not so easy, however, to find interlingual concepts relevant to a given query in such a multilingual thesaurus. One possibility is to search interlingual concepts (i.e., labels) for the given query using their descriptions given in the thesaurus, and to rank interlingual concepts according to a retrieval model. For example, we may be able to identify correctly the relevant concept “train/1” for the query “trains on a railroad” if an English text in the explanation of “train/1” is searched for the query and this label has the highest score for ranking (see [Eichmann et al., 1998] for details).

2.4. 2.4.1.

Cognate Matching Fuzzy Matching

For BLIR between two similar languages (e.g., English and French), it is possible to identify document terms equivalent to a given query term by using a fuzzy matching technique without any MT. This method is sometimes called cognate matching, and in the most na¨ıve case, untranslatable terms such as proper nouns or technical terminology are left unchanged in the stage of translation. The unchanged terms are expected to match successfully a corresponding term in another language if the two languages have a close linguistic relationship. 1

http://multiwordnet.itc.it/english/home.php

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Methods for Cross-Language Information Retrieval

249

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

A useful device for more effective matching cognates of two different languages is edit distance, which measures similarity between two character strings [Davis, 1997]. For example, Levenshtein distance is defined as the minimum number of deletions, insertions, or substitutions required to transform one string into another string [Galvez and Moya-Aneg´on, 2007]. That is, since the English word “family” can be converted into the French word “famille” by inserting “l” and substituting “e” for “y”, the distance between them is 2. If two terms between which such distance is less than a threshold are heuristically regarded as identical terms, term matching in BLIR can be operated with no translation. Interestingly, Buckley et al. pointed out that “English query words are treated as potentially misspelled French words”, and attempted to treat English words as variations of French words according to lexicographical rules [Buckley et al., 1998]. Furthermore, a kind of rule for transformation, e.g., if a Spanish term starts with “es”, “es” should be replaced with “e” for converting it into an English word, would be helpful for fuzzy matching with edit distance. Such rules may be automatically generated from statistical analysis of language resources (e.g., see [Pirkola et al., 2003] for details). An alternative approach to fuzzy matching is to decompose each word in both the query and documents into n-grams (more specifically, character-based overlapping n-grams), and to perform matching operations between the two sets of n-grams [Hedlund et al., 2002, McNamee and Mayfield, 2002b, McNamee and Mayfield, 2004]. For example, when n = 2 (i.e., bi-gram), “family” and “famille” are decomposed into { fa, am, mi, il, ly } and { fa, am, mi, il, ll, le }, respectively, and the similarity between them is computed. Since they have four common bi-grams, the Dice coefficient is calculated as (2×4)/(5+6) = 0.727..., which can be used as a metric to identify corresponding terms. It is also possible to extract and compare tri-grams (n = 3), quad-grams (n = 4) and so on. In order to enhance matching probabilities, it may be effective to transform the term beforehand according to a heuristic rule such as the Spanish-English example described above (see [Toivonen et al., 2005]). 2.4.2.

Machine Transliteration

When two languages are very different, e.g., English and Japanese, techniques based on edit distance or n-gram inevitably will not work well. However, in such cases, phonetic transliteration from English words may be effective for cognate matching. Gey stated that “...we can often find that many words, particularly in technology areas, have been borrowed phonetically from English and are pronounced similarly, yet with phonetic customization in the borrowing language” [Gey, 2001]. Accordingly, we can use machine transliteration [Knight and Graehl, 1998] for the operation of matching terms in very different languages. There are two methods of machine transliteration: modeling of transliteration and extracting transliterations from a parallel corpus [Kuo et al., 2008]. In a typical transliteration modeling, phonetic coincidence between two terms is examined. For example, the English word “America” corresponds to a Japanese word consisting of four Katakana characters pronounced “a”, “me”, “ri” and “ka”, respectively (see Figure 3). This means that the English term “America” can be identified from the combination of four sound representations “ame-ri-ka” in Japanese if a heuristic rule for converting “ka” to “ca” is introduced (see [Fujii and Ishikawa, 2001, Qu et al., 2003a] for details).

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

250

Kazuaki Kishida

Japanese word:

ࠕ ࡔ ࡝ ࠞ

phonetic elements:

a

me

ri

ka

English word:

A

me

ri

ca

Figure 3. Example of machine transliteration. On the other hand, it is possible to estimate a conditional probability P (t|s) from parallel corpora as discussed later, where s is a transliteration and t indicates an original term. Such kinds of probabilities or alternative statistics indicating the relationship between t and s enable us to produce a list of transliterations [Kuo et al., 2008]. A combination of the rule-based method using phonetic elements and corpus-based method would be a promising strategy for correctly detecting transliterations in CLIR.

3.

Translation Methods It is widely recognized that there are three main approaches to translation in CLIR: • Machine translation • Translation by bilingual machine-readable dictionary

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

• Parallel or comparative corpora-based methods In addition, some researchers have recently attempted to use Internet resources for obtaining translation equivalents.

3.1. 3.1.1.

MT and Dictionary-Based Methods Machine Translation System

As already stated, intuitively, the MT system seems to be a fine tool for CLIR, and if good MT software were available, the CLIR task would become easier. However, in query translation, the MT approach has not always shown better performance than simpler dictionary-based translation. For example, an experiment [Ballesteros and Croft, 1998] indicated the dominance of dictionary-based techniques over a popular commercial MT system (of course, which method is dominant depends highly on the quality of the dictionary and MT system used for each experiment). One of the reasons is that queries are often short and do not provide sufficient contextual information for translation. In particular, a query is often represented as only a set of terms, and it may be difficult to expect MT systems to work well with such poor representation. Also, MT systems usually try to select only one translation from many candidates that each source word may have, which may lead to removing synonyms or related terms from the set of translations, and to missing relevant documents [Nie et al., 1999].

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Methods for Cross-Language Information Retrieval 3.1.2.

251

Dictionary-Based Methods

Simple replacement of source terms using a bilingual machine-readable dictionary (MRD) is a general approach for CLIR when no MT system with an established reputation is available. In the world, numerous languages are spoken, and there are many pairs of languages for which effective MT software has not yet been developed. Therefore, it is still important to explore CLIR methods based on bilingual MRD because such dictionaries are easier to prepare than MT software. Most retrieval systems are still based on the so-called ‘bag-of-words’ architecture, in which both the query and document texts are decomposed into a set of words (or phrases) through a process of indexing. Thus we can translate a query easily by replacing each query term with its translation equivalents appearing in a bilingual dictionary or a bilingual term list. Due to its convenience, the dictionary-based method is also used in document translation as described above. Unlike standard MT systems, it is not difficult in dictionary-based methods to remain multiple translations for each source term. If they contain synonyms or effective related terms relevant to the given query, search performance may be improved by entering them into the IR system (of course, this strategy does not always have positive effects). However, there are some difficulties when executing this method as follows [Ballesteros and Croft, 1997]: • Specialized vocabulary not contained in the dictionary will not be translated. • Dictionary-based translation is inherently ambiguous and may add extraneous information.

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

• Failure to translate multi term concepts such as phrases reduces effectiveness. These defects are the main reasons for degradation of CLIR performance in comparison with that of monolingual retrieval. Hull and Grefenstette stated that “...we learn that translation ambiguity and missing terminology are the two primary sources of error...” [Hull and Grefenstette, 1996]. Also, they reported that manual translation of multi word noun phrases improves retrieval performance. This suggests the importance of translation of multi term concepts. For example, a combination of independent translations from “hot” and “dog” would produce highly inappropriate translations of “hot dog” in many cases. Various methods have been developed for solving problems of out-of-vocabulary, term disambiguation and phrasal translation, as will be discussed later. Practically, it is necessary for implementing dictionary-based translation to compare a string of each source term with that of headwords in the dictionary. A stemming algorithm that removes a suffix from each string is often employed before the matching operation (e.g., “libraries” is automatically converted into “librar”) in order to find successfully the corresponding terms in the dictionary. For enhancing the performance of this process, it may be effective to apply backoff translation in which both the surface form and its stems are taken into account [Oard et al., 2001, Levow et al., 2005]. For example, in four-stage backoff translation, four different matching operations are performed: (1) matching of the surface form of a source term to the surface form of headwords in the dictionary, (2) matching of the stem of a source term to the surface form of headwords, (3) matching of the

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

252

Kazuaki Kishida

surface form of a source term to the stems of headwords and (4) matching of the stem of a source term to the stems of headwords. Since there is no completely perfect stemmer, it is important to use a technique such as backoff translation to increase the possibility of finding the corresponding term in the dictionary.

3.2.

Parallel Corpora-Based Method

Parallel or comparable corpora are useful resources enabling us to extract beneficial information for CLIR. As described already, the cross-lingual LSI approach uses this kind of corpus to construct a multidimensional indexing space. We can also obtain translation equivalents directly from a parallel or comparable corpus by the following methods: • Use of search results from parallel corpus • Construction of bilingual term lists 3.2.1.

PRF-based Method Using Parallel Corpus

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Suppose that we execute English to French bilingual searches using a documentaligned parallel corpus of English and French languages. The first approach [Davis and Dunning, 1995, Yang et al., 1998] tries to extract French terms appearing frequently in French documents obtained from the parallel corpus by searching for the given English query (see Figure 4). That is, since each French document is aligned with an English document in the corpus, we can identify French documents corresponding to English documents searched for the given English query. Naturally, such French documents are expected to include search terms relevant to the query.

English query

English-French parallel corpus Search results

search

Final results

search

French terms

French document collection (target)

Figure 4. Example of BLIR using parallel corpus. This approach can be regarded as a kind of pseudo-relevance feedback (PRF), which is often used for query expansion (i.e., adding new search terms to the original query) in IR experiments. Its basic assumption is that top-ranked documents searched for a given query tend to be relevant and contain effective terms other than those included in the initial query.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Methods for Cross-Language Information Retrieval

253

Table 1. Term co-occurrence frequency in parallel corpus

s appears s does not appears Total

t appears nts nt − nts nt

t does not appears ns − nts N − ns − nt + nts N − nt

Total ns N − ns N

According to a standard PRF technique, such effective terms can be identified from a list of terms ranked by scores computed for each term t such that rt × log

(rt + 0.5)(N − R − nt + rt + 0.5) , (R − rt + 0.5)(nt − rt + 0.5)

(3)

where N is the total number of documents in the collection, nt is the number of documents including the term t (nt ≤ N ), R is the presupposed number of top-ranked documents (e.g., R = 30) and rt is the number of top-ranked documents including the term t (rt ≤ R). By using this score, it is possible to select some top-ranked French terms from the result of searching the parallel corpus.

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

3.2.2.

Estimation of Association between Terms

In the second approach, a bilingual term list is generated from parallel or comparable corpora based on empirical term association computed from co-occurrence statistics as shown in Table 1. The generated list can be used as a bilingual MRD. Suppose that t indicates a French word and s an English word. In this case, nts is the number of alignments including both t and s in a parallel corpus. From the data in Table 1, it is possible to calculate association τts between t and s as the Jaccard’s coefficient τts = nts /(nt + ns − nts ) [Adriani, 2002] or mutual information (MI) [McNamee and Mayfield, 2002a] τts = log

P (t, s) N nts = log P (t)P (s) nt ns

(4)

in which it is assumed that P (t, s) = nts /N , P (t) = nt /N and P (s) = ns /N . Also the logarithm of a likelihood ratio −2 log λ [Dunning, 1993] such that −2 log λ = 2 log

ζ(p1 , nts , ns )ζ(p2 , nt − nts , N − ns ) ζ(p, nts , ns )ζ(p, nt − nts , N − ns )

(5)

can be used as association τts where ζ(x, y, z) ≡ xy (1 − x)y−z Cp1 = nts /ns Cp2 = (nt − nts )/(N − ns )Cand p = nt /N . Furthermore, it is possible to employ other metrics such as the Dice coefficient, χ2 statistics and so on, which are computed from a contingency table like Table 1. Otherwise, a method for constructing a so-called similarity thesaurus can be used to compute term association between t and s if a concatenation of two aligned documents in different languages is assumed to be a single document [Sheridan and Ballerini, 1996, Braschler and Sch¨auble, 2000]. According to the standard vector space model for IR, it is

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

254

Kazuaki Kishida (t)

possible to define an N -dimensional term vector of which i-th element wi is the weight of the term in di , e.g., (t)

wi = xi (t) × log N/nt ,

i = 1, ..., N,

(6)

where xi (t) is the frequency of term t in document di , and nt is the number of documents including t in this case 2 . If term similarity is computed as a cosine measure between the two vectors of t and s, query terms in the target language can be selected for each source term s according to their cosine values 3 . Also, as another method for constructing a bilingual term list from a parallel or comparable corpus, a more complicated technique based on cognate matching and morpho-semantic analysis has been proposed by [Mark´o et al., 2005]. 3.2.3.

Estimation of Translation Probability

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Some researchers in the CLIR field (e.g., [Nie et al., 1999]) have attempted to estimate translation probability P (t|s) from parallel corpora according to a well-known algorithm developed by a research group at IBM [Brown et al., 1993]. By executing the algorithm for a set of sentence alignments included in a parallel corpus, a bilingual term list with a set of probabilities that a term is translated into equivalents in another language is automatically generated. The algorithm includes five models, Model 1 through Model 5, of which Model 1 is the most basic and is often used for CLIR. In particular, researchers employing the language modeling approach for CLIR (see below) have used the IBM Model 1 for computing translation probabilities (e.g., [Xu et al., 2001, Kraaij, 2002]). The fundamental idea of Model 1 is to estimate each translation probability so that the probability defined as Pr(t|s) =

m X l Y ε P (tj |sk ), (l + 1)m

(7)

j=1 k=0

is maximized, where t is a sequence (usually a sentence) of terms t1 t2 . . . tm , s is the corresponding sequence of terms s1 s2 . . . sl and ε is a parameter (s0 indicates an empty in s). That is, the translation probability P (tj |sk ) in Equation (7) is determined so that Pr(t|s), which is the probability that sentence s is translated into t, takes a maximum value. Actually, each P (tj |sk ) is estimated by iterative computation as an EM algorithm (see the appendix for details). The algorithm based on the IBM Model has become widely available with the release of a software package, the GIZA++ toolkit [Och and Ney, 2003], incorporating it as a component, but other models for statistical machine translation have also been developed. For example, another statistical translation model by [Melamed, 2000] was used for generating Japanese-English bilingual thesauri from bilingual corpora [Tsuji and Kageura, 2006]. In addition, estimation techniques based on EM algorithms have been proposed by some researchers (e.g., [Hiemstra, 1998b, Koehn and Knight, 2000, Cao and Li, 2002]). 2

Note that log N/nt is a typical idf (inverse document frequence) factor, which is generally assumed to be higher for more effetive index terms. Equation (6) is a kind of tf-idf weighting, where ‘tf’ indicates term frequency (for details, see IR textbooks, e.g., [Manning et al., 2008]) p 3 In general, the cosine measure of two vectors x and y is defined as cos(x, y) = xT y/ xT x × yT y.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Methods for Cross-Language Information Retrieval 3.2.4.

255

Availability of Parallel Corpus

Terms used in current affairs and people’s names are often not registered in MRDs or dictionaries incorporated into MT systems, in which case the correct translation can not be acquired. If a parallel corpus consisting of current news articles is available, it may be possible to construct a bilingual term list including current terminology. This is a remarkable advantage of parallel corpus-based methods. However, it would not always be possible to obtain such parallel corpus relevant to each CLIR situation, thus preventing the generalized application of parallel corpus-based methods. Even if a parallel corpus in two languages corresponding to a BLIR task is available, the search performance will inevitably decrease when the subject domains covered by the parallel corpus do not match the BLIR situation [Rogati and Yang, 2004]. One means of overcoming this problem may be to generate automatically parallel corpora from Internet resources. For example, official web sites of organizations or institutions often have English and non-English pages with almost the same contents, from which parallel or comparable corpora could be produced (see [Nie et al., 1999, Resnik, 1999, Nie, 2000] for details). Also, automatic construction of document alignments from independent collections in two different languages has been explored by [Talvensaari et al., 2006]. The fundamental operation for detecting such alignments is BLIR using an MT system or MRD, i.e., we have to specify corresponding documents in one collection by searching for translations of key terms extracted from a document in another collection. If appropriate pairs of documents are detected by the BLIR, it may be possible to find useful target terms other than translations of the key terms used as a query from the aligned documents.

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

3.3. 3.3.1.

Out-of-Vocabulary Problem Detecting Unknown Translations from Web Documents

If the dictionary used for translation does not include the target entries, i.e., the outof-vocabulary (OOV) situation, a special device is needed if cognate matching or machine transliteration can not be applied. Suppose that entries of a Japanese-English dictionary do not contain a particular Japanese term represented by Kanji characters, for which machine transliteration does not work. A promising method for resolving the OOV problem is to extract translations from Japanese Web documents, in which English equivalents in parentheses are often provided for proper nouns or technical terms [Chen and Gey, 2003]. It is possible to identify such English equivalents by examining frequencies of cooccurrence with the un-translated term in top-ranked documents fetched automatically from a search engine (e.g., Google) via its application program interface (API) by searching for the un-translated term [Zhang and Vines, 2004]. A simple heuristic rule is to select the English term co-occurring most frequently in a presupposed range of word sequence (i.e., text window) with the un-translated term. Of course, we can measure the association between two terms by using a metric such as χ2 statistics and so on, and select translations based on its values [Cheng et al., 2004]. Another approach for measuring the term association is to compare context vectors of two terms [Cheng et al., 2004]. We denote the context vector of term t by wt , of which

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

256

Kazuaki Kishida

the j-th element represents the weight of the j-th term appearing in the top-ranked Web documents. For example, the j-th element can be defined as x(t, tj ) × log N/ntj where x(t, tj ) means the number of times that term tj appears with t in a text window of fixed size. Since the un-translated term and its translation possibly share common contextual terms in the top-ranked Web documents, the similarity of their context vectors is expected to be relatively higher. The similarity can be computed as a cosine measure of two vectors wt and ws according to the standard vector space model for IR. 3.3.2.

Combining Multiple Language Resources

A method for reducing the possibility of the OOV problem occurring in the translation process is to combine multiple translation resources. For example, we can merge translation results from distinct types of resources such as MRD, MT system and parallel corpus (e.g., [Xu et al., 2001]). For merging the results, it is possible to employ techniques of data fusion or query combination [Jones and Lam-Adesina, 2002]. Suppose that we have results from two distinct MT systems. In the case of data fusion, two document scores computed from outputs by the two MT systems respectively were summed for each document. On the other hand, in query combination, before the estimation of document scores, a single query was formed by taking the unique terms from two outputs.

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

3.4.

Pivot Language Approach

As already mentioned, it is not always possible to obtain bilingual resources for a particular pair of languages in a BLIR task. A promising technique to circumvent the problem of limited availability of linguistic resources would be the pivot language approach, in which an intermediate language acts as a mediator between two languages for which no bilingual resources are available. Suppose that a BLIR task between Japanese and Dutch is requested by a user. Even if any language resource between Japanese and Dutch is not available, it would be easier to find Japanese-English and English-Dutch resources since English is widely used as an international language. Thus BLIR between Japanese and Dutch can be performed via English as an intermediary without direct bilingual resources between Japanese and Dutch. The pivot language approach would also alleviate the problem of explosive combinations of languages, i.e., if we have to execute BLIR tasks between each pair of n languages, O(n2 ) resources are needed. However, the pivot language approach enables us to handle the complex job with only O(n) resources [Gey, 2001]. A basic pivot language approach is transitive translation of a query using two bilingual dictionaries [Ballesteros, 2000]. In the case of searches from Japanese to Dutch via English, if Japanese-English and English-Dutch MRDs are available, CLIR can be performed by replacing Japanese query terms with the corresponding English equivalents and successively substituting the English equivalents with the Dutch equivalents. Of course, if JapaneseEnglish and English-Dutch MT systems can be used, a similar transitive translation is also feasible. In dictionary-based transitive translation, translation ambiguity becomes a more serious problem. Suppose that a Japanese source query consists of three terms, and every term

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Methods for Cross-Language Information Retrieval

257

has three English equivalents. If every English equivalent has three Dutch equivalents, simple replacements will produce 27 (= 33 ) search terms in total from only three source terms, and the final set of search terms will inevitably contain some irrelevant translations [Kishida and Kando, 2004] (see Figure 5). Therefore, it is important to apply a translation disambiguation technique (discussed later) to the set of intermediary or target terms obtained from the bilingual dictionaries. Actually, an experiment [Kishida et al., 2005] reported that translation disambiguation for intermediary terms did not have a remarkable effect in German-Italian bilingual searches via English and it was enough to disambiguate only final Italian terms. Japanese query terms (source) relevant translation

irrelevant translation







irrelevant translations















Dutch query terms (target)

… irrelevant translation

: English query term (intermediary)

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Figure 5. Example of dictionary-based query translation via pivot language. Retrieval experiments on the dictionary-based query translation via a pivot language have been conducted for various combinations of languages, e.g., English > French > German [Franz et al., 1999], French > English > German [Gey et al., 1999], German > English > Italian [Hiemstra and Kraaij, 1999], Japanese > English > Chinese [Lin and Chen, 2003], Chinese > English > Japanese [Chen and Gey, 2003], and so on (these are only a portion of many experiments). In particular, Franz et al. proposed some interesting techniques for searching German documents with English queries [Franz et al., 1999] as follows (the intermediary is French). • Convolution of translation probability: Estimating translation probability from an English term s to a German term t through French terms f such that P (t|s) = P f P (t|f )P (f |s).

• Automatic query generation from the intermediate language corpus: Generating French queries automatically by simply merging all non-stopwords in the top-ranked French documents searched by the English-French BLIR system, and entering the French query into the French-German BLIR system.

3.5.

Translation Quality

The quality or correctness of the translation greatly affects CLIR performance. For example, experiments [McNamee and Mayfield, 2002a, Xu and Weischedel, 2005] have shown that the size or lexical coverage of MRD has an influence on the effectiveness of Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

258

Kazuaki Kishida

CLIR, i.e., as the number of entry terms in the dictionary is reduced and as its translation quality becomes lower, search performance gradually deteriorates. CLIR performance can be represented as a regression model with two independent variables, translation quality z1 and ‘ease of searching for query’ z2 , i.e., y = a + b 1 z 1 + b2 z 2

(8)

where y denotes CLIR performance, and a, b1 , b2 are parameters [Kishida, 2008]. The ‘ease of searching for query’ is an inherent nature of a given query, which is independent of the translation process. For example, if the query does not contain any specific concept that explicitly designates its content, it will be difficult to find relevant documents even if the translation is perfectly correct. Therefore, it is necessary to take the ‘ease of searching for query’ into consideration in the regression model [Kishida, 2008]. In retrieval experiments in the laboratory, the ‘ease of searching for query’ z2 can be measured by evaluation indicators (e.g., average precision) representing the performance of monolingual searches for which correct translations of queries are given by human experts. Similarly, it is possible to gauge translation quality z1 by a metric for assessing automatically translation results such as well-known BLEU [Papineni et al., 2002] based on the correct translations. In particular, a metric called WAMU (weighted average for matching unigrams) that is specifically designed to evaluate translations in CLIR situations was developed [Kishida, 2008]. A simpler formula of WAMU is: 

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

W AM U = 

m X j=1

−1

ωj 

m X j=1

ωj

min(cj , c˜j ) , cj

(9)

where cj is the total number of the j-th unigram (i.e., a word) in the translation, c˜j is the total number of the j-th unigram in an answer given by a human expert (i.e., a correct translation), ωj is the weight of the j-th unigram, and m is the number of unigrams in the translation. The weight ωj is calculated such that ωj = log(N/ntj ) where ntj is the number of documents including the j-th unigram tj (in CLIR situations, ntj can be obtained from the target document collection). Suppose that two translations are obtained with an answer as follows. - Translation 1: Database management system issue - Translation 2: Database administration system problem - Answer: Database management system problem In either of the translations, only one term is different from the corresponding term in the answer, i.e., “issue” in Translation 1 and “administration” in Translation 2. If these translations are to be used as a query, Translation 1 is clearly better because it represents “database management system” correctly. If the numbers of documents including each term are as shown in Table 2 and N = 1000, the WAMU scores for Translations 1 and 2 are computed as follows.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Methods for Cross-Language Information Retrieval

259

Table 2. Sample data for calculating WAMU j 1 2 3 4

1 2 3 4

Terms tj Translation 1 database management system issue Total Translation 2 database administration system problem Total

min(cj , c˜j )

ntj

log(N/ntj )

1 1 1 0 -

5 50 50 100 -

5.30 3.00 3.00 2.30 13.59

1 0 1 1 -

5 50 50 100 -

5.30 3.00 3.00 2.30 13.59

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Translation 1: (5.30 + 3.00 + 3.00)/13.59 = 0.83 Translation 2: (5.30 + 3.00 + 2.30)/13.59 = 0.78 As this example indicates, WAMU differentiates adequacy of translation between specific and non-specific terms, i.e., mistranslation of a specific term reduces its WAMU score more largely than non-specific terms due to idf factor log(N/ntj ). An experiment in [Kishida, 2008] showed that the regression model containing two independent variables, ‘ease of searching for query’ and translation quality, can explain approximately 60% of the variation in CLIR performance. In this result, translation quality has a statistically significant effect, which means that translation quality is crucial for enhancing CLIR effectiveness, as would be expected.

4. 4.1.

Term Disambiguation Techniques Translation Ambiguity

For improving translation quality, it is indispensable to select a correct translation from a set of candidates by disambiguating the sense of a given source term. In general, word sense disambiguation (WSD) is a key element in various applications such as machine translation, information retrieval and hypertext navigation systems, content and thematic analysis, grammatical analysis, speech processing, text processing and so on [Ide and V´eronis, 1998]. For CLIR, it is often necessary to disambiguate translations enumerated under each headword in a bilingual MRD or a term list generated from a parallel corpus. If all translations listed in bilingual resources are straightforwardly adopted as search terms, extraneous or superfluous terms irrelevant to the original query usually diminish the effectiveness of the search. Thus it is desirable that only relevant terms will be automatically

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

260

Kazuaki Kishida

or semi-automatically selected from a set of translation candidates. A simple method is to take only translations corresponding to the first sense listed in the dictionary. Alternatively, we could investigate the frequency of each translation within a corpus and use only the most frequent translation. However, such heuristic strategies would be insufficient to resolve the ambiguity of words that are highly homonymous. Several more sophisticated methods have been explored in the field of CLIR as follows. 1. Use of part-of-speech (POS) tags 2. Use of parallel corpora 3. Use of co-occurrence statistics in the target corpus 4. Use of query expansion technique It should be noted that the term ‘corpus-based disambiguation’ is often used in literature for collectively referring to techniques 2 through 4. As another solution, the ‘structured query model’ has also been investigated for improving search performance in cases where multiple translations are obtained from a bilingual dictionary. This model will be discussed in another section.

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

4.2.

Use of Part-of-Speech Tags

The basic idea of using part-of-speech (POS) tags for translation disambiguation is to select only translations having the same POS tag as that of the source query term [Davis, 1997, Davis and Ogden, 1997, Ballesteros and Croft, 1998]. For example, in the case of applying it to the BLIR task from English to Spanish, the translation is selected as a search term only if the POS tag of a Spanish equivalent listed in an English-Spanish dictionary coincides with that of the English query term. This technique requires POS tagging tools for both languages.

4.3.

Parallel Corpus-Based Methods

It is possible to determine the best translations among the candidates according to the result of searching a parallel corpus for the original query [Davis, 1997, Davis, 1998, Boughanem et al., 2002]; specifically, a typical procedure is as follows (see also Figure 6). 1. Identify a set of translations for each term in a given source query by using an MRD. 2. Search a part of the parallel corpus written in the target language for each translation respectively, and save each set of documents in the target language. 3. Search the other part of the parallel corpus written in the source language for the source query. 4. Select a translation for which the set of documents is the closest to the set of documents searched by the source query (e.g., choose a translation for which documents are mostly included in the set of search results by the source term). This procedure is repeated for each query term until a final set of the best translations is obtained.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Methods for Cross-Language Information Retrieval Search results

A parallel corpus

comparison

261

A source query term

translation

: source language

: target language

Figure 6. Outline of parallel corpus-based disambiguation.

4.4.

Disambiguation Based on Term Co-occurrence Statistics

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

A basic assumption underlying disambiguation techniques based on term co-occurrence is that “the correct translations of query terms should co-occur in target language documents and incorrect translations should tend not to co-occur” [Ballesteros and Croft, 1998]. For example, in the case of a French to English search, if the source query includes a concept of “database management system”, the combination of translations, “database”, “management” and “system” is reasonably expected to appear more frequently in the target document collection than that of “database”, “administration” and “system”. Thus, if we appropriately employ statistics on co-occurrence frequencies, “management” would be able to be identified as a correct translation in the context 4 . This kind of methods based on term co-occurrences does not need any extra corpus because the statistics can be obtained from the target documents. This is a remarkable merit in practice. 4.4.1.

Best Pairs Selection

When a given query includes many terms, the frequency of their combination may become very low or zero. Thus, the query is usually decomposed into a set of term pairs, and co-occurrence frequencies of the corresponding two translations are used for determining final query terms {t˜1 , ..., t˜m } in the target language. For example, we can select them such that [ t˜j = arg max sim(t, t′ ), t′ ∈ Tk ; j = 1, ..., m, (10) t∈Tj

k6=j

where Tj indicates a set of translations in the target language for j-th source term sj (j = 1, ..., m) and sim(t, t′ ) is a term association or similarity between the t and t′ computed from term co-occurrences. Equation (10) simply means that we choose repeatedly a pair of translations with the highest similarity from those excluding pairs that have already been selected [Bian and Chen, 1998], i.e., if a pair with the highest similarity includes a translation of the source term for which a translation was previously determined, this translation is ignored. As a variation of this method, we may be able to consider only similarities with the translations selected in the previous step [Jang et al., 1999]. 4

Moreover, it is possible to incorporate term co-occurrence statistics into the process of estimating translation probabilities (see [Liu et al., 2005] for details). Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

262

Kazuaki Kishida

The term similarity sim(t, t′ ) can be measured as MI, the Dice coefficient and so on, which are computed from term occurrence and co-occurrence statistics in the target document collection. For example, if we employ MI in Equation (4), its probabilities are operationally defined such that P (t, s) = nts /N , P (t) = nt /N and P (s) = ns /N where nt and ns are the numbers of sentences including term t and s, respectively, nts is the frequency of appearance of both terms in the same sentence, and N is the total number of sentences included in the target collection. While the number of observations is normalized by the size of the corpus according to [Church and Hanks, 1990] in these formulae, it is possible to adopt other definitions (e.g., see [Gao et al., 2001]). The term co-occurrence statistics required for computing the similarity can be compiled in a process of constructing index files for searching the target collection, or obtained from a search engine on the Internet via API [Maeda et al., 2000]. Fortunately, the computational complexity of selecting the best pairs based on Equation (4) is not so high. If the similarity measure used in this process is symmetric such as MI, the number of pairs to be measured for the selection, Mp , amounts to Mp =

m−1 X

m X

|Tk | × |Th |,

(11)

k=1 h=k+1

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

(see [Kishida, 2007]). It should be noted that Mp does not exceed b2 m(m − 1)/2 where m(m − 1)/2 corresponds to the number of all combinations of two source terms and b indicates the maximum number of translations for a source term, i.e., b = maxj=1,...,m |Tj |. For example, if there are five source query terms and every set of translations includes five words respectively, we obtain Mp = (5 × 5) × (5 × 4/2) = 250. 4.4.2.

Best Sequence Selection

However, this algorithm may yield erroneous translations because it checks only a local relationship between just two translations at each step for selecting a pair. Suppose that, in the above example, the pair of “administration” and “system” has the highest degree of similarity in the target document collection even though the collection contains some documents on “database management system”. In this case, the correct translation, “management”, is never obtained by the algorithm based on Equation (10). Since this algorithm looks at only a very limited range (i.e., a span of just two terms) in the query, a few pairs having strong relationships out of the context of the query tend to impact excessively on the result. A straightforward solution to this problem is to focus on all relationships between the query terms [Seo et al., 2005]. We denote a sequence of translations by τ = {t1 , ..., tj , ..., tm } where tj ∈ Tj , i.e., τ is constituted by taking arbitrarily a term from each translation set, respectively (see Figure 7). If the sum of degrees of similarity between all translations included in such a sequence, i.e., U (τ ) =

m−1 X

m X

sim(tk , th ),

tk , th ∈ τ,

k=1 h=k+1

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

(12)

Methods for Cross-Language Information Retrieval

263

can be computed, the sequence with the highest value of the sum should be selected as the final set of translations (similar strategies were adopted by [Maeda et al., 2000, Qu et al., 2003b]). Since Equation (12) contains all relationships between possible pairs of translations in each sequence, we can avoid errors caused by the locality of range in Equation (10). For example, the possibility that “management” is correctly selected remains even if the pair of “administration” and “system” has the highest similarity. Source terms

Sequence Translations Pair

Figure 7. Sequences and pairs of translations. However, the computational complexity of processing the algorithm based on Equation (12) is very high. The total number of sequences to be processed is given by Mτ = |T1 | × ... × |Tm | =

m Y

|Tj |,

(13)

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

j=1

and the upper limit of Mτ amounts to bm . In addition, each sequence includes m(m − 1)/2 pairs as expressed in Equation (12), and therefore, in total, we need to deal with bm m(m − 1)/2 pairs at most in the process of disambiguation (note that a single pair is repeatedly counted). If we have five source query terms and every set of this translation includes five words respectively, the number of pairs amounts to 31,250 (= 55 × 5 × 4/2). As this example shows, it takes much longer to select translations when using Equation (12). 4.4.3.

Approximation for Best Sequence Selection

One way to reduce the computational difficulty of the best sequence method is to apply the maximal value of similarity between a given translation and those of another source query term, i.e., C(t, Tk ) = max sim(t, t′ ), t 6∈ Tk , (14) ′ t ∈Tk

which was employed in some experiments (e.g., [Adriani, 2000, Gao et al., 2001]). C(t, Tk ) is often called ‘cohesion’. In order to solve ambiguity of translations using this quantity, for each term t, it is necessary to compute the sum of C(t, Tk ) over all sets Tk except the set that includes the term t itself, and then to select the term with the highest score from each set of translations respectively [Adriani, 2000, Gao et al., 2001], i.e., X t˜j = arg max C(t, Tk ), j = 1, ..., m. (15) t∈Tj

k6=j

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

264

Kazuaki Kishida

The number of translation pairs to be processed in this algorithm for determining the P translation of the j-th source term is |Tj | × k6=j |Tk |, and therefore, the total number amounts to   m X X |Tj | × Mp = |Tk | , (16) j=1

k6=j

which does not exceed m×b×(m−1)b = b2 m(m−1). If we have five source query terms and every set of translations includes five words respectively, we obtain Mp = 5×5×4×5 = 500. Although its degree of computational complexity is almost the same as that of the method in Equation (10), the algorithm using C(t, Tk ) avoids making local judgments in the selection of translations to some degree unlike Equation (10). That is, a translation of a given source term sj is chosen using information on relationships with all other source terms through the C(t, Tk ) in Equation (14), not with only a single source term. In our example, even if the similarity score between “administration” and “system” is the highest one, it is possible that “management” is correctly selected as a translation, i.e., the final sum of C(t, Tk ) for “management” may become greater than that for “administration” due to the contribution of a high degree of similarity between “management” and “database”. One experiment [Kishida, 2007] has reported that the search performance of the disambiguation technique based on Equation (15) is compatible with that of the best sequence method using Equation (12).

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

4.5.

PRF-based Techniques for Disambiguation

We can apply PRF techniques to translation disambiguation [Ballesteros and Croft, 1997, Ballesteros and Croft, 1998]. In CLIR situations, two kinds of PRF are feasible: • Pre-translation feedback • Post-translation feedback Suppose that we have a corpus in the source language which is independent of the target document collection. First, this corpus is searched for a given source query, and resulting documents are analyzed prior to translation for CLIR in order to add a set of new terms to the source query (pre-translation feedback). The new terms can be selected based on term weights, e.g., Equation (3). Second, after translation, standard PRF can be applied using the target document collection (post-translation feedback). Inevitably, for executing the pre-translation feedback, an extra corpus in the source language is needed unlike the post-translation feedback working on the target collection. The pre-translation feedback may improve the precision (see [Ballesteros and Croft, 1997]), because PRF is basically executed using the entire query, not each source term respectively. That is, synonyms or related terms corresponding to the correct meaning of each source term within the context of the query are expected to be automatically added through the PRF process. An experiment [McNamee and Mayfield, 2002a] has suggested that the pre-translation query expansion is useful when lexical coverage of

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Methods for Cross-Language Information Retrieval

265

translation resources is poor. On the other hand, the post-translation feedback is just a standard PRF, and therefore, the recall ratio would increase by applying it as many IR experiments have shown. It is also possible to determine explicitly a final translation t˜j for each source term (j = 1, ..., m) based on frequencies of term occurrence in output from the initial search of post-translation feedback [Kishida and Kando, 2004, Kishida, 2007]. In the first stage of this process, the target document collection is searched for a set of all translation candidates, and the number of documents including each translation candidate in the list of top-ranked documents by this search is calculated. Finally, t˜j = arg max rt ,

j = 1, ..., m

(17)

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

t∈Tj

is selected as a final translation for the j-th source term where rt indicates the number of documents including a translation candidate t within the top-ranked documents. As Yamabana et al. pointed out, unexpected false combinations of translations may be generated by disambiguation techniques based on term co-occurrences discussed in the previous section because it is possible that two translations having no relation within the context of a given source query tend to co-occur frequently in the whole document collection [Yamabana et al., 1998]. That is, suggestions from macro-statistics compiled using the whole collection are not always valid in the sense implied by a particular query. A similar problem occurs when applying query expansion techniques to general IR situations. It is widely known that query expansion techniques using statistical thesauri generated through term co-occurrence statistics cannot achieve better performance than PRF in which new search terms are locally identified from a restricted set of top-ranked documents searched for the given query. Similarly, in the context of translation disambiguation for CLIR, local analysis of the target document collection such as Equation (17) may yield better results than global analysis. In addition, it should be noted that the technique based on Equation (17) is easier to implement in IR systems having a standard PRF function. This is a practical advantage of this technique.

4.6.

Disambiguation for Phrasal Translation

As Ballesteros and Croft pointed out that “...failure to translate multi term concepts as phrases reduces effectiveness” [Ballesteros and Croft, 1997], phrasal translation is certainly significant for CLIR. The basic technique is to search a bilingual dictionary or a term list including phrases or compound words as its headwords. That is, they can be automatically identified in the source query by simple matching operations against headwords of such a language resource. Also, if a part-of-speech tagger is available in this process, a word combination of ‘noun-noun’ or ‘adjective-noun’ would be reasonably assumed to be a compound word. Inevitably, the coverage of lexical resources to be used will not always be sufficient, i.e., all phrases or compound words are not always found in the resources as headwords. When an untranslatable phrase is included in a given source query, there is no way other than to attempt word-by-word translation, which may cause a term ambiguity problem

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

266

Kazuaki Kishida

[Ballesteros and Croft, 1997, Ballesteros and Croft, 1998]. Therefore, one of the disambiguation techniques discussed above is required in this word-by-word translation process. However, it should be noted that there are two kinds of compound words as follows [Pirkola et al., 2001]. • Compositional compounds: The meaning can be derived from meanings of the component words, e.g., “database management system”. • Non-compositional compounds: The meaning cannot be derived from meanings of the component words, e.g., “hot dog”. In many cases, disambiguation methods based on term co-occurrences may have limitations for detecting correctly translations of non-compositional compounds. It is thus indispensable to augment the coverage of bilingual dictionaries in order to enhance the quality of phrasal translations. This could be done by some techniques for detecting OOV from the web (see the above section), for extracting phrasal representations from parallel or comparable corpora (see [Lopez-Ostenero et al., 2005]), or for disambiguating directly noun phrases not at the level of distinct words (see [Gao and Nie, 2006]).

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

4.7.

Other Disambiguation Techniques

In a technique called bi-directional translation, backward translations in which translation results are automatically re-translated into the original language are used for ranking translation candidates [Boughanem et al., 2002]. For example, when French terms are translated into English ones, first a set of English equivalents for each French term is extracted from a French-English bilingual dictionary. Next, using an English-French dictionary, each English equivalent is reversely translated into a set of French terms. Basically, if the set includes the original source term, the English translation equivalent is chosen as a preferred translation. When a pivot language approach to translation is used, we can apply lexical triangulation to translation disambiguation [Gollins and Sanderson, 2001, Lehtokangas et al., 2004]. In this technique, two pivot languages are used independently, and an attempt is made to remove erroneous translations by taking only translations in common obtained from both ways of transitive translation using the two pivot languages respectively. Also, a disambiguation technique via dynamic clustering of search results has been explored by [Lee et al., 2004]. In the first stage of this method, all translation candidates are used as the search query without any disambiguation like the PRF-based method, and in the next stage, top-ranked documents searched for the initial query are divided into some clusters by a clustering algorithm. Finally, the degree of similarity between the query and a cluster to which each document belongs is reflected in its document score for ranking, and documents are re-ranked according to the new scores. If each cluster correctly corresponds to one of the multiple senses and appropriate clusters for the given query can be identified, search performance may be enhanced by the re-ranking.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Methods for Cross-Language Information Retrieval

5.

267

Formal Models for CLIR

If representations in two different languages of a given query and documents become comparable after the translation process, the BLIR problem reduces to that of standard monolingual IR. That is, after the translation is completed, the remaining task is how to compute document scores for ranked output, in which a retrieval model such as Equation (1) is applied in a standard manner of monolingual IR. On the other hand, some researchers have tried to incorporate directly the translation process into a retrieval model.

5.1.

Language Modeling for CLIR

It is relatively easy to incorporate translation probabilities P (t|s) into Equation (1), which is a typical formula derived from language modeling for IR. Figure 8 shows the basic idea of the incorporation, i.e., we assume that P (t|di ) =

X

P (t|sk )P (sk |di ),

(18)

k

where sk indicates a term included in document di . It should be noted that t is a query term in a different language from that of documents in this case. Equation (18) means that P (t|di ), which is the probability that query term t is generated from document di , is decomposed into term occurrence probabilities P (sk |di ) and translation probabilities P (t|sk ), and that its value is computed by summing the products of them for all sk . By directly substituting Equation (18) into Equation (1), we finally obtain:

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

P (Ωq |di ) =

Y

t∈Ωq

α

X

P (t|sk )P (sk |di ) + (1 − α)P (t),

(19)

k

according to [Miller et al., 1999, Xu et al., 2001, Xu and Weischedel, 2005]. translation prob. P(t|s)

䋫 P(t|d)

t is a query term.

㸠 information 㬍 t 㸠 report 㬍 t 㸠 intelligent 㬍 t

term occurrence prob. P(s|d) information report intelligent

……………... information.. …report….... …………….. …intelligent.. …………..

Document

Figure 8. Language modeling with translation probabilities. In Equation (1), for preventing P (Ωq |di ) from automatically becoming zero when P (t|di ) = 0 for term t, a general probability P (t) is added to P (t|di ) with a mixing parameter α. This modification is generally called Jelinek-Mercer smoothing, in which the term αP (t|di ) + (1 − α)P (t) is considered to be the ‘true probability’ that t appears in di .

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

268

Kazuaki Kishida

If translation probabilities are applied to the true probability that term sk appears in di , we obtain another formula: YX P (Ωq |di ) = P (t|sk )[αP (sk |di ) + (1 − α)P (sk )], (20) t∈Ωq

k

for language modeling in BLIR [Hiemstra and Kraaij, 1999]. While P (t) in Equation (19) needs to be estimated using an extra corpus in the query language (e.g., Pˆ (t) = nt /N ), an estimation of P (sk ) in Equation (20) can be obtained from the target document collection. As discussed above, translation probability P (t|sk ) can be estimated by the IBM model if an appropriate parallel corpus is available. Fortunately, the GIZA++ toolkit including a module for the IBM model has been developed, and we can use it without writing source code. If no parallel corpus is available, other methods have to be employed. A simple one is to count the number of translations for each source term in a bilingual dictionary. For example, if a source term s has ms translations t1 , ..., tms , the translation probabilities can be assumed such that P (tj |s) = 1/ms (j = 1, ..., ms ) uniformly [Xu et al., 2001]. As a more sophisticated method, an EM algorithm computing iteratively the translation probabilities based on term co-occurrence statistics has been proposed by [Monz and Dorr, 2005] (the initial values of translation probabilities are estimated as the above P (tj |s) = 1/ms from a bilingual dictionary). Since term co-occurrence statistics can be compiled from the target document collection, this algorithm allows us to estimate translation probabilities without any parallel corpus.

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

5.2.

Relevance Model

The relevance model (RM) developed by [Lavrenko and Croft, 2001] can be theoretically applied to CLIR situations [Lavrenko et al., 2002, Larkey and Connell, 2005] since essentially RM is a variation of LM. In RM, the document score is often calculated as relative entropy E(R k di ) between P (t|di ) in Equation (1) and P (t|R), which is the average probability that t is included in a relevant document, i.e., X P (t|R) (21) E(R k di ) = P (t|R) log P (t|di ) t∈ΩD

where ΩD indicates a set of all terms appearing in the database D. If a set of relevant documents is unknown, P (t|R) can be approximately estimated by P (t|Ωq ) = P (t, Ωq )/P (Ωq ), and we can write: X Y P (t, Ωq ) = P (di )P (t|di ) P (t′ |di ), (22) t′ ∈Ωq

i:di ∈D

under a assumption [Lavrenko and Croft, 2001] 5 . Also, X P (Ωq ) = P (t, Ωq ).

(23)

t∈ΩD

5

P First, P (t, Ωq ) = di ∈D P (di )P (t, Ωq |di ). If we assume that all terms are sampled independently and Q identically from di , P (t, Ωq |di ) = P (t|di ) t′ ∈Ωq P (t′ |di ). By substituting it into the first equation, Equation (22) is obtained. Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Methods for Cross-Language Information Retrieval

269

In order to apply RM to CLIR, it is necessary to distinguish terms in documents from those in queries by denoting a document term by sk in Equations (21) and (22), i.e., sk ∈ ΩD . For example, Equation (22) is re-written as X Y P (sk , Ωq ) = P (di )P (sk |di ) P (t|di ), (24) t∈Ωq

i:di ∈D

Q

in which P (t|di ) can be estimated using Equation (19) and usually P (di ) = 1/N . While only query terms appearing in the document contribute to computation of the document score in standard IR models including LM, RM takes all terms in the database D into consideration as suggested by Equation (21). This means that much more information on term occurrences is used in RM than other IR models, which may enhance search performance. However, more computational time is required for calculating document scores based on RM.

5.3.

Hidden Markov Model for CLIR

Suppose that s is a sequence of terms included in a source query and t is its translation, and that the number of source terms is equal to that of translations (e.g., m = l in Equation (7)). From the definition of conditional probability, we can write Pr(t, s) such that Pr(t, s) = Pr(t1 , ..., tm , s1 , ..., sm ) = Pr(s1 , ..., sm |t1 , ..., tm )Pr(t1 , ..., tm ),

(25)

where Pr(t1 , ..., tm ) = Pr(tm |tm−1 , ..., t1 ) × ... × Pr(t3 |t2 , t1 ) × Pr(t2 |t1 ) × Pr(t1 ), and if the translation process is interpreted as a Markov model,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Pr(t1 , ..., tm ) = P (tm |tm−1 )...P (t3 |t2 )P (t2 |t1 )P (t1 ).

(26)

Similarly, Pr(s1 , ..., sm |t1 , ..., tm ) = P (sm |tm ) × ... × P (s2 |t2 ) × P (s1 |t1 ), and therefore, Pr(t, s) = P (t1 )P (s1 |t1 )

m Y

P (sj |tj )P (tj |tj−1 ).

(27)

j=2

This is a query translation model based on HMM (Hidden Markov Model) proposed by [Federico and Bertoldi, 2002, Bertoldi and Federico, 2004], in which ‘reverse’ translation probability P (sj |tj ) and transitive probability P (tj |tj−1 ) of translations (target terms) are included. The transitive probability in this model may work as a device for translation disambiguation. For example, P (“management”|“database”) × P (s2 |“management”) is expected to be greater than P (“administration”|“database”) × P (s2 |“administration”) where s2 is the second term of a query representing “database management system” in a language other than English.

5.4.

Structured Query Model

The INQUERY system developed by a research group at Massachusetts University [Callan et al., 1995a, Turtle and Croft, 1991] provides an important function for CLIR. In Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

270

Kazuaki Kishida

principle, INQUERY is a retrieval model based on a probabilistic Bayesian network, in which a ‘belief score’ B(q|di ) measuring the degree to which a document di is relevant to a given query q is estimated. Although some variations can be derived dependently on assumptions to be selected, B(q|di ) is basically computed from B(t|di ) of each search term included in q, i.e.,   log[(N + 0.5)/nt ] xi (t) × , (28) B(t|di ) = 0.4 + 0.6 log(N + 1) xi (t) + 0.5 + 1.5li /¯l

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

where li is the length of di , ¯l is the average of li in the database and xi (t) is the frequency of t in the document di . For example, Pmwhen ‘#sum( )’ operator is used such as #sum(t1 , t2 , ..., tm ), B(q|di ) is calculated as j=1 B(tj |di ). INQUERY has several operators like #sum( ), which clearly distinguish this system from other retrieval models. Among the operators, the #syn( ) operator for dealing with synonyms is often used in CLIR tasks. Suppose that there are three source terms s1 , s2 , s3 , and translations of these terms are obtained from a bilingual dictionary such that T1 = {t11 , t12 , t13 }, T2 = {t21 , t22 } and T3 = {t31 , t32 }. We can enter a structured query, #sum ( #syn (t11 t12 t13 ) #syn (t21 t22 ) #syn (t31 t32 )), into the INQUERY system without any translation disambiguation [Pirkola, 1998]. This is often called Pirkola’s method, in which the belief score of a set of translations Tk for the source term sk is computed based on the #syn operator such that   x′i (Tk ) log[(N + 0.5)/n′ (Tk )] B(Tk |di ) = 0.4 + 0.6 × , (29) log(N + 1) x′i (Tk ) + 0.5 + 1.5li /¯l P ′ where x′i (Tk ) = t∈Tk xi (t) and n (Tk ) indicates the number of documents including at least one translation in Tk [Kek¨al¨ainen and J¨arveline, 1998, Sperer and Oard, 2000, Lehtokangas et al., 2004, Hedlund et al., 2004]. Pirkola’s method has been slightly modified by some studies [Darwish and Oard, 2003, Wang and Oard, 2006]. For example, in an experiment [Darwish and Oard, 2003], translation probabilities are incorporated into P ′ (T ) and n′ (T ) such as x′ (T ) = the computation of x k k k i i t∈Tk P (t|sk ) × xi (t) and P n′ (Tk ) = t∈Tk P (t|sk ) × nt in order to increase infulence of probable target terms with higher translation probabilities.

6.

Method for Multilingual Information Retrieval

6.1.

Approaches to MLIR

Suppose that we have a multilingual document collection in which two or more languages are mixed (not a parallel corpus), and that a user wishes to search the collection for a query expressed in a single language (see Figure 1). This MLIR task is more complicated than simple BLIR. Basically, there are two strategies for MLIR as follows [Lin and Chen, 2003]. • Distributed architecture in which the document collection is separated by language, and each part is indexed and retrieved independently.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Methods for Cross-Language Information Retrieval

271

• Centralized architecture in which the document collection in various languages is viewed as a single document collection and is indexed in one huge index file.

indexing

search query

multilingual collection

index

result

translation

merge final list

Figure 9. Distributed architecture for MLIR.

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

6.2.

Merging Technique

In the distributed architecture, a standard bilingual search is repeatedly performed for each separate language sub-collection respectively, and several ranked document lists are generated by each run (see Figure 9). Then the problem becomes how to merge the results of each run into a single ranked list so that all relevant documents in any language are successfully located on upper ranks. Essentially, the merging strategy is a general research topic of IR when searching distributed resources (i.e., distributed IR), in which it is necessary to merge ranked lists obtained from each resource. In CLIR, the following merging strategies have been investigated 6 . • Raw score: straightforwardly using document scores estimated in each run • Round robin: interleaving each document list in a round robin fashion by assuming that the distribution of relevant documents is identical among the lists • Normalized score: normalizing document scores of each run in order to remove the effects of collection-dependent statistics used for estimating the scores • Rank-based score: mathematically converting ranks in each run into scores by assuming a relationship between probabilities of relevance and the ranks • Modified score: modifying raw scores in each run so as to reduce the effects of collection-size dependency, translation ambiguity, and so on If the retrieval model employed for each run can estimate the relevance probability of each document correctly, it would be reasonable to re-rank all documents together according 6

Experimental comparison of the search performance of these methods has been attempted by [Savoy, 2004]. Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

272

Kazuaki Kishida

to values of the probability (i.e., raw score strategy). For example, we can use the logistic regression model for IR as such a retrieval model (see [Chen and Gey, 2003] for details). However, in most cases, it would be too difficult to interpret each document score as a pure probability of relevance even though a probabilistic retrieval model was actually used because the probability is often approximately estimated for convenience of calculation in the model. In such cases, if it can be assumed that relevant documents are distributed in the same way within every separate language sub-collection, we can employ round robin-based merging in which only the rank of each document is taken into account. Otherwise, an alternative method is to use normalized document scores such that v˜i = (vi − vmin )/(vmax − vmin ),

(30)

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

where vi is the raw score of document di , and vmin and vmax are the minimum and maximum in each search run respectively [Powell et al., 2000]. One widely-known method for normalization in distributed IR is the CORI (collection retrieval inference) algorithm [Callan et al., 1995b]. In this algorithm, first, the score of a sub-collection for the given query is computed based on Equation (28) by considering each sub-collection Dk (k = 1, ..., L) as a single huge document where L is the number of sub-collections included in the multilingual collection (i.e., the number of languages). We denote the score by B(q|Dk ). Second, the score vi (= B(q|di )) of a document di in a sub-collection Dk (i.e., di ∈ Dk ) is converted such that  ¯  B(q|Dk ) − B(q) , (31) v˜i = vi × 1 + L × ¯ B(q) P ¯ where B(q) = L−1 L k=1 B(q|Dk ). For more details on applying the CORI algorithm to MLIR, see [Savoy, 2002] or [Moulinier and Molina-Salgado, 2003]. Another possibility is to predict the relevance probability of a document ranked in a position using training data sets [Franz et al., 1999, Kraaij et al., 2000, Savoy, 2003]. For example, we may calculate this probability such that Pˆ (R|di ) = a + b log(ρi )

(32)

where P (R|di ) indicates the relevance probability of document di , ρi is its rank in the output list, and a and b are parameters to be estimated from training data (a more complicated regression model was used in [Savoy, 2003]). Meanwhile, researchers have explored other techniques of modifying raw scores so as to remove the effect of collection-size dependency (e.g., [Hiemstra et al., 2001]), or of reducing them based on the degree of translation ambiguity according to the assumption that a good translation may give much more relevant documents (e.g., [Lin and Chen, 2003]).

6.3.

Searching Heterogeneous Collections

On the other hand, in the centralized architecture, the set of multilingual documents is not divided into sub-collections for each language. In order to search such a heterogeneous collection we need either Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Methods for Cross-Language Information Retrieval

273

1. to translate a given source query into all languages included in the document collection and to merge all translations into a single query, or 2. to translate all documents in a single language used in the query. In general, the first method has been adopted (e.g., [Gey et al., 1999, Nie and Jin, 2003]). Since documents in a language having fewer documents may take advantage of weighting by document frequency, it may be necessary in this method to adjust the idf factor [Lin and Chen, 2003]. For example, in Equation (28), {log[(N + 0.5)/nt ]}/ log(N + 1) is an idf factor, and its value becomes larger as nt is smaller. Meanwhile, in the second method, it is not necessary to add such adjustment of parameters because a single index registering query language words is created although the document translation is a very time-consuming task. This method was attempted by [Chen and Gey, 2004], and the search performance of various MLIR techniques discussed in this section was experimentally compared.

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

7.

Concluding Remarks

As is widely recognized, research on developing CLIR techniques can be traced back to G. Salton’s paper in 1970 [Salton, 1970]. Since then, many research results have been published in various journals and conferences in different fields such as IR and NLP, especially since the mid-1990s. This review has not covered all the techniques reported in such publications, and excludes studies focusing on only a particular language (e.g., Chinese, Japanese, etc.). Other review articles [Oard and Hackett, 1998, Peters and Sheridan, 2001, Kishida, 2005] provide further useful information on CLIR research efforts. The rapid development of CLIR techniques since the mid-1990s is largely due to CLIR experiment workshops such as TREC [Harman, 2005], CLEF [Braschler and Peters, 2004], and NTCIR [Kishida et al., 2007], in which research programs on CLIR have been incorporated and many researchers worldwide have participated. Furthermore, test collections for CLIR constructed in these projects allow a wider range of researchers including those not participating in the projects to test new ideas on CLIR techniques. New applications on the Internet have been continually appearing, and accordingly, IR techniques tailored to them have been required (e.g., blog searching). CLIR techniques will need to be developed in parallel in order to satisfy users’ emerging and diverse needs.

Appendix: IBM Model 1 Let a = am 1 ≡ a1 a2 . . . am be an alignment, where aj indicates mapping of a target word to one or more source words. For example, aj = k means that the word in position j of the target string is connected to the word in position k of the source string. The IBM Model [Brown et al., 1993] is Pr(t|s) =

X

Pr(t, a|s),

a

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

(33)

274

Kazuaki Kishida

where Pr(t, a|s) = Pr(m|s)

m Y

j−1 j j−1 Pr(aj |aj−1 1 , t1 , m, s)Pr(tj |a1 , t1 , m, s).

(34)

j=1

In the simplest Model 1, we put the following assumptions: • Pr(m|s) is independent of s j−1 −1 • Pr(aj |aj−1 1 , t1 , m, s) depends only on l, and must be (l + 1)

• Pr(tj |aj1 , tj−1 depends only 1 , m, s) j j−1 Pr(tj |a1 , t1 , m, s) = P (tj |saj )

on

tj

saj ,

and

and

therefor,

Hence, Equation (34) becomes Pr(t, a|s) =

m Y ε P (tj |saj ), (l + 1)m

(35)

j=1

where ε ≡ Pr(m|s). By substituting it into Equation (33), we have l l m X X Y ε Pr(t|s) = ··· P (tj |saj ), (l + 1)m

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

a1 =0

(36)

am =0 j=1

where aj = 0 means that there is no corresponding term in the source string. Since these sums are interchangeable with the product in Equation (36), Equation (7) can be obtained. The probability P (t|s) can P be maximized by introducing the Lagrange multipliers λs . That is, under the condition t P (t|s) = 1, we set m X l Y X X ε P (t|s) − 1), P (t |s ) − λ ( H(P, λ) = j s k (l + 1)m s t

(37)

j=1 k=0

and ∂H(P, λ) ∂P (t|s)

=

Qm Pl m X l X P (th |su ) ε δ(t, tj )δ(s, sk ) h=1 − λs Pl u=0 (l + 1)m u=0 P (t|su ) j=1 k=0

=

ε (l + 1)m

Qm Pl

u=0 P (th |eu )

h=1

Pl

u=0 P (t|su )

m X

δ(t, tj )

j=1

l X

δ(s, sk ) − λs ,

k=0

where δ(·, ·) is the Kronecker’s delta. From the equation ∂H(P, λ)/∂P (t|s) = 0, it follows that P (t|s) = λ−1 s

m X l m l Y X X ε P (t|s) P (t |s ) δ(t, t ) δ(s, si ) P j h u l (l + 1)m u=0 P (t|su ) h=1 u=0

= λ−1 s Pr(t|s) Pl

P (t|s)

j=1

m X

u=0 P (t|su ) j=1

δ(t, tj )

l X

i=0

δ(s, sk ).

k=0

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

(38)

Methods for Cross-Language Information Retrieval

275

If we replace λs by λs Pr(t|s), and introduce the expected number of times that s connects to t in the translation (t|s) as m

l

j=1

k=0

X X P (t|s) c(t|s; t, s) = δ(t, tj ) δ(s, sk ), P (t|s0 ) + · · · + P (t|sl )

(39)

then the Equation (38) can be written compactly as P (t|s) = λ−1 s c(t|s; t, s)

(40)

In practice, we have N translations, (t(1) |s(1) ), ..., (t(N ) |s(N ) ) as training data, so this equation becomes N X P (t|s) = λ−1 c(t|s; t(i) , s(i) ). (41) s i=1

We can estimate P (t|s) iteratively using Equation (39) and (41) as follows. 1. Choose initial values for P (t|s). 2. Compute Equation (39) for each sentence (1 ≤ i ≤ N ). 3. For each s, estimate λs as λs =

N XX t

c(t|s; t(i) , s(i) )

(42)

i=1

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

4. For each t, use Equation (41) to obtain a new value for P (t|s) 5. Repeat steps 2 to 4 until the values of P (t|s) have converged to the desired degree.

References [Adriani, 2000] Adriani, M. (2000). Using statistical term similarity for sense disambiguation in cross-language information retrieval. Information Retrieval, 2(1):71–82. [Adriani, 2002] Adriani, M. (2002). English-Dutch CLIR using query translation techniques. In et al., C. P., editor, Evaluation of Cross-Language Information Retrieval Systems, pages 219–225. Springer-Verlag. (LNCS 2406). [Ballesteros and Croft, 1997] Ballesteros, L. and Croft, W. B. (1997). Phrasal translation and query expansion techniques for cross-language information retrieval. In Proceedings of the 20th Annual International ACM SIGIR Conference on Research and Development in Information Retrieval, pages 84–91. [Ballesteros and Croft, 1998] Ballesteros, L. and Croft, W. B. (1998). Resolving ambiguity for cross-language retrieval. In Proceedings of the 21st Annual International ACM SIGIR Conference on Research and Development in Information Retrieval, pages 64–71.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

276

Kazuaki Kishida

[Ballesteros, 2000] Ballesteros, L. A. (2000). Cross-language retrieval via transitive translation. In W. B. Croft et al., editor, Advances in Information Retrieval: Recent Research from the Center for Intelligent Information Retrieval, pages 203–234. Kluwer. [Berry and Young, 1995] Berry, M. W. and Young, P. G. (1995). Using latent semantic indexing for multilanguage information retrieval. Computers and the Humanities, 29(6):413–429. [Bertoldi and Federico, 2004] Bertoldi, N. and Federico, M. (2004). Statistical models for monolingual and bilingual information retrieval. Information Retrieval, 7(1-2):53–72. [Bian and Chen, 1998] Bian, G. W. and Chen, H. H. (1998). Integrating query translation and document translation in a cross-language information retrieval system. In D. Farwell et al., editor, Machine Translation and the Information Soup, pages 250–265. Springer. (LNCS 1529). [Bian and Lin, 2001] Bian, G. W. and Lin, C. C. (2001). Trans-EZ at NTCIR-2 : synset cooccurrence method for English-Chinese cross-lingual information retrieval. In Proceedings of the Second NTCIR Workshop on Research in Chinese & Japanese Text Retrieval and Text Summarization. Available from: http://research.nii.ac.jp/ntcir/workshop/.

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

[Boughanem et al., 2002] Boughanem, M., Chrisment, C., and Nassr, N. (2002). Investigation on disambiguation in CLIR: aligned corpus and bi-directional translation-based strategies. In et al., C. P., editor, Evaluation of Cross-Language Information Retrieval Systems, pages 158–168. Springer-Verlag. (LNCS 2406). [Braschler, 2004] Braschler, M. (2004). Combination approaches for multilingual text retrieval. Information Retrieval, 7(1-2):183–204. [Braschler and Peters, 2004] Braschler, M. and Peters, C. (2004). Cross-Language Evaluation Forum: objectives, results, achievements. Information Retrieval, 7(1-2):7–31. [Braschler and Sch¨auble, 2000] Braschler, M. and Sch¨auble, P. (2000). Using corpusbased approaches in a system for multilingual information retrieval. Information Retrieval, 3(3):273–284. [Braschler and Sch¨auble, 2001] Braschler, M. and Sch¨auble, P. (2001). Experiments with the Eurospider retrieval system for CLEF 2000. In Peters, C., editor, Cross-Language Information Retrieval Evaluation, pages 140–148. Springer-Verlag. (LNCS 2069). [Brown et al., 1993] Brown, P. F., Della Pietra, S. A., Della Pietra, V. J., and Mercer, R. L. (1993). The mathematics of statistical machine translation: parameter estimation. Computational Linguistics, 19(2):263–311. [Buckley et al., 1998] Buckley, C., Walz, J., Mitra, M., and Cardie, C. (1998). Using clustering and superconcepts within SMART: TREC-6. In Proceedings of TREC-6. National Institute of Standards and Technology. Available from: http://trec.nist.gov/pubs.html. Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Methods for Cross-Language Information Retrieval

277

[Callan et al., 1995a] Callan, J. P., Croft, W. B., and Broglio, J. (1995a). TREC and TIPSTER experiments with INQUERY. Information Processing & Management, 31(3):327– 343. [Callan et al., 1995b] Callan, J. P., Lu, Z., and Croft, W. B. (1995b). Searching distributed collections with inference networks. In Proceedings of the 18th Annual International ACM SIGIR Conference on Research and Development in Information Retrieval, pages 21–28. [Cao and Li, 2002] Cao, Y. and Li, H. (2002). Base noun phrase translation using web data and the EM algorithm. In Proceedings of the 19th International Conference on Computational Linguistics, volume 1, pages 1–7. [Chen and Gey, 2003] Chen, A. and Gey, F. C. (2003). Experiments on cross-language and patent retrieval at NTCIR-3 workshop. In Proceedings of the Third NTCIR Workshop on Research in Information Retrieval, Automatic Text Summarization and Question Answering. Available from: http://research.nii.ac.jp/ntcir/. [Chen and Gey, 2004] Chen, A. and Gey, F. C. (2004). Multilingual information retrieval using machine translation, relevance feedback and decompounding. Information Retrieval, 7(1-2):149–182.

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

[Cheng et al., 2004] Cheng, P.-J., Teng, J.-W., Chen, R.-C., Wang, J.-H., Lu, W.-H., and Chien, L.-F. (2004). Translating unknown queries with Web corpora for cross-language information retrieval. In Proceedings of the 27th Annual International ACM SIGIR Conference on Research and Development in Information Retrieval, pages 146–153. [Church and Hanks, 1990] Church, K. W. and Hanks, P. (1990). Word association norms, mutual information, and lexicography. Computational Linguistics, 16(1):22–29. [Darwish and Oard, 2003] Darwish, K. and Oard, D. W. (2003). Probabilistic structured query methods. In Proceedings of the 26th Annual International ACM SIGIR Conference on Research and Development in Information Retrieval, pages 338–344. [Davis, 1997] Davis, M. (1997). New experiments in cross-language text retrieval at NMSU’s Computing Research Lab. In Proceedings of TREC-5. National Institute of Standards and Technology. Available from: http://trec.nist.gov/pubs.html. [Davis and Dunning, 1995] Davis, M. and Dunning, T. (1995). A TREC evaluation of query translation methods for multi-lingual text retrieval. In Proceedings of TREC-4. National Institute of Standards and Technology. Available from: http://trec.nist.gov/pubs.html. [Davis, 1998] Davis, M. W. (1998). On the effective use of large parallel corpora in crosslanguage text retrieval. In Grefenstette, G., editor, Cross-Language Information Retrieval, pages 12–22. Kluwer. [Davis and Ogden, 1997] Davis, M. W. and Ogden, W. C. (1997). QUILT: implementing a large-scale cross-language text retrieval system. In Proceedings of the 20th Annual Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

278

Kazuaki Kishida

International ACM SIGIR Conference on Research and Development in Information Retrieval, pages 92–98. [Deerwester et al., 1990] Deerwester, S., Dumais, S. T., Furnas, G. W., Landauer, T. K., and Harshman, R. (1990). Indexing by latent semantic analysis. Journal of the American Society for Information Science, 41(6):391–407. [Diekema et al., 1999] Diekema, A., Oroumchian, F., Sheridan, P., and Liddy, E. D. (1999). TREC7 evaluation of conceptual interlingua document retrieval (CINDOR) in English and French. In Proceedings of TREC-7. National Institute of Standards and Technology. Available from: http://trec.nist.gov/pubs.html. [Dumais et al., 1996] Dumais, S. T., Landauer, T. K., and Littman, M. L. (1996). Automatic cross-linguistic information retrieval using latent semantic indexing. In Proceedings of the 19th Annual International ACM SIGIR Conference on Research and Development in Information Retrieval, pages 16–23. [Dunning, 1993] Dunning, T. (1993). Accurate methods for the statistics of surprise and coincidence. Computational Linguistics, 19(1):61–74.

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

[Eichmann et al., 1998] Eichmann, D., Ruiz, M. E., and Srinivasan, P. (1998). Crosslanguage information retrieval with the UMLS metathesaurus. In Proceedings of the 21st Annual International ACM SIGIR Conference on Research and Development in Information Retrieval, pages 72–80. [Federico and Bertoldi, 2002] Federico, M. and Bertoldi, N. (2002). Statistical crosslanguage information retrieval using N-best query translations. In Proceedings of the 25th Annual International ACM SIGIR Conference on Research and Development in Information Retrieval, pages 167–174. [Franz et al., 1999] Franz, M., Scott McCarley, J., and Roukos, S. (1999). Ad hoc and multilingual information retrieval at IBM. In Proceedings of TREC-7. National Institute of Standards and Technology. Available from: http://trec.nist.gov/pubs.html. [Franz et al., 2000] Franz, M., Scott McCarley, J., and Todd Ward, R. (2000). Ad hoc, cross-language and spoken document information retrieval at IBM. In Proceedings of TREC-8. National Institute of Standards and Technology. Available from: http://trec.nist.gov/pubs.html. [Fujii and Ishikawa, 2001] Fujii, A. and Ishikawa, T. (2001). Japanese/English crosslanguage information retrieval: exploration of query translation and transliteration. Computers and the Humanities, 35:389–420. [Galvez and Moya-Aneg´on, 2007] Galvez, C. and Moya-Aneg´on, F. (2007). Approximate personal name-matching through finite-state graphs. Journal of the American Society for Information Science and Technology, 58(13):1960–1976. Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Methods for Cross-Language Information Retrieval

279

[Gao and Nie, 2006] Gao, J. and Nie, J.-Y. (2006). A study of statistical models for query translation: finding a good unit of translation. In Proceedings of the 29th Annual International ACM SIGIR Conference on Research and Development in Information Retrieval, pages 194–201. [Gao et al., 2001] Gao, J., Nie, J.-Y., Zhang, J., Xun, E., Su, Y., Zhou, M., and Huang, C. (2001). Improving query translation for cross-language information retrieval using statistical models. In Proceedings of the 24th Annual International ACM SIGIR Conference on Research and Development in Information Retrieval, pages 96–104. [Gey, 2001] Gey, F. C. (2001). Research to improve cross-language retrieval: position paper for CLEF. In Peters, C., editor, Cross-Language Information Retrieval Evaluation, pages 83–88. Springer-Verlag. (LNCS, 2069). [Gey et al., 1999] Gey, F. C., Jiang, H., Chen, A., and Larson, R. R. (1999). Manual queries and machine translation in cross-language retrieval and interactive retrieval with Cheshire II at TREC-7. In Proceedings of TREC-7. National Institute of Standards and Technology. Available from: http://trec.nist.gov/pubs.html. [Gollins and Sanderson, 2001] Gollins, T. and Sanderson, M. (2001). Improving cross language information retrieval with triangulated translation. In Proceedings of the 24th Annual International ACM SIGIR Conference on Research and Development in Information Retrieval, pages 90–95.

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

[Harman, 2005] Harman, D. (2005). Beyond English. In Voorhees, E. M. and Harman, D. K., editors, TREC: Experiment and Evaluation in Information Retrieval, pages 153– 181. MIT Press. [Hedlund et al., 2004] Hedlund, T., Airio, E., Keskustalo, H., Lehtokangas, R., Pirkola, A., and J¨arveline, K. (2004). Dictionary-based cross-language information retrieval: learning experiences from CLEF 2000-2002. Information Retrieval, 7(1-2):99–119. [Hedlund et al., 2002] Hedlund, T., Keskustalo, H., Pirkola, A., Airio, E., and J¨arvelin, K. (2002). Utaclir@CLEF2001: effects of compound splitting and n-gram techniques. In et al., C. P., editor, Evaluation of Cross-Language Information Retrieval Systems, pages 118–136. Springer-Verlag. (LNCS 2406). [Hiemstra, 1998a] Hiemstra, D. (1998a). A linguistically motivated probabilistic model of information retrieval. In Nikolaou, C. and Stephanidis, C., editors, Research and Advanced Technology for Digital Libraries (LNCS 1513), pages 569–584. Springer. [Hiemstra, 1998b] Hiemstra, D. (1998b). Multilingual domain modeling in Twenty-One: automatic creation of a bi-directional translation lexicon from a parallel corpus. In Proceedings of the 8th CLIN Meeting, pages 41–58. [Hiemstra and Kraaij, 1999] Hiemstra, D. and Kraaij, W. (1999). Twenty-one at TREC7: ad-hoc and cross-language track. In Proceedings of TREC-7. National Institute of Standards and Technology. Available from: http://trec.nist.gov/pubs.html. Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

280

Kazuaki Kishida

[Hiemstra et al., 2001] Hiemstra, D., Kraaij, W., Pohlmann, R., and Westerveld, T. (2001). Translation resources, merging strategies, and relevance feedback for cross-language information retrieval. In Peters, C., editor, Cross-Language Information Retrieval Evaluation, pages 102–115. Springer-Verlag. (LNCS 2069). [Hull and Grefenstette, 1996] Hull, D. A. and Grefenstette, G. (1996). Querying across languages: a dictionary-based approach to multilingual information retrieval. In Proceedings of the 19th Annual International ACM SIGIR Conference on Research and Development in Information Retrieval, pages 49–57. [Ide and V´eronis, 1998] Ide, N. and V´eronis, J. (1998). Introduction to the special issue on word sense disambiguation: the state of the art. Computational Linguistics, 24(1):1–40. [Jang et al., 1999] Jang, M.-G., Myaeng, S. H., and Park, S. Y. (1999). Using mutual information to resolve query translation ambiguities and query term weighting. In Proceedings of the 37th Annual Conference of the Association of Computational Linguistics, pages 223–229. [Jones and Lam-Adesina, 2002] Jones, G. J. F. and Lam-Adesina, A. M. (2002). Experiments with machine translation for bilingual retrieval. In et al., C. P., editor, Evaluation of Cross-Language Information Retrieval Systems, pages 59–77. Springer-Verlag. (LNCS 2406).

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

[Kang et al., 2005] Kang, I.-S., Na, S.-H., and Lee, J.-H. (2005). Combination approaches in information retrieval: words vs. n-grams and query translation vs. document translation. In Proceedings of the Fourth NTCIR Workshop on Research in Information Access Technologies Information Retrieval, Question Answering and Summarization. Available from: http://research.nii.ac.jp/ntcir/workshop/OnlineProceedings4/. [Kek¨al¨ainen and J¨arveline, 1998] Kek¨al¨ainen, J. and J¨arveline, K. (1998). The impact of query structure and query expansion on retrieval performance. In Proceedings of the 21st Annual International ACM SIGIR Conference on Research and Development in Information Retrieval, pages 130–137. [Kishida, 2005] Kishida, K. (2005). Technical issues of cross-language information retrieval: a review. Information Processing & Management, 41(3):433–455. [Kishida, 2007] Kishida, K. (2007). Term disambiguation techniques based on target document collection for cross-language information retrieval: an empirical comparison of performance between techniques. Information Processing & Management, 43(1):103– 120. [Kishida, 2008] Kishida, K. (2008). Prediction of performance of cross-language information retrieval using automatic evaluation of translation. Library and Information Science Research, 30(2):138–144. [Kishida et al., 2007] Kishida, K., Chen, K.-H., Lee, S., Kuriyama, K., Kando, N., and Chen, H.-H. (2007). Overview of CLIR task at the Sixth NTCIR Workshop. In Proceedings of the 6th NTCIR Workshop Meeting on Evaluation of Information Access TechBilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Methods for Cross-Language Information Retrieval

281

nologies: Information Retrieval, Question Answering and Cross-Lingual Information Access. Available from: http://research.nii.ac.jp/ntcir/workshop/OnlineProceedings6/. [Kishida and Kando, 2004] Kishida, K. and Kando, N. (2004). Two-stage refinement of query translation in a pivot language approach to cross-lingual information retrieval: an experiment at CLEF 2003. In C. Peters et al., editor, Comparative Evaluation of Multilingual Information Access Systems, pages 253–262. Springer-Verlag. (LNCS 3237). [Kishida and Kando, 2006] Kishida, K. and Kando, N. (2006). A hybrid approach to query and document translation using a pivot language for cross-language information retrieval. In C. Peters et al., editor, Accessing Multilingual Information Repositories, pages 93–101. Springer-Verlag. (LNCS 4022). [Kishida et al., 2005] Kishida, K., Kando, N., and Chen, K. H. (2005). Two-stage refinement of transitive query translation with english disambiguation for cross-language information retrieval: an experiment at CLEF2004. In C. Peters et al., editor, Multilingual Information Access for Text, Speech and Images, Lecture Notes in Computer Science, pages 135–142. Springer-Verlag. (LNCS 3491). [Knight and Graehl, 1998] Knight, K. and Graehl, J. (1998). Machine transliteration. Computational Linguistics, 24(4):599–612.

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

[Koehn and Knight, 2000] Koehn, P. and Knight, K. (2000). Estimating word translation probabilities from unrelated monolingual corpora using the EM algorithm. In Proceedings of the Seventeenth National Conference on Artificial Intelligence and Twelfth Conference on Innovative Applications of Artificial Intelligence, pages 711–715. [Kraaij, 2002] Kraaij, W. (2002). TNO at CLEF-2001: comparing translation resources. In et al., C. P., editor, Evaluation of Cross-Language Information Retrieval Systems, pages 78–93. Springer-Verlag. (LNCS 2406). [Kraaij et al., 2000] Kraaij, W., Pohlmann, R., and Hiemstra, D. (2000). Twentyone at TREC-8: using language technology for information retrieval. In Proceedings of TREC-8. National Institute of Standards and Technology. Available from: http://trec.nist.gov/pubs.html. [Kuo et al., 2008] Kuo, J., Li, H., and Yang, Y. (2008). Active learning for constructing transliteration lexicons from the Web. Journal of the American Society for Information Science and Technology, 59(1):126–135. [Landauer and Littman, 1990] Landauer, T. and Littman, M. L. (1990). A statistical method for language-independent representation of the topical content of text segments. In Proceedings of the 6th Annual Conference of the UW Centre for the New Oxford English Dictionary and Text Research. Available from: http://www.cs.duke.edu/˜mlittman/docs/x-lang.ps. [Landauer et al., 2007] Landauer, T. K., McNamara, D. S., Dennis, S., and Kintsch, W., editors (2007). Handbook of Latent Semantic Analysis. Lawrence Erlbaum Associates. Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

282

Kazuaki Kishida

[Larkey and Connell, 2005] Larkey, L. S. and Connell, M. E. (2005). Structured queries, language modeling, and relevance modeling in cross-language information retrieval. Information Processing & Management, 41(3):457–473. [Lavrenko et al., 2002] Lavrenko, V., Choquette, M., and Croft, W. B. (2002). Crosslingual relevance models. In Proceedings of the 25th Annual International ACM SIGIR Conference on Research and Development in Information Retrieval, pages 175–182. [Lavrenko and Croft, 2001] Lavrenko, V. and Croft, W. B. (2001). Relevance-based language models. In Proceedings of the 24th Annual International ACM SIGIR Conference on Research and Development in Information Retrieval, pages 120–127. [Lee et al., 2004] Lee, K.-S., Kageura, K., and Choi, K.-S. (2004). Implicit ambiguity resolution using incremental clustering in cross-language information retrieval. Information Processing & Management, 40(1):145–159. [Lehtokangas et al., 2004] Lehtokangas, R., Airio, E., and J¨arveline, K. (2004). Transitive dictionary translation challenges direct dictionary translation in CLIR. Information Processing & Management, 40(6):973–988. [Levow et al., 2005] Levow, G.-A., Oard, D. W., and Resnik, P. (2005). Dictionary-based techniques for cross-language information retrieval. Information Processing & Management, 41(3):523–547.

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

[Lin and Chen, 2003] Lin, W. C. and Chen, H. H. (2003). Description of NTU approach to NTCIR3 multilingual information retrieval. In Proceedings of the Third NTCIR Workshop on Research in Information Retrieval, Automatic Text Summarization and Question Answering. Available from: http://research.nii.ac.jp/ntcir/. [Littman et al., 1998] Littman, M. L., Dumais, S., and Landauer, T. K. (1998). Automatic cross-language information retrieval using latent semantic indexing. In Grefenstette, G., editor, Cross-Language Information Retrieval, pages 51–62. Kluwer. [Liu et al., 2005] Liu, Y., Jin, R., and Chai, J. Y. (2005). A maximum coherence model for dictionary-based cross-language information retrieval. In Proceedings of the 28th Annual International ACM SIGIR Conference on Research and Development in Information Retrieval, pages 536–543. [Lopez-Ostenero et al., 2005] Lopez-Ostenero, F., Gonzalo, J., and Verdejo, F. (2005). Noun phrases as building blocks for cross-language search assistance. Information Processing & Management, 41(3):549–568. [Maeda et al., 2000] Maeda, A., Sadat, F., Yoshikawa, M., and Uemura, S. (2000). Query term disambiguation for Web cross-language information retrieval using a search engine. In Proceedings of the 5th International Workshop on Information Retrieval with Asian languages, pages 25–32. [Manning et al., 2008] Manning, C. D., Raghavan, P., and Sch¨utze, H. (2008). Introduction to Information Retrieval. Cambridge. Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Methods for Cross-Language Information Retrieval

283

[Markey, 2007] Markey, K. (2007). Twenty-five years of end-user searching, part 1: research findings. Journal of the American Society for Information Science and Technology, 58(8):1071–1081. [Mark´o et al., 2005] Mark´o, K., Schulz, S., Medelyan, O., and Hahn, U. (2005). Bootstrapping dictionaries for cross-language information retrieval. In Proceedings of the 28th Annual International ACM SIGIR Conference on Research and Development in Information Retrieval, pages 528–535. [McNamee and Mayfield, 2002a] McNamee, P. and Mayfield, J. (2002a). Comparing cross-language query expansion techniques by degrading translation resources. In Proceedings of the 25th Annual International ACM SIGIR Conference on Research and Development in Information Retrieval, pages 159–166. [McNamee and Mayfield, 2002b] McNamee, P. and Mayfield, J. (2002b). JHU/APL experiments at CLEF: translation resources and score normalization. In et al., C. P., editor, Evaluation of Cross-Language Information Retrieval Systems, pages 193–208. SpringerVerlag. (LNCS 2406). [McNamee and Mayfield, 2004] McNamee, P. and Mayfield, J. (2004). Character n-gram tokenization for European language text retrieval. Information Retrieval, 7(1-2):73–97.

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

[Melamed, 2000] Melamed, I. D. (2000). Models of translational equivalence among words. Computational Linguistics, 26(2):221–249. [Miller et al., 1999] Miller, D. R. H., Leek, T., and Schwartz, R. M. (1999). A hidden Markov model information retrieval system. In Proceedings of the 22nd Annual International ACM SIGIR Conference on Research and Development in Information Retrieval, pages 214–221. [Monz and Dorr, 2005] Monz, C. and Dorr, B. J. (2005). Iterative translation disambiguation for cross-language information retrieval. In Proceedings of the 28th Annual International ACM SIGIR Conference on Research and Development in Information Retrieval, pages 520–527. [Moulinier and Molina-Salgado, 2003] Moulinier, I. and Molina-Salgado, H. (2003). Thomson Legal and Regulatory experiments for CLEF2002. In et al., C. P., editor, Advances in Cross-Language Information Retrieval, pages 155–163. Springer-Verlag. (LNCS 2785). [Nie, 2000] Nie, J.-Y. (2000). CLIR using a probabilistic translation model based on Web documents. In Proceedings of TREC-8. National Institute of Standards and Technology. Available from: http://trec.nist.gov/pubs.html. [Nie and Jin, 2003] Nie, J.-Y. and Jin, F. (2003). A multilingual approach to multilingual information retrieval. In et al., C. P., editor, Advances in Cross-Language Information Retrieval, pages 101–110. Springer-Verlag. (LNCS 2785). Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

284

Kazuaki Kishida

[Nie et al., 1999] Nie, J.-Y., Simard, M., Isabelle, P., and Durand, R. (1999). Crosslanguage information retrieval based on parallel texts and automatic mining of parallel texts from the Web. In Proceedings of the 22nd Annual International ACM SIGIR Conference on Research and Development in Information Retrieval, pages 74–81. [Oard and Diekema, 1998] Oard, D. W. and Diekema, A. R. (1998). Cross-language information retrieval. Annual Review of Information Science and Technology, 33:223–256. [Oard and Hackett, 1998] Oard, D. W. and Hackett, P. (1998). Document translation for cross-language text retrieval at the University of Maryland. In Proceedings of TREC-6. National Institute of Standards and Technology. Available from: http://trec.nist.gov/pubs.html. [Oard et al., 2001] Oard, D. W., Levow, G. A., and Cabezas, C. I. (2001). CLEF experiments at Maryland: statistical stemming and backoff translation. In Peters, C., editor, Cross-Language Information Retrieval Evaluation, pages 176–187. Springer-Verlag. (LNCS 2069). [Och and Ney, 2003] Och, F. J. and Ney, H. (2003). Systematic comparison of various statistical alignment models. Computational Linguistics, 29(1):19–51.

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

[Papineni et al., 2002] Papineni, K., Roukos, S., Ward, T., and Zhu, W.-J. (2002). BLEU: a method for automatic evaluation of machine translation. In Proceedings of the 40th Annual Meeting of the Association for Computational Linguistics, pages 311–318. ACL. [Peters and Sheridan, 2001] Peters, C. and Sheridan, P. (2001). Multilingual information access. In M. Agosti, et al., editor, Lectures on Information Retrieval, pages 51–80. Springer-Verlag. (LNCS 1980). [Pirkola, 1998] Pirkola, A. (1998). The effects of query structure and dictionary-based cross-language information retrieval. In Proceedings of the 21st Annual International ACM SIGIR Conference on Research and Development in Information Retrieval, pages 55–63. [Pirkola et al., 2001] Pirkola, A., Hedlund, T., Keskustalo, H., and J¨arvelin, K. (2001). Dictionary-based cross-language information retrieval: problems, methods, and research findings. Information Retrieval, 4(3-4):209–230. [Pirkola et al., 2003] Pirkola, A., Toivonen, J., Keskustalo, H., V., K., and J¨arvelin, K. (2003). Fuzzy translation of cross-lingual spelling variants. In Proceedings of the 26th Annual International ACM SIGIR Conference on Research and Development in Information Retrieval, pages 345–352. [Powell et al., 2000] Powell, A. L., French, J. C., Callan, J., Connell, M., and Viles, C. L. (2000). The impact of database selection on distributed searching. In Proceedings of the 23rd Annual International ACM SIGIR Conference on Research and Development in Information Retrieval, pages 232–239. Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Methods for Cross-Language Information Retrieval

285

[Qu et al., 2003a] Qu, Y., Grefenstette, G., and Evans, D. A. (2003a). Automatic transliteration for Japanese-to-English text retrieval. In Proceedings of the 26th Annual International ACM SIGIR Conference on Research and Development in Information Retrieval, pages 353–360. [Qu et al., 2003b] Qu, Y., Grefenstette, G., and Evans, D. A. (2003b). Resolving translation ambiguity using monolingual corpora. In et al., C. P., editor, Advances in CrossLanguage Information Retrieval, pages 223–241. Springer. (LNCS 2785). [Rehder et al., 1998] Rehder, B., Littman, M. L., Dumais, S., and Landauer, T. K. (1998). Automatic 3-language cross-language information retrieval with latent semantic indexing. In Proceedings of the TREC6. National Institute of Standards and Technology. Available from: http://trec.nist.gov/pubs.html. [Resnik, 1999] Resnik, P. (1999). Mining the web for bilingual text. In Proceedings of the 37th Annual Meeting of the Association for Computational Linguistics, pages 527–534. [Rogati and Yang, 2004] Rogati, M. and Yang, Y. (2004). Resource selection for domainspecific cross-lingual IR. In Proceedings of the 27th Annual International ACM SIGIR Conference on Research and Development in Information Retrieval, pages 154–161. [Salton, 1970] Salton, G. (1970). Automatic processing of foreign language documents. Journal of the American Society for Information Science, 21(3):187–194.

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

[Savoy, 2002] Savoy, J. (2002). Report on CLEF-2001 experiments: effective combined query-translation approach. In et al., C. P., editor, Evaluation of Cross-Language Information Retrieval Systems, pages 27–43. Springer-Verlag. (LNCS 2406). [Savoy, 2003] Savoy, J. (2003). Report on CLEF-2002 experiments: combining multiple sources of evidence. In et al., C. P., editor, Advances in Cross-Language Information Retrieval, pages 66–90. Springer-Verlag. (LNCS 2785). [Savoy, 2004] Savoy, J. (2004). Combining multiple strategies for effective monolingual and cross-language retrieval. Information Retrieval, 7(1-2):121–148. [Scott McCarley, 1999] Scott McCarley, J. (1999). Should we translate the documents or the queries in cross-language information retrieval? In Proceedings of the 37th Annual Meeting of the Association for Computational Linguistics, pages 208–214. [Seo et al., 2005] Seo, H.-C., Kim, S.-B., Rim, H.-C., and Myaeng, S.-H. (2005). Improving query translation in English-Korean cross-language information retrieval. Information Processing & Management, 41(3):507–522. [Sheridan and Ballerini, 1996] Sheridan, P. and Ballerini, J. P. (1996). Experiments in multilingual information retrieval using the SPIDER system. In Proceedings of the 19th Annual International ACM SIGIR Conference on Research and Development in Information Retrieval, pages 58–64. Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

286

Kazuaki Kishida

[Sperer and Oard, 2000] Sperer, R. and Oard, D. W. (2000). Structured translation for cross-language information retrieval. In Proceedings of the 23rd Annual International ACM SIGIR Conference on Research and Development in Information Retrieval, pages 120–127. [Talvensaari et al., 2006] Talvensaari, T., Laurikkala, J., J¨arveline, K., and Juhola, M. (2006). A study on automatic creation of a comparable document collection in crosslanguage information retrieval. Journal of Documentation, 62(3):372–387. [Toivonen et al., 2005] Toivonen, J., Pirkola, A., Keskustalo, H., Visala, K., and J¨arveline, K. (2005). Translating cross-lingual spelling variants using transformation rules. Information Processing & Management, 41(4):859–872. [Tsuji and Kageura, 2006] Tsuji, K. and Kageura, K. (2006). Automatic generation of Japanese-English bilingual thesauri based on bilingual corpora. Journal of the American Society for Information Science and Technology, 57(7):891–906. [Turtle and Croft, 1991] Turtle, H. R. and Croft, W. B. (1991). Evaluation of an inference network-based retrieval model. ACM Transactions on Information Systems, 9(3):187– 222.

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

[Wang and Oard, 2006] Wang, J. and Oard, D. W. (2006). Combining bidirectional translation and synonymy for cross-language information retrieval. In Proceedings of the 29th Annual International ACM SIGIR Conference on Research and Development in Information Retrieval, pages 202–209. [Xu and Weischedel, 2005] Xu, J. and Weischedel, R. (2005). Empirical studies on the impact of lexical resources on CLIR performance. Information Processing & Management, 41(3):475–487. [Xu et al., 2001] Xu, J., Weischedel, R., and Nguyen, C. (2001). Evaluating a probabilistic model for cross-lingual information retrieval. In Proceedings of the 24th Annual International ACM SIGIR Conference on Research and Development in Information Retrieval, pages 105–110. [Yamabana et al., 1998] Yamabana, K., Muraki, K., Doi, S., and Kamei, S. (1998). A language conversion front-end for cross-language information retrieval. In Grefenstette, G., editor, Cross-Language Information Retrieval, pages 93–104. Kluwer. [Yang et al., 1998] Yang, Y., Carbonell, J., Brown, R. D., and Frederking, R. E. (1998). Translingual information retrieval: learning from bilingual corpora. Artificial Intelligence, 103:323–345. [Zhang and Vines, 2004] Zhang, Y. and Vines, P. (2004). Using the web for automated translation extraction in cross-language information retrieval. In Proceedings of the 27th Annual International ACM SIGIR Conference on Research and Development in Information Retrieval, pages 162–169.

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

INDEX

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

A abnormalities, 10 academic, ix, 67, 68, 71, 81, 82, 103, 169, 188 academic settings, 169 ACC, 8, 11, 15, 19 access, viii, ix, 35, 40, 41, 42, 43, 44, 48, 49, 52, 61, 62, 63, 78, 85, 87, 94, 95, 98, 100, 105, 111, 114, 117, 118, 119, 123, 124, 125, 126, 136, 137, 138, 144, 145, 151, 160, 163, 171, 196, 203, 209, 225, 243, 284 acculturation, 182, 198 accuracy, viii, 66, 68, 75, 77, 82, 83, 86, 94, 97, 113, 116, 124, 126, 128, 129, 131, 134, 135, 150, 153, 156, 162, 176, 185, 205, 221, 222, 223, 224, 226 achievement, viii, 81, 82, 83, 89, 111, 154, 162, 170, 172, 181, 183, 186, 195 achievement scores, viii, 81, 82 achievement test, 195 ACL, 242, 284 ACM, 241, 242, 244, 275, 277, 278, 279, 280, 282, 283, 284, 285, 286 acquisitions, 176 activation, viii, 20, 32, 35, 39, 40, 46, 47, 50, 51, 52, 57, 60, 61, 63, 86, 93, 104, 105, 118, 120, 123, 125, 131, 136, 137, 138, 164, 170, 171, 172, 174, 176, 185, 195, 206, 207, 208, 209, 210, 211, 214 acute, 108 Adams, 148, 162 adaptation, 123, 125, 126, 189, 190 adjustment, 273 administration, 73, 169, 172, 174, 186, 188, 190, 192, 196, 258, 259, 261, 262, 263, 264, 269 adolescence, 19, 29, 104 adolescents, 2, 11, 17, 212 adult, 18, 64, 82, 85, 94, 113, 136, 137, 151, 152, 153, 178, 182, 186, 187, 189, 207, 216 adulthood, 13, 19, 204 adults, x, 2, 6, 7, 8, 9, 11, 12, 13, 14, 15, 17, 18, 20, 22, 23, 25, 28, 52, 60, 86, 125, 136, 137, 142, 153, 155, 163, 171, 181, 184, 188, 193, 196, 197, 198, 207, 208, 212, 213, 216, 217

affirmative action, 195 Africa, 242 African American, 182, 183 African Americans, 182 age, 2, 12, 13, 19, 22, 26, 41, 43, 67, 68, 69, 70, 71, 72, 73, 77, 87, 106, 109, 110, 111, 112, 127, 128, 129, 130, 131, 152, 161, 169, 170, 171, 177, 179, 182, 189, 190, 191, 202, 203, 204, 205, 207, 211, 213, 216 agent, 6, 10, 20, 21 agents, 2 aging, 60, 171, 177, 178, 179 aid, 219 aiding, 171 algorithm, 242, 247, 251, 254, 262, 263, 264, 266, 268, 272, 277, 281 alphabets, x, xi, 147, 229, 231, 232 alternative, 12, 16, 183, 186, 206, 207, 230, 231, 249, 250, 272 ambiguity, xi, 87, 92, 98, 100, 107, 117, 150, 230, 240, 243, 246, 251, 256, 260, 263, 265, 271, 272, 275, 282, 285 American culture, 21 American Psychiatric Association, 78 American Psychological Association, 182 American Sign Language, 37, 46, 60, 62, 63, 64, 207 Amsterdam, 29, 62, 99, 100, 102, 119, 122, 144, 164, 179 amygdala, 10, 11, 26 analysis of variance, 219, 221, 222 anatomy, 210 animations, 17 ANOVA, 133, 189, 192 anterior cingulate cortex, 3, 6, 8, 11, 172 aphasia, 104, 106, 108, 110, 120, 121, 170, 179, 203, 205, 211, 212, 214 API, 255, 262 APL, 283 appendix, 254 application, xi, 169, 174, 176, 234, 235, 236, 243, 255 Arab world, 98, 107, 117, 148, 149, 155 Argentina, 188 argument, 24, 31, 59, 153, 184

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

288

Index

Ariel, 79 arithmetic, 32 Army, 225 arousal, 101 articulation, 38, 70 ASD, 11 Asian, 5, 6, 9, 21, 27, 223, 242, 282 Asian cultures, 5 assessment, 69, 78, 99, 118, 163, 172, 177, 182, 187, 190, 192, 196, 197, 198 assessment techniques, 190 assignment, 218 assimilation, 170 assumptions, 171, 270, 274 asymmetry, 86, 95, 113 Atlas, 213 attribution, 2, 5, 7, 24, 30, 32, 154 atypical, 2, 102, 122, 202, 211 auditory domain, 59 auditory modality, 41, 56, 86, 93, 154 autism, 2, 10, 11, 16, 23, 24, 25, 26, 27, 29, 31, 33, 71 automatization, 13, 207 availability, 66, 69, 71, 73, 167, 256 avoidance, 98, 117 awareness, 30, 63, 68, 69, 83, 84, 87, 88, 89, 98, 99, 100, 111, 118, 119, 120, 124, 125, 141, 142, 143, 148, 152, 155, 156, 160, 163, 164, 169, 185, 216, 217, 227

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

B babies, 208 background information, 36 Bahrain, 152 Balearic Islands, 67 barriers, 177 batteries, 174 battery, 102, 120, 122 Bayesian, 270 behavior, 2, 6, 14, 25, 31, 32, 71, 164, 183 beliefs, 14, 15, 26, 27, 28, 30, 32, 33 bell, 229 benefits, viii, 65, 68, 80, 145, 217 bias, 61, 181, 190, 196, 198 bilingual, vii, viii, ix, x, 11, 32, 35, 36, 37, 42, 43, 44, 46, 47, 48, 49, 50, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 65, 66, 67, 68, 69, 70, 72, 73, 74, 76, 77, 78, 79, 80, 82, 84, 85, 86, 87, 88, 89, 93, 97, 98, 99, 100, 101, 102, 103, 104, 105, 106, 109, 110, 111, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 126, 127, 128, 129, 130, 131, 132, 135, 136, 137, 138, 142, 143, 144, 155, 156, 163, 168, 169, 170, 171, 173, 177, 178, 179, 181, 184, 185, 186, 187, 188, 191, 192, 193, 194, 195, 196, 197, 198, 202, 203, 204, 205, 206, 207, 208, 209, 210, 211, 212, 214, 244, 247, 248, 250, 251, 252,

253, 254, 255, 256, 257, 259, 260, 265, 266, 268, 270, 271, 276, 280, 285, 286 bilingualism, vii, viii, x, 35, 36, 47, 55, 57, 60, 61, 63, 65, 66, 67, 68, 71, 78, 79, 84, 85, 87, 88, 89, 98, 99, 117, 118, 137, 142, 171, 177, 178, 179, 181, 182, 183, 184, 185, 186, 187, 188, 190, 194, 195, 196, 197, 202, 203, 205, 207, 209, 210, 211, 214 birth, 71, 155, 213 blocks, 95, 113, 148 blog, 272, 273 blood, 104, 206 BOLD, 177 borrowing, 175, 249 Boston, 109, 120, 187, 188, 197, 198 bottom-up, 36, 45, 46, 47, 52 boys, 69, 71, 72, 127, 191 Braille, 176 brain, vii, ix, x, 1, 2, 3, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 22, 23, 24, 26, 27, 29, 57, 58, 76, 78, 79, 103, 104, 105, 106, 108, 109, 110, 111, 116, 117, 119, 120, 121, 167, 168, 170, 171, 172, 175, 176, 177, 178, 179, 182, 183, 201, 202, 203, 204, 205, 206, 207, 208, 209, 210, 211, 212, 214 brain activity, 3, 7, 11, 13, 14, 15, 79, 175 brain damage, 106, 109, 110, 111, 116, 117, 202 brain imaging techniques, 78 brain injury, 106, 184 brain structure, 9 brain tumor, 106, 120 Brazil, 167 Brazilian, 5 Bronx, 70 buffer, 41, 51 building blocks, 282 Bureau of the Census, 190, 198

C Canada, 167 candidates, 50, 127, 230, 250, 259, 260, 265, 266 capacity, 14, 42, 59, 67, 98, 117, 174, 202, 216, 235, 236 case study, 29, 95, 110, 114 categorical perception, 41, 42, 58 categorization, 56, 157 category a, 55, 71, 109, 223 category d, 182 causal reasoning, 5 ceiling effect, 184, 186 cement, 183, 195 Census, 67, 80, 190, 198 Census Bureau, 67, 80 central executive, 174 centralized, 272 cerebellum, 11 cerebral blood flow, 206

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Index cerebral hemisphere, 95, 101, 164 channels, 36, 97 childhood, vii, 13, 17, 19, 29, 30, 66, 68, 88, 89, 90, 163, 201, 202, 204, 210, 214 children, vii, viii, ix, x, 1, 2, 3, 4, 5, 6, 7, 8, 10, 11, 12, 13, 15, 16, 17, 18, 19, 20, 22, 23, 25, 26, 27, 28, 29, 30, 32, 33, 41, 44, 52, 62, 63, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 83, 84, 87, 88, 89, 92, 98, 99, 100, 106, 117, 118, 119, 123, 125, 126, 137, 138, 141, 142, 143, 149, 153, 155, 156, 157, 159, 161, 162, 163, 164, 173, 181, 182, 183, 184, 185, 186, 188, 190, 191, 192, 193, 196, 197, 198, 202, 208, 209, 210, 211, 212, 213, 216, 223, 227 chimpanzee, 2, 31 China, 33 classes, 159, 183, 191 classical, 17, 104, 155, 168 classification, 184, 186, 187, 197 classroom, 70, 161 classrooms, 82, 142 clients, 196 clinician, 69 clustering, 266, 276, 282 clusters, 56, 57, 121, 266 coding, 24, 98, 117, 144, 163, 205 cognition, x, 2, 15, 18, 19, 23, 24, 27, 29, 32, 100, 102, 119, 122, 151, 181, 183, 184, 185, 186, 187, 190, 194 cognitive, vii, viii, ix, 1, 2, 4, 8, 9, 10, 11, 12, 13, 14, 15, 16, 18, 19, 22, 23, 25, 27, 28, 36, 37, 39, 41, 42, 59, 60, 65, 68, 69, 77, 78, 81, 84, 85, 86, 87, 88, 92, 94, 98, 100, 103, 104, 105, 111, 115, 116, 117, 118, 120, 142, 155, 161, 162, 164, 165, 167, 168, 169, 171, 172, 173, 174, 175, 176, 177, 184, 185, 186, 190, 193, 195, 196, 197, 203, 211, 212, 227 cognitive abilities, 168, 173, 174 cognitive ability, 173 cognitive capacities, 9 cognitive deficit, 77 cognitive deficits, 77 cognitive development, 2, 22, 27, 212 cognitive flexibility, 68, 173 cognitive function, vii, 1, 12, 16, 18, 173, 176, 184, 197 cognitive level, 37, 39, 42 cognitive load, 161 cognitive process, 19, 39, 69, 92, 100, 120, 155, 162, 164, 165, 171, 176 cognitive processing, 19, 39, 100, 120, 162, 164, 171 cognitive psychology, 227 cognitive science, 10 cognitive system, 41, 86, 87, 88, 94, 104, 115 cognitive tasks, 171, 173, 186 cognitive test, 177, 186 cognitive testing, 177 coherence, 10, 26, 282 cohesion, 263

289

collaboration, 78 College Station, 201 Colombia, 99, 118, 191 Colorado, 215, 217, 225, 226 colors, 3 Columbia, 98, 117 communication, 29, 44, 63, 66, 79, 92, 107, 137, 148, 155, 168, 176 communication systems, 63 communicative intent, 33 communities, 66 community, xi, 92, 195, 217, 236, 243 competence, 13, 16, 27, 30, 82, 108, 109, 110, 172, 196, 212 competition, 47, 61, 62, 63 competitor, 45, 47 complement, 16 complexity, 23, 47, 60, 84, 89, 90, 93, 95, 96, 107, 111, 112, 113, 114, 115, 151, 153, 156, 157, 161, 169, 171, 174, 175, 178, 179, 196, 209, 247, 262, 263, 264 components, 12, 41, 104, 109, 152, 167, 169, 171, 177, 231, 245 compounds, 266 comprehension, ix, x, 24, 33, 36, 45, 46, 57, 63, 71, 81, 83, 84, 89, 92, 107, 108, 109, 110, 127, 152, 154, 162, 165, 167, 169, 172, 187, 195, 205, 209, 210, 215, 216, 217, 218, 219, 221, 223, 224, 225, 226 computation, 163, 254, 269, 270 computer use, 128 computing, 245, 254, 262, 268 conception, 23, 24 conceptual model, 36 concrete, 105 concreteness, 85, 122 confidence, 82 configuration, 227 conflict, 94, 196 confrontation, 110 confusion, 108, 154 Congress, 231, 237, 238, 239, 240, 241, 242 conjecture, 185, 195 connectionist, 62 connectivity, 161 consciousness, 2, 30, 108, 227 consensus, 17, 84, 88, 184, 208 consent, 70, 190 constraints, 43, 58, 209, 212 construction, 109, 110, 255 contingency, 253 control, 6, 7, 8, 10, 12, 13, 17, 43, 44, 45, 50, 60, 62, 68, 78, 83, 88, 120, 121, 128, 129, 137, 153, 168, 169, 170, 171, 172, 173, 175, 176, 177, 178, 179, 185, 196, 209, 210, 212, 216, 217, 218, 219, 220, 221, 222, 223, 225 control condition, 6, 7, 12, 13, 17, 129, 218, 219, 220, 221, 223, 225 control group, 83, 170, 176, 216, 217

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

290

Index

controlled research, 186 convergence, 60 conversion, 90, 112, 124, 286 coordination, 62, 174 correlation, 11, 12, 16, 73, 75, 76, 77 correlations, 75, 76, 77, 153, 173 cortex, 3, 6, 7, 11, 24, 26, 33, 122, 174, 202, 204, 207, 210 cosine, 254, 256 Costa Rica, 188 cost-effective, 59 costs, 63 cough, 215 covariate, 186 coverage, 257, 264, 265, 266 covering, 226 craniotomy, 109 credit, 226 critical period, 170, 212 criticism, 13, 14 cross-cultural, vii, viii, 1, 2, 4, 9, 20, 22, 65, 66, 67, 69, 70, 75, 76 cross-cultural differences, 9, 76 cues, 11, 40, 203 cultural differences, 5, 6, 20 cultural influence, 1, 33 cultural psychology, 23 culture, vii, 1, 2, 3, 4, 5, 9, 14, 16, 20, 21, 22, 32, 33, 69, 99, 102, 118, 176 curriculum, 188

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

D data analysis, 7 data collection, 233 data set, x, 181, 188, 189, 194, 272 database, x, 128, 142, 229, 230, 258, 259, 261, 262, 264, 266, 268, 269, 270, 284 database management, 258, 261, 262, 266, 269 deafness, 63 decision task, 97, 115, 136 decisions, 105 decoding, 14, 83, 84, 152, 153, 160, 216, 227 decomposition, 97, 116, 247 defects, 251 deficit, viii, 23, 65, 68, 69, 70, 77, 99, 108, 116, 118, 120, 160, 185, 201, 202, 203 deficits, 30, 31, 67, 68, 69, 71, 77, 95, 108, 110, 111, 114, 185, 201, 202, 203, 214 definition, vii, 15, 66, 87, 98, 117, 168, 182, 183, 186, 194, 269 degradation, 251 degrading, 283 demand, 89, 116 dementia, 68, 78, 177 demographics, 188 density, 207 dependent variable, 90, 112, 205, 217

depressed, 223 derivatives, 106 designers, 175 desire, 18, 19, 20, 22 detection, 14, 15, 87, 155, 156, 159, 161, 227 determinism, 15, 16, 19, 21, 22, 26 developing brain, 24, 209 developmental change, 13 developmental dyslexia, 144, 145, 163 developmental psychology, 31 developmental theories, vii, 1, 9, 18, 21 dichotic listening studies, 205 dichotomy, 9, 20, 47 differentiation, 82, 92 disability, 183 disabled, 152, 162 discourse, 6, 25, 82, 169, 174, 175, 178 discrimination, 14, 60, 69, 95, 111 disorder, 10, 11, 29, 31, 109, 110, 211 disposition, 6 dissociation, ix, 103, 104, 106, 108, 109, 110, 120 distraction, 218 distribution, 171, 185, 186, 195, 219, 271 divergence, 60 diversity, 144, 148, 167 division, 91, 95, 96, 114, 115 division of labor, 91, 95, 96, 114, 115 dogs, 71 dominance, ix, 123, 130, 133, 137, 138, 179, 186, 187, 188, 190, 202, 203, 204, 205, 206, 207, 208, 210, 212, 213, 250 donor, 139 dorsolateral prefrontal cortex, 12 drugs, 109 dual task, 205 duration, 90, 93, 112, 114, 206 dyslexia, 142, 145, 163, 165

E East Asia, 27, 223 ecological, 92 Education, 1, 35, 63, 65, 67, 78, 81, 82, 101, 103, 123, 143, 147, 167, 181, 189, 190, 196, 201, 215, 229, 243 educational system, 82 educators, 83 Egypt, 152 elaboration, 160 elderly, 171 elderly population, 171 election, 67 electroencephalography, 167, 206, 208, 213 electrophysiological study, 102, 121 elementary school, x, 83, 127, 157, 162, 215 emission, 7, 24, 25, 167, 206, 210, 211 emotion, 15, 27, 29, 30, 31, 32 emotional, 5, 8, 9, 13, 19, 24, 25

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Index emotions, 5, 13, 15, 19, 25 empathy, 15, 17, 27, 28, 32, 33 encapsulated, 3, 10 encephalomyelitis, 108 encoding, 28, 44, 62, 94, 144 England, 98, 117 English as a second language, 188, 190, 216 English language proficiency, x, 181, 182, 186, 191, 194 enterprise, 168 enthusiasm, 14 entropy, 268 environment, 25, 37, 92, 161, 169, 172, 202, 225 epidemiologic studies, 69 epilepsy, 109, 213 episodic memory, 30 equipment, 128 ERPs, 206 ESL, 215, 216, 217, 219, 221, 223, 224, 225, 226, 227 ESO, 78 estimating, 188, 261, 271 ethnicity, 173, 188 Europe, viii, 65 Europeans, 2, 5, 6, 21 event-related brain potentials, 210 event-related potential, 31 evolution, 27, 29, 136 exclusion, 127 execution, 244 executive function, 12, 26, 168, 171, 183, 184, 185 executive functioning, 184 executive functions, 168, 171, 183 executive processes, 68 experimental condition, 217, 218 experimental design, 167 expert, 258 expertise, 137 explicit memory, 170 exposure, 66, 68, 71, 76, 77, 83, 84, 87, 88, 89, 90, 91, 97, 106, 112, 117, 128, 130, 137, 158, 159, 161, 169, 170, 172, 173, 202, 204, 207, 217 extinction, 69 extraction, 96, 114, 286 eye movement, 45, 142 eye-gaze, 17 eyes, 60, 62

F facial expression, 38, 43, 52 factorial, 219, 221, 222 failure, 10, 93, 95, 111, 265 false belief, 26, 29, 30, 31, 32, 33 family, 6, 66, 71, 82, 174, 249 Far East, 223, 224 fax, 201 feedback, 57, 60, 219, 252, 264, 265, 277, 280

291

feeding, 47 feelings, 30 females, 189 fever, 108, 139 filters, 7 fire, 55 first language, vii, 70, 71, 76, 77, 82, 85, 86, 90, 104, 106, 109, 130, 131, 137, 153, 189, 223, 227, 248 first language (L1), 104, 109 fixation, 96, 99, 114, 119, 129 flexibility, 97 flow, 104, 206, 231 focusing, 37, 124, 130, 147, 151, 273 foreign language, 87, 179, 190, 243, 285 formal education, 169, 170, 171, 173 Fox, 179 France, 209, 234 frontal cortex, 30, 178, 204, 208 frontal lobe, 7, 183, 184, 198 frontal lobes, 183, 184, 198 fruits, 71 functional architecture, 59 functional imaging, 26, 27, 29 functional magnetic resonance imaging, 7, 8, 11, 14, 23, 26, 27, 28, 30, 32, 104, 105, 122, 125, 167, 174, 177, 178, 179, 202, 206, 207, 208, 209, 210, 211, 212, 213, 214 functional separation, 208 fusiform, 11 fusion, 256

G Gallup, 185, 198 games, 4, 161 gauge, 258 gender, 72, 73, 99, 119, 191, 202 general intelligence, 68 generalization, 144, 185, 186 generation, 104, 109, 120, 170, 174, 178, 211, 257, 286 genetics, 182 Geneva, 80 geography, 30 Germany, 22, 28 gestures, 42, 49 gifted, 183, 197 girls, 69, 71, 72, 127, 191 grades, 67, 68, 152, 153, 158, 159, 160, 161, 162, 191 grain, 124, 125, 126, 145 grants, 78 gray matter, 207 group membership, 186 grouping, 185, 188, 191, 192, 195 groups, viii, x, 3, 6, 7, 13, 16, 18, 20, 22, 36, 41, 42, 43, 47, 58, 59, 60, 67, 69, 70, 73, 74, 75, 76, 77, 81, 88, 90, 91, 94, 96, 112, 113, 115, 150, 154,

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

292

Index

155, 156, 157, 159, 161, 170, 173, 174, 176, 197, 198, 215, 216, 217, 223 growth, 136 guilt, 8, 32 gyri, 104, 174 gyrus, 6, 8, 11, 17, 18, 19

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

H habituation, 10 Haifa, 81, 103, 147, 164 handedness, 99, 119 handling, 40, 230, 238 hands, 11, 37, 41 handwriting, 207 hardships, 82 harmful effects, x, 215 Harvard, 197 harvest, 188 Hawaii, 27 headache, 108 hearing, 42, 53, 60, 68, 71, 207 heart, 36 Hebrew, ix, x, 68, 76, 82, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, 100, 102, 103, 105, 106, 107, 108, 109, 110, 111, 112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 145, 147, 148, 149, 150, 151, 153, 154, 155, 156, 157, 158, 159, 160, 161, 162, 163, 164 hemiplegic, 211 hemisphere, 18, 109, 121, 170, 171, 172, 212 hemispheric asymmetry, 171 hemodynamic, 122, 208 hemorrhage, 106, 108 herpes, 108 herpes simplex, 108 heterogeneity, 176 heterogeneous, 244, 272 heuristic, 249, 255, 260 high school, 108 high scores, 187 high-level, 68, 197 high-speed, 247 hippocampus, 11, 183 Hippocampus, 178 Hispanic, viii, x, 65, 67, 80, 181, 182, 183, 185, 187, 188, 189, 190, 191, 193, 194, 195, 196, 197, 199 Hispanic population, 67 Hispanics, 182 Holland, 62, 99, 100, 102, 118, 119, 122, 164, 203, 211, 213 homework, 161 homogeneity, 169, 172, 173 homogenous, 177 homograph, 149, 152, 179 homophones, 148 Hong Kong, 4, 24 hospital, 108

hospitalized, 108, 109 human, 2, 5, 12, 14, 15, 22, 23, 25, 27, 28, 40, 101, 121, 164, 168, 173, 176, 213, 258 human behavior, 5 human brain, 25, 27, 213 human development, 101, 121, 164 human intentionality, 25 humans, 2, 3, 15 hybrid, 246, 281 hypertext, 259 hypothesis, 1, 2, 3, 4, 5, 10, 12, 13, 15, 16, 18, 19, 21, 22, 29, 63, 75, 86, 89, 90, 94, 95, 104, 105, 111, 113, 135, 159, 160, 170, 171, 173, 204, 205, 208, 210, 223

I IBM, 247, 254, 268, 273, 278 identification, viii, 61, 66, 89, 90, 91, 92, 93, 96, 107, 112, 113, 114, 122, 129, 131, 132, 133, 134, 135, 145, 151, 153, 160, 161, 164, 230 identity, 51, 94, 113, 127 idiosyncratic, 84 IFG, 11, 14, 15, 17, 19, 206 Illinois, 71, 79, 123 illiteracy, 81, 98, 117, 176 illusion, 63 imageability, 175 images, 7, 29, 38, 42, 45, 167 imagination, 27 imaging, vii, 1, 2, 3, 6, 7, 9, 10, 12, 13, 14, 16, 17, 18, 19, 20, 25, 26, 27, 28, 29, 32, 104, 122, 167, 175, 178, 206, 208, 212 imitation, 11, 14, 17, 27, 28, 30 immersion, ix, 67, 123, 127, 137, 170, 214 immigrants, 173, 182, 195 impairments, 10, 110 implementation, 53, 59, 174 implicit memory, 170 in transition, 142 incidence, 69 inclusion, 45, 55, 92, 95, 97, 114, 116, 187, 188, 191 independence, 85, 100, 120, 234 independent variable, 217, 258, 259 indexing, 247, 248, 251, 252, 271, 276, 278, 282, 285 Indian, 4 indicators, 258 indices, 183, 191, 192 individual character, x, 168, 169 individual characteristics, x, 168, 169 individual differences, 24, 145, 165, 198 Individuals with Disabilities Education Act, 197 infancy, 7, 169, 191 infants, 10, 22, 30, 32, 41, 42, 60, 207, 208, 209, 210 Infants, 30, 210 inferences, 173, 175 inferior frontal gyrus, 11, 14, 170

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Index inferior parietal region, 14 information retrieval, xi, 230, 241, 242, 243, 244, 259, 275, 276, 277, 278, 279, 280, 281, 282, 283, 284, 285, 286 Information System, 286 infrared, 167, 206 infrared spectroscopy, 167, 206 inhibition, ix, 27, 40, 50, 52, 57, 81, 97, 103, 116, 143, 184, 185, 194, 196 inhibitory, 10, 12, 50, 168, 173, 185, 196 injury, 202, 211, 213 insight, 30, 43, 49 institutions, 255 instruction, 67, 70, 83, 84, 124, 126, 175, 188, 192, 216, 219, 225 instruments, 77, 176, 189, 190 integration, viii, 35, 46, 47, 52, 59, 62, 96, 114, 115, 161, 208 intellectual development, 24 intelligence, x, 23, 69, 73, 79, 181, 182, 183, 186, 188, 190, 192, 194, 195, 196, 198, 199 intelligence tests, x, 181, 182, 183, 190, 194 intentionality, 23, 29 intentions, 14, 15, 20, 24, 25, 33 interaction, ix, 18, 33, 41, 43, 44, 49, 52, 54, 55, 57, 94, 103, 115, 132, 133, 134, 135, 150, 170, 176, 192, 219, 221, 222, 223 interactions, 9, 27, 39, 53, 54, 59, 62, 132, 133, 134, 135 interdependence, 142 interface, 62, 84, 255 interference, 16, 44, 50, 51, 54, 94, 95, 99, 101, 102, 113, 121, 179, 185, 195, 205 internalization, 2 internalizing, 170 Internet, xi, 243, 246, 250, 255, 262, 273 interpretation, x, 86, 102, 122, 137, 168, 198, 201 interval, 176 intervention, viii, 65, 69, 70, 76, 78, 79, 97, 142, 154 intracranial, 108 intrinsic, 31, 207 intrusions, 50 ions, 15, 17, 31 isolation, 172 Israel, viii, 81, 82, 85, 98, 103, 108, 117, 147, 154, 157, 162, 163, 164 Italy, 144

J Japan, 6, 30, 243 Japanese, 2, 3, 4, 5, 6, 7, 8, 9, 11, 12, 17, 20, 21, 22, 25, 26, 27, 28, 29, 30, 90, 174, 217, 223, 243, 244, 246, 249, 250, 255, 256, 257, 273, 276, 278 Jerusalem, 101, 162 Jordan, 81 judge, 21, 183 judgment, 3, 7, 8, 14, 21, 169, 170, 236

293

K KAP, 242 kindergarten, 69, 80, 83, 88, 164 kindergarten children, 69, 80, 164 kindergartners, 88 Korean, 8, 9, 25, 122, 130, 145, 165, 217, 223 Kurdish, 234

L labor, 171 Lagrange multipliers, 274 language, vii, viii, ix, x, xi, 1, 2, 3, 4, 5, 8, 9, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 32, 33, 35, 36, 37, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 92, 93, 94, 95, 96, 97, 98, 99, 100, 101, 103, 104, 105, 106, 107, 108, 109, 110, 111, 112, 113, 114, 116, 117, 118, 120, 121, 122, 123, 124, 125, 126, 127, 128, 129, 130, 131, 132, 134, 135, 137, 138, 142, 143, 145, 148, 149, 150, 151, 152, 153, 154, 155, 156, 157, 158, 160, 161, 162, 163, 167, 168, 169, 170, 171, 172, 173, 174, 175, 176, 177, 178, 179, 181, 182, 183, 184, 185, 186, 187, 188, 189, 190, 191, 192, 194, 195, 197, 201, 202, 203, 204, 205, 206, 207, 208, 209, 210, 211, 212, 213, 214, 216, 217, 218, 223, 224, 227, 229, 230, 231, 232, 234, 241, 242, 243, 244, 245, 246, 247, 248, 249, 254, 256, 257, 260, 261, 264, 265, 266, 267, 268, 269, 270, 271, 272, 273, 277, 278, 279, 281, 282, 283, 284, 285, 286 language acquisition, 25, 40, 67, 69, 79, 168, 189, 206, 209, 212, 214 language development, x, 41, 67, 201, 203, 208, 212 language impairment, 67, 70, 77, 78, 79, 80, 109, 110, 204 language lateralization, 205, 211, 213 language processing, vii, viii, ix, x, 9, 18, 20, 26, 35, 36, 37, 39, 40, 43, 46, 47, 48, 52, 55, 57, 58, 59, 60, 62, 81, 88, 94, 122, 142, 156, 167, 168, 170, 171, 173, 176, 177, 181, 182, 185, 186, 188, 189, 191, 194, 195, 202, 204, 206, 207, 208, 212, 213, 244 language proficiency, 98, 118, 130, 172, 177, 185, 186, 187, 188, 189, 206 language skills, viii, 69, 70, 81, 82, 192 large-scale, 277 laterality, 204, 205, 206, 211 late-stage, 51 Latin America, 71 lead, 58, 110, 137, 172, 173, 195, 225, 230, 250 leadership, 234 learners, ix, x, 81, 122, 125, 142, 145, 152, 198, 210, 215

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

294

Index

learning, 41, 42, 44, 55, 67, 68, 82, 83, 88, 90, 97, 98, 100, 105, 117, 118, 119, 130, 131, 143, 145, 148, 149, 152, 153, 155, 161, 162, 165, 169, 170, 183, 190, 197, 202, 216, 227, 279, 281, 286 learning difficulties, 152 learning environment, 161 learning process, 97 Lebanon, 235 left hemisphere, 16, 104, 106, 110, 172, 201, 202, 205, 210, 212 left visual field, 90, 94, 112, 113 left-handed, 213 Lesion, 214 lesions, 101, 106, 121, 201, 202, 212 lexical decision, 39, 40, 63, 93, 96, 101, 114, 116, 120, 142, 154, 160 lexical processing, 99, 118, 164 lifespan, 184, 186, 196, 204 likelihood, 253 limitation, 22, 43 limitations, 19, 266 linear, 247 lingual, 67, 230, 282 linguistic, ix, 2, 4, 5, 6, 9, 15, 16, 17, 19, 21, 22, 28, 29, 36, 37, 39, 41, 42, 47, 49, 58, 66, 67, 78, 81, 82, 84, 85, 86, 87, 90, 91, 94, 98, 99, 104, 105, 106, 108, 109, 112, 118, 119, 128, 141, 142, 143, 156, 163, 169, 172, 173, 174, 175, 176, 177, 178, 185, 186, 196, 197, 208, 248, 256 linguistic information, 39, 49, 58 linguistic processing, 17, 109 linguistic task, 91, 109, 112, 175 linguistically, 36, 42, 49, 104, 198, 248, 279 linguistics, 99, 119, 164, 178 links, 52, 55, 86, 87 listening, vii, 36, 73, 163, 169, 172, 175, 179, 202, 205, 223 literacy, ix, 82, 83, 84, 88, 93, 95, 99, 111, 119, 123, 124, 125, 126, 130, 137, 142, 144, 149, 152, 162, 163, 177, 196 liver, 139 localization, 11, 104, 121, 202, 210 location, 2, 37, 39, 40, 45, 89, 107, 148, 151, 161, 206 locus, viii, 35, 39, 57, 94, 112, 113, 179 London, 23, 26, 78, 142, 163, 164, 211 long period, 2, 109 longitudinal studies, 216 longitudinal study, 16, 23, 164 long-term, 68, 79, 217 lover, 139 low-level, 51 LSI, 247, 252

M machine-readable, 250, 251 magnetic, 122

magnetic resonance, 32, 104, 167, 178, 206 Magnetic Resonance Imaging, 28, 32, 98, 104, 105, 117, 122, 167, 177, 178, 179, 122, 206, 207, 209, 210, 211, 213 magnetoencephalography, 167 maintenance, 68 males, 189 management, 241, 258, 259, 261, 262, 263, 264, 266, 269 manipulation, 37, 224 manners, 124, 125, 126 mapping, 17, 55, 121, 178, 207, 212, 213, 223, 225, 231, 232, 233, 237, 241, 273 market, 82 Markov, 269, 283 Markov model, 269, 283 marriage, 207 Mars, 60 Maryland, 235, 241, 284 masking, 49, 144, 145, 164 Massachusetts, 231, 234, 237, 269 mastery, 204 mathematics, ix, 81, 82, 276 matrix, 48, 247 Mauritania, 81 meanings, 17, 24, 85, 86, 150, 175, 247, 266 measurement, 187, 205 measures, x, 69, 77, 88, 89, 90, 112, 127, 152, 153, 155, 156, 159, 181, 182, 183, 187, 188, 192, 195, 206, 207, 208, 216, 217, 227, 249 media, 169 medial prefrontal cortex, vii, 1, 2, 3, 4, 6, 7, 8, 9, 10, 11, 12, 13, 15, 16, 18, 19, 20, 21, 22, 25, 29 mediation, 85, 92, 160, 165 Mediterranean, 101, 121, 164 MEG, 69, 167, 177 membership, 104, 105 memory, vii, viii, 24, 27, 29, 30, 35, 41, 58, 61, 64, 65, 66, 68, 69, 73, 74, 76, 77, 79, 98, 99, 101, 111, 117, 118, 120, 121, 122, 130, 152, 161, 168, 170, 173, 174, 176, 179, 183, 185, 189, 195, 197, 204, 212, 213, 214, 227 memory deficits, 69 men, 23 mental ability, 182 mental activity, 26 mental age, 182 mental health, 191 mental representation, 102, 122 mental retardation, 183 mental state, vii, 1, 2, 4, 5, 7, 14, 20, 32 mental states, vii, 1, 2, 14, 20 messages, 50 meta-analysis, 2, 4, 13, 29, 205 metaphors, 175 metric, 249, 255, 258 Mexican, 189, 191, 192 Mexican-Americans, 68 Mexico, 127, 138, 188, 189

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Index Microsoft, 218 migrant, 188 migration, 171, 187 milk, 25 mining, 60, 284 Ministry of Education, 82, 101, 154 minorities, 183 minority, 67, 69, 77, 80, 181, 186, 197, 198 minority students, 197 mirror, 14, 15, 23, 25, 30 misleading, 173, 184 MIT, 23, 26, 29, 32, 33, 61, 62, 79, 122, 143, 179, 212, 279 mixing, 174, 175, 176, 209, 245, 267 mobility, 69 modalities, viii, 35, 36, 43, 46, 52, 55, 58, 59, 90, 108, 110, 112 modality, vii, 19, 35, 36, 37, 39, 41, 42, 43, 44, 46, 48, 56, 57, 58, 59, 62, 86, 93, 154 modeling, xi, 36, 43, 53, 55, 58, 59, 243, 245, 249, 254, 267, 268, 279, 282 models, vii, viii, xi, 1, 4, 9, 21, 22, 31, 35, 36, 37, 39, 40, 47, 49, 52, 55, 57, 58, 59, 62, 63, 85, 92, 93, 105, 150, 152, 153, 154, 203, 231, 237, 243, 244, 254, 269, 270, 276, 279, 282, 284 moderates, 182 modulation, 179 modules, 12, 13 MOG, 11 mole, 141 monkeys, 14, 15 mood, 92 morning, 226 Morocco, 83, 102 morphemes, 91, 96, 106, 115, 141 morphological, ix, 37, 85, 86, 91, 92, 95, 96, 97, 103, 106, 108, 111, 114, 115, 116, 119, 145, 148, 151, 152, 153, 154 morphology, 37, 69, 77, 91, 93, 94, 95, 97, 106, 109, 110, 114, 115, 119, 142, 144, 153, 154, 162, 163, 164 mother tongue, vii, 76, 79 motion, 25, 37, 38, 45, 46, 207 motivation, 29, 172, 187 motor coordination, 52 motor skills, 183 mouth, 38 movement, 11, 37, 39, 40, 41, 100, 102, 119, 122 MRD, 230, 251, 253, 255, 256, 257, 259, 260 multicultural, 196, 198 multidimensional, 186, 252 multiple-choice questions, 218, 219, 226

N naming, 39, 63, 101, 105, 108, 109, 110, 111, 121, 150, 154, 160, 161, 170, 177, 178, 185, 187, 197, 206

295

national, viii, 78, 81, 82, 189, 191 National Academy of Sciences, 26, 210, 213 National Institute of Standards and Technology, 277, 278, 281, 283, 284, 285 National Research Council, 183, 197 natural, 158, 169 natural environment, 169 navigation system, 259 near-infrared spectroscopy, 209, 214 neonates, 208, 214 nervous system, 214 Netherlands, 100, 101, 102, 119, 120, 122, 164, 165 network, 18, 28, 30, 86, 203, 208, 270 neural development, 19 neural mechanisms, 27, 174 neural network, 24, 30, 125, 138, 203 neural networks, 24, 125, 138 neural systems, 30 neural tissue, 204 neuroanatomy, 24, 211, 213 neurobiological, 18, 171 neurobiology, 111 neuroimaging, vii, 1, 4, 6, 7, 8, 9, 10, 11, 12, 14, 15, 17, 18, 19, 20, 22, 31, 167, 168, 169, 171, 173, 174, 175, 176, 177, 179, 202, 206, 207, 208, 209, 214 neuroimaging techniques, 167, 176, 207 neurological condition, 184 neurological deficit, 108 neurological disorder, 71 neuronal systems, 30 neurons, 15 neurophysiology, viii, 26, 65, 69, 78 neuropsychological assessment, 196 neuropsychological tests, 108, 181, 186 neuropsychology, 167, 168, 177, 184, 196, 212 neuroscience, 23, 25, 211 neurosurgery, 121 New Orleans, 241 New York, 23, 27, 29, 30, 31, 33, 62, 70, 78, 79, 98, 117, 120, 164, 165, 177, 182, 196, 197, 198, 209, 213, 214, 242 newspapers, 159 Newton, 17, 30 NIRS, 204, 206, 207, 208, 209, 214 NIST, 241 nodes, 52, 53, 56, 59, 105 noise, 68, 79 non-English speaking, 2, 4 non-human, 22 non-human primates, 22 non-native, 40, 41, 63, 176, 206, 217 nonverbal, 16, 24, 26, 28, 32, 68 normal, 23, 69, 72, 93, 97, 130, 156, 162, 179, 186, 195, 213, 217 normal distribution, 186, 195 normalization, 272, 283 norms, 23, 69, 71, 72, 76, 109, 110, 184, 188, 189, 277

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

296

Index

NTU, 282

O

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

obligatory secondary education, 78 observations, 201, 239, 262 oculomotor, 93, 114 oligodendroglioma, 109 one dimension, 184 one-to-one mapping, 237 online, 46, 96, 114, 175, 178, 230 opacity, 124 operator, 235, 237, 270 opposition, 18, 83 optical, 213 oral, 39, 44, 77, 84, 90, 112, 121, 130, 155, 169, 172, 175, 187 orbitofrontal cortex, 3 organ, 168 organization, ix, 26, 84, 85, 86, 100, 103, 104, 106, 111, 116, 120, 121, 170, 171, 174, 176, 178, 179, 202, 203, 204, 206, 209, 210, 212 organizations, 255 orientation, 45, 125 orthographic codes, 54, 93 orthography, ix, 55, 57, 59, 83, 84, 89, 90, 91, 92, 93, 94, 97, 98, 100, 102, 106, 107, 109, 111, 114, 116, 117, 119, 120, 122, 144, 145, 147, 149, 151, 152, 153, 154, 155, 156, 157, 161, 162, 164, 165, 224, 231 oxygen, 177

P pain, 27 Panama, 188 paper, 24, 101, 121, 139, 209, 210, 231, 273, 279 parallel processing, 57, 59 parameter, 39, 245, 254, 267, 276 parameter estimation, 276 parents, 6, 66, 67, 70, 71, 83, 161, 188, 191 Parietal, 26 parietal cortex, 172 parietal lobe, 207 Paris, 178, 209 passive, 66, 67, 71, 73, 106, 175 patents, 243 pathologists, 70 pathology, 68, 202, 210 pathways, 54, 55, 59, 204, 206, 213 patients, ix, 7, 68, 101, 103, 104, 106, 121, 171, 178, 201, 202, 210, 212 pears, 101, 120 Pearson correlations, 75 pedagogical, 82, 97 pedagogy, 83 pediatric, 2

peers, 47, 79, 152 Pennsylvania, 121, 142 percentile, 189 perception, viii, 4, 35, 37, 40, 42, 44, 46, 52, 54, 55, 58, 59, 60, 63, 64, 79, 86, 87, 108, 109, 110, 111, 148, 156, 159, 164, 206, 209, 210, 211, 213 perceptions, 5, 25 performance, viii, x, 2, 4, 5, 6, 7, 10, 13, 16, 20, 33, 41, 65, 68, 70, 73, 76, 77, 83, 86, 87, 88, 89, 90, 91, 94, 95, 100, 106, 108, 109, 110, 111, 112, 113, 116, 120, 133, 144, 152, 155, 156, 157, 158, 159, 163, 169, 170, 171, 173, 174, 176, 179, 181, 182, 184, 185, 188, 196, 197, 198, 215, 216, 217, 221, 223, 224, 225, 229, 230, 233, 234, 235, 236, 239, 240, 241, 244, 250, 251, 255, 257, 258, 259, 260, 264, 265, 266, 269, 271, 273, 280, 286 personal, 6, 27, 137, 185, 191, 278 personality, 8, 32, 185 PET, 7, 8, 24, 27, 31, 120, 167, 174, 178, 206, 207, 212 PFC, 19 Philadelphia, 120, 121, 142 phone, 201 phonemes, 39, 40, 41, 42, 44, 56, 57, 88, 99, 107, 118, 124, 126, 141, 145, 147, 151, 152, 156, 216 phonetic perception, 108 phonological, vii, viii, ix, 30, 35, 36, 37, 39, 40, 41, 44, 45, 46, 47, 48, 49, 51, 52, 53, 54, 55, 56, 57, 58, 59, 63, 64, 65, 66, 67, 68, 69, 73, 74, 76, 77, 84, 85, 87, 88, 89, 91, 92, 93, 94, 96, 98, 100, 106, 107, 111, 114, 117, 118, 123, 124, 125, 126, 127, 137, 138, 141, 142, 143, 144, 148, 149, 150, 152, 153, 154, 155, 156, 159, 160, 161, 163, 164, 174, 185, 189, 203, 206, 216, 217, 219, 223, 225, 227 phonological codes, 94, 164 phonological form, 92, 106, 107, 149 phonology, 37, 39, 40, 44, 55, 56, 59, 92, 94, 106, 118, 124, 125, 137, 142, 144, 149, 152, 160, 161, 164, 169, 223 physicians, 182 physiology, 100, 102, 119, 122 PISA, viii, 81, 82, 102 pitch, 104 planning, 51, 78, 185 planum temporale, 179 plastic, 204 plasticity, x, 201, 202, 203, 212 play, 9, 20, 69, 148 poor, 2, 4, 67, 77, 82, 83, 97, 109, 110, 116, 124, 126, 137, 150, 162, 216, 250, 265 poor performance, 77, 109 poor readers, 83, 150, 162 population, 7, 67, 69, 77, 127, 130, 168, 169, 171, 174, 175, 176, 182, 186, 187, 190, 217 population density, 190 positive correlation, 11, 12, 13 positron, 7, 167, 206, 210, 211 positron emission tomography, 7, 210 poststroke, 211

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Index poverty, 196 power, 25, 104 pragmatic, 16, 32, 154, 169, 175, 204 prayer, 155 prediction, 10, 12, 13, 14, 18, 45, 50 predictive validity, 182 predictors, 152, 162, 197 preference, 111, 126, 208 prefrontal cortex, vii, 1, 2, 3, 7, 11, 12, 25, 29, 172 prejudice, 173 preschool, 83, 99, 119, 141 preschool children, 83, 99, 119 preschoolers, 24 preservative, 110 pressure, 185 Pretoria, 242 prevention, 171 PRI, 191 primary school, 67, 83, 153 primates, 27 priming, vii, ix, 35, 39, 40, 41, 42, 52, 58, 61, 62, 63, 85, 86, 87, 100, 101, 108, 111, 120, 123, 127, 128, 129, 132, 133, 135, 142, 144, 154, 160, 163, 164 priming paradigm, 41, 160 prior knowledge, 151 private, 85, 127 probability, 75, 128, 233, 235, 237, 238, 245, 250, 254, 257, 267, 268, 269, 271, 272, 274 probe, 44 procedural memory, 170 production, viii, 35, 37, 42, 43, 47, 48, 49, 50, 51, 52, 55, 59, 61, 68, 69, 71, 94, 105, 108, 110, 118, 144, 153, 167, 177, 206, 207, 208, 210, 211 prognosis, 211 program, viii, 67, 68, 69, 81, 82, 101, 210, 216, 217, 255 promote, 40, 87, 225 pronunciation, 83, 126, 148, 152, 215, 216, 219 protection, 78, 177 pseudo, 157, 159, 160 PSI, 191, 192 psycholinguistics, 82 Psychological Perspective, 60 psychologist, 71, 77 psychology, 12, 13, 14, 18, 19, 22, 23, 27, 121, 142, 198, 226 puberty, 105 public, 70, 127, 155, 188, 190, 191 public schools, 70, 188, 190 Puerto Rican, 5 pupils, 161

Q qualitative differences, 6 query, x, xi, 229, 230, 231, 232, 234, 235, 236, 237, 239, 241, 243, 244, 245, 246, 247, 248, 249, 250, 251, 252, 254, 255, 256, 257, 258, 259, 260, 261,

297

262, 263, 264, 265, 266, 267, 268, 269, 270, 271, 272, 273, 275, 276, 277, 278, 279, 280, 281, 283, 284, 285 questionnaire, 70, 71, 73, 130, 169 questionnaires, 169, 172 quizzes, 217, 225

R race, 148, 188, 198, 199 radioactive tracer, 207 random, 153, 186, 218 random assignment, 186 range, 43, 71, 73, 129, 154, 187, 188, 189, 190, 192, 194, 207, 209, 239, 243, 255, 262, 263, 273 reaction time, 39, 93, 154, 159, 205 reading, vii, viii, ix, x, 14, 15, 17, 30, 32, 68, 80, 81, 82, 83, 84, 88, 89, 91, 92, 93, 94, 95, 96, 97, 98, 99, 100, 102, 103, 108, 109, 110, 111, 114, 115, 116, 117, 118, 119, 120, 122, 123, 124, 125, 126, 127, 128, 129, 130, 131, 136, 137, 141, 142, 143, 144, 145, 148, 149, 150, 151, 152, 153, 154, 155, 156, 160, 161, 162, 163, 164, 165, 169, 172, 173, 174, 175, 179, 184, 185, 187, 205, 215, 216, 217, 218, 219, 220, 222, 223, 224, 225, 226, 227 reading comprehension, 82, 83, 84, 97, 117, 127, 150, 151, 153, 162, 217, 218 reading comprehension test, 83, 217 reading skills, 82, 83, 84, 88, 136, 153 reality, 33, 85 reasoning, 5, 14, 15, 16, 23, 24, 26, 31, 32, 33, 62, 109, 110, 168 reasoning skills, 109, 110 recall, 41, 44, 62, 205, 235, 236, 265 recognition, 23, 30, 45, 46, 56, 59, 61, 62, 63, 83, 84, 90, 92, 93, 95, 96, 97, 98, 99, 100, 113, 114, 115, 117, 119, 120, 121, 124, 125, 131, 132, 133, 135, 136, 141, 142, 143, 144, 150, 154, 155, 162, 163, 164, 205, 213, 216, 227 reconcile, 204 recovery, 109, 170, 203, 204, 214 recruiting, 19, 173 redistribution, 202 reduction, 171 redundancy, 162 refining, 176, 203, 208 regional, 206 regression, 258, 259, 272 regular, 127, 144, 151, 152, 191, 230 rehabilitation, 108, 109, 168, 177 reinforcement, 171 rejection, 96, 115 relationship, vii, 1, 9, 12, 13, 16, 21, 30, 37, 58, 70, 77, 84, 86, 88, 97, 111, 124, 147, 151, 152, 153, 155, 183, 186, 195, 248, 250, 262, 271 relationships, 25, 39, 84, 156, 262, 263, 264 relativity, 29

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

298

Index

relevance, 27, 67, 183, 211, 236, 268, 271, 272, 277, 280, 282 reliability, 168 remediation, 79, 216 repetition effect, 86 repetition priming, 84, 86, 111 reputation, 251 research, vii, x, xi, 1, 2, 4, 7, 15, 21, 22, 30, 32, 35, 36, 37, 41, 46, 47, 51, 52, 57, 58, 60, 66, 67, 68, 69, 76, 77, 78, 83, 87, 92, 93, 95, 102, 111, 123, 124, 125, 126, 136, 137, 138, 151, 153, 154, 156, 159, 160, 161, 162, 163, 165, 167, 168, 169, 170, 171, 172, 173, 174, 175, 176, 177, 186, 187, 195, 196, 203, 207, 209, 210, 214, 225, 235, 242, 243, 244, 247, 254, 269, 271, 273, 276, 277, 280, 281, 282, 283, 284 Research and Development, 242, 275, 277, 278, 279, 280, 282, 283, 284, 285, 286 researchers, 4, 9, 59, 68, 82, 85, 86, 87, 92, 93, 97, 104, 105, 106, 116, 156, 160, 167, 168, 172, 173, 174, 175, 182, 207, 208, 246, 248, 250, 254, 267, 272, 273 reserves, 171 resolution, 7, 105, 106, 196, 282 resources, 77, 89, 96, 115, 174, 241, 243, 249, 250, 252, 255, 256, 259, 265, 271, 280, 281, 283, 286 response time, 51, 94, 111, 113, 158, 159, 219, 220, 222, 224 restructuring, 57 retention, 227 returns, 244 right hemisphere, 18, 32, 94, 99, 104, 110, 112, 113, 118, 119, 172, 202, 209 right visual field, 90, 91, 112 Royal Society, 26 Russian, 68, 76, 156

S SAC, 242 sample, 127, 130, 133, 134, 135, 169, 172, 173, 182, 187, 189, 190, 191, 219 sampling, 170, 183 scarcity, 18, 19 schema, 127 schizophrenia, 24 scholastic achievement, viii, 81, 82, 197 school, ix, x, 67, 70, 73, 78, 81, 82, 83, 85, 108, 127, 128, 137, 141, 152, 153, 154, 155, 157, 159, 161, 162, 169, 183, 188, 189, 190, 191, 215 schooling, 148, 152, 182, 188 scientific theory, 3 scientists, 2, 14 scores, viii, 68, 70, 72, 74, 75, 76, 81, 82, 88, 89, 111, 127, 130, 156, 172, 181, 185, 186, 187, 188, 189, 191, 192, 194, 195, 246, 253, 256, 258, 266, 267, 269, 271, 272

scripts, ix, 89, 107, 124, 125, 126, 147, 151, 152, 156, 162, 242 search, xi, 25, 97, 104, 115, 177, 178, 182, 232, 234, 235, 236, 237, 239, 243, 244, 245, 246, 248, 251, 252, 255, 257, 258, 259, 260, 261, 262, 264, 265, 266, 269, 270, 271, 272, 273, 282 search engine, 232, 234, 243, 246, 255, 262, 282 search terms, 252, 257, 265 searches, 252, 256, 257, 258 searching, xi, 243, 252, 253, 255, 257, 258, 259, 260, 262, 271, 273, 283, 284 second language, 63, 66, 70, 76, 77, 85, 86, 87, 88, 89, 90, 94, 97, 99, 100, 104, 106, 108, 109, 111, 112, 116, 118, 120, 121, 124, 130, 138, 142, 145, 162, 163, 170, 178, 179, 186, 187, 188, 190, 202, 203, 205, 209, 210, 211, 214, 216, 223, 227, 248 secondary school students, 78 segmentation, 124, 126, 137 seizures, 210 selecting, 196, 246, 262 Self, 6, 8, 27, 57 self-awareness, 185 self-evaluations, 172 self-knowledge, 28, 30 self-organizing, 62 self-report, 172, 182, 187, 195 semantic, 16, 18, 24, 26, 28, 30, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 57, 58, 59, 62, 66, 84, 85, 86, 87, 91, 92, 93, 100, 101, 102, 104, 105, 107, 109, 110, 111, 114, 116, 120, 122, 144, 150, 155, 161, 163, 170, 175, 176, 177, 205, 247, 248, 276, 278, 282, 285 semantic activation, 105, 176 semantic information, 114, 150 semantic memory, 111, 161 semantic priming, 28, 51, 85, 86, 100, 101, 120, 163 semantic processing, 105, 120, 170 semantics, 16, 104, 107, 169 sensitivity, 29, 69, 96, 115, 124, 125, 126, 136, 137, 141, 145, 159, 207, 208 sentence comprehension, 26, 209 sentences, 16, 17, 18, 37, 43, 52, 71, 175, 210, 246, 262 separation, 41, 56, 88, 121, 148, 208 series, 10, 42, 109, 129, 153, 156, 160, 184 SES, 67, 70, 71, 76, 77, 186, 195 shape, 24, 89, 97, 107, 151 shares, 45 sharing, 92, 107, 149, 155, 171 Short-term, 62 SIGIR, 242, 244, 275, 277, 278, 279, 280, 282, 283, 284, 285, 286 sign, vii, 30, 35, 36, 37, 38, 39, 40, 41, 43, 44, 45, 46, 50, 51, 53, 55, 58, 59, 63, 64, 176, 207 signals, 149 signs, 36, 38, 39, 41, 42, 43, 44, 49, 50, 51, 55, 62, 126, 129, 149, 164, 171

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

Index similarity, 15, 41, 44, 45, 55, 62, 63, 85, 86, 87, 92, 101, 104, 107, 120, 202, 230, 233, 239, 245, 248, 249, 253, 254, 256, 261, 262, 263, 264, 266, 275 simulation, 3, 4, 14, 15, 18, 21, 25, 27, 30, 31 singular, 92, 107, 247 sites, 208, 243, 255 skills, viii, ix, 67, 68, 69, 70, 77, 81, 82, 83, 84, 88, 99, 103, 109, 110, 118, 125, 130, 136, 143, 148, 152, 153, 156, 162, 163, 182, 183, 185, 187, 192, 194, 195, 216 smoothing, 267 social behavior, 184 social cognition, vii, 1, 4, 6, 7, 13, 17, 19, 20, 21, 22, 29, 32 social development, 98, 118 social norms, 23 social problems, 82 social situations, 185 socialization, 25, 30, 164 sociocultural, 4 socioeconomic, 32, 173, 182, 186 socioeconomic status, 32, 173, 182, 186 software, 128, 246, 247, 250, 251, 254 solutions, 145 sorting, 121, 184 sounds, 41, 42, 43, 69, 98, 109, 118, 149, 151, 152, 153, 215 Spain, viii, 65, 66, 67, 70, 71, 72, 74, 75, 76, 77, 78 spatial, 24, 37, 41, 49, 62, 90, 174 spatial array, 24 spatial information, 174 special education, 183, 191, 195, 197 specialization, 17, 30, 91, 99, 112, 119, 182 specificity, 12 spectroscopy, 209, 211, 214 spectrum, 11, 26, 33, 186 speculation, 18 speech, 36, 40, 42, 43, 44, 46, 48, 49, 51, 55, 58, 59, 61, 62, 64, 68, 70, 79, 107, 108, 109, 120, 126, 130, 164, 201, 206, 207, 208, 209, 210, 211, 212, 213, 259 speech perception, 40, 55, 59, 64, 108, 209, 210, 213 speed, 82, 89, 90, 112, 124, 156, 176, 185, 205, 221, 223 spelling, x, xi, 62, 98, 109, 117, 153, 160, 162, 215, 216, 218, 227, 229, 230, 231, 232, 234, 237, 239, 241, 242, 284, 286 St. Louis, 101, 121, 123 stages, 59, 83, 95, 172 standard deviation, 72, 127, 130 standard error, 132, 220, 221, 222, 224 standardization, 182, 190, 191, 194, 195, 196 Standards, 71, 78, 276, 277, 278, 279 statistical analysis, 74, 249 statistics, 81, 82, 101, 183, 250, 253, 255, 260, 261, 262, 265, 268, 271, 278 Stimuli, 138, 210 stimulus, 44, 95, 113, 156, 157, 184, 206, 208 storage, 41, 145

299

story comprehension, 26 strategic, 6, 129 strategies, xi, 13, 87, 93, 94, 99, 100, 114, 116, 118, 119, 144, 174, 203, 204, 210, 243, 244, 245, 260, 263, 270, 271, 276, 280, 285 strength, 85, 87 stress, 89, 127, 151 string matching, x, 229, 230, 231, 232 strokes, 89, 149 structuring, 64 students, viii, ix, x, 67, 78, 81, 82, 85, 90, 97, 112, 115, 138, 144, 161, 188, 197, 198, 215, 223, 224, 226 subcortical structures, 209 subgroups, 74 subjective, 20, 31, 86, 145, 172 substitution, 197 substrates, x, 29, 170, 201, 204, 211 suffering, 108, 110 sugar, 139 summer, 225, 226 superior temporal gyrus, 11, 17, 23, 206 superiority, 68, 143, 156, 159, 161, 184 supply, 53 suppression, 41, 50, 184 surgery, 106, 108, 109, 110 surprise, 278 survival, 31 susceptibility, 212 switching, 32, 63, 86, 102, 111, 116, 121, 168, 175, 176, 204, 206 symbolic, 4, 32, 98, 117, 196 symbolic activity, 32 symbolic systems, 4 symbols, 2, 23 symmetry, 85, 189 symptom, 110, 214 symptoms, 78, 108, 171, 177 syndrome, 11, 25, 27, 31 syntactic, 16, 17, 18, 24, 26, 37, 48, 52, 62, 63, 84, 89, 92, 99, 107, 118, 124, 148, 153, 170, 174, 175, 179, 203, 205, 213 syntax, 16, 37, 48, 62, 89, 92, 94, 109, 110, 150, 152, 169, 205 systems, viii, ix, 31, 35, 43, 46, 49, 55, 57, 58, 81, 82, 87, 89, 90, 93, 98, 108, 111, 117, 124, 145, 147, 150, 155, 162, 168, 170, 174, 176, 179, 183, 202, 204, 208, 212, 250, 251, 255, 256, 265

T talent, 187 tangles, 226 targets, 39, 40, 85, 86, 129, 156 task demands, 54, 93 task force, 182 task performance, 6, 10, 13 teachers, 70, 83, 97, 161

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

300

Index

teaching, 44, 67, 70, 97, 116, 126, 137, 161, 168, 227 technology, ix, 81, 82, 249, 281 Tel Aviv, 98, 117, 163 television, 161 temporal, 3, 6, 7, 8, 11, 26, 38, 39, 41, 54, 104, 108, 175, 207, 208 temporal lobe, 108 test scores, 182, 183, 194, 195 Texas, 181, 188, 191, 194, 195, 201 textbooks, 254 Thai, 4 thalamus, 11 theory, vii, 1, 3, 4, 10, 12, 13, 14, 15, 18, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 40, 84, 104, 118, 124, 125, 142, 145, 165, 171, 184, 185, 196, 198, 212, 247 therapy, 70, 76, 108, 109, 110 thesaurus, 248, 253 thinking, 13, 31, 62, 162, 207, 212 Thomson, 283 three-way interaction, 17, 222, 223 threshold, 78, 145, 172, 176, 230, 235, 249 threshold level, 172, 176 time, x, 4, 19, 24, 38, 40, 41, 43, 47, 49, 50, 51, 54, 82, 90, 93, 112, 114, 128, 130, 131, 136, 163, 170, 174, 184, 185, 188, 195, 201, 215, 217, 218, 219, 220, 222, 223, 224, 225, 269 time pressure, 185 timing, 43, 49, 51, 206, 219 tin, 254 tip-of-the-tongue, 185, 197 tissue, 204 title, 234 ToM, vii, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22 top-down, viii, 36, 45, 46, 47, 57 tracking, 207 trade, 82 tradition, 16, 44 training, x, 29, 137, 177, 192, 215, 216, 217, 218, 219, 221, 222, 223, 224, 225, 227, 272, 275 traits, 5, 37 trajectory, 4, 20, 21 transcription, 216 transfer, 52, 68, 76, 223, 227 transformation, 230, 231, 232, 249, 286 transgression, 10 transition, 32, 153 translation, x, xi, 50, 51, 84, 85, 86, 87, 98, 99, 100, 101, 111, 117, 119, 121, 164, 192, 197, 229, 230, 231, 235, 240, 242, 243, 244, 245, 246, 247, 248, 249, 250, 251, 252, 254, 255, 256, 257, 258, 259, 260, 261, 262, 263, 264, 265, 266, 267, 268, 269, 270, 271, 272, 273, 275, 276, 277, 278, 279, 280, 281, 282, 283, 284, 285, 286 translational, 283 transparency, 96, 115, 143 transparent, ix, x, 92, 96, 106, 115, 123, 124, 138, 149, 151, 152, 215, 216

trauma, 203 traumatic brain injury, 184 trees, 101, 121 trend, 69, 182 trial, 121, 129, 160 triangulation, 266 tribal, 99, 118 tumor, 106, 109, 120 tumors, 202, 210 two-way, 67 typology, 174

U UCB, 225 undergraduate, 225 undergraduates, 186, 218, 224, 226 underlying mechanisms, 60 UNESCO, 81, 101 UNICEF, 82, 101 United States, 67, 80, 144, 189, 190, 191, 210, 213 universality, 4, 13, 14 university students, 90, 112, 115

V Valencia, 67, 71 validation, 142 validity, 90, 168, 182, 205 values, 96, 104, 115, 177, 187, 236, 238, 254, 255, 268, 272, 275 variability, 2, 83, 118, 152, 172, 178, 209 variable, x, 73, 90, 112, 125, 131, 181, 205, 217 variables, x, 43, 84, 88, 94, 104, 150, 168, 173, 175, 176, 186, 205, 217, 236 variance, 37, 74, 83, 156, 219, 221, 222 variation, 3, 12, 37, 62, 63, 152, 202, 259, 261, 268 vector, 245, 248, 253, 254, 255, 256 Verbal IQ, 182, 190 Vietnamese, 190 virus, 108 visual attention, 46 visual field, 46, 58, 91, 94, 95, 96, 112, 113, 114, 205 visual field studies, 112 visual memory, 152 visual modality, 62 visual perception, 101, 110, 121, 159 visual processing, 164 visual stimuli, 95, 113 Visual-spatial, 25 visuospatial, 41, 64, 108, 109 vocabulary, 68, 78, 88, 150, 169, 170, 185, 187, 215, 223, 224, 251 voice, 175, 213 vomiting, 108 Vygotsky, 1, 3, 4, 16, 33

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,

Index

W

workers, 188 working memory, vii, viii, 27, 35, 41, 58, 64, 65, 66, 68, 69, 73, 74, 76, 77, 79, 99, 118, 152, 174, 183, 185, 189, 204 World Bank, 101, 121, 164 World Health Organization, 80 World War, 182 World War I, 182 writing, vii, xi, 33, 44, 79, 81, 83, 89, 90, 98, 108, 109, 110, 117, 124, 128, 129, 147, 150, 151, 155, 159, 161, 162, 164, 165, 169, 172, 173, 187, 243, 268 writing to dictation, 108, 109, 110

Y yes/no, 156 yield, 204, 241, 262, 265 younger children, 2, 19, 22

Copyright © 2010. Nova Science Publishers, Incorporated. All rights reserved.

warrants, 208 water, 140, 150 WCST, 110 web, 243, 255, 266, 277, 285, 286 web sites, 243, 255 Wechsler Intelligence Scale, 182, 193, 198, 199 Western culture, 6 windows, 128 Wisconsin, 110, 184, 197 women, 23 word format, 98, 117 word frequency, 128, 175, 176 word naming, 143, 164, 178 word processing, 152, 154, 174, 175 word recognition, 61, 62, 83, 90, 93, 95, 98, 100, 117, 119, 120, 124, 125, 136, 141, 142, 143, 144, 150, 155, 163, 164, 227 word-frequency, 154

301

Bilinguals: Cognition, Education and Language Processing : Cognition, Education and Language Processing, Nova Science Publishers, Incorporated,