AUTHOR ""
TITLE "Precursors of Functional Literacy"
SUBJECT "Studies in Written Language and Literacy, Volume 11"
KEYWORDS ""
SIZE HEIGHT "220"
WIDTH "150"
VOFFSET "4">
Precursors of Functional Literacy
Studies in Written Language and Literacy The aim of this series is to advance insight into the multifaceted character of written language, with special emphasis on its uses in different social and cultural settings. It combines interest in sociolinguistic and psycholinguistic accounts of the acquisition and transmission of literacy. The series focusses on descriptive and theoretical reports in areas such as language codification, cognitive models of written language use, written language acquisition in children and adults, the development and implementation of literacy campaigns, and literacy as a social marker relating to gender, ethnicity, and class. The series is intended to be multi-disciplinary, combining insights from linguistics, psychology, sociology, education, anthropology, and philosophy.
Editors
Brian Street
Ludo Verhoeven
King’s College, London
University of Nijmegen
David Bloome Vanderbilt University, Nashville
Editorial Board Charles Bazerman
David Olson
University of California, Santa Barbara
Ontario Institute for Studies in Education, Toronto
David Barton
Mastin Prinsloo
Lancaster University
University of Cape Town
Florian Coulmas
Dorit Ravid
Chuo University, Tokyo
University of Tel Aviv
Peter Freebody
David Reinking
Griffith University
University of Georgia
Eve Gregory
Linda Siegel
University of London
University of British Columbia
Nancy Hornberger
Daniel Wagner
University of Pennsylvania
University of Pennsylvania
Charles Kinzer Vanderbilt University, Nashville
Volume 11 Precursors of Functional Literacy Edited by Ludo Verhoeven, Carsten Elbro and Pieter Reitsma
Precursors of Functional Literacy Edited by
Ludo Verhoeven University of Nijmegen
Carsten Elbro University of Copenhagen
Pieter Reitsma Free University Amsterdam
John Benjamins Publishing Company Amsterdam/Philadelphia
8
TM
The paper used in this publication meets the minimum requirements of American National Standard for Information Sciences – Permanence of Paper for Printed Library Materials, ansi z39.48-1984.
Library of Congress Cataloging-in-Publication Data Precursors of Functional Literacy / edited by Ludo Verhoeven, Carsten Elbro and Pieter Reitsma. p. cm. (Studies in Written Language and Literacy, issn 0929–7324 ; v. 11) Includes bibliographical references and indexes. 1. Language acquisition. 2. Literacy. 3. Language awareness in children. I. Verhoeven, Ludo Th. II. Elbro, Carsten. III. Reitsma, P. (Pieter) IV. Series. P118.7 P74 2002 302.2’244-dc21 isbn 90 272 1806 4 (Eur.) / 1 58811 228 4 (US) (Hb; alk. paper)
2002074691
© 2002 – John Benjamins B.V. No part of this book may be reproduced in any form, by print, photoprint, microfilm, or any other means, without written permission from the publisher. John Benjamins Publishing Co. · P.O. Box 36224 · 1020 me Amsterdam · The Netherlands John Benjamins North America · P.O. Box 27519 · Philadelphia pa 19118-0519 · usa
(JB) SWL II_f1_iv-vii
9/26/02
2:23 PM
Page v
Table of contents Introduction Functional literacy in a developmental perspective Ludo Verhoeven, Carsten Elbro and Pieter Reitsma
Part 1: Roots of phonological awareness The quality of phonological representations and phonological awareness: A causal link? Carsten Elbro and Bolette R. Pallesen
Precursors of phonemic awareness Pieter Reitsma
The role of orthographic onset-rime units in Dutch beginning readers Astrid Geudens and Dominiek Sandra
Manifestations of phonological deficits in dyslexia Vera C. S. Messbauer, Peter F. de Jong and Aryan van der Leij
Metaphonological awareness in monolingual and bilingual kindergartners Vincent Goetry, Régine Kolinsky and Philippe Mousty
Does speech manipulation make word discrimination easier? Eliane Segers and Ludo Verhoeven
Part 2: Factors in reading and writing efficiency The acquisition of untaught orthographic regularities in French Sébastien Pacton, Michel Fayol and Pierre Perruchet
Subsyllabic units in reading: A difference between Korean and English Hye Kyung Yoon, Donald J. Bolger, Oh-Seek Kwon and Charles A. Perfetti
Orthography, phonology and semantics: Concerted action in word perception Anna M. T. Bosman and Janet G. van Hell
(JB) SWL II_f1_iv-vii
9/26/02
2:23 PM
Page vi
Table of contents
The lexical quality hypothesis Charles A. Perfetti and Lesley Hart Relationships between reading and writing skills in the intermediate grades Hanna S. Mäki, Marinus J. M. Voeten, Marja M. S. Vauras and Pekka Niemi Task-related factors in reading efficiency of dyslexic children Aryan van der Leij, Victor van Daal and Peter de Jong
Part 3: Attaining functional literacy Parental and teacher commitment to emergent literacy development Judith Stoep, Joep Bakker and Ludo Verhoeven Sociocultural differences in reading skills, reading motivation, and reading strategies Willy van Elsäcker and Ludo Verhoeven
Bilingualism and reading Linda S. Siegel
Age and gender differences in reading engagement Liliane Kjellman
Promoting at-risk pupils’ foreign language literacy learning Mia Dufva and Marja Vauras
Predictors of adult functional reading skills Elisabeth Arnbak
Subject index
357
(JB) SWL II_f1_iv-vii
9/26/02
2:23 PM
Page vii
Acknowledgements
From August 17–21 1999 an international expert meeting on precursors of functional literacy took place in Nijmegen, The Netherlands. The meeting was organized by Ludo Verhoeven (University of Nijmegen), Carsten Elbro (University of Copenhagen) and Pieter Reitsma (Peadologisch Instituut — Free University Amsterdam), and funded by the Dutch National Science Foundation (NWO), and the Faculty of Social Sciences of the University of Nijmegen. A selection of papers being presented is now integrated into a single academic reference, after being edited and updated. The editors wish to thank all contributors to this volume for redrafting their original papers, Caroline van der Laan for her editorial assistance, and the funding agencies for their support. Ludo Verhoeven, Carsten Elbro and Pieter Reitsma
(JB) SWL II_BLANK
9/26/02
3:38 PM
Page i
(JB)SWL II_f2_1-13
9/27/02
11:02 AM
Page 1
Introduction
(JB)SWL II_f2_1-13
9/24/02
2:36 PM
Page 2
(JB)SWL II_f2_1-13
9/24/02
2:36 PM
Page 3
Functional literacy in a developmental perspective Ludo Verhoeven, Carsten Elbro and Pieter Reitsma
Advancements in research during the past decades has paved the way for models of functional literacy that incorporate personal and social needs (e.g., Verhoeven, 1994; Pontecorvo, 1997). This new line of research has greatly contributed to our insight into the distribution, the consequences and the causes of literacy and problems with literacy. At the same time there has been an increasing awareness of the multifaceted nature of written language. The focus of written language and literacy has moved from concerns with structural aspects of reading and writing towards the acceptance of broader definitions that take into account also the functions of written language in social contexts. The term ‘functional literacy’ was introduced in order to refer to the demands of literacy in the complex world (see Gray, 1956). Literacy has come to be viewed as a complex of skills which is defined in terms of the print demands of occupational, educational, civic, community and personal functioning. With respect to education, the ultimate question is how children’s oral and written language can be fostered from their origins in early infancy to their mastery as systems of representation for communication with others and for the inner control of thinking and feeling. In the present volume an attempt is made to provide insight into the precursors of functional literacy by bridging cognitive and sociocultural points of view. The chapters of the book address the critical issues in the development of literacy from the early stages of reading, writing and the uses of literacy as an instrument for learning and critical thinking. There is clear research evidence that word identification in learning to read requires a phonological mechanism that generates phonological word forms. A phonological constituent applies as soon as the child begins to treat the letters of a word as having speech associated with them. With respect to sociocultural variation, the development of literacy is related to social markers, such as gender, ethnicity, and class.
(JB)SWL II_f2_1-13
9/24/02
2:36 PM
Page 4
Precursors of Functional Literacy
Defining functional literacy Taking a sociocultural approach of literacy as a starting point, the question is what psycholinguistic abilities underlie a functional literacy level in the individual. By incorporating the sociolinguistic concept of communicative competence (cf. Hymes, 1971), a more elaborated conceptualization of literacy behavior can be arrived at. In the context of language teaching Canale and Swain (1980) defined communicative competence as: a synthesis of knowledge of basic grammatical principles, knowledge of how language is used in social settings to perform communicative functions, and knowledge of how utterances and communicative functions can be combined according to the principles of discourse.
Verhoeven (1994, 1997) shows how the construct of functional literacy could be defined by taking the framework of communicative competence as a starting point. A distinction is made between the following types of competences: – – – – –
Grammatical competence Discourse competence (De)coding competence Strategic competence Sociolinguistic competence
Grammatical competence covers the mastery of phonological rules, lexical items, morphosyntactic rules and rules of sentence formation. Discourse competence refers to the knowledge of conventions regarding the cohesion and coherence of various types of discourse. Grammatical and discourse competence refer to those abilities involved in controlling the formal organization of written discourse. The competence to code and decode written text comprises the technical abilities of writing and reading. Strategic competence refers to the ability to perform planning, execution and evaluative functions to implement the communicative goal of the written text. Sociolinguistic competence comprises the literacy conventions which are appropriate in a given culture and in varying social situations, and the mass body of cultural background knowledge. The present model makes it possible to operationalise functional literacy in more or less concrete terms. Coding and decoding abilities relate to the mastery of the essentials of the written language code itself. It has been claimed by many educators that orthographies differ in degree of learnability. From comparative studies of writing systems (see Perfetti, 1998) it can be concluded that all systems represent spoken language at one level or another and that readers activate speech codes during the decoding process — even in morphemic writing systems such as the Chinese. Alphabetic codes have the advantage of a small number number of symbols (letters) needed to map the phoneme inventory of a language. However, from an extensive body of research (see Shankweiler & Liberman, 1989), it has
(JB)SWL II_f2_1-13
9/24/02
2:36 PM
Page 5
Functional literacy in a developmental perspective
been shown that alphabetical codes are difficult in the sense that phonemes as the constituent units can hardly be perceived. The acquisition of phonographic systems appears to be affected by the ‘goodness off it’ between oral and written language units which is relatively high for Finnish and relatively low for English. Given the continuities between oral and written language, the abilities involved in grammatical and discourse competence constitute basic components of functional literacy. Though the linguistic devices used to comprehend or produce written language are not completely identical to those involved in oral discourse, a close relationship can still be expected. Strategic competence refers to metacognitive abilities which are involved in planning, executing and evaluating written text. Planning and evaluation (revision) turn out to be crucial abilities in writing (Levy & Ransdell, 1996), whereas monitoring plays an important role during the execution of the reading process (Butler & Winne, 1995). Sociolinguistic competence enables the individual to cope with literacy situations in everyday life. Sociolinguistic competence comprises both literacy conventions and cultural background knowledge which are interrelated. Literacy conventions refer to the types of documents that are used in the social institutions of a society, such as letters, forms, legal briefs, political tracts, religious texts, novels and poems. Documents often require specialized knowledge about particular document formats. Moreover, different types of documents may also call for different types of cultural background knowledge (Goldman & Rakestraw, 2000) as well as different values and beliefs (Gee, 1990). It is important to see that the notion of sociolinguistic competence makes literacy a relative measure, depending on the social and cultural context. For instance, being able to fill in a form may be functional in one context, but less relevant in another. With respect to modeling the competence of functional literacy, the particular sociolinguistic position of ethnic minorities should be recognized (e.g., Geva & Verhoeven, 2000). Grammatical and discourse abilities become very critical for people from ethnic minorities who have to learn to read and write in an unfamiliar (second) language. Children who acquire literacy in a second language are faced with a dual task: besides the written code they have to learn the grammatical and discourse competence of the second language. In many cases children learning to read in a second language (L2) are less proficient in the target language than their native language speaking peers are. Thus, it can be hypothesized that L2 learners have difficulty in using (meta)linguistic cues while reading. Limited oral proficiency in a second language may influence the various subprocesses of reading (see Koda, 1996). With respect to word recognition and word spelling, there can be difficulties in phonic mediation, resulting in a slow rate of acquisition of graphemephoneme correspondency rules. There can also be difficulties in the use of orthographic constraints, due to a restricted awareness of phoneme distribution rules in the second language. Furthermore, there may be differences between first and second language readers as to higher order processes which follow the identification of words. Due to restricted lexical and syntactic knowledge,
(JB)SWL II_f2_1-13
9/24/02
2:36 PM
Page 6
Precursors of Functional Literacy
or limited background knowledge, L2 learners may have difficulty in parsing sentences into their constituents and in finding their underlying propositions.
Precursors of literacy development Research on emergent literacy has shown that interactive activities, such as storybook reading, communicative writing and language games, have some impact on children’s oral and written language development (see Yaden, Rowe & MacGillivray, 2000; Scarborough & Dobrich, 1994). The interaction with symbols in their environment with literate others helps children to learn that print carries meaning, that written texts may have various forms and functions, and that ideas can be expressed with (non)conventional writing. Moreover, from interactive storybook reading children learn new vocabulary and gain insight into the structure of narrative text. Conditions that strengthen the relevance and purpose of literacy turned out to be quite important for the development of preschool literacy. Empirical studies have made clear that the attainment of literacy can be stimulated and extended by offering children a school environment where valid understandings about literacy can continue to emerge (cf. Snow, Burns & Griffin, 1998). In such an environment, children have the opportunity to enhance the positive literacy experiences they have had prior to school. However encouraging, these findings should not overshadow the crucial role of direct instruction in the alphabetic code (National Reading Panel, 2000). It is one of the most well documented facts in educational psychology that direct instruction in the orthographic code is more helpful for children than indirect instruction where children are left to infer the grapheme-phoneme mappings on their own. This is true especially for children with relatively poor language abilities while other children appear to be able to learn to read with practicaly any method of teaching. A large body of research has been conducted on the relation between phonological awareness and learning to read. Strong support has been provided that lack of phonological awareness can cause difficulties with the acquisition of reading and writing (Brady & Shankweiler, 1991; Elbro & Scarborough, in press; Scarborough, 1998; Torgesen, Wagner & Rashotte, 1997). Being able to distinguish and identify the different phonemes in a word is part of this awareness. Research in the past decades has provided ample evidence that dyslexic children have problems with phonological awareness and certain other aspects of phonological processing. There is a general agreement that this initial processing deficit has to do with problems in phonological encoding (see Brady, 1997; Snowling, 2000). Poor readers are less precise in phonemic discrimination, they have problems on a variety of phoneme segmentation and awareness tasks (Wagner, Torgeson, & Rashotte, 1994), and they are slower in rapid naming of objects, digits and letters (Wolf & Obregon, 1992; Wolf & Bowers, 2000), as well as in producing rhyming words (Høien & Lundberg, 2000). It can be hypothesized that dyslexia is fundamentally a linguistic problem which involves a deficit in phonological encoding. Elbro,
(JB)SWL II_f2_1-13
9/24/02
2:36 PM
Page 7
Functional literacy in a developmental perspective
Borstrøm, and Petersen (1998) tested this hypothesis by predicting dyslexia from phonological processing abilities of kindergartners. Children from dyslexic parents were followed in a longitudinal study from kindergarten through the second grade. Regression analysis showed that three language measures contributed independently to the prediction of dyslexia: letter naming, phoneme identification, and distinctness of phonological representations. The results further indicated that the quality of phonological representations in the child’s mental lexicon may also be a determinant of the development of phonemic awareness. As mentioned above, accumulated research evidence indicates that children need sequentially structured activities that are mediated by a teacher or by skilled peers in order to acquire accurate and automatic coding and decoding skills in writing and reading (see Adams, 1990; National Reading Panel, 2000). Through experience with literacy tasks in guided participation with skilled partners the child’s repertoire of relevant strategies can be gradually expanded. Breaking the alphabetic code is a very strong motivation in the child during the initial phases of reading development. However, in order for the child to develop a sustained motivation and interest in literacy it is important to focus on meaningful experiences, and to stimulate critical thinking in reading and creative expression in writing. Advanced reading and writing demands the development of vocabulary, insight into the structure of sentences and larger textual structures, such as episodes and paragraphs, and knowledge of rules for punctuation. Comparisons between expert and novice learners have also called attention to the importance of control processes, such as planning and monitoring reading and writing processes. Literacy in advanced classes is fostered by teachers who plan lessons that have a clear conceptual focus. Students should be given time to reflect, to practice relevant strategies, and to achieve depth of meaning and understanding. Instruction should focus on principles and ideas that help children making connections between prior knowledge and the new information in the text. However, from observation studies we know that very little time is devoted to explicit or direct instruction of reading and writing strategies. Strategies, such as comprehension monitoring, using graphic organizers and activating prior knowledge must be taught not just as recipes for learning but as flexible learning devices (see Pressley, 2000). Students should come to realize that they can use written language as a foundation for building new concepts and new structures of meaning. By doing so, they will gain more and more inner control and become less dependent on others and more confident in using their own strategies for reading and writing. Finally, the social context of literacy should be emphasized, taking into account sociocultural aspects of development and the concerns of different communities and individuals. It is clear that the development of literacy cannot be seen as an autonomous process of learning universal cognitive or technical skills independently of specific contexts or cultural frameworks. From national surveys it has indeed become clear that the cultural and socioeconomic background of children is an important predictor of their degree of success in school. Literacy seems to correlate with unequal structures and experiences of poverties in societies
(JB)SWL II_f2_1-13
9/24/02
2:36 PM
Page 8
Precursors of Functional Literacy
throughout Europe (e.g., Barton & Hamilton, 1990). Wells (1981, 1985, 1990) has shown that a match between linguistic experience in children’s home and the linguistic demands in the classroom is essential for academic progress. He found that the degree of experience with literate practices in the home had a positive influence on the understanding of the functions and the mechanisms of literacy. In a longitudinal study by Snow, et al. (1991) on the literacy development of lower socioeconomic children it was shown that different home factors predict various literacy skills. The most powerful predictors of children’s word recognition and vocabulary development were the literacy environment of the home, the mother’s education and the mother’s expectations for the child. Variables relating to the emotional organizational dimensions of the family strongly predicted the children’s writing skills. Reading comprehension was related to a wide range of home variables. Furthermore, contacts between parents and teachers regarding academic matters turned out to be related with improved schoolwork and progress in reading. However, this and similar studies are purely correlational. Hence, it is a yet unresolved issue how much of the variance in the home interaction is brought about by the child rather than by the parents. Some children are much more interested in being read to and more eager to engage in language interactions than other children. Children actively “pick their niche”, they are not passive subjects to variation in their environments. Consequently, the traditional correlational studies are likely to overestimate the effects of differences in environmental variation, because they tend to contribute all relevant variance in the environment — to the environment rather than the child. A research paradigm which allows for independent assessement of the children’s contribution and the parents’ contribution to the variance in the home environment would seem to be called for. Of course, what appears as child related variance, may (almost) always be partly an effect of earlier environmental influence.
The present volume The purpose of this volume is to present recent research in the field of the acquisition of functional literacy and its precursors. The volume aims to capture the state of the art in this rapidly expanding field. An attempt is made to clarify the vague and often inconsistent definitions of functional literacy from the perspective of development. Cognitive, linguistic, educational, and social factors of literacy development are all taken into account. The volume consists of three subsequent parts. The first part goes into phonological precursors of literacy development. In this part the focus is on the development of early language precursors of of reading and writing. The cultural foundations of these precursors are explored, and their links with reading development are dealt with in detail. Different psycholinguistic approaches are also proposed to explain the occurrence of literacy problems. In the second part, the scope is on the constraints of reading and writing efficiency at the word level and beyond.
(JB)SWL II_f2_1-13
9/24/02
2:36 PM
Page 9
Functional literacy in a developmental perspective
The acquisition of reading and writing is seen as a result from the interaction between phonological, orthographic, and semantic processes. A crosslinguistic perspective is taken on the role of writing system factors in the acquisition of literacy skills. The final part deals with the role of social and educational factors in literacy acquisition. Starting from a crosscultural perspective, the central issue is how the attainment of functional literacy is dependent on sociocultural variation. The predictors of more advanced levels of literacy development are considered, including foreign language literacy and adult literacy.
Part 1: Roots of phonological awareness Part 1 focuses on the roots of phonological awareness. In the opening chapter, Carsten Elbro and Bolette Pallesen explore the roots of phonological awareness. Their first point is that several factors in both the child and in the child’s cultural environment may contribute to the child’s development of phonological awareness. Their second point is that the child’s phonological represeantations of words set the stage for phonological awareness development. This second point is supported by results from a small scale training study, the first of its kind. In the following chapter, Pieter Reitsma examines the precursors of phonemic awareness in Dutch which is considered to have a shallow orthography — as opposed to the deep English and Danish orthographies. He made an attempt to use some implicit and epilevel tasks at early kindergarten age for predicting later reading achievements in Grade 1. The two tasks chosen were: gating, and nonword repetition. The predicting power turns out to be low which can be explained from the fact that instruction in school with its emphasis on phonics outbalance the individual differences in kindergarten with respect to literacy and phonological processing skills. Astrid Geudens and Dominiek Sandra consider the role of orthographic rime units in Dutch beginning readers in the following chapter. Their results question the role of rimes as functional reading units for beginning readers of Dutch, not even when they are emphasized in the children’s reading curriculum. Hence, their results partly contrast and partly extend previous studies in English into a shallow orthography. In the next chapter, children with manifestations of phonological deficits in developmental dyslexia are considered by Vera Messbauer, Peter de Jong, and Aryan van der Leij. They compared the encoding of phonological representations of familiar and new words in normally developing and dyslexic Dutch children. They found that the Dutch dyslexic children showed phonological deficits similar to those found in studies in English. However, for there must always be a however, long term retention of neither real words nor pseudowords did differentiate the groups. This finding contrasts with a part of the previsous literature and suggests that dyslexia may be associated with difficulties only in the initial phases of phonological encoding.
(JB)SWL II_f2_1-13
9/24/02
2:36 PM
Page 10
Precursors of Functional Literacy
The following chapter by Vincent Goetry, Régine Kolinsky, and Philippe Mousty is concerned with the development of phonological awareness in mono- and bilingual children. They present empirical evidence of the bilinguals’ meta-phonological advantage. This advantage is discussed in the light of the contrasts between the two sets of phonological structures present in bilingual children. In the final chapter of Part 1, Eliane Segers and Ludo Verhoeven question to what extent the quality of phonological representations in language normal and specific language impaired children can be improved by means of natural speech manipulation. In a replication study of Tallal and colleagues, they found that the normal language children performed better than the SLI children in all conditions with no evidence for a positive effect of speech manipulation on children’s word discrimination.
Part 2: Factors in reading and writing efficiency Part two focuses on determining factors in learning to read and write at the word level and beyond. Attempts are presented to bring to light the various constraints on the development of efficient reading and writing. To start with, Sébastien Pacton, Michel Fayol, and Pierre Perruchet explore the acquisition of untaught regularities in French. They report on two studies exploring the impact of graphotactic and morphological regularities on French children’s spelling of nonwords. They show how children learn to extract orthographic knowledge from exposure to written language. In addition, Hye Kyung Yoon, David Bolger, Oh-Seek Kwon, and Charles Perfetti discuss reading in the light of factors that depend on language and writing systems. By comparing Korean and English they provide evidence for language specific variations in orthographic processing. Their findings suggest that while the sensitivity to subsyllabic units may be universal, the form of preferred linguistic units may be language specific. Then, Anna Bosman and Janet van Hell go into the process of spelling of morphologically complex words. In an analysis of the spelling behavior of Dutch students in the intermediate grades they show how phonological, orthographic and semantic constraints are in interaction during word processing. They found changing orthographic appearance and maintaining the phonology of words to be rather flexible adjustment processes in more advanced reading. Such results could be explained in terms of the phonologic coherence model, a recurrentnetwork account. Charles Perfetti and Lesley Hart continue this part of the book. They argue for the importance of lexical (word-level) knowledge in reading comprehension. Specifically, they provide evidence that poor comprehension may be caused by slowness in semantic activation at the lexical level, and they demonstrate how such a simple explanation may render more convoluted hypotheses superfluous, e.g., Gernsbacher’s (1990) hypothesis about failure to suppress previously activated lexical information.
(JB)SWL II_f2_1-13
9/24/02
2:36 PM
Page 11
Functional literacy in a developmental perspective
In the next chapter, Hanna Mäki, Marinus Voeten, Marja Vauras, and Pekka Niemi examine the relationships between reading and writing skills in the intermediate grades. Path analysis was used to study how word recognition and reading comprehension can be used to explain the spelling and composition coherence of young beginning writers. In the final chapter, Aryan van der Leij, Victor van Daal, and Peter de Jong highlight the role of task factors in reading efficiency. They contrast three positions: the task-specific position that restricts dyslexia to impairments in reading only, the task-related position that extends that view to deficits in phonological processing also with spoken language, and the task-independent position based on the assumption that a wider range of processes must be affected in dyslexics, including processes outside the area of reading and phonological processes. In an empirical study they found evidence for the task-specific position.
Part 3: Attaining functional literacy The final part concerns literacy in relation to social and educational context. In the opening chapter of this part, Judith Stoep, Ludo Verhoeven, and Joep Bakker go into the role of parental and teacher commitment in emergent literacy development. They show that children’s literacy skills can be substantially predicted from the literacy environment at home and from the parents’ educational level. They also demonstrate that the responsivity of both parents and teachers significantly contributes to the child’s literacy learning. In the following chapter, Willy van Elsäcker and Ludo Verhoeven examine sociocultural differences in reading skills, reading motivation, and reading strategies. They show significant differences between sociocultural strata for reading comprehension, but not for word decoding. With respect to motivation, minority children showed a more positive reading attitude than their monolingual peers. The minority students also reported a greater use of reading strategies. However, closer inspection revealed an interaction between reading ability and reading strategies: poor readers used more routine strategies whereas good readers used more monitoring strategies. Linda Siegel continues this section with a review of research on the acquisition of literacy in a bilingual context. She explores the role of phonological and orthographic processing in second language reading. Moreover, she addresses the level of phonological and syntactic awareness in second language readers in relation to learning to read. Her data from Canadian studies show that second language learners have no major difficulty in learning to decode. In the next chapter, Liliane Kjellman further addresses the role of motivation in reading acquisition. She shows that younger students tend to value reading more than older students, and that girls have a more positive reading attitude than boys at various age levels. Linguistic background did not seem to affect the students’ self-concept as readers. In a subsequent chapter, Mia Dufva and Marja Vauras go into the English as
(JB)SWL II_f2_1-13
9/24/02
2:36 PM
Page 12
Precursors of Functional Literacy
a foreign language reading of Finnish students who are at risk of reading difficulties. They make clear that by focusing at-risk students’ attention on the sound structure of the foreign language it is possible to enhance foreign language literacy learning. Students showed the largest gains in skills tapping the understanding of grapheme to phoneme correspondences and the phonology of English. In the final chapter, Elisabeth Arnbak examines predictors of adult functional reading skills. In a study of Danish adults and young adults she found the educational level and amount of literacy experience to be the most powerful predictors of reading competence with everyday texts. An interaction between word decoding and text type was also evidenced, indicating different effects of genre in good and poor decoders.
References Adams, M. J. (1990). Beginning to read: Learning and thinking about print. Cambridge, MA: MIT Press. Barton, D. & Hamilton M. E. (1990). Researching literacy in industrialised countries: Trends and prospects. UIE Reports 2. Hamburg: Unesco Institute for Education. Borstrøm, I. & Elbro, C. (1997). Prevention of dyslexia in kindergarten: effects of phoneme awareness training with children of dyslexic parents. In C. Hulme & M. Snowling (Eds.), Dyslexia: Biology, cognition, and intervention (235–253). London: Whurr Publishers. Brady, S. (1997). Abilities to encode phonological representations: an underlying difficulty of poor readers. In B. Blachman (Ed.), Foundations of reading acquisition and dyslexia. Implications for early intervention. Mahwah, NJ: Lawrence Erlbaum. Brady, S. & Shankweiler, D. P. (1991). Phonological processes in reading. Hillsdale, NJ: Erlbaum. Butler, D. L. & Winne, P. H. (1995) Feedback and selfregulated learning: A theoretical analysis. Review of Educational Research, 65, 245–281. Canale, M. & Swain, M.: 1980, ‘Theoretical bases of communicative approaches to second language testing and teaching’, Applied Linguistics, 1, 1–47. Elbro, C., Borstrøm, I. & Petersen, D. K. (1998). Predicting dyslexia from kindergarten. The importance of distinctness of phonological representations of lexical items. Reading Research Quarterly, 33, 36–60. Elbro, C. & Scarborough, H. S. (in press). Early identification. In P. Bryant & T. Nunes (Eds.), International handbook of children’s reading. Dordrecht: Kluwer. Gee, J. (1990). Social linguistics and literacies: Ideology in discourse, Falmer Press, London. Gernsbacher, M. A. (1990). Language comprehension as structure building. Hillsdale, NJ: Erlbaum. Geva, E. & Verhoeven, L. (2000). Basic processes in early second language reading. Scientific Studies of reading, 4, 261–353. (Special issue). Goldman, S. R., & Rakestraw, J. A. (2000). Structural aspects of constructing meaning from text. In M. L. Kamil, P. B. Mosenthal, P. D. Pearson & R. Barr (Eds.), Handbook of reading research, Vol III (311–336). Mahwah, NJ: Erlbaum. Gray, W. S. (1956), The teaching of reading and writing, Scott Foreman, Chicago. Høien, T. & Lundberg, I. (2000). Dyslexia: From theory to intervention. Dordrecht: Kluwer Academic Publishers. Hymes, D. (1971). On communicative competence. Philadelphia: Philadelphia University Press. Koda, K. (1996). L2 word recognition research: A critical review. The Modern Language Journal, 80, 450–460.
(JB)SWL II_f2_1-13
9/24/02
2:36 PM
Page 13
Functional literacy in a developmental perspective Levy, C. & Ransdell, S. (1996). The science of writing: Methods, individual differences and applications. Hillsdale, NJ: Erlbaum. National Reading Panel (2000). Teaching children to read: An evidence-based assessment of the scientific research literature on reading and its implications for reading instruction. Washington, DC: The National Institute of Child Health and Human Development. Palin, M. & Cleese, J. (1974). Bolton is not a palindrome. International Journal of Linguistic Nonsense, 13, 17–35. Perfetti, C. (1998). Learning to read. In: P. Reitsma & L. Verhoeven (Eds.), Literacy problems and interventions (15–48). Dordrecht: Kluwer. Pontecorvo, C. (1997). Writing development. Amsterdam/Philadelphia: John Benjamins. Pressley, M. (2000). What should comprehension instruction be the instruction of? In M. L. Kamil, P. B. Mosenthal, P. D. Pearson & R. Barr (Eds.), Handbook of reading research, Vol III (311–336). Mahwah, NJ: Erlbaum. Scarborough, H. S. (1998). Children at risk for reading disabilities. Phonological awareness and some other promising predictors. In B. K. Shapiro, P. J. Accardo & A. J. Capute (Eds.), Specific reading disability: A view of the spectrum (75–119). Timonium, MD: York Press. Scarborough, H. S. & Dobrich (1994). On the efficacy of reading to preschoolers. Developmental Review, 14, 245–302. Shankweiler, D. & Liberman I. (1989). Phonology and reading disability. Ann Arbor, MI: University Press. Snow, C. E., Barnes, W., Chandler, J., Goodman, I. & Hemphill, L. (1991). Unfulfilled expectations: Home and school influences on literacy. Boston: Harvard University Press. Snow, C. E., Burns, M.S. & Griffin, P. (1998). Preventing reading difficulties in young children. Washington, DC: National Academy Press. Snowling, M. J. (2000). Language and literacy skills: Who is at risk and why? In D. V. M. Bishop & L. B. Leonard (Eds.), Speech and language impairment in children: Causes, characteristics, interventions and outcome (245–260). Hove, UK: Psychology Press. Torgesen, J. K., Wagner, R. K. & Rashotte, C. A. (1997). Prevention and remediation of severe reading disabilities: Keeping the eye in mind. Scientific Studies of Reading, 1, 217–234. Verhoeven, L. (1994). Modeling and promoting functional literacy. In: L. Verhoeven (Ed.), Functional literacy. Theoretical issues and educational implications. Amsterdam/Philadelphia: John Benjamins. Verhoeven, L. (1997). Functional literacy. In V. Edwards & D. Corson (Eds.), Encyclopedia of language and education (Vol. 2): Literacy (127–132). Dordrecht/Boston: Kluwer. Wagner, R. K., Torgesen, J. K. & Rashotte, C. A. (1994). Development of readingrelated phonological processing abilities: new evidence of bidirectional causality from a latent variable longitudinal study. Developmental Psychology, 30, 1, 73–87. Wells, G. (1981). Learning through interaction. Cambridge: University Press. Wells, G. (1985). Language development in the preschool years. Cambridge: University Press. Wells, G. (1990). Talk about text: Where literacy is learned and taught. Curriculum Inquiry, 20, 4, 369–405. Wolf, M. & Bowers, P. (2000). The question of naming speed deficits in developmental reading disabilities: an introduction to the double-deficit hypothesis. Journal of Learning Disabilities, 33, 322–324. Wolf, M. & Obregon, M. (1992). Early naming deficits, developmental dyslexia, and a specific deficit hypothesis. Brain and Language, 42, 219–247. Yaden, D. B., Rowe, D. W. & MacGillivray, L. (2000). Emergent literacy: A matter of perspectives. In M. L. Kamil, P. B. Mosenthal, P. D. Pearson & R. Barr (Eds.), Handbook of reading research, Vol III (425–454). Mahwah, NJ: Erlbaum.
(JB)SWL II_f3_14-32
9/24/02
2:37 PM
Page 14
(JB)SWL II_f3_14-32
9/24/02
2:37 PM
Page 15
Part 1 Roots of Phonological Awareness
(JB)SWL II_f3_14-32
9/24/02
2:37 PM
Page 16
(JB)SWL II_f3_14-32
9/24/02
2:37 PM
Page 17
The quality of phonological representations and phonological awareness: A causal link?1 Carsten Elbro and Bolette R. Pallesen University of Copenhagen
Linguistic awareness is conscious attention to linguistic form rather than the contents that are conveyed by language. Linguistic awareness is thus awareness of language itself, its structure, variations etc., rather than the actual uses of language. Grammars and other less formal descriptions of languages are based on the authors’ linguistic awareness and knowledge and are meant to inspire and qualify the linguistic awareness and knowledge of their readers. However, linguistic awareness exists in many variants and with varying cognitive depths. Just to mention a few early examples, the three year old child is linguistically aware when he or she enjoys or plays with the sounds of his or her language in nursery rhymes, or when the child pays attention to an unusual pronunciation of a word and perhaps even corrects the pronunciation. Language can be analysed at different, partially independent levels. Each of these levels gives rise to a particular form of linguistic awareness: syntactic, lexical, morphemic etc. (figure 1). Linguistic awareness
Lexical
Morphemic
Syllable
Phonological
Rhyme
Figure 1. Some types of linguistic awareness
Syntactic
Phoneme
Pragmatic
(JB)SWL II_f3_14-32
9/24/02
2:37 PM
Page 18
Precursors of Functional Literacy
Phonological awareness is the kind of linguistic awareness that is directed towards the sound of language. Since the sound of language can be analysed into many different units, it is possible to distinguish between many types of phonological awareness, e.g., awareness of syllables, rhyme, or of the smallest significant units, the phonemes. Similarly, the other types of linguistic awareness may be decomposed into more specific forms. For example, lexical awareness may be analysed into awareness of associative word meanings (connotations), awareness of synonyms and antonyms, awareness of word history (etymology) etc. The variety of linguistic units is not the only dimension of variance in linguistic awareness. There are also important differences between various levels of awareness. As pointed out by Gombert (1992) and others, there is a low-level, or implicit, awareness implied in many tasks and language games, such as appreciation of rhymes, or in judgements of similarities between the sounds of words. This is what Gombert refers to as epi-linguistic awareness because it appears to be a more or less implicit consequence of learning a language. Towards the other end of the level-of-consciousness scale there are higher levels of linguistic awareness that require explicit access to linguistic forms and segments of language — such as, for example, phoneme deletion (what is left if you take away the final sound in sex? Expected answer: “sec”) or explicit formulation of linguistic regularities (e.g., which are the differences between the uses of the pronouns who and which?) These kinds of ‘meta-linguistic’ awareness are normally not developed without some kind of formal or informal instruction. A third distinction between types of linguistic awareness stems from the various types of cognitive operations involved. For example, the cognitive operations may be synthesis (e.g., of phonemes into words), segmentation (e.g., of words into syllables), or identification (e.g., which word contains the sound [p], and what type of sound is a [p]?) Language awareness is mostly a relatively useless, albeit entertaining aspect of knowing a language. Children have no obvious use for explicit knowledge of the grammar or phonology of their language. One exception is learning a second language. The learner may find it helpful to know about irregular verbs in French, or about which prepositions that govern which cases in German. There is one other notable exception from the rule that language awareness is useless. That exception is learning to read.
The role of phonological awareness in learning to read One of the insights that are most well documented by reading research is the fact that phonological awareness is an important component in learning to read. Literally dozens of longitudinal prediction studies show that children who enter school with low levels of phonological awareness are much more prone to reading difficulties than children with high levels of phonological awareness. Apart from letter knowledge and precocious reading abilities, phonological aware-
(JB)SWL II_f3_14-32
9/24/02
2:37 PM
Page 19
Phonological representations and phonological awareness
ness is among the strongest predictors of initial success with reading — at least for English (see surveys in Elbro & Scarborough, in press a; Scarborough, 2001). It is also the case that early intervention aimed to improve preschoolers’ phonological awareness have proven very beneficial for their initial reading development (see surveys in National Reading Panel, 2000; Elbro & Scarborough, in press b). These positive effects of intervention strongly suggest that phonological awareness is somehow causally linked to reading. The explanation of the importance of phonological awareness is straight forward. All alphabetic orthographies use letters to represent segments of sounds at the phoneme level. Some orthographies, like the English, are used to represent relatively many words of foreign origin (e.g., Greek, Latin, and French) with spellings that do not correspond in a simple manner to the modern spoken form of the words. But the basic principle remains the same across all alphabetic orthographies: letters represent units of sound at the phoneme level. Notice also that the pronunciations of so-called irregular words like sword, have, yacht etc. are not entirely unpredictable. Only a single or two letters in each word have unusual sounds. Therefore, in order to understand what the single letters represent, the child must have some awareness of the corresponding units of sound in his or her spoken language. Some degree of phoneme awareness is called for. Phoneme awareness does not have to be an explicit, ‘meta-linguistic’ awareness in Gombert’s terms, although such explicit awareness is helpful when learning to read. But the segments of spoken sounds must at least be accessible if complete mappings between orthography and phonology are to be established. This access is necessary even for the development of a “sight word” vocabulary (Ehri, 1999). The remainder of this chapter will mainly deal with the more broad term phonological awareness, rather than with its specific forms. There are two reasons for this. First, various forms of phonological awareness are moderately to strongly intercorrelated. Secondly, and more importantly, the studies concerning the bases of phonological awareness are yet too few to make it possible to draw a detailed picture of the various bases of separate types of phonological awareness. Whilst phonological awareness is one of the strongest predictors of initial success with reading, it is certainly also the case that experience with written language boosts the development of phonological awareness. ‘Shallow’ orthographies with mostly regular words may be better sources for the development of phonological awareness than ‘deep’ orthographies like the English or the Danish. The reason is that children who learn to read in German or Spanish or Finnish — or in most other languages — are offered a very direct insight into the phoneme structure of their spoken words. The precise phoneme structure is directly visible in the written words. Hence, the causal relationship between phonological awareness and learning to read is certainly bi-directional. The importance of phonological awareness for the development of reading raises the obvious questions: where does phonological awareness come from? What determines the development in preschool children? And how come that
(JB)SWL II_f3_14-32
9/24/02
2:37 PM
Page 20
Precursors of Functional Literacy
some children enter school with a much higher degree of awareness of the sounds of words than other children? These questions are the topic of the remainder of this chapter. The structure of the chapter is as follows. First, the possible sources of phonological awareness are discussed. Some sources are suggested which are not usually explicitly mentioned in the reading literature. Special attention is given to the idea that there may be underlying individual differences in the quality with which the sounds of whole lexical items (e.g., words) are stored in the mental lexicon. Secondly, a small scale experiment is summarised which aimed to study a possible causal relationship between quality of phonological representations and the development of phonological awareness.
Possible sources of phonological awareness At the first step of the analysis, one may distinguish between the linguistic material and the cognitive processes involved in phonological awareness (see figure 2). Phonological awareness is based upon and operates on the phonological representations already stored in the child’s mental lexicon. Hence, phonological representations in the mental lexicon are the material that phonological awareness is directed towards. Yet, these representations are not by themselves the process of becoming aware of phonological segments. This process involves a shift of attention from word meanings to the sounds of words; and it is reasonable to expect individual variation in the ease with which children undertake this shift and possibly also in other linguistic processes involved. The processing aspect may become clearer if we take a look at the possible sources that fuel it. Two such sources will be considered here. They are not mutually exclusive, and they do not exclude other sources.
Morphological structure as a source of phonological awareness Certain language structures may be particularly stimulating for phonological awareness development (see also Mattingly, 1987). In English the plural ending -s is a so-called productive component of word formation. Productivity means that the ending may be used in the formation of new word forms (within certain limits). For example, if there were such a thing as a seng, then two of them would probably be two sengs. Even small children use such productive morphemes (as -s or past tense -ed) to construct new word forms. Even though children may not be consciously aware of the segments, let alone the conventions for their application, the segments must have some independent existence in the linguistic system of the child. Notice, however, that all such segments have independent meanings (they are morphemes) and there is certainly no guarantee that the child can abstract sound from meaning (Byrne & Liberman, 1999).
(JB)SWL II_f3_14-32
9/24/02
2:37 PM
Page 21
Phonological representations and phonological awareness
There are two potential advantages of the productive -s. First, somewhere in the language faculty of the child there is a segment of speech which is only one phoneme in size. The point is that the segment is already there, it is not intractably imbedded in words in the mental lexicon of the child. The second potential advantage is that the child can already add or remove the segment from the ends of words. Hence, the morphological status of the -s offers potential support for the development of awareness of the /s/ phoneme, at least in certain contexts. Similarly, awareness of the lax /i/ may be extracted from the -y ending (rainy), /t/ from -ed (wished), and /n/ from -en (rotten). It is well known that English speaking children find it much harder to become aware of sounds within consonant clusters (e.g., the /t/ in stuck) than of consonants in initial or final positions in words and syllables (e.g., the final /t/ in sucked) (e.g., Bowey & Francis, 1991). There may be several reasons for this difference. One is that the second consonant in an initial consonant cluster is never an independent, productive morpheme. The point — that morphological structure may be a determinant of the development of phonological awareness — may perhaps be made more clear by means of a comparison with a very different language. Mla’bri’ is a language spoken by small groups of people in the rain forest in the north of Thailand (Rischel, 1995). In this language infixion is a device in word formation. For example, from the word ɯh (‘to be ablaze’) a related word, rnɯh (‘flames’) may be derived by the insertion of the infix -rn-. Similarly, from laʔ (‘to speak’) it is possible to derive rlaʔ (‘speech, words’) by insertion of a variant, -r-, of the same infix. If morphological structure influences the development of phonological awareness, then the prediction would be that Mla’bri’ speaking children may find it Reading
Phonological awareness
Quality of phonological representations in memory
?
?
Phonological processing
Language structure
Figure 2. Some causes of variation in phonological awareness
Poetry and language games
(JB)SWL II_f3_14-32
9/24/02
2:37 PM
Page 22
Precursors of Functional Literacy
relatively easy to become aware of segments within initial consonant clusters. Presumably, consonant clusters are more transparent for these children because the morphological structure of their language makes them so. However, this prediction has yet to be tested.
Poetry and language games as sources of phonological awareness In addition to the morphological bases of phonological awareness there are cultural traditions that directly highlights certain segments of speech. Rhyme as a poetic device may be practised by individuals and brought to very high levels of mastery — even in illiterates. Some Brazilian street poets, so-called repentistas, show marked abilities to generate rhymes and to judge whether words rhyme, even though they have not attended school and cannot read or write (e.g., Roazzi, Dowker & Bryant, 1993). But notice that such language abilities are highly specific, and do not generalise to, for example, segmentation of words into single phonemes. Rhyming is often seen as a first “natural” step into awareness of sub-lexical units. It may be argued that rhymes are relatively salient parts of words and that most children spontaneously catch on to rhymes once they are exposed to childrens poetry and language games. However, as indicated by results from a series of studies by Seymour, Duncan, and Bolik, and others (1999), children in the first grade generally find it hard to say the common unit (the rhyming part) of rhyming words. It is relatively easier to for them to identify initial sounds shared by two words. Alliteration, not rhyme, was a widely used poetic device in old Germanic poetry from before the introduction of the Romance tradition of rhymes. An example is given here from 10th century Iceland. Völuspá
From Velu’s prophesy
Ár var alda þats ekki var, vara sandr né sær né svalar unnir; iörä fannsk æva né upphiminn, gap var Ginnunga, en gras hvergi.
At the dawn of time when nothing was, was neither sand nor sea no swallow waved, earth did not exist nor heaven gaps and barren emptiness, of grass nothing.
Poems like the above were created in a mainly illiterate culture. The poems were not written down until several generations after their conception. It is conceivable that at least some of the poets knew the rune alphabet. But neither their audience nor the subsequent performers, who kept the poems alive in an oral tradition, were likely to be literate. This is interesting given the advanced poetic devices
(JB)SWL II_f3_14-32
9/24/02
2:37 PM
Page 23
Phonological representations and phonological awareness
employed by this and other early Germanic poems. The lines are connected in pair by means of alliterations between stressed syllables (the whole words are in italics). Two stylistic delicacies may be spotted in the above poem. First, the alliteration in the first two lines is not a simple repetition of phonemes, but a chain of gradually narrowing vowels: á [au], a, and e, compare bim, bam, boom, clink, clank, clonk etc. The assonance is intended and well known from other contemporary poems. Secondly, and most importantly, notice that the alliteration between lines 3 and 4 rests on single consonants, rather than the whole onset: sandr, sær, svalar. Hence, this poem is a brief demonstration of the fact that awareness of single phonemes may occur in a largely illiterate population. Such phonemic awareness is not strictly dependent on exposure to an alphabetic script. There are also present day evidence to support the idea that several types of phonological awareness develop side by side in pre-literates, i.e. children. Dowker (1989) found that 26% of 606 poems produced by children aged 2 to 6 years contained alliteration. It is also noteworthy that the frequency of alliteration declined with age. In sum, there is ample evidence that awareness of both rhymes and single phonemes may develop outside the context of reading. In some cultures, rhyme awareness will develop rapidly because rhyme is a common poetic device. In other cultures, alliteration may take the precedence. Variations in the exposure to poetry, songs, and children’s rimes are, therefore, a possible source of individual variation in the development of phonological awareness.
The quality of phonological representations as a determinant of phonological awareness We may now turn to the material itself on which phonological awareness operates. The material is the phonological representations of words and other lexical items in the mental lexicon. For example, in order to say only the first sound in bus, the word bus must be represented somewhere in the mind of the speaker. It seems reasonable to expect that possible variations in the quality of the phonological representations may vary between words. Early acquired words may be better represented than late acquired words (Garlock, Walley & Metsala, 2001). There may also be important variation between individuals. The hypothesis is that the quality with which the sounds of words are represented in the speaker’s mental lexicon is a determinant (among others) for how easily segments of these representations are accessed (e.g., Elbro, 1996; Elbro, Borstrøm & Petersen, 1998; Metsala & Walley, 1998; Swan & Goswami, 1997). There are many ways to conceptualise quality of phonological representations. Two of these ways have received some attention in the literature from the latest decade. One is presented by the distinctness hypothesis (e.g., Elbro, 1996; Elbro, 1998; Elbro et al., 1998). The distinctness of a phonological representation depends
(JB)SWL II_f3_14-32
9/24/02
2:37 PM
Page 24
Precursors of Functional Literacy
on how much phonetic information that distinguish the representation from its neighbours. Obviously, the amount of phonetic information stored with an item in the mental lexicon is a good indicator of its distinctness. For example, secretary represented as “secretry” is less distinct than “secretary”. Similarly, reading represented as “readn” is less distinct than “reading”. According to the distinctness hypothesis the development of phonological awareness is to some extent determined by the distinctness of phonological representations in the mental lexicon. The study reported below focussed on the distinctness hypothesis. The other conceptualisation to be mentioned here is the segmentation hypothesis (e.g., Fowler, 1991; Metsala & Walley, 1998), which proposes that the level of segmentation of lexical items determines the ease with which children become consciously aware of the corresponding sublexical segments. This hypothesis assumes that phonological representations in small children are initially holistic. Through the pressure from an increasing vocabulary, representations are gradually restructured into smaller segments. An increasing number of minimal pairs (same, tame, fame etc.), for example, may drive a restructuring towards phonemic segments. There is some evidence that the quality of phonological representations is related to reading ability. For example, Fowler and Swainson (1999) reported that good readers were much better than poor readers (in Grades 1 and 4) at identifying and correcting slight mispronunciations of single words — even when receptive vocabulary was accounted for. Further, Elbro, Borstrøm, and Petersen (1998) reported that the distinctiveness with which preschoolers pronounced long, familiar words was predictive of initial reading development — also when individual differences in letter knowledge and phonological awareness were accounted for. More direct evidence for a link between quality of phonological representations and the development of phonological awareness comes from another part of the Elbro et al. (1998) study. It was found that preschoolers’ accuracy of pronunciation of long, familiar words was a relatively strong predictor of the development of phonological awareness during the kindergarten grade. Pronunciation accuracy was also the strongest predictor (out of a dozen pre-school measures of language abilities) of how well children at risk for reading difficulties responded to intensive training of awareness of word-initial phonemes. However, so far the evidence for a link between quality of phonological representations and development of phonological awareness has only been correlational. Experimental studies are needed in order to shed light on a possible causal relationship. But before we turn to such a study, the issue of how to find measures that tap into phonological representations should be given some consideration. After all, phonological representations are not directly accessible, it is not possible to apply a stethoscope directly to the brain and listen to the phonological representations of words.
(JB)SWL II_f3_14-32
9/24/02
2:37 PM
Page 25
Phonological representations and phonological awareness
Measures of quality of phonological representations How to measure the quality of phonological representations is an important issue in testing hypotheses about the relations between quality of phonological representations and the acquisition of phonological awareness. Nonword repetition has been used to tap into the quality of phonological representations (e.g., Wesseling, 1999). By nature, nonwords have no lexical representation, however. So nonword repetition or similar tasks with nonwords are not direct measures of quality of phonological representations. Unfortunately, there is not even a generally agreed upon theory about how already existing lexical representations are used when novel words are encountered and encoded in memory. So it is unclear exactly how nonword tasks may measure quality of phonological representations. The gating paradigm is another measure of phonological representations with a respectable track record. In the gating paradigm, subjects listen to the initial bit of a real word and are asked to try to guess the word. Increasing amounts (gates) of the word are presented until the word is guessed (e.g., Metsala & Walley, 1998; Wesseling, 1999). Whilst it is clear that a high quality representation is more likely than a low quality representation to be activated by a partial cue, several other factors may influence the response to the task, e.g., familiarity with the word and with what it represents, and general willingness to offer a guess. In previous studies we and others have used the ‘poor parrot paradigm’ to elicit the most distinct representations of lexical items in children (Elbro et al., 1998; Fowler & Swainson, 1999). In this task the experimenter names pictured words on behalf of a hand-held parrot at a very low level of distinctness, e.g., “coco” for crocodile. The child’s task is to correct the pronunciation of the parrot. The experimenter repeats the child’s pronunciation to make sure that the child’s intentions are met. Occasionally children know pronunciations that they are unable to produce. We have found that the task is readily understood by every preschool child we have tested so far. Most children enjoy the task; apparently they like to correct the pronunciation of others. So we have continued to use this paradigm because we believe that it taps into the actual speech sounds of the surface structure of the phonological representations. However, children may have somewhat variable ideas about how to “say words as clearly as possible” (as instructed) and they may have different ideas about what it takes to help other persons correct their pronunciations. So we certainly do not claim that the ‘poor parrot paradigm’ is a perfectly valid measure of quality of phonological representations. A comparison between several potential measures of quality of phonological representations is called for.
How can already existing phonological representations be improved? Apart from the problems involved in measuring the quality of phonological representations, another issue is whether it is at all possible to improve the quality
(JB)SWL II_f3_14-32
9/24/02
2:37 PM
Page 26
Precursors of Functional Literacy
of phonological representations of lexical items. In the present case, the question was whether it would be possible to help children increase the level of distinctness of words that they are already familiar with. A major complication here is that common practice when correcting children’s phonological representations targets both phonological representations and segmental awareness simultaneously, e.g., “it is not ‘particliar’ but ‘particular’”. Such a practice cannot be used to study effects of an improvement of phonological representations on segmental phonological awareness because the procedure in itself induces segmental awareness directly. To our knowledge, this issue has not been dealt with at all in the literature. So to avoid a confound between training of phonological representations and phonological awareness the study reported below employed a new technique where children were asked to imitate an artificial voice. In our study, training of the quality of phonological representations was presented to the children as a game in which they should try to learn to talk like a robot. The model voice was computer generated synthetic speech (di-phone synthesis of high intelligibility). The synthetic speech stimuli were individual words pronounced at the highest normal (lexical) distinctness level. The stimuli were tape recorded and played to the children. The task of the child was simply to try to imitate the robot’s voice as closely as possible. All participants were given five trials with each of 13 words. No corrective feedback was given, nor were any sublexical segments enhanced at any time.
The quality of phonological representations affects phonological awareness: An experimental study of a causal link To our knowledge, we report here a first attempt to establish whether quality of phonological representations may play a causal role in the development of phonological awareness, i.e., whether an improvement in the quality of the phonological representations may cause an improvement in phonological awareness. The causal role was studied in a short term, small-scale experiment with preschool children. The study had two main aims: 1. To clarify whether it is possible to help children improve the quality (distinctness) of their phonological representations through short term training. 2. To study the effects of an improvement in the quality of phonological representations on phonological awareness. A further question concerning possible transfer effects of the training was also raised in the study: does distinctness training affect the phonological awareness of the trained words only, or may the effect generalise to other words that bear some structural resemblance to the trained words? For example, improvements of the phonological representation of secretary may generalise to secretarial,
(JB)SWL II_f3_14-32
9/24/02
2:37 PM
Page 27
Phonological representations and phonological awareness
secretariat or private secretary via morphological analogy. The current study included a list of possible transfer words in order to study such generalisation effects. The untrained transfer words shared a root morpheme with the trained words. The study employed a within subjects design. Preschool children’s phonological awareness and the level of distinctness of their phonological representations were assessed individually with words from a pool of 39 words judged to be within the children’s vocabularies. In a second session a week later, the distinctness of the children’s pronunciation of 13 of the words was trained in an imitation task mentioned earlier (“try to says the words in the same way as the robot voice”). The same tests as before training were administered again immediately following training. Thirteen of the 26 untrained words shared a root morpheme with the 13 trained words and thus worked as transfer words. The remaining 13 control words bore no resemblance to the trained words. Twelve children from Danish kindergarten classes participated in the study. They were six boys and six girls aged 6:7 to 7:3 years all native speakers of Danish with no previous history of language delays. In Denmark formal reading instruction does not begin before the age of 7 when children enter first grade. One child was already reading at an elementary level and was thus excluded from the study. Another did not finish post-testing. This left us with ten children from whom we had complete data sets. Before and after training with the ‘robot’ speech, the children’s level of distinctness was measured in the ‘poor parrot’ paradigm with each of the 13 trained words. The child’s most distinct response to each word was selected for scoring. Scoring was based on the vowel segments that are most likely to be reduced in ordinary pronunciation, e.g., the second vowel in chocolate. The child’s pronunciation was judged against the maximally distinct pronunciation as listed in the Unabridged dictionary of Danish pronunciation (Brink, Heger, Lund & Jørgensen, 1991). Fully distinct pronunciations of words were scored as 100%, while minimally distinct, but yet standard pronunciations, were scored as 0%. Non-standard responses were excluded from the analysis. A child’s score was the average distinctness score across the 13 words. For further details of the scoring see Elbro et al. (1998). Training resulted in a dramatic and significant improvement in the level of distinctness (Wilcoxon signed rank test Z = –2.80, p < .01). All children improved their level of distinctness with the 13 trained words. Averages are pictured in figure 3. This increase indicated that it was indeed possible to help preschool children improve their quality of phonological representations — without special attention to single segments of the words. It should be noted that the pretest distinctness level was not associated with short term memory span (digit span) (r = .1, n.s.) The children’s phonological awareness was assessed with five different tests:
(JB)SWL II_f3_14-32
9/24/02
2:37 PM
Page 28
Precursors of Functional Literacy Distinctness (pct.) 80
70
60
50
40
30
20
10
0 Pretest
Posttest
Figure 3. Direct average effect of training on quality of phonological representations (level of distinctness)
– – – – –
syllable tapping, count the number of syllables, syllable deletion, say a word without a certain syllable, syllable identification, say which syllable has been omitted from a spoken word, phoneme deletion, say a word without the initial or final phoneme, and phoneme identification, identify the word (from a selection of pictured words) that has the same the same initial or final phoneme as a spoken word.
Each test had a total of 12 items of which four were selected from the trained words, four from the transfer words (that shared a root morpheme with the trained words), and four from the control words (unrelated to the trained words). No corrective feedback was given. The score was the total number of correct responses. The changes in phonological awareness across all five tasks from pre- to posttest are shown in figure 4. The gain in phonological awareness was significantly greater for the trained words than for the control words (Wilcoxon signed rank test of log odds trans-
(JB)SWL II_f3_14-32
9/24/02
2:37 PM
Page 29
Phonological representations and phonological awareness Items correct 12
10
8
6
4
Trained words Transfer words Control words
2 Pretest
Posttest
Figure 4. Phonological awareness development from pre- to post-test with three types of items. Mean number correct (of 20 of each type of item)
formed scores Z = –2.1, p < .05, see Allerup and Elbro (1998) for details of the statistical procedure). There was a similar increase in phonological awareness across all five subtests. However, the difference between gains with trained words and transfer words was not significant. Neither was the difference between gains with transfer words and control words significant. For some unknown reason the trained words were more difficult at pretest than the transfer and control words. This difference means that at least some of the interaction effect between word type and time of testing may be caused by a regression towards the mean. However, the sizes of the individual gains in distinctness correlated significantly with individual gains in phonological awareness across all five tasks (r = .64, p < .05). Both of these results — the differential effects on trained versus untrained words and the correlation between gains in distinctness and in phonological awareness — suggest that improvements in phonological representations are indeed associated with improvements in phonological awareness — at least for individual words.
(JB)SWL II_f3_14-32
9/24/02
2:37 PM
Page 30
Precursors of Functional Literacy
Discussion This chapter had three major aims. The first, and simplest, was to present the concept of phonological awareness and its critical role in the initial phases of learning to read. The second and not so simple aim was to discuss some likely causes of variation in the development of phonological awareness. We suggested several possible factors. One cause was associated with differences between orthographies, most notably between transparent (‘shallow’) orthographies and less transparent (‘deep’) orthographies such as the English and the Danish orthographies. As others have done before us, we argued that phonological awareness, especially phoneme awareness, is supported more readily by transparent orthographies than by less transparent ones. Another possible source of variation in phonological awareness is morphological structure (and productivity) of the spoken language. This is not a new idea either, but we took it a little further when we suggested that there may not be any universal order in which children become aware of various segments of speech. There may even be languages, such as Mla’bri’, in which morphology supports awareness of the internal structure of consonant clusters. Children speaking these languages may have little difficulty with the analysis of consonant clusters which are almost impossible to analyse for preliterate English speaking children. Thirdly, we pointed out that there are potentially important variations between cultural traditions for poetry and language games. For instance, rhyme is not a universal poetic device, and children may gain access to and awareness of single segments of speech through alliteration. An implication of these sources of variation in phonological awareness is that even awareness of single phonemes may certainly develop outside the context of learning to read or other direct instruction. This is contrary to frequent claims of the opposite (e.g., Bowey & Francis, 1991). Finally, we suggested that there may be important variation in phonological awareness both within and between children associated with the quality of the phonological representations of words and other lexical items. This variation was implied by the third and final aim. The third aim was to discuss ways in which quality of phonological representations may set the stage for the development of phonological awareness — and to present a first study of a possible causal link between the two. The study found some support for the view that the quality of phonological representations of lexical items plays a causal role in phonological awareness development — with the same lexical items. To our knowledge, this is the first experimental study of this role. In the study, we were first able to induce substantial improvements in the quality (distinctness) of phonological representations of already existing lexical items. A simple modelling procedure in which children were asked to imitate a very distinct (computer generated) pronunciation seemed to work well. Next, we found that improvements in phonological awareness with trained words were
(JB)SWL II_f3_14-32
9/24/02
2:37 PM
Page 31
Phonological representations and phonological awareness
significantly greater than improvements with untrained words, suggesting that at least a significant part of the improvement in phonological awareness is caused by a word-specific improvement of the quality of phonological representations. The laboratory nature of the study does not permit long term generalisations to be made, nor does it allow us to make any strong claims about generalisations between words. Further research is needed to shed light on the stability over time of improvements of the quality of phonological representations. Future research might also address the question of whether the causal link between quality of phonological representations and phonological awareness generalises to untrained words. With respect to reading difficulties it would also seem worth while to study whether at-risk children can benefit from training of the quality of phonological representations.
Note . An earlier version of the first part of this chapter was presented at the Second European Graduate School on Literacy Acquisition, Nijmegen, 23–27 August 1999. The experimental work reported in the second half of the chapter was presented at the meeting of the Society for the Scientific Study of Reading, July 22–23, 2000 in Stockholm. This research was supported by a grant from the Nordic Research Councils (contract NOS-S 124811/541) to the first author. The authors would like to thank the children who so readily and with such joy learned to speak in a new way.
References Allerup, P. & Elbro, C. (1998). Comparing differences in accuracy across conditions or individuals: An argument for the use of log odds. The Quarterly Journal of Experimental Psychology, 51A, 409–424. Bowey, J. A. & Francis, J. (1991). Phonological analysis as a function of age and exposure to reading instruction. Applied Psycholinguistics, 12, 91–121. Brink, L., Heger, S., Lund, J. & Jørgensen, J. N. (1991). Den Store Danske Udtaleordbog [‘The Unabridged Dictionary of Danish Pronunciation’]. København: Munksgaard. Byrne, B. & Liberman, A. M. (1999). Meaninglessness, productivity, and reading: Some observations about the relation between the alphabet and speech. In J. V. Oakhill & R. Beard (Eds.), Reading Development and the Teaching of Reading. A Psychological Perspective (157–173). Oxford: Blackwell. Dowker, A. D. (1989). Rhyme and alliteration in poems elicited from young children. Journal of Child Language, 16, 181–202. Ehri, L. C. (1999). Phases of development in learning to read words. In J. Oakhill & R. Beard (Eds.), Reading Development and the Teaching of Reading. A Psychological Perspective (79–108). Oxford: Blackwell Publishers. Elbro, C. (1996). Early linguistic abilities and reading development: A review and a hypothesis. Reading and Writing: An Interdisciplinary Journal, 8, 453–485. Elbro, C. (1998). When reading is “readn” or somthn. Distinctness of phonological representations of lexical items in normal and disabled readers. Scandinavian Journal of Psychology, 39, 149–153.
(JB)SWL II_f3_14-32
9/24/02
2:37 PM
Page 32
Precursors of Functional Literacy Elbro, C., Borstrøm, I. & Petersen, D. K. (1998). Predicting dyslexia from kindergarten. The importance of distinctness of phonological representations of lexical items. Reading Research Quarterly, 33, 36–60. Elbro, C. & Scarborough, H. S. (in press a). Early identification. In P. Bryant & T. Nunes (Eds.), International Handbook of Children’s Reading. Dordrecht: Kluwer. Elbro, C. & Scarborough, H. S. (in press b). Early intervention. In P. Bryant & T. Nunes (Eds.), International Handbook of Children’s Reading. Dordrecht: Kluwer. Fowler, A. E. (1991). How early phonological development might set the stage for phoneme awareness. In S. Brady & D. Shankweiler (Eds.), Phonological processes in literacy: A tribute to Isabelle Y. Liberman (97–118). Hillsdale, NJ: Erlbaum. Fowler, A. E. & Swainson, B. (1999, April). Phonological representation of lexical items in good and poor readers: Evidence from three experimental measures. Paper presented at. The biannual meeting of the Society of Research in Child Development. NM, Albuquerque). Garlock, V. M., Walley, A. C. & Metsala, J. L. (2001). Age-of-acquisition, word frequency and neighborhood density effects on spoken word recognition by children and adults. Journal of Memory and Language, 45, 468–492. Gombert, J. E. (1992). Metalinguistic development. New York: Harvester Wheatsheaf. Mattingly, I. G. (1987). Morphological structure and segmental awareness. Cahiers de Psychologie Cognitive, 7, 488–493. Metsala, J. L. & Walley, A. C. (1998). Spoken vocabulary growth and the segmental restructuring of lexical representations: Precursors to phonemic awareness and early reading ability. In J. L. Metsala & L. C. Ehri (Eds.), Word Recognition in Beginning Literacy (89–120). Mahwah, NJ: Erlbaum. National Reading Panel (2000). Teaching children to read: An evidence-based assessment of the scientific research literature on reading and its implications for reading instruction. Washington, DC: The National Institute of Child Health and Human Development (http://www.nichd.nih. gov/publications/nrp/smallbook.htm). Rischel, J. (1995). Minor Mlabri: a hunter-gatherer language of Northern Indochina. Copenhagen: Museum Tusculanum Roazzi, A., Dowker, A. & Bryant, P. (1993). Phonological abilities of Brazilian street poets. Applied Psycholinguistics, 14, 4, 535–551. Scarborough, H. S. (2001). Connecting early language and literacy to later reading (dis)abilities: Evidence, theory, and practice. In S. Neuman & D. Dickinson (Eds.), Handbook for Research in Early Literacy. New York: Guilford Press. Seymour, P. H. K., Duncan, L. G. & Bolik, F. M. (1999). Rhymes and phonemes in the common unit task: replications and implications for beginning reading. Journal of Research in Reading, 22, 2, 113–130. Swan, D. & Goswami, U. (1997). Phonological awareness deficits in developmental dyslexia and the phonological representations hypothesis. Journal of Experimental Child Psychology, 66, 18–41. Wesseling, R. (1999). Quality of phonological representations: Etiology and remediation of dyslexia. Amsterdam: Free University. Doctoral dissertation. Address University of Copenhagen Dept. of General and Applied Linguistics 86 Njalsgade, DK-2300 Copenhagen S Denmark email:
[email protected]
(JB)SWL II_f4_33-48
9/24/02
2:40 PM
Page 33
Precursors of phonemic awareness Pieter Reitsma1 PI Research – Free University Amsterdam
For many people in Western countries it is entirely commonsensical that writing is a graphic device for transcribing speech. The alphabet is seen as the culmination of a long evolution of writing systems in attempts to represent everything one can say. Spoken languages can be analysed into discrete sound units and in alphabetic writing letters are used to denote these elementary sounds of the spoken language. In order to write a spoken word one only has to analyse the speech sound of the word into its constituent sound segments and to assign the correct letter to each piece. Similarly, reading a written word basically is a matter of sounding out the graphic signs by producing the sounds indicated by the individual letters. Although this conventional view of reading and writing an alphabetic script seems to be appealingly simple or even logical, it is fundamentally wrong. It is indisputable that the development of an alphabet was a significant cultural accomplishment, but other claims referred to above are clearly incorrect. Instead of the invention of a device for transcribing the known properties of speech, the development of the alphabet can also be considered as a means of discovering the basic structures of speech. The alphabetic writing system offered a new model for speech and provided new concepts and categories (e.g., phonemes) for thinking about the structure of spoken language (Harris, 1986; Coulmas, 1989). It is not a way of describing in detail the sounds of a word, rather the letters represent a word by capturing the phonological structure of the word. Phonemes are not sounds themselves but abstract categories of language. Thus, a series of discrete sounds that are commonly associated with letters are not the basic constituents of a spoken word. For example, “buh-oh-kuh-ss” never will be “box” even when spoken at high speed or as smoothly as possible. Nevertheless, the research literature amply demonstrates that the processes involved in learning to read, and the possible explanations for unexpected difficulties in the acquisition of reading skills, are intimately related to learning to use the correspondences between letters and ‘sounds’. This chapter briefly reviews some research with attention to the role of phonemic awareness for beginning reading. Then it is argued that some common tasks for assessing phonemic awareness are not suitable for evaluating a causal role of phonemic awareness for literacy
(JB)SWL II_f4_33-48
9/24/02
2:40 PM
Page 34
Precursors of Functional Literacy
acquisition. Some alternative measures are suggested and a two year longitudinal study is presented in which these measures are used. The results show that there is a dramatic increase in explicit phonemic awareness as soon as reading instruction begins. But previous measures of phonemic awareness do not strongly predict this first grade performance. Only nonword repetition measured at the age of five year does bear some relationship with phonemic awareness at nearly two years later. It is concluded that explicit manipulation of phonological structures appears to be crucially dependent on experiences with letters and written words. Instruction in reading and writing in the Netherlands seems to neutralize or overrule early individual differences in emergent literacy skills and phonological awareness.
The phonological core in a shallow orthography Research of more than two decades has affirmed the importance of phonological awareness (PA) and its relation to reading acquisition. Reviews indicated that the presence of PA is a hallmark characteristic of good readers while its absence is a consistent characteristic of poor readers (see Adams, 1990; Brady & Shankweiler, 1991; Stanovich, 1992). Moreover, there are now many studies linking early phonological awareness or phonological sensitivity to later reading acquisition. Predictive studies compare the relation between PA at an earlier age with subsequent reading achievement at a later age. If consistent and strong correlations are found, then PA is considered to predict later reading achievement. In fact, the findings show that phonological abilities seem to be stronger predictors than such important correlates as intelligence, vocabulary, and listening comprehension, and remain significant predictors of reading achievement even after such factors as intelligence and verbal ability are partialled out (Stanovich, 1992; Wagner & Torgesen, 1987). Of course, predictive evidence does not unambiguously establish causal relations because other variables may be an explanatory or mediating factor. Intervention studies provide another source of support for a causal relation between PA and reading (e.g., Ball & Blachman, 1991; Byrne & Fielding-Barnsley, 1990; Lundberg et al., 1988; Reitsma & Wesseling, 1998). In this type of study, the effect of PA instruction or training on subsequent PA development and reading achievement is assessed with pre- and posttest comparisons. Most intervention studies now have demonstrated that PA instruction indeed has a significant influence on subsequent measures of reading achievement. However, the strength of the conclusions varied. One of the most influential studies is that of Bradley and Bryant (1983). They tested whether difficulties in sound categorization at 4 or 5 year of age, before children started to learn to read, were related to the later development of reading ability. Performance on the sound categorization task was predictive of later reading scores, even when measures of intelligence and memory were taken into account. Bradley and Bryant also provided some children who initially were poor at sound categorization with training in the very same task.
(JB)SWL II_f4_33-48
9/24/02
2:40 PM
Page 35
Precursors of phonemic awareness
Another group of children received in addition to this training in letter-sound correspondences and exercises relating the sound structure of words to their spelling patterns using plastic letters. After training, which was spread over 2 years, the group that had been taught both sound categorization and letter-sound correspondences was in reading scores some 9 months ahead of a control group and this differences was statistically significant. The group that only had been taught to categorize sounds was 4 months ahead, and this was not significantly different. Thus according these findings, training in PA in isolation from reading and spelling skills may be much less effective than training that forms explicit links between phonological skills and experiences in learning to read. Indeed, learning to read is a nontrivial cause in the development of PA as is evidenced by the work on adult illiterates (Morais, Cary, Alegria & Bertelson, 1979). Portuguese illiterate adults were much less sensitive to the phonemic structure of words than their literate colleagues. Thus, it seems that learning to read and write is an effective way to become proficient in PA tasks. In correspondence with this conjecture there is evidence that at least in shallow orthographies, such as Dutch, preschool children generally have no awareness of the phonemic nature of spoken language. However, their first contact with formal reading instruction acts as a powerful trigger for development of PA (Wesseling & Reitsma, 1998). Moreover, PA skills in kindergarten do not effectively predict performance on PA tasks in Grade 1 after reading instruction has began. This absence of predictive power of early PA skills to later reading achievement in Dutch children was also reported by Bast and Reitsma (1998), and De Jong and Van der Leij (1999). Because these results deviate from what commonly is reported in research in English speaking countries, it can be hypothesized that perhaps the specific orthography or the way children are instructed in beginning reading at school may turn out to be crucial. Dutch is an orthography with many simple regular relations between graphemes and phonemes (cf. Reitsma & Verhoeven, 1990). Also, initial reading instruction systematically introduces children to use letter-sound correspondences for word decoding. In this context PA may have less predictive value than when a deeper orthography like English has to be deciphered and when the focus in beginning reading is less on decoding instruction than in the Netherlands. A first goal of the present research then is to see whether these previous findings can be replicated. Basically, the longitudinal study of Wesseling and Reitsma (1998) is extended over a larger period of time and some interesting new measures are added.
Dimensions of phonological awareness and possible precursors Research has shown that PA can be validly and reliably measured through a variety of tasks (e.g., Yopp, 1988). The following tasks have been used in research as indicators of PA: auditory discrimination, blending, counting, deletion, isolation, rhyme, segmenting, substitution, sound categorization, tapping, reversing order
(JB)SWL II_f4_33-48
9/24/02
2:40 PM
Page 36
Precursors of Functional Literacy
of sounds, and word to word matching (cf. Ball & Blachman, 1991, Lundberg et al., 1988; Yopp, 1988). Although many of these tasks have been shown to relate to reading performance, little attention has been paid to the question of processing relationships among the various tasks. Another question is whether phonological processing is a general ability or a compilation of independent abilities? For example, does rhyme belong to the same processing category as more difficult dimensions, such as segmentation. Research has shown that based on significant interrelations among the component abilities the answer seems to be: to some degree, phonological processing is general across tasks (cf. Wagner & Torgesen, 1987; Yopp, 1988; Stanovich, 1992). Important differences among tasks include the degree of abstractness of the phonological representation that are involved, the phonological unit or segment that is in focus, and the degree of conscious access to these units. Also, tasks may be differentially sensitive or age-appropriate. Not all children readily know how to respond to some PA tests. For example, some tests may falsely assume children have experience pronouncing phonemes in isolation. Other tasks may overtax the children’s working memory; for example, Bradley and Bryant’s oddity task requires children to hold several different words in memory while making mental comparisons of phonemes. A developmentally appropriate PA test should have simple instructions and should not assume familiarity with isolated phonemes nor overreach working memory. In this respect, it is also important to distinguish between epiphonological and metaphonological awareness (Gombert, 1992). A similar distinction is made by Stanovich (1992) who suggested that there may be some continuum or hierarchy ranging from ‘shallow’ to ‘deep’ levels of PA. Deeper levels are thought to require more explicit analysis involving smaller-sized phonological units and shallow levels are shallower form of analysis involving larger units. Rhyming skills could be regarded to represent the shallow end of the continuum and segmentation the deep end. Similarly, epiphonological awareness refers to an implicit level of awareness of sublexical segments, but that is not accessible to consciousness or intentional control. In contrast, metaphonological awareness refers to a more explicit organization in which segments become accessible to consciousness and manipulation. The epi-level is normally formed during preschool period and may progress from syllable to onset/rime to phonemes. Gombert’s claim is that a meta-level representation must build on both prior epi-developments and external demand, e.g., literacy acquisition. One example of a deep task or meta-task probably is the phonemic segmentation task. Stahl and Murray (1994) found evidence that segmentation ability is a result of learning to read rather than a cause. Similarly, in a previous study in prereaders we did find a close relationship between knowing some letter-sound correspondences and the ability to segment words (Wesseling & Reitsma, 1998). In the present study, explicit phonemic awareness was tested with two types of tasks, phoneme blending and phoneme segmentation. But blending and seg-
(JB)SWL II_f4_33-48
9/24/02
2:40 PM
Page 37
Precursors of phonemic awareness
menting are prime examples of deep or meta-level PA tasks. Also, as indicated before, one limitation of many studies finding positive evidence for a strong relationship between PA and reading achievement is that performance on some PA tasks seem to be reciprocally related to the development of reading ability. And explicit tasks are regarded to increase the possibility for these reciprocal influences. Therefore, these tasks are inappropriate to predict later reading ability in prereading kindergarten children who are largely unaware of the phonological structure of spoken language. The question then arises: Is it possible, in the absence of reading ability and letter knowledge, to measure the potential to become explicitly phonologically aware? Is it possible to uncover the precursors of explicit PA skills? One way to resolve this problem is, for example, to study children at high risk of dyslexia early in their development to find precursors of later reading disabilities. Scarborough (1990) took this approach and compared retrospectively the early language skills of children who later were diagnosed as dyslexics with those of normal readers. These analyses revealed a changing pattern of language difficulties over time for children who later became dyslexic. At 30 months, there were no differences in vocabulary, but children who went on to be dyslexic used a more restricted range of syntactic devices and made more speech production errors. At 36 and 42 months, both receptive and expressive vocabulary were less well developed than those of controls and syntactic difficulties persisted. At 60 months deficiencies in letter knowledge and PA appeared, naming difficulties persisted but syntactic deficiencies were no longer apparent. Whereas the sample studied was fairly small (20 dyslexics), Scarborough carried out this pioneering study by examining a wide range of language variables over quite a long period of time. At present, there still is relatively little known about the linguistic precursors of later PA and reading difficulties. The purpose of the present study was therefore also to attempt to use some implicit and epi-level tasks at early kindergarten age for predicting later reading achievements in Grade 1. The two tasks chosen were: gating, and nonword repetition. The gating task was previously used by Metsala (1997) to show that spoken word recognition ability of poor readers was worse than that of normally developing readers. The gating paradigm is a task of spoken word recognition that gauges the amount of verbal stimuli a listener requires in order to identify a spoken word. The listener is presented with a partial auditory sample (first sample usually 100 ms) from the beginning of a word and asked to identify which word the sample may be from. After each gate the duration of the sound sample is increased making it easier to correctly identify the word presented. Young children have a smaller vocabulary (lexicon) than older children and adults, so one could predict that younger children would find it easier to recognise words in a gating task due to fact that they have less items in their lexicon to search. However, the consistent finding is that older children and adults require less aural input for correct
(JB)SWL II_f4_33-48
9/24/02
2:40 PM
Page 38
Precursors of Functional Literacy
identification of words than young children. This is taken as evidence that gating tasks indicate the level of lexical differentiation (Metsala, 1997; Walley, 1993). Metsala (1997) also showed that poor readers and young children require more verbal input in order to correctly identify a word. The difference in the amount of input required was most pronounced for words that came from sparse lexical neighbourhoods. Neighbourhood density is a measure of the number of similar sounding words that border specific lexical items. Lexically unique words may identified with less auditory information than words from lexically dense neighbourhoods. This is because for words that have many neighbours it remains unclear which particular word it will be until a point of uniqueness is achieved. For example, it is easier to identify /voice/ than /bag/, because there are less words that begin with /voi/ than with /ba/. Metsala reported that performance on the gating task when using words from sparse lexical neighbourhoods was a good predictor of concurrent reading ability. The gating task seems a promising task to measure the (epi-) level of lexical segmentation. Another task that may be suitable is nonword repetition (Fowler, 1991; Metsala, 1999). Nonwords are in essence just like real words that have not been learned yet. In theory one characteristic of a good quality lexical system is the ability to easily assimilate new and novel words. In order to do this the listener must have accurate encoding, retention and articulation ability, exactly the skills required for repeating nonwords. The nonword repetition task was first used by Snowling (1981) and she found that children with a lower than expected reading ability were worse at repeating nonwords than reading matched younger children. Performance differences on a nonword repetition test are probably related to individual differences in the quality of speech encoding, storage, and articulation, and therefore also to the construction of stable phonological specifications in long-term memory during vocabulary acquisition (Gathercole, 1995; Metsala, 1999). The nonword repetition therefore is an alternative measure of differentiation of the lexicon, relatively unconfounded by letter-sound knowledge, and an age-appropriate task for kindergarten children. In summary, in this longitudinal study an attempt was made to answer the following questions. First, is it possible to replicate the previous findings of Wesseling and Reitsma (1998) who showed a small increase in PA during kindergarten, but a quite substantial increase as soon as reading instruction has began? Second, is the predictive relation of PA measures in kindergarten for later reading achievement relatively low? Third, are the proposed alternative measures for emerging phonological awareness, nonword repetition, and gating, adequate predictors for later scores on explicit PA tasks as blending and segmentation? Additionally, do such measures have predictive value for later reading achievement?
(JB)SWL II_f4_33-48
9/24/02
2:40 PM
Page 39
Precursors of phonemic awareness
Method Participants In this study the data of 37 kindergarten children are presented, 21 boys and 18 girls. Their mean chronological age was 4 years and 11 months (SD = 4.7 months) at the first moment of testing. More children (56) were tested initially, but in this two year longitudinal study some attrition occurred due to various reasons and some children learned to read in kindergarten while no formal instruction in school was provided. The first and second year of testing (T1 and T2) was carried out about 3 and 4 months, respectively, before the end of each kindergarten school year. The third moment of testing (T3) was about 22 months after the first tests, after 5 months of teaching in Grade 1. In the Netherlands in general no instruction in reading or related skills is given during kindergarten, and the teachers of the children participating in this study confirmed this. At the beginning of first grade, however, formal teaching of reading began for all children. The teachers all used the same balanced method with a strong focus on learning to decode the relatively shallow orthography of Dutch language. The children came from four different kindergarten classes from district schools situated in or near the city of Amsterdam. Only those children that spoke Dutch as their first language participated.
Materials and procedure Tests in kindergarten Eight tests were administered to the children in kindergarten each year. The tests were administrated individually in two sessions. Each session took approximately 25 minutes to complete. In the first session the Letter Sound Correspondences, Reading, Phoneme blending and segmentation, and Nonword Repetition were presented. In the second session the Gating and Vocabulary were administrated. Vocabulary. A standardised 98 item receptive vocabulary test was administrated (Verhoeven, Vermeer & Van de Guchte, 1986). Each item consisted of 4 picture alternatives and the child was asked to point to the picture that best represents a word spoken by the experimenter. If the child failed to get more than 2 items correct, out of 8 consecutive items, the test was halted. The score for this test consists of the number of correctly identified items. Letter Sound Correspondences. On a single page 27 lower case graphemes used in Dutch were presented. The graphemes were the 26 letters of the alphabet and included the ij commonly used in Dutch writing. The children were asked to provide the sounds of the letters without a time limit. The score of this test is the number of correctly sounded out or named letters.
(JB)SWL II_f4_33-48
9/24/02
2:40 PM
Page 40
Precursors of Functional Literacy
Reading. Ten common words used in early children’s books were arranged on a single page. The children were asked to read these words. The score consisted of the sum of words correctly identified. Phoneme Blending. Ten digitally recorded (16 bit at 22 Khz) segmented words were presented aurally via a computer. The segments of one word were presented at one second intervals and the child was asked to respond with the whole word. Two items in each of the following word structures were used: VC, CVC, CCVC, CVCC, CCVCC, and in this order the difficulty of items was assumed to increase. Because the presentation did not provide articulatory cues that normally are present when viewing a human speaker, material was chosen for its clarity and ease of recognition. For example, in an attempt to minimise perception difficulties words with difficult to distinguish nasal sounds like /m/ and /n/ were not used. The words used were single syllable, high frequency words that are considered to be known by more than 95% of Dutch children at this age level (Kohnstamm, Schaerlaekens, de Vries, Akkerhuis & Froonincksx, 1981). Before testing began three (C)V(C) words were practiced. The reliability score (Cronbach’s alpha) for this test is 0.91. Phoneme Segmentation. Ten digitally recorded whole words, different from those used in the blending task, were aurally presented by a computer and the child was asked to verbally respond with the phoneme segments. The words had the same structure as the words in the blending test and were all known by children at this age. Three practice items were included. The Cronbach’s alpha test of reliability is 0.89. Gating. Children were asked to identify a word from a partial acoustic signal presented via headphones attached to a computer (Grosjean, 1980; Metsala, 1997). Six different, high frequency, single syllable (c)cvc(c) words known by all 5 and 6 year olds were used. These words had few or no lexical neighbours, i.e. words that had the same onset and vowel but another final consonant (cluster). After all items in the stimulus array were played through, the duration of the audio samples played (the gate) increased. The first presentation was set at 150 ms and the rest of the word was divided into even sections so that on the 8th trial the entire word would be heard. The average step size was 40 ms. The mean number of times a word was identified correctly was used in further analyses. The maximum score was thus 8. The entire test took 8–11 minutes to complete. Cronbach’s alpha is 0.68. Nonword Repetition. The children were asked to repeat as accurately as possible 15 pseudowords. The words were either 2 or 3 syllables in length and each syllable 2 to 4 phonemes. A response was considered correct if all phonemes were articulated. Differences in pronunciation due to dialect were accounted for and were scored as correct. The score for this test is the total number of correctly repeated words. Cronbach’s alpha is 0.69.
(JB)SWL II_f4_33-48
9/24/02
2:40 PM
Page 41
Precursors of phonemic awareness
Tests in Grade 1 In Grade 1 most tests of the pretest were administrated again. The reading test, however, now consisted of a standardized decoding test with CVC words and the score is the total number of correctly identified words within one minute. The letter-sound knowledge test now also included vowel graphemes that are represented by digraphs, e.g., aa, or oe and some other digraphs. It should be noted that the children will not have been taught all these grapheme-phoneme correspondences by the time of testing. The gating test was not presented again. The remaining tests were administered in a single session that took about 18 minutes on average. Results Descriptive statistics for the three test moments are presented in Table 1. The vocabulary scores at the three moments show that there is a steady increase in vocabulary scores. But the means at each moment are completely according to expectations based on published norm data. At T1 in kindergarten the children apparently have little or no knowledge of reading or of letters. They know on average only two letters and cannot read a single word. The blending and segmentation tasks appear too difficult for the majority of first year kindergarten children, 62% are unable to complete a single item on the phoneme blending task and 86% score zero on the phoneme segmentation task. In fact, on the latter test the highest score obtained was only 1, out of 10 items. Nonword repetition appeared not to be too difficult, as was the gating task. At T2 the average scores on all the tests are higher than at T1. All differences between T1 and T2 are statistically significant (p < .01), except the reading and gating scores (both p’s > .10), and phoneme segmentation (F (1,36) = 4.83, p = .034). The kindergarten children now know on average five letters and more children are able to name or sound out at least one letter, only 14% still scoring zero on the letter task. The proportion of children unable to read a single word remains high at 78%. Tasks requiring explicit phonological awareness, such as the blending and the segmentation task, continue to present great difficulty. On the phoneme blending task 34% scored zero, and on the phoneme segmentation task 63% were unable to complete a single item. But there is a slight increase overall. The average scores on nonword repetition and the gating tasks also have increased over those of 12 months earlier, but as noted above the change in gating scores was not significant. At T3 the children then had received reading and writing instruction in Grade 1 for approximately five months. The data in Table 1 show that the children are reading on average almost 22 words per minute correctly and have mastered most of the letters. The scores on the two phoneme awareness tasks, blending and segmentation, show that apart from the most difficult items the children are able to complete these tasks. The minimum score shows that some children are still finding
(JB)SWL II_f4_33-48
9/24/02
2:40 PM
Page 42
Precursors of Functional Literacy Table 1. Descriptive statistics for the three test moments (T1, T2 & T3) Test (number of items or max. score)
M
SD
Min. Max. Observed Observed
T1 (K1)
Vocabulary (98) Letter sound knowledge (27) Reading (10) Phoneme blending (10) Phoneme segmentation (10) Nonword repetition (25) Gating (8)
40.59 2.00 0.03 0.68 0.14 17.30 3.95
17.52 2.71 0.16 1.11 0.35 2.90 0.84
12.0 0.0 0.0 0.0 0.0 10.0 2.0
80.0 9.0 1.0 5.0 1.0 22.0 5.0
T2 (K2)
Vocabulary (98) Letter sound knowledge (27) Reading (10) Phoneme blending (10) Phoneme segmentation (10) Nonword repetition (25) Gating (8)
51.73 5.19 0.14 1.97 .62 20.73 4.15
16.78 4.05 0.42 2.35 1.30 3.47 0.97
25.0 0.0 0.0 0.0 0.0 12.0 2.0
84.0 14.0 2.0 9.0 7.0 25.0 6.0
T3 Vocabulary (98) (Grade 1) Letter sound knowledge (36) Word decoding (1 minute) Phoneme blending (10) Phoneme segmentation (10) Nonword repetition (25)
69.48 31.35 21.62 7.93 7.45 17.52
16.21 5.52 13.86 1.90 1.78 3.13
39.0 13.0 1.0 2.0 3.0 12.0
92.0 36.0 69.0 10.0 10.0 24.0
these tasks difficult though. The average score for the nonword repetition task was lower than the year before and at the same level as the first moment of testing. Scores for all variables are statistically different from the mean score at T2 (p < .01). With regard to the first research question the answer can thus definitely be positive. In Figure 1 the increase of scores for phonological awareness (PA), as the sum of score for blending and segmentation, is shown as a function of moment of testing. It is evident that during kindergarten years the increase in PA is only modest, whereas entrance to Grade 1 does boost the PA scores dramatically. In the same Figure the next panels subsequently show the development of vocabulary, letter-sound knowledge, and nonword repetition. It is interesting to see that the form of developmental changes for PA, i.e. initially little improvement followed by a steep increase, is almost exactly paralled by the results of the lettersound knowledge. Vocabulary at the other hand gradually develops and nonword repetition remains at about the same level. In order to more easily examine the interrelationships among the variables some of the variables were consolidated into one variable. First, for each moment of testing a Literacy variable was created as a composite score of both reading and letter-sound knowledge. Second, Phonological Awareness was considered as a composite of phonemic blending scores and phonemic segmentation. In Table 2 the Pearson correlation coefficients between the experimental measures are presented.
(JB)SWL II_f4_33-48
9/24/02
2:40 PM
Page 43
Precursors of phonemic awareness 20
80
PA
40
Voc
30
LS
NW
60 20
10
40
20
10 20
0
1
2
3
0
1
2
0 3 1 2 moment of testing
3
0
1
2
3
Figure 1. The average scores of Phonological Awareness (PA), Vocabulary, Lettersound knowledge, and Nonword repetition, as a function of moment of testing: (1) Kindergarten 1, (2) Kindergarten 2, and (3) Grade 1
Except gating, all variables are measured at each of the three moments of testing. The correlation coefficients in Table 2 show that the individual differences in vocabulary are quite stable with coefficients near .80. Furthermore, vocabulary is significantly related to nonword repetition in K1. In addition, the cross-age coefficients also demonstrate relationships between vocabulary either measured in K1 or K2, and nonword repetition, and PA in Grade 1. Thus, the data strongly suggest that general language proficiency does have a facilitatory effect on nonexplicit PA tasks such as nonword repetition both in early and later phases and on explicit PA task such as blending and segmentation in Grade 1. Note that the latter relationships can be established across a developmental period of almost two years. Both vocabulary scores in K1, in K2 and in Grade 1 are positively related to the three variables mentioned. However, in intermediate phases at the second half of the second kindergarten year (T2), there are no clear interrelationships between these variables. Performance on the gating task at T1 is related to nonword repetition, but no other relationships can be discerned. At T2, however, the gating scores are not associated with other variables. This corroborates our hypothesis that gating may reveal early phonological processing, but as soon as literacy and more explicit phonological processing procedures come into play, the relationship between gating and other indices of the quality of phonological processing will evaporate.
(JB)SWL II_f4_33-48
9/24/02
2:40 PM
Page 44
Precursors of Functional Literacy Table 2. Pearson’s correlation coefficients for all variables over the 3 moments of testing 1 1 2 3 4 5 6 7 8 9 10 11 12 13 14
2
3
4
Voc1 Lit1 23 Pa1 17 –14 Nw1 42* 01 11 Gat1 01 –18 18 52** Voc2 78** 31 –03 45** Lit2 –07 56** –29 –26 Pa2 21 03 32 22 Nw2 29 30 20 39* Gat2 25 12 08 12 Voc3 82** 34* 16 29 Lit3 19 17 –12 08 Pa3 46** 18 18 44** Nw3 38* 05 10 31
5
02 –11 08 13 11 –12 –01 15 –01
6
7
–02 22 –07 24 03 29 10 77** 08 17 28 48** 04 33* –10
8
20 06 21 21 33* 22
9
10 11
12
13
–12 39* 32 25 –22 06 61** –03 49** 61** 38* –17 32 21 45**
* – p < .05, ** – p < .01 (2–tailed). Voc – vocabulary, Lit – lettersound knowledge and reading, Pa – blending and segmentation, Nw – nonword repetition, Gat – gating.
An interesting finding is that the literacy scores at the two age levels in kindergarten are interrelated (r = .56), but have no relationship with any variable measured in Grade 1. In fact, in the present study it was impossible to predict reading scores in Grade 1 on basis of the measures gathered in kindergarten. Only the explicit PA scores on blending and segmenting and the less direct measures as nonword repetition could be predicted on basis of the scores previously obtained in kindergarten. For example, the PA scores in Grade 1 are closely related (r = .61) to nonword repetition scores one year before. In fact, using all variables in a multiple regression analysis it was possible to account for 71 per cent of the variance of PA in Grade 1, but only vocabulary and nonword repetition in K2 have significant beta weights, .52 and .41, respectively. The pattern of concurrent measures in Grade 1 reveal, however, that literacy and the various PA measures, direct or indirect, appear to be closely related.
Discussion The results show that there is a dramatic increase in the scores on PA tests as soon as reading instruction in Grade 1 begins. These findings corroborate the previous observations of Wesseling and Reitsma (1998). Apparently, explicit manipulation of phonological structures appears to be crucially dependent on experiences with letters and words and on learning about the way they represent sound structures. The individual differences in scores of preceding measures of PA in kindergarten do not appear to predict first grade performance of PA tests. Only nonword repetition measured at the age of 5 does bear some relationship with phonemic aware-
(JB)SWL II_f4_33-48
9/24/02
2:40 PM
Page 45
Precursors of phonemic awareness
ness at nearly two years later. The predictive value of vocabulary for later PA appears to be more stable and generalized. The present pattern of results suggest an important role for both vocabulary and nonword repetition and is compatible with the developmental account of PA by Fowler (1991), Metsala (1999), and Walley (1993). The segmentation theory by Fowler (1991) states that lexical items in the mental lexicon undergo change from holistic sound units in early childhood to a more segmented, ultimately phonemic level of representation in older children. An important event that prompts the reorganisation of lexical representations is the “vocabulary growth spurt” (Walley, 1993) in early childhood. At the age of 2 most children know around 50 words, by age 6, however, children know 8000 to 14000 words. Vocabulary growth thus drives development in spoken word recognition from processes based on relatively holistic lexical representations to adult-like recognition based on segmental/phonemic representations of lexical items. As new words are learned existing lexical items must be refined into a more segmented form in order to ensure sufficient distinguishing features are available to for the purposes of accurate aural recognition. On the assumption that many PA tests require the analysis of words in long-term memory, these tests will also be affected by variables that facilitate segmental development. Thus, a relation between performance on the PA tests and size of children’s vocabulary can reasonably be expected. Moreover, knowledge about the structure of existing words can be used to support the accuracy of nonword repetition. If a larger vocabulary results in more segmentalized representations, this system may lead to better flexibility in rearranging individual phonemes in new patterns and thus more robust representations of nonwords (Metsala, 1999). This theoretical perspective then provides an account for the interrelations between vocabulary, nonword repetition, and explicit PA tests. This line of reasoning may also explain why the performance of nonword repetition is increasing during kindergarten years, along with the growth in vocabulary. However, it does not provide an explanation for the change to Grade 1 where there again is a significant increase in vocabulary, but a significant decrease in nonword repetition. I can only speculate on why children are less accurate at the final test than the one an year before. Although it was noted that the most common error in repeating nonwords are due to substituting (a segment of) a word in place of a nonlexical syllable, no record of errors was made. I would have expected that the first graders made another type of error. The reason could be that the segmental specificity in the representations supporting spoken word recognition and production made enormous improvements due to the acquisition of reading skills, as evidenced by the increases in PA performance, that nonwords are now more fully analyzed in all phonemic segments leading to higher probability of more but smaller errors instead of fewer but larger errors (i.e. whole syllables). The auditory gated word recognition task was an inconsistent predictor of Grade 1 PA or reading acquisition. Nor could a relation with other concurrent variables be found. It should be noted though that at the first moment of testing
(JB)SWL II_f4_33-48
9/24/02
2:40 PM
Page 46
Precursors of Functional Literacy
an significant relation was found between gating scores and nonword repetition. Also, the gating task appeared to be unreliable in a test-retest situation, because no relation was found between gating scores at T1 and at T2. The gating task used in the current study therefore appeared to be inconsistent and previous results of Metsala (1997) could not be extended. The relationships between variables do vary quite a bit over the years. There is no perfect and stable pattern of either high or low correlation coefficients obtained. One of the reasons may be that the measures are not sufficiently precise and reliable. Another is that the measures may be sensitive for certain periods in the development only. Yet, another is that there is no obvious reason to assume that the development of segmental reorganization of the mental lexicon is a constant, serial or gradual process. Variation in development between subjects would results in low correlations between test moments. In addition, the process of reorganization may be another example where a non-linear, dynamic system approach may provide an explanation for instability and turbulence, for discontinuous levels of complexity, and temporary periods of extreme variability (cf. Van Geert, 1994). An important finding is that reading achievement in Grade 1 is only directly related to concurrent measures of PA. No direct predictors from kindergarten could be found. Thus, measures of emerging literacy in kindergarten also appear not to be related to individual differences in reading skill in Grade 1. This finding is completely in line with previous Dutch findings (Bast & Reitsma, 1998; De Jong & Van der Leij, 1999; Wesseling & Reitsma, 1998). Again, the only explanation that can reasonably be provided is that instruction in school with its emphasis on phonics outbalance the individual differences in kindergarten with respect to literacy and phonological processing skills completely. Thus, in a context of an orthography with relatively simple relationships between graphemes and phonemes and an initial reading instruction in which the letter-sound correspondences are stressed and practised in the context of the decoding of words, these instructions tend to convincingly nullify previous differences in emergent literacy skills and PA.
Note . Thanks to the pupils and the staff of the elementary schools involved in this research, and to Ralph Wesseling and some of his students for their help in gathering the data.
References Adams, M. J. (1990). Beginning to read: Thinking and learning about print. Cambridge, MA: M.I.T. Press. Ball, E. W. & Blachman, B. A. (1991). Does phoneme segmentation training in Kindergarten make a difference in early word recognition and developmental spelling? Reading Research Quarterly, 26, 49–66. Bast, J. W. & Reitsma, P. (1998). Analyzing the development of individual differences in
(JB)SWL II_f4_33-48
9/24/02
2:40 PM
Page 47
Precursors of phonemic awareness terms of Matthew effects in reading: results from a Dutch longitudinal study. Developmental Psychology, 34, 1373–1399. Bradley, L. & Bryant, P. E. (1983). Categorizing sounds and learning to read — a causal connection. Nature, 301, 419–421. Brady, S. & Shankweiler, D. (Eds.). (1991). Phonological processes in literacy. Hillsdale, N.J.: Erlbaum. Byrne, B. & Fielding-Barnsley, R. (1990). Acquiring the alphabetic principle: A case for teaching recognition of phoneme identity. Journal of Educational Psychology, 82, 805–812. Coulmas, F. (1989). The writing systems of the world. Oxford: Basil Blackwell. De Jong, P. F. & Van der Leij, A. (1999). Specific contributions of phonological abilities to early reading acquisition: results from a Dutch latent variable longitudinal study. Journal of Educational Psychology, 91, 450–476. Fowler, A. E. (1991). How early phonological development might set the stage for phoneme awareness. In S. Brady & D. Shankweiler (Eds.), Phonological processes in literacy. A tribute to Isabelle Y. Liberman (97–118). Hillsdale, NJ: Erlbaum. Gathercole, S. E. (1995). Nonword repetition: More than just a phonological output task. Cognitive Neuropsychology, 72, 857–861. Gombert, J. E. (1992). Metalinguistic development. London: Harvester Wheatsheaf. Grosjean, F. (1980). Spoken word recognition and the gating paradigm. Perception & Psychophysics, 28, 267–283. Harris, R. (1986). The origin of writing. London: Duckworth. Kohnstamm, G. A., Schaerlaekens, A. M., de Vries, A. K., Akkerhuis, G. W. & Froonincksx, M. (1981). Nieuwe Streeflijst Woordenschat voor 6–jarigen.[New word vocabulary list for six year olds]. Lisse: Swets & Zeitlinger. Lundberg, I., Frost, J. & Peterson, O.-P. (1988). Effects of an extensive program for stimulating phonological awareness in preschool children. Reading Research Quarterly, 23, 263–284. Metsala, J. L. (1997). Spoken word recognition in reading disabled children. Journal of Educational Psychology, 89, 159–169. Metsala, J. L. (1999). Young children’s phonological awareness and nonword repetition as a function of vocabulary development. Journal of Educational Psychology, 91, 3–19. Morais, J., Cary, L., Alegria, J. & Bertelson, P. (1979). Does awareness of speech as a sequence of phones arise spontaneously? Cognition, 7, 323–331. Reitsma, P. & Verhoeven, L. (Eds.), Acquisition of reading in Dutch. Dordrecht: Foris. Reitsma, P. & Wesseling, R. (1998). Effects of computer-assisted training of blending skills in kindergartners. Scientific Studies of Reading, 2, 301–320. Scarborough, H. S. (1990). Very early language deficits in dyslexic children. Child Development, 61, 1728–1743. Snowling, M. J. (1981). Phonemic deficits in developmental dyslexia. Psychological Research, 43, 219–234. Stahl, S. A. & Murray, B. A. (1994). Defining phonological awareness and its relationship to early reading. Journal of Educational Psychology, 86, 221–234. Stanovich, K. E. (1992). Speculations on the causes and consequences of individual differences in early reading acquisition. In P. Gough, L. Ehri & R. Treiman (Eds.), Reading Acquisition (307–342). Hillsdale, NJ: Erlbaum. Van Geert, P. (1994). Dynamic systems of development: Change between complexity and chaos. New York: Prentice Hall/Harvester Wheatsheaf. Verhoeven, L., Vermeer, A. & Van de Guchte, C. (1986). Taaltoets Allochtone Kinderen [Language test for Children]. Tilburg: Zwijsen. Wagner, R. K. & Torgesen, J. K. (1987). The nature of phonological processing and its causal role in acquisition of reading skills. Psychological Bulletin, 101, 192–212. Walley, A. C. (1993). The role of vocabulary development in children’s spoken word recognition and segmentation ability. Developmental Review, 13, 286–350.
(JB)SWL II_f4_33-48
9/24/02
2:40 PM
Page 48
Precursors of Functional Literacy Wesseling, R. F., & Reitsma, P. (1998). Phonemically aware in a hop, skip, and a jump. In P. Reitsma & L. Verhoeven (Eds.) Problems and Interventions in Literacy Development (81–94). Dordrecht: Kluwer. Yopp, H. (1988). The validity and reliability of phonemic awareness tests. Reading Research Quarterly, 23, 159–177. Address PI Research PO Box 366 1115 ZH Duivendrecht The Netherlands
(JB)SWL II_f5_49-67
9/24/02
2:41 PM
Page 49
The role of orthographic onset-rime units in Dutch beginning readers Astrid Geudens* Dominiek Sandra** University of Antwerp
One of the basic skills each beginning reader has to acquire is a process of mapping letters onto sounds, a process that is called phonological decoding (Adams, 1990). An important research question concerns the nature of the orthographic units which form the input to this decoding process. Onsets and rimes have often been proposed as candidate units (see Goswami & Bryant, 1990). The (optional) onset of a syllable is the initial consonant or consonant cluster (e.g., /spl/ in split), the rime is the sequence of the vowel and subsequent consonants (e.g., /it/ in split). The research on onsets and rimes as units in reading was stimulated by the demonstration that onsets and rimes are natural units in spoken words (Goswami & Bryant, 1990; Treiman, 1989, 1992). For instance, children find it much easier to pick out the odd item in a spoken word list like /bi:n/, /mi:n/, /pi:k/, /li:n/, in which similarities are based on the rime, than in a spoken word list like /bi:n/, /bi:f/, /li:k/, /bi:t/, in which similarities are based on the onset and the vowel, the so-called body (cf. Bradley & Bryant, 1983; Kirtley, Bryant, MacLean & Bradley, 1989). Further support for these phonological units was offered by McClure, Bisanz, and Ferreira (1996), who found that kindergarten, first and second grade children could blend speech segments more easily when they coincided with the onset and the rime than when they were individual phonemes. Subsequent research with English readers suggested that onsets and rimes are also orthographic recognition units,1 i.e. that readers parse written syllables at the onset-rime boundary. A number of experiments converged on this conclusion across a variety of techniques, such as visual distortion (Santa, 1976; Treiman & Chafetz, 1987), partial identity priming (Bowey, 1996) and frequency manipulation (Bowey & Hansen, 1994; Treiman, Goswami & Bruck, 1990). For instance, children read pseudowords with high-frequency rimes like tain more accurately than pseudowords with low-frequency rimes like taich. Body frequency did not contribute to the orthographic rime effect.
(JB)SWL II_f5_49-67
9/24/02
2:41 PM
Page 50
Precursors of Functional Literacy
Different explanations have been offered for the existence of onset-rime effects. According to one account the salience of onsets and rimes in children’s phonological awareness triggers a preference for their orthographic counterparts. This will lead beginning readers to parse words before the vowel grapheme (Goswami & Bryant, 1990; Goswami, 1993). In a series of experiments Goswami demonstrated that children determine the pronunciation of a novel word (e.g., peak) by making an analogy between its orthographic rime and the rime of a visually and auditorily presented clue word (e.g., beak). Such rime-based analogies would precede phoneme-based analogies because the phonological units involved are more easily accessible than phonemes. Recently, however, a number of studies suggest that Goswami’s clue word experiments may overestimate the extent to which orthographic rime analogies are used in spontaneous and normal reading and that effects may be task dependent (e.g., Booth & Perfetti, 2002; Bowey, Vaughan & Hansen, 1998; Bowey, 1999; Brown & Deavers, 1999; Savage & Stuart, 1998). According to a different explanation, orthographic onsets and rimes emerge from the pattern of relationships between the orthographic and phonological structures of a language (Treiman, Mullenix, Bijeljac-Babic & Richmond-Welty, 1995). If phonological decoding of written English had to operate on a graphemeby-grapheme basis, it would suffer greatly as a result of the one-to-many mappings of graphemes onto phonemes (compare beak and head). Treiman et al. (1995) reported lexical-statistic data for English, showing that the relationships between orthography and phonology are most consistent at the level of the orthographic rime, making this substring a useful decoding unit (see also Stanback, 1992). In line with these data, several studies have found that the reading speed and accuracy of skilled adult English readers is influenced by the (pronunciation) consistency of the orthographic rime (Glushko, 1979; Jared, McRae & Seidenberg, 1990). Moreover, the use of the orthographic rime seems to increase with reading skill and experience with the writing system. Children start to pick up the advantages of grouping and classifying words on an onset-rime basis once they have built up an adequately extensive sight vocabulary (Bowey & Underwood, 1996; Leslie & Calhoon, 1995; Treiman et al., 1995). The orthography of Dutch offers an ideal case for testing these two accounts. Dutch, unlike English, is fairly transparent at the grapheme-phoneme level, such that most words can be decoded grapheme by grapheme. If the nature of the orthography-phonology interface determines the decoding units, Dutch spelling would not cause the emergence of orthographic onsets and rimes (cf. Goswami, Gombert & De Barrera, 1998; Wimmer and Goswami; 1994, Martensen, Maris & Dijkstra, 2000). This is not to say that higher-order orthographic substrings could not emerge in Dutch, for instance as the result of mere exposure frequency. However, this does not automatically imply a special decoding status for onsets and rimes. On the other hand, if orthographic rimes result from the “projection” of phonologically salient units onto the orthographic forms of words, onset-rime effects can be found with Dutch readers as much as with English readers. Dis-
(JB)SWL II_f5_49-67
9/24/02
2:41 PM
Page 51
Onset-rime units in Dutch
tributional evidence by Martensen et al. (2000) supports this hypothesis. In a statistical study, Martensen and her colleagues calculated the possible phoneme combinations for sublexical units in Dutch. In their corpus bodies covered 28% of all possible onset-nucleus combinations, whereas rimes covered only 13% of all possible nucleus-coda combinations. This means that in Dutch distributional constraints on the rime are stronger than those on the body (cf. Treiman et al., 1995 in English). Previous experimental findings suggest that Dutch beginning readers do not use orthographic onset-rime decoding units (Geudens & Sandra, 1999; Reitsma, 1990, 1997; Sandra & Geudens, 1999; Theloosen & Van Bon, 1993; Van Daal, Reitsma & Van der Leij, 1994; Van den Bosch, 1991; but see Assink, Kattenberg & Wortmann, 1998). Van den Bosch (1991), for instance, did not find any preference for an onset-rime structure in a visual segmentation task. First and third graders were equally disturbed by the presence of an asterisk or a shift in colour and letter size at the onset-rime boundary (e.g., b*oot) as at the body-coda boundary (e.g., boo*t). However, a potentially important factor that has often been neglected in previous research is the influence of the reading method. This instructional factor may contribute to the emergence of reading units as well. In a study by Duncan, Seymour, and Hill (1997), beginning readers decoded pseudowords by using their letter-sound knowledge rather than using rime-sized units, independently of their pre-school rhyming skills. The authors attributed this effect to the teachers’ emphasis on the identification of words by their initial letters. Given that Dutch spelling does not require children to decode at a level above graphemes (see Martensen et al., 2000), orthographic units may nevertheless emerge if they are emphasised in reading manuals. As it happens, most phonics methods for Dutch make abundant use of so-called structured word lists, in which multiletter strings are emphasised. Words in such lists differ in one letter position only (e.g., rime-based series: hand, sand, band; body-based series: cat, cap, can). Apart from focusing the child’s attention on individual graphemes, such lists may also underscore the value of multiletter groups, and thus cause the development of higher-order orthographic substrings (but see Reitsma, 1990; Van den Broeck, 1997). In previous research with Dutch readers the children were taught to read with a method in which structured word lists are about equally often body-based as rime-based (i.e., Veilig Leren Lezen; Mommers, Verhoeven & Van der Linden, 1990). Such a method might train children on two types of subsyllabic letter groupings, thus masking a possible effect of phonologically salient rimes.2 Experiment 1 was designed to test whether instructional effects may account for the earlier null effects in Dutch. We compared children who were taught to read with the same method as those in previous experiments with children who used a method in which rime-based word lists by far outnumber body-based lists (Leeslijn; De Baar, 1995). The orthographic rime preference in Leeslijn is undeniable and even stressed in the teacher’s manual (with explicit references to the international literature). If the
(JB)SWL II_f5_49-67
9/24/02
2:41 PM
Page 52
Precursors of Functional Literacy
effect cannot even be obtained with children whose instructional method strongly biases them towards using orthographic rimes, this would be a strong demonstration that young readers of Dutch do not use such strings as decoding units (see Geudens & Sandra, 1999). Besides reading instruction we will also focus on the children’s reading expertise. As pointed out above, orthographic rimes may appear only after sufficient reading experience. We will operationalise this concept by distinguishing between fluent and non-fluent readers.
Experiment 1: Visual segmentation Experiment 1 was based on the assumption that breaking up reading units should be more disruptive to the reading process than segmenting between such units. The purpose was to compare children’s naming speed on intact pseudowords (e.g., wot) to their speed on pseudowords that were segmented either at the onset-rime boundary (e.g., w ot) or at the body-coda boundary (e.g., wo t).
Method Participants Three groups of first graders (mean age: 6 years and 6 months), participated in the experiment after having received 8–9 months of formal reading instruction. All children were native speakers of Dutch. We tested a group of children (n = 27) who were taught to read with the method Veilig Leren Lezen (Mommers et al., 1990). This method stresses grapheme-to-phoneme correspondences (GPCs) and uses many structured word lists in which onset-rime and body-coda structures are equally emphasized. We will refer to this group as the unbiased group. We also tested two groups (n = 36 each) who were taught to read with the phonics method Leeslijn (De Baar, 1995), which stresses GPCs and has a strong preference for orthographic onset-rime units in its structured word lists. We refer to this group as the onset-rime biased group. We tested two groups in order to replicate the effect with two independent samples (cf. below). Materials and design The experiment was built on a 2 × 3 (Reading Method × Segmentation Condition) design, in which Reading Method was a variable defining participant groups and Segmentation Condition a variable defining materials groups. The items were CVC and CCVC pseudowords that could be read by applying simple GPC rules. The pseudowords were presented in three conditions: 1) intact (INT; e.g., kis), 2) segmented by two spaces between onset and rime (O/R; e.g., k is), and 3) segmented by two spaces between body and coda (B/C; e.g., ki s). A double space was chosen as a segmentation marker since previous research by Van den Bosch (1991)
(JB)SWL II_f5_49-67
9/24/02
2:41 PM
Page 53
Onset-rime units in Dutch Table 1. Composition of the experimental items in Experiment 1 kis
zis
his
zif
hif
kif
hig
kig
zig
had shown that children sometimes interpret visual markers like double slashes (//; see Treiman & Chafetz, 1987) and asterisks (*; see Van den Bosch, 1991) as missing letters (thus turning the test into a cloze task). The double space would not elicit such perceptions. To compose the items we selected one vowel, (e.g., 〈i〉) and constructed three pseudowords with different bodies and different rimes (kis, zif, hig). By rotating the onsets we obtained a matrix of nine items. Table 1 gives an example of one matrix. We constructed six matrices of nine pseudowords, avoiding repetition of onsets, rimes, bodies, and codas across matrices, resulting in a total of 54 items. Conditions were assigned to the three columns of each matrix. Thus, the same phonological material (same onsets, rimes, bodies and codas) appeared in each condition. Each child saw the nine items of a matrix but the assignment of conditions to columns was counterbalanced across three lists, each of which was presented to one third of the participants in a group.
Procedure To determine the children’s reading level we first administered a standard reading test (Een-minuut-test, [One-minute-test]; Brus & Voeten, 1972). This test consists of a list of unrelated real words (n = 116), which have to be read aloud. The number of words read correctly in one minute is the reading score. On the basis of these test scores children were assigned to the three lists so as to guarantee close matching on this measure. The pseudowords were randomly presented one at a time on a computer screen (Arial, 36 pt.). Each trial started with a short auditory signal, immediately followed by a pseudoword. The experimenter told the children that they would have to read “funny made up words” with no meaning and that these had to be read aloud as a whole, ignoring possible spaces between letters. In a practice session the children were acquainted with the materials and urged to name the items as fast and accurately as possible. Naming latencies were measured with millisecond accuracy by a voice-key attached to the computer. The experimenter registered whether the stimulus was read correctly and equipment failures were noted. Time-outs occurred at 5 seconds.
(JB)SWL II_f5_49-67
9/24/02
2:41 PM
Page 54
Precursors of Functional Literacy Table 2. Mean latencies (M in ms), error percentages (enclosed in parentheses) and standard errors (SE) in the three segmentation conditions for all participants in the unbiased reading method (n = 27) and the two biased reading methods (n = 36 each) in Experiment 1 Reading Method
Segmentation Condition INT
Unbiased Biased Sample 1 Sample 2
O/R
B/C
M
SE
M
SE
M
SE
1874 (9.4%)
133
1849 (8.3%)
120
2022 (9.4%)
136
2003 (10%) 1624 (9.4%)
164 115
1991 (8.3%) 1709 (8.9%)
94 116
2126 (7.8%) 1787 (8.9%)
107 133
Note. Unbiased refers to the children who are taught to read with a reading method in which orthographic bodies and rimes are equally stressed. Biased refers to the two samples of children who are taught to read with a reading method that has a strong preference for orthographic onset-rime structures above body-coda structures. INT = intact condition; O/R = onset-rime condition; B/C = body-coda condition.
Results Table 2 shows mean latencies and error percentages. In order to measure the effect of reading method we performed Analyses of Variance (ANOVAs) on the participant and item data for a 2 × 3 (Reading Method × Segmentation Condition) design, analysing both latencies and error rates. Reading Method was a betweenparticipants and within-items factor. Segmentation Condition was a within-participants and within-items factor. Two ANOVAs were performed; one for each rime-biased group (the unbiased control group being the same in both analyses). Because all theoretically relevant effects were the same in the two analyses, we will report the data for one comparison only. The main effect of Segmentation Condition was significant, F1(2, 122) = 19.65, p < .0001; F2(2, 106) = 7.85, p < .001. The interaction effect between Reading Method and Segmentation Condition was non-significant (Fs < 1). This nonsignificance, and its replication across two groups of rime-biased readers, demonstrates that there is no reason to keep these reader groups separate. Hence, we performed all subsequent analyses on the entire sample of 99 participants. One-way ANOVAs, treating Segmentation Condition as a within-participants and within-items factor, were performed on the mean latencies. The main effect of Segmentation Condition was significant, F1(2,196) = 28.26, p < .0001; F2(2,106) = 10.50, p < .0001. Planned comparisons showed that responses were 112 ms faster in the O/R condition than in the B/C condition, F1(1,196) = 36.76, p < .0001; F2(1,106) = 15.57, p < .0001. Mean latencies in the O/R condition and the INT baseline condition differed by one millisecond only (Fs < 1). Responses in the B/C condition were 113 ms slower than in the INT condition, F1(1,196) = 47.36, p < .0001; F2(1,106) = 15.92, p < .0001. The effect of Segmentation Condition was non-significant in the error analysis (Fs < 1).
(JB)SWL II_f5_49-67
9/24/02
2:41 PM
Page 55
Onset-rime units in Dutch
During the actual experiment the experimenter (AG) had noted that there were two types of readers: those who read rather fluently and those who were still labouring hard at the grapheme level. The latter could be called letter-by-letter readers since they often subvocally said the letters of the item before pronouncing it.3 Several researchers have pointed out that orthographic rime units only emerge after sufficient reading experience (e.g., Bowey & Underwood, 1996; Leslie & Calhoon, 1995; Treiman et al., 1995). In order to find out whether reader fluency was an important factor in this experiment, we performed a median-split analysis with a group of fluent (n = 50) and a group of non-fluent (n = 49) readers (Reading Speed).4 We actually performed two such analyses, one based on the results of the standard reading test and another one based on the average latencies in the INT condition. Both analyses converged on the same pattern of results. We will report the analyses based on the INT latencies only. There are two reasons for this choice: reading latencies make finer discriminations among readers than accuracy scores in a one-minute test, and the INT scores discriminate among readers on the basis of pseudoword performance whereas the standardised reading test measured word performance.5 An ANOVA for a 2 × 3 (Reading Speed × Segmentation Condition) design revealed a significant interaction between Reading Speed and Segmentation Condition, F1(1,194) = 10.82, p < .0001; F2(2, 106) = 8.99, p < .0005. The separate ANOVA on the means of the fluent readers did not show a preference for the O/R condition in this reader subgroup at all: both segmentation conditions took longer to read than the INT items (INT versus O/R: F1(1, 98) = 10.28, p < .005, F2(1,106) = 10.99, p < .005; and INT versus B/C: F1(1,98) = 16.14, p < .0001; F2(1,106) = 14.56, p < .0005) with a similar disturbance for O/R and B/C (Fs < 1). In contrast, non-fluent readers benefited from the O/R condition in comparison to the B/C condition, F1(1, 96) = 47.96, p < .0001; F2(1, 106) = 21.82, p < .0001, and in comparison to the INT baseline condition, F1(1, 96) = 10.43, p < .005; F2(1, 106) = 5.35, p < .05. Figure 1 clearly illustrates that the O/R condition — not the B/C condition — was the cause of the interaction between the non-fluent and the fluent reader group: the onset-rime effect changed from inhibitory for fluent readers to facilitatory for non-fluent readers.
Discussion This experiment addressed the question whether an instructional effect may have played a role in previous research on the use of orthographic onsets and rimes in Dutch. The results do not indicate such an effect. Onsets and rimes were treated equally by readers who were taught to read with an onset-rime biased reading method and by readers who were taught to read with an unbiased reading method. One would expect different effects of the O/R and B/C segmentation conditions in the former method. Furthermore, the dissociation between fluent and non-fluent readers emphasises the importance of individual differences when studying decoding units. The pattern of results is easily summarised: visual segmentation delays pronunciation
(JB)SWL II_f5_49-67
2:41 PM
Page 56
Precursors of Functional Literacy
2500 2300 Mean RT (ms)
9/24/02
2100
INT O/R B/C
1900 1700 1500 1300 1100 fluent readers
non-fluent readers
Reader type Figure 1. Mean latencies (in ms) in each Segmentation Condition for fluent and nonfluent readers in Experiment 1 Note. INT = intact condition; O/R = onset-rime condition; B/C = body-coda condition.
times relative to the intact control condition, with the single exception of the O/R condition in the non-fluent reader group. We shall discuss the findings for the fluent and non-fluent reader groups respectively. The rationale behind visual segmentation is that segmenting letter strings in accordance with readers’ decoding units will result in faster processing than segmenting within reading units. Hence, the finding of a non-significant difference between the two segmentation conditions in the fluent reader group indicates that these readers have no preference for an onset-rime structure. However, the data are compatible with an account in which both the (C)CV and VC parts in (C)CVC pseudowords play a functional role. Such an account predicts that segmenting pseudowords either before the vowel (O/R condition) or after the vowel (B/C condition) will destroy a multiletter pattern, leading to delayed processing relative to the INT condition (in which no multiletter pattern is broken up). Note that such an account is still incompatible with the onset-rime hypothesis for Dutch which predicts a processing advantage for the O/R segmentation condition relative to the B/C condition. Experiment 2 will be explicitly designed to address the idea that both initial and final multiletter strings are activated in reading. In the non-fluent reader group we did find a processing advantage for the O/R segmentation condition, even compared to the INT condition. This result is not incompatible with a preference for orthographic onsets and rimes in this reader group. The effect might mean that orthographic rimes are the first multiletter units to be developed in reading acquisition (hence are the only decoding units to be discovered in the non-fluent readers), and that other multiletter units (like CVs) emerge later. However, there are reasons to doubt whether the effect in the O/R condition reveals the existence of orthographic onset-rime decoding units in
(JB)SWL II_f5_49-67
9/24/02
2:41 PM
Page 57
Onset-rime units in Dutch
the non-fluent readers’ mental lexicon. As mentioned earlier, many of the readers in this group were letter-by-letter readers, which makes their use of higherorder decoding units rather unlikely. As a matter of fact, if individual letters form their primary orthographic decoding units, one can account for the processing advantage in the O/R condition without having to take recourse to higher-order orthographic units. In the O/R condition, the first segment matches the size of the reading unit used by these readers, whereas a mismatch occurs in the B/C condition (hence: facilitation versus inhibition). Once on the right track, the readers in the O/R condition may experience less difficulties segmenting the remaining rime string than in the B/C condition, where the leftmost segment in the visual stimulus suggests the wrong-sized decoding unit. An alternative account would also be predicated on the letter-by-letter decoding strategy of non-fluent readers. These readers’ advantage in the O/R condition might show that the phonemes which are activated during letter-by-letter decoding are blended more easily when the visual stimulus suggests an integration according to the phonological onset-rime distinction than when it suggests a different integration (see Geudens & Sandra, 1999). In summary, the present experiment offers no evidence for a privileged status of orthographic onsets and rimes. It does suggest an alternative hypothesis: Readers in a phonologically transparent orthography may develop several types of multiletter strings rather than the single type predicted by the onset-rime hypothesis. Experiment 2 was explicitly designed to investigate this hypothesis.
Experiment 2: Frequency manipulation Experiment 2 investigated which types of multiletter strings are involved in the reading process of young children. We used frequency manipulation (with intact presentation) as the experimental technique. The rationale behind this technique is that frequency effects in reading reveal orthographic representations6 in the mental lexicon. When applied to the orthographic onset-rime hypothesis, rime frequency effects should be stronger than body frequency effects. In many models of reading development, familiarity with recurrent letter combinations is seen as an important factor in the development of higher-order orthographic representations. Reitsma (1983) provided evidence that even beginning readers of Dutch use their knowledge of specific spelling patterns when naming printed words. After a training session with unfamiliar words, children pronounced these trained words faster than their untrained corresponding pseudohomophones, i.e., pseudowords with the same pronunciation (e.g., brain versus brane in English). Only four repetitions in the training session were sufficient to yield the effect. Coenen, Van Bon, and Schreuder (1997) as well reported word frequency effects at different reading levels of Dutch readers (cf. Theloosen & Van Bon, 1993 with poor readers). Additionally, these frequency effects do not seem to be restricted to whole word levels. Although the evidence is not clear-cut, it has been demonstrated
(JB)SWL II_f5_49-67
9/24/02
2:41 PM
Page 58
Precursors of Functional Literacy
that young readers are also sensitive to the repetition of sublexical multiletter patterns when reading novel words (see Reitsma, 1990, 1997; Van den Broeck, 1997; but Theloosen & Van Bon, 1993). Research in English has shown that beginning readers are especially sensitive to the frequency of the rime unit (Bowey & Hansen, 1994; Bowey & Underwood, 1996; Leslie & Calhoon, 1995). Bowey and Hansen (1994) showed that third grade children make fewer pronunciation errors on pseudowords containing frequent rime units (e.g., vess and fip) than on pseudowords containing infrequent rime units (e.g., fiss and vep). If children use GPCs to read pseudowords, performance in the high-frequency and low-frequency conditions should be identical, as the same GPCs occurred within both conditions. Similar evidence for the use of body units could not be found in English (Treiman et al., 1990). Rather than manipulating the frequency of the entire letter string or a single substring, we varied the frequency of both the body and rime letter patterns simultaneously. The following experiment investigated two questions: (i) do beginning readers of Dutch use multiletter strings in naming pseudowords? and (ii) if they do, do these strings always correspond to orthographic onsets and rimes?
Method Participants Twenty-eight first graders participated (mean age: 6 years and 7 months). All were native speakers of Dutch and had been taught to read for nine months. The reading method that was used in the school was the unbiased method we described in the previous experiment (Veilig Leren Lezen, Mommers et al., 1990). Materials A set of 32 monosyllabic pseudowords (CVC and CVVC) was made up on the basis of the orthogonal combination of two frequency factors: CV (body) frequency and VC (rime) frequency. This resulted in four conditions: 1) high-frequency body and high-frequency rime (H/H), 2) high-frequency body and low-frequency rime (H/L), 3) low-frequency body and high-frequency rime (L/H), and 4) low-frequency body and low-frequency rime (L/L). As there is no up-to-date corpus of written word frequencies for Dutch children we composed a written word frequency count by scanning the books that had been read by all the children during reading lessons at school. On the basis of this count we calculated body and rime frequencies. The dataset amounted to 1447 types and 38677 tokens. For each rime and body string in the corpus we determined a type frequency measure, which refers to the number of different word forms in which the substring appeared, and a token frequency measure, which refers to the actual number of word occurrences in the corpus in which the substring was contained.
(JB)SWL II_f5_49-67
9/24/02
2:41 PM
Page 59
Onset-rime units in Dutch Table 3. Example of orthogonal item composition in Experiment 2
H rime L rime
H body
L body
wet wef
set sef
Note. H = high frequency, L = low frequency. Pseudoword items are composed on the basis of the orthogonal combination of body and rime frequency.
Due to the limitations of our database, an orthogonal combination of type and token frequency was impossible. Since both type and token frequency may play a significant role (see Reitsma, 1990), we selected our items in such a way that the type and token frequency measures of the CV (body) and VC (rime) strings covaried, i.e., a high/low value on the type frequency measure also had a high/low-frequency value on the token frequency measure. The mean type and token frequencies on the selected high frequency (H) and low frequency (L) rimes were matched as closely as possible to the mean type and token frequencies of the selected H and L bodies (for type frequencies: 10 [H rimes], 8.9 [H bodies], 1.3 [L rimes], and 1.3 [L bodies]; for token frequencies: 271.8 [H rimes], 265.9 [H bodies], 9.8 [L rimes], and 8.5 [L bodies]). The items were constructed in sets of four pseudowords, all sharing the vowel. Two rimes (H/L) and two bodies (H/L) were chosen. By making all four combinations of these units we filled the cells of a matrix which was defined by the orthogonal combination of body and rime frequency. An example of an item set is given in Table 3. Given the limitations of the frequency count and the demands of the orthogonal procedure we were able to construct 16 matrices, yielding a total of 64 pseudowords. To avoid that participants would see the same body or rime more than once, the items of a set were divided over two lists (e.g., sef, wet in List 1; wef, set in List 2), with all four conditions occurring equally often in each list. Thus each list contained 32 items (eight per condition). Each list was presented in two random orders. Prior to the experiment, 10 practice items were presented.
Procedure Pseudowords were presented one by one in the centre of a computer screen (Arial, 36 pt.). Each trial started with a short auditory signal, followed immediately by the presentation of the pseudoword. The children were tested individually and asked to read the “funny made up” items as fast and accurately as possible. Upon reading the pseudoword, an asterisk was presented to initiate the next trial. Naming latencies were measured with millisecond accuracy by a voice-key attached to the computer. The experimenter registered whether the stimulus was read correctly and noted equipment failures. Time-outs occurred at 5 seconds.
(JB)SWL II_f5_49-67
9/24/02
2:41 PM
Page 60
Precursors of Functional Literacy Table 4. Mean latencies (M in ms), error percentages (in parenthesis), and standard errors (SE) in each condition for all participants (n = 28) in Experiment 2 Condition
M
SE
H/H H/L L/H L/L
1506 (8 %) 1731 (8 %) 1725 (9 %) 1714 (10 %)
140 175 161 149
Note. H = high frequency, L = low frequency. The initial part before the slash refers to the CV (body), the final part after the slash refers to the VC (rime).
Results ANOVAs were performed on the mean latencies for correct responses. Four items (one matrix) had to be removed from the analyses because of very high error rates (the orthographic sequence 〈ji〉 being read as 〈ij〉, a high frequency digraph in Dutch). In analysing the error data the frequency effects and their interaction were non-significant (Fs < 1). In Table 4 the mean naming latencies and error percentages for each condition are presented. An ANOVA for a 2 × 2 × 2 (Body Frequency [H-L] × Rime Frequency [H-L] × List) design was performed on the participant means, treating the two frequency factors as within-participant factors and the list factor as a betweenparticipants factor. The main effect of Body Frequency was significant, F1(1, 26) = 6.26, p < .05, as well as the main effect of Rime Frequency, F1(1,26) = 9.81, p = .01. The two frequency factors Body and Rime interacted significantly, F1(1, 26) = 8.04, p < .01. An ANOVA was also performed on the item means, treating all factors as between-items factors. The main effect of Body Frequency was nearly significant, F2(1, 52) = 3.52, p = .06, as well as the main effect of Rime Frequency, F2(1, 52) = 5.98, p < .01. Again the ANOVA yielded a significant interaction between the two frequency factors, F2(1, 52) = 8.20, p < .01. Even though we orthogonally combined the two factors Body Frequency and Rime Frequency, the relatively small (first-grade) frequency count made it impossible to match the onsets of items in the H and L body conditions. This mismatch may have introduced considerable extra variance in the naming data (e.g., acoustic differences in onsets may affect voice-key triggering). As a result, an unequivocal interpretation of the effect of Body Frequency is impossible. It also follows that only two planned comparisons were allowed to further analyse the interaction effect: H/H versus H/L and L/H versus L/L, as these are the only comparisons where conditions are matched for onsets. The comparison between H/H and H/L was statistically significant, F1(1, 26) = 14.59, p < .001; F2(1, 52) = 15.26, p < .001. The comparison between L/H and L/L failed to reach significance (Fs < 1). In Figure 2 the comparisons between H/H versus H/L and L/H versus L/L are illustrated. Correlation analyses were performed to determine whether reader fluency, measured by a standardised reading test (Een-minuut-test, [One-minute-test]; Brus
(JB)SWL II_f5_49-67
9/24/02
2:41 PM
Page 61
Onset-rime units in Dutch
1750
1725
Mean RT (ms)
1700
1731 1714
1650 1600 1550 1500
1506 H Body
1450
L Body
1400 H Rime
L Rime Frequency
Figure 2. Mean latencies (in ms) of the planned comparisons H/H versus H/L and L/H versus L/L matched for onsets in Experiment 2 Note. H Body and H Rime = H/H; H Body and L Rime = H/L; L Body and H Rime = L/H; L Body and L Rime = L/L
& Voeten, 1972), predicted the children’s naming latencies in the four frequency conditions. The difference H/L minus H/H correlated significantly with the children’s reading score, r = –.452, p < .01. The correlation was negative, indicating that poor performance on the reading test was related to high naming latencies on the H/L condition. The difference L/L minus L/H did not correlate with reading score, r = .096; p > .1.
Discussion The present experiment explored the effect of beginning readers’ familiarity with sublexical orthographic patterns on their naming latencies. To that effect we manipulated the frequency of the rime string (VC) and the body string (CV) in pseudowords. The outcome with the largest theoretical significance is the interaction effect between Body Frequency and Rime Frequency. An effect of Rime frequency was observed with high frequency bodies but no such effect was obtained with low frequency bodies. Since the rime effect is conditioned by the frequency of the body, we can conclude that both types of sublexical strings affect the children’s pseudoword processing. The results of a follow-up experiment with second-graders, in which onsets were matched across the four frequency conditions, confirm this conclusion (see Geudens & Sandra, submitted). This outcome suggests that the CV letter pattern determines the children’s reading strategy. H bodies quickly and automatically activate their orthographic pattern, whereas L bodies initiate a grapheme-by-grapheme decoding process. In the case of H/H pseudowords readers can continue to rely on the activation of
(JB)SWL II_f5_49-67
9/24/02
2:41 PM
Page 62
Precursors of Functional Literacy
the rime part, whereas in the case of H/L pseudowords, they fail to find a sublexical representation of the L rime and have to take recourse to a decoding strategy. The correlations showed that especially non-fluent readers had problems with this grapheme-by-grapheme decoding strategy, caused by the presence of a L string. On the other hand, pseudowords with a L body cause beginning readers, both fluent and non-fluent, to engage in decoding from the beginning of the word. Apparently, once this decoding process is started it is followed throughout the pseudoword, i.e. the presence of a H rime does not lead beginning readers to use the sublexical representation of this string. As a result, in this age group (firstgrade) L/H pseudowords are functionally identical to L/L words. One could as well interpret our data pattern in terms of a parallel processing account in which children process CV and VC substrings simultaneously. If a naming response can only be initiated when both substrings have been processed, the lowest frequency string will determine response times. Such a view would also predict the outcome of our experiment, i.e., H/H being the fastest condition and L/H, H/L, and L/L clustering together. To summarise, the present experiment suggests that beginning readers of Dutch do not assign a special representational status to orthographic rimes. They develop sublexical representations for high-frequency orthographic patterns, independently of their position in the word (body or rime position). This finding fits in with the idea that the systematic relationship between graphemes and phonemes in Dutch does not cause readers to treat rimes as more helpful word segments for the purpose of lexical processing than other sublexical orthographic patterns (see Geudens & Sandra, 1999; Martensen et al., 2000; Reitsma, 1990; 1997; Sandra & Geudens, 1999; Theloosen & Van Bon, 1993; Reitsma & Van der Leij, 1994; Van den Bosch, 1991).
General discussion The purpose of the two experiments in this paper was to find out whether beginning readers in a phonologically transparent orthography make use of orthographic onset-rime units. The orthography of Dutch offers a way to distinguish between two possible accounts for the existence of onset-rime effects. If they reflect the saliency of onsets and rimes in children’s phonological awareness, they should turn up in Dutch as well as in English. However, if orthographic onset-rime units emerge as a result of inconsistencies in grapheme-to-phoneme mapping (Treiman et al., 1995) multiletter representations in Dutch should not reflect an onset-rime preference (Martensen et al., 2000). Even though earlier experiments with Dutch beginning readers suggested no involvement of orthographic onset-rime (Reitsma, 1990; 1997; Theloosen & Van Bon, 1993; Reitsma & Van der Leij, 1994; Van den Bosch, 1991; with the exception of Assink et al., 1998), the reading method used for children in these experiments may have masked onset-rime effects. The reading manuals contained many structured word lists, i.e., lists in which a single
(JB)SWL II_f5_49-67
9/24/02
2:41 PM
Page 63
Onset-rime units in Dutch
grapheme is varied between successive words, and emphasised a body-coda structure of words (varying coda) to the same extent as an onset-rime structure (varying onset). Experiment 1 was explicitly designed to study this possible effect of instruction. Using a visual segmentation technique with pseudowords, we compared readers who were taught to read with the same reading method as in previous research to readers who were taught to read with a reading method that strongly favoured an onset-rime structure. We found no effect of instruction. We also studied the difference between fluent and non-fluent readers. Whereas non-fluent readers showed an advantage of an onset-rime segmentation relative to a body-coda segmentation, fluent readers did not. The letter-by-letter decoding strategy of the non-fluent readers makes it rather unlikely to interpret their preference for onsetrime segmented items in terms of multiletter decoding units. As far as the fluent readers are concerned, the absence of an onset-rime advantage does not deny their use of multiletter representations. It only demonstrates that they have no preference for onset and rime segments. These data are compatible with an account in which different types of sublexical multiletter strings are simultaneously involved. We investigated this hypothesis in Experiment 2. The results of Experiment 2 show that readers at the end of the first grade indeed make use of multiletter representations and that they have no preference for orthographic onsets and rimes. This was revealed by an interaction between the effect of Body Frequency and Rime Frequency, indicating that both types of substrings are involved in processing. Indeed, Perfetti (1992) and Ehri (1992) have argued that one of the hallmarks of fluent reading is the development of precise and redundant mappings between orthography and phonology at different word and subword levels. Note that our first-graders developed these multiletter representations even though they were learning to read with a phonics approach in which the high level of consistency in individual grapheme-to-phoneme mapping was explicitly pointed out and trained. In addition, our correlations indicated that non-fluent readers were more slowed down by the presence of low frequency multiletter strings than fluent readers. It is interesting to observe that this finding at a sublexical level is completely in line with findings at the lexical level reported by Perfetti and Hogaboam (1975) in an experiment in which whole-word frequency was manipulated. In this experiment the differences on naming latencies of good and poor readers were more outspoken for low-frequency words and pseudowords than for high-frequency words. The combined set of results for the two experiments indicates that readers of Dutch at the end of the first grade have developed multiletter representations but that these representations are not restricted to units predicted by the onset-rime hypothesis. Thus, the present study extends the existing evidence against the use of onsets and rimes in Dutch and adds the finding that exposure frequency of multiletter strings is a major determinant of higher-order orthographic representations in a language with a phonologically transparent orthography.
(JB)SWL II_f5_49-67
9/24/02
2:41 PM
Page 64
Precursors of Functional Literacy
Notes * [The first author is Research Assistant of the Fund for Scientific Research — Flanders (Belgium) (F.W.O.). The study was supported by a research grant by the Ministry of the Flemish Community (Concertated Research Action, contract # 98/3). ** We are grateful to pupils and staff of the primary schools Delta, Reuzepas and Zwaneven (Oud-Turnhout), who cooperated so generously. We thank Helena Taelman for her assistance in composing the word frequency count. We also would like to thank Kate Nation, James Booth and the reviewers for the helpful comments on an earlier draft of the manuscript. . For spelling, the rime unit seems less influential than is the case for reading. With a lexical priming paradigm Nation and Hulme (1996) observed robust priming effects regardless of whether the prime-target pair shared a rime unit, a body or a vowel. . One might argue that when rime-based and body-based lists are used equally frequently in a reading method this eliminates the instructional effect for the study of these units (the effect being constant across unit types). However, one might just as well argue that sufficient (equal) training on two types of subsyllabic orthographic structure overrules possible effects from phonology. . Note that responses on which individual letter pronunciations triggered the voice-key were removed from the analysis. . As reading speed is a continuum and the median-split analysis divides this continuum into two regions, the terminology “fluent” and “non-fluent” readers suggests a dichotomy which does not correspond to the natural distinctions within the group of readers. Even though the distinction between the two subgroups is statistically warranted, the readers must keep in mind that the terms have no absolute meaning but are defined with respect to the median of this sample, and that there is also considerable variability within each subgroup. . Many researchers have argued that reading speed is a powerful predictor of reading skill in a phonemically transparent language (Van den Bos & Spelberg 1997; Wimmer, 1996). . The term “representation” does not reflect a theoretical commitment to either a symbolic or a connectionist approach to language, nor does the concept “orthographic” representation exclude an intimate connection between orthographic and phonological levels.
References Adams, M. (1990). Beginning to read: Thinking and learning about print. Cambridge, MA: MIT Press. Assink, E. M. H., Kattenberg, G. & Wortmann, C. (1998). Exploring the boundaries of sublexical word identification units: The use of onsets and rimes and reading ability. Journal of Psycholinguistic Research, 27, 6, 639–659. Booth, J. R. & Perfetti, C. A. (2002). Onset and rime structure influences naming but not early word identification in children and adults. Scientific Studies of Reading, 6, 1–23. Bowey, J. A. (1996). Phonological recoding of nonword orthographic rime primes. Journal of Experimental Psychology: Learning, Memory, and Cognition, 22, 1, 117–131. Bowey, J. A. (1999). The limitations of orthographic rime analogies in beginners’ word reading: a reply to Goswami (1999). Journal of Experimental Child Psychology, 72, 220–231.
(JB)SWL II_f5_49-67
9/24/02
2:41 PM
Page 65
Onset-rime units in Dutch Bowey, J. A. & Hansen, J. (1994). The development of orthographic rimes as units of word recognition. Journal of Experimental Child Psychology, 58, 465–488. Bowey, J. A. & Underwood, N. (1996). Further evidence that orthographic rime usage in nonword reading increases with word-level reading proficiency. Journal of Experimental Child Psychology, 63, 526–562. Bowey, J. A., Vaughan, L. & Hansen, J. (1998). Beginning reader’s use of orthographic analogies in word reading. Journal of Experimental Child Psychology, 68, 108–133. Bradley, L. & Bryant, P. (1983). Categorizing sounds and learning to read — a causal connection. Nature, 310, 419–421. Brown, G. D. A. & Deavers, R. P. (1999). Units of analysis in nonword reading: Evidence from children and adults. Journal of Experimental Child Psychology, 73, 208–242. Brus, B. T. & Voeten, M. J. M. (1973). Een-minuut-test vorm A en B. Verantwoording en handleiding [One-minute-test form A and B. Manual]. Nijmegen: Berkhout Testmateriaal. Coenen, M. J. W. L., Van Bon, W. H. J. & Schreuder, R. (1997). Reading and spelling in Dutch first and second graders: Do they use an orthographic strategy? In C. K. Leong & R. M. Joshi (Eds.), Cross-language studies of learning to read and spell (213–233). Dordrecht: Kluwer Academic Publishers. De Baar, K. (1995). Leeslijn — De leesweg [Reading line — The reading path]. Amsterdam: Meulenhof Educatief. Duncan, G. L., Seymour P. H. K. & Hill, S. (1997). How important are rhyme and analogy in beginning reading? Cognition, 63, 171–208. Ehri, L. C. (1992). Reconceptualizing the development of sight word reading and its relationship to recoding. In P. B. Gough, L. C. Ehri & R. Treiman. Reading Acquisition (107–143). Hillsdale, NJ: Erlbaum. Geudens, A. & Sandra, D. (1999). Onsets and rimes in a phonologically transparent orthography: Differences between good and poor beginning readers of Dutch. Brain and Language, 68, 284–290. Geudens, A. & Sandra, D. (submitted). Exposure frequency determines the use of orthographic multiletter strings in beginning readers of Dutch. Glushko, R. J. (1979). The organization and activation of orthographic knowledge in reading aloud. Journal of Experimental Psychology: Human Perception and Performance, 5, 674–691. Goswami, U. (1993). Toward an interactive analogy model of reading development: Decoding vowel graphemes in beginning reading. Journal of Experimental Child Psychology, 56, 443–475. Goswami, U. & Bryant, P. E. (1990). Phonological skills and learning to read. Hillsdale, NJ: Erlbaum. Goswami, U., Gombert, J. E. & 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. Jared, D., McRae, K. & Seidenberg, M. S. (1990). The basis of consistency effects in word naming. Journal of Memory and Language, 29, 687–715. Kirtley, C., Bryant, P., MacLean, M. & Bradley, L. (1989). Rhyme, rime, and the onset of reading. Journal of Experimental Child Psychology, 48, 224–245. Leslie, L. & Calhoon, A. (1995). Factors affecting children’s reading of rimes: Reading ability, word frequency, and rime-neigborhood size. Journal of Educational Psychology, 87, 4, 576–586. Martensen, H., Maris, E. & Dijkstra, T. (2000). When does inconsistency hurt? On the relation between phonological-consistency effects and the reliability of sublexcial units. Memory & Cognition, 28, 648–656. McClure, K. K., Bisanz, G. L. & Ferreira, F. (1996). Effects of grade, syllable segmentation,
(JB)SWL II_f5_49-67
9/24/02
2:42 PM
Page 66
Precursors of Functional Literacy and speed of presentation on children’s word-blending ability. Journal of Educational Psychology, 88, 4, 670–681. Mommers, M. J. C., Verhoeven, L. & Van der Linden, S. (1990). Veilig leren lezen [Learning to read safely]. Tilburg: Zwijsen. Nation, K. & Hulme, C. (1996). The automatic activation of sound-letter knowledge: An alternative interpretation of analogy and priming effects in early spelling development. Journal of Experimental Child Psychology, 63, 416–435. Perfetti, C.A. (1992). The representation problem in reading acquisition. In P. B. Gough, L. C. Ehri, & R. Treiman. Reading Acquisition (145–174). Hillsdale, NJ: Erlbaum. Perfetti, C. A. & Hogaboam, T. (1975). Relationship between single word decoding and reading comprehension skill. Journal of Educational Psychology, 67, 461–469. Reitsma, P. (1983). Printed word learning in beginning readers. Journal of Experimental Child Psychology, 36, 321–339. Reitsma, P. (1990). Development of orthographic knowledge. In P. Reitsma & L. Verhoeven (Eds.), Acquisition of reading in Dutch (43–64). Dordrecht: Foris Publications. Reitsma, P. (1997). How to get friends in beginning word recognition. In C. K. Leong and R. M. Joshi (Eds.), Cross-language studies of learning to read and spell (213–233). Dordrecht: Kluwer Academic Publishers. Sandra, D. & Geudens, A. (1999). On the use of subsyllabic reading units in a language with a phonologically transparent orthography: The case of Dutch. In S. Bagnara (Ed.), Proceedings of the European Conference on Cognitive Science 1999, Siena (363–368). Siena: Multimedia & Communication Laboratory, Università di Siena. Santa, C. M. (1976). Spelling patterns and the development of flexible word recognition strategies. Reading Research Quarterly, 12, 2, 125–144. Savage, R. & Stuart, M. (1998). Sublexical inferences in beginning reading: Medial vowel digraphs as functional units of transfer. Journal of Experimental Child Psychology, 69, 85–108. Stanback, M. L. (1992). Syllable and rime patterns for teaching reading: Analysis of a frequency-based vocabulary of 17.602 words. Annals of Dyslexia, 42, 196–221. Theloosen, G. & Van Bon, W. H. J. (1993). Herhaling van woorden en lettergroepen: Oefeneffecten bij zwakke lezers [Repetition of words and letter clusters: Practice effects in poor readers]. Pedagogische Studiën, 70, 180–194. Treiman, R. (1989). The internal structure of the syllable. In G. N. Carlson & M. K. Tanenhaus (Eds.), Linguistic structure in language processing (27–52). Dordrecht: Kluwer Academic Publishers. Treiman, R. (1992). The role of intrasyllabic units in learning to read and spell. In P. Gough, L. C. Ehri & R. Treiman (Eds.), Reading acquisition (65–106). Hillsdale, NJ: Erlbaum. Treiman, R. & Chafetz, J. (1987). Are there onset- and rime-like units in printed words? In M. Coltheart (Ed.), Attention and performance XII. The psychology of reading (281–198). Hillsdale, NJ: Erlbaum. Treiman, R., Goswami, U. & Bruck, M. (1990). Not all pseudowords are alike: Implications for reading development and theory. Memory & Cognition, 18, 559–567. Treiman, R., Mullenix, J., Bijeljac-Babic, R. & 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, 107–136. Van Daal, V. H. P., Reitsma, P. & Van der Leij, A. (1994). Processing units in word reading by disabled readers. Journal of Experimental Child Psychology, 57, 180–210. Van den Bosch, K. (1991). Poor readers’ decoding skills: Effects of training, task, and word characteristics. Doctoral dissertation, University of Nijmegen, The Netherlands. Van den Bos, K. P. & Spelberg, H. C. L. (1997). Measuring word identification skills and related variables in Dutch children. In C. K. Leong & R. M. Joshi (Eds.), Cross-language studies of learning to read and spell: Phonological and orthographic processing, 271–282. Norwell: Kluwer Academic Publishers.
(JB)SWL II_f5_49-67
9/24/02
2:42 PM
Page 67
Onset-rime units in Dutch Van den Broeck, W. (1997). De rol van fonologische verwerking bij het automatiseren van de leesvaardigheid. Doctoral dissertation, University of Leiden, The Netherlands. Wimmer, H. (1996). The nonword reading deficit in developmental dyslexia: Evidence from German children. Journal of Experimental Child Psychology, 61, 80–90. Wimmer, H. & Goswami, U. (1994). The influence of orthographic consistency on reading development: Word recognition in English and German children. Cognition, 51, 91–103. Address University of Antwerp Center for Psycholinguistics Prinsstraat 13 B – 2000 Antwerp Belgium e-mail:
[email protected] Fax: ++32 (0)3 220 42 59 Tel: ++32 (0)3 220 42 60
(JB)SWL II_f6_68-88
9/26/02
11:52 AM
Page 68
(JB)SWL II_f6_68-88
9/26/02
11:52 AM
Page 69
Manifestations of phonological deficits in dyslexia: Evidence from Dutch Children Vera C. S. Messbauer, Peter F. de Jong, and Aryan van der Leij University of Amsterdam
In the past few decades, much research has been done on dyslexic children’s problems with phonological information processing. This research has yielded much evidence that dyslexic children have difficulty with phonological awareness, i.e. awareness of sound units in words (Goswami & Bryant, 1990; Wagner & Torgesen, 1987). Phonological awareness enables a child to gain access to the phonological structure of words. Words can be segmented in large phonological units such as rhymes and syllables but also in smaller phonological units such as phonemes. Access to the phonemes in words has been assumed to play a critical role in learning to read where written letter strings must be accurately translated into strings of phonemes. Dyslexic children, however, are less sensitive to the sound segments in spoken words and are found to have sustaining problems with the detection and manipulation of phonemes in words. An underlying factor giving rise to these phonological awareness problems could be the quality of the phonological representations in the lexicon (Elbro, 1996; Fowler, 1991). Both the completeness and accuracy of the phonological representations of words in long-term memory are considered to be lower in dyslexics (Elbro, 1998). In addition to problems with phonological awareness, many studies have found evidence that dyslexic children have shorter memory spans for phonological material as compared to children without reading problems (Brady, 1991; De Jong, 1998; Jorm, 1983; McDougall, Hulme, Ellis & Monk, 1994; Stone & Brady, 1995; Wagner & Torgesen, 1987). Compared to children without reading problems dyslexic children have more difficulty to reproduce a sequence of verbal items (words or digits) in the order they were presented. Memory for non-verbal material, however, is not inferior in poor readers (Steger, Vellutino & Meshoulam, 1972; Vellutino, Steger, Harding & Phillips, 1975; Vellutino, Steger & Pruzek, 1973). Both Jorm (1983) and Brady (1991) conclude that the inferior verbal memory span of poor readers is primarily due to inefficient phonological coding in short-term memory. This inefficient phonological coding may be the result of inaccurate phonological
(JB)SWL II_f6_68-88
9/26/02
11:52 AM
Page 70
Precursors of Functional Literacy
representations in long-term memory. Hulme, Maughan, and Brown (1991) have argued that long-term memory representations of the phonological form of words is important in supporting the retrieval of partially decayed words held in a rehearsal loop during memory tasks. When phonological representations are inaccurate this will result in slower retrieval and may in turn lead to less efficient rehearsal in short-term memory, resulting in inferior recall performance. Naming studies have also provided evidence that dyslexic children have inaccurate or underspecified phonological representations in long-term memory. They have consistently found that the majority of poor readers has subtle and pervasive lexical retrieval difficulties for familiar symbols such as objects, digits, letters, and colors (see for a review Wolf & Bowers, 1999). These difficulties are most clearly manifested in dyslexic readers by their performance on continuous naming or naming-speed tasks, in which they are required to provide names for common, serially presented stimuli under time constraints. Denckla and Rudel (1974; 1976a,b) found that the speed with which names were retrieved, rather than accuracy in naming itself, differentiated dyslexic readers from others. Later research on the source of the naming differences between dyslexic readers and controls has ruled out differences in articulation rate, short-term memory difficulties, and visual scanning problems (Obregón, 1994; Wimmer, 1993). The slower naming speed of poor readers as compared to normal readers of the same age has been considered to reflect the lower accuracy and distinctness of their phonological representations. Phonologically related words with underspecified representations tend to overlap each other to a great extent which makes lexical access more difficult and, hence, slower (Korhonen, 1995; Manis, Seidenberg & Doi, 1999). Virtually all studies on rapid naming (RAN) are based upon the performance of dyslexics and normal readers of the same age. The studies including a readingage control group to examine whether naming speed differentiates dyslexic readers from these younger normal readers are scarce and the results are mixed (Ackerman & Dykman, 1993; Badian, 1997; Olson, 1995).
Acquisition of new phonological representations The research mentioned above indicates that dyslexic children have deficits in phonological awareness, verbal short-term memory, and rapid naming (Frith, 1997). These deficits are assumed to be caused by less accurate or underspecified phonological representations of words in long-term memory. But what if new, phonologically unfamiliar words are encountered in language? Are dyslexic children hampered to a greater extent than normal reading children in setting up representations of new words? A number of studies support a relationship between problems in phonological processing abilities and the acquisition of new phonological representations. Gathercole and Baddeley (1990a) found that 5-year-old children with poor phonological memory skills were slower at learning phonologically unfamiliar words as
(JB)SWL II_f6_68-88
9/26/02
11:52 AM
Page 71
Phonological deficits in dyslexia
compared to children with good phonological memory skills. In addition, the two groups also differed one day later in their retention of the new words, suggesting that immediate memory processes are directly involved in the learning of new vocabulary items in young children. No differences were found between the two groups in learning phonologically familiar words. Michas and Henry (1994) also reported that the ability to accurately produce a new word was strongly related to phonological memory. They also found that 5-year-old children with good phonological memory skills were better at reproducing the names of new words. The authors concluded that the construction of a stable representation of the phonological structure of the sounds of new words depends critically on the adequacy of the temporary representations of the items in phonological short-term memory. More recently, De Jong, Seveke, and Van Veen (2000) have reported two studies, which examined the relationship between phonological awareness and the acquisition of new words. The first study revealed that phonological awareness of 5-year-old kindergartners was related to the paired associate learning of phonologically unfamiliar words, but not to the learning of familiar words. In the second study a group of non-reading 5-year-old children received a phonological awareness training. After this training the children appeared to learn phonologically unfamiliar words more easily as compared to children who had received no training. These findings suggest that phonological awareness can support the acquisition of novel words. In contrast to the influences of verbal memory and phonological awareness on the acquisition of new phonological representations, to our knowledge, nothing is known about the relationship between RAN and learning new phonological representations. On account of the phonological processing problems of dyslexic children it can only be inferred from the studies just mentioned that poor readers might have difficulty with the acquisition of new phonological representations. More direct evidence on this relationship is provided by a study by Aguiar and Brady (1991). They hypothesized that the vocabulary deficits often reported in disabled readers are not likely to be merely the consequence of less reading experience because differences in vocabulary knowledge have been observed in very young poor readers, raising questions about other factors in vocabulary acquisition. They argued that since poor readers have been found to have difficulties in accurate perception, storage, and retrieval of words “they might be expected to demonstrate difficulties in vocabulary acquisition, even when words are encountered outside of text, or aurally” (Aguiar & Brady: 226). Aguiar and Brady developed a vocabulary learning task of six new words in order to examine if poor readers had more difficulty acquiring auditorily presented words. Indeed, poor readers were found to need more trials to learn the new, phonologically unfamiliar words, and to make a greater number of phonological errors as compared to chronological age-controls. No differences between the groups were found in the ability to learn the semantic attributes of words.
(JB)SWL II_f6_68-88
9/26/02
11:52 AM
Page 72
Precursors of Functional Literacy
More recently, Windfuhr and Snowling (2001) found a relationship between reading ability and paired associate learning performance in 6- to 11-year-old children. Poor readers were found to have more difficulty with learning to pair nonwords to abstract figures than good readers. These differences could only be partially accounted for by phonological awareness skills. Vellutino and colleagues have reported results similar to those of Aguiar and Brady and Windfuhr and Snowling previously. In several studies they compared the performance of poor and normal readers on verbal and non-verbal paired associate learning tasks (Vellutino, Steger, Harding & Phillips, 1975; Vellutino, Bentley & Phillips, 1978; Vellutino, Scanlon & Bentley, 1983; Vellutino & Scanlon, 1989; Vellutino, Scanlon & Spearning, 1995). In all instances poor and normal readers were differentiated only on measures involving a verbal component. Poor readers were found to have more difficulty with learning unfamiliar words as compared to their chronological-age controls (Vellutino et al., 1975, 1995; Vellutino & Scanlon, 1989). On learning familiar words, however, mixed results were found. Early studies revealed that poor readers made more errors than the controls in learning to pair familiar words to Chinese characters (Vellutino et al., 1978, 1983). Several later studies, however, did not differentiate poor readers from peers without reading problems on learning words, which were high in meaning. The performance of poor readers on learning familiar words approximated that of normal readers (Vellutino & Scanlon, 1989; Vellutino et al., 1995). Vellutino and colleagues suggested that poor readers rely more heavily on semantic attributes and make less use of a word’s phonological attributes, i.e. phonological coding, to aid in remembering the newly learned words (Vellutino et al., 1995). In addition to paired associate learning tasks, Vellutino and colleagues administered a free recall task. The performance of the subjects on these two tasks was highly correlated. The authors reported that these results suggest that the associative learning difficulties observed in poor readers are to some extent attributable to a dysfunction in the storage and retrieval of word names. Finally, Wimmer, Mayringer, and Landerl (1998) administered a paired associate learning task to German dyslexic children and chronological-age controls. A significant difference was found between the reading groups. The dyslexic children needed more trials to learn the three new words than their chronologicalage controls. This result indicates that poor readers have difficulty with learning unfamiliar words in languages with regular orthographies as well.
Aims and general design of the study The scope of the present study is twofold. First, we wanted to extend the existing findings on phonological processing problems of dyslexics into the Dutch language, which is a language with more straightforward grapheme-phoneme correspondences compared to the English language. In Dutch only a few studies on the manifestations of dyslexia have been performed, which included both a
(JB)SWL II_f6_68-88
9/26/02
11:52 AM
Page 73
Phonological deficits in dyslexia
chronological-age control group and a reading-age control group. With respect to phonological awareness De Gelder and Vroomen (1991) found that Dutch retarded readers performed poorer on consonant deletion than both normal readers of the same age and reading-age controls. They also found that good readers are more sensitive to common phoneme relations between word stimuli than poor readers. These differences in (the manipulation of) phonological representations were found to persist into adulthood (De Gelder & Vroomen, 1991). In a study on working memory, De Jong (1998) found that Dutch reading disabled children performed worse on several measures of working memory capacity in both the language and the numerical domain. Finally, Van Bon and Van der Pijl (1997) reported that Dutch dyslexic children have deficits in non-word repetition and non-word recall. To our knowledge, no Dutch studies are available on the rapid naming performance of dyslexic children as compared to their peers without reading problems and their reading-age controls. In addition, the various aspects of phonological processing have not yet been included in one study. To examine a broader range of phonological processing problems in Dutch dyslexic children, we administered tasks for phonological awareness, verbal short-term memory and rapid naming in the present study. The performance of the dyslexic children was compared to both a chronological-age control group and a reading-age control group. Sound deletion was used to assess phonological awareness. The prediction was that the dyslexic children would perform worse than both their chronological-age controls without reading problems and their reading-age controls (De Gelder & Vroomen, 1991). Verbal short-term memory was assessed with a non-word span task. The dyslexic children were expected to maintain less phonological material in verbal short-term memory. As a result their scores on the non-word memory task would be considerably lower than those of their chronological-age controls. Besides, extra attention is needed for the processing of unfamiliar phonological information, which yields an extra demand to the task. Therefore, the dyslexic children were also expected to perform worse than their reading-age controls on this task. Finally, rapid automatic naming was included to measure the retrieval of phonological representations from long-term memory. Dyslexic children were expected to have a lower naming speed for objects, digits, and letters than their chronological-age controls (for a review, see Wolf & Bowers, 1999). No predictions were made here for the performances of dyslexic children versus their reading-age controls because mixed results have been reported in previous studies. The second aim of this study was to examine the acquisition of new phonological representations in dyslexic and normal readers. If dyslexic children have poorer phonological abilities than their peers without reading problems, does this hamper them to a greater extent in the establishment of new phonological representations? In line with the study by Aguiar and Brady (1991) a verbal learning task was administered. The prediction was that dyslexic children have more difficulty with learning phonologically unfamiliar words than their peers without reading
(JB)SWL II_f6_68-88
9/26/02
11:52 AM
Page 74
Precursors of Functional Literacy
problems. As an extension to the study by Aguiar and Brady we included a readingage control group. On the basis of the results of Vellutino and colleagues (Vellutino & Scanlon, 1989; Vellutino et al., 1995) it was expected that dyslexic children performed similar to their reading-age matched controls on the verbal learning of unfamiliar words. Unlike most previous studies, we also examined the types of errors made in the verbal learning task. Because of the phonological processing difficulties of the dyslexic children, we expected that they would make relatively more phonologically based errors than the other two groups. In addition to a learning task with unfamiliar words, we incorporated a verbal learning task with familiar words to examine if dyslexic children also experience difficulty on this task. As discussed earlier, learning well-known words is less dependent on phonological processing. The findings of the studies by Vellutino et al. (1978, 1983, 1995; Vellutino & Scanlon, 1989) on differences between dyslexic children and controls in learning familiar words, however, are mixed. Based on their latest findings where poor readers were found to approximate peers without reading problems in learning familiar words, we expected the dyslexic children to perform similar to their chronological-age controls on the verbal learning of familiar words. Vellutino and colleagues also regularly incorporated a readingage control group in their design and found that poor readers performed similar to this control group of younger children with the same reading level on learning familiar words (Vellutino et al., 1978, 1983, 1995; Vellutino & Scanlon, 1989). Hence the dyslexic children were expected to perform similar to their reading-age controls on learning phonologically familiar words as well.
Method Participants Three groups of 20 children each participated in this study: a group of dyslexic children (DYS), a group of reading-age (RA) controls, and a group of chronological-age (CA) controls. Each group consisted of 12 boys and 8 girls. The dyslexic children were individually matched with the reading-age control group on reading ability and with the chronological-age control group on vocabulary and age. All participants were administered the Een-Minuut-Test (EMT) (Brus & Voeten, 1979), a Dutch standardized test of single word reading. This test is commonly used to determine the reading level of children in primary school. The test consists of 116 unrelated words of increasing difficulty. The participants are required to read the words aloud as quickly as possible, and without making mistakes. The score was the number of correctly read words within one minute. A reading lag of at least two years compared to their chronological-age was used as an indication of dyslexia. Receptive vocabulary was assessed by means of the Passive Vocabulary Test,
(JB)SWL II_f6_68-88
9/26/02
11:52 AM
Page 75
Phonological deficits in dyslexia Table 1. Characteristics of the Dyslexic children (DYS), the Reading-Age (RA) control, and the Chronological-Age control (CA) group Variable Age (in months) Word decoding (EMT) Reading Grade Vocabulary (raw score)
DYS
RA
CA
122.95 (3.63) 30.20 (7.33) 2.4 34.75 (4.68)
96.70 (6.10) 32.10 (7.81) 2.5 30.65 (3.05)
121.95 (3.03) 68.65 (9.55) 4.5 35.60 (2.48)
Note. The standard deviations are shown between brackets.
a standardized subtest of the Dutch Taaltest voor Kinderen (Van Bon & Hoekstra, 1982). The construction of the test corresponds with the Peabody Picture Vocabulary Test (Dunn, 1959). The participants had to choose the correct picture from a selection of four, which matched the given word best. The test consisted of 40 items. The vocabulary score is the number of correctly chosen pictures. Children with a vocabulary score beneath the 50th percentile according to the age-norms were excluded from the study. The 20 children in the DYS group were selected from a lager group of 30 children ranging in age from 8.8 to 10.8 years. All except three children, who attended primary schools, were dyslexic children who attended schools for primary learning disabled children. The IQ of these children was 85 or above. Children with hearing or articulatory problems, neurological deficits, or children for whom Dutch was not their native language were excluded from the study. In addition, children who had been diagnosed as ADHD were omitted from the study. The teachers of the children had access to the school records of the children and were asked to register which child met one or more of the above exclusion criteria. The children assigned to the two control groups all attended primary schools. The children in the CA group were selected from a larger group of 72 children ranging in age from 9.3 to 11.3 years. The children in the RA group were also selected from a larger group of 86 children ranging in age from 5.4 to 8.10 years. The characteristics of the groups are presented in Table 1.
Instruments Phoneme awareness A sound deletion task was used to assess phoneme awareness. This task was based on the principle outlined by McDougall et al. (1994). The test consisted of 24 CCVC and CVCC non-words that were derived from the non-words used by Van Bon and Van der Pijl (1997). The non-words were presented one by one by the experimenter. The child was asked to repeat the non-word to make sure the child had perceived it correctly and could pronounce the non-word accurately. The child was then asked to delete a sound indicated by the experimenter. The initial,
(JB)SWL II_f6_68-88
9/26/02
11:52 AM
Page 76
Precursors of Functional Literacy
middle or final sound had to be deleted on alternate trials. The non-words used in this task consisted of so-called wordlike and non-wordlike non-words. Six examples preceded the test. Correct deletion never resulted in a word. No corrective feedback was given. The maximum score was 24.
Verbal short-term memory A non-word memory task was constructed using 12 monosyllabic non-words derived from the non-word repetition test (De Jong & Van der Leij, 1999). None of these non-words were already used in the sound deletion task. With these nonwords 28 lists were formed varying in length from two to eight non-words. Four trials of each list length were presented on audio-tape. The participants had to repeat the non-word lists presented in the correct order and without making pronunciation errors. The test was stopped when three or more trials of the same length were incorrectly repeated. For each correctly repeated list of non-words one point was awarded. The maximum score was 28. Rapid automatic naming This test measured the naming speed of symbols, which is an indication for the rapidity at which phonological information is retrieved from long-term memory. The test consisted of three parts: objects, digits, and letters. The participants were shown two cards with 32 and 28 objects respectively (knife, eye, book, door, and jacket) (De Jong & Van der Leij, 1999), 24 digits, or 24 letters. The children were required to name the symbols as fast as they could without making mistakes. The naming times on both cards were added for each task and subsequently divided by the total number of symbols, resulting in the naming time per symbol. A low naming score per symbol on these tests represents a high naming speed. Verbal learning Two verbal learning tasks were administered. The tasks measured the ease and accuracy with which children are able to learn to pair words to pictures. On one task, the participants had to learn phonologically familiar words and on the other task unfamiliar words paired to pictures. Hence, a set of four boy-names and a set of four girl-names were composed. The familiar boy-names were Thomas, Stefan, Martin, and Robbert. The unfamiliar names were constructed by rearranging the phoneme sequences across these names in such a way that the new sound strings were not current in Dutch language, yet easily pronounceable. The resulting non-names were Sarne, Tamro, Stomes (pronounced as ‘Stomus’), and Rafin. The familiar girl-names were Karin, Hester, Laura, and Judith. The unfamiliar names derived from these names were Itnau, Juttar, Tudil, and Haske. Half of the participants in each group were taught the boy-names and their corresponding non-names, the other half learned the girl-names and non-names. The verbal learning tasks were administered in two sessions. To avoid sequence effects, half of the participants learned the phonologically familiar words in the first session and the phonologically unfamiliar words in the second session. The
(JB)SWL II_f6_68-88
9/26/02
11:52 AM
Page 77
Phonological deficits in dyslexia
remaining participants learned the words in the other order. The (non-) words had to be paired with pictures of cats or dogs. Each verbal learning task started with a presentation-trial. The child was asked to listen carefully and to try to remember the (non-) name of each animal. One by one, the experimenter showed the four pictures of the cats or dogs, and named them aloud. After each (non-) name the child was asked to repeat the name to make sure the child had perceived it correctly and could pronounce the (non-) name accurately. Subsequently, a recall-trial took place in which the child was asked to pronounce the (non-) name corresponding to the picture shown. Of the child’s verbal response a written transcription was made by the experimenter. Next, another presentation-trial took place, followed by five successive recalltrials. Corrective feedback was given after each response. The maximum score was 24 (4 names × 6 trials).
Procedure Each participant was tested individually in his/her school setting in two sessions in a quiet room. The first session took about 15 minutes and started with the administration of the rapid automatic naming tests of objects and digits, followed by the first verbal learning task and the rapid automatic naming of letters. The second session consisted of the sound deletion task, the second verbal learning task, and finally, the non-word memory task. This session took about 35 minutes. Results First, the results of the phonological processing tests are presented. Then, the results of the verbal learning tasks are given.
Phonological information processing In Table 2 the means and standard deviations of the scores of the three reading groups on the tests of sound deletion, verbal short-term memory, and rapid automatic naming are presented. The scores on the sound deletion task were subjected to an ANOVA. This analysis revealed a significant effect of reading group (F (2, 57) = 4.86, p < .05). Contrasts showed that the mean score of the CA group was significantly higher than the mean score of the DYS group (t (57) = 3.11, p < .01). No significant mean score differences were found between the DYS group and the RA group. The scores of the participants on non-word memory were also subjected to an ANOVA. No significant effect of reading group was obtained (F < 1). All three groups performed similar on this task. The scores on the three RAN-tests were subjected to a multivariate analysis of variance (MANOVA) with objects, digits, and letters as dependent variables and reading group (DYS, RA, and CA) as a between-subjects factor. Subsequently,
(JB)SWL II_f6_68-88
9/26/02
11:52 AM
Page 78
Precursors of Functional Literacy Table 2. Means and Standard Deviations on the Phonological Processing Tests for the Dyslexic (DYS), the Reading-Age control (RA), and the Chronological-Age control (CA) group DYS Test Sound deletion Non-word memory Rapid Automatic Naminga Objects Digits Letters a
RA
CA
M
SD
M
SD
M
SD
17.00 6.00
6.17 3.39
18.90 6.70
4.36 2.54
21.45 7.10
2.11 3.14
.75 .49 .62
.10 .08 .16
.78 .57 .62
.12 .14 .11
.70 .42 .44
.12 .06 .08
Naming time per symbol (objects, digits, and letters) in seconds.
two orthogonal contrasts were specified: one comparing the CA and DYS group, and the other comparing the DYS and RA group. If the multivariate statistics indicated significant overall differences, then the univariate statistics were considered. A significant effect of reading group on RAN was found (F (6, 110) = 6.70, p < .001). Univariate statistics revealed a significant effect of reading group on digits (F (2, 57) = 12.17, p < .001), and on letters (F (2, 57) = 15.21, p < .001). The effect of reading group on naming speed of objects approached significance (F (2, 57) = 2.94, p = .06). Contrasts showed that the mean naming speed for digits of the DYS group was significantly lower compared to the mean naming speed of the CA group (F (1, 57) = 5.13, p < .05), but significantly higher compared to the mean naming speed of the RA group (F (1, 57) = 7.09, p = .01). Thus, the dyslexic children needed significantly more time to name digits than their peers without reading problems. Compared to their reading-age controls, however, dyslexic children named digits faster. Furthermore, group contrasts showed that the mean naming speed for letters of the DYS group was significantly lower than the mean naming speed of the CA group (F (1,57) = 22.81, p < .001). Unlike the results for digits, no differences were found between the DYS and RA group. Finally, contrasts showed that naming speed of objects did not differentiate between the DYS and RA group (F < 1). The group contrast between the CA and DYS group (F (1, 57) = 2.61, p = .06 one-tailed) approached significance.
Acquisition of new phonological representations In Table 3 the means and standard deviations of the three reading groups on the verbal learning tasks are presented. The scores on the verbal learning tasks were subjected to a repeated measures analysis of variance with reading group (DYS, RA, and CA) as a between-subjects factor, and type of (non-)word learning task as a within-subjects factor. Subsequently, two orthogonal contrasts were specified: one comparing the CA and DYS group, and the other comparing the DYS and RA group.
(JB)SWL II_f6_68-88
9/26/02
11:52 AM
Page 79
Phonological deficits in dyslexia Table 3. Means and Standard Deviations on Verbal Learning for the Dyslexic (DYS), the Reading-Age control (RA), and the Chronological-Age control (CA) group DYS
RA
CA
Test
M
SD
M
SD
M
SD
Verbal Learning Words Non-words
15.85 7.50
3.50 4.61
15.35 7.05
5.08 4.70
19.85 10.75
3.47 5.87
A significant effect of reading group was found, F (2, 57) = 7.22, p < .01. Contrasts showed that the verbal learning scores of the CA group were significantly higher than the scores of the DYS group, F (1, 57) = 9.43, p < .01. No significant differences were found between the RA and the DYS group (F < 1). Verbal learning in general appeared to be more difficult for dyslexic children than for their peers without reading problems. In addition, learning phonologically unfamiliar words was significantly more difficult than learning phonologically familiar words for all reading groups (F (1, 57) = 150.29, p < .001). Finally, contrary to our prediction, the interaction between reading group and type of word learning task was not significant (F < 1). Next, we considered the errors made on the word learning tests. Written transcriptions had been made of the children’s responses during the verbal learning tasks. Therefore, it was possible to conduct an analysis of the types of errors. Of six dyslexic children no detailed written transcriptions were obtained due to notation errors during testing. The deletion of these children hardly altered the mean scores of the dyslexic group on the control variables age, word reading and vocabulary. The data presented below are thus based on 14 dyslexic children, 20 chronological-age controls and 20 reading-age controls. Errors could be made in the pronunciation of the word. These errors were considered as phonological errors. In addition, errors could be made in the pairing of a word to a picture, i.e. general learning errors. The combination of these error types resulted in the distinction of the following types of errors: 1. Pronunciation of the (non-) word incorrect, but paired to the correct picture (pr –/ pi +); 2. Pronunciation of the (non-) word incorrect, and paired to the incorrect picture (pr –/ pi –); 3. Pronunciation of the (non-) word correct, but paired to the incorrect picture (pr + / pi –); 4. Other errors, mostly “don’t know” or mentioning a completely different, but existing word (other). Each error a child had made was classified in one of the four error categories. Subsequently, for each child the percentage of the errors made within each category was calculated on the basis of the total number of errors made by that child.
(JB)SWL II_f6_68-88
9/26/02
11:52 AM
Page 80
Precursors of Functional Literacy Table 4. Error types in percentages in both Verbal Learning conditions for the Dyslexic (DYS), the Reading-Age Control (RA), and the Chronological-Age Control (CA) group % Errors
DYS
RA
CA
pr –/ pi + pr –/ pi – pr + / pi – other
24.33 26.65 14.64 34.39
15.34 13.22 25.27 46.17
12.78 19.75 17.40 50.06
pr –/ pi + pr –/ pi – pr + /pr – other
3.88 3.68 64.14 28.30
3.59 1.63 45.67 49.12
0 0 67.70 32.30
Non-words
Words
The mean percentages of errors per category for the three reading groups are displayed in Table 4. To test for differences among the groups in the type of errors made two variables were considered. One variable was the overall percentage of phonological errors, which was the sum of the percentages of errors on the two phonological error categories (pr –/ pi + and pr –/ pi –). The other variable was the percentage of general learning errors (pr + / pi –). The category ‘other errors’ was not used because the scores on this variable, and accordingly the results, would be fully dependent on the scores on the other two variables. The scores of both variables were rescaled according to an arc-sinus transformation. This was done because group means and variances of scores that reflect percentages tend to be related, which would violate the assumption of their independence underlying analysis of variance. Finally, one outlier in the CA group was excluded from the analysis. For the learning of phonologically unfamiliar words, a multivariate analysis with phonological errors and learning errors as dependent variables and reading group as a between-subjects factor revealed a significant effect of group (F (4, 98) = 2.67, p < .05). Univariate statistics revealed a significant effect of reading group on phonological errors (F (2, 50) = 4.40, p < .05) but not on general learning errors (F (2, 50) = 2.19, p > .10). Examination of the group contrasts revealed that the dyslexic children made more phonological errors in learning the phonologically unfamiliar words as compared to both their chronological-age controls (F (1, 50) = 5.17, p < .05), and their reading-age controls (F (1, 50) = 8.25, p < .01). Although further investigation of group contrasts on general learning errors is disputable because the univariate statistics did not reveal a significant effect of reading group, we were still interested if dyslexic children differed from their reading-age controls on general learning errors. Dyslexic children were found to make a similar amount of general learning errors (i.e., correctly pronounced non-words paired to incorrect pictures) as their peers without reading problems (F < 1). The difference between
(JB)SWL II_f6_68-88
9/26/02
11:52 AM
Page 81
Phonological deficits in dyslexia Table 5. Definitions of Phonological Error Types, for ‘Stomus’ S—T—O—M—U—S C1–C2–V1–C3–V2–C4 Errors on phoneme level C1 changing the initial consonant only ‘S’ C2 changing the second consonant only ‘T’ C3 changing the third consonant only ‘M’ C4 changing the fourth consonant only ‘S’ V1 changing the first vowel only ‘O’ V2 changing the second vowel only ‘U’ Errors on syllable level C1V1C2 changes in the first syllable only ‘. . . mus’ C3V2C4 changes in the last syllable only ‘sto . . .’ Errors on word level C1C2C3C4 changes in the consonants of the word respecting the vowels WWI whole word intrusions; changes in both consonants and vowels in both syllables
‘. . o . u .’ ‘. . . . .’
Note. Each error can be made in a non-word paired to the correct picture (pi +) or to the incorrect picture (pi –).
the dyslexic children and their reading-age controls on general learning errors, however, approached significance (F (1, 50) = 3.77, p = .06); the reading-age controls made slightly more of these general learning errors. In the condition of learning phonologically familiar words, the dyslexic children and their reading-age controls made a negligible amount of phonological errors while the chronological-age controls made no phonological errors at all. Therefore, we only tested the differences in general learning errors between the reading groups. The ANOVA analysis revealed a significant effect of reading group (F (2, 50) = 5.22, p < .01). Examination of the group contrasts revealed that, again, no notable difference was found in the amount of general learning errors made by the dyslexic children and their peers without reading problems (t (50) = 0.29, p > .05). Reading-age controls, however, made less general learning errors compared to the dyslexic children (t (50) = –2.45, p < .05). Since a negligible amount of phonological errors was made in learning familiar words, only the incorrect pronunciations (i.e., the phonological errors) made in learning the phonologically unfamiliar words were examined in more detail. The errors were divided into three categories: errors on single phoneme level, on syllable level, and on word level could be made (based on Morais, Castro, ScliarCabral, Kolinsky, & Content, 1987). In Table 5 the definitions of these phonological errors are displayed. Each phonological error a child had made was now classified in one of the three phonological error categories. Subsequently, for each child the percentage of the errors made within each category was calculated on the basis of the total number of phonological errors made by that child. In Table 6 the mean percentage of errors per category for the three reading groups are displayed.
(JB)SWL II_f6_68-88
9/26/02
11:52 AM
Page 82
Precursors of Functional Literacy Table 6. Mean percentages of Phonological Error Types in the Non-word Learning condition for the Dyslexic (DYS), the Reading-Age Control (RA) and the Chronological-Age Control (CA) group % Errors
DYS
RA
CA
Non-word learning Phoneme Syllable Word
36.75 17.87 45.38
43.24 16.84 39.45
17.87 33.82 48.31
Again, the scores of the variables were rescaled according to an arc-sinus transformation. The multivariate analysis with errors on phoneme, syllable, and word level as dependent variables and reading group as a between-subjects factor revealed a significant effect of group (F (6, 82) = 2.39, p < .05). Univariate statistics revealed a significant effect of reading group on errors on phoneme level (F (2, 43) = 3.59, p < .05) but not on syllable level (F (2, 43 = 2.42, p > .05), and word level (F < 1). Examination of the group contrasts revealed that the dyslexic children did not differ from their reading-age controls in the amount of errors made on phoneme (F (1, 43) = 1.79, p > .05), syllable (F < 1), and word level (F < 1). Unexpectedly, the dyslexic children also did not differ from their peers without reading problems in the amount of errors made on the phoneme (F (1, 43) = 1.57, p > .05), and on the global word level (F < 1). In contrast, the difference between dyslexic children and their chronological-age controls in the amount of errors made on syllable level approached significance (F (1, 43) = 3.40, p = .07); the dyslexic children made fewer errors on the syllable level.
Discussion The results on the phonological processing tasks indicate that Dutch dyslexic children have deficits in phonological awareness and rapid automatic naming as compared to their chronological-age controls, but not as compared to their reading-age controls. Concerning phonological awareness the prediction was that dyslexic children would perform worse than their peers without reading problems and their reading-age controls on the sound deletion task (De Gelder & Vroomen, 1991). In line with previous research, the dyslexic children performed significantly worse on the sound deletion task, but, contrary to our expectations, only as compared to their chronological-age controls. These results implicate that dyslexic children encounter difficulties related to their reading-level in the processing of phonological information, especially in the manipulation of sound sequences. A reason for the observed difference between the present results and those of De Gelder and Vroomen (1991) might be that in the latter study the dyslexic children attended regular schools while most of the dyslexic children of the present study were in special education. Possibly, the dyslexic children attending special
(JB)SWL II_f6_68-88
9/26/02
11:52 AM
Page 83
Phonological deficits in dyslexia
education received remediation programs focusing on phonological awareness, which might have had a generalization effect. With respect to RAN, the main prediction was that naming speed on all three tasks (objects, digits, and letters) would differentiate dyslexic children from their chronological-age controls and probably also from their reading-age controls. Their naming speed was expected to be considerably lower than the naming speed of the other children without reading problems. This turned out to be a far too simple assumption. Indeed, the dyslexic children had a lower naming speed for digits and letters as compared to their chronological-age controls. The difference in naming speed of objects between the two groups just failed to reach significance, although there was a tendency for objects to be named slower by dyslexics than by their peers without reading problems. These results are in accordance with previous research (see Wolf & Bowers, 1999). The differences between the dyslexic children and their reading-age controls, however, were quite diverse. The speed of naming objects and letters was similar for the dyslexics and their reading-age controls. Thus, not only letter naming speed appears to be related to reading grade, but naming speed of pictures of well known words as well. The naming speed for digits, in contrast, was higher in dyslexic children. This might be caused by the fact that the dyslexic children had more instruction in and experience with mathematics, compared to their reading-age controls. Besides, dyslexic children named digits significantly faster than letters in contrast to both their reading-age and their chronological-age controls. An explanation of this finding could be that digits, compared to letters, are additionally stored in a visuo-spatial manner in longterm memory. This extra coding component of digits might enhance retrieval speed. Although contradicting results have been reported concerning the performance of dyslexics and reading-age controls on RAN-tasks, the results of the present study are in accordance with the results of Badian (1997) and Olson (1995). For non-word memory the prediction was that dyslexic children would have more difficulty to hold (unfamiliar) phonological material in verbal short-term memory. They were expected to obtain lower scores on this task as compared to both control groups. Unexpectedly, and in contrast with other research assessing verbal memory span (De Jong, 1998; Gathercole & Baddeley, 1990b; Stone & Brady, 1995), performance of the dyslexic children on the verbal memory task did not differ from their chronological-age controls. Also unexpected, the dyslexic children performed similar to their reading-age controls. The processing of unfamiliar phonological information did not appear to be an extra constraint for the dyslexic children. A possible explanation for the aberrant findings might be that the task was very difficult for the children without reading problems as well and therefore unable to discriminate between the groups. The total score that could be obtained was 28 correctly recalled strings of non-words varying from two to eight nonwords per string. All reading groups obtained a mean score of six or seven correctly recalled strings. This indicates that the correctly recalled string length was maximally three non-words. Especially for the chronological-age controls this is low (see for example McDougall et al., 1994; Windfuhr & Snowling, 2001).
(JB)SWL II_f6_68-88
9/26/02
11:52 AM
Page 84
Precursors of Functional Literacy
To sum up, Dutch dyslexic children were found to have deficits in phonological awareness and rapid automatic naming as compared to their peers without reading problems. In contrast, dyslexic children performed similar to their reading-age controls on these tasks, with the exception that dyslexics named digits faster than their younger reading-age controls. The second aim of this study was to examine whether these phonological processing problems are accompanied by problems in the formation of new phonological representations in dyslexic children. Since dyslexic children have problems with various phonological processing skills that are involved in learning novel phonological material, dyslexic children were expected to have more difficulty than chronological-age controls with learning phonologically unfamiliar words. The present findings support previous research indicating that skilled and less-skilled readers differ on a task likely to approximate vocabulary learning. Dyslexic children indeed had more difficulty with the acquisition of new, phonologically unfamiliar words, in spite of similar performance of dyslexics to chronological-age controls on a verbal short-term memory task. This suggests that dyslexic children need more exposure to and rehearsal of new words to include these words in their lexicon (also see Aguiar & Brady, 1991; Kamhi, Catts & Mauer, 1990). Additionally, the results also suggest that poor readers encounter difficulties in the acquisition of familiar words. This result is in contrast with our expectations and partially in contrast with prior research. Vellutino and colleagues (1978, 1983, 1995; Vellutino & Scanlon, 1989) found mixed results concerning differences between older poor and normal readers on paired associate learning of familiar word stimuli. In the two latest studies of Vellutino et al. (1995) and Vellutino & Scanlon (1989) no differences were found between poor en normal readers in learning words of high familiarity and meaning. Remarkably, this was not the case in the present study. One possible explanation for this finding may be sought in the words used. The familiar words used in the present study might have been not so familiar for the participants as expected. Unlike in Vellutino’s studies, no preceding check was made to insure that the familiar words were indeed familiar for each individual child. Consequently, it is possible that not all words were entirely familiar to every child. On both verbal learning tasks dyslexic children performed similar to their reading-age controls. This is in line with the studies of Vellutino et al. (1978, 1983, 1995) and Vellutino and Scanlon (1989). Their poor readers did not perform any better than the reading-age controls on the verbal learning tasks. These outcomes provide additional support for the suggestion that poor and reading level matched normal readers are comparable in phonological processing ability and in the ability to form new phonological representations. However, in contrast to the control groups, the dyslexic children had particular difficulty with the phonological aspects of the acquisition of the new words. Dyslexic children made more phonological errors while learning the new words than both control groups. These findings are also in line with the results reported
(JB)SWL II_f6_68-88
9/26/02
11:52 AM
Page 85
Phonological deficits in dyslexia
by Vellutino and Scanlon (1989) and Vellutino et al. (1995). In addition, no differences were found between the dyslexic children and their peers without reading problems in general learning ability. Neither for the non-words, nor for the words any differences were found between the dyslexic children and their normal reading peers in the ability to pair (non-) words to the correct pictures. It has been suggested that dyslexic children form less specified phonological representations of words (Elbro, Nielsen & Petersen, 1994; Elbro, 1996). As a result, poor readers would only have access to the general acoustic form of the word, and mispronunciations are made. The phonological representation is not sufficiently specific to enable them to recall the correct form. Others (e.g., Metsala & Walley, 1998) have suggested that the representations of dyslexic children did not develop into (fully) segmentalized representations at the level of phonemes. The latter would imply that during the acquisition of novel words especially global errors are to be expected. However, in the present study dyslexic children made an equal amount of phonological errors on the word level, and even less errors on the syllable level compared to their peers without reading problems. Instead, dyslexic children made more single phoneme errors. These results suggest that the dyslexic children formed new phonological representations that are quite detailed, that is phonological representations that are segmented at the level of phonemes. However, at this level their phonological representations tended to be persistently underspecified. Thus, the results suggest that dyslexic children’s relatively slow acquisition of phonological unfamiliar words might be due to phonological processing problems and in particular to the acquisition of phonological representations in which each phoneme is fully specified. Finally, considering the paired associate learning with phonological familiar words, the evaluation of the types of errors revealed that the dyslexic children made somewhat more phonological errors than both control groups. This might indicate that the dyslexic children possessed less distinct phonological representations of these known words. However, even in the dyslexic children the percentage of phonological errors was very low. Therefore, whether differences in the distinctness of phonological representations between dyslexic children and their normal reading peers can also account for the observed differences in the paired associate learning of known words is not yet clear.
References Ackerman, P. T., & Dykman, R. A. (1993). Phonological processes, confrontation naming, and immediate memory in dyslexia. Journal of Learning Disabilities, 26, 597–609. Aguiar, L., & Brady, S. (1991). Vocabulary acquisition and reading ability. Reading and Writing, 3, 413–425. Badian, N. A. (1997). Dyslexia and the double deficit hypothesis. Annals of Dyslexia, 47, 69–87. Brady. S. A. (1991). The role of working memory in reading disability. In S. A. Brady, & D. P. Shankweiler (Eds.), Phonological processes in literacy: A tribute to Isabelle Y. Liberman. Hillsdale, NJ: Erlbaum.
(JB)SWL II_f6_68-88
9/26/02
11:52 AM
Page 86
Precursors of Functional Literacy Brus, B. Th., & Voeten, M. J. M. (1979). Een-minuut-test; Vorm A en B [One-Minute Test]. Nijmegen: Berkhout B.V. De Gelder, B., & Vroomen, J. (1991). Phonological deficits: Beneath the surface of readingacquisition problems. Psychological Research, 53, 88–97. De Jong, P. F. (1998). Working memory deficits of reading disabled children. Journal of Experimental Child Psychology, 70, 75–96. De Jong, P. F., Seveke, M. J., & Van Veen, M. (2000). Phonological sensitivity and the acquisition of new words in children. Journal of Experimental Child Psychology, 76, 275–301. De Jong, P. F., & Van der Leij, A. (1999). Specific contributions of phonological abilities to early reading acquisition: Results from a Dutch latent variable longitudinal study. Journal of Educational Psychology, 91, 450–476. Denckla, M. B., & Rudel, R. G. (1974). “Rapid Automatized Naming” of pictured objects, colors, letters, and numbers by normal children. Cortex, 10, 186–202. Denckla, M. B., & Rudel, R. G. (1976a). Naming of objects by dyslexic and other learning disabled children. Brain and Language, 3, 1–15. Denckla, M. B., & Rudel, R. G. (1976b). Rapid Automatized Naming (R.A.N.): Dyslexia differentiated from other learning disabilities. Neuropsychologia, 14, 471–479. Dunn, L. M. (1959). Peabody Picture Vocabulary Test. Minnesota: American Guidance Service. Elbro, C. (1996). Early linguistic abilities and reading development: A review and a hypothesis. Reading and Writing, 8, 453–486. Elbro, C. (1998). When reading is “readn” or somthn. Distinctness of phonological representations of lexical items in normal and disabled readers. Scandinavian Journal of Psychology, 39, 149–153. Elbro, C., Nielsen, I., & Petersen, D. K. (1994). Dyslexia in adults: Evidence for deficits in non-word reading and in the phonological representation of lexical items. Annals of Dyslexia, 44, 205–226. Fowler, A. E. (1991). How early phonological development might set the stage for phoneme awareness. In S. A. Brady, & D. P. Shankweiler (Eds.), Phonological processes in literacy. Hillsdale, NJ: Erlbaum. Frith, U. (1997). Brain, mind, and behavior in dyslexia. In Ch. Hulme, & M. Snowling, Dyslexia; Biology, cognition, and intervention. London: Whurr Publishers Ltd. Gathercole, S. E., & Baddeley, A. D. (1990a). The role of phonological memory in vocabulary acquisition: A study of young children learning new names. British Journal of Psychology, 81, 439–454. Gathercole, S. E., & Baddeley, A. D. (1990b). Phonological memory deficits in language disordered children: Is there a causal connection? Journal of Memory and Language, 29, 336–360. Goswami, U., & Bryant, P. (1990). Phonological skills and learning to read. Hove, UK: Erlbaum. Hulme, Ch., Maughan, S., & Brown, G. D. A. (1991). Memory for familiar and unfamiliar words: Evidence for a long-term memory contribution to short-term memory span. Journal of Memory and Language, 30, 685–701. Jorm, A. F. (1983). Specific reading retardation and working memory: A review. British Journal of Psychology, 74, 311–342. Kamhi, A. G., Catts, H. W., & Mauer, D. (1990). Explaining speech production deficits in poor readers. Journal of Learning Disabilities, 23, 10, 632–636. Korhonen, T. T. (1995). The persistence of rapid naming problems in children with reading disabilities: A nine-year follow-up. Journal of Learning Disabilities, 28, 4, 232– 239. Manis, F. R., Seidenberg, M. S., & Doi, L. M. (1999). See Dick RAN: Rapid naming and the longitudinal prediction of reading subskills in first and second graders. Scientific Studies of Reading, 3, 2, 129–157.
(JB)SWL II_f6_68-88
9/26/02
11:52 AM
Page 87
Phonological deficits in dyslexia McDougall, S., Hulme, Ch., Ellis, A., & Monk, A. (1994). Learning to read: The role of short-term memory and phonological skills. Journal of Experimental Child Psychology, 58, 112–133. Metsala, J. L., & Walley, A. C. (1998). Spoken vocabulary growth and the segmental restructuring of lexical representations: Precursors to phonemic awareness and early reading ability. In J. L. Metsala, & L. C. Ehri (Eds.), Word recognition in beginning literacy (89–120). Mahwah, NJ: Erlbaum. Michas, I. C., & Henry, L. A. (1994). The link between phonological memory and vocabulary acquisition. British Journal of Developmental Psychology, 12, 147–163. Morais, J., Castro, S. L., Scliar-Cabral, L., Kolinsky, R., & Content, A. (1987). The effects of literacy on the recognition of dichotic words. Quarterly Journal of Experimental Psychology, 39 A, 451–465. Obregón, M. (1994). Exploring naming timing patterns by dyslexic and normal readers on the serial RAN task. Unpublished Master’s Thesis. Tufts University, Boston, MA. Olson, R. K. (1995). Language deficits in “specific” reading disability. In M. Gernsbacher (Ed.), Handbook of Psycholinguistics. San Diego, C.A.: Academic Press. Steger, J. A., Vellutino, F. R., & Meshoulam, U. (1972). Visual-tactile and tactile-tactile pairedassociate learning by normal and poor readers. Perceptual and Motor Skills, 35, 263– 266. Stone, B., & Brady, S. E. (1995). Evidence for phonological processing deficits in less-skilled readers. Annals of Dyslexia, 45, 51–78. Van Bon, W. H. J., & Hoekstra, J. G. (1982). Taaltoets voor kinderen [Language Test for Children]. Lisse: Swets & Zeitlinger. Van Bon, W. H. J., & Van der Pijl, J. (1997). Effects of word length and wordlikeness on pseudoword repetition by poor and normal readers. Applied Psycholinguistics, 18, 101–114. Vellutino, F. R., Bentley, W. L., & Phillips, F. (1978). Inter- versus intra-hemispheric learning in dyslexic and normal readers. Developmental Medicine & Child Neurology, 20, 71–80. Vellutino, F. R., & Scanlon, D. M. (1989). Auditory information processing in poor and normal readers. In J. J. Dumont, & H. Nakken (Eds.), Learning Disabilities (Vol. 2): Cognitive, Social and Remedial Aspects. Amsterdam/Lisse: Swets & Zeitlinger. Vellutino, F. R., Scanlon, D. M., & Bentley, W. L. (1983). Interhemispheric learning and speed of hemispheric transmission in dyslexic and normal readers: A replication of previous results and additional findings. Applied Psycholinguistics, 4, 209–228. Vellutino, F. R., Scanlon, D. M., & Spearing, D. (1995). Semantic and phonological coding in poor and normal readers. Journal of Experimental Child Psychology, 59, 76–123. Vellutino, F. R., Steger, J. A., Harding, C. J., & Phillips, F. (1975). Verbal vs non-verbal pairedassociates learning in poor and normal readers. Neuropsychologia, 13, 75–82. Vellutino, F. R., Steger, J. A., & Pruzek, R. M. (1973). Inter- vs intrasensory deficit in paired associate learning in poor and normal readers. Canadian Journal of Behavioural Science, 5, 2, 111–123. Wagner, R. K., & Torgesen, J. K. (1987). The nature of phonological processing and its causal role in the acquisition of reading skills. Psychological Bulletin, 101, 2, 192– 212 Wimmer, H. (1993). Characteristics of developmental dyslexia in a regular writing system. Applied Psycholinguistics, 14, 1–33. Wimmer, H., Mayringer, H., & Landerl, K. (1998). Poor reading: A deficit in skill-automatization or a phonological deficit? Scientific Studies of Reading, 2, 4, 321–340. Windfuhr, K. L., & Snowling, M. J. (2001). The relationship between paired associate learning and phonological skills in normal and dyslexic readers. Journal of Experimental Child Psychology, 80, 160–173.
(JB)SWL II_f6_68-88
9/26/02
11:52 AM
Page 88
Precursors of Functional Literacy Wolf, M., & Bowers, P. G. (1998). The Double-Deficit Hypothesis for the developmental dyslexias. Journal of Educational Psychology, 91, 3, 415–438. Address University of Amsterdam Dept. of Education Wibautstraat 4 1091 GM Amsterdam
[email protected]
(JB)SWL II_f7_89-108
9/26/02
11:53 AM
Page 89
Metaphonological awareness in monolingual and bilingual kindergartners* Vincent Goetry, Régine Kolinsky & Philippe Mousty Université Libre de Bruxelles
Since more than half of the world’s population speaks more than one language in everyday life (Grosjean, 1982), the proportion of bilinguals is certainly not marginal compared to those of monolinguals. Therefore, the proposal that “the basic models of language production should concern bilingualism, with an optional monolingual version” (De Bot, 1992, p. 2) should be applied to other domains of psycholinguistics as well. As we will argue, literacy development and its consequences at the level of metaphonological abilities constitute one of these domains, which has been relatively neglected so far. Indeed, most of the studies relating bilingualism to metalinguistic awareness have been conducted on lexical and syntactic awareness. The advantages generally reported in favour of bilinguals as compared to monolinguals seem to support the hypothesis that bilingualism in childhood promotes metalinguistic awareness (e.g., Ben-Zeev, 1977; Bialystok, 1986, 1988, 1997; Campbell & Sais, 1995; Cummins, 1978; but see Rosenblum & Pinker, 1983, for contradictory data). Bialystok has proposed to consider the metalinguistic advantage of bilinguals in light of two processing requirements for metalinguistic and other language tasks, i.e. control of linguistic processing and analysis of linguistic knowledge (see Bialystok, 1986; 1988; Bialystok & Ryan, 1985). Bilinguals would be expected to display more advanced level of control of linguistic processing than monolinguals because their experience of two languages would make the arbitrary connection between forms and meanings more apparent (see also Leopold, 1949; Vygotsky, 1962). In addition, bilinguals would be expected to be more advanced than monolinguals in the analysis of linguistic knowledge because their experience of structuring and analysing two linguistic systems would accelerate the extraction of abstract linguistic structures, rules or concepts. Contrasting with this tradition devoted to the lexical and syntactic components of metalinguistic awareness, only a few studies, to our knowledge, have explored the association between bilingualism and metaphonological awareness, as reflected by the ability to manipulate sublexical phonological units like syllables,
(JB)SWL II_f7_89-108
9/26/02
11:53 AM
Page 90
Precursors of Functional Literacy
onsets and rimes, or phonemes. This is surprising given the attention devoted to phonological awareness in the developmental literature and the constant finding that these abilities to segment utterances into phonological fragments and to represent them explicitly as concatenations of such fragments are critical components of early reading skills within an alphabetic system (see reviews in Adams, 1990; Liberman & Liberman, 1990; Wagner, 1988). Moreover, many children grow up in the first few years of life learning to speak one language and then learn to read and write in another language as a result of immigration, ethnic background or parents’ decision (e.g., Verhoeven, 1994). This provides further reasons to study the consequences of bilingualism on metaphonological development in children. As Walley suggests (1993), we might think that these young bilinguals would develop heightened sensitivity to the phonological units of words, since they need to use appropriate phonological repertories in each language and map appropriate phonological patterns to lexical representations. However, the hypothesis that bilingualism promotes general metaphonological awareness might be too broad. First, metaphonological abilities themselves seem to involve several components that develop at different times and under the influence of different factors (e.g., Bertelson, 1986; Goswami & Bryant, 1990; Treiman & Zukowski, 1991; 1996). In particular, it may be worth considering the possible implications of the structural levels of the syllable described by autosegmental phonology for metaphonological development. Hierarchical views of the syllable typically divide it into two primary units: the onset and the rime — and within rime, between peak and coda (e.g., Halle & Vergnaud, 1980; Fudge, 1969). The onset is the initial consonant or consonant cluster of the syllable. The rime is the vowel and any consonants that comes after it. The rime contains an obligatory peak, i.e. the vowel nucleus, and an optional coda, i.e. the final consonant or consonant cluster. Within such a hierarchical description, the higher level would thus correspond to the syllable, the lower level to the phonemes, and the intermediate level to the onset and rime (see a detailed presentation of this conceptualisation in Treiman, 1988). Developmental studies seem to support the psycholinguistic validity of such a hierarchical description in English. As a matter of fact, awareness of onsets and rimes seems intermediate between awareness of syllables and awareness of phonemes in terms of both age of emergence and degree of dependence on instruction (Kirtley, Bryant, McLean & Bradley, 1989; Treiman & Zukowski, 1991; see Treiman, 1992; Goswami & East, 2000, for comprehensive reviews, but see Seymour & Evans, 1994; Seymour & Duncan, 1997, for an opposite view). Second, it is quite plausible that the monolingual developmental pattern of metaphonological awareness is at least partly modulated by the most salient phonological characteristics of the child’s language. Such a view is supported by crosslanguages studies. For example, Cossù, Shankweiler, Liberman, Katz, and Tola (1988) ascribe the greater syllable awareness of Italian children compared to English children to the higher proportion of open syllables and the clearer syllabic boundaries in Italian than in English. Similarly, Caravolas and Bruck (1993) relate the higher complex onset awareness in Czech children compared to English chil-
(JB)SWL II_f7_89-108
9/26/02
11:53 AM
Page 91
Metaphonological awareness in kindergartners
dren to the more diversified and more frequent complex onsets in the Czech than in the English phonological input. Under this view, the results observed by Treiman and Zukowski (1991), and Kirtley et al. (1989) on monolingual English-speaking children could reflect language-specific rather than universal tendencies (Morais, 1995). If the varying phonologies of the languages affect the relative saliency of various phonological properties, it is possible that bilingualism affects the development of specific phonological awareness as a function of the salient structures of the phonological inputs bilinguals are exposed to. Although Rubin and Turner (1989) observed better metaphonological abilities for English-speaking bilingual children attending French schools than for English monolinguals and interpreted this difference in terms of a general metaphonological advantage, they did not address the possible differential effects of bilingualism on metaphonological awareness. Indeed, the metaphonological abilities of the children were assessed with a modified version of the Auditory Analysis test (Rosner, 1972) containing less phoneme deletion items and more syllable deletion items than the original version. Bruck and Genesee (1995) pointed out that such a task may have required much syllable segmentation but little phoneme segmentation, and propose a more conservative interpretation of Rubin and Turner’s data in terms of heightened syllable awareness for English-French bilinguals compared to English monolinguals. Bruck and Genesee (1995) also compared English-French bilinguals to English monolinguals, but used a metaphonological awareness test battery including different phonological units, i.e. syllable, onset-rime and phoneme, in order to examine the general versus specific influence of bilingualism on metaphonological awareness. In kindergarten, the bilinguals showed heightened levels of onset-rime awareness than did the monolinguals. One year later, there was no difference between the two groups for onset-rime awareness, but the bilinguals showed better syllable awareness than the monolinguals, while the reverse pattern was observed for phoneme awareness. These data were interpreted in terms of selective effects of bilingualism on the development of metaphonological awareness, owing to the salient phonological structures of both languages (the syllable being more salient in French than in English, the onset-rime distinction being highly salient in English). However, as acknowledged by the authors, most of the samedifferent judgement tasks used in this study to assess metaphonological awareness did not require high levels of explicit manipulation of the linguistic units under consideration. Stated differently, these tasks required low levels of analysis of linguistic knowledge, defined as the skill component responsible for making explicit those representations that had previously been implicit or intuitive (see KarmiloffSmith, 1986, 1992; Bialystok, 1986). Consequently, some children could have succeeded at least some onset-rime items on the basis of either larger, perhaps syllabic, representations, or global evaluation of the similarity between the two items. It is therefore not clear to what extent the performance of the bilingual children can be attributed to the salient phonological structures of both languages or to those of the second language only. We will return to this issue in the discussion.
(JB)SWL II_f7_89-108
9/26/02
11:53 AM
Page 92
Precursors of Functional Literacy
According to Bialystok (1988), higher levels of analysis of linguistic knowledge, and consequently, higher level of metaphonological awareness, is confined to bilinguals whose linguistic knowledge of both languages has been analytically acquired through instruction. Within such a view, the benefit of bilingualism should be limited to the salient phonological structures of the analysed language(s) only, and a language that remains unanalysed, such as one used only for conversation, would not be expected to yield better analysis of linguistic knowledge. To test this hypothesis, we compared in the present study the metaphonological abilities of two bilingual groups of kindergartners attending school in their second language, i.e. French native children attending Dutch schools (henceforth, French-DUTCH bilinguals) and Dutch native children attending French schools (henceforth, Dutch-FRENCH bilinguals), to those of French and Dutch monolinguals. The children’s metaphonological abilities was addressed with a phonological unit deletion task that requires high levels of analysis of linguistic knowledge (e.g., Yopp, 1988). The role of the analysis of linguistic knowledge on the bilingual children’s metaphonological abilities was addressed on the basis of the phonological differences between French and Dutch that may potentially be relevant (cf. Cossù et al., 1988; Caravolas & Bruck, 1993). To summarise these differences, it should be first noted that French has less diversified syllabic structures than Dutch: Whereas Goldman, Content, and Frauenfelder (1996) identified 21 different types of syllables in French, Schiller, Meyer, and Levelt (1997) counted no less than 34 syllable types in Dutch. Second, the syllabic boundaries are much clearer in French than in Dutch: As a consequence of both fixed stress and absence of vowel length contrast in French, the frequency of ambisyllabic consonants, i.e. consonants that forms the coda of a syllable and at the same time the onset of the following syllable (see Schiller et al., 1997; Booij, 1995), is lower than in Dutch (e.g., Kager, 1989; Tranel, 1987). Third, stress falls on the last non-schwa syllable of words irrespective of their segmental structure in French, while stress position depends on the relative syllabic weight1 of the different syllables of the same word in Dutch (Van der Hulst, 1984; Kager, 1989; but see Booij, 1995). The combination of these phonological differences between the two languages could induce a greater saliency of the syllabic units in the French phonological input, and a greater saliency of subsyllabic units such as onsets and rimes in the Dutch phonological input. In line with the notion that the effects of bilingualism are confined to the phonological structures of the analysed language(s) only (Bialystok, 1988), the impact of these salient phonological units should be confined to the second language of the bilingual children who participated to the present study. Within the submersion context characterising these children, the second language is indeed the only one analytically acquired via the instruction provided at school, as opposed to the native language, which is not. Thus, at a general level, the salient syllabic structures of French should induce better syllabic awareness only for the children attending French schools, i.e. the
(JB)SWL II_f7_89-108
9/26/02
11:53 AM
Page 93
Metaphonological awareness in kindergartners
French monolinguals and the Dutch-FRENCH bilinguals, in comparison with the children attending Dutch schools, i.e. the Dutch monolinguals and French-DUTCH bilinguals. Similarly, the effects of the salient subsyllabic phonological structures of Dutch should positively influence onset-rime awareness in the monolinguals and (French-DUTCH) bilinguals attending Dutch schools, as compared to the monolinguals and (Dutch-FRENCH) bilinguals attending French schools. The effects of analysis of the linguistic knowledge on bilingual children’s metaphonological awareness can be further addressed through specific comparisons between the bilinguals and the respective monolingual groups sharing their native language. Because of the relative importance of the syllabic units in their instruction language, we expected the Dutch-FRENCH bilinguals to show better syllable awareness than the Dutch monolinguals, while no difference was expected between the French-DUTCH bilinguals and the French monolinguals for that unit. Similarly, the French-DUTCH bilinguals were expected to show better onset-rime awareness than the French monolinguals, while the Dutch-FRENCH bilinguals should not differ significantly from the Dutch monolinguals for these units.
Method Participants Forty-three French native children attending Dutch schools (French-DUTCH bilinguals) and 39 Dutch native children attending French schools (Dutch-FRENCH bilinguals) took part in the study. These children were fully educated in their second language for three years. They spoke this language with teachers and other children at school, where they were forbidden to speak their native language. All the bilinguals were asked to answer a few questions about their linguistic background (language spoken with parents, siblings, others members of the family and friends) and linguistic habits (preferred language when they watch TV, listen to stories, etc). At home, the children generally used only their native language or both languages when one of the parents was also sufficiently bilingual. We also examined 43 French monolinguals and 43 Dutch monolinguals, also schooled for three years and taken from the same classes as the bilingual children when it was possible. These 168 children were all in kindergarten and came from 35 schools of Brussels and suburbs. Testing was conducted between March and June. In Belgium, principals and teachers are not encouraged to provide any form of literacy instruction before the first grade, and kindergartners are typically not able to read or write any word except their first name. Most of the kindergarten’s educational activities are aimed at developing motor skills as well as spatial-temporal organisation. The first author observed that some linguistic activities consisting mainly in nursery rhymes learning were given in the classrooms. However, none of the interviewed teachers related
(JB)SWL II_f7_89-108
9/26/02
11:53 AM
Page 94
Precursors of Functional Literacy
these activities to phonological development or preparing activities for literacy acquisition, and all of them considered that explicit prereading exercises fell beyond the intended pedagogic aims of kindergarten classes. The interviews and observation of the children’s activities in the classroom also showed no obvious pedagogic differences between the French and Dutch schools, nor between the linguistic exercises proposed to the children in both sets of schools. Since the children who participated in the present study came from many different schools, we controlled for the potential effects of several factors that have been shown to influence the development of metaphonological awareness, such as general language ability (e.g., Bowey & Francis, 1991), general cognitive ability (e.g., Bialystok, 1988; Tunmer, Herriman & Nesdale, 1988) and letter knowledge (e.g., Stuart & Coltheart, 1988). Thus, all children were presented with three additional tests assessing their vocabulary knowledge, non-verbal reasoning abilities and knowledge of the alphabet. Vocabulary knowledge was assessed with the French adaptation of the Peabody Picture Vocabulary Test-Revised -PPVTR- (Dunn, Thériault-Whalen & Dunn, 1993) for the French monolinguals, and with a Dutch translation of this test for the Dutch monolinguals. The bilinguals were presented with both versions of the test. Non-verbal reasoning abilities were assessed with the coloured version of the Raven’s Progressive Matrices -MATRIX- (Schutzenberg & Mavré, 1981; Kline, 1993). Letter knowledge was verified by presenting the children with all letters of the alphabet. A description of the three additional tests is provided in the Tasks section. Two subject selection criteria were used, based on the vocabulary performance of the children. First, in order to allow comparison between bilinguals and monolinguals sharing the same native language, we did not include those bilinguals whose native language vocabulary score fell below the lowest score observed in the corresponding monolingual group. Four bilinguals, one French-DUTCH and three Dutch-FRENCH, were discarded on the basis of this criterion. Second, we considered the relative level of first language/second language proficiency as a rough estimation of level of bilingualism (e.g., Hakuta & Diaz, 1985). In the vocabulary test, some of the 82 bilingual children obtained much better normalised scores in their instruction than in their native language. When this difference was greater than 15 (one standard deviation), the child was considered as “false bilingual” and thus discarded from further analyses. This was the case for four FrenchDUTCH bilinguals and ten Dutch-FRENCH bilinguals. The questionnaires revealed that these children were those who spoke the same language at school and with their mother, which was probably the reason why their “native” language knowledge, spoken only with the father, was relatively poor. The data of nine other children (three French monolinguals, two Dutch monolinguals, three French-DUTCH bilinguals and one Dutch-FRENCH bilingual) were not analysed because they were incomplete, i.e. three children refused to participate to the second session and six children were not presented with the second session because of lack of time or absence.
(JB)SWL II_f7_89-108
9/26/02
11:53 AM
Page 95
Metaphonological awareness in kindergartners Table 1. Mean ages, age ranges, and additional tests results for the monolingual and bilingual groups. Standard deviations are indicated between brackets French monolinguals Mean age Minimum Maximum PPVTa French Dutch MATRIXb Lettersc
5; 11 5; 4 6; 6
Dutch monolinguals 5; 10 5; 3 6; 4
French-DUTCH bilinguals 5; 11 5; 5 6; 7
Dutch-FRENCH bilinguals 6; 0 5; 4 6; 5
116.1 (15.8) –
– 111.1 (12.7)
110.3 (15.5) 93.5 (11.1)
99.0 (11.9) 106.2 (7.0)
52.6 (9.5)
53.7 (10.0)
55.2 (12.6)
51.7 (12.1)
6.7 (4.1)
6.2 (4.4)
6.5 (3.3)
7 (3.8)
Note. PPVT = Peabody Picture Vocabulary Test; MATRIX = Raven’s Progressive Matrices. Normalised score. b Percent correct responses. c Number of letters (out of 26).
a
Thus, for all analyses, the data of the remaining 139 children were considered: 39 French monolinguals, 40 Dutch monolinguals, 35 French-DUTCH bilinguals and 25 Dutch-FRENCH bilinguals. Table 1 presents the mean ages and results on the additional tests of the four groups. On these remaining children, no significant difference was observed between the vocabulary scores of the French and Dutch monolinguals, nor between the native language vocabulary scores of the two bilingual groups and the performance of their respective monolingual peers. There was also no group difference on the Raven’s Progressive Matrices or number of known letters. In all groups, the few letters that were identified, i.e. named, were in most cases those of the first name of the child and some vowels (a, i, o, u). The four groups were thus similar in overall verbal and non-verbal reasoning abilities, as well as in knowledge of the letters of the alphabet.
Tasks Metaphonological awareness task A deletion task was constructed to assess children’s metaphonological awareness. The children were asked to delete phonological units from disyllabic pseudo-words that formed legal sequences in French and Dutch, and contained “neutral” phonemes (i.e., phonemes that were specific neither to French nor to Dutch, such as /r/, nasals or diphthongs). These items were recorded in an anechoic room once by a French native speaker and once by a Dutch native speaker. The two lists were compared so that for each item the more neutral version, from either the French or the Dutch native speaker, was chosen, with the constraint that in the final list half of the items were pronounced by the French speaker and the other half by
(JB)SWL II_f7_89-108
9/26/02
11:53 AM
Page 96
Precursors of Functional Literacy
the Dutch speaker. Stimuli were then recorded on MiniDisc with an optical cable and at a sampling frequency of 44,1 kHz, in order to display a high digital acoustic quality. They were individually presented to the children with loudspeakers. The task was divided across two sessions. In the first session, the child had to remove some final part of the items. The final part could be the CV or CVC final syllable, e.g., saying ta when presented with tafu or di when presented with dimut; the VC rime, e.g., saying viz when presented with vizuf; the V rime, e.g., saying bof when presented with bofa; or the final consonant, e.g., saying méda when presented with médas. In the second session the child had to remove some initial part of the items. The initial part could be the initial CV syllable, e.g., saying bi when presented with mubi; the initial consonant, e.g., saying uda when presented with muda; or the initial CC cluster, e.g., saying oba when presented with kloba. For sake of clarity, the unit to be repeated and the unit to be deleted were instanced by two wooden blocks. To make it clear that the units to be deleted were of decreasing size throughout the several parts of the test, these units were represented with blocks of decreasing size (the biggest block for the syllable, the smallest for the phoneme). For each part of the test, the child was first presented with four examples. For each of them, the experimenter first displayed the two blocks in front of the child and then removed either the first or the second block, depending on the final or initial position of the phonological unit. He then pointed to the block remaining on the table, while uttering the remaining part of the item. From the second example on, the child was encouraged to remove the block representing the phonological unit to be deleted and to guess what would remain after deletion of this unit. Correction and help were provided when necessary. The examples were followed by four trials with corrective feedback and then by six trials without corrective feedback. If the child failed to give at least two correct answers for the first five items, the five last items concerning this unit were not presented and the subtest corresponding to the next unit type began.
Additional tests French and Dutch adaptations of the Peabody Picture Vocabulary Test-Revised. The child was asked to identify from among four pictorial alternatives the one that corresponded to an orally presented word. The test lasted until the child failed to give less than three correct responses out of eight successive items. All the bilinguals were presented first with the test in their second language, and then, in a different session occurring several days later, with the test in their first language. Raven’s Progressive Matrices (coloured version). The child was presented with a pattern of figural/geometric forms with a portion missing. He had to complete the pattern by selecting the appropriate missing portion from among six alternatives. The set contained 36 items ordered in increasing complexity. Letters knowledge. The child was presented with four cards presenting six or eight letters of the alphabet in random order. For each card, the experimenter
(JB)SWL II_f7_89-108
9/26/02
11:53 AM
Page 97
Metaphonological awareness in kindergartners
pointed to each letter and asked to the child what this letter was or meant. Both letter names and phonetic renderings were considered as correct responses.
Procedure Each child was tested individually in either two sessions (monolinguals) or three sessions (bilinguals) of approximately 30 minutes each, which were separated by at least one day. The tests reported in this paper formed part of a larger battery that included other phonological tests. For all children, the first session began with the Progressive Matrices test and ended with the final unit deletion test. The second session began with the Vocabulary test (in the second language for bilinguals) and ended with the initial unit deletion test. The monolinguals were presented with the letter knowledge test at the end of the second session. The bilinguals were presented with the Vocabulary test in their native language, the letter knowledge test and the questions about linguistic background in a third session.
Results It has been repeatedly found that it is more difficult, at least for children, to delete an initial than a final phonological unit (e.g., Content, Kolinsky, Morais & Bertelson, 1986; Morais, Bertelson, Cary & Alegria, 1986). Therefore, the analysis of variance (ANOVA) performed on the percentages of correct responses for the 10 items corresponding to each phonological unit considered the position of the units to be manipulated as an additional factor, with final units including the final syllable, consonant, VC or V rime, and initial units including the initial syllable, consonant, or consonant cluster. The group factor (French monolinguals vs. Dutch monolinguals vs. French-DUTCH bilinguals vs. Dutch-FRENCH bilinguals) was also considered. As expected, the ANOVA shows a highly significant unit effect (F(2,270) = 460.89, p < . 0001), and a highly significant position effect (F(1,135) = 176.22, p < .0001). Indeed, children performed much better for syllabic than for subsyllabic units (on the average, 73% vs. 22% of correct responses, respectively), and much better for the final than for the initial units (on the average, 50% vs. 28% of correct responses, respectively). Moreover, the analysis shows significant position by unit (F(2,270) = 30.88, p < .0001) and position by unit by group (F(6,270) = 2,86, p = .01) interactions. Therefore, separate analyses were performed on the final and initial units.
Final units The percentages of correct responses for each final unit are presented in Figure 1, separately for each group. The ANOVA performed on these data shows a highly significant unit effect (F(3,405) = 83.58, p < .0001), with all groups performing
(JB)SWL II_f7_89-108
9/26/02
11:53 AM
Page 98
Precursors of Functional Literacy
better for syllabic than for subsyllabic final units (on the average, 75% vs. 32% to 38% of correct responses, respectively). More interesting for our purpose, the group effect is not significant (F(3,135) = 1.31; p > .10), but there is a highly significant group by unit interaction (F(9,405) = 2.80; p < .005). We thus examined performance for each phonological unit separately. In order to examine whether the effects of bilingualism are confined to the salient phonological structures of the second language only, we performed two complementary types of planned comparisons. First, we examined the effects of phonological salient structures displayed in the instruction language on metaphonological development in children through general comparisons between the two groups attending French schools, i.e. the Dutch-FRENCH bilinguals and the French monolinguals, and the two groups attending Dutch schools, i.e. the French-DUTCH bilinguals and the Dutch monolinguals. Second, in order to test precisely whether the bilinguals’ metaphonological advantage was confined to the language of instruction, we compared for each unit both bilingual groups with their monolingual peers, i.e. the Dutch-FRENCH bilinguals with the Dutch monolinguals and the French-DUTCH bilinguals with the French monolinguals.
Final syllable As shown in Figure 1, the four groups of children did not reach the same level of performance when asked to delete the final syllable. Coherently, the ANOVA performed on the percentages of correct responses shows a significant group effect (F(3,135) = 3.60, p = .01). As predicted, the children attending French schools displayed significantly more correct responses than the children attending Dutch schools (F(1,137) = 8.76, p < .001). This suggests that the salient syllabic structures in the French input positively influence syllabic awareness in children receiving instruction in that language. This was confirmed by the specific comparisons between bilinguals and monolinguals showing that the effect of the salient syllabic structures of French on syllabic awareness is limited to the bilinguals having French as instruction language. Indeed, the Dutch native bilinguals attending French schools (Dutch-FRENCH) displayed better syllable awareness than the Dutch monolinguals (F(1,63) = 6.02; p = .01), while the French native bilinguals attending Dutch schools (FrenchDUTCH) did not differ significantly from the French monolinguals (F(1,72) = 2.90, p < .10). Rime As can be seen in Figure 1, the four groups did not perform at the same level on rime deletion. However, the group differences are reversed in comparison to those observed on syllable deletion, with the two groups attending Dutch schools performing better for rimes than the two groups attending French schools. The ANOVA performed on the two types of rimes (VC vs. V, which are grouped in Figure 1) across the four groups shows a significant group effect (F(3,135) = 2.96; p < .05), and a significant rime type effect (F(1,135) = 6.24; p < .025), with all
(JB)SWL II_f7_89-108
9/26/02
11:53 AM
Page 99
Metaphonological awareness in kindergartners 100 90 80
Percent correct
70 60 50 40 30 20 10 0 Syllable
Rime
Consonant
Figure 1. Mean proportions of correct responses, separately for each final unit and for each group (in black, French monolinguals; with horizontal lines, Dutch-FRENCH bilinguals; in light grey, Dutch monolinguals; with diagonal lines, French-DUTCH bilinguals)
groups performing better for the VC than for the V rimes (on the average, 38% vs. 32% of correct responses, respectively). Since the rime type by group interaction is not significant (F(3,135) = 1.48; p > .1), we performed the planned comparisons with the two rimes conditions confounded. The comparisons show that the children attending Dutch schools display better rime awareness than the children attending French schools (F(1,137) = 7.74; p < .01). This is in agreement with the idea that subsyllabic units such as rimes are more salient in the Dutch phonological input and enhance rime awareness in children receiving instruction in Dutch. Moreover, the French-DUTCH bilinguals performed better than the French monolinguals (F(1,72) = 4.76; p < .03), while the Dutch-FRENCH bilinguals tended to perform lower than the Dutch monolinguals (F(1,63) = 3.15; p < .10). This shows that the effects of the saliency of rimes on metaphonological development are limited to the bilinguals having Dutch as instruction language.
Final consonant No significant Group effect (F < 1) nor any significant planned comparison (F ≅ 1 in all cases) were observed for the final consonant. Compared with the differences observed for the syllabic and rime units, this suggests that neither phonological
(JB)SWL II_f7_89-108
9/26/02
11:53 AM
Page 100
Precursors of Functional Literacy
structures of the language nor bilingualism influence awareness of phonemic units in kindergarten, which is not surprising if we consider the great dependence of phonemic awareness on formal instruction.
Initial units The percentages of correct responses for each initial unit are presented in Figure 2, separately for each group. As can be seen, all groups performed better for initial syllable deletion than for onset deletion (on the average across groups, 71% for the syllable, 8% for the initial consonant and 5% for the initial consonantal cluster). The ANOVA performed on these data, with the factors unit (syllable, CC onset, C onset) and group shows a significant unit effect (F(2,270) = 547.03; p < .0001), with all children performing better for the syllable than for the two onsets. However, neither the group effect nor the group by unit interaction are significant (F < 1 in both cases). The planned comparisons conducted for each unit separately show no difference between groups, either (F < 1 in all cases). Thus, contrary to our prediction, these results suggest that all groups performed approximately at the same level when asked to delete the initial syllable or the onset of disyllabic pseudo-words. Nevertheless, the absence of difference between groups for the two onsets can be attributed to a floor effect induced by the combined difficulty of the initial deletion task and the subsyllabic nature of the unit to be deleted. As a matter of fact, all children obtained very low scores when asked to delete onsets, with a majority of children within each group failing to produce even one correct answer. The most common error consisted in repeating the entire item without deleting any segment. Yet, the absence of difference between groups for the initial syllable contrasts with the previous group effect observed for final syllable deletion. One possible reason that could account for these discrepant results is the increased level of analysis required by the initial position of the syllable to be deleted. As Content et al. (1985) argued, one way of performing a final unit deletion task is to monitor one’s articulation and stop just before that segment, given that the articulatory program for the pronunciation of an utterance that has just been heard is readily available. For initial segment deletion, however, the articulation program can not be used directly and some further analysis of the item is required to locate the appropriate starting point of the remaining sequence. Indirect evidence for the use of such a strategy to perform the initial syllable deletion task can be found in the most common error produced across the four groups of children. This error consisted in deleting the second syllable instead of the first (e.g., answering mu instead of bi when presented with mubi), and represented about 64% of the total number of incorrect responses. Whereas such responses were noted as incorrect, they may nevertheless reflect the child’s correct understanding of the syllabic nature of the unit to be deleted. As can be seen in Figure 3, the percentage of such “syllabic” errors was more important for the monolinguals and bilinguals attending French schools than for the monolinguals and bilinguals attending Dutch schools.
(JB)SWL II_f7_89-108
9/26/02
11:53 AM
Page 101
Metaphonological awareness in kindergartners 100 90 80
Percent correct
70 60 50 40 30 20 10 0 Syllable
Consonant cluster
Consonant
Figure 2. Mean proportions of correct responses, separately for each initial unit and for each group (in black, French monolinguals; with horizontal lines, Dutch-FRENCH bilinguals; in light grey, Dutch monolinguals; with diagonal lines, French-DUTCH bilinguals)
To test the relevance of this difference, we conducted a Chi-square test comparing the percentage of “syllabic” errors among the total number of errors observed for the French schooled children to those observed for the Dutch schooled children. The result shows a significant relationship between the distributions of errors and the groups (x2(1, N = 354) = 7.73, p < .01), indicating that the children attending French schools produced more “syllabic” errors than did the children attending Dutch schools.2 Moreover, the specific comparisons between groups show that the Dutch-FRENCH bilinguals produced more “syllabic” errors than did the Dutch monolinguals (x2(1, N = 173) = 10.49, p < .01), while no significant difference is observed between the French-DUTCH bilinguals and the French monolinguals (x2(1, N = 181) = 0.38, p > .10). In spite of the absence of differences at the level of correct performance, the error analysis thus supports the idea that the bilinguals and monolinguals attending French schools benefit from the salient syllabic structures displayed in their instruction language. As such, this error analysis corroborates the results observed on the percentages of correct responses for the final syllable.
(JB)SWL II_f7_89-108
9/26/02
11:53 AM
Page 102
Precursors of Functional Literacy 100 90
Percent syllable errors
80 70 60 50 40 30 20 10 0 French monolinguals
Dutch-FRENCH billinguals
Dutch monolinguals
French-DUTCH billinguals
Figure 3. Relative percentages of “syllabic” errors among the total number of errors for initial syllable deletion, separately for each group
Discussion The primary aim of this study was to examine the effects of bilingualism on metaphonological awareness in kindergarten children. Of specific interest was to specify the bilinguals’ metaphonological advantage reported in the literature (e.g., Rubin & Turner, 1989). A previous study of Bruck and Genesee (1995) suggests that bilingualism has specific rather than general effects on metaphonological development, depending on the salient phonological structures bilinguals are exposed to. We further hypothesised that these specific effects of bilingualism should be confined to the phonological units corresponding to the salient structures of the language(s) that are analytically acquired through education (Bialystok, 1988), i.e. to the language(s) of schooling. To this aim, we compared the abilities to segment syllables, onsets, rimes and phonemes of children acquiring a second language within a submersion context, i.e. French native children attending Dutch schools (French-DUTCH bilinguals) and Dutch native children attending French schools (Dutch-FRENCH bilinguals), to those of French and Dutch monolinguals. The structural differences between French and Dutch led us to predict a greater saliency of the syllabic structures in the French phonological input than in the
(JB)SWL II_f7_89-108
9/26/02
11:53 AM
Page 103
Metaphonological awareness in kindergartners
Dutch phonological input, while subsyllabic units like onsets and rimes were expected to be more salient in Dutch than in French. Since we predicted these contrasted salient phonological structures to positively affect the awareness of the corresponding units of the analysed languages only (Bialystok, 1988), we expected to observe better syllable awareness for the children attending French schools, i.e. the Dutch-FRENCH bilinguals and the French monolinguals, than for the children attending Dutch schools, i.e. the FrenchDUTCH bilinguals and the Dutch monolinguals. In contrast, the bilingual and monolingual children attending Dutch schools were expected to show better onsetrime awareness than the bilingual and monolingual children attending French schools. At the level of the specific comparisons between bilinguals and monolinguals, the restricted effects of bilingualism to the salient phonological structures of the instruction language should only induce better syllabic awareness for the DutchFRENCH bilinguals than for the Dutch monolinguals, while the French-DUTCH bilinguals should not differ from the French monolinguals. Similarly, the Dutch schooled (French-DUTCH) bilinguals were expected to show better onset-rime awareness than the French monolinguals, while the French schooled (DutchFRENCH) bilinguals should not outperform the Dutch monolinguals for these units. These predictions were corroborated by the better final syllable awareness of the two groups attending French schools compared to the two groups attending Dutch schools, together with the reverse group difference observed for rime awareness. Although there was no difference between groups concerning the percentages of correct responses for initial syllables, the higher percentages of errors consisting in a syllable deletion for the children attending French schools than for the children attending Dutch schools may also suggest that children attending French schools benefit from the salient syllabic structures of the French phonological input. It should also be noted that no difference between groups was observed for onsets deletion. However, the subtests involving onsets led to floor effects in all four groups. It can be argued that a much higher level of analysis of the linguistic structure, only provided by literacy acquisition, is required for such manipulations implying initial phoneme or initial consonant clusters (see e.g., Ehri, 1979; Morais, Alegria & Content, 1987; Morais, Cary, Alegria & Bertelson, 1979). Furthermore, the specific comparisons between bilinguals and monolinguals revealed that the bilinguals attending French schools showed better syllable awareness than the Dutch monolinguals, while no difference was observed between the two groups for rime awareness. Also as predicted, the bilinguals attending Dutch schools showed better rime awareness than the French monolinguals, while no difference was observed between the two groups in regard to syllable awareness. Taken together, our results support the idea that the effects of bilingualism on metaphonological awareness are not general but are confined to the salient phonological structures of the analysed language(s) only (Bialystok, 1986, 1988).
(JB)SWL II_f7_89-108
9/26/02
11:53 AM
Page 104
Precursors of Functional Literacy
These results cannot be attributed to overall differences between the bilingual and the monolingual groups in terms of reasoning abilities, language or letter knowledge. Indeed, bilinguals and monolinguals showed similar levels of vocabulary knowledge, non-verbal reasoning abilities and letter recognition. In addition, the very nature of our results eliminates such a possibility, since the metaphonological superiority of the bilingual children over the monolingual children was not general, but restricted to the salient phonological structures of the instruction language. It is worth noting that the present results were observed with kindergarten children. Thus, our findings extent the idea that bilingualism has restrictive effects on metaphonological awareness to children who at school do not yet benefit from any formal analysis of their linguistic knowledge. Such precocious effects were predicted on the basis of the qualitative asymmetry between second and native language acquisition in the bilinguals examined in the present study. Indeed, the linguistic structures of the second language would be acquired in a much more analytic way that the linguistic structures of the native language, within the submersion setting characterising the education of these children. However, such submersion setting may also have induced a quantitative asymmetry between both languages in terms of amount of exposure. While no measure of this variable has been undertaken in the present study, we can hypothesise that at least some bilingual children were more exposed to their second than to their native language. As such, this difference may have influenced the present pattern of results. Further studies controlling for the potential effect of this factor are therefore needed. Our interpretation may appear discrepant with the results of Bruck and Genesee (1995) who observed in kindergarten better onset-rime awareness for English children acquiring French in a submersion setting (i.e., English-FRENCH bilinguals) than for English monolinguals. Such bilinguals’ advantage would indeed not be expected for onset-rime units in that these correspond to salient phonological unit of English (cf. Treiman, 1989; Caravolas & Bruck, 1993), which was the native but not the instruction language of these bilinguals. However, to assess onset-rime awareness, Bruck and Genesee mostly used same-different judgements with highly contrasted riming and non-riming items in terms of number of shared phonemes (e.g., /darp/ — /jarp/ vs. /lisk/ — /farp/). Thus, some correct responses may have been based on a rather global evaluation of the similarity between the two members of each pair relying on larger phonological representations (see for example Bertelson, De Gelder, Tfouni & Morais, 1989; Yopp, 1988). Moreover, Bruck and Genesee observed that the proportion of children attending preschool programme was significantly higher in their bilingual than in their monolingual group. This difference may have been at least partly responsible for the better onset-rime awareness of the bilinguals compared to the monolinguals. Indeed, children’s early knowledge of nursery rhymes and word games, which are typically acquired during preschool, would affect the development of their phonological sensitivity, especially to rimes (e.g., Bryant, Bradley, Maclean & Crossland, 1989). As such, the results of Bruck and Genesee do not speak against our own research findings.
(JB)SWL II_f7_89-108
9/26/02
11:53 AM
Page 105
Metaphonological awareness in kindergartners
It should also be reminded that Bruck and Genesee (1995) observed better syllabic awareness for the bilinguals than for the monolinguals in the first grade, and interpreted these results, together with those of Rubin and Turner (1989), as showing that English-speaking bilingual children attending French schools are advantaged in syllable awareness, as compared to English monolinguals. These two sets of results, together with the ones reported on kindergartners in the present study, thus show that bilingual children attending school in their second language within a submersion context display heightened awareness than their monolingual peers for the most salient phonological units of their instruction language only. The hypothesis that the schooling language may dramatically influence metaphonological development in bilingual children conveys the implicit assumption that differences in educational settings would lead to differences in metaphonological awareness. More precisely, we would predict that bilingual children attending school in their two languages would display heightened levels of metaphonological awareness, as well as of general metalinguistic awareness (cf. Bialystok, 1988), compared to monolinguals. Campbell and Sais (1995) provided indirect support for this idea by comparing bilingually schooled (English-Italian) preschool children to English monolinguals. The results of this study showed indeed that the bilinguals, who manipulated both Italian and English analytically since the two languages were spoken at school by the teachers, helpers and children, were largely superior to the monolinguals on a set of metalinguistic tasks including semantic sorting, morpheme deletion, and phonological (presumably, phonemic) classification. This contrasts with the restricted analysed knowledge providing specific metaphonological advantages to the bilinguals examined in the present study, as well as in those of Rubin and Turner (1989). The validity and scope of this hypothesis needs to be further confirmed, for example by a direct comparison between bilinguals attending school in their second language, as typically done in submersion programs, and bilinguals attending schools in their two languages, as typically done in immersion programs. Such a comparison, as well as the longitudinal study of the potential effects of each of these two educational settings in the acquisition of literacy skills, will be matters of further research given their great importance for theoretical and educational purposes.
Notes * This study was supported by the National Fund for Scientific Research and by the Ministry of Education of the Belgian French-speaking Community (Concerted Research Action, convention 91/96–148). The first author is a Research Assistant of the Belgian Fonds National de la Recherche Scientifique (F.N.R.S.), and the second author is Research Associate of the F.N.R.S. We are grateful to the staff and children from the schools who participated to this study. Many thanks are also due to Linda Siegel as well as to the editors for helpful comments and suggestions on an earlier version of this paper.
(JB)SWL II_f7_89-108
9/26/02
11:53 AM
Page 106
Precursors of Functional Literacy . The syllable weight is determined in Dutch by the values of the onset and coda. That is, the more onset and coda positions are filled with consonants, the more a syllable is heavy. Syllabic weight affects stress placement in that heavy syllables and superheavy syllables are more likely to bear stress than light syllables (Van der Hulst, 1984; Trommelen & Zonneveld, 1999). . It should be noted that a significant relationship between percentage of “syllabic” errors and group is also observed when the four groups are considered separately (x2(3, N = 354) = 11,46, p < .01).
References Adams, M. J. (1990). Beginning to read: Thinking and learning about print. Cambridge, MA: MIT Press. Ben-Zeev, S. (1977). The influence of bilingualism on cognitive strategy and cognitive development. Child Development, 48, 1009–1018. Bertelson, P. (1986). The onset of literacy: Liminal remarks. Cognition, 24, 1–30. Bertelson, P., De Gelder, B., Tfouni, L. V. & Morais, J. (1989). Metaphonological abilities of adult illiterates: New evidence of heterogeneity. European Journal of Cognitive Psychology, 4, 239–250. Bialystok, E. (1986). Factors in the growth of linguistic awareness. Child Development, 57, 498–510. Bialystok, E. (1988). Levels of bilingualism and levels of linguistic awareness. Developmental Psychology, 24, 4, 560–567. Bialystok, E. (1997). Effects of bilingualism and biliteracy on children’s emerging concepts of print. Developmental Psychology, 33, 429–440. Bialystok, E. & Ryan, E. B. (1985). A metacognitive framework for the development of first and second language skills. In D. L. Forrest-Pressley, G. E. MacKinnon & T. G. Waller (Eds.), Meta-cognition, cognition, and human performance (207–252). New York: Academic Press. Booij, G. (1995). The phonology of Dutch. Oxford: Clarendon Press. Bowey, J. A. & Francis, J. (1991). Phonological analysis as a function of age and exposure to reading instruction. Applied Psycholinguistics, 12, 91–121. Bruck, M. & Genesee, F. (1995). Phonological awarness in young second language learners. Child Language, 22, 307–324. Bryant, P. E., Bradley, L., Maclean, M. & Crossland, J. (1989). Nursery rhymes, phonological skills and reading. Journal of Child Language, 16, 407–428. Campbell, R. & Sais, E. (1995). Accelerated metalinguistic (phonological) awareness in bilingual children. British Journal of Developmental Psychology, 13, 61–68. Caravolas, M. & Bruck, M. (1993). The effect of oral and written language input on children’s phonological awarness: a cross-linguistic study. Journal of Experimental Child Psychology, 55, 1–30. Content, A., Kolinsky, R., Morais, J. & Bertelson, P. (1986). Phonetic segmentation in prereaders: Effects of corrective information. Journal of Experimental Child Psychology, 42, 113–121. Cossù, G., Shankweiler, D., Liberman, I. Y., Katz, L. & Tola, G. (1988). Awareness of phonological segments and reading ability on Italian children. Applied Psycholinguistics, 9, 1–16. Cummins, J. (1978). Bilingualism and the development of metalinguistic awareness. Journal of cross-cultural psychology, 9, 131–149. De Bot, K. (1992). A bilingual production model: Levelt’s speaking model adapted. Applied Linguistics, 13, 1–24.
(JB)SWL II_f7_89-108
9/26/02
11:53 AM
Page 107
Metaphonological awareness in kindergartners Dunn, L. M., Thériault-Whalen, C. M. & Dunn, L. M. (1993). Echelle de Vocabulaire en Images Peabody. Toronto (Ontario): Psycan. Ehri, L. C. (1979). Linguistic insight: Threshold of reading acquisition. In T. G. Waller & G. E. MacKinnon (Eds.), Reading research: Advances in theory and practice (vol. 1, 63–114). New York: Academic Press. Fudge, E. C. (1969). Syllables. Journal of Linguistics, 5, 253–286. Goldman, J.-P., Content, A. & Frauenfelder, U. (1996). Comparaison des structures syllabiques en français et en anglais. Proceedings of the XXIèmes Journées d’Etude sur la Parole (315–318). Avignon, France. Goswami, U. & East, M. (2000). Rhyme and analogy in beginning reading: Conceptual and methodological issues. Applied Psycholinguistics, 21, 63–93. Goswami, U. & Bryant, P. (1990). Phonological skills and learning to read. Hillsdale, NJ: Erlbaum. Grosjean, F. (1982). Life with two languages. An introduction to bilingualism. Harvard University Press. Hakuta, K. & Diaz, R. M. (1985). The relationship between degree of bilingualism and cognitive ability: A critical discussion and some new longitudinal data. In K. E. Nelson (Ed.), Children’s language (vol. 5, 319–344). Hillsdale, NJ: Erlbaum. Halle, M. & Vergnaud, J.-R. (1980). Three dimensional phonology. Journal of Linguistic Research, 1, 83–105. Karmiloff-Smith, A. (1986). From metaprocess to conscious access: Evidence from children’s metalinguistic and repair data. Cognition, 28, 95–147. Karmiloff-Smith, A. (1992). Beyond modularity: A developmental perspective on Cognitive Science. Cambridge, MA: MIT Press/Bradford Books. Kager, R. (1989). A metrical theory of stress and destressing in English and Dutch. Dordrecht: Foris. Kline, P. (1993). The handbook of psychological testing. London: Routledge. Kirtley, C., Bryant, P., Maclean, M. & Bradley, L. (1989). Rhyme, rime, and the onset of reading. Journal of Experimental Child Psychology, 48, 224–245. Leopold, W. F. (1949). Speech development of a bilingual child. New York: Academic Press. Liberman, I. Y. & Liberman, A. M. (1990). Whole language vs code emphasis: Underlying assumptions and their implications for reading instruction. Annals of Dyslexia, 40, 51–76. Morais, J. (1995). Do orthographic and phonological peculiarities of alphabetically written languages influence the course of literacy acquisition? Reading and Writing: An Interdisciplinary Journal, 7, 1–7. Morais, J., Alegria, J. & Content, A. (1987). The relationship between segmental analysis and alphabetic literacy: An interactive view. Cahiers de Psychologie Cognitive, 7, 415–438. Morais, J., Cary, L., Alegria, J. & Bertelson, P. (1979). Does awareness of speech as a sequence of phones arise spontaneously? Cognition, 7, 323–331. Morais, J., Bertelson, P., Cary, L. & Alegria, J. (1986). Literacy training and speech segmentation. Cognition, 24, 45–64. Rosner, J. (1972). The development and validation of an individualized perceptual skills curriculum. Learning Resource and Development Center, University of Pittsburgh. Rosenblum, T. & Pinker, S. A. (1983). Word magic revisited: Monolingual and bilingual children’s understanding of the word-object relationship. Child Development, 54, 773–780. Rubin, H. & Turner, A. (1989). Linguistic awareness skills in grade one children in a French immersion setting. Reading and Writing: An Interdisciplinary Journal, 1, 73–86. Stuart, M. & Coltheart, M. (1988). Does reading develop in a sequence of stages? Cognition, 30, 139–181. Schiller, N. O., Meyer, A. S. & Levelt, W. J. M. (1997). The syllabic structure of spoken words: Evidence from the syllabification of intervocalic consonants. Language and Speech, 40, 103–140.
(JB)SWL II_f7_89-108
9/26/02
11:53 AM
Page 108
Precursors of Functional Literacy Schutzenberger, A. & Mavré, D. (1981). Traduction du manuel Progressive Matrices de J. Raven. Issy-les-Moulineaux: Editions Scientifiques et Psychologiques. Seymour, P. H. K. & Duncan, L. G. (1997). Small vs. large unit theories of reading acquisition. Dyslexia, 3, 125–134. Seymour, P. H. K. & Evans, H. (1994). Levels of phonological awareness and learning to read. Reading and Writing: An Interdisciplinary Journal, 6, 221–250. Tranel, B. (1987). The sounds of French. New York: Cambridge University Press. Treiman, R. (1988). Distributional constraints and syllable structure in English. Journal of Phonetics, 16, 221–229. Treiman, R. (1989). The internal structure of the syllable. In G. Carlson & M. Tannenhaus (Eds.), Linguistic structure in language processing (27–52). Dordrecht: Kluwer. Treiman, R. (1992). The role of intrasyllabic units in learning to read and spell. In P. Gough, P. L. Ehri & R. Treiman (Eds.), Linguistic structure in language processing (65–106). Dordrecht: Kluwer. Treiman, R. & Zukowski, A. (1991). Levels of phonological awareness. In S. Brady and D. Shankweiler (Eds.), Phonological processes in literacy: A tribute to Isabelle Y. Liberman (67–83). Hillsdale, NJ: Erlbaum. Treiman, R. & Zukowski, A. (1996). Children’s sensitivity to syllables, onsets, rimes, and phonemes. Journal of Experimental Child Psychology, 61, 193–215. Trommelen, M., & Zonneveld, W. (1999). Word stress in West-Germanic languages. In H. van der Hulst (Ed.), Word prosodic systems in the language of Europe (478–515). Berlin: Mouton de Gruyter. Tunmer, W. E., Herriman, M. L. & Nesdale, A. R. (1988). Metalinguistic abilities and beginning reading. Reading Research Quarterly, 23, 134–158. Van der Hulst, H. (1987). Syllabe structure and stress in Dutch. Dordrecht: Foris. Verhoeven, L. (1994). Linguistic diversity and literacy development. In L. Verhoeven (Ed.), Functional Literacy (199–220). Amsterdam: Benjamins. Vygotsky, L. (1962). Thought and language. Cambridge, MA: MIT Press. Wagner, R. K. (1988). Causal relations between the development of phonological processing abilities and the acquisition of reading skills: a meta-analysis. Merrill-Palmer Quarterly, 34, 261–279. Walley, A. C. (1993). The role of vocabulary growth in children’s spoken word recognition and segmentation ability. Developmental Review, 13, 286–350. Yopp, H. K. (1988). The validity and reliability of phonemic awareness tests. Reading Research Quarterly, 23, 159–177. Address Laboratoire de Psychologie Expérimentale Université Libre de Bruxelles 50 Av. Franklin Roosevelt 1050 Bruxelles Belgium
(JB)SWL II_f8_109-118
9/24/02
2:45 PM
Page 109
Does speech manipulation make word discrimination easier? Eliane Segers and Ludo Verhoeven University of Nijmegen
A large body of research has been conducted on the relation between phonological awareness and learning to read. Strong support has been provided that lack of phonological awareness can cause beginning reading problems (Adams, 1990; Liberman & Shankweiler, 1985). Being able to distinguish phonemes in a word is part of this awareness. Especially children with Specific Language Impairment (SLI) have problems extracting the phoneme units from the stream of spoken language (Bishop, 1997). Discrimination experiments, in which participants have to indicate whether two stimuli are different or the same, are often used to test this ability. More than twenty-five years ago, Tallal and Piercy (1973) found that if the time between two non-verbal stimuli was prolonged, or the stimuli themselves were stretched, an experimental group of 12 SLI children performed as well as other children on discrimination tasks. The same 12 SLI children also performed as well as other children in discriminating between synthetic vowel contrasts, but not in discriminating between the synthetic /ba/ and /da/ utterances (Tallal & Piercy, 1974). Tallal and Piercy (1974) suggested that SLI children have a central auditory processing problem which causes difficulty in processing fast transitional elements (in speech). In consonant-vowel stimuli like /ba/, there is a fast formanttransition between the /b/ and the /a/. If this fast transition is stretched from 43 msec to 95 msec, the same 12 SLI-children from the previous experiments performed as well as other children (Tallal & Piercy, 1975). In a follow-up experiment Tallal and Stark (1981) concluded that SLI’s discriminate poorly between syllables containing consonants, particularly stop consonants. They are not impaired in discriminating between syllables with different vowels. Different Inter Stimulus Intervals (ISI’s) or artificial prolongation of fast formant transitions were not tested in this experiment. Reed (1989) used tests with different ISI’s and found that a longer ISI facilitated the discrimination task. However, according to Mody, Studdert-Kennedy, and Brady (1997), the positive effects of formant transition lengthening reported by Tallal and Piercy (1975)
(JB)SWL II_f8_109-118
9/24/02
2:45 PM
Page 110
Precursors of Functional Literacy
have not yet been replicated. Frumkin and Rapin (1980) though, did test this on what they called dysphasic children. Twenty children were placed in three groups: seven in a group with normal phonological production, nine in a group with abnormal phonological production and four in a group with verbal auditory agnosia. Two children from the first group, five from the second group and two from the third group could not discriminate synthesized /ba/ and /da/ syllables with a 40 msec formant-transition. They did reach the criterion when the formanttransition was 80 msec. In experiments reported by Tallal et al. (1996) and Merzenich et al. (1996) a speech modification algorithm was used, which not only slowed down the complete speech signal, but also amplified fast transitional elements up to 20 dB. This algorithm was specifically designed to provide learning effects. Computer games containing modified sounds or speech signals were designed. Children in the 1996 experiments (Tallal et al., 1996; Merzenich et al., 1996) who were trained for six weeks, 1.5 hours per day, showed great improvements in among others discrimination abilities. It was not tested whether slowing down or amplification separately would have provided similar results. Details of the speech modification algorithm were described by Nagarajan et al. (1998). In discrimination experiments by Tallal and Stark (1981) the effect of slowing down fast formant transitions in speech was tested on different synthetic speech contrasts. To our knowledge, no similar experiment has been done to test the effects of (1) slowing down the complete speech signal, (2) amplification of fast formant transitions or (3) the effect of both slowing down and amplification. The purpose of the present study was thus to separate out the effects of slowing down speech and amplification of fast formant transitions on discriminating speech sounds, for both LN and SLI children.
Method Participants Twenty-one Dutch kindergartners (6 girls, 15 boys) who entered a special school for children with severe language and speech problems and a group of 24 kindergartners (14 girls, 10 boys) from a regular school participated in the experiment. The age range of the children with language and speech problems was from 4 years 9 months to 6 years 7 months (average age: 67.5 months). The children had no hearing problems or other impairments. The language normal children were aged between 4 years 8 months and 5 years 6 months (average age: 60.3 months). None of the children in the two groups had been exposed to formal reading education yet. As a control measure of mental ability, both groups of children were tested with the Coloured Progressive Matrices test (Raven, 1965). There were no differences between the two groups on the standard scores (t(43) = .246, p > .05).
(JB)SWL II_f8_109-118
9/24/02
2:45 PM
Page 111
The effects of speech manipulation
Stimulus set We chose five consonant contrasts, which are most difficult in a Dutch discrimination test (Verhoeven & Vermeer, 1986) according to Stoep and Verhoeven (2001). Contrasts on which the children had an error percentage of over 20% were: /b/–/p/ /d/–/t/ /v/–/w/ /h/–/g/ /b/–/d/ Sixty pairs of meaningful Dutch consonant-vocal-consonant (CVC) words were chosen. Half of the pairs consisted of two equal words. The stimuli were recorded in an audio studio and spoken by a professional female speaker. Audio format was PCM, 22,050 kHz, 16 bits, mono. All words in the stimulus set were recorded twice, to allow us to present for example /beer/–/beer/, by not using the same audio file twice. Some SLI children have a deficit in processing stimuli in rapid succession. To make sure that this would not influence the results, we set the time between two words in a stimulus pair at 500 msec. The inter trial interval was 2000 msec. Each presentation of a stimulus pair was preceded by a beep (1000 Hz, 200 msec) followed by a 1000 msec pause. The stimulus set is presented in the Appendix.
Speech manipulation The words in the stimulus set were manipulated in several ways: slowed down, amplified or both. To enhance fast transitional elements in speech, an algorithm similar to the one described by Nagarajan et al. (1998) was designed (Boersma & Weenink, 1998). The authors agreed on the equality of both algorithms. A PitchSynchronous Overlap-and-Add (PSOLA) algorithm was used to lengthen the speech signal. We made four different presentations of each pair of words in the stimulus set: 1. 2. 3. 4.
normal speech (NS); amplification of fast transitional elements up to 20 dB (AM); speech lengthened by 50% (LT); a combination of (LT) and (AM):(LTAM).
In Figure 1 two oscillograms of the words /b∧gk/ (belly) — /d∧gk/ (dive) are presented. In the top half of the figure, the NS condition is presented in the lower half the AM condition, which clearly shows that the intensity of the signal is altered by the algorithm.
(JB)SWL II_f8_109-118
9/24/02
2:45 PM
Page 112
Precursors of Functional Literacy 1
0
–1
1.82
0 Time (s)
1
0
–1
0
1.82 Time (s)
Figure 1. Oscillograms of the word /b∧gk/ (belly) — /d∧gk/ (dive). In the top half of the picture, normal speech is presented, in the lower half of the figure, fast transitional elements are enhanced up to 20dB
Procedure Each child was tested twelve times; three times for each type of speech manipulation. One child for example was first tested in the NS condition, two days later with AM, then LT, then LTAM then NS again and so on. The twelve tests were conducted within seven weeks. Children had to make same-different judgements on each of the 60 word pairs. The pairs were presented to the child in random order. The experiment was conducted in a quiet room inside the schools. The children wore a headphone (Monacor MD-4100); auditory stimuli were presented at
(JB)SWL II_f8_109-118
9/24/02
2:45 PM
Page 113
The effects of speech manipulation
a comfortable listening level. At the regular school, a Toshiba 310CDT laptop with Toshiba Soundblaster card was used, at the SLI school an IBM computer with a 16B card. The 60 stimuli were presented in random order. Subjects responded by pressing either of two buttons on a button box The meaning of the two buttons was indicated by symbols. Near the button, which represented two different words, two differently shaped and colored symbols were shown; near the other button two symbols with the same shape and color were shown. Before each test, we made sure that the child understood the same-different concept. The experimenter could interrupt the test when she noticed that the child lost attention. Each test took 10 to 15 minutes, including a training of 8 easy-to-discriminate pairs before the actual test. Children, who did not understand the samedifferent concept or were not motivated enough, were excluded from the experiment. Of an initial group of 29 SLI-children, 21 were tested 12 times. No children from the normal school had to be excluded.
Analysis Both error percentages and reaction times were analyzed, using the multivariate approach to the analysis of a repeated measures design (Maxwell & Delaney, 1990). The between subjects factor was school type (regular school or SLI). As within subjects factors, we used speech manipulation (NS, AM, LT, LTAM) and speech contrast (/b/–/p/, /d/–/t/, /v/–/w/, /h/–/g/, /b/–/d/). Reaction times were measured from directly after the inter-stimulus-interval until the answer-button was pressed, since from the onset of the second word, a same-different decision can be made. Reaction times below zero and above 7 seconds (three standard deviations from the mean) were discarded. Reaction times were analyzed only for correct answers. To keep the results of all children in the analysis, differences in cell means of the reaction times were tested for significance. In comparing reaction times, we separately compared the two conditions where speech was lengthened and the two conditions where speech was not lengthened, to prevent lengthening to cause a main effect.
Results All subjects Error percentages A between subjects effect of school type (F(1,43) = 14.245, p < .01) indicates the difference between the regular and the SLI school: the SLI children have higher error percentages. Multivariate tests of the error percentages showed a main effect of speech contrast (F(4,40) = 6.309, p < .01). Paired samples t-tests group the contrasts /b/–/p/,
(JB)SWL II_f8_109-118
9/24/02
2:45 PM
Page 114
Precursors of Functional Literacy
/d/–/t/ and /v/–/w/ to be more difficult than /h/–/g/ and /b/–/d/ (with an exception for v-w and h-g after correction for significance level): (/b/–/p/, /d/–/t/: t(44) = .436, p > 0.01; /b/–/p/, /v/–/w/: t(44) = 2.035, p > .01; /b/–/p/, /h/–/g/: t(44) = 4.282, p < .01; /b/–/p/, /b/–/d/: t(44) = 4.399, p < .01; /d/–/t/, /v/–/w/: t(44) = 2.020, p > .01; /d/–/t/, /h/–/g/: t(44) = 4.868, p < .01; /d/–/t/, /b/–/d/: t(44) = 4.352, p < .01; /v/–/w/, /h/–/g/: t(44) = 2.706, p = .01; /v/–/w/, /b/–/d/: t(44) = 3.1–1, p < .01; /h/–/g/, /b/–/d/: t(44) = .422, p > .01). In general, voiced/voiceless contrasts prove to be more difficult than place-contrasts. We also found an interaction of speech contrast and speech manipulation (F(12,32) = 2.220, p < .05). Further analysis of the interaction effect show that for the /h/–/g/ contrast, the normal speech condition (LN) causes less errors than the LTAM condition (t(44) = –4.971, p < .001). This effect is hard to explain, since we only found an effect in this specific combination of contrast and speech manipulation.
Reaction times Multivariate analysis of the reaction times showed a main effect of speech manipulation when comparing normal and enhanced speech F(1,42) = 4,428, p < .05). Children reacted faster in the enhanced speech condition (AM). No effects were found when comparing delayed speech (LT) and delayed and enhanced speech (LTAM). Subjects performing above chance Part of the children performed on or below chance, so the task might have been too difficult for them, which may make the analysis unreliable. Using binomial tests (alpha = 0.05), we excluded subjects which performed on (6 regular, 9 SLI) or below chance (6 SLI). This left 17 children in the regular group and 6 in the SLI group which performed above chance. The number of subjects who fail to perform above chance is in line with results from e.g., Tallal, Stark, Kallman, and Mellits (1981). Their subjects varied in age between 5 and 9 years. Twenty out of 34 dysphasics failed to reach criterion on discrimination between /ba/, and /da/ (the least difficult contrast in our study). Four of the thirty-eight normal subjects had difficulties with the task. The larger percentage in our experiment can be explained by the fact that our subjects were young: in the study by Tallal et al. (1981), younger children performed worse than older children.
Error percentages We did not find the between-subjects effect which was found in the analysis of the complete group. This can be explained by the fact that we now only had the well-performing children in the analysis. We also did not find the interaction effect, which was already an outlier in the first analysis. We again found a main effect of speech contrast in the multivariate tests on percentages correct (F4,18) = 4,346, p < .05). Further analysis on the effect of
(JB)SWL II_f8_109-118
9/24/02
2:45 PM
Page 115
The effects of speech manipulation 100
30
25
Error Percentage
20
15
0
/b/-/p/
/d/-/t/
/v/-/w/
/h/-/g/
/b/-/d/
Figure 2. Mean error percentages of the group of children performing above chance
speech contrasts again showed that /b/–/p/, /d/–/t/ and /v/–/w/ were more difficult than /h/–/g/ and /b/–/d/ (/b/–/p/, /d/–/t/: t(22) = –0.338, p > 0.05; /b/–/p/, /v/–/w/: t(22) = .888, p > .05; /b/–/p/, /h/–/g/: t(22) = 3.294, p < .01; /b/–/p/, /b/–/d/: t(22) = 3.532, p < .01; /d/–/t/, /v/–/w/: t(22) = 1.320, p > .05; /d/–/t/, /h/–/g/: t(22) = 3.840, p < .01; /d/–/t/, /b/–/d/: t(22) = 3.850, p < .01; /v/–/w/, /h/–/g/: t(22) = 2.163, p < .05; /v/–/w/, /b/–/d/: t(22) = 2.868, p < .01; /h/–/g/, /b/–/d/: t(22) = .813, p > .05). Voiced/voiceless contrasts prove to be more difficult than place-contrasts. This result is found when taking all subjects in the analysis, or when only using subjects who perform above chance. The results are in line with the error-percentages in the Dutch discrimination test described in de methodsection. Figure 2 shows the error percentages of the group of subjects which perform above chance.
Reaction times The analysis of the results of this group showed no effects on reaction times. The effect found in the analysis with the complete group of subjects can therefore be attributed to results of children that score on or below chance and is therefore less reliable.
(JB)SWL II_f8_109-118
9/24/02
2:45 PM
Page 116
Precursors of Functional Literacy Appendix. Stimulus set (Dutch cvc-words) b-p
d-t
v-w
h-g
b-d
1. 2. 3. 4.
/bet/ – /pet/ /pet/ – /bet/ /bet/ – /bet/ /pet/ – /pet/
1. 2. 3. 4.
/dak/ – /tak/ /tak/ – /dak/ /tak/ – /tak/ /dak/ – /dak/
1. 2. 3. 4.
/ver/ – /wer/ /wer/ – /ver/ /ver/ – /ver/ /wer/ – /wer/
1. 2. 3. 4.
/has/ – /gas/ /gas/ – /has/ /gas/ – /gas/ /has/ – /has/
1. 2. 3. 4.
/buk/ – /duk/ /duk/ – /buk/ /duk/ – /duk/ /buk/ – /buk/
5. 6. 7. 8.
/ber/ – /per/ /per/ – /ber/ /ber/ – /ber/ /per/ – /per/
5. 6. 7. 8.
/dæp/ – /tæp/ /tæp/ – /dæp/ /dæp/ – /dæp/ /tæp/ – /tæp/
5. 6. 7. 8.
/væl/ – /wæl/ /wæl/ – /væl/ /wæl/ – /wæl/ /væl/ – /væl/
5. 6. 7. 8.
/hut/ – /gut/ /gut/ – /hut/ /gut/ – /gut/ /hut/ – /hut/
5. 6. 7. 8.
/b∧gk/ – /d∧gk/ /d∧gk/ – /b∧gk/ /d∧gk/ – /d∧gk/ /b∧gk/ – /b∧gk/
9. 10. 11. 12.
/bot/ – /pot/ /pot/ – /bot/ /bot/ – /bot/ /pot/ – /pot/
9. 10. 11. 12.
/dau/ – /tau/ /tau/ – /dau/ /dau/ – /dau/ /tau/ – /tau/
9. 10. 11. 12.
/vat/ – /wat/ /wat/ – /vat/ /vat/ – /vat/ /wat/ – /wat/
9. 10. 11. 12.
/hel/ – /gel/ /gel/ – /hel/ /gel/ – /gel/ /hel/ – /hel/
9. 10. 11. 12.
/bœs/ – /dœs/ /dœs/ – /bœs/ /dœs/ – /dœs/ /bœs/ – /bœs/
Discussion In this experiment we tested the effect of four different types of speech manipulation on discrimination ability of LN (language normal) and SLI (speech and language impaired) kindergartners in five different speech contrasts. A difference between the two groups of children was found. SLI kindergartners performed worse than LN kindergartners in all conditions. When looking at error percentages of children performing above chance, we found differences in the five speech contrasts, which indicates that some contrasts are more difficult than others. No differences were found between the four speech manipulations. There are several possible explanations that could lead to these results. First, the enhancement algorithm is meant to be a learning algorithm, which only has positive effects after intensive training. Second, we used natural speech instead of synthesized speech. Usually, synthesized speech is used in these kinds of experiments, because the stimulus set can be manipulated more easily. Differences between the speech conditions may therefore be larger in the traditional experiments, especially when fast transitional elements are enhanced. We still would expect an effect of slowing down the speech signal though. This was not found, neither in error percentages nor in reaction times. Slowing down the complete speech signal is not what was tested in earlier experiments, where only the fast transitional elements were slowed down. In the 1996 Science articles (Tallal et al., 1996; Merzenich et al., 1996), for the first time an algorithm is described which slows down the entire speech signal. In the Fast ForWord program, this algorithm is only used in part of the program. Maybe this part of the program is not what causes the positive effects. Leonard, McGregor, and Allen (1992) mention that the difficulty for SLI children is the fast formant transition relative to the rest of the speech signal.
(JB)SWL II_f8_109-118
9/24/02
2:45 PM
Page 117
The effects of speech manipulation
When the entire signal is slowed down, the formant transition is still relatively fast compared to the rest of the speech signal. It then may be concluded that slowing down fast transitional elements is in line with enhancing these elements. Both manipulations emphasize the difficult part of the speech signal. This could be investigated in future research.
References Adams, M. (1990). Beginning to read: Thinking and Learning about print. Cambridge MA: MIT Press. Bishop, D. V. M. (1997). Uncommon Understanding: Development and Disorders of Language Comprehension in Children. Hove: Psychology Press. Boersma, P. & Weenink, D. (1998). Praat, a system for doing phonetics by computer. (Version 3.8beta). Amsterdam: University of Amsterdam. Frumkin, B. & Rapin, I. (1980). Perception of vowel and consonant-vowel of varying duration in language impaired children. Neuropsychologica, 18, 4, 434–454. Leonard, L. B., McGregor, K. K. & Allen, G. D. (1992). Grammatical morphology and speech perception in children with specific language impairment. Journal of Speech and Hearing Research, 35, 1076–1085. Liberman, I. Y. & Shankweiler, D. (1985). Phonology and the problems of learning to read and write. Rase: Remedial and Special Education, 6, 6, 8–17. Merzenich, M. M., Jenkins, W. M., Johnston, P., Schreiner, C., Miller, S. L. & Tallal, P. (1996). Temporal processing deficits of language-learning impaired children ameliorated by training. Science, 271, 5245, 77–81. Maxwell, S. E. & Delaney, H. D. (1990). Designing experiments and analyzing data. A model comparison perspective. Brooks/Cole Publishing Company: Pacific Grove, California. Mody, M., Studdert Kennedy, M. & Brady, S. (1997). Speech perception deficits in poor readers: Auditory processing or phonological coding? Journal of Experimental Child Psychology, 64, 2, 199–231. Nagarajan, S. S., Wang, X., Merzenich, M. M., Schreiner, C. E., Johnston, P., Jenkins, W. M., Miller, S. & Tallal, P. (1998). Speech Modification Algorithms Used for Training Language Learning-Impaired Children. IEEE Transactions on rahabilitation engineering. (Vol.6, No. 3). Raven, J. C. (1965). Coloured Progressive Matrices. London: H. K. Lewis&Co. Reed, M. A. (1989). Speech perception and the discrimination of brief auditory cues in reading disabled children. Journal of Experimental Child Psychology, 48, 2, 270–292. Stoep, J. & Verhoeven, L. (2001). Family and classroom predictors of children’s early language and literacy development. In: T. Shanahan and F. V. Rodriguez-Brown (Eds.), National Reading Conference Yearbook, 49, 209–221. Tallal, P. & Piercy, M. (1973). Developmental aphasia: Impaired rate of non-verbal processing as a function of sensory modality. Neuropsychologica, 11, 4, 389–398. Tallal, P. & Piercy, M. (1974). Developmental aphasia: Rate of auditory processing and selective impairment of consonant perception. Neuropsychologica, 12, 1, 83–93. Tallal, P. & Piercy, M. (1975). Developmental aphasia: The perception of brief vowels and extended stop consonants. Neuropsychologia, 13, 1, 69–74. Tallal, P. & Stark, R. E., Kallman, C. & Mellits, D. (1981). Perceptual constancy for phonemic categories: a developmental study with normal and language impaired children. Applied psycholinguisitcs, 1, 49–64. Tallal, P. & Stark, R. E. (1981). Speech acoustic cue discrimination abilities of normally
(JB)SWL II_f8_109-118
9/24/02
2:45 PM
Page 118
Precursors of Functional Literacy developing and language-impaired children. Journal of the Acoustical Society of America, 69, 568–574. Tallal, P., Miller, S. L., Bedi, G., Wang, X., Nagarajan, S. S., Schreiner, C., Jenkins, W. M. & Merzenich, M. M. (1996). Language comprehension in language-learning impaired children improved with acoustically modified speech. Science, 271, 5245, 81–84. Verhoeven, L., & Vermeer, A. (1986). Taaltoets Allochtone Kinderen. Tilburg: Zwijsen. Address University of Nijmegen Dept. of Special Education PO Box 9104 6525 HE Nijmegen, The Netherlands Phone: +31-(24)-3615901 Fax: +31-(24)-3616211 E-mail:
[email protected]
(JB)SWL II_f9_119-137
9/26/02
9:09 AM
Page 119
Part 2 Factors in Reading and Writing Efficiency
(JB)SWL II_f9_119-137
9/26/02
9:09 AM
Page 120
(JB)SWL II_f9_119-137
9/26/02
9:09 AM
Page 121
The acquisition of untaught orthographic regularities in French* Sébastien Pacton Laboratoire Cognition et Développement, CNRS-PARIS
Michel Fayol Université Clermont 2 and CNRS
Pierre Perruchet Université de Bourgogne and CNRS
Many alphabetic languages like English or French do not have one to one mapping between phonemes and graphemes. Those linguistic systems are not completely irregular however and there are various ways to solve the phoneme-to-grapheme inconsistencies. In certain cases, spellers can use graphotactic regularities (i.e., regularities at the level of the graphemes that are independent of regularities at the level of the phonemes, Jaffré & Fayol, 1997). For instance, English spellers can use their knowledge of the fact that /k/ is never transcribed “ck” in the beginning of words and French spellers can use their knowledge that the phoneme /o/, which can be spelled among other “o”, “au”, “ot”, “eau”, is never spelled “eau” in the initial position of words. In other cases, spellers can solve phoneme-to-grapheme inconsistencies by using morphological information. For instance, English spellers can use their knowledge of the spelling of the word “heal” in order to spell the related word “health”. Likewise, French spellers can use their knowledge of the fact that the sound /et/ is transcribed “ette”, rather than “ète”, “aite” or “ête” when it corresponds to a diminutive suffix (e.g., “une fille” means a girl; une fillette means a little girl). Classical models of spelling development have depicted children’s initial spellings as an attempt to spell words strictly on the basis of sound, without regard for acceptable letter sequence or other conventions of their written language (e.g., Ehri, 1986; Frith, 1985; Gentry, 1982; Henderson, 1985; Marsh, Friedman, Welch & Desberg, 1980). It would be later only that children reach an “orthographic stage” in which they grasp the higher order, more sophisticated, aspects of the nature of written language. However, recent studies have shown that children use a wide range of sources of information in spelling, albeit imperfectly, very early during spelling development (e.g., for a naturalistic study, Treiman, 1993; for
(JB)SWL II_f9_119-137
9/26/02
9:09 AM
Page 122
Precursors of Functional Literacy
experimental studies: Cassar & Treiman, 1997; Nation, 1997; Nation & Hulme, 1996; Pacton, Fayol & Perruchet, 1998; Pacton, Perruchet, Fayol & Cleeremans, 2001). These studies show the importance to develop theories of spelling acquisition that are less stage like and take into account the relationships between the different sources of information (e.g., phonological, morphological and lexical) that influence spelling (e.g., Lennox & Siegel, 1994, 1998; Seymour & Evans, 1994; Snowling, 1994). In what follows, we first review studies investigating the role of graphemic and morphological regularities in spelling acquisition. Then, we report two studies that explore the impact of graphotactic and morphological regularities, as well as the interactions between those two orthographic features on French children’s spelling of nonwords including phonemes that can be transcribed with various graphemes.
Graphemic regularities in spelling acquisition In Frith’s (1985) model, after an alphabetic stage characterized by the strict use of phoneme-grapheme correspondences, children would move into an orthographic stage of development when extensive reading and spelling experience has been provided. A first set of studies aimed at testing this hypothesis investigated whether and when the spelling of nonwords can be lexically biased. Early experimental studies have suggested that analogies are used only late in development (Campbell, 1985; Marsh et al., 1980). Campbell (1985) used an experimental technique called “lexical priming” in which children heard a mixed list of words (e.g., crane or brain) and nonwords (e.g., /prein/) and were asked to ignore the words but attempt to spell the nonwords. She found that the spelling of the nonwords was biased by the words previously heard only for children with a reading age of greater than 11 years. For instance, the nonword /prein/ was more often spelled “prane” when children had previously heard the word “crane” and “prain” when they had previously heard the word “brain”. An important limit to these studies, however, is that they did not ensure that the children do spell words such as brain or crane correctly. Contrary to these studies, Goswami (1988) reported that children could much earlier (i.e. 6-year-old) use the spelling pattern of a clue word remaining in children’s view in order to spell a target word, especially when the analogies were based on the rime unit. For instance, the presentation of the word “beak” helped children to spell the word “peak”. Using the same paradigm, Nation and Hulme (1996) also found that 6-year-old children made analogies between a visible clue word and a similar sounding target nonword (Experiment 1). However they reported that analogies occurred to the same extent regardless of whether the unit shared by the clue and the target words was a rime, a consonant + a vowel or a vowel alone. Moreover, they obtained similar results when the clue word was not visible (i.e., as in Campbell’s experiment), showing that children did not use analo-
(JB)SWL II_f9_119-137
9/26/02
9:09 AM
Page 123
The acquisition of untaught orthographic regularities in French
gies when spelling unfamiliar words simply because the clue word acts as a visual prime (Experiment 2). Extending this finding, Nation (1997) showed that 8–9 year-old children were sensitive to rime unit frequency when spelling monosyllabic words (Experiment 1) and nonwords (Experiment 2) without using a priming paradigm. In a naturalistic study, Treiman (1993) showed that adherence to simple orthographic conventions or regularities begins very early by examining writings produced over the course of a school year by first graders whose teacher encouraged creative writing but did not stress correct spelling. For instance, children’s erroneous doubling involved more often frequently doubled letters (e.g., ll, ee) than letters that are never doubled in English (e.g., hh, kk). Likewise, children rarely used double consonants or ck in initial position where they never occur in English. This shows that untaught orthographic regularities that are without phonological counterpart influence children’s spellings earlier than previously thought. Similar results have been reported in studies involving nonwords judgment tasks in English (e.g., Cassar & Treiman, 1997; Treiman, 1993) and in French (Pacton et al., 2001). For instance, Pacton et al showed that, as early as in Grade 1, children’s judgments of nonwords reflected their sensitivity to the identity of the consonants that can (or can not) be doubled and to the legal position of double consonants. Thus, there is convincing evidence that, very early in spelling development, children’s orthographic behavior is influenced by graphemic regularities that do not have a phonological counterpart and that are not explicitly taught.
Morphological regularities in spelling acquisition According to Henderson (1985), the role of meaning would become conspicuous for morphological aspects such as the past tense ending from the third grade and above. However, it would be at best at the end of the elementary grades that the spelling relationships among words in terms of roots, origins, and meanings are used to assist spelling. Beers and Beers (1992) found that children’s productions were initially massively alphabetic and that their ability to use morphological information concerning three spelling patterns (-s for plurals, -ed for past regular verbs, and -ing for the continuous) develops only later. These results have been recently confirmed by Nunes, Bryant, and Bindman (1997) as concerns the -ed inflection. In their longitudinal study, children first spelled past regular verbs with little regard for their morphological basis. In a second phase, children generalized the written form “ed” to grammatically inappropriate words (e.g., writing sofed for soft) as if they would consider the “ed” spelling pattern merely as an exception to the phonemic system. Then, in a third phase, children’s over-generalizations were confined to the right grammatical category (e.g., keped for kept). Finally, in a fourth phase, children used “ed” only for regular verbs. The initial concentration on letter-sound correspondences, rather than on
(JB)SWL II_f9_119-137
9/26/02
9:09 AM
Page 124
Precursors of Functional Literacy
morphology, has also been observed in Treiman’s (1993) naturalistic study in which US first graders spelled the past regular with the correct “ed” ending in only about 12% of the time. However, in subsequent experimental studies, Treiman, Cassar, and Zukowski (1994), and Treiman and Cassar (1996) have reported results that could be indicative of an early use of morphology. For example, Treiman et al. (1994) found that children were more likely to spell correctly the “t” of twomorpheme words based on a stem ending with /t/ plus a suffix (e.g., dirt-y), than the “t” of one-morpheme words which contain no smaller related word (e.g., city). Recently, with a pseudo-word spelling task, Bryant, Nunes, and Snaith (2000) have investigated whether 8- to 11-year-old children learn implicitly that English verbs whose stems sound the same in the present and past forms (e.g., clear and peel) are given the “ed” spelling (e.g., cleared; peeled) whereas verbs whose stems sound different in the present and past have phonetically spelled endings (e.g., heard; slept). Regular pseudo-verbs, whose stems sound the same in the present and past (e.g., [/krel/ — /kreld/] and irregular pseudo-verbs, whose stems sound different in the present and past (e.g., [/prel/ — /prold/] were embedded in sentences such as “My friend always prells at bedtime. We usually prell in the morning, but last week we /prold/ in the afternoon”. Children spelled regular pastpseudo-verbs with an “ed” ending more often than they did the irregular ones and, conversely, spelled the endings of the irregular past pseudo-verbs phonetically more often than they did the regular ones.
What types of linguistic information do French spellers use in spelling? The transcription of certain phonemes (e.g., /o/ in French) can be constrained by both graphotactic and morphological regularities. This characteristic is interesting because it allows to study the interactions between those two types of constraints, the influence of which is most of the time explored independently. The experiments reported in this chapter focus on the transcription of /o/ in French for four reasons. First, there are many possible transcriptions of /o/ (e.g., “o”, “au”, “eau”, “ot”, “aut”, “aud”, “os”, “aux”, “eaux”, “ho”, “hau”) that vary in terms of frequency. The most frequent transcription of /o/ is “o”. Transcriptions such as “aud” or “os” are far more rare. Secondly, the transcription of /o/ depends on its position within words. For instance, /o/ is transcribed “ho” and “hau” only in initial position of words. The transcriptions “ot”, “aut”, “aud”, “os” and “aux” occur only in final position of words. Note that, using “aut” or “ot” instead of “o” or “au” does not lead to a modification in the phonological form of words in final position but does in non-final position. However, the grapheme “eau” which occurs frequently at the end of words, infrequently in medial position of words and never at the beginning of words is pronounced /o/ wherever it occurs within words. Thirdly, the transcription of /o/ varies as a function of its consonantic context. For instance, in medial position, /o/ is more often spelled “o” than “au” between “b” and “r” but is more often spelled “au” than “o” between “p” and “v”. Likewise,
(JB)SWL II_f9_119-137
9/26/02
9:09 AM
Page 125
The acquisition of untaught orthographic regularities in French
/o/ is frequently transcribed “eau” after “r” or “t” but is never transcribed “eau” after “f ” at the end of words. Fourthly, the transcription of /o/ can be guided by morphology. In certain polymorphemic words, /o/ set in the final position, corresponds to a diminutive suffix (Catach, 1986). For instance, “éléphanteau” (baby elephant) and “renardeau” (fox cub) are two-morpheme words based on a stem “éléphant” (elephant) and “renard” (fox) followed by the diminutive suffix “eau”. The key point is that /o/ is transcribed “eau” when it corresponds to a diminutive morpheme.
The influence of graphotactic constraints on children’s transcription of /o/ We have asked 20 second graders, 20 third graders and 20 fourth graders to spell 48 tri-syllabic nonwords in which the phoneme /o/ was located in initial (16), medial (16) or final (16) position. We used a nonword spelling task because the use of a word spelling task to explore children’s sensitivity to orthographic regularities is problematic. Indeed, a child who has already seen a given word should retrieve it in his orthographic lexicon whereas another child who has never seen this word would spell it using regularities. In those conditions, it is not possible to determine whether a child writes a word such as “ordre” (order) correctly instead of “eaurdre” because he/she knows the spelling of this specific word or because he/she is sensitive to the fact that /o/ is never transcribed “eau” in the beginning of words. We first explored the variety of the spellings of /o/ used by children at each grade level. We also assessed the influence of the position of /o/ within nonwords on its transcription. For instance, we explored whether and when children used the grapheme “eau” more often in final than in initial and medial positions. Finally, we investigated the impact of the consonantic context in which /o/ occurred on its transcription. We explored whether children used differently the graphemes “o” and “au” as a function of the consonants which follow /o/ when /o/ occurred in initial position and as a function of the consonants which precede and follow /o/ when /o/ occurred in medial position. For instance, we assessed whether /o/ was more often spelled “au” for nonwords such as /povila/ (/o/ is frequently spelled “au” between “p” and “v”) than for nonwords such as /borile/ (/o/ is infrequently spelled “au” between “b” and “r”). We also explored whether and when children’s use of the grapheme “eau” differed as a function of the preceding consonants when /o/ occurred in final position by assessing whether children transcribe /o/ “eau” more often for nonwords such as /bitavo/ (/o/ is frequently transcribed “eau” after “v”) than for nonwords such as /bylefo/ (/o/ is never transcribed “eau” after “f ”). The orthographic regularities exploited in this study have been obtained using a computerized database for written and spoken French Brulex (Content, Mousty & Radeau, 1990). Although this database may be unrepresentative of the words children are exposed to, we have assumed that the infra-lexical orthographic regularities present in this database do not differ from those present in children books. In 16 nonwords, /o/ occurred in the beginning of nonwords, followed by one
(JB)SWL II_f9_119-137
9/26/02
9:09 AM
Page 126
Precursors of Functional Literacy
of the eight following consonants /b/, /d/, /f/, /g/, /m/, /p/, /r/ and /t/ (e.g., /obevy/; /obidar/). The phoneme /o/ always formed the first syllable of those items and the consonants which followed /o/ always belonged to the second syllable. Sixteen nonwords included the phoneme /o/ in medial position (i.e. neither at the beginning, nor at the end). In eight nonwords, called “o > au”, /o/ occurred in a consonantic context (the consonants that precede and follow /o/) in which /o/ is more frequently spelled “o” than “au” in French (e.g., /borile/; /ribore/: /o/ is more often spelled “o” than “au” between “b” and “r”). In eight other nonwords called “au > o”, /o/ occurred in a consonantic context in which /o/ is more frequently spelled “au” than “o” in French (e.g., /povari/; /ripove/: /o/ is more often spelled “au” than “o” between “p” and “v”). The phoneme /o/ and the following consonants of these 16 nonwords belonged to two different syllables. In 16 other nonwords, the phoneme /o/ was in final position. In eight nonwords, called “eau Frequent”, /o/ occurred in a consonantic context (i.e. the consonant that precede /o/) in which /o/ is frequently transcribed “eau” in French (e.g., /bitavo/: /o/ is frequently transcribed “eau” after “v”). In eight other nonwords, called “eau Never”, /o/ occurred in a consonantic context in which /o/ is never transcribed “eau” in French (e.g., /bylefo/: /o/ is never transcribed “eau” after “f ”). Children were told that the experimenter had made up new words that no one had ever seen or heard before and that their task consisted in listening to and writing these “new words” as they would do in a dictation, when they encounter new words they ignore the spelling form.
Number of different graphemes used by children to transcribe /o/ Table 1 shows the number of different spellings of /o/ used by children. Only one participant, a second grader, used one unique grapheme (“o”). Among the other second graders, six used two graphemes and 13 used at least three different graphemes. Four third graders used only two different graphemes; the 16 others used at least three different graphemes. Every fourth graders used at least three different graphemes. Thus, the variety of the spellings of /o/ increased with grade level, with differences according to the position of /o/ within nonwords. Influence of the position of /o/ within words Table 1 shows that participants used more different graphemes in initial position, and even more in final position, than in medial position. This corresponds to the distribution of the possible transcriptions of /o/ in French: four alternatives in initial position (“o”, “au”, “ho” or “hau”); three in medial position (“o”, “au” and, very rarely, “eau”); at least eight in final position (“o”, “au”, “eau”, “ot”, “aut”, “aud”, “os”, “aux”). Figure 1 indicates the mean percents of occurrences of graphemes — used to transcribed /o/ — as a function of the grade level and the position of /o/ within the items. At each grade level, “o” was the most frequent transcription of /o/ (73.4% in Grade 2; 47.6% in Grade 3; 52.2% in Grade 4) and “au” was the second most frequent transcription of /o/ (21.8% in Grade 2; 35.8% in Grade 3; 30.8% in Grade
(JB)SWL II_f9_119-137
9/26/02
9:09 AM
Page 127
The acquisition of untaught orthographic regularities in French Table 1. Number of graphemes used by children to transcribe /o/ (20 children per grade) Number of graphemes 1
2
3
4
>4
Overall positions
grade 2 grade 3 grade 4
1 0 0
6 4 0
11 4 7
1 6 8
1 6 5
Initial Position
grade 2 grade 3 grade 4
7 1 1
10 11 10
3 5 8
0 3 1
0 0 0
Medial Position
grade 2 grade 3 grade 4
3 0 0
14 11 10
3 9 10
0 0 0
0 0 0
Final Position
grade 2 grade 3 grade 4
2 3 0
10 1 0
6 6 12
1 9 4
1 1 4
4). Those two graphemes occurred in the three (initial, medial and final) positions. The other transcriptions were very less frequent and varied as a function of the position of /o/ within the items (see Figure 1). At each grade level, children spelled /o/ “ho” or “hau” only at the beginning of the nonwords. Likewise, they spelled /o/ “ot”, “aut”, “aux”, “aud” and “os” only in final position. However, this could be due to the fact that those graphemic forms end by consonants that are silent in final position but are pronounced in non-final positions. The case of the grapheme “eau”, which is the third most frequent transcription of /o/ at each grade level (3.7% in Grade 2; 9.0% in Grade 3; 11.3% in Grade 4), is interesting because “eau” is pronounced /o/ irrespective of its position within words. The grapheme “eau” was more often used in final position (18%) than in initial (2.2%) and medial (4.1%) positions. This effect was significant as early as in Grade 2 where children used “eau” more often in final position (7.0%) than in initial (2.2%) and medial position (2.0%) and the size of this position effect (initial and medial vs. final) increased with grade level (5.0% in Grade 2; 15.0% in Grade 3 and 23.7% in Grade 4). Note however that spellers who used “eau” became more and more numerous (7/20 in Grade 2, 15/20 in Grade 3 and 20/20 in Grade 4) but, that at each grade level, about one third of them (2/7 in Grade 2; 4/15 in Grade 3 and 5/20 in Grade 4) used “eau” even in initial position where it never occurs in French. This suggests that many of them did not rely on a rule specifying that /o/ is never transcribed “eau” at the beginning of French words.
Influence of the consonantic context In French, the relative proportions of “o” and “au” to transcribe /o/ in initial position vary as a function of the consonants that follow /o/. The characteristics of
(JB)SWL II_f9_119-137
9/26/02
9:09 AM
Page 128
Precursors of Functional Literacy Initial position
60
38.5
33.6
59.1
71.6
40
25.6
80
33.6
eau
eau
ot
au eau
o
0.0 0.0 0.0
0.0 0.0 0.0
0
0.6 4.4 4.7
20 2.2 2.9 1.6
% of occurrences of graphemes
100
aut, aux, aud, os
Medial position 79.4
80
66.0 65.6
28.9 29.2
60 40
o
eau
ho, hau
ot
au
0.0 0.0 0.0
0.0 0.0 0.0
0
0.0 0.0 0.0
20
1.9 5.1 5.2
18.7
% of occurrences of graphemes
100
aut, aux, aud, os
Final position
69.2
80 60
o
au Grade 2
eau Grade 3
13.4
ot
3.2
ho, hau
1.3
1.6 5.3 9.2
0
0.0 0.0 0.0
7.0
20
19.0 27.1
40
21.0 19.3 24.7
35.8
43.1
% of occurrences of graphemes
100
aut, aux, aud, os
Grade 4
Figure 1. Proportion of occurrences of the different transcriptions of /o/ as a function of the position within nonwords
(JB)SWL II_f9_119-137
9/26/02
9:09 AM
Page 129
The acquisition of untaught orthographic regularities in French
the French orthography prevent to make a clear distinction between contexts in which “o” is more frequent than “au” versus contexts in which “au” is more frequent than “o” in initial position. Because this feature is gradual, rather than dichotomic, children’s sensitivity to the consonantic context has been assessed by computing correlations between the relative use of “o” and “au” as a function of the consonants that follow /o/ in Brulex database on the one hand and in children’s spellings on the other hand across the eight consonantic contexts (/ob/, /od/, /of/, /og/, /om/, /op/, /or/, /ot/). In order to compute those correlations, we averaged the proportion of /o/ transcribed “au”, rather than “o”, (a) in children’s spellings for the two nonwords in which /o/ occurred in the same consonantic context and (b) in Brulex for all the words in which /o/ occurred in the same consonantic context — for both Type and Token frequency. The correlation between the distribution of “o” and “au” as a function of the consonants that follow /o/ in children’s spellings and in Brulex database increased from r = .07 in Grade 2 to r = .53 in Grade 3 and r = .92 in Grade 4 when correlations were performed with Brulex Type frequency. Similar results were observed when correlations were performed with Brulex Token frequency (r = .21 in Grade 2; r = 56 in Grade 3 and r = .88 in Grade 4). This shows that children’s spelling of the initial /o/ became more and more influenced by the consonant that followed /o/ even though /o/ and this following consonant belonged to different syllables. In the medial position of French words, the relative proportions of “o” and “au” (to transcribe /o/) vary as a function of the consonants that precede and follow /o/. Figure 2 represents the mean proportion of /o/ spelled “au” (rather than “o”) as a function of grade level and consonantic context. The phoneme /o/ was more often transcribed “au” in Grade 3 (31.7%) and in Grade 4 (32.6%) than in Grade 2 (20.2%). There was a main effect of the consonantic context showing that /o/ was more often transcribed “au” for “au > o” items (43.1%) than for “o > au” items (13.3%). The size of this consonantic effect, that was significant as early as in Grade 2, increased with grade level (from 16.2% in Grade 2 to 32.3% in Grade 3 and 41.1% in Grade 4). In French, “eau” is used to transcribe /o/ mainly in final position. Previous analyses showed that children often used “eau” in final position. The use of “eau” further varies as a function of the consonants that precede /o/. Figure 3 represents the mean proportion of /o/ transcribed “eau” in final position as a function of the consonantic context in which /o/ occurred. The mean proportion of /o/ transcribed “eau” increased as a linear function of the grade level. There was a main effect of the consonantic context, indicating that /o/ was more often spelled “eau” for “eau Frequent” nonwords (30.8%) than for “eau Never” nonwords (4.6%). The amplitude of this context effect further increased with grade level (6.0% in Grade 2; 27.0% in Grade 3 and 45.8% in Grade 4). To summarize, as early as in Grade 2, children used many different written forms to transcribe /o/, varied their transcriptions of /o/ as a function of its position and its consonantic context and those effects increased with grade level. Those results show that the size of the sound-to-spelling correspondences on which
(JB)SWL II_f9_119-137
9/26/02
9:09 AM
Page 130
Precursors of Functional Literacy 50.0 45.0 40.0 % of “au”
35.0 30.0 25.0 20.0 15.0 10.0 5.0 0.0 Grade 2
Grade 4
Grade 3 o > au
au > o
Figure 2. Proportion of /o/ transcribed “au” (rather than “o”) in medial position as a function of consonantic context (“o > au” nonwords vs. “au > o” nonwords) and Grade level
children base their spellings — when they spell a new word — is larger than the phoneme-grapheme unit. This confirms and extends previous studies showing that children’s spellings can not be simply described as an attempt to represent the sounds of their language without regard for orthographic regularities or conventions (e.g., Nation, 1997; Pacton et al., 2001; Treiman, 1993). It is worth stressing that the impact of the graphotactic constraints on children’s spellings has been obtained with a methodology different from the common priming paradigm (e.g., Campbell, 1985; Nation & Hulme, 1996) and with polysyllabic nonwords that differed from words more extensively than in most existing studies (e.g., /zisk/ that differs slightly from “disk”, Nation, 1997).
The influence of both graphotactic and morphological constraints on the transcription of /o/ We (Pacton, Fayol & Perruchet, 1999, submitted) have took advantage of the fact that, in French, the transcription of /o/ can be constrained by both graphotactic regularities and derivational morphology in order to assess how those orthographic constraints are integrated during spelling acquisition. Concerning graphotactic regularities, we exploited again the fact that the probability to transcribe “eau” the final /o/ varies as a function of the consonants that precede /o/. The influence of the graphotactic constraints has been assessed by
(JB)SWL II_f9_119-137
9/26/02
9:09 AM
Page 131
The acquisition of untaught orthographic regularities in French 60
% of “eau”
50
40
30 20 10 0 Grade 2
Grade 3 eau Never
Grade 4 eau Frequent
Figure 3. Proportion of /o/ transcribed “eau” in final position as a function of the consonantic context (“eau Frequent” nonwords vs. “eau Never nonwords”) and Grade level
asking children to spell nonwords such as /vitaro/ and /vitafo/ that differ only regarding the consonants that precede the final /o/. An effect of the graphotactic regularities should result in a wider use of “eau” for /vitaro/ (/o/ is frequently spelled “eau” after “r” in French) rather than for /vitafo/ (/o/ is never spelled “eau” after “f ” in French). Concerning morphological constraints, we exploited the fact that /o/ is spelled “eau” when it corresponds to a diminutive suffix. One week after having performed the above-mentioned task, the impact of this untaught morphological dimension — that can be described with a rule — on children’s spelling has been assessed by asking children to spell the same nonwords embedded within a sentence that provided information about the morphological structure of the nonword (i.e. a stem followed by the diminutive suffix /o/, e.g., “a little /vitar/ is a /vitaro/”). We postulated that a morphological effect would result in a wider use of “eau” in the “diminutive” (second) condition rather than in the “base” (first) condition.1 We further addressed the question of whether morphological constraints would reduce, or even suppress, the impact of graphotactic constraints in the “diminutive” condition. We hypothesized that, if spellers relied on an abstract rule such as ‘if the word ends in /o/ and if the word is a diminutive, then /o/ is transcribed “eau”’, the impact of the graphotactic constraints expected in the first “base” condition should no longer be observed in the second, “diminutive” condition.
(JB)SWL II_f9_119-137
9/26/02
9:09 AM
Page 132
Precursors of Functional Literacy
We have constructed six pairs of nonwords in order to test the impact of graphotactic and morphological constraints on the transcription of /o/. The two nonwords of each pair differed only with regard to their final consonants. One nonword ended with a consonant after which /o/ is frequently transcribed “eau” in French (e.g., /vitar/). The other nonword ended with a consonant after which /o/ is never transcribed “eau” in French (e.g., /vitaf/). A diminutive corresponding to each of those 12 nonwords was elaborated by adding /o/ after the final consonant of those nonwords. For example, the diminutive of /vitaro/ was composed of the stem /vitar/ followed by the diminutive suffix /o/. Hereafter, nonwords in which /o/ followed a consonant after which /o/ is frequently transcribed “eau” in French are labelled “eau Frequent” nonwords. Those in which /o/ followed a consonant after which /o/ is never transcribed “eau” in French are labelled “eau Never” nonwords. The experiment included two sessions separated by a one week interval. In the two conditions (“base” and “diminutive”), children had to spell the nonword preceded by the indefinite article “un” (a or an). The difference between the two conditions was that children heard “a vitaro” in the “base” (first) condition but heard “a little /vitar/ is a /vitaro/” in the “diminutive” (second) condition. Figure 4 represents the mean proportion of /o/ transcribed “eau” as a function of graphotactic constraints, morphological constraints and grade level. The use of “eau” differed as a function of the grade level (21.0% in grade 2, 17.9% in grade 3 and 34.6% in grade 5). There was a main effect of graphotactic constraints, showing that /o/ was more often transcribed “eau” for “eau Frequent” nonwords (33.6%) than for “eau Never” nonwords (15.4%). There was also a main effect of the morphological constraints, showing that /o/ was more often transcribed “eau” in the “diminutive” condition (29.6%) than in the “base” condition (19.4%). The Grade level by Morphological constraints interaction indicated that while “eau” was not significantly more often used in the “diminutive” condition — rather than in the “base” condition — in Grade 2 (+1.3%, both Fs < 1), this effect was significant in Grade 3 (+5.8%) and even more in Grade 5 (+23.3%). Importantly, the effect of the graphotactic constraints did not differ significantly according to whether nonwords were spelled in the “base” or in the “diminutive” condition (no Graphotactic constraints by Morphological constraints interaction) and this result was stable across grade levels (no Grade level by Graphotactic constraints by Morphological constraints interaction). To summarize, as in the previous study, graphotactic constraints influenced children’s spellings: as early as in Grade 2, the transcription of the final /o/ differed as a function of the consonants that precede /o/. With regard to morphology, from Grade 3 onward, /o/ was more often spelled “eau” in the “diminutive” condition than in the “base” condition. A major result was that the effect of the graphotactic constraints persisted in the “diminutive” condition and, further, that the size of the graphotactic effect did not differ according to whether nonwords were spelled in the “base” condition or in the “diminutive” condition in a very stable way across grade levels. The persistence of the effect of the graphotactic
(JB)SWL II_f9_119-137
9/26/02
9:09 AM
Page 133
The acquisition of untaught orthographic regularities in French “diminutive condition”
60
60
50
50 % of eau
% of eau
“base” condition
40 30
40 30
20
20
10
10
0
Grade 2
Grade 3
Grade 4
0
Grade 2
Grade 3
Grade 4
“eau Never” nonwords “eau Frequent” nonwords
Figure 4. Proportion of /o/ transcribed “eau” as a function of morphological constraints (base vs. diminutive), graphotactic constraints (“eau Frequent” nonwords vs. “eau Never nonwords”) and Grade level
constraints in the “diminutive” condition in spite of the possibility to rely on an orthographic rule suggests that, even after at least five years of exposure to print, children did not rely on a rule specifying how to transcribe /o/ when it stands for a diminutive suffix. Indeed, reliance on such a (morphological) rule would predict that morphological constraints should suppress or, at least, reduce the effect of the graphotactic constraints in the “diminutive” condition.
Discussion In spite of the facts that “o” is the most frequent transcription of /o/, that “o” is the simplest graphemic form of /o/ (one single letter) and that reading instruction starts with the grapheme-phoneme association “o” → /o/, 19 out of the 20 first graders of our experiment did not systematically transcribe /o/ “o” but, on the contrary, used at least two different graphemes. This inclination to vary the possible transcriptions of /o/, which increased with grade level, echoes Treiman’s (1993) observation that American first graders used different graphemic forms (e.g., “c”, “ck” and “k”) in order to transcribe /k/. Our study also revealed that children’s spellings were influenced by positional regularities. For instance, children used “eau” more frequently in final position (where “eau” is frequent in French) rather than in initial and medial positions (where “eau” infrequently or never occurs in French) and this effect increased with grade level. This sensitivity to positional regularities is congruent with Treiman’s (1993) findings that American first graders used “ck” (to transcribe /k/) as well as double letters in medial and final positions, where they are legal in English, but rarely used “ck” or double
(JB)SWL II_f9_119-137
9/26/02
9:09 AM
Page 134
Precursors of Functional Literacy
letters in initial position where they never occur in English. It is worth stressing however that, in spite of the possibility to rely on a rule specifying that /o/ is never transcribed “eau” in initial position in French, about one third of the spellers of each grade level used “eau” sometimes at the beginning of the items. In the two experiments reported in this chapter, children’s spelling of /o/ was influenced by the consonantic context in which it occurred. Those contextual effects indicate that children based their sound-to-spelling correspondences on units that are larger than the phoneme-grapheme unit when they spell new words. These effects were significant from the second grade onward in medial and final positions. Furthermore, in Grade 3 and even more in Grade 4, children’s spelling of the initial /o/ was influenced by the consonants that follow /o/ even though /o/ and the following consonants belonged to different syllables. This thus suggests that children’s spelling were influenced by regularities of the written language that go beyond the syllabic unit. With regard to the influence of morphology on children’s spelling, we showed that, from Grade 3 onward, children’s use of “eau” increased when the nonwords were embedded within sentences that make their morphological structure clear (i.e. a stem followed by the diminutive suffix “eau”). This is much in accord with Bryant et al.’s (2000) study in which English children’s spellings of pseudo-words were influenced by inflectional morphology. Note that in a control experiment, we have shown that embedding the same nonwords within sentences such as “a tall /vitar/ is a /vitaro/” did not increase the use of “eau”. This indicates that the wider use of “eau” in the “diminutive” condition, rather than in the “base” condition, did reflect children’s use of the morphological information provided by the “a little . . .” sentences and not the insertion of nonwords within any sentences. Getting an effect of morphology with nonwords so early is all the more relevant since participants may be more inclined to use phoneme-grapheme correspondences with unfamiliar and meaningless items than with words. Therefore, one might expect the impact of morphology to be even more precocious when children write words. A specificity of the second experiment reported in this paper was to assess the joint influence of derivational morphology — an orthographic feature that can be described with a rule — and graphotactic regularities — an orthographic feature that is probabilistic — on the transcription of the same phoneme. We showed that the effect of the graphotactic constraints did not differ significantly according to whether nonwords were spelled in the “base” or in the “diminutive” condition and that this effect was stable across grade levels. Importantly, the persistence of the effect of the graphotactic constraints in the “diminutive” condition can not be explained by children’s trend to spell the nonwords as they previously did in the “base” condition. Indeed, in a control experiment (Pacton et al., submitted), similar results were obtained when the experiment was practiced in the reverse order namely, the “diminutive” condition followed by the “base” condition one week after. Thus, the persistence of the effect of the graphotactic constraints in the “diminutive” condition in spite of the possibility to rely on an orthographic rule shows that, even after at least five years of exposure to print,
(JB)SWL II_f9_119-137
9/26/02
9:09 AM
Page 135
The acquisition of untaught orthographic regularities in French
children did not rely on a rule specifying that /o/ is transcribed “eau” when it stands for diminutive suffixes. The persistence of the graphotactic effects in the “diminutive” condition of the present study meets the results reported by Pacton et al. (2001) in the case of the learning of regularities about the use of double letters in French. They assessed, for instance, whether children acquire genuine knowledge of the fact that consonants can only be doubled in medial position in French. They asked first to five graders to choose between one nonword including a doublet in legal (medial) position and another nonword including a doublet in illegal (initial or final) position the one which looked most like a word. The crucial point was that in half of the nonword pairs, doublets were formed with consonants that are doubled in French (e.g., “tiffol” and “ttifol”) whereas, in the other half, doublets were formed with consonants that are never doubled (e.g., “xihhel” and “xxihel”). The results showed that (a) children’s knowledge of the legal position of double consonants in French extended to never doubled consonants as early as in Grade 1 but (b) children’s performance remained greater with frequently doubled consonants rather than with never doubled consonants, without any trend towards a reduction of the magnitude of this effect over the 5 years of training they examined. Those results echo some of the results obtained in implicit learning studies in which transfer to novel material is never perfect, but, instead, depends on the familiarity of the material (the so-called transfer decrement effect, see Redington & Chater, in press, for a recent review). Taken as a whole, those results challenged the idea that learners acquire implicitly an abstract rule-based knowledge because an essential prediction of any system using abstract rules to represent its knowledge about some domain is that its transfer performance on novel items should be just as good as its performance on familiar items (e.g., Anderson, 1993; Smith, Langston & Nisbett, 1992; Whittlesea & Dorken, 1997). Note to conclude that the lack of reliance on orthographic rule does not seem to be restricted to untaught orthographic features. Indeed, Fayol and his colleagues (Fayol, Hupet & Largy, 1999; Largy, Fayol & Lemaire, 1996; Totereau, Barrouillet & Fayol, 1998; Totereau, Thévenin & Fayol, 1997) have shown that, instead of relying on the verb and noun agreement rule previously taught, 7 to 10 year-old children as well as adults retrieved instances — composed of word stem and its most frequent inflection. This leads both French children and adults to use in an erroneous way the nominal inflection “s” instead of the verbal inflexion “nt” for verbs which have nominal homophones than for verbs which do not, especially for homophones of which the nominal form is more frequent than their verbal counterpart.
Notes * This research was supported by the C.N.R.S (UMR 5022). . Experiment 2 also investigated the issue of the interactions between graphotactic regularities and morphological constraints by assessing children spell the sound /et/ which transcription varies as a function of the consonants that precede it and which is
(JB)SWL II_f9_119-137
9/26/02
9:09 AM
Page 136
Precursors of Functional Literacy systematically spelled “ette” when it corresponds to a diminutive suffix. However, because similar results were obtained for /o/ and /et/, only results relative to the transcription of /o/ are reported in this chapter.
References Anderson, J. R. (1993). Rules of mind. Hillsdale, NJ: Erlbaum. Beers, C. S. & Beers, J. W. (1992). Children’s spelling of English inflectional morphology. In S. Templeton & D. R. Bear (Eds.), Development of orthographic knowledge and the foundations of literacy (231–251). Hillsdale, NJ: Erlbaum. Bryant, P., Nunes, T. & Snaith, R. (2000). Children learn an untaught rule of spelling. Nature, 13, 157–158. Campbell, R. (1985). When children write nonwords to dictation, Journal of Experimental Child Psychology, 40, 133–151. Cassar, M. & Treiman, R. (1997). The Beginnings of Orthographic Knowledge: Children’s Knowledge of Double Letters in Words, Journal of Educational Psychology, 89, 631–644. Catach, N. (1986). L’orthographe française [The French spelling]. Paris: Nathan. Content, A., Mousty, P. & Radeau, M. (1990). Brulex. Une base de données lexicales informatisée pour le Français écrit et parlé. Année Psychologique, 90, 551–566. Ehri, L. C. (1986). Sources of difficulty in learning to spell and read. In M. L. Wolraich & D. Routh (Eds.), Advances in developmental and behavioral pediatrics (Vol. 7, 121–195). Greenwich, CT: JAI Press. Fayol, M., Largy, P. & Hupet, M. (1999). The acquisition of subject-verb agreement in written French: From novices to experts’ errors. Reading and Writing, 11, 153–174. Frith, U. (1985). Beneath the surface of developmental dyslexia, In K. E. Patterson, J. C. Marshall & M. Coltheart (Eds.), Surface Dyslexia: Neuropsychological and cognitive studies of phonological reading (301–330). London: Routledge & Kegan Paul. Gentry, J. R. (1982). Analysis of developmental spelling in GNYS AT WORK. The Reading Teacher, 36, 192–200. Goswami, U. (1988). Children’s use of analogy in learning to spell, British Journal of Developmental Psychology, 6, 21–33. Henderson, L. (1985). Teaching spelling. Boston: Houghton Mifflin. Jaffré, J. P. & Fayol, M. (1997). Orthographe: Des systèmes aux usages [Spelling: From systems to use]. Paris: Flammarion. Largy, P., Fayol, M. & Lemaire, P. (1996). On confounding verb/noun inflections. A study of subject-verb agreement errors in French. Language and Cognitive Processes, 11, 217–255. Lennox, C. & Siegel, L. S. (1994). The role of phonological and orthographic processes in learning to spell. In G. D. A Brown & N. Ellis (Eds.), Handbook of spelling: Theory, process and intervention (93–109). Toronto: Wiley. Lennox, C. & Siegel, L. S. (1998). Phonological and orthographic processes in good and poor spellers. In C. Hulme & R. M. Joshi (Eds.), Reading and spelling: Development and disorders (395–404). Hillsdale, NJ: Erlbaum. Marsh, G., Friedman, M. P., Welch, V. & Desberg, P. (1980). The development of strategies in spelling. In U. Frith (Ed.), Cognitive processes in spelling (339–353). New York: Academic Press. Nation, K. & Hulme, C. (1996). The automatic activation of sound-letter knowledge: An alternative interpretation of analogy and priming effects in early spelling development. Journal of Experimental Child Psychology, 63, 416–435. Nation, K. (1997). Children’s sensitivity to rime unit frequency when spelling words and nonwords, Reading and Writing: An Interdisciplinary Journal, 9, 321–338.
(JB)SWL II_f9_119-137
9/26/02
9:09 AM
Page 137
The acquisition of untaught orthographic regularities in French Nunes, T., Bryant, P. & Bindman, M. (1997). Morphological spelling strategies: Developmental stages and processes. Developmental Psychology, 33, 637–649. Pacton, S., Fayol, M. & Perruchet, P. (1998). How do young French spellers use different graphemes to transcribe one phoneme? The case of the phoneme /o/. European Writing Conferences (S.I.G. Writing, E.A.R.L.I.), Poitiers (France). Pacton, S., Fayol, M. & Perruchet, P. (1999). Implicit learning of morphology and graphotactic regularities affect children’s and adult’s judgments of nonwords. XIIth Congress of the European Society for Cognitive Psychology, Gand, Belgium. Pacton, S., Fayol, M. & Perruchet, P. (submitted). Integrating graphotactic and morphological constraints in spelling acquisition. Pacton, S., Perruchet, P., Fayol, M. & Cleeremans, A. (2001). Implicit learning out of the lab: The case of orthographic regularities. Journal of Experimental Psychology: General, 130, 401–426. Redington, M. & Chater, N. (in press). Knowledge representation and transfer in artificial grammar learning. In R. A. French & A. Cleeremans (Eds.), Implicit Learning. London: Routledge, Psychology Press. Seymour, P. H. K. & Evans, H. M. (1994). Sources of constraint and individual variations in normal and impaired spelling. In G. D. A. Brown & N. Ellis (Eds.), Handbook of spelling: Theory, process and intervention (129–153). Toronto: Wiley. Smith, E. E., Langston, C. & Nisbett, R. E. (1992). The case for rules in reasoning. Cognitive Science, 16, 1–40. Snowling, M. (1994). Towards a model of spelling acquisition; the development of some component skills. In G. D. A. Brown & N. Ellis (Eds.), Handbook of spelling: Theory, process and intervention (111–128). Toronto: Wiley. Totereau, C., Barrouillet, P. & Fayol, M. (1998). Overgeneralizations of number inflections in the learning of written French. The case of nouns and verbs. British Journal of Developmental Psychology, 16, 447–464. Totereau, C., Thévenin, M. G. & Fayol, M. (1997). The development of the understanding of number morphology in French. In C. Perfetti, L. Rieben & M. Fayol (Eds.), Learning to spell (97–114). Mahwah: Erlbaum. Treiman, R. (1993). Beginning to spell: A study of first-grade children. New York: Oxford University Press. Treiman, R. & Cassar, M. (1996). Effects of morphology on children’s spelling of final consonant clusters. Journal of Experimental Child Psychology, 63, 141–170. Treiman, R., Cassar, M. & Zukowski, A. (1994). What types of linguistic information do children use in spelling? The case of flaps. Child Development, 65, 1310–1329. Whittlesea, B. W. A. & Dorken, M. D. (1997). Implicit learning: indirect, not unconscious, Psychonomic Bulletin and Review, 4, 63–67. Addresses Laboratoire Cognition et Dèveloppement, UMR 8605, CNRS-PARIS V, 71, avenue Edouard Vaillant, 92774 Boulogne-Billancourt Cedex, France, Institut de
[email protected] LAPSCO/CNRS, 34, avenue Carnot, 63000 Clermont Ferrand, France, Michel.Fayol@ srvpsy.univ-bpclermont.fr LEAD/CNRS, Faculté des sciences, Université de Bourgogne, 6 Bd Gabriel, 21000 Dijon, France,
[email protected]
(JB)SWL II_f10_138-163
9/27/02
11:00 AM
Page 138
(JB)SWL II_f10_138-163
9/27/02
11:00 AM
Page 139
Subsyllabic units in reading A difference between Korean and English Hye-Kyung Yoon*, Donald J. Bolger**, Oh-Seek Kwon** and Charles A. Perfetti** * Inje University Korea ** University of Pittsburgh
Onset-rime structures appear to be important in learning to read English. They provide, in the spoken language, a subsyllabic structure that is accessible to children prior to their ability to reliably access phonemes (Goswami & Bryant, 1990; Treiman, 1992). This highly accessible rime structure then can play a role in the beginning stages of reading, as children learn to map written words onto spoken units. Goswami (1993), for example, found that presenting children with rimebased analogies facilitates learning a set of written words. This support for learning comes from the possibility that beginning readers recognize that the spoken word plum has (at least) two sounds, the onset /pl/ and the rime /um/ without necessarily being aware that each further contains two segments. Beginning readers take advantage of these accessible units, in particular allowing the child to recognize that the written words plum, tum, and yum all share a rime in their um spelling. Evidence for this process came from the fact that children, having learned to read one word, could best read words that shared rimes, rather than other syllable parts, with the learned word. But is the value of onset-rime structure a universal feature of language (Fudge, 1969, 1987) or is it a linguistic-specific property? Correspondingly, is its value in supporting early reading a general property of how reading builds on spoken language or is a more specific property of reading certain languages written in certain orthographies? Research on the development of phonological awareness and word recognition has led to a number of theories about influential features of both the spoken language and writing systems, and the process of mapping the latter onto the former. McBride-Chang, Wagner, and Chang (1997) demonstrate that speech perception is an important precursor to phonological awareness, and that its effects on word reading may be mediated by phonological processing skill. Cross-linguistic research
(JB)SWL II_f10_138-163
9/27/02
11:00 AM
Page 140
Precursors of Functional Literacy
and analyses have become a necessary methodology to isolate effects of each of these systems on the development of phonological awareness and word recognition ability (see for example Cossu, Shankweiler, Liberman, Katz & Tola, 1988; Wimmer & Goswami, 1994). The Korean writing system provides an interesting comparison with studies done in English and European languages. In contrast to other writing systems, Hangul was more invented than developed, and its letter-phoneme correspondences are completely transparent. The main writing system of Korean, the Hangul, is alphabetic; however, unlike the Roman alphabetic system it is nonlinear. The composition of its letters follows a square structure, or Kulja, in which the letters are arranged left-to-right, top-to-bottom, as illustrated in Figure 1. Each square pattern contains up to 4 letters and corresponds to a single syllable. Linguistically, there are a number of differences between Korean and English that could be important for the issues we raise here. For one, Korean has fewer phonemes than English, an inventory of 19 consonants, 10 vowels (19 including diphthongs), and 2 glides. Also in contrast to English and European languages, there is no voicing contrast in Korean. Thus /k/ and /g/ are not distinguished. However, Korean distinguishes among three manners of stop consonants in terms of vocal tract constriction (tenseness); for example, /p/, /pp/, and /p’/ are three different levels of the voiceless bilabial. A tendency toward open syllables and a lack of consonant clusters provide further contrast with English and European languages. With these and other linguistic differences as potentially important for the issue we address, we can focus on the central question of whether this alphabetically written language, like English, awards some privilege in the use of onsetrime in reading. This question was addressed in a dissertation by H. K. Yoon (1997), who studied 4, 5, and 6-year-old Korean children’s preference for subsyllabic units. In her grapheme substitution task, Yoon taught a child to read a ‘clue word’, e.g., (/kal/). Then, with the clue word still visible, the child heard other words that share some part of the syllable /kal/. Some shared the rime, e.g., /dal/, others shared the first and final consonant, e.g., /kul/, and others shared the initial consonant and the vowel, /kam/. The child’s task was to select which part must be changed to produce a given target word from /kal/. For the /dal/ example, the child should select the first phoneme ‘k’, for the /kul/ example, the ‘a’, and for /kam/, the ‘l’. Note that in the first case, going from kal to dal, the rime unit is preserved, and should be the easiest case if the onset-rime structure is salient. However, Yoon (1997) found that subjects performed more accurately in substituting the final consonant grapheme task (kal to kam) than the middle vowel (kal to kul) or initial consonant grapheme (kal to dal). Although the grapheme substitution task is not identical to Goswami’s analogy task, there are enough similarities (e.g., a learned cue word to guide children’s responses) to encourage the conclusion that Korean children, unlike English children, assign no privilege to the onset-rime structure. Instead, they seem more sensitive to the syllable body, the onset plus the vowel.
(JB)SWL II_f10_138-163
9/27/02
11:00 AM
Page 141
Subsyllabic units in reading pattern 1
Nonterminal
pattern 2
pattern 3
C1
C1
V
V
C1 V
pattern 4
C1 V Terminal
C2
V
pattern 5
pattern 6
C1
C1
V
V
V
C1
C2
Figure 1. The 6 patterns of Kulja. Experiment 1 used patterns 4 and 5. Experiment 3 used 4, 5, and 6
Korean linguists Kim (1981) and Kwon (1987) first postulated the idea of the primacy of the body/coda structure in Korean. Similarly, evidence from word games and errors in speech and writing (Kim, 1981) have also lent support to the body/coda hypothesis. A number of empirical studies using adult Korean speakers further establish the use and preference for this pattern. In a Korean adapted replication of Fowler, Treiman, and Gross (1993), Yi (1998) asked subjects to perform a phoneme shift task. Subjects were presented with the pseudowords pang and sep, and then asked to substitute either the initial, middle, or final letter of the second syllable with the corresponding phoneme from the first syllable. Response time and error data suggested a preference for the body/coda structure in both visual and auditory presentation of the stimuli. Yi found similar effects in Korean using disyllabic (CVC-CVC) pseudowords, suggesting that this effect of syllable structure is not limited to word-level effects. In considering the differences between Korean and English, however, it is also useful to keep in mind what these observations share — the observation that there are intermediate units between the syllable and the phoneme that may be functional in supporting early reading. It is clear that the rime is well established as one such unit. However, it may not be universal. Both the spoken language and the writing system might serve to influence which intermediate units will be most functional. In the case of a Korean-English comparison, the writing systems are both alphabetic. However, the unique spatial arrangement of the Hangul system, along with other features of the system1, may encourage different processing strategies. Or it might be that because English is relatively unreliable in its grapheme-phoneme mappings, it encourages the onset-rime division in the interest of greater decoding reliability. Indeed, when rime units, rather than
(JB)SWL II_f10_138-163
9/27/02
11:00 AM
Page 142
Precursors of Functional Literacy
grapheme-phoneme pairings are the units of analysis, English turns out to be highly reliable (Treiman, Mullenix, Bijeljac-Babic & Richmond-Welty, 1995). Thus, it would be adaptive for children to focus on the more reliable rime units than on the less reliable grapheme-phoneme mappings. Korean is more reliable in its mappings than is English. With these several possibilities in mind, we summarize below experiments aimed at learning more about the basis of this difference between English and Korean. We start by attempting to replicate H. K. Yoon’s (1997) body-coda preference using a task which has reliably been used to establish the onset-rime unit, Goswami’s analogy task. Experiment 1 seeks to establish that the differences between Korean (Yoon, 1997) and English (Goswami, 1993) children are not due solely to differences between word analogy and grapheme substitution tasks. Experiments 2 and 3 examine the possibility that script differences between the two writing systems influences the process of acquiring grapheme-phoneme correspondence rules. Experiment 2 examined the performance on the grapheme-substitution task of English-speaking adults learning Korean Hangul as an artificial orthography. Experiment 3 examined Korean children to learn whether their syllable body (C1V2) preference emerges also in an analogy task. Experiment 4 turned to spoken language, using a sound similarity judgement task to directly test whether differences in subsyllabic structure preferences are also present in perceptions of the spoken languages.
Experiment 1 Before we can address our primary research questions, we felt that it was necessary to establish the consistency of the body-coda preference in Korean children as shown in Yoon’s (1997) Grapheme Substitution task. The findings of this study directly contradict those of Goswami’s (1993) English-speaking children learning to read novel words in an Analogy task. While the findings from both studies appear reliable in their own right, task demands in either case may have contributed to the main effects. To be conclusive on this point, a cross-methodological design was developed to establish the psychological validity of the body-coda effect. Experiment 1 tested Korean-speaking children with the Analogy task used by Goswami (1993). Because many Korean 5-year olds can read words, we used monosyllabic CVC nonwords rather than real words. Table 1 illustrates the comparison between the Grapheme Substitution task used in the previous experiments and Goswami’s Analogy task.
(JB)SWL II_f10_138-163
9/26/02
9:10 AM
Page 143
Subsyllabic units in reading Table 1. Comparing Methodologies: The Grapheme Substitution Task and Reading by Analogy Visual stimuli
Auditory stimuli
Visual stimuli
Auditory stimuli
CVC orthographic pattern
Corresponding pronunciation
CVC orthographic pattern
Corresponding pronunciation
mug
/mug/
mug
Grapheme Substitution Task (Yoon, 1997) ?
/mug/
Reading by Analogy (Goswami, 1993)
/dug/
dug
?
Table 2. Experiment 1: Analogy task Korean Hangul test stimuli and pronunciations CVC Pattern
Clue Word
Shared Unit C1V
VC2
V
C1C2
/jyal/
/jyap/
/syal/
/syap/
/jyul/
/byuk/
/byun/
/syuk/
/syun/
/byok/
/guap/
/guam/
/suap/
/suam/
/guop/
Pattern 4 Pattern 5 Pattern 6
Method Participants Twenty eight preschool children from the same kindergarten as Experiment 2 participated. Their mean age was 4 years and 5 months and all were native Korean speakers. Materials Korean CVC nonwords were developed to represent three of the basic six patterns of Kulja. These were Patterns 4, 5, and 6 illustrated in Figure 1. There were 3 clue words for each of the CVC Kulja forms. Each of the nine clue words was paired with 4 test nonwords. Each test word shared some part in common with the clue word; i.e. CV, VC, V, or CC. Table 2 shows examples of the nonwords used as tests for each type of clue word. Procedure An experimental session began with an explanation of the English word structure that included the idea that the initial consonant, the middle vowel, and the final
(JB)SWL II_f10_138-163
9/26/02
9:10 AM
Page 144
Precursors of Functional Literacy Table 3. Proportion of correct responses on each task condition Analogy condition (unit shared with CVC clue word) CVC Pattern
CC
VC
CC
V
sum
4 ( ) 5 ( ) 6 ( )
.494 (.3888) .512 (.4580) .345 (.3205)
.280 (.3982) .256 (.3941) .107 (.2877)
.238 (.3725) .274 (.4063) .155 (.3205)
.262 (.3887) .179 (.3040) .071 (.1839)
.3185 .3051 .1696
.4504
.2143
.2222
.1706
consonant part can be changed to make a new word. The experiment consisted of three phases. In the pre-reading phase, a child’s ability to read was tested on 4 nonwords. If the child was able to read a particular test word, it was replaced by another word of the same condition to insure novel stimuli were used in the analogy phase. In the word learning phase, the child received the printed clue word along with its pronunciation (e.g., /jyal/) and encouraged to learn to read the clue word. Finally, in the analogy phase, the test words were given to the child with a request to use the clue word to help read any of the nonwords that he or she could not read. The order of the test words was randomized for each child.
Results The main result concerns the comparison of analogy performance for rimes (shared VC) and bodies (shared CV). When clue words and test nonwords shared the CV, children correctly read the test nonwords on 45% of trials. This success rate was 21% when the clue word and test nonwords shared the VC, about the same as when they shared the CC, 22%. Nonwords that shared only the vowel produced a 17% success rate. Table 3 shows these data in proportion of correct responses. Planned comparisons among the 4 Analogy Types verified that CV analogies produced a higher success rate than the VC analogies. There were no reliable differences among VC, CC, and V analogies. The effect of Pattern was localized in the contrast between Pattern 6 and the other two types; the difference between Patterns 4 and the Pattern 5 condition was not reliable. The difficulty of Pattern 6 is arises either from a greater visual complexity (it presents 4 elements rather than 3) or a greater phonemic complexity (it contains two medial vowels) or both. The source of Kulja pattern differences is beyond our main purpose here. What is important is the generality of the CV > VC advantage over all 3 Kulja patterns. Discussion We now have a straightforward comparative result. Whereas English speaking children in the beginning stages of reading do better at reading rime-based analogies, Korean children at the beginning stages of reading do not; instead, Korean chil-
(JB)SWL II_f10_138-163
9/26/02
9:10 AM
Page 145
Subsyllabic units in reading
dren do better at reading syllable body (CV) based analogies. This replicates the results of Yoon (1997) with the kind of task used by Goswami (1993). These results cast doubt on the idea that the rime as a functional orthographic unit is universal. The patterns found in Yoon’s Grapheme Substitution results are not likely to be driven by task demands. Moving forward we are now able to ask questions about the influence of features of the written language in the development of functional rime or body units.
Experiment 2 One possibility for the Korean-English contrast lies in their scripts. Although both are alphabetic writing systems, the Korean script, in presenting a syllable unit in a recognizable spatial array, may promote a syllable parsing preference that favors the first and second graphemes. (Alternatively, one might say that the linear arrangement of letters in English discourages parsing the first two graphemes together.) An inspection of the 6 Kulja types in Figure 1 illustrates that possibility. Although most of the types cannot be said to group syllable bodies more than rimes, Pattern 4 for CVCs presents C1 and V on the same line, with C2 on the bottom line. Pattern 6 also separates the final consonant from the vowel. It is possible that the Korean preference for body + coda parsing arises from the script: C1 and V are more often grouped visual than are V and C2. To examine this possibility, Experiment 2 presented Korean children with linear English script, using English CVC words and nonwords. If the children show the same results with English script as in Hangul, the Korean script, this would at least suggest that it is not the script by itself that is responsible for the contrast between Korean and English.
Method Participants Twenty seven preschool children (7 age-four and 20 age 5) with a mean age of 5 years, 5 months, participated. All were kindergarten students in a central South Korean elementary school. All were native Korean speakers. Materials English CVC words and nonwords were used in a Grapheme Substitution task. There were 12 CVC clue word (in English/Roman script) and 12 nonword clues and their corresponding 12 spoken syllables. None of the syllables were words in Korean, although all could be considered legal nonwords. Four problems were given for each substitution location (initial, middle, and final substitution). An example in each task condition is shown in Table 4.
(JB)SWL II_f10_138-163
9/26/02
9:10 AM
Page 146
Precursors of Functional Literacy Table 4. Experiment 2: Grapheme Substitution task stimuli in English orthography Substitution condition
word
nonword
Initial consonant
bar/bar/ ➝ /par/ bog/bog/ ➝ /jog/ mug/mug/ ➝ /dug/ sum/sum/ ➝ /gum/
deg/deg/ ➝ /jeg/ lum/lum/ ➝ /hum/ sup/sup/ ➝ /jup/ zep/zep/ ➝ /lep/
Middle vowel
jig/jig/ ➝ /jag/ sin/sin/ ➝ /sun/ bar/bar/ ➝ /bur/ gut/gut/ ➝ /got/
bap/bap/ ➝ /bip/ mol/mol/ ➝ /mel/ dit/dit/ ➝ /dat/ lod/lod/ ➝ /led/
Final consonant
dog/dog/ ➝ /dot/ sum/sum/ ➝ /sun/ mug/mug/ ➝ /mut/ zip/zip/ ➝ /zig/
dit/dit/ ➝ /dil/ min/min/ ➝ /mig/ sup/sup/ ➝ /sut/ jep/jep/ ➝ /jeg/
Table 5. Grapheme Substitution procedure for CVC syllable in English orthography ➝
bun ➝ initial consonant substitution sal min ➝ middle vowel substitution dol dal mul ➝ final consonant substitution dam ➝ ➝
mun
Procedure Subjects were tested individually in the Grapheme Substitution task adapted from Yoon (1997) illustrated in Table 5. The first part of the experiment was an explanation of English writing. The explanation emphasized the location of the initial consonant, the middle vowel, and the final consonant part and was followed by an explanation of the substitution game. The experimenter then taught the child to read a clue word (e.g., mun ‡/mun/) on a card. While this card remained visible as a clue word, other words were presented. One shared a rime, and required the initial consonant to be substituted; e.g., /bun/. Another required vowel substitution /min/, an a third required final consonant substitution, e.g., /mul/. In the test phase, children were presented with a clue word that was written on a card and simultaneously spoken. The card remained in front of them while the new target word was spoken. They were then asked which part of the clue word should be changed to produce the spoken target word. Results The main result is a confirmation of the syllable body-coda preference. On average, target words that required final consonant substitutions, the indicator of bodycoda preferences, were correct 79% of trials. By contrast, words that required initial consonant substitution, the indicator of onset-rime preference, were correct on
(JB)SWL II_f10_138-163
9/26/02
9:10 AM
Page 147
Subsyllabic units in reading Table 6. Experiment 2: Mean number of correct response on each substitution task Substitution Position Lexicality
Initial consonant
Middle vowel
Final consonant
sum
Word Nonword
1.222 (1.423) 1.000 (1.468)
0.370 (0.839) 0.296 (0.869)
2.482 (1.397) 2.259 (1.678)
1.358 1.185
1.111
0.333
2.370
37% of trials. Poorest performance was for medial vowel substitution, which produced correct responses on only 11% of trials. Table 6 summarizes these data in terms of the mean number of correct responses. Planned comparisons confirmed that there were two differences in the Substitution Location effect. Initial consonant and final consonant substitution were reliably different (p < .05) as was the difference between middle vowel and final consonant substitution (p > .05).
Discussion These results further replicate those of Yoon (1997). Thus, whether in Korean with written Hangul or in English (unknown words for Korean children) with linear Roman letters, Korean children at the early stages of learning to read show a preference for the body-coda structure over the onset-rime structure. These results tend to cast doubt on the possibility that script differences as the source of the Korean-English contrast. It not only suggests that the Korean body-coda preference is not likely driven by the Hangul script, but it also suggests that the English script fails to drive an onset-rime preference in these speakers. To drive this point home, Experiment 3 asks if the reverse is true. That is, will Native-English speakers exhibit the onset-rime preference in the presence of the Korean script?
Experiment 3 Our first question was whether we would observe the rime preference in English speaking subjects using the Grapheme Substitution task. If so, then we would have increased reason to believe that the Korean-English contrast in subsyllabic structure is a genuine language or writing system difference rather than a result of task driven, methodological differences. The second question asked was whether these differences were driven from script rather than spoken language differences. If the onset-rime pattern persists in English subjects’ performance on the Grapheme Substitution task, using the Korean Hangul script, we can infer that the body-coda preference exhibited by Korean speakers is not driven by the script. For practical reasons, this study was carried out using English-speaking adults. Comparing this
(JB)SWL II_f10_138-163
9/26/02
9:10 AM
Page 148
Precursors of Functional Literacy
task with the data from Yoon’s (1997) Korean children may present some obvious differences with children learning to read. However, by utilizing Hangul as an artificial orthography, some general features of learning in an unfamiliar writing system can be mimicked. For instance, Artificial Orthographies use novel alphabet forms that confront adult learners with some of the challenges faced by a child learning to read (McCandliss, Schneider & Smith, 1997). Although the composition of artificial orthographies is arbitrary and any system can be used to study certain aspects of learning, the use of the Korean Hangul as an artificial orthography allows a comparison with Korean learners. When presented with a Grapheme Substitution task in an artificial orthography, would English adults show the Korean preference pattern (preference for onset + vowel) or the English (rime) pattern? Experiment 1 used the materials of Yoon (1997) with only minor modification in a Grapheme Substitution task. Three possible outcomes for these data arise when compared with the original results from Korean speakers. American adults could produce (1) a different pattern than Korean speakers consistent with English onset-rime syllable structure (Kessler & Treiman, 1997); (2) a pattern similar to the Korean subjects, implying that the demands of the task discourage an onsetrime preference; (3) a pattern of equally high performance across all units, reflecting the fact that the English adults’ high level of literacy and related phonological sensitivity allows easy access to all units, phonemes included.
Method Participants Thirty native English speakers, ages 18–24, participated for partial satisfaction of an undergraduate course requirement at the University of Pittsburgh. Material An artificial orthography was created from the Korean Hangul system. In the full Hangul system, there are 6 different syllable patterns (configurations of letters) that can represent the two basic Korean syllable types: consonant + vowel (C1V1) and consonant + vowel + consonant (C1V1C2). A syllable is encoded in a Kulja (visual syllabic unit) that contains 2–4 letters. The six patterns of Kulja vary according to the placements of consonant and vowel (see Figure 2.) In this experiment, only two Kulja patterns representing a C1V1C2 type syllable were used, pattern 4 (ex. ) and 5 (ex. ). For the experiment, there were 12 visually presented clue Kuljas and their corresponding 12 spoken syllables. The clue stimuli were equally divided into the two Kulja patterns, half used pattern 4 and half used pattern 5. For each grapheme position in the C1V1C2 Kulja, three grapheme substitution problems were created: initial substitution, middle substitution, and final substitution. As illustrated in Figure 1, a Pattern 4 Kulja is a left-right and bottom form (for example, ); Pattern 5 is top-middle-bottom form (for example, ). The following examples in Table 7
(JB)SWL II_f10_138-163
9/26/02
9:10 AM
Page 149
Subsyllabic units in reading Table 7. Grapheme Substitution procedure for C1V1C2 syllable in Hangul as an Artificial Orthography ➝ ➝ ➝
➝ initial consonant substitution ➝ middle vowel substitution ➝ final consonant substitution
Table 8. Experiment 3: Graphem Substitution task procedure in Korean Hangul Task
C1V1C2 Kulja
Example
Initial-consonant substitution
We read it /myung/. To become / /(/gyung/), which part is substituted or deleted? (with pointing to the initial, middle and final part)
Middle-vowel substitution
We read it /gyul/. To become / /(/gyol/), which part is substituted or deleted? (with pointing to the initial, middle and final part)
Final-consonant substitution
We read it /johng/. To become / (/johp/),which part is substituted or deleted? (with pointing to the initial, middle and final part)
illustrate how the Grapheme Substitution task employed the Kulja. In each case, the Kulja on the left (the clue Kulja) is converted to the Kulja on the right (the target Kulja) by the substitution of a single letter.
Procedure Subjects were tested individually, beginning with an explanation of the artificial orthography. The experimenter explained that in the Kulja word structure, the initial consonant, the middle vowel, and the final consonant part can be substituted to create a new word. The participant was then taught to read a clue word, e.g., (/kang/), which remained available on a card as a clue. In an experimental trial, presentation of the clue word was paired with the corresponding spoken syllable and then followed by a spoken syllable that shared segments with the clue word. For example, the clue word /kang/, could be followed by the spoken syllables /dang/ (initial consonant substitution), /kung/ (middle vowel substitution) and /kal/ (final consonant substitution). The participant was asked to indicate which part of the visual clue word would need to be changed in order to produce the new spoken syllable. For each of the three substitution patterns, there were four problems (2 in Pattern 4, 2 in Pattern 5). An example in each task condition is shown in Table 8. All twelve problems were presented in random order.
(JB)SWL II_f10_138-163
9/26/02
9:10 AM
Page 150
Precursors of Functional Literacy Table 9. Experiment 3: Mean number of correct responses in each substitution task Substitution Position Initial con(C1) kulja pattern M SD Sum
4
Middle vow(V)
5
1.62** (.622)
1.76 (.511)
4
5
1.10 (.817)
1.69
1.31 (.850) 1.21
Final con(C2) 4
5
1.24 (.831)
1.14 (.875)
1.19
** Maximum in each category = 2 Table 10. Experiment 3: Number of subjects who showed above 3 of 4 correct responses in each condition Substitution Position initial con.(C1) 26*
middle vow.(V) 14
final con(C2) 15
* Maximum in each category = 29
Results The important result is that when the target Kulja was produced from the Clue word by the substitution of the initial consonant, performance was superior to the other two substitution conditions. Thus, when the written syllable and the spoken syllable shared a rime, the task was easier. The average percent correct was 84.5% for initial consonant substitution, but only 60.5% and 59.5% for middle vowel and final consonant substitution, respectively. Planned comparisons reveal that this was a reliable effect. The mean number of correct responses on the substitution task are summarized in Table 9. Because there is 1/3 chance of guessing correctly, we also examined the results from a criterion perspective. If we take 3 out 4 problems correctly solved a criterion for individual success, we can ask how many individuals were successful for each substation type. Of 29 participants, these numbers were 26 for initial consonant, 14 for medial vowel, and 15 for final consonant as shown in Table 10. From both overall means and individual participants, the results support the conclusion that the rime is a privileged unit for English adults in the grapheme substitution artificial orthography task.
Discussion The results demonstrate that the Grapheme Substitution task is sensitive to more than one kind of subsyllabic structure. For Korean children, the task produces a
(JB)SWL II_f10_138-163
9/26/02
9:10 AM
Page 151
Subsyllabic units in reading 90.0%
Correct response
80.0% 70.0% Kor(Bgnr) 60.0%
American
50.0% 40.0% 30.0%
initial
middle
final
Replaced Grapheme Figure 2. Experiment 1. The Grapheme Substitution Task Comparing English Speaking Note. Adults and Korean Speaking Children (Yoon, 1997)
preference for body-coda structures; for English speaking adults, the task produces a preference for an onset-rime structure. This contrast can be seen in Figure 2, which shows Yoon’s (1997) Korean data and the results of Experiment 3. Of course, we must keep in mind that children and literate adults may differ in learning to read. It is likely that literacy supported the adult performance in Experiment 1. Although the spoken syllables were Korean, hence unfamiliar to our American participants, the Grapheme Substitution problems could have been solved by mapping the letters of the novel alphabet onto English alphabetic representations, which would serve as a basis for the decision. Also, the Korean speech sounds used in this experiment are represented much differently in the English orthography than in Korean. For example the word /myung/ is represented with 5 graphemes in English and the onset ‘my’ violates phonotactic regularities in English. However the task was performed in detail, there is no obvious explanation for the onset-rime preference. A letter strategy would make all letters available (subject to perhaps serial order effects) and a phonetic “segment” strategy leaves open the question to be addressed: Is some subsyllabic unit larger than the phoneme privileged in this processing task? Thus, whatever the processing details for this task, it appears that they entail a preference for the onset-rime structure, as suggested for other English language studies (Treiman, 1992; Goswami, 1993). An explanation for why Korean and English are different in this respect was the goal of the remaining experiments. We now have three results for reading that point to a genuine language related difference. Experiment 1 found that Korean children replicated the Korean bodycoda pattern using the Analogy task, previously demonstrated to produce the
(JB)SWL II_f10_138-163
9/26/02
9:10 AM
Page 152
Precursors of Functional Literacy
onset-rime pattern in English speaking children. Experiment 2 found that Korean children showed what we will now call the “Korean pattern” even when presented with English words in the linear Romanized format. Experiment 3 found that the English-speaking adults showed what we will now call the “English pattern” in a Grapheme Substitution Task using the spatially organized Korean Hangul. Given the results of Experiment 1–3, the Korean-English differences appear not to lie in the script. This result leads to the hypothesis that the differences originate in linguistic factors within the two spoken languages. Experiment 4 examines this hypothesis.
Experiment 4 The question turns now from the writing system to the language system. Evidence from various speech-related tasks is at least consistent with the assumption that spoken English syllables are perceived as having onset/rime (C-VC) structures (Fowler, 1987; Kessler & Treiman, 1977; Treiman, 1983, 1986). More recently, evidence similarly points to the possibility that Korean syllable structure is CV-C in spoken language tasks rather than C-VC (Yoon, 1995; Yoon & Derwing, 1994; Yi, 1998). If so, this fact would be sufficient to account for our observed contrasts in reading tasks. Thus, the Linguistic Hypothesis is that differences observed between Korean and English arise not in script differences but in differences between the two languages. The possible sources for phonological differences between Korean and English are numerous. To consider two that are especially relevant for differences in subsyllabic structures, we note (1) that Korean has a smaller inventory of syllable types, essentially only CV and CVC (V and VC also occur with less frequency), compared with English, (2) that Korean and English differ in the degree to which syllable boundaries are clear and unambiguous. The first of these facts means that Koreans do not encounter complex consonants in onsets. Accordingly, the distinction between onset and initial phoneme collapses — onsets are phonemes. Also, Korean rimes tend to end in simple codas. To the extent that onset-rime structures are especially useful only in syllables with complex onsets (or complex codas), Korean gains no value. However, because there is no reason to think that the onset-rime structure in English is restricted to its complex consonants, the role of syllable type would have to be mediated by some other factor. For example, the existence of complex consonants might make phoneme segmentation more difficult and allow syllable nuclei (vowel peaks) plus syllable closings to be perceived as more coherent. With generalization across syllable types, this would result in a highly general functional value for rime units in English. The second fact may be more important. English, but not Korean, is a syllable stressed language. In general, syllable-stress languages have greater variation in syllable “weight” than non-stress languages; a strong syllable is more perceptible than a weak syllable. This may serve to weaken syllable boundaries, as inter-
(JB)SWL II_f10_138-163
9/26/02
9:10 AM
Page 153
Subsyllabic units in reading
nal stressed vowel nuclei become perceptually salient in strong syllables, while weak syllables, and their boundaries, become perceptually invisible. Like other languages without syllable stress alternation, Korean has very clear syllable boundaries. With clear boundaries and even weighting on all syllables may come a stronger functioning of the syllable unit. Subsyllabic structures would be less salient.2 These differences illustrate the potential for speech-based perceptual differences between the two languages. The present study was designed not to localize the exact source of any such difference, but rather to discover whether the differences we observed in reading have a parallel in the spoken languages. Would Korean and English listeners, confronted with identical syllables, show differences in how they perceive these syllables? In particular, in judging the extent to which two spoken syllables are similar, would their judgments follow on rime preference or a syllable body preference? The Linguistic Hypothesis is that Korean speakers should show a syllable body preference and English speakers, a rime preference.
Method Participants Korean and American undergraduates students participated. Forty-nine students at the University of Pittsburgh (all native speakers of English) and 49 from the Inje University (all native speakers of Korean) served in partial fulfillment of introductory psychology requirements at their respective universities. Materials Pairs of monosyllabic CVC words and nonwords of were constructed such that a given CVC was a word in both Korean and English, or else was a nonword in both Korean and English. Pairs of disyllabic CVCVC nonword stimuli were similarly constructed. Forty-eight word or nonword pairs were constructed for each condition. As shown in Table 11, the stimulus pairs varied in the number of phonemes shared by the two members of the pair: CVCVC pairs shared 0-phonemes, 1 phoneme matched, 2-phonemes or all 3 phonemes. Members of a CVCVC pair shared 0, 1, 2, 3 or all 5 phonemes. CVCs were the main focus, because CVCs had been used in the other experiments. Thus CVC pairs could share CV (syllable bodies) or VC (rimes). Procedure Stimulus pairs were tape-recorded in a single random order and played to participants in individual experimental sessions. As shown in Figure 3, the words within a stimulus pair had an ISI of one second, and each pair was repeated once with an ISI of two seconds. After the pair was repeated, participants rated the sound similarity of the pair on a 7-point scale. Prior to the experimental sequence, participants received four practice CVC
(JB)SWL II_f10_138-163
9/26/02
9:10 AM
Page 154
Precursors of Functional Literacy Time:
1s
1s
2s
1s
4s
Tape: “Number 1”–/mug/–/mut—/mug/–/mut/——“Number 2” Subject: ————————————————— (Response) —— Figure 3. Presentation rate of Sound Similarity Judgement stimulus pairs Table 11. Examples of CVC stimulus pairs used in Experiment 4 Example pairs (C1VC2–C1VC2) Number of matched phonemes 0 1 1 1 2 2 2 3
Units matched
Word
nonword
none C1 C2 V C1V VC2 C1C2 all
/bag/-/sun/ /tug/-/tar/ /bam/-/sum/ /tug/-/bud/ /mug/-/mut/ /sum/-/gum/ /bum/-/bam/ /jip/-/jip/
/nig/-/wom/ /dit/-/dag/ /mol/-/pel/ /min/-/nig/ /jeb/-/jec/ /dep/-/zep/ /dit/-/dat/ /nig/-/nig/
Table 12. Examples of CVCVC stimulus pairs used in Experiment 4 Number of matched phonemes 5 4 4 4 3 3 3 3 2 2 2 0
Mismatched units
Example /CVCVC/-/CVCVC/
Num. of matched unit CV VC
none C1 C2 C3 C1V1 V1C2 C2V2 V2C3 C1V1C2 V1C2V2 C2V2C3 all
/sodak/-/sodak/ /tosam/-/posam/ /sodak/-/somak/ /cokam/-/cokan/ /kopak/-/tupak/ /mupak/-/mokak/ /tomik/-/tosak/ /konok/-/kokun/ /sutam/-/cokam/ /tocin/-/tusan/ /konok/-/kokun/ /kopak/-/tusin/
1 1 2 1 0 1 1 0 0 1
2 1 1 1 1 0 1 1 0 0
pairs with 0 to 3-phoneme overlap and five practice CVCVC pairs representing 0 to 5-phoneme overlap. Subjects were instructed to focus on the global impression of sound only and to rate the similarity in sound on a 7-point scale ranging from 0 (completely different) to 6 (exactly the same). The stimuli were presented in blocks with CVC word pairs first followed by CVC nonword pairs and then CVCVC nonword pairs.
(JB)SWL II_f10_138-163
9/26/02
9:10 AM
Page 155
Subsyllabic units in reading Table 13. Mean Sound Similarity rating on CVC Word/nonword Number of shared phoneme
Word
Nonword
Korean
English
Korean
English
sum
0 1 2 3
0.429 1.382 3.100 5.830
0.187 1.186 3.351 5.912
0.361 1.720 3.127 5.765
0.228 1.139 3.329 5.803
0.301 1.357 3.227 5.827
sum
2.685
2.659
2.743
2.678
Results Because our main purpose is to test the Linguistic Hypothesis on CVC syllables, we report here only those results. An important result is that both Korean and English speakers were sensitive to the number of shared phonemes. As can be seen in Table 13, as the number of shared phonemes between members of a pair increased, the mean similarity rating did also. The reliability of this effect was confirmed in a 2 (Word vs. Nonword) × 2 (Language) × 4 (Number of Shared Phonemes) Analysis of Variance (not reported here). With the general meaningfulness of the similarity ratings established, we can turn to the key question about subsyllabic structures. The data were then analyzed according to shared subsyllabic unit. These results for the critical rime comparison can be seen in Figure 4. For both words and nonwords, English speakers rated pairs that shared rimes as the most similar. Korean speakers rated words that shared syllable bodies as most similar. Consider first the VC (rime) results. Korean speakers gave a mean rating of 2.8 for word and 2.4 for nonwords (6 maximum) that shared VC; English speakers produced a mean rating of 3.9 and 3.7 for pairs sharing VC. The opposite pattern is seen for CV. Korean speakers produced a mean similarity rating of 4.1 and 4.0 for words and nonwords (respectively) sharing CV. English speakers produced mean similarity ratings of 3.23 and 3.08 for CV sharing word and nonword pairs, respectively. Thus, English speakers judged more similarity for pairs sharing a rime than pairs sharing a syllable body. Korean speakers judged more similarity for pairs sharing a syllable body than for pairs sharing a rime. Pairs that shared C_C were the lowest rated for words, but not for nonwords. Korean speakers rated nonwords with shared C_C higher than nonwords with shared VC. And English speakers rated C_C pairs no lower than CV pairs. A 2 (Word vs. Nonword) × 2 (Language) × 3 (subunit type) Analysis of Variance confirmed the reliability of the mean differences seen in Table 8. The effect of the subunit that shared phonemes (CV, VC, C_C) was reliable (p < .001) and the interaction between the subunit and the language user was also significant (p < .001). Lexicality, whether a stimulus pair was a word or nonword, had no main effect.
(JB)SWL II_f10_138-163
9/26/02
9:10 AM
Page 156
Precursors of Functional Literacy
Sound Similarity Rating
6 5 4 3
C1-C2 Shared
2
C-V Shared V-C Shared
1 0 English: Words
Nonwords
Korean: Words
Non-words
Subject Language by Word Pair Figure 4. Experiment 4: Mean similarity ratings for CVCs that shared two phonemes (CV) (VC) (C_C) for Korean speakers and English speakers
However, Lexicality did reliably interact with subunit, (p < .001). This interaction is due not to the rime and body ratings, which were general over words and nonwords, but to the C_C ratings, which were affected by whether the stimulus pair was a word. Planned comparisons found the expected differences between shared subunits and language user (English speakers: VC > CV, VC > C_C, CV = C_C; Korean speakers: CV > VC, CV > C_C, VC = C_C) to be reliable.
Discussion The results provide clear support for the Linguistic Hypothesis, that subsyllabic pattern preferences observed in reading arise from language differences, rather than script differences or other factors. Korean speakers and English speakers heard the same spoken syllables — and the syllables had the same lexical status in each language — but produced different similarity patterns. English language subjects rated syllables more similar if they shared a rime (VC); Korean language subjects rated syllables more similar if they shared a syllable body (CV). This pattern is the same as observed in the reading experiments. Of course, there is a big qualification that must be added to the interpretation of these results. The language-specific preferences observed in the experiment may arise from the structure of the spoken languages, the straightforward interpretation consistent with the Linguistic Hypothesis. Alternatively, they may arise because written representations influence performance of literate adult speakers,
(JB)SWL II_f10_138-163
9/26/02
9:10 AM
Page 157
Subsyllabic units in reading Table 14. Mean similarity scores for the eight types of CVC test pairs C1VC2 word
C1VC2 nonword
Matched unit
Korean
English
Korean
English
none C1 C2 V C1V VC2 C1C2 all
0.429 (0.566) 1.449 (0.902) 0.922 (0.635) 1.776 (0.844) 4.088 (0.834) 2.843 (1.043) 2.371 (1.113) 5.830 (0.317)
0.187 (0.415) 1.463 (0.775) 1.112 (0.746) 0.983 (0.687) 3.323 (0.813) 3.901 (0.832) 2.830 (0.982) 5.912 (0.415)
0.361 (0.499) 2.197 (0.739) 0.840 (0.787) 2.122 (0.865) 4.017 (0.924) 2.412 (0.889) 2.952 (1,132) 5.765 (0.412)
0.228 (0.508) 1.520 (0.697) 1.037 (0.744) 0.861 (0.754) 3.082 (0.804) 3.742 (0.873) 3.163 (1.186) 5.803 (0.608)
even when the task involves spoken language. In short, we cannot rule out the possibility that the fundamental difference between Korean and English comes from different reading strategies that have some other origin than the spoken language. We will take up this point again in the general discussion. So, until we learn about the performance of pre-literate speakers in spoken language tasks, the allowable conclusion is that subsyllabic preferences observed in reading are also observed in speech judgments.
General discussion Our goal was to investigate the characteristics of spoken and written language that lead to the development of subsyllabic units (onset-rime or body-coda) in reading. We began by asking if subsyllabic units in early reading, namely the onset and rime, were universal or language-specific. Secondly, we asked whether these units are indicative of a general early mapping process of orthography onto the spoken language or a property of particular orthographies in particular writing systems. The results of our experiments directly suggest that while the sensitivity to subsyllabic units may be universal, the form of preferred linguistic units may be language specific. Whereas native English speakers develop a preference for onset-rime units, native Korean speakers show a preference for a body-coda structure. Just as studies of English children find support for an onset-rime structure that emerges prior to acquiring literacy (see Goswami, 1993), evidence exists of Korean children’s use of the body-coda structure prior to acquiring reading skill (Yoon, 1997). In Experiment 1, we found further support for Korean children’s use of the body-coda pattern in Goswami’s Analogy task. Thus, we could no longer assume our findings were due to secondary, task-driven differences. A similar study conducted by Lee and Yi (1999) investigated Korean children’s body/coda preferences at different points in reading acquisition. Lee and Yi presented the aural
(JB)SWL II_f10_138-163
9/26/02
9:10 AM
Page 158
Precursors of Functional Literacy
version of Phoneme Shift task to three groups of children: kindergartners with no reading ability, kindergartners with some reading proficiency, and first graders with proficient reading and writing skill. The results of this study suggest that the body-coda preference is somewhat diminished as children acquire proficiency in reading. However, similar to the onset-rime in English, the body-coda pattern is present at a very early stage of reading development. Now that we have well established the differences between Korean and English patterns of performance, we turn our attention to the origination of these preferences. Both Goswami’s (1993) Analogy task and Yoon’s (1997) Grapheme Substitution task require subjects to utilize some knowledge of the orthographic system in the child’s native language. Similarly, Lee and Yi’s (1999) study seems to suggest that orthographic knowledge and reading skill may play a role in the body-coda pattern preference. The uniqueness of the kulja pattern of the Korean Hangul led us to question whether the script may be driving the effects seen in Korean children. In Experiment 2, we asked Korean children to perform the Grapheme Substitution task in a foreign script (Roman/English alphabet) where orthographic knowledge would not play a direct role in performance. Again, the body preference persisted. While it is possible that children may have mapped the English orthography onto their knowledge of Korean orthography, it is highly doubtful that this would occur given such little experience with their native script. Thus, our Korean pattern is seemingly not driven by experience with Hangul, or the Kulja pattern within the writing system. We speculate that the Roman-English orthography fails to drive the onset-rime pattern — since it had not for these children. The results of Experiment 3 confirm, at least for adult speakers of English, that the onset-rime preference persists in the presence of a novel orthography. This result also rules out the influence of the spatial arrangement of the Kulja driving body-coda preferences. In light of these findings, there is great cause to doubt the hypothesis that the rime structure is universal (Fudge, 1987). Since the Korean-English difference in Experiments 1–3 are not driven by the particular scripts of each language, we assume that differences must be driven by experience with spoken language. In Experiment 4, adult native language users of both Korean and English were asked to perform a speech perception task. In our Sound Similarity Judgment task, Korean subjects reliably rated item pairs according to a body-coda preference and American subjects according to an onset-rime structure regardless of the stimuli voice (Korean vs. English speaker). These results suggest that subsyllabic patterns are fairly well established phonological representations in adult language users. Because the voice of the speaker had little effect on similarity ratings, we conclude that subsyllabic units emerge as linguistic elements of speech perception in a speaker’s native language. Although we cannot entirely rule-out the possibility of an effect of literacy on speech judgements, we believe that such an effect would be greatly minimized. Thus, we assert a Linguistic Hypothesis that the emergence of subsyllabic units as functional elements in reading are language-specific and are derived from features of the spoken language system mapping onto the writ-
(JB)SWL II_f10_138-163
9/26/02
9:10 AM
Page 159
Subsyllabic units in reading
ing system. To further understand how these linguistic differences arise, we look at how these subsyllabic units emerge in reading performance. One general explanation for the use of sub-syllabic units in reading is that larger orthographic units of recognition develop in response to children’s experience with the writing system. In particular, Orthographic Rime units are argued to arise (1) because of the lack of transparency in the English orthography, (2) because of the frequency of the occurrence and use of these units from the child’s instructional/cultural background. According to the orthographic depth hypothesis, the fact that the English orthography lack transparency, i.e. one grapheme maps to multiple phonemes, encourages the use of larger and more reliable units in order to read proficiently. Treiman et al. (1995) analyzed the predictability of phonology from orthography and found greater regularity among rime units in the English language then in the body (CV) or vowel unit (V) alone. The second argument for Orthographic Rime units is that these units emerge from mere frequency of exposure (Treiman, Goswami & Bruck, 1990; Bowey, 1996; Bowey & Hansen, 1994). According to this view, children necessarily acquire graphemephoneme correspondences at the early stages of reading development, but as they become proficient at decoding, children utilize these larger units. This idea is supported by the orthographic rime frequency effect (Treiman, Goswami & Bruck, 1990; Bowey & Underwood, 1996), i.e. the fact that children read words like tain (high frequency rime) more accurately than taich (low frequency rime). What is important to note about these theories is that the emergence of subsyllabic orthographic units occurs as a function of experience with a particular orthography in a particular writing system. A second general explanation for the acquisition of subsyllabic units is that they develop as phonological structures in the process of gaining phonological awareness (Goswami & Bryant, 1990; Treiman, 1992). According to Goswami’s (1993) model, children at the early stages of reading instruction come equipped with phonological knowledge of onsets and rimes onto which orthographic recognition units are then mapped. This idea is supported by a number of findings identifying the onset-rime preference in spoken language tasks in adults and preliterate children (see Bradley & Bryant, 1983; Treiman, 1983, 1986; Treiman, Goswami & Bruck, 1990). Goswami’s theory thus assumes that subsyllabic phonological units arise prior to reading instruction, and that the phonological onsetrimes are the bootstrapping units for acquiring subsequent orthographic knowledge. The phonological structure hypothesis is consistent with our Linguistic Hypothesis of the Korean body-coda preference. As we stated earlier, speech perception is an important precursor to phonological awareness, and that its effects on word reading may be mediated by phonological processing skill (McBride-Chang, Wagner & Chang, 1997). The majority of studies on subsyllabic units have focussed on the onset-rime hypothesis in English and European languages. Because English orthography lacks transparency, we cannot rule out an orthographic factor in the emergence of these units, nor can we discriminate between phonological and orthographic accounts.
(JB)SWL II_f10_138-163
9/26/02
9:10 AM
Page 160
Precursors of Functional Literacy
However, studies in relatively transparent writing systems (e.g., Dutch, Turkish, German, and Korean) may allow for stronger conclusions. A series of experiments in Dutch reported in this volume, Geudens and Sandra (1999, this volume) investigated the occurrence of orthographic rime units in a relatively transparent orthography, in which the grapheme-phoneme mapping is nearly one-to-one. They argue, based on previous results, that the transparent Dutch writing system does not require onset-rime orthographic units, but that preferences for such units might arise as a function of frequency of exposure to onset-rime orthographic patterns in instructional materials and word games. The results of their experiments, however, suggest that at least low-skilled Dutch readers (Grade 1) benefit from viewing words segmented into onset-rime units (see Geudens & Sandra, this volume, Experiment 1, Figure 1: Mean latencies by fluency). While seemingly contradictory to a hypothesis that asserts the existence of functional subsyllabic units in reading, the Geudens and Sandra findings may prove consistent with our Linguistic Hypothesis. In McBride-Chang et al. (1997), the effects of speech perception on phonological awareness and word reading seem to occur relatively early in the acquisition of word recognition skill. As a child gains fluency in their decoding skill, the impact of spoken language on phonological skill is minimized as the influence of the writing system acts to finely tune phonological awareness. The fact that the low skilled Dutch readers did show evidence of onset-rime orthographic structures but high skilled readers did not, suggests a developmental difference in the salience of these units. Similarly, the Yi and Lee findings seem to lend further evidence to this developmental shift in phonological skill in Korean children. Furthermore, the functionality of these subsyllabic units as fluency is achieved may be relegated to processes of production and manipulation in reading rather than identification/recognition processes. For instance, Booth and Perfetti (in press) found no evidence for the use of the onsetrime structure in early word identification processes after Grade 2, but did so in a masked-prime word-naming task.
Summary So, what have we learned? Aspects of the spoken language have a strong effect on the shape and development of phonological awareness. As spoken languages, English and Korean have highly contrasting features. As opposed to English, Korean has a much simpler syllable structure and vowel harmony. Korean has definite syllable boundaries and lacks stress in multisyllabic words. English has complex syllables with consonant clusters at syllable onsets and codas. English is a stress language with unclear syllable boundaries. Importantly, English has relatively more flexibility between onsets and vowels than between vowels and codas (Chomsky & Halle, 1968). Thus, the development of a rime as a phonological unit in English and the body in Korean are likely tied to the linguistic structure of the spoken language. Children gain the ability to manipulate these phonological features (gain
(JB)SWL II_f10_138-163
9/26/02
9:10 AM
Page 161
Subsyllabic units in reading
‘awareness’) prior to encountering instruction with the writing system. Kindergarten children in both English and Korean show an appropriate knowledge of subsyllabic units. How do these units persist into orthographic representations? While we assume that phonological subsyllabic units are functional prior to reading skill, we have not ruled out a possible interaction with the writing system. The evidence from English, Dutch, and Korean suggests that as children gain fluency in reading performance, reliance on onset-rime units may diminish. As children become more skilled at reading they appear to rely less on these units. Somewhat contradictory to this is the orthographic rime frequency effect in English, where subsyllabic units play a strong role in word recognition skill after fluency is achieved (see Bowey, Vaughan & Hansen, 1998). However, the effects of rime frequency after fluency is achieved may be task dependent (see Booth & Perfetti, in press). The orthographic depth hypothesis fails to capture why subsyllabic units appear in Korean reading performance (a transparent orthography) or in perception of speech. Similarly, this hypothesis fails to account for onset-rime effects in low-skilled Dutch readers (Geudens & Sandra, this volume) and German normals and dyslexics (Landerl & Wimmer, 2000), both languages having transparent orthographies as well. Mere exposure-based accounts of the development of orthographic recognition units do not account for Korean children’s performance in the English writing system or English adults performance using Hangul as an artificial orthography. Also, this hypothesis fails to explain Geudens and Sandra (this volume) onset-rime persistence in the unbiased curriculum. In summary, the appearance of subsyllabic units in reading according to the data in Korean and English is most likely due to the early mapping of orthography onto salient phonological units that exist in the spoken language system. Such a conclusion not only accounts for onset-rime units, but similarly may account for morpho-syntactic features as well (e.g., the relative ease of recognizing and manipulating “d/ed” and “s/es” as final phonemes/graphemes). We conclude that our Linguistic Hypothesis identifies orthographic units such as onsets and rimes, or bodies and codas, as properties of a mapping system of salient visual features to phonological features that reflect the deep linguistic structures of the spoken language.
Notes . For example, consonants and vowels are systematically different in their visual appearance and consonant letters often include visual information relating to place of articulation. . It is important to keep in mind that subsyllabic structures can co-exist with other units. To raise the possibility that two languages contrast in sub-syllabic structures is not to suggest that speech perception uses different mechanisms or even different units in the two languages. Phonemes, for example, can function as perceptual units, even if syllables or unites intermediate to phonemes and syllables are also functional in some sense.
(JB)SWL II_f10_138-163
9/26/02
9:10 AM
Page 162
Precursors of Functional Literacy
References Booth, J. R. & Perfetti, C. A. (in press). Onset and rime structure influences naming but not early word identification in children and adults. Journal of Memory and Language. Bowey, J. A. (1996). Phonological recoding of nonword orthographic rime primes. Journal of Experimental Psychology: Learning, Memory, & Cognition, 58, 134–159. Bowey, J. A. & Hansen, J. (1994). The development of orthographic rimes as units of word recognition. Journal of Experimental Child Psychology, 58, 465–488. Bowey, J. A. & Underwood, N. (1996). Further evidence of orthographic rime usage in nonword reading increases with word-level proficiency. Journal of Experimental Child Psychology, 63, 526–562. Bowey, J. A., Vaughan, L. & Hansen, J. (1998). Beginning reader’s use of orthographic analogies in word reading. Journal of Experimental Child Psychology, 68, 108–133. Bradley, L. & Bryant, P. E. (1983). Categorizing sounds and learning to read — a causal connection. Nature, 301, 419–421. Chomsky, N. & Halle, M. (1968). The Sound Pattern of English. New York: Harper & Row. Cossu, G., Shankweiller, D., Liberman, I.Y. & Gugliotta, M. (1988). Awareness of phonological segments and reading ability in italian children. Applies psycholinguistics, 9, 1–16. Fowler, C. A. (1987). Consonant vowel cohesiveness in speech production as revealed by initial and final consonant exchanges. Speech Communication, 6, 231–244. Fowler, C. A., Treiman, R. & Gross, J. (1993). The structure of English syllable and polysyllables. Journal of Memory and Language, 32, 115–140. Fudge, E. C. (1969). Syllables. Journal of Linguistics, 5, 253–286. Fudge, E. C. (1987). Branching structure within the syllable. Journal of Linguistics, 23, 359–377. Goswami, U. (1993). Toward an interactive analogy model of reading development: Decoding vowel graphemes in beginning reading. Journal of Experimental Child Psychology, 56, 2443–2475. Goswami, U. & Bryant, P. (1990). Phonological skills and learning to read. Hillsdale, NJ: Erlbaum. Geudens, A. & Sandra, D. (1999). Onsets and rimes in a phonologically transparent orthography: Differences between good and poor beginning readers of Dutch. Brain & Language, 68, 284–290. Kim, C. K. (1981). Theory of syllables and phonological rules in Korean. The Journal of Humanities, 8, 1–45. Kessler, B. & Treiman, R. (1997). Syllable structure and the distribution of phonemes in English Syllables. Journal of Memory and Cognition. 37, 295–311. Kwon, I. H. (1987). On the psychological reality in Korean Phonology. M.A. Thesis submitted to Seoul National University. Landerl, K. & Wimmer, H. (2000). Deficits in phoneme segmentation are not the core problem of dyslexia: Evidence from German and English children. Applied Psycholinguistics, 21, 243–262. Lee, H. & Yi, K. (1999). The structure of Korean syllables: Evidence from Korean children. Poster presented at the 107th Convention of the American Psychological Association, Boston, Massachusettes, August 21, 1999. Levelt, C. C., Schiller, N. O. & Levelt, W. J. M. (1999). A developmental grammar for syllable structure in the production of child language. Brain & Language, 68, 291–299. McBride-Chang, C., Wagner, R. K. & Chang, L. (1997). Growth modeling of phonological awareness. Journal of Educational Psychology, 89, 621–630. McBride-Chang, C. (1995). What is phonological awareness? Journal of Educational Psychology, 87, 179–192.
(JB)SWL II_f10_138-163
9/26/02
9:10 AM
Page 163
Subsyllabic units in reading McCandliss, B., Schneider, W. & Smith, T. (1997). Learning to new visual symbols as integrated wholes or component parts. Paper Presented at the 38th Annual Meeting of the Psychonomic Society, Philadelphia, PA. Treiman, R. (1983). The structure of spoken syllables: Evidence from novel word games. Cognition, 15, 49–74. Treiman, R. (1986). The division between onsets and rimes in English syllables. Journal of Memory and Language, 25, 476–491. Treiman, R. (1992). The role of intrasyllabic units in learning to read and spell. In P. B. Gough, L. C. Ehri & R. Treiman (Eds.), Reading acquisition (153–198). Hillsdale, NJ: Erlbaum. Treiman, R. & Chafetz, J. (1987). Are there onset- and rime-like units in printed words? In M. Coltheart (Ed.), Attention and Performance (Vol. 12, 281–298). Hillsdale, NJ: Erlbaum. Treiman, R., Goswami, U. & Bruck, M. (1990). Not all nonwords are alike: Implications for reading development and theory. Memory & Cognition, 18, 559–567. Treiman, R., Mullenix, J., Bijeljac-Babic, R. & 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, 107–136. Wimmer, H. & Goswami, U. (1994) The influence of orthographic consistency on reading development: Word recognition in English and German children. Cognition, 51, 91–103. Yi, K. (1995). The Internal Structure of Kulja and Its Relation to Syllable in Korean. Korean Journal of Experimental and Cognitive Psychology, 7, 57–69. Yi, K. (1998). The internal structure of Korean syllables: Rhyme or body? Korean Journal of Experimental and Cognitive Psychology, 10, 67–83. Yi, K. (1999). The internal structure of Korean syllables. To appear in The proceedings of the 2nd ICCS held at Waseda University, Tokyo, Japan. Yoon, H. K. (1997). A study on the Hangul reading development: acquisition of graphemephoneme correspondence rule. Unpublished doctoral dissertation. Pusan National University, Pusan, Korea. Yoon, Y. B. (1995). Experimental studies of the syllable and the segment in Korean. Doctoral dissertation, University of Alberta. Yoon, Y. B. & Derwing, B. L. (1994). Sound similarity judgments for Korean words by Korean and English speakers. Proceedings of the 1994 Annual Conference of the Canadian Linguistic Association, 657–666. Address University of Pittsburgh 651 Learning Research & Development Center 3939 O’Hara Street Pittsburgh, PA 15260 USA
(JB)SWL II_f11_164-187
9/26/02
9:14 AM
Page 164
(JB)SWL II_f11_164-187
9/26/02
9:14 AM
Page 165
Orthography, phonology, and semantics Concerted action in word perception* Anna M. T. Bosman and Janet G. van Hell University of Nijmegen
A couple of decades of scientific research on printed word perception led to the emergence of a coherent picture with respect to the fundamental aspects that govern reading. Most researchers agree with the assumption that orthography (the spelling of a word), phonology (the sound of a word), and semantics (the/a meaning(s) of a word) all play a role in processing printed materials. In this paper, we focus on the concerted action of orthography, phonology, and semantics, and propose that they should be understood as emergent properties in a highly interactive, interdependent, dynamical system We will first present a short overview of the evidence that each of these three aspects is actually contributing to the processes involved in printed word perception and then report three empirical studies. In the General Discussion, we interpret the findings in terms of the phonologic coherence model of Van Orden and his colleagues (e.g., 1990, 1994, and 1997).
Orthography The role of orthography in reading is almost self-evident. The clearest example of how orthography exerts an essential influence on reading is the fact that experienced readers can distinguish between homophones. A homophone is a word with identical phonology to and dissimilar spelling and meaning from another word. For example, THERE and THEIR, DEAR and DEER, or WEAK and WEEK are all homophones of each other. Only orthographic information represented in these words provides unequivocal evidence of their meaning. Another example of the role of orthography in reading is that experienced readers (Bosman, unpublished material, see Experiment 1) unlike novice readers (Bosman & de Groot, 1991; Reitsma, 1983), process words faster than pseudohomophones. A pseudohomophone is a letter string that is phonologically identical
(JB)SWL II_f11_164-187
9/26/02
9:14 AM
Page 166
Precursors of Functional Literacy
to, but orthographically different from a word (e.g., BERD is a pseudohomophone of the word BIRD). Coltheart and Freeman (1974) also demonstrate the role of orthography in reading. They found that words presented iN AlTeRNatIng cASe were more difficult to read than regularly written words (see also Smith, Lott, & Cronnell, 1969). A final example of how orthography affects printed word perception is that words with many neighbours are processed more quickly and accurately than words with few neighbours (e.g., Andrews, 1989, 1992; Grainger, Spinelli, & Ferrand, 2000; Peereman & Content, 1995). Neighbourhood size is defined as the number of words that share all but one letter with the target word. For example, BARK, DARK, LARK, MARK, SARK, PERK, PORK, PACK, PARA, PARD, PARE, PARR, PART, are all neighbours of the word PARK (see for a critical remark on the partial confound between orthographic and phonographic neighbourhood, Peereman & Content, 1997).
Phonology That phonology plays a role in (beginning) reading was never doubted, but whether and when it is an obligatory aspect of printed word perception is still a big issue among psycholinguists (see for discussions, Berent & Perfetti, 1995; Frost, 1998; Humphreys & Evett, 1985; Van Orden, Pennington, & Stone, 1990). We could fill this entire volume with empirical evidence of the role of phonology in reading, but restrict ourselves to a couple of examples. Convincing evidence is provided in experiments in which pseudohomophone processing is compared to pseudoword processing. A pseudoword is an orthographically legal letter string with phonology dissimilar from words For example, BELF, JARL, and SKONK are English pseudowords. Naming latencies of pseudohomophones like ‘BERD’ are often faster than naming latencies of pseudowords like ‘BELF’ (e.g., Besner & Davelaar, 1983; McCann & Besner, 1987; Seidenberg, Petersen, MacDonald, & Plaut, 1996; Taft & Russell, 1992). Lexical-decision experiments show that rejection latencies of pseudohomophones are often longer and more error prone than rejection latencies of pseudowords (e.g., Bosman & de Groot, 1996; Seidenberg et al., 1996; Stone & Van Orden, 1993; Van Orden et al., 1992). It is believed that the familiar phonology of pseudohomophones interferes with its unfamiliar orthography (e.g., JALE), which in turn, causes a delay in the decision to reject the letter string. Such interference does not occur in pseudowords, because both orthography and phonology are unfamiliar (e.g., LOIP). A recent study suggests that such effects extend to second language processing (Bosman, van Hell, & Van Orden, 1998). In a proofreading task in English, in which Dutch-English bilinguals were asked to detect errors in an ongoing text, we observed that pseudohomophones were harder to detect than pseudowords. A most elegant demonstration of the effect of phonology is a study by Stone and Van Orden (1993). They conducted experiments that involved the so-called
(JB)SWL II_f11_164-187
9/26/02
9:14 AM
Page 167
Word perception
ideal strategy manipulation, in which the same target words were embedded in a pseudohomophone context or in a pseudoword context. It appeared that correct “yes”-response latencies to words were slowed down in their lexical-decision task that included pseudohomophones as compared to one that included pseudowords. Thus, processing-time difference between these two conditions can only be attributed to the differential contexts (see also Gibbs & Van Orden, 1998; James, 1975; Vanhoy & Van Orden, 2001 and Pugh, Rexer, & Katz, 1994 for a failure to find a pseudohomophone effect). Homophone effects are also a signature of phonology. Many readers, both beginning and experienced readers, often mistakenly judge ROWS to be a member of the category FLOWERS in lexical-decision and semantic-categorisation tasks, despite their near-perfect spelling knowledge (e.g., Coltheart, Laxon, Rickard, & Elton, 1988; Jared & Seidenberg, 1991; Starr & Fleming, 2001; Van Orden, 1987; 1991). An extensive overview on this subject is presented in Van Orden, Holden, Podgornik, and Aitchison (1999). A final example of phonology at work outside the realm of pseudohomophones is provided by results obtained in the first-letter-naming task. In one version of this task, participants are presented with words or word-like letter strings and are requested to name the first letter of each stimulus as quickly and as accurately as possible. If the onset of a stimulus is identical to the letter name (e.g., the O in OVER), both beginning and experienced readers name the first letter of these stimuli more quickly than in a condition in which they are not identical (e.g., the O in OTTER). If, however, both beginning and experienced readers are asked to use phonemes to identify the first letter in these conditions, the O in OTTER is named faster than the O in OVER (Bosman & de Groot, 1995; Bosman, van Leerdam, & De Gelder, 2000). Thus, in a condition in which the pronunciation of the first letter is phonologically congruent with the onset pronunciation of the stimulus readers are faster than in a condition in which it is incongruent. We believe that this indicates that the phonology of the stimulus is activated while performing first-letter naming, even though it sometimes hinders performance.
Semantics The third and final factor involved in printed word perception is a word’s meaning. Many empirical examples convincingly show the influence of semantic factors on visual word identification. The most widely studied effect is that of semanticpriming, ignited by the seminal paper of Meyer and Schvaneveldt (1971). They showed that participants decided more quickly that two related stimuli (e.g., BREAD and BUTTER) were English words than two unrelated stimuli (e.g., BREAD and DOCTOR). Two types of word relatedness are distinguished. Semantic relatedness refers to a word pair from the same semantic category (e.g., DOCTOR and NURSE) and associative relatedness refers to word pairs that stem from different semantic
(JB)SWL II_f11_164-187
9/26/02
9:14 AM
Page 168
Precursors of Functional Literacy
categories (e.g., WINTER and COLD). Reliable priming effects for both types of relatedness exist (for an extensive overview, see Neely, 1991). Another factor showing the involvement of semantic effects is imageability. Naming latencies of low-frequency low-imageability words (like DOSE or SUAVE) are longer than naming latencies of low-frequency high-imageability words like COMB or PEAR (Cortese, Simpson, & Woolsey, 1997; de Groot, 1989; Strain, Patterson, & Seidenberg, 1995). Likewise, lexical-decision performance on lowimageability words is slower than on high-imageability words, both in the native language and in the second language for more details on the relationship between imageability, concreteness, and context availability (van Hell & de Groot, 1998). Two final examples of semantic variables affecting visual word perception are semantic ambiguity and semantic synonymy. Words that have several related meanings like SAFE lead to faster lexical decision times than words with unrelated meanings like FAST (Azuma & Van Orden, 1997). Words without clear semantic synonyms like BRIDGE are processed faster in lexical-decision and naming tasks than words with clear semantic synonyms like JAIL, which has PRISON as alternative (Pecher, 2001).
Overview of the present study Three experiments are presented. The goal of all three is to show what experienced as well as beginning readers do when presented with pseudohomophones. In the first experiment, experienced readers learn to read alternative spellings (pseudohomophones) of low-frequency Dutch words. These alternative spellings differ in one major respect: Half of the spellings deviate slightly from the baseword (have small orthographic changes) and the other half deviate a lot from the baseword (have large orthographic changes). If orthography plays a role, pseudohomophones with a small orthographic change should be learned more quickly than pseudohomophones with a large orthographic change. Furthermore, if phonology and semantics are also involved, pseudohomophones should be read as quickly as their basewords after sufficient presentations. In the second experiment, both beginning and experienced readers perform a sounds-like-a-word lexical decision task after having read basewords in various frequency conditions. If experimental presentation frequency of basewords affects decisions on pseudohomophones, then phonology and/or semantics are at work in printed word perception (see General Discussion for details). If prior presentation frequency of basewords does not affect decisions on pseudowords then orthography has a negligible effect in this task. In the third experiment, experienced readers perform a sounds-like-a-word lexical decision on pseudohomophones derived from high-frequency basewords and low-frequency basewords. If baseword frequency affects the decision on pseudo-
(JB)SWL II_f11_164-187
9/26/02
9:14 AM
Page 169
Word perception
homophones, it suggests that phonology and/or semantics is a crucial factor in printed word perception. Furthermore, if orthographic appearance of pseudohomophones also affects the decision latencies, another demonstration of the role of orthography in reading becomes apparent.
Experiment 1 In this experiment, we investigate the speed with which experienced readers establish new orthographic knowledge and whether the acquisition rate depends upon orthographic complexity.
Method Participants A group of 25 Dutch PhD-students of the Department of Psychology of the University of Amsterdam volunteered to participate in the experiment. All participants were native speakers of Dutch. Materials Thirty low-frequency Dutch words served as basewords. Each of the basewords was changed into a pseudohomophone. Two sets of pseudohomophones with equal average length (8.5 letters) were created. One set consisted of stimuli with a large orthographic change (mean number of grapheme changes was 3.2, SD = .68), whereas the other set consisted of stimuli with a small orthographic change (mean number of grapheme changes was 2.4, SD = .51). This difference in orthographic change was significant, F(1, 28) = 13.44, p < .001. The stimuli are listed in Appendix A. Five different training lists and one test list were created. In each training list, six pseudohomophones were not used at all (Frequency 0) six pseudohomophones appeared once (Frequency 1), six pseudohomophones appeared twice (Frequency 2), six pseudohomophones appeared three times (Frequency 3), and six pseudohomophones appeared six times (Frequency 6). Each pseudohomophone appeared in each condition and within each frequency condition, three pseudowords contained a large orthographic change and the remaining three had a small orthographic change. Each training list contained 72 pseudohomophones. The test list contained all 30 pseudohomophones and their corresponding basewords. Procedure The experiment was run on a Macintosh computer. The training stimuli were presented on the screen one by one. The participants were told that wrongly spelled words would be presented on the screen, and that they were to read these words as fast and as accurately as possible. Each participant was presented with one of
(JB)SWL II_f11_164-187
9/26/02
9:14 AM
Page 170
Precursors of Functional Literacy
the training lists. Presentation of the test materials followed the training list immediately. The experimental training stimuli were preceded by 8 pseudohomophone practice trials to familiarise the participants with the task. Naming times were registered with a voice key and a millisecond timer. Each response was evaluated on correctness by the experimenter by pressing a key on the computer keyboard.
Results and discussion The results of the training stage and the test stage will be discussed separately. Before the data were subjected to statistical analyses, responses based on naming errors (2.9%), errors due to apparatus failure (2.8%), extremely slow (larger than 4000 ms: .1%) and extremely fast responses (less than 250 ms: 0) were removed from the data set.
Training stage A 2 (orthographic change: large vs. small) × 6 (frequency: 1 vs. 2 vs. 3 vs. 4 vs. 5 vs. 6) ANOVA revealed significant main and interaction effects. The interaction effect between orthographic change and frequency was F(5, 120) = 5.18, p < .001. The difference in naming times between pseudohomophones with a small orthographic change and pseudohomophones with a large orthographic change was significant in Frequency conditions 1 and 2, F(1, 24) = 11.78, p < .001, and F(1, 24) = 4.99, p < .05, respectively. The differences in the other frequency conditions were not significant. The main effect of frequency showed up in both the large and small orthographic change conditions, F(5, 120) = 47.60, p < .001. The results are depicted in Figure 1. Test stage A 2 (stimulus: pseudohomophone vs. baseword) × 5 (frequency: 0 vs. 1 vs. 2 vs. 3 vs. 6) ANOVA on subjects’ mean naming times yielded significant main and interaction effects. The interaction effect, F(4, 96) = 16.84, p < .001, indicated a significant effect of frequency in the pseudohomophone condition, F(4, 96) = 21.72, p < .001), but not in the baseword condition (F < 1). The effect of stimulus was significant in all frequency conditions (main effect: F(4, 96) = 50.09, p < .001), except in Frequency 6 (F < 1). Thus, in all frequency conditions the participants read the basewords faster than the pseudohomophones, but after 6 presentations they read the pseudohomophones as fast as the basewords. The results are presented in Figure 2. The results of this experiment are clear-cut. Experienced, fluent readers learn alternative spellings of extant low-frequency words in six presentations. The learning curve, however, depends on the complexity of the alternative spelling. Stimuli that deviate more strongly from the baseword (pseudohomophones with a large orthographic change) are learned more slowly than stimuli that deviate less from the baseword (pseudohomophones with a small orthographic change).
(JB)SWL II_f11_164-187
9/26/02
9:14 AM
Page 171
Word perception 1000 Small orthographic change Large orthographic change mean naming times in ms
900
800
700
600
500 1
2
4
3
5
6
Training frequency
Figure 1. Mean naming times and standard errors of pseudohomophones in the training stage of Experiment 1 750
pseudohomophones
mean naming times in ms
700 650
600
550
500 base words 450 0
1
2
3
6
Training frequency
Figure 2. Mean naming times and standard errors of pseudohomophones and basewords in all frequency conditions in the test stage of Experiment 1
(JB)SWL II_f11_164-187
9/26/02
9:14 AM
Page 172
Precursors of Functional Literacy
Experiment 2 In this experiment, we investigate the effect of reading familiar words on the decision latencies of derived pseudohomophones and pseudowords in beginning and experienced readers.
Method Participants From a population of 246 Dutch children of Grade 1, we selected a group of 24 advanced readers (8 boys and 16 girls) with a mean age at the time of testing of 88 months. Their reading level was assessed with the Eén-minuut-test, a standardised reading-decoding test (Caesar, 1975; One-minute test). The score on this test is the number of words read correctly in one minute. The mean score of the children who participated in the experiment was 35.8 (SD = 1.9). The mean score of the population was 29.7 (SD = 10.9). Twenty students from the Department of Psychology of the University of Amsterdam also participated in the experiment for course credits. These participants did not take part in Experiment 1. All children, referred to as beginning readers, and psychology students, referred to as experienced readers, were native speakers of Dutch. Materials A set of 20 monosyllabic basewords, used in the reading stage of the experiment, was selected from the list of Kohnstamm, Schaerlaekens, de Vries, Akkerhuis, and Froonincksx (1981). All words are familiar to six-year old children, but are not used as instruction words in the first three books of the reading course. Ten words consisted of three letters and ten words had four letters. All three-letter words had a CVC-structure (C is consonant, V is vowel). Five of the four-letter words followed a CVVC- and five a CCVC-structure. Four different reading lists were constructed. Each of these lists contained five words that appeared only once (Frequency 1), five words appeared twice (Frequency 2), five words appeared six times (Frequency 6), and five words did not appear at all (Frequency 0). Thus, each baseword occurred in each condition, and each list consisted of 45 basewords. The distribution of the basewords within a list was not random, but such that a word’s repetition was evenly distributed over the list, without the order being predictable. A second list of stimuli was constructed to serve in the decision stage of the experiment. The list consisted of 20 pronounceable pseudohomophones and 20 pseudowords. The pseudohomophones were derived from the basewords of the training list by changing one letter or grapheme. The pseudowords were derived from the pseudohomophones by changing one letter or grapheme. The spelling of the pseudowords was orthographically legal, but the phonology was dissimilar from extant words, whereas the spelling of the pseudohomophones was also ortho-
(JB)SWL II_f11_164-187
9/26/02
9:14 AM
Page 173
Word perception
graphically legal, but their phonology was identical to extant words. Appendix B lists the set of basewords, pseudohomophones, and pseudowords.
Procedure The experiment was run on a Macintosh computer. All participants (both beginning and experienced readers) were presented with one of the four reading lists. They were asked to read aloud the words of the list as quickly and as accurately as possible. Naming times were registered with a voice key and a millisecond timer. Each response was evaluated on correctness by the experimenter by pressing a key on the computer keyboard. Before the experimental session started each participant received five practice trials. In the second part of the experiment, participants were confronted with the decision list. They were told that letter strings would appear on the screen, but that none of the words as such exists according to the way they were spelled. The participants were asked to decide as quickly and as accurately as possible whether the word if pronounced was a Dutch word (yes) or not (no) by pressing the proper button. A millisecond timer registered the time it took the participants to make the decision. Results and discussion The results of the beginning and experienced readers will be analysed separately but discussed simultaneously. The results of the beginning readers, the children (CH) will be presented first, followed by those of the experienced readers, the adults (AD).
Reading stage Before the data were subjected to analysis responses based on naming errors (CH: .9%, AD: .3%), errors due to apparatus failure (CH: 7.2%, AD: 1.4%), extremely slow (more than three standard deviations above the mean; CH: 1.8%, AD: .6%) and extremely fast responses (less than 250 ms, CH: 0%, AD: .3%) were removed from the data set. A 6 (exposure: 1, 2, 3, 4, 5, 6) ANOVA was performed on the data of the beginning and the adult readers separately. The main effect of exposure was significant for both the beginning and the experienced readers, CH: F(5, 115) = 6.51, p < .001; AD: F(5, 95) = 2.32, p < .05. A post-hoc analysis (NewmanKeuls, p < .01) on the data of the beginning readers revealed that the mean of 1st Exposure (813 ms) was significantly lower than those of 3rd Exposure (717 ms), 4th Exposure (711 ms), 5th Exposure (707 ms), and 6th Exposure (735 ms). The 1st Exposure (813 ms) did not differ significantly from the 2nd Exposure (769 ms). Further exploration of the data of the experienced readers showed a significant difference between Frequency 1 and 6, F(1, 19) = 4.73, p < .05. The mean naming times of Frequencies 1 to 6 were 447, 447, 445, 437, 441, and 435 ms respectively. Thus, both beginning and experienced readers showed a frequency effect.
(JB)SWL II_f11_164-187
9/26/02
9:14 AM
Page 174
Precursors of Functional Literacy Table 1. Mean decision latencies of the beginning and experienced readers in the decision stage of Experiment 2 Pseudohomophones1 Baseword frequency 0 1 2 6 Total 1
Pseudowords2
Beginning
Experienced
Beginning
Experienced
2478 (112) 2323 (126) 2388 (166) 2186 (109) 2343 (106)
739 (28) 675 (23) 671 (19) 681 (23) 691 (20)
3007 (213) 3112 (203) 2971 (182) 3031 (181) 3030 (176)
975 (60) 938 (50) 924 (44) 926 (45) 941 (46)
Yes-response required, 2 No-response required. Standard errors in parentheses.
Decision stage Before the data were analysed, decision errors (CH: 6.4%, AD: 6.0%), extremely slow (more than three standard deviations above the mean; CH: 1.9%, AD: .8%) and extremely fast responses (less than 250 ms, CH: 0%, AD: 0%) were removed from the data set. Both the beginning and the experienced readers were faster responding yes to pseudohomophones than responding no to pseudowords, CH: F(1, 47) = 26.71, p < .001; AD: F(1, 39) = 47.66, p < .001. The number of errors made on the two types of stimuli did not differ significantly from each other in both groups, both F’s < 1. The results are presented in Table 1. An analysis on the decision times for pseudohomophones for the beginning readers with reading frequency as within factor showed no significant effect, CH: F(3, 69) = 2.02, p = .12. A planned contrast between mean decision times of Frequency 0 and Frequency 6 revealed a significant decrease, CH: F(1, 23) = 12.93, p < .001. Pseudohomophones of which the basewords were read 6 times in the preceding reading stage were decided on faster than pseudohomophones of which the basewords were not read at all in the reading stage. No significant effects emerged in the same analyses on the mean correct no-response latencies of the pseudowords, F < 1. The same analysis for the group of experienced readers revealed a significant effect of reading frequency, AD: F(3, 57) = 4.32, p < .01. A post-hoc analysis on the means for the experienced readers showed that decision times of pseudohomophones in Frequency 0 were significantly longer (Newman-Keuls, p < .05) than those of Frequency 1, Frequency 2 and Frequency 6. Pseudohomophones derived of basewords that were not read in the reading stage showed longer decision latencies than pseudohomophones that were read at least once in the reading stage. Again, no significant effects emerged in the same analyses on the mean correct no-response latencies of the pseudowords, F(3, 57) = 1.22, p = .31. The results of the reading stage revealed that both beginning and experienced readers read highly familiar words more quickly after several presentations. The effect of repetition is obvious. The drop in reading time occurred in the begin-
(JB)SWL II_f11_164-187
9/26/02
9:14 AM
Page 175
Word perception
ning readers after the third presentation, whereas the drop in the reading time of the experienced readers occurred after six presentations. Yes-decisions (i.e., to pseudohomophones) were faster than no-decisions (i.e., to pseudowords) in both beginning and experienced readers. This converges with most other research. Yes-responses are usually faster than no-responses. More interestingly is that both beginning and experienced readers decide more quickly that a pseudohomophone sounds like a word when the corresponding basewords has been presented more often in a preceding reading stage. The nodecisions on the pseudowords of both the beginning and the experienced readers were not affected by the frequency of the corresponding baseword.
Experiment 3 In this experiment, we investigate the effect of baseword frequency and orthographic change on processing pseudohomophones and pseudowords in experienced readers.
Method Participants Twenty first-year students of the Department of Psychology of the University of Amsterdam participated in the experiment, fulfilling course requirements. They were native speakers of Dutch and had not participated in Experiments 1 and 2. Materials The set of 80 stimuli consisted of 40 pseudohomophones and 40 pseudowords. The pseudohomophones were constructed by changing one letter or letter cluster of the baseword into one that represents the same phoneme, or by leaving out a silent letter (i.e., h). The changes were, t-d, ei-ij, ou-au, ch-g, h-Ø. Half of the pseudohomophones were derived from high-frequency basewords (equal to or greater than 52 and smaller than or equal to 1446 occurrences per million; Uit den Boogaart, 1975). The other half were derived from low-frequency basewords (equal to or greater than 4 and smaller than or equal to 18 occurrences per million). The difference in mean frequencies between the two groups was significant, F(1,38) = 12.72, p < .001. The pseudohomophones were also matched on length: Half of the high-frequency pseudohomophones and half of the low-frequency pseudohomophones contained four letters, and the remaining halves had five letters. The set of pseudowords was developed according to Dutch rules of orthography. The construction of the pseudowords was such that 20 of them had a high (equal to or greater than 144 and smaller than or equal to 2741) Cumulative Positional Trigram frequency (CPT-frequency) and the other half a low CPTfrequency (equal to or greater than 0 and smaller than or equal to 48). The CPT-
(JB)SWL II_f11_164-187
9/26/02
9:14 AM
Page 176
Precursors of Functional Literacy
frequency (Rolf & van Rijnsoever, 1984) is a standard for the incidence of a particular combination of letters. A high CPT-frequency indicates that the sum of frequencies of trigrams occurs relatively often in Dutch orthography. A low CPT-frequency indicates that its occurrence is less likely. The difference in CPTfrequency between the two groups was significant, F(1,38) = 18.19, p < .001. The pseudowords were also matched on length. The stimuli used in this experiment are presented in Appendix C.
Procedure The experiment was run on a Macintosh computer. Each participant received the list of stimuli in a different random order. They were told that words would appear on the screen, but none of the words as such existed according to the way they were spelled. Half of the words when pronounced according to Dutch graphemephoneme correspondence rules actually sound like a Dutch word but the other half of the words do not. The participants were asked to decide as quickly and as accurately as possible whether the word when pronounced was a Dutch word (yes) or not (no) by pressing the proper button. A millisecond timer registered the time it took the participants to make the decision. Results and discussion Before the data were subjected to analysis, erroneous responses (5.6%), extremely long responses (more than 3 standard deviations above the mean; .8%), and extremely short responses (< 250 ms; 0) were removed from the data set. A 2 (type of stimulus: pseudohomophones vs. pseudowords) × 2 (length: 4 vs. 5) × 2 (frequency: high vs. low) ANOVA on subjects’ mean response latencies showed a significant main effect of type of stimulus, F(1,19) = 42.67, p < .001. Mean yes-response latencies to pseudohomophones (M = 782 ms; SD = 158) were shorter than mean no-response latencies to pseudowords (M = 1091 ms; SD = 359). The same analysis on the mean number of errors showed a significant difference, F(1,19) = 6.37, p < .05. The mean number of errors to pseudohomophones (M = .14, SD = .07) was larger than to pseudowords (M = .09, SD = .06). Thus, it took the participants longer to say no to the pseudowords than yes to pseudohomophones, but they made fewer errors on pseudowords than on pseudohomophones. A significant three-way interaction emerged between type of stimulus, length, and frequency on both the mean response latencies and mean number of errors, Fv1,19) = 5.60, p < .05, and F(1,19) = 4.76, p < .05, respectively. For the sake of clarity, the subsequent analyses of the data on pseudohomophones and pseudowords will be discussed separately.
Pseudoword no-decisions A 2 (length: 4 vs. 5) × 2 (CPT-frequency: high vs. low) ANOVA on subjects’ mean latencies and on subject’s mean number of errors showed no significant main or interaction effects.
(JB)SWL II_f11_164-187
9/26/02
9:14 AM
Page 177
Word perception 1200
mean yes-decision latencies to pseudohomophones
1100
1000
4-letters 5-letters
900
800
700
600 Low
High
Base-word frequency
Figure 3. Mean yes-response latencies and standard errors of pseudohomophones in Experiment 3
Pseudohomophone yes-decisions A 2 (length: 4 vs. 5) × 2 (frequency: high vs. low) ANOVA on subjects’ mean latencies revealed a significant interaction effect, F(1,19) = 6.47, p < .05. The frequency effect appeared to be significant in the 5–letter condition only, F(1,19) = 25.25, p < .001. The length effect was significant in the high-frequency condition only, F(1,19) = 19.55, p < .001. The same analysis on mean number of errors revealed a similar pattern. The interaction effect between length and frequency did not reach significance, F(1,19) = 2.44, p = .14. Further analyses showed that the frequency effect was only significant in the 5-letter condition, F(1,19) = 13.75, p < .001. The effect of length was again only significant in the high-frequency condition, F(1,19) = 1.88, p < .01. The results are depicted in Figure 3. As in Experiment 2, participants decided more quickly that pseudohomophones sound like words than that pseudowords do not sound like words. Unlike the results of Experiment 2, participants made more errors on pseudohomophones (incorrectly deciding that they do not sound like words) than on pseudowords (incorrectly deciding that they sound like words). Pseudoword-decision times were not affected by length or by CPT-frequency. Stated differently, the no-decisions do not appear to be affected by otherwise relevant orthographic characteristics. Pseudohomophone-decision times, on the other hand, are affected by orthographic properties of word (i.e., length and baseword
(JB)SWL II_f11_164-187
9/26/02
9:14 AM
Page 178
Precursors of Functional Literacy
frequency). Pseudohomophones derived from high-frequency basewords are accepted more quickly than pseudohomophones derived from low-frequency basewords. Words that are more familiar (semantically, orthographically, and/or phonologically?) facilitate the processing of pseudohomophones. Thus, it seems that in this experimental set-up semantics (and phonology) overrules orthography. However, the fact that length affected the decision times of the pseudohomophones suggests that orthography also plays a role. Participants decided more quickly that long pseudohomophones (i.e., five letters) sounded like words than short pseudohomophones (i.e., four letters). This is an interesting result, because, generally, shorter stimuli are processed more quickly. The reason that pseudohomophones with five letters are processed more quickly is probably because their appearance deviated only 20% from the baseword, whereas the pseudohomophones with four letters changed 25% of their appearance.1
General discussion Changing orthographic appearance and maintaining the phonology of words elicits a rather quick and flexible adjustment process in the behaviour of experienced readers. In the first experiment, experienced readers learned to read pseudohomophone spellings of extant words in six presentations. The flexibility of the adjustment process was demonstrated by the finding that pseudohomophones that deviated orthographically more from the original basewords were a little harder to learn to read than pseudohomophones that deviated less. Orthography also played a role: Pseudohomophones more similar to the baseword were accepted more quickly than pseudohomophones less similar to the baseword. Baseword frequency, irrespective of whether this was induced experimentally (via repeated exposure) or derived from natural occurrence (high frequency of occurrence in language), affected pseudohomophone processing, but did not affect pseudoword processing. Beginning and experienced readers responded more quickly to pseudohomophones derived from high-frequency basewords than to pseudohomophones derived from low-frequency basewords. This result converges with other findings of English-speaking and French-speaking participants (Borowsky & Masson, 1999; Grainger et al., 2000; Marmurek & Kwantes, 1996; Taft & Russell, 1992; Van Orden, 1991; Van Orden et al., 1992; see for null-findings Herdman, LeFevre, & Greenham, 1996; McCann & Besner, 1987). An interesting and hotly debated issue is the explanation of the (baseword)frequency effect (see Monsell, 1991 for an excellent discussion of this topic). There is good reason to believe that frequency effects are at least partly reflecting meaningrelated aspects. Standard frequency effects are often much stronger in lexical decision than in naming and Monsell (1991) reports the results of a semanticcategorisation experiment in which he finds robust and strong frequency effects. Both lexical decision and semantic categorisation involve semantics. Frequency effects may also be attributed to the phonologic aspect (and maybe
(JB)SWL II_f11_164-187
9/26/02
9:14 AM
Page 179
Word perception
even to the orthographic aspect) of words. Although we are not aware of a study in which this was investigated, repeated readings of pseudowords must in the end lead to faster reading times (i.e., cause a frequency effect) despite the absence of semantics. Whether we would have to attribute this effect to the orthographic or phonologic aspect is difficult to ascertain and can only be addressed within the theoretical standpoint one has taken. Theorists, however, do not need to put themselves in this awkward position if they are willing to give up the search for causal components in human cognitive architecture. An alternative account, in which effects are interpreted as the results of interactive dynamics between three aspects that seem to govern reading, circumvents the problem of relating an effect to a particular cognitive process, representation, or strategy (see Van Orden & Paap, 1997; Van Orden et al., 1999). An attractive alternative, but by no means the only solution, is the phonologic coherence model, a recurrent-network account. The reason for using Van Orden’s account is twofold. First, his proposal not only assumes that orthography, phonology, as well as semantics contribute to the processes involved in printed word perception, but also explicitly describes (and models, Farrar & Van Orden, 2001) how these aspects interact. That is, the phonologic coherence model is an account in which input and output form an irreducible interdependent whole. Uni-directional causality is given up for bi-directional causality. This implication of the account solves the problem encountered when applying traditional causality theory, in which we are forced to attribute an empirical phenomenon to one particular aspect or process. Second, the account is a metaphor. Generally, theorists consider nodes in a network to be mental, symbolic representations. This notion is abandoned in Van Orden’s network model. Nodes are not psychologically real units of cognition. Their explanatory power resides in the dynamics of the model and should not be attributed causal properties outside of these dynamics. In fact, nodes serve a narrative function only; they are pragmatic notations for modelling or illustration purposes. The remainder of this chapter is devoted to a severely abridged description of the model. More detailed information on the account is presented in several publications by Van Orden and others (1990, 1994, and 1997, in press). Figure 4 summarises the description presented below. Three families of fully interdependent nodes (i.e., orthographic nodes, phonologic nodes, and semantic nodes) are sufficient to describe reading and spelling performance in a large variety of laboratory reading tasks, for example the notorious frequency by regularity or consistency interaction (Seidenberg, Waters, Barnes & Tanenhaus, 1984) and the counterintuitive feedback consistency effect (Stone et al., 1997), and it explains natural reading and spelling phenomena, for example, why spelling is more difficult than reading (Bosman & Van Orden, 1997). An essential aspect of the model is that all node families are connected recurrently or bi-directionally (Note the relation with the assumption of bi-directional causality). That is, there is a connection from each of the orthographic nodes to
(JB)SWL II_f11_164-187
9/26/02
9:14 AM
Page 180
Precursors of Functional Literacy
Phonologic nodes
Semantic nodes
Orthographic nodes Figure 4. Macro dynamics in the phonologic coherence model
each of the phonologic and each of the semantic nodes, but there are also backward connections from each of the phonologic and semantic nodes to the orthographic nodes, and similarly so for the connections between phonologic and semantic nodes. This is represented in Figure 4 by the double arrows between all three node families. Upon presentation of a printed word the orthographic nodes get activated, which in turn activate phonologic and semantic nodes (feedforward activation). The recurrent connections cause the phonologic and semantic nodes to activate the orthographic nodes again (feedback activation). Whenever the feedback activation pattern matches the feedforward activation pattern, a temporarily stable, coherent dynamic whole emerges. It is important to note that the overall strength between node families differ, as illustrated by the relative boldness of the arrows. Connection strength indicates the speed with which dynamics cohere. In the present case, the connections between orthographic and phonologic nodes are strongest. In alphabetic languages, letters and phonemes correlate rather strongly. The letter P is almost always pronounced as /p/ (exceptions are the P in PHOENIX or PSYCHO), and the phoneme /p/ is almost always written with a P. The relations between phonemes and semantics are less strong. Knowing that a word starts with the phoneme P does not provide us with a lot of information on its meaning (e.g., PAGE, PURE, or PRACTICAL do not share a lot of meaning). Although phonologic and semantic nodes share weak connections, they are stronger than those between semantic and orthographic nodes. This is primarily because we learned to speak before we learn to read, and we speak more often than we read. Thus, this recurrent network predicts that dynamics involving the relation between orthography and phonology cohere before the dynamics between phonology and semantics, which, in turn, cohere before the dynamics between seman-
(JB)SWL II_f11_164-187
9/26/02
9:14 AM
Page 181
Word perception
tics and orthography. If this is true, it should not surprise us that phonology is an early and omnipresent constraint in printed word perception. Numerous findings demonstrate the validity of this assumption of the model. Recurrent feedback or reciprocal constraints on the relation between orthography and phonology are presented by Stone, Vanhoy, and Van Orden (1997) for English and by Ziegler, Montant, and Jacobs (1997) for French (see Peereman, Content & Bonin, 1998 for null-findings in French). A recurrent network is, simply put, a system that is designed to detect the statistical relationships present in the stimulus.2 A learned network favours consistent relations above inconsistent ones. Thus, words with consistent spellings (e.g., RAT or SAVE) are generally read more quickly than words with inconsistent spellings (e.g., CHOIR or PINT). The network (and the reader alike) can overcome slow processing of inconsistent words (e.g., HAVE) through frequent presentations. In Experiments 1 and 3, we showed a) an effect of orthographic change on printed word perception and b) an effect of presentation frequency. Pseudohomophones with less common orthographic-phonologic relations were read slightly slower than those with more common relations between orthography and phonology. In the first experiment readers overcame the fairly uncommon (irregular) spellings of the pseudohomophones after only six presentations. The semantic aspect became apparent, at least partly, in the final experiment. Pseudohomophones derived from high-frequency basewords were processed more quickly than pseudohomophones derived from low-frequency basewords. In terms of the model, semantic-phonologic dynamics and probably also semantic-orthographic dynamics cohere more quickly when presented with “high-frequency” pseudohomophones than with “low-frequency” pseudohomophones. A similar explanation holds for the findings with experimentally manipulated baseword frequency in Experiment 2. Thus, the results of the above-presented experiments suggest interdependent roles for orthography, phonology, and semantics in the reading behaviour of beginning and experienced readers. The results can be explained in terms of the phonologic coherence model and complement other (better) empirical examples of the interdependence between orthography, phonology, and semantics presented in, for example, Farrar, Van Orden, and Hamouz (2001), Vanhoy and Van Orden (2001), and in Gottlob, Goldinger, Stone, and Van Orden (1999). This very short description of the phonologic coherence model was meant to demonstrate that effects found in experiments need not be (in fact, cannot be) attributed to a single cause. They are better understood as emergent properties in a highly interactive, interdependent, dynamical system.
(JB)SWL II_f11_164-187
9/26/02
9:14 AM
Page 182
Precursors of Functional Literacy Appendix A. Stimuli used in Experiment 1 Baseword lacherig nauwelijks tijdelijk dagelijks houterig vrouwelijk veiligheid mogelijk schrijven vijandig lichtelijk brouwsels godsvrucht onmiddellijk eigenschap drijfhout schildpad kwaadheid rijstbouw slechtheid damschijf beschijnen schaakbord dwaasheid achtvoudig schrijnend landschap slagveld achterlijf koudvuur
Pseudohomophone
Letters
Changes
laggurich nouwuluks teiduluk dachuluks hauturich vraueluk vijlichhijt moochuluk sgreivun veiandich ligtuluk brausuls gotsvrugd omiduluk ijgunsgap dreifhaud sgiltpat kwaathijt reisdbauw slegthijt damsgeif besgeinen sgaakbort dwaashijt agtvaudig sgreinent lantsgap slachvelt agterleif kautvuur
9 9 8 9 9 8 9 9 8 9 8 8 9 8 8 9 8 8 9 8 8 9 9 8 9 9 8 9 9 8
3 3 3 3 3 3 4 4 3 2 3 3 3 5 3 3 3 3 3 3 2 2 2 2 2 3 2 2 2 2
Orthographic Change large large large large large large large large large large large large large large large small small small small small small small small small small small small small small small
Appendix B. Stimuli used in Experiment 2 Baseword kous vijf rijd god kwijt wijs geit bed fout stad prijs vouw
Pseudohomophone kaus veif reid got kweit weis gijt bet faut stat preis vauw
Pseudoword kaul veim reif gof kweis weil gijm bem faus stas preil vauk
(JB)SWL II_f11_164-187
9/26/02
9:14 AM
Page 183
Word perception Appendix B. (cont.) Baseword
Pseudohomophone
tijd bad pijn lijm glad hoed grijs rijm
teid bat pein leim glat hoet greis reim
Pseudoword tein baf peim leil glap hoel greik reil
Appendix C. Stimuli used in Experiment 3 Baseword goed groot vrouw weg klein woord juist kort vrij slecht thuis zwart recht prijs kracht breed fijn fout paus grijs joods knecht touw spijt grauw vaart pand blijk plicht klauw saus plein zuid jacht
Pseudohomophone goet grood vrauw wech klijn woort juisd kord vrei slegt tuis zward regt preis kragt breet fein faut pous greis joots knegt tauw speit grouw vaard pant bleik pligt klouw sous plijn zuit jagt
Baseword Frequency 1446 1296 585 492 484 357 298 240 239 144 144 143 121 117 97 64 62 58 53 52 18 17 17 15 15 15 13 12 12 11 11 10 10 10
Pseudoword
CPT-frequency
bijns ficht wocht techt zerd vroud vant plijfd keeg krous megt bist peids kest beisd rebt sleig deit glopt treid krugd bloch densd slad slout flijp grech gauf praul zweis drad pach daafd mous
2741 1363 1291 1261 1171 1113 750 550 504 496 466 266 258 226 219 188 163 152 145 144 140 130 120 104 88 79 73 63 55 42 20 18 13 11
(JB)SWL II_f11_164-187
9/26/02
9:14 AM
Page 184
Precursors of Functional Literacy Appendix C. (cont.) Baseword trant dag geit wreed brein wijf
Pseudohomophone trand dach geid wreet brijn weif
Baseword Frequency 9 8 8 8 5 4
Pseudoword flund jots taun teif zauk wreg
CPT-frequency 11 3 0 0 0 0
Notes * We are grateful to all children and teachers of various schools in Purmerend and Almere in the Netherlands for their participation in Experiment 2. We also thank Rineke Keijzer who conducted Experiment 3. Special thanks to “boss” Ludo Verhoeven who provided us with the facilities to pursue our careers. . A post-hoc analysis of the difference in orthographic similarity, a measure developed by Van Orden (1987) and derived from Weber’s measure of graphic similarity (1970), showed that pseudohomophones with four letters were significantly less similar to their basewords (.66) than pseudohomophones with five letters (.72), F(1,38) = 11.22, p < .002. . Feedforward networks, pattern-associators, and simple recurrent networks also pick up the statistical regularities present in the stimulus (see for example, Mcleod, Plunkett & Rolls, 1998).
References Andrews, S. (1989). Frequency and neighborhood size effects on lexical access: Activation or search? Journal of Experimental Psychology: Learning, Memory, and Cognition, 15, 802–814. Andrews, S. (1992). Frequency and neighborhood effects on lexical access: Lexical similarity or orthographic redundancy? Journal of Experimental Psychology: Learning, Memory, and Cognition, 18, 234–254. Azuma, T. & Van Orden, G. C. (1997). Why SAFE is better than FAST: The relatedness of a word’s meanings affects lexical decision times. Journal of Memory and Language, 36, 484–504. Berent, I. & Perfetti, C. A. (1995). A Rose is a REEZ: The two-cycles model of phonology assembly in reading English. Psychological Review, 102, 146–184. Besner, D. & Davelaar, E. (1983). Suedohomofoan effects in visual word recognition: Evidence for phonological processing. Canadian Journal of Psychology, 37, 300–305. Borowsky, R. & Masson, M. E. J. (1999). Frequency effects and lexical access: On the interpretation of null pseudohomophone base-word frequency effects. Journal of Experimental Psychology: Human Perception and Performance, 25, 270–275. Bosman, A. M. T. & de Groot, A. M. B. (1991). De ontwikkeling van woordbeelden bij beginnende lezers en spellers. [The development of orthographic images in beginning readers and spellers]. Pedagogische Studiën, 68, 199–215. Bosman, A. M. T. & de Groot, A. M. B. (1995). Evidence for assembled phonology in beginning and fluent readers as assessed with the first-letter-naming task. Journal of Experimental Child Psychology, 59, 234–259.
(JB)SWL II_f11_164-187
9/26/02
9:14 AM
Page 185
Word perception Bosman, A. M. T. & de Groot, A. M. B. (1996). Phonologic mediation is fundamental to reading: Evidence from beginning readers. Quarterly Journal of Experimental Psychology, 49A, 715–744. Bosman, A. M. T., van Hell, J. G. & Van Orden, G. C. (1998). How do you reed errur and ermor? Bilingual proofreading. Eleventh Congress of the European Society for Cognitive Psychology. Jerusalem, Israel, 13–17 september. Bosman, A. M. T., van Leerdam, M. & De Gelder, B. (2000). The /O/ in OVER is different from the /O/ in OTTER: Phonological effects in Dutch children with and without dyslexia. Developmental Psychology, 36, 817–825. Caesar, F. B. (1975). Een-minuut-test voor de technische leesvaardigheid. [One-minute test for reading decoding]. Tilburg, The Netherlands: Zwijsen. Coltheart, M. & Freeman, R. (1974). Case alternation impairs word recognition. Bulletin of the Psychonomic Society, 3, 102–104. Coltheart, V., Laxon, V., Rickard, M. & Elton, C. (1988). Phonological recoding in reading from meaning by adults and children. Journal of Experimental Psychology: Learning, Memory, and Cognition, 14, 387–397. Cortese, M. J., Simpson, G. B. & Woolsey, S. (1997). Effects of association and imageability on phonological mapping. Psychonomic Bulletin & Review, 4, 226–231. de Groot, A. M. B. (1989). Representational aspects of word imageability and word frequency as assessed through word association. Journal of Experimental Psychology: Learning, Memory, & Cognition, 15, 824–845. Farrar, W. T., & Van Orden, G. C. (2001). Errors as multistable response options. Nonlinear Dynamics, Psychology, and Life Sciences, 5, 223–265. Farrar, W. T. IV, Van Orden, G. C. & Hamouz, V. (2001). When SOFA primes TOUCH: Interdependence of spelling, sound, and meaning in “semantically mediated” phonological priming. Memory & Cognition, 29, 530–539. Frost, R. (1998). Toward a strong phonological theory of visual word recognition: True issues and false trails. Psychological Bulletin, 123, 71–99. Gibbs, P. & Van Orden, G. C. (1998). Pathway selection’s utility for control of word recognition. Journal of Experimental Psychology: Human Perception and Performance, 24, 1162–1187. Gottlob, L. R., Goldinger, S. D., Stone, G. O. & Van Orden, G. O. (1999). Reading homographs: Orthographic, phonologic, and semantic dynamics. Journal of Experimental Psychology: Human Perception and Performance, 25, 561–574. Grainger, J., Spinelli, E. & Ferrand, L. (2000). Effects of baseword frequency and orthographic neighborhood size in pseudohomophone naming. Journal of Memory and Language, 42, 88–102. Herdman, C. M., LeFevre, J. & Greenham, S. L. (1996). Base-word frequency and pseudohomophone naming. Quarterly Journal of Experimental Psychology, 49A, 1044–1061. Humphreys, G. W. & Evett, L. J. (1985). Are there independent lexical and nonlexical routes in word processing? An evaluation of the dual-route theory of reading. The Behavioral and Brain Sciences, 8, 689–740. James, C. T. (1975). The role of semantic information in lexical decision. Journal of Experimental Psychology: Human Perception and Performance, 1, 130–136. Jared, D. & Seidenberg, M. S. (1991). Does word identification in reading proceed from spelling to sound to meaning. Journal of Experimental Psychology: General, 120, 358–394. Kohnstamm, G. A., Schaerlaekens, A. M., De Vries, A. K., Akkerhuis, G. W. & Froonincksx, M. (1981). Nieuwe streeflijst woordenschat voor 6-jarigen. [New target list vocabulary for 6-year olds]. Lisse, The Netherlands: Swets & Zeitlinger. Marmurek, H. H. C. & Kwantes, P. J. (1996). Reading words and wirds: Phonology and lexical access. Quarterly Journal of Experimental Psychology, 49A. 696–714. McCann, R. S. & Besner, D. (1987). Reading pseudohomophones: Implications for models
(JB)SWL II_f11_164-187
9/26/02
9:14 AM
Page 186
Precursors of Functional Literacy of pronunciation assembly and the locus of word-frequency effects in naming. Journal of Experimental Psychology: Human Perception and Performance, 13, 14–24. McLeod, P., Plunkett, K. & Rolls, E. T. (1998). Introduction to connectionist modelling of cognitive processes. Oxford, UK: Oxford University Press. Meyer, D. E. & Schvaneveldt, R. W. (1971). Facilitation in recognizing pairs of words: Evidence of a dependence between retrieval operations. Journal of Experimental Psychology, 90, 227–234. Monsell, S. (1991). The nature and locus of word frequency effects in reading. In D. Besner & G. W. Humphreys (Eds.), Basic processes in reading (148–197). Hillsdale, NJ: Erlbaum. Neely, J. H. (1991). Semantic priming effects in visual word recognition: A selective review of current findings and theories. In D. Besner & G. W. Humphreys (Eds.), Basic processes in reading (264–336). Hillsdale, NJ: Erlbaum. Pecher, D. (2001). Perception is a two-way junction: Feedback semantics in word recognition. Psychonomic Bulletin and Review, 8, 545–551. Peereman, R. & Content, A. (1995). Neighborhood size effect in naming: Lexical activation or sublexical correspondences. Journal of Experimental Psychology: Learning, Memory, and Cognition, 21, 409–421. Peereman, R. & Content, A. (1997). Orthographic and phonological neighborhoods in naming: Not all neighbors are equally influential in orthographic space. Journal of Memory and Language, 37, 382–410. Peereman, R., Content, A. & Bonin, P. (1998). Is perception a two-way street? The case of feedback consistency in visual word recognition. Journal of Memory and Langauge, 39, 151–174. Pugh, K. R., Rexer, K. & Katz, L. (1994). Evidence of flexible coding in visual word recognition. Journal of Experimental Psychology: Human Perception and Performance, 20, 807–825. Reitsma, P. (1983). Word-specific knowledge in beginning reading. Journal of Research in Reading, 6, 41–56. Rolf, P. C. & van Rijnsoever, R. J. (1984). Positionele letterfrekwenties van het Nederlands. [Positional letter frequencies of Dutch]. Lisse, The Netherlands: Swets & Zeitlinger. Seidenberg, M. S., Petersen, A., MacDonald, M. C. & Plaut, D. C. (1996). Pseudohomophone effects and models of word recognition. Journal of Experimental Psychology: Learning, Memory, and Cognition, 22, 48–62. Seidenberg, M.S., Waters, G. S. Barnes, M. A. & Tanenhaus, M. K. (1984). When does irregular spelling or pronunciation influence word recognition? Journal of Verbal Learning and Verbal Behavior, 23, 383–404. Smith, F., Lott D. & Cronnell, B. (1969). The effect of type size and case alternation on word identification. American Journal of Psychology, 82, 248–253. Starr, M. S. & Fleming, K. K. (2001). A rose by any other name is not the same: The role of orthographic knowledge in homophone confusion errors. Journal of Experimental Psychology: Learning, Memory, and Cognition, 27, 744–760. Stone, G. O., Vanhoy, M. & Van Orden, G. C. (1997). Perception is a two-way street: Feedforward and feedback phonology in visual word recognition. Journal of Memory and Language, 36, 337–359. Stone, G. O. & Van Orden, G. C. (1993). Strategic control of processing in word recognition. Journal of Experimental Psychology: Human Perception and Performance, 19, 744–774. Strain, E., Patterson, K. & Seidenberg, M. S. (1995). Semantic effects in single-word naming. Journal of Experimental Psychology: Learning, Memory, and Cognition, 21, 1140–1154. Taft, M. & Russell, B. (1992). Pseudohomophone naming and the word frequency effect. The Quarterly Journal of Experimental Psychology, 45A, 51–71. Uit den Boogaart, P. C. (1975). Woordfrequenties in geschreven en gesproken Nederlands
(JB)SWL II_f11_164-187
9/26/02
9:14 AM
Page 187
Word perception [Word frequencies in written and spoken Dutch]. Utrecht, The Netherlands: Oosthoek, Scheltema & Holkema. van Hell, J. G. & De Groot, A. M. B. (1998). Disentangling context availability and concreteness in lexical decision and word translation. Quarterly Journal of Experimental Psychology, 51A, 41–63. Vanhoy, M. & Van Orden, G. C. (2001). Pseudohomophones and word recognition. Memory & Cognition, 29, 522–529. Van Orden, G. C. (1987). A ROWS is a ROSE: Spelling, sound, and reading. Memory & Cognition, 15, 181–198. Van Orden, G. C. (1991). Phonologic mediation is fundamental to reading. In D. Besner & G. W. Humphreys (Eds.), Basic processes in reading (77–103). Hillsdale, NJ: Erlbaum. Van Orden, G. C., Bosman, A. M. T., Goldinger, S. D. & Farrar IV, W. T. (1997). A recurrent-network account of reading, spelling, and dyslexia. In J. W. Donahoe & V. Packard Dorsel (Eds.), Neural-network models of cognition (522–538). Amsterdam, The Netherlands: Elsevier Science BV. Van Orden, G. C. & Goldinger, S. D. (1994). Interdependence of form and function in cognitive systems explains perception of printed words. Journal of Experimental Psychology: Human Perception and Performance, 20, 1269–1291. Van Orden, G. C., Holden, J. G., Podgornik, M. C. & Aitchison, C. S. (1999). What swimming says about reading: Coordination, context, and homophone errors. Ecological Psychology, 11, 45–79. Van Orden, G. C. & Paap, K. R. (1997). Functional neuroimages fail to discover pieces of mind in the parts of the brain. Philosophy of Science, 64, S85–S94. Van Orden, G. C., Pennington, B. F. & Stone, G. O. (1990). Word identification in reading and the promise of subsymbolic psycholinguistics. Psychological Review, 97, 488–522. Van Orden, G. C., Stone, G. O., Garlington, K. L., Markson, L. R., Pinnt, G. S., Simonfy, C. M. & Brichetto, T. (1992). “Assembled” phonology and reading: A case study in how theoretical perspectives shapes empirical investigation. In R. Frost & L. Katz (Eds.), Orthography, phonology, morphology, and meaning (249–292). Amsterdam, The Netherlands: North-Holland. Weber, R. M. (1970). A linguistic analysis of first-grade reading errors. Reading Research Quarterly, 5, 427–451. Ziegler, J. C., Montant, M. & Jacobs, A. M. (1997). The feedback consistency effect in lexical decision and naming. Journal of Memory and Language, 37, 533–554. Address University of Nijmegen Dept. of Special Education PO Box 9104 6500 HE Nijmegen The Netherlands
(JB)SWL II_f12_188-213
9/26/02
10:33 AM
Page 188
(JB)SWL II_f12_188-213
9/26/02
10:33 AM
Page 189
The lexical quality hypothesis Charles A. Perfetti and Lesley Hart University of Pittsburgh
Reading is partly about words. Or, to begin the argument more forcefully, it is mainly about words. Either way, it follows that knowledge about words is important to understand skill in reading. This claim is more interesting if we take reading skill to be, in its definition, not about word skill but about comprehension skill. So there it is: Our argument is that skill in reading comprehension rests to a considerable extent on knowledge of words. In what follows, we try to explain what we mean by this and why we conclude that it’s true. Is so doing, we re-state arguments and illustrations that we have made in Perfetti & Hart (in press). A precursor of our argument is contained in earlier work (Perfetti, 1985) in the following observation: When one compared children who differed in assessed comprehension skill, they always divided themselves in word reading and decoding in the same way. The less skilled comprehenders were not just not so good in comprehension, they were also not so good in word reading. From such a pervasive correlation — and it is very pervasive — came a hypothesis that seized the causal high ground, with only modest justification: Comprehension skill depends on word reading skill. The bases of this conclusion lay in a program of research that consistently found that when either children or adults were separated by their scores on a reading comprehension test, they sorted themselves also on their speed of written word and pseudo-word identification. The causal link is not established in a clear way even now. But the causal relationship is complex enough to partly excuse this lack of progress. Lexical skills allow comprehension, comprehension allows reading practice, reading practice strengthens lexical skills, etc. The recurrent nature of component interactions in the process assured an intimate connection between lexical and comprehension skills. But the natural (first cause) privilege belonged to the lexical processes. Word reading gets things started, and it was more difficult to imagine a reader who comprehends the texts without reading the words than vice-versa. Verbal Efficiency Theory was the name we gave to this simple idea — which was that many — not all — problems in comprehension are caused by ineffective word identification processes. Rapid and perhaps modular (context-free) word identification was important for a comprehension system of limited capacity.
(JB)SWL II_f12_188-213
9/26/02
10:33 AM
Page 190
Precursors of Functional Literacy
We have revisited this idea recently, adding some ideas about what kinds of word processes are at issue. The original formulation of Verbal Efficiency emphasized the speed and automaticity of word processes, such that by virtue of being automatic they allowed processing resources to be devoted to comprehension. This idea, we think, was correct but incomplete. What is it about words or about readers such that word processing can vary in a way that affects comprehension? One tentative answer to this was a precursor to the Lexical Quality Hypothesis based on linguistic code manipulation: The basic idea was that, in any modality, efficiency is the rapid retrieval, from inactive memory, of codes that are part of a stored linguistic symbol. And “to the extent retrieval is effortful and the retrieved codes low in quality, the processing is inefficient” (Perfetti, 1985: 118). Whether by spoken language or by written language, a low quality code retrieved with effort would jeopardize comprehension processes that depend on a high quality representation. However, the additional linguistic step in reading — mapping an orthographic form onto a lexical representation — could make the problem more noticeable in reading.
Lexical quality The question of code quality, in this account, becomes central. The original suggestion was that the retrieval of a lexical representation is high in quality “to the extent that it contains both semantic and phonetic information sufficient to recover its memory location . . . This quality must be retained long enough for subsequent processes to perform their work. Thus a “name” without meaning and a meaning without a “name” are both low quality.” (Perfetti, 1985: 118). This idea was developed further in a theory of reading acquisition by reference to lexical specificity and redundancy (Perfetti, 1992). A lexical representation has high quality to the extent that it has a fully specified orthographic representation (a spelling) and redundant phonological representations (one from spoken language and one recoverable from orthographic-to-phonological mappings). If a lexical representation is specific and redundant, its retrieval is more likely to be coherent and reliable. By coherent, we mean that the constituents are available synchronously at retrieval, giving the impression of a unitary word. (The contrast is constituent asynchrony, a fragmented appearance of constituents and parts of constituents, as when a speech recoding occurs followed by meaning retrieval. See Perfetti, 1985: 114). The consequence of reliability is that multiple encounters with a given word tend to produce a common core representation consisting of a nexus of orthographic, phonological, and semantic information. To put it approximately, these defining features of high quality allow the reader to get exactly the word that is printed rather than parts of it that may also be parts of other words. Confusion about word meaning as well as word form is minimized by high quality representations. Variation in lexical quality can be observed in spelling, in retrieval of pro-
(JB)SWL II_f12_188-213
9/26/02
10:33 AM
Page 191
Lexical quality
OR {gate}
PH {geyt}
SE {WiiGM}
Figure 1. The word “gate” illustrates the concept of a lexical entry with three well-integrated constituents of orthography (OR), phonology (PH), and semantics (SE)
nunciations, and in identifying meanings. Reliability in performance is part of the assessment of quality. A reliable, coherent, high quality representation is retrieved easily and consistently. Figure 1 represents a high quality representation by indicating a tightly bonded set of word constituents — the orthographic (OR), phonological (PH), and semantic (SE) specifications of the word. The identification of the word is the retrieval of these constituents.1 Referring to the components of words as “constituents” has certain implications. A constituent in a linguistic or algebraic representation is not merely a part of a larger whole; it is a defining symbol or variable. The specification of each constituent is subject to constraints from a relevant system. A word is a triple of PH, OR, and SE, each of which is a constituent whose specification is constrained by systems of phonology, orthography, and syntax-semantics, respectively. Reading research has largely ignored the systematic nature of phonological constraints, but Berent & Shimron (1999) present an interesting example of constraints on phonology that may be relevant for reading. Similarly, SE constituents reflect basic conceptual and grammatical constraints. One consequence of the constituent idea is that it encourages the view that word identity is both unitary and compositional. Thus, the lexical representation of the word “cat” is the (unitary) linguistic object such that it has spelling C-A-T, phonology [kaet] and meaning (whatever it is that cat means). In general terms, all words are triples of {PH, OR, SE} specification. The SE constituent, however, is incomplete in the “cat” example, because it ignores grammatical information. To avoid a fourth constituent, we can stipulate that SE is actually a collection of meaning and grammatical information. Thus, it includes the information of the lemma, the abstract grammatical frame that appears necessary to understand the production of speech (Bock & Levelt, 1994), as well as a conceptual structure that links to the lemma. We can now state the key idea about individual differences: People vary in the quality of their lexical representations. Any representation that does not specify the value of one if its constituents is low quality. Consider the following examples:
(JB)SWL II_f12_188-213
9/26/02
10:33 AM
Page 192
Precursors of Functional Literacy
a. Presented with the word “incarcerate”, the reader can pronounce it accurately and knows it has some negative meaning, but is not sure what that meaning is. b. Presented with “incarcerate”, the reader stumbles on its pronunciation, producing something like in-cark-rate. c. Presented with “incarcerate”, the reader can pronounce it and indicate that it means something like “to confine in prison”. When attempting to speak the word to produce a message about someone going to jail, however, he sometimes produces “incarcerate” and sometimes something more like “incarsate”. d. A reader can perform all the tasks failed variably in a, b, and c above but can spell the word correctly only on some attempts. The examples indicate several unreliable representations of the word {incarcerate, [inkarsreyt], (verb trans; put in jail)}. They will have differential consequences in different tasks. Case (d), for example, is one familiar to many individuals of high literacy — the feeling of semantic and phonological competence coupled with a spelling block. Case (a) may represent a skilled reader of limited experience who can “decode” a word that he doesn’t really know, although he has heard it pronounced and thus has some phonological representation. By illustrating the basic idea with a low frequency word, we intend to suggest a highly general concept that applies to skilled readers as well as less skilled readers. Skilled readers can have low quality representations for many words, both low frequency words from general vocabulary and words from specialized vocabularies. What skilled readers have by way of an advantage are the foundational resources that get the most out of impoverished representations and support learning given a novel exposure. A skilled reader has a better chance of adding new information (about spelling, pronunciation or meaning) to an impoverished representation. Our analysis, however, focuses on words more than individuals. Not individuals, but word representations vary in quality. A skilled reader, in addition to having foundational resources (decoding, spelling, and grammatical skills), is one who has many high quality word representations. A less skilled reader is one who has fewer high quality word representations. The genesis of the number of quality word representations is a matter of experience that produces relevant knowledge. Certainly, basic and explicit phonological knowledge supports early progress in reading, and subsequent practice in reading, listening, speaking, and writing promote more high quality word representations. The profound differences in the amount of reading among individuals are sufficient to produce the degree of lexical quality differences observable at any given age. Such experience affects the quality of a given word and the number of words of a high quality. There is, in effect, a continuum of lexical knowledge and it is this continuum that is bisected when researchers refer to skilled and less skilled readers.
(JB)SWL II_f12_188-213
9/26/02
10:33 AM
Page 193
Lexical quality
Threats to lexical quality One way to observe the consequences of lexical quality is to create conditions that expose variation in quality and to test the effects of this variation on simple comprehension. One threat to lexical quality arises from properties of written English that lead to words sharing forms — orthographic form or phonological form. Words with multiple meanings on single forms risk lexical quality because there is a lack of one-to-one mapping among one or more constituents. This occurs in three distinct ways: (1) A single orthographic form maps to a single phonological form, but two multiple meanings (more than one morpheme). This case, word ambiguity, has been the staple of research addressing the meaning selection process. Example: count (1. enumeration; 2. feudal title). (2) A single orthographic form maps onto two different phonological forms and hence two meanings (two morphemes). Example, bass (1. /baes/, type of fish; 2. /beys/, voice or instrument of the lower pitch.) (3) Two orthographic forms maps onto a single phonological form and two meanings (1. Seed. 2. Cede). These three cases of meaning polysemy, homography, and homophony, respectively, each depart from the one-toone-to-one mappings among orthographic form, phonological form and meaning. In sharing forms, words in each of the three cases risk lexical quality. Homophones, the third case, provide the demonstration case for our experiments. Figure 2 illustrates this case by adding the homophone gait to the representation of gate in Figure 1. The words gait and gate in Figure 2 are represented as high quality, as they might be for a reader high in knowledge about words. Nevertheless, there is potential for confusion. If one encounters the spoken form [geyt] both gait and gate might be activated, along with their different meanings. Context will select the right one and the confusion will be of no real consequence for comprehension. Less obviously, even in reading, confusion can result, despite the fact that spelling uniquely identifies the word. Thus gait can cause its homophonic partner gate to be activated, leading to momentary subconscious activation of both words. So the case for a skilled reader is that the written presentation of homophones can lead to the activation (retrieval) of two words. Although the representations are as good as they can be, the momentary retrieval of these representations can yield confusion, even for skilled readers, as Gernsbacher and Faust (1991) have shown. If the potential for confusion exists for the skilled reader, it is even stronger for the less skilled reader. According to the Lexical Quality Assumption, a low quality representation could be observed for gait, for gate, or for both. Imagine a reader whose skill is low enough that the word gait is completely unknown. No spelling, no meaning. This reader has no confusion when gate is presented. Or, more carefully, the only problem with gate is due to its representation quality independent of gait. It should be no different from a control word, one that is not a homophone. However, a more typical case might be a reader with an unreliable low quality representation for gait. Now there is the increased potential of confusion beyond what one might see from a nonhomophone. This is because
(JB)SWL II_f12_188-213
9/26/02
10:33 AM
Page 194
Precursors of Functional Literacy
OR {gate}
PH {geyt}
SE {WiiGM} PH {geyt}
OR {gait}
SE {WiiGM}
Figure 2. With “gate’s” homophone, “gait,” the lexical entries are interconnected. Through PH, the activation of one lexical entry will partially activate the competing homophone, causing brief interference
the representation for gait, no matter how impoverished, can be partially activated by a spelling, a pronunciation or a meaning — when gate is presented. Word frequency is an important determinant of word processing and this applies as well to homophones. Reading the more frequent member of a pair of homophones may occur without leading to access of the less frequent member of the pair, a finding that has long been observed in research with ordinary semantically ambiguous words (Duffy, Morris & Rayner, 1988; Hogaboam & Perfetti, 1978). According to the Lexical Quality Hypothesis, this frequency result is a consequence of the benefits of experience for the higher frequency meaning, making its meaning and spelling a more stable constituent, compared with the lower frequency meaning. Moreover, if we consider word representation quality to be distributed across individuals the way we have suggested, then high frequency words may be less a source of confusion for skilled readers than for less skilled readers. This is because a word that is high frequency according to a corpus count may have actually been experienced less frequently by a less skilled reader. Similarly, a word that is low frequency for a skilled reader may be functionally even lower for a low skilled reader. Consistent with this assumption are data reported in Perfetti and Roth (1981) showing that whether context facilitates word identification depends on both skill (less skilled readers more likely to be affected than skilled readers) and basic word identification speed (words identified quickly in isolation
(JB)SWL II_f12_188-213
9/26/02
10:33 AM
Page 195
Lexical quality
are less likely to be affected than words identified more slowly in isolation). Thus, it is the functional identifiability of a word (not the reader’s skill in absolute terms) that is critical, and this can vary for the same word across readers and for the same reader across words. These assumptions about lexical representation and functional identifiability provide for some non-obvious predictions about homophone confusions that have been tested in our experiments. First, because a skilled reader has higher quality representations for more words than does a less skilled reader, the mutual activation of homophones should occur more quickly than for the less skilled reader. Because both members of a homophone pair, e.g., gate and gait, are more identifiable for the skilled reader, activation will spread more quickly from the one to the other. This can lead to a homophone confusion more quickly for a skilled reader than for a less skilled reader. However, the same assumption — that the skilled reader has better quality representations of homophones as for all words — leads to the prediction that this confusion will be short lived. The presented word will quickly have more activation than its unpresented homophone mate will. For the less skilled reader, confusion should build more slowly (because of a lower functional identifiability of both homophones) and release more slowly. This description referring to confusion and release from confusion is more general (and more theoretically neutral) than one referring to suppression (Gernsbacher & Faust, 1991). We leave open the possible operation and failure of a suppression mechanism, although it is not required by our analysis of the process of homophone meaning selection, for which the assumptions of the Lexical Quality Hypothesis are sufficient. The effect of word frequency is also implicated by this analysis. A skilled reader, whose knowledge of the high frequency member of the pair (e.g., gate) is of very high quality, should show little interference from a presentation of that high frequency word. For the low frequency member of the pair (e.g., gait), the skilled reader’s representation is of lower quality; so the presentation of gait allows confusion to confusion to spread to the better known gate before comprehension of gait is complete. Compare this with a less skilled reader, who, by assumption, has lower quality representations for both the higher and lower frequency member of a homophone pair. The presentation of a high frequency member of the pair will now allow confusion. Because its functional identifiability is lower than it is for the high skilled reader, this allows activation of gate to spread to the incomplete representation of gait. For this to occur, we need to assume that gait is partially represented not entirely absent. Whether confusion occurs for the lower frequency member of the pair (e.g., gait) depends on the quality of its representation as well. For a reader who has no representation of gait there is no confusion intrinsic to that word when it is presented. Any confusion comes from activation of the unreliable representation of gate, which will be a level of confusion characteristic of nonhomophone words. On average, what all this means is that we should observe more interference from gait for skilled readers than for less skilled readers. But we should see more interference from gate for less skilled readers.
(JB)SWL II_f12_188-213
9/26/02
10:33 AM
Page 196
Precursors of Functional Literacy
Tests of these predictions come from experiments in which readers make simple meaning judgments (Perfetti & Hart, in press; Hart & Perfetti, under review). Sometimes the words are related in meaning (king-royalty), sometimes not (kingevening). Sometimes the first word is a homophone for both yes trials (nightevening) and no trials (knight-evening). To vary the stress on lexical quality, we varied the SOA between the onset of the first word and the onset of the second word at 150, 450, and 2000 ms. At 150 ms, representation quality should matter more than at longer SOAs, because there is little time to activate the word, and we expect only skilled readers and high frequency words to show this activation. At 2000 ms, even less skilled readers and less frequent words have time to achieve activation and selective meaning encoding. With subjects sorted on the basis of reading comprehension skill, as measured by the Nelson-Denny comprehension test, several predictions of the Lexical Quality Hypothesis were supported. First, less skilled readers showed slower decisions across the board, for control words as well as homophones, confirming a basic prediction of the Lexical Quality Hypothesis. Second, skilled readers showed homophone confusions more quickly than less skilled readers. Third, skilled readers showed confusions mainly for the less frequent member of a homophone pair. Figures 3–5 show some of the critical results. Figure 3 shows that less skilled readers are slower than skilled readers on control words, an obvious but theoretically important result. According to the Lexical Quality Hypothesis, the general lexicon of skilled readers contains more high quality representations than that of the less skilled readers. All other lexically based differences, including those from homophones, arise from this fact. And indeed homophone differences are seen. Relative to control words, skilled readers showed longer decision times at the shortest SOA of 150 ms. The homophone effect disappeared by 450 ms and did not return. The less skilled readers showed no homophone confusion (beyond their confusion for control words) at 150 ms. However, they did at 450 ms. By 2100 ms they show release from confusion. Note that there is a small difference remaining at 2100 ms for the less skilled readers, but it is not reliable statistically. However, these nonsignificant differences can be traced to individual reader and word differences that affect variance estimates. Because such differences are the heart of individual differences research, we return to these in the next section. For now, the general point is that rapid confusion leads to rapid release and delayed confusion leads to delayed release. The rapid confusion/release pattern is characteristic of the skilled readers and the delayed confusion/release pattern is characteristic of the less skilled readers. A single highly general activation/deactivation function is sufficient to explain this pattern. Figures 4 and 5 show the results for high frequency words and low frequency words, respectively, as difference scores between decision times to homophones and controls. Thus, a confusion effect is represented by a score above the zero baseline. When the higher frequency member of a pair of homophones was presented (e.g., night), less skilled readers showed confusion at the middle SOA of
(JB)SWL II_f12_188-213
9/26/02
10:33 AM
Page 197
Lexical quality 1000 Homophone Control
RT (ms)
900
800
700
600 150
450
2000
Less-skilled Less-skilled
150
SOA (ms) (ms) SOA
450
2000
More-Skilled More-Skilled
Figure 3. Skilled and less-skilled comprehenders differ in the time course of interference due to the activation of a competing homophone. Skilled readers are faster overall, and show both interference and its resolution more quickly
Homophone-Control Difference High Frequency
Difference (ms)
80
40
0
–40 150
450
Less-skilled Less-Skilled
2000
150
SOA (ms) (ms) SOA
450
2000
More-Skilled More-Skilled
Figure 4. When reaction time to reject a high-frequency control word is subtracted from reaction time to reject a dominant homophone foil of the same average frequency, lessskilled comprehenders show evidence of interference, but skilled comprehenders do not
(JB)SWL II_f12_188-213
9/26/02
10:33 AM
Page 198
Precursors of Functional Literacy
Homophone-Control Difference Low Frequency
Difference (ms)
80
40
0
– 40 150
450
Less-Skilled
2000
150
SOA (ms)
450
2000
More-Skilled
Figure 5. When reaction time to reject a low-frequency control word is subtracted from reaction time to reject a subordinate homophone foil of the same average frequency, skilled comprehenders show evidence of interference, but less-skilled comprehenders do not
450ms (and a nonsignificant effect at 2100 ms.). But no homophone confusion occurred for skilled readers at any SOA. In contrast, for the lower frequency homophones, skilled readers showed the clearest evidence for confusion. And they showed this confusion at the shortest SOA, 150 ms. Low skilled readers show some confusion, but the homophone-control difference was not reliable. To put these results in terms of our gate/gait example, skilled readers were confused by gait but not gate. Less skilled readers were confused by gate more reliably than by gait. And the time courses follow those shown in Figure 3: A rapid confusion/release effect for skilled readers, a slower confusion/release effect for low skill readers. These results are consistent with the general assumptions of the Lexical Quality Hypothesis, but more interesting is that they are consistent with two specific nonobvious predictions derived from the hypothesis: that skilled readers show earlier confusion than less skilled readers and that their confusion is restricted to the less frequent member of a homophone pair.
Homophones in context Our strategy in the meaning judgment experiment was to create a simple comprehension situation that put stress on lexical representations. In the absence of
(JB)SWL II_f12_188-213
9/26/02
10:33 AM
Page 199
Lexical quality
any supporting context, the only way for the reader to make a judgment about meaning is to use knowledge of word form and meaning. What happens when the reader can use context? Ordinarily, the constraint of sentence meaning reinforces the lexical information, adding evidence, for example, that the word gait requires some meaning other than that of gate. Other research demonstrates that top-down comprehension strategies are available to children who are less skilled readers (Perfetti, 1985; Stanovich, 1980), and we should expect this to be the case for our adult readers. We should see a reduction of homophone interference in context. Indeed this is what we found in a second experiment carried out with a new sample from the same population of college students, divided on reading comprehension according to the Nelson-Denny (Hart & Perfetti, under review). In this experiment, a constraining sentence ended in a homophone, and then, after a variable SOA, a meaning probe appeared. For example, Because of his leg injury, the man walked with an unusual gait. The key comparison was between a decision to a word related to the meaning of the homophone mate (fence) and a control word. The design was complete and fully counterbalanced, replicating the first experiment. One view of the results is seen in Figure 6, which shows the decision time difference between control words and homophones for correct “no” judgments. For both groups of readers, homophones produced no problems in context. There was an interesting but unreliable tendency for skilled readers to show confusion only at the shortest 150 ms SOA, consistent with the general pattern that skilled readers, if confused at all, show early confusions that disappear quickly. Not visible in Figure 6 is the important fact that skilled readers were faster than less skilled readers in decision times for control words as well as homophones, as expected. Also as expected, the difference in decision times between the skill groups was much smaller than when the subjects had no context to support their decision-making and had to rely on form knowledge. Although this is not the same group of subjects as those in the first experiment, it is striking to see the disappearance of a confusion effect in context. This is certainly consistent with the findings that even less skilled readers use context. Its implications, when coupled with the results of the word pair experiment, are clear: Homophone confusion depends on comprehension skill. However, the effect of comprehension skill is mediated solely by lexical factors and is not visible in comprehension. This seeming paradox is resolved as follows: The comprehension task here was targeted to the single final word. In effect, the semantic constituent of the word is reinforced sufficiently to override any unstable orthography and phonology. Less skilled readers may be more dependent on word meaning (because of weakness elsewhere in their lexical representations) and providing more cues to the meaning is very helpful. More directly persuasive, however, is the clear fact that this group does not comprehend as well as the high skill group. They were defined this way. Thus, their general comprehension problem has a lexical basis. It can be overcome only in specific tasks that are not typical of ordinary comprehension.
(JB)SWL II_f12_188-213
9/26/02
10:33 AM
Page 200
Precursors of Functional Literacy
Biased Context Homophone-Control Difference
Difference (ms)
80
40
0
– 40 150
450
Less-Skilled
2000
150
SOA (ms)
450
2000
More-Skilled
Figure 6. Homophone interference (control words minus homophones) disappeared in sentence contexts. Neither skill group showed a difference between control and homophone foils when the sentence fit only one of the homophones
Incidentally, there is nothing in these data to suggest that less skilled readers have a problem in suppression. In terms of the Structure Building Framework (Gernsbacher, 1990) one could say that our context effects demonstrate the enhancement of structures is something that less skilled readers can do well. The Verbal Efficiency Theory makes the same prediction, based not on enhancement but on basic lexical identification. The slower the process (because of lexical quality) the more room for context facilitation. Moreover, this idea generalizes to account for the facts of homophone confusion when there is no context. When the reader is solely dependent on word form, weaknesses in word form are revealed. This results in the pattern of experimental results we observed: Slower activation and deactivation by the less skilled reader and confusion restricted to more familiar word forms. Thus, a separate suppression mechanism (Gernsbacher, 1990) appears not to be needed to explain the pattern of results. One would have to say both that less skilled readers have low quality word representations and that they have faulty suppression mechanisms. Alternatively, suppression may be nothing more than the time course of competing activations. A failure to suppress would be a prolonged course (with lexical knowledge setting the upper limit) and a successful suppression would be a shorter time course of competing activations. The quality of the lexical representation would be the determining factor, both for the initial process of word identification and the time course of the resolution of the competition.
(JB)SWL II_f12_188-213
9/26/02
10:33 AM
Page 201
Lexical quality
Reading skill and the structure of lexical knowledge Given some experimental evidence that supports predictions of the Lexical Quality Hypothesis, it is of interest to demonstrate the meaningfulness of the concept of lexical quality by using assessments of word knowledge and comprehension. For this demonstration, we have data on 445 University of Pittsburgh undergraduates who participated in a range of assessments. The point was whether, by fortunate sampling of lexical knowledge constituents, we could quantify a basic idea of the Lexical Quality Hypothesis — that reading skill consists of coherent and reliable representations of constituents: reliable in that they survive threats to their stability and coherent in that the constituents are inter-correlated to the point of functional redundancy. We take our homophone experiments to have demonstrated the reliability of representations, in that threats to quality are demonstrable when words share form. In this study, we want to demonstrate the coherence of representations. By coherence, we mean roughly that given either a pronunciation or a spelling, the other two constituents of a word are immediately available. Coherence is a natural by-product of fully specified and redundant representations, as those ideas were defined in Perfetti (1991). To test the interconnections among the constituents of the model, we administered tasks that tap spelling, phonology and meaning, and sought the factor structures that underlay performance across the tasks. For a skilled reader, scores on tests of orthography, phonology, and semantics should be firmly inter-correlated. Equally important, tests for constituent knowledge should be correlated with text comprehension, as well. Although text comprehension requires skills beyond lexical knowledge, a strong lexical representation system provides the scaffolding upon which to build a representation of the text. The more coherent the word representations, the more efficient the word identification system, and the more resources are available for combining identified words into a meaningful representation of the text’s message. Thus, our expectations were that we could find evidence in the correlational structure of task performance consistent with the Lexical Quality Hypothesis that skilled readers show more coherent representations of more words than do less skilled readers.
Methods of the study Undergraduate students who participated in a variety of language experiments, also completed a number of tasks designed to tap the orthographic, phonological, and semantic lexical constituents as well as a standardized reading comprehension test. Several of the lexical knowledge tasks were computer-administered, allowing measures of speed as well as accuracy. The number of students who completed all assessments was 445. The tasks, which are listed below, cannot be “pure” measures of orthographic, phonological, or semantic knowledge. They assess performances that are supported by multiple knowledge sources. The most one can say
(JB)SWL II_f12_188-213
9/26/02
10:33 AM
Page 202
Precursors of Functional Literacy
is that a given task at least requires knowledge about one or more targeted constituents. We have provided a brief indication of whether we take a task to assess primarily orthographic (OR), phonological (PH), or semantic (SE) knowledge or mixes of two or more. 1. The Baroff Spelling Test (Wood, personal communication), adapted for computer administration. This is a spelling discrimination task in which subjects are presented with one correct and four incorrect spellings of irregular, easily misspelled words (e.g., nuisance, nuisence, newsance, newcense, newsince). Both speed and accuracy were recorded. This can be considered a direct reflection of orthographic knowledge, relatively free from semantic and phonological influences. (OR) 2. The Test of Auditory Analysis Skills (Rosner, 1979), with three additional, more difficult items. This is a phoneme elision task, that varies from removing syllables (say ‘cowboy’ without the ‘boy,’ to splitting blends (say ‘smack’ without the /m/). Only accuracy was recorded. This is phonological knowledge assessment, with some potential for support from orthographic knowledge (PH). 3. A homophone choice task. Subjects saw sixteen sentences with the last word missing, and chose one of two homophones to complete the sentence. For example, “The woman lived in the elegant brick 1. Manner, 2. Manor.” Half of the sentences were best completed by dominant homophones, and half were best completed by subordinate homophones. Both speed and accuracy were recorded. This can be considered an assessment of orthographicsemantic knowledge, relatively free from phonological influences. (OR, SE) 4. The Woodcock-Johnson Psychoeducational Battery, Word Attack Subtest = (Woodcock & Johnson, 1977), adapted for computer administration. This is a pseudoword decoding task, in which the items go from easy ‘jox’ to difficult ‘phigh.’ Both speed and accuracy were recorded. This is an assessment of orthographic-phonological knowledge, free of semantic influences. (OR, PH) 5. The Woodcock-Johnson Psychoeducational Battery, Word Identification Subtest (Woodcock & Johnson, 1977), adapted for computer administration. This is a single word reading task, in which the items go from easy ‘must’ to difficult ‘enceinte.’ Both speed and accuracy were recorded. This assesses a mix of orthographic, phonological, and semantic knowledge. (OR, PH, SE) 6. The Nelson-Denny Vocabulary Test (Nelson & Denny, 1973). Twenty items were chosen, spanning the range of difficulty of the task. Subjects were given two minutes to circle the correct definitions of these twenty words, in a multiple-choice format. Both speed (number of items attempted) and accuracy (on only those items attempted) were recorded. This test assesses semantic knowledge, although with some influence of orthography and phonology (both of which support the word identification necessary to perform the task. (SE) 7. The Nelson-Denny Comprehension Test (Nelson & Denny, 1973). This is a text comprehension test, in which eight paragraphs are followed by four to five questions, for a total of 36 questions. Subjects were given 15 minutes to
(JB)SWL II_f12_188-213
9/26/02
10:33 AM
Page 203
Lexical quality Table 1. Hypothesized demands of 6 tasks on orthographic (OR), phonological (PH), and semantic (SE) knowledge TASK
OR
Baroff spelling Homophone choice Auditory Analysis Word attack Word identification Vocabulary
+ + + +
PH
SE +
+ + +
+ +
Note. Pluses indicate the hypothesized lexical knowledge sources tapped by a given task. Arguably, all reading tasks involve all three knowledge sources, so our + marks the importance of the knowledge source relative to others for that task.
complete the test, instead of the usual 20 minutes. Both speed (number of items attempted) and accuracy (on only those items attempted) were recorded. According to our assessment, of the six tasks that measure lexical and sublexical knowledge (all but the Nelson-Denny comprehension test), one is OR, one is PH, and one is SE. The other three are mixed: one OR/PH, one OR/SE, and one is OR/PH/SE. Thus OR is involved in 4 tasks, PH in 3, and SE in 3. Our assumptions on this issue are represented in Table 1. Subjects were identified as less-skilled, average, or more-skilled readers based on the number of correct items on the Nelson-Denny text comprehension test. Each group consisted of a third of the total distribution. Figure 7 shows the distribution of comprehension scores for this group of 445 subjects. Figures 8 and 9 show the mean test scores for each group. Separate factor analyses for each comprehension group allowed us to examine the degree to which the correlation structures were comparable in their expression of lexical knowledge and comprehension. Factors were orthogonally rotated such that factors were not correlated with each other. Within this general approach, we can look for differences among the groups in the number of factors required to account for a given level of variance, the loading of different measures on a given factor, and the extent to which variables load onto more than one factor, all of which can be taken as reflecting the coherence of lexical knowledge within a group of readers of comparable comprehension skill. We separate speed and accuracy measures.2
Accuracy With a criterion of Eigenvalues > 1 and factor loadings > 0.4, the accuracy analysis achieved the following solutions: Two factors accounted for 54% of the variance for the 145 skilled and the 150 average comprehenders, and three factors accounted for 63% of the variance for the 150 less-skilled comprehenders.3
(JB)SWL II_f12_188-213
9/26/02
10:33 AM
Page 204
Precursors of Functional Literacy 45
36
27
18
9
0 0
6
12
18
24
30
36
Figure 7. Distribution of Nelson-Denny comprehension scores for the norming group of 445 subjects. The breaks in the normal distribution correspond to divisions between passages and groups of comprehension questions
Speed The speed score analyses included the six independent variables scored for speed, and achieved the following solutions: Two factors accounted for 66% of the variance for the skilled comprehenders, and three factors accounted for 82% and 80% of the variance for the average and less skilled comprehenders, respectively.
Interpretation of the factor analyses One interesting result is that for both speed and accuracy, two factors were sufficient for skilled readers. For less-skilled readers, three factors were required to account for both speed and accuracy. For average readers, two factors accounted for accuracy but three were needed to account for speed. In interpreting the factor structures, we refer to the hypothesized lexical knowledge sources — Orthographic (OR), Phonological (PH), and Semantic (SE). Comprehension as measured by the Nelson Denny is not a lexical factor, but by including it in our analyses, we hoped to be able to relate lexical factors with comprehension. Consider first accuracy measures, shown in Table 2. For skilled readers, only 2 factors emerge. Factor 1 includes all tasks except vocabulary and comprehension. It has an OR task as its highest loading task, and three of the other tasks include PH. Thus Factor 1 appears to be a factor dominated by form knowledge,
– 1.2
– 0.9
– 0.6
Auditory Analysis
Spelling Choice
Homophone Choice
Pseudoword Decoding
Word Identification
Accuracy
Vocabulary Text Recognition Comprehension
10:33 AM
– 0.3
0
Less-Skilled Less-Skilled Average Average More-Skilled More-Skilled
9/26/02
0.3
0.6
0.9
1.2
(JB)SWL II_f12_188-213 Page 205
Lexical quality
Figure 8. Means and standard deviations, between comprehension groups, for standardized accuracy scores
(JB)SWL II_f12_188-213
9/26/02
10:33 AM
Page 206
Speeling Choice – 1.2
– 0.9
– 0.6
– 0.3
0
0.3
0.6
0.9
1.2
Less-Skilled Average More-Skilled
Homophone Choice
Pseudoword Decoding
Speed
Word Identification
Vocabulary Recognition
Text Comprehension
Precursors of Functional Literacy
Figure 9. Means and standard deviations, between comprehension groups, for standardized reaction time scores
(JB)SWL II_f12_188-213
9/26/02
10:33 AM
Page 207
Lexical quality Table 2. Factor loadings of 7 tasks for orthographic (OR), phonological (PH), and semantic (SE) accuracy SKILL TASK/FACTOR Baroff spelling Homophone choice Auditory Analysis Word attack Word identification Vocabulary Comprehension VARIANCE EXPLAINED
High 1 .8 .5 .6 .6 .6 30
Average 2
.4 .9 .7 24
1 .7 .5 .6 .7 .7 32
2
Low 1 .8 .7 .7
.8 .8 22
24
2 .7 .8
3
.5
21
.8 .7 18
OR and PH. Factor 2 links word identification, vocabulary and text comprehension. We can characterize Factor 2 as an SE factor. But interestingly, word identification loads on both factors. Thus for the skilled readers, there is a Form Factor (OR/PH) and a Meaning Factor, with word identification linking the two. For less skilled readers (n = 150), Factor 1 is dominated by PH, with the auditory analysis task being the highest loader, followed by word identification and word attack. Factor 2 was a semantic-comprehension factor, represented only by the two Nelson-Denny tests. Factor 3 appears to be a form factor with orthography dominant, grouping homophone choice, spelling choice, and word attack. Whereas the skilled readers had word identification as a linking variable, the less skilled readers had word attack as a linking variable linking a PH-dominant factor (1) with an OR-dominant factor (3). Average comprehenders (n = 150) had a factor structure similar to moreskilled comprehenders, with two factors sufficient. Factor 1 included 5 variables that included OR and PH, and, like the first factor for the skilled group, can be considered a Form factor. Factor 2 picked up the two Nelson-Denny tasks, and can be considered a meaning-comprehension factor. Unlike the skilled readers, however, word identification did not serve as a linking variable. The middle group showed no linking variables. For lexical speed measures, Table 3 indicates that, for skilled readers, two factors are again sufficient. Factor 1 includes word identification and word attack as the two strongest loadings, adding two OR tasks, spelling and homophone choice. This factor thus includes tasks that tap OR, PH, and SE knowledge, a general lexical factor that includes sublexical phonology. Factor 2 includes again the OR tasks of homophone choice and spelling choice, adding comprehension and vocabulary. This Factor can be interpreted as an orthographic-semantic-comprehension factor that picks out the connection between word familiarity and comprehension. The less skilled readers required three factors again. The first factor was loaded by word attack and word identification strongly, with homophone choice last. This can be characterized as a general lexical factor similar to Factor 1 for skilled readers,
(JB)SWL II_f12_188-213
9/26/02
10:33 AM
Page 208
Precursors of Functional Literacy Table 3. Factor loadings of 6 tasks for orthographic (OR), phonological (PH), and semantic (SE) speed SKILL TASK/FACTOR Baroff spelling Homophone choice Word attack Word identification Vocabulary Comprehension VARIANCE EXPLAINED
High 1 .4 .4 .9 .9 33
2 .7 .7
Average 1
2 .9 .9
Low 3
.4 .9 .9
.9 .9 .6 .7 33
31
29
1
.8 .8 22
31
2 .9 .8
.5 29
3
.8 .7 20
except that spelling did not load on this factor. Semantic and phonological knowledge may be more represented in this factor than orthographic knowledge for less skilled readers. For Factor 2, spelling and homophone choice were high loaders, with comprehension an additional variable. Thus Factor 2 is primarily an OR factor at the lexical level. Factor 3 was again the combination of comprehension and vocabulary from the Nelson-Denny. The middle group of readers, like the less skilled readers, required a threefactor solution. In an unusually tidy structure, Factor 1 was word identification and word attack, Factor 2 spelling and homophone choice, and Factor 3, vocabulary and comprehension, in each case with very high factor loadings. Thus, the average college reader’s factor structure for speeded tasks can be characterized as (1) General Lexical processing, including pseudoword reading. (2) Orthography. (3) Semantics and Comprehension. A summary of these interpretations is in Table 4, which refers to the hypothesized lexical knowledge sources — Orthographic (OR), Phonological (PH), and Semantic (SE) in addition to comprehension. Table 4 also shows links between factors where a variable loaded on more than one factor. If we ignore the middle group of readers for now, we can summarize the interesting contrasts between high and low skill groups as follows, using highly inferential descriptors: Skilled readers show a highly generalized lexical processing factor, one we characterized as an integration of lexical knowledge sources (OR, PH, and SE). When accuracy is measured, this factor is one of lexical form, combining orthographic and phonological knowledge; when speed is measured and we have one fewer phonological tasks (the Auditory Analysis is accuracy only), this lexical factor includes semantic information, so is Form + Meaning. Second, two factors account for more of the variance for skilled readers than for less skilled readers. For accuracy, skilled readers can be said to reflect two factors, lexical form and meaning-comprehension. For speed, the two factors are better characterized as lexical knowledge (form + meaning) and comprehension (meaning + comprehension). Third, skilled readers show linking variables, variables that load on more than one factor. In accuracy, this variable is word identification, which is
(JB)SWL II_f12_188-213
9/26/02
10:33 AM
Page 209
Lexical quality Table 4. Inferred structure of lexical knowledge and comprehension for three levels of comprehension skill Accuracy Measures
Factor 1
Factor 2
Factor 3
High Skill Low Skill
OR-PH PH
SE-Comp SE-Comp
OR
Medium Skill
OR-PH
SE-Comp
High Skill
OR-PH-SE
OR-SE-Comp
Low Skill
PH-SE
OR-Comp
SE-Comp
Medium Skill
PH-SE-OR
OR
SE-Comp
Links Between Factors Word Identification Pseudoword reading, 1& 3 None
Speed Measures Spelling & Homophone knowledge Homophone Knowledge, 1& 2 None
Note. The cell entries are the inferred lexical knowledge sources based on the factor analyses: PH = phonological, OR = orthographic, SE = semantic. All tasks allowed both accuracy and speed measures, except the Auditory Analysis task, a PH task. Accordingly the potential to observe a PH component in speed measures is greatly reduced. Links between factors are tasks that loaded on more than factor. Such links can be said to mark increased lexical coherence.
important for the lexical form factor and the meaning-comprehension factor. For speed, the links are made by orthographic tasks that tap spelling knowledge (presumably linked to word meaning knowledge). Word familiarity drives response speed and supports both rapid word identification (Factor 1) and comprehension (Factor 2). The evidence for coherence among knowledge sources — in the factor structures and the links — is less for low skill readers. They require 3 factors instead of 2 and their links are different. In accuracy, their first factor lacks orthography, instead being mainly phonology. Orthography comes in as a separate third factor. They show a link, but instead of word identification, it is pseudoword reading (word attack), linking their first and third factors. In speed measures, the picture is similar — no OR component to their primary lexical factor, which instead is PH-SE. For the low skilled readers, orthographic knowledge may not be integrated with other lexical knowledge as powerfully as it is for skilled readers. We do not suggest sharp qualitative differences here. Indeed, the Lexical Quality Hypothesis is an explicitly graded one. Our contrasts between high and low skill readers pick out small differences in correlational structures that get reflected in factor structure and interpretation. But there are no dramatic reversals in the pattern of correlations that are at the bottom of the pyramid. Only from the top can we see hints that the extent to which pieces of knowledge cohere in the two groups may be different. Of course, part of what gets reflected in these analyses are differences within groups, which is what cause correlations to be lower. The less skilled readers are more variable. Still, our theoretical argument takes variability
(JB)SWL II_f12_188-213
9/26/02
10:33 AM
Page 210
Precursors of Functional Literacy
as part of the story. That is what it means to have lower quality representations. The chances that two words for the same reader will vary in quality are higher and the chances that two readers of the same word will vary are also higher. Whether we can identify which lexical constituents are most likely to be the weak constituents and thus be the cause of low quality is not yet clear. Probing a reader’s knowledge of a word is seldom if ever carried out in the way that is necessary to really assess lexical quality. The tasks we used here, although they were sensible and sensitive (providing speed of processing as well as accuracy), were far from the ideal implied by the concept of lexical quality. They were not repeated tests of the same word, and they were not particularly sensitive to just certain constituents. Still, the results are suggestive of the following: Low skill readers have less well integrated (or less coherent) lexical representations. The first factor for both speed and accuracy showed this difference, and the piece missing seems to be orthography. To head off an unwarranted conclusion, this is not the same as saying that the low skill readers didn’t know the spellings of words. In fact, they differed the least from the average and skilled readers on the Baroff spelling and Homophone choice tasks. They also do not know meanings, how to decode difficult pseudowords, and they are low on phonological awareness. Rather, it is that their orthographic knowledge is doing a little less for their overall word knowledge. It is less interconnected with both their vocabulary knowledge and their decoding ability, which are actually highly related. If we assume that the main thing this group lacks is enough experience in reading, things fall into place. They are behind in developing the high quality orthographic representations that come with lots of reading experience with specific words (Stanovich & West, 1989). The mid-level readers fill in the picture. They are more similar to skilled readers in some ways, more similar to less skilled readers in others. Like high skill readers, their accuracy performance can be described by two factors; but unlike the skilled readers, word identification is not a linking variable. Like skilled readers, their first factor for both speed and accuracy is a generalized lexical form factor for accuracy, and a generalized form + meaning factor for speed. Like less skilled readers, their speed performance requires 3 factors; but they show no linking variables. These average readers can be characterized as on their way to fully integrated lexical knowledge, but their lower loadings on orthographic measures in speed tasks and the absence of linking variables in both speed and accuracy suggests that their present level of reading experience falls short of the skilled readers. We must apply caution to this exercise of interpretation. Not to be forgotten is that underlying these factor structures are variability and co-variability between subjects on the same task and between tasks for the same subject. The factor structures are not about skill levels, but about the configurations of co-variance that are expressed in task performances. Even in drawing attention to the different roles played by some task, e.g., orthographic performance, for skilled and less
(JB)SWL II_f12_188-213
9/26/02
10:33 AM
Page 211
Lexical quality
skilled readers, we emphasize that it is the inter-task correlations, not the level of orthographic knowledge that is interesting.
Summary and conclusions The Lexical Quality Hypothesis includes some highly general assumptions about the nature of reading and some specific claims about effective word knowledge. Skilled reading rests on high quality word representations that include well specified orthographic, phonological, and semantic-syntactic information. The effects of variation in quality are seen in comprehension, both in comprehension as assessed in standard ways, and in simple scaled-down comprehension tasks such as the word meaning judgment tasks we described in the first part of this chapter. Our results in simple word pair comprehension produced non-obvious results consistent with the Lexical Quality Hypothesis. When words that shared phonological forms with other words (homophones) were presented for meaning judgments, readers briefly showed confusions, as if the unpresented homophone were being considered in the meaning judgment. Especially interesting, however, was that skilled readers actually showed this confusion more quickly than did less skilled readers. And their confusion also ended more quickly. Also consistent with the Lexical Quality Hypothesis is that skilled readers showed no confusion when presented with the high-frequency member of a homophone pair. Thus, the homophone experiments provide evidence that high quality representations (higher for high frequency words, higher for high skilled readers, higher for words that aren’t homophones) have predictable consequences in simple comprehension tasks. They demonstrate also the low knowledge that less skilled readers have about word forms. When context could support the meaning decisions, the confusions they showed for homophones disappeared. To further illustrate the concept of lexical quality, we analyzed the factor structure extractable from the performance of 445 college students on tasks that tap mixes of orthographic, phonological, and semantic knowledge. The factor structures that emerged suggest a more coherent lexical knowledge structure for skilled readers than less skilled readers. Orthographic and phonological structures are closely linked with each other for skilled readers, and less so for less skilled readers. The lack of integration of orthographic performance for low skilled readers suggests that spelling knowledge is not serving word reading in the same way it is for skilled readers, as a group. It is clearly true, without factor structures, that skilled comprehenders know more about words and word forms — their spelling, their meaning, their pronunciations — than do less skilled comprehenders. This is the simple general idea underlying the lexical quality hypothesis. However, an implication of our analysis is not that there is some constituent of word knowledge, e.g., spelling, that needs to be fixed for a less skilled reader. The implication is that experience with words, retrieving and using their spellings, pronunciations
(JB)SWL II_f12_188-213
9/26/02
10:33 AM
Page 212
Precursors of Functional Literacy
and meanings is a critical foundation of reading skill. The coherence and reliability of constituent knowledge is a consequence of such lexical experiences, repeated generously.
Notes . The word in Figure 1 is actually a homophone. However, this makes no difference for the general analysis. Just assume that only this word and not its homophone mate is known to the reader. . Another way to examine coherence of shared variance is to use an oblique rotation and examine correlations between factors. We chose to rotate orthogonally to obtain independent factors. Oblique rotation produces the same number of factors with very similar structures, for both speed and accuracy analyses. Skilled comprehenders have higher correlations between factors than do average comprehenders, and average comprehenders have higher correlations between factors than do less-skilled comprehenders. . There is a tradeoff between number of factors and variance explained. In the optimal solution (with two factors for skilled and average comprehenders and three factors for lessskilled comprehenders), the optimal ratio of number of factors to variance is achieved; the skilled and average comprehenders have nearly the same amount of variance explained with only two factors as the less-skilled comprehenders have with three factors. When a two-factor or three-factor solution is forced, the amount of variance explained for each group is equated. Three factors explain 67%, 67%, and 63% of the variance and two factors explain 53%, 54%, and 49% of the variance for the skilled, average, and less-skilled groups, respectively. This applies similarly to the speed analysis. Three factors explain 80%, 82%, and 80%, and two factors explain 66%, 64%, and 63% of the variance for the skilled, average, and less-skilled groups, respectively.
References Berent, I. & Shimron, J. (1999). The representation of Hebrew words: Evidence from the obligatory contour principle. Cognition, 64, 39–72. Bock, K. & Levelt, W. (1994). Language production: grammatical encoding. In M. A. Gernsbacher (Ed.), Handbook of Psycholinguistics (945–984). San Diego: Academic Press. Duffy, S. A., Morris, R. K. & Rayner, K. (1988). Lexical ambiguity and fixation times in reading. Journal of Memory and Language, 27, 429–446. Gernsbacher, M. A. (1990). Language Comprehension as Structure Building. Hillsdale, NJ: Erlbaum. Gernsbacher, M. A. & Faust, M. E. (1991). The mechanism of suppression: A component of general comprehension skill. Journal of Experimental Psychology: Learning, Memory, and Cognition, 17, 245–262. Hart, L. A. & Perfetti, C. A. Quality of lexical representations affects reading comprehension. Under review. Hogaboam, T. W. & Perfetti, C. A. (1978). Reading skill and the role of verbal experience in decoding. Journal of Educational Psychology, 70, 717–729. Nelson, M. J. & Denny, E. C. (1973). The Nelson-Denny Reading Test. Houghton Mifflin Company.
(JB)SWL II_f12_188-213
9/26/02
10:33 AM
Page 213
Lexical quality Perfetti, C. A. (1985). Reading Ability. New York: Oxford University Press. Perfetti, C. A. (1991). Representations and awareness in the acquisition of reading competence. In R. Laurence & C. A. Perfetti (Eds.), Learning to Read: Basic Research and its Implications. Hillsdale, NJ: Erlbaum. Perfetti, C. A. (1992). The representation problem in reading acquisition. In P. B. Gough, L. C. Ehri & R. Treiman (Eds.), Reading Acquisition (145–174). Hillsdale, NJ: Erlbaum. Perfetti, C. A. & Hart, L. (in press). The lexical bases of comprehension skill. In D. Gorfien (Ed.), On the consequences of meaning selection. Washington: American Psychological Association. Perfetti, C. A. & Roth, S. F. (1981). Some of the interactive processes in reading and their role in reading skill. In A. M. Lesgold & C. A. Perfetti (Eds.), Interactive Processes in Reading (269–297). Hillsdale, NJ: Erlbaum. Rosner, J. (1979). Test of Auditory Analysis Skills. Navato, CA: Academic Therapy Publications. Stanovich, K. E. (1980). Toward an interactive-compensatory model of individual differences in the development of reading fluency. Reading Research Quarterly, 16, 32–71. Stanovich, K. E. & West, R. F. (1989). Exposure to print and orthographic processing. Reading Research Quarterly, 24, 402–433. Woodcock, R. W. & Johnson, M. B. (1977). Woodcock-Johnson psycho-educational battery. Hinghan, MA: Teaching Resources. Address University of Pittsburgh 651 Learning Research & Development Center 3939 O’Hara Street Pittsburgh PA 15260, USA
[email protected],
[email protected]
(JB)SWL II_f13_214-228
9/26/02
10:35 AM
Page 214
(JB)SWL II_f13_214-228
9/26/02
10:35 AM
Page 215
Relationships between reading and writing skills in the intermediate grades Hanna S. Mäki* University of Turku
Marinus J. M. Voeten University of Nijmegen
Marja M. S. Vauras University of Turku
Pekka Niemi University of Turku
Present theoretical models of writing (see Hayes, 1996; Rijlaarsdam & Van den Bergh, 1996) suggest that the ability to read is one fundamental sub-process needed in skilled writing. Reading already written text can be used in generating and formulating more text. Also, writing output can only be revised on the basis of information gained by evaluative reading or text interpretation (Hayes, 1996; Hayes, Flower, Schriver, Stratman & Carey, 1987). This implies that writers must be able and willing to read their own text during the writing in order to find out if something is wrong with either spelling or text structure, and in order to decide what to write next and how to do it. In this respect, however, different theoretical models for mature and immature writing have been proposed. Models explicitly recognising reading as one sub-process or cognitive skill needed in writing have usually been tested with more advanced writers (Hayes, 1996), whereas simpler models have been proposed for immature writers. The researchers argue that models of skilled writing necessarily include fine-grained sub-skills and their functional relationships that cannot yet be detected among immature writers (the Simple View by Juel, Griffith & Gough, 1986; the Knowledge Telling Strategy by Scardamalia & Bereiter, 1987). For example, it has been proposed that reading does not have a role in the immature writing process. Some studies have indeed shown that beginning writers rarely read their own text during the writing process (Scardamalia & Bereiter, 1987). This does not, however, allow us to conclude that the sub-process of reading should be excluded from the model of immature writing. We argue that if the model of writing is made too simple (for example, by excluding the text interpretation process), important
(JB)SWL II_f13_214-228
9/26/02
10:35 AM
Page 216
Precursors of Functional Literacy
aspects of the beginning writing process may go undetected. Therefore, we suggest that even with immature writers, it is important to test models that take into account the complexity of the writing skill. This enables us, also, to conclude at which stage the patterns observed with more advanced writers emerge in the course of writing skill development. In the present longitudinal study, we explored a path model that assumes an association between writing and reading skills in young beginning writers (grades 2 and 3). In particular, we assumed that, at this early stage of writing development reading skills may influence writing skills. Two components of reading skill were studied, namely word recognition speed and reading comprehension. Problems with either of these aspects of reading skill can be expected to make writing difficult. Poor word recognition forces a writer to read his text with effort, which taxes attentional capacity to a degree that too little attention may be allocated to evaluation of micro-level text properties, such as spelling (Kellogg, 1996), or the quality of the produced text. Significant positive correlations have indeed been found between basic word attack skills (decoding or word recognition) and spelling proficiency in primary grades (Boland, 1993), in intermediate grades (Berninger, Cartwright, Yates, Swanson & Abbott, 1994; Zutell & Rasinski, 1989), and even at high school level (Shankweiler, Lundquist, Dreyer & Dickinson, 1996). Another aspect of writing, the quality of composition, has also been found to correlate with general reading skills. Abbott and Berninger (1993) reported correlations ranging from .25 to .56 between reading of real words and quality of written stories in the primary and intermediate grades (see also Berninger et al., 1994). Poor reading comprehension can also interfere with the writing process and lead to poor writing products. Hayes (1996) sees text revision as a process in which a task schema is used to control several cognitive sub-processes, one of which is critical reading. Hayes and others (1987) related the critical reading of text outputs to the process of reading comprehension as described by Just and Carpenter (1980). They suggested that revising writers read to comprehend and criticise their text at different levels: spelling, grammar, logic of the message, text structure and tone. Thus, poor reading comprehension could seriously inhibit effective text revision. Some empirical studies have lent support to this suggestion. For example, in the study of Cox, Shanahan, and Sulzby (1990), students with proficient reading comprehension scored significantly more highly in tasks measuring the cohesive ties and cohesive harmony of their writings. Also Juel (1988) reported significant positive correlations between reading comprehension and composition coherence among beginning writers. Thus, the positive influence of reading skills on the writing skills of beginning writers has gained support in several studies. However, there is great variation in how reading and writing skills have been measured and in which skills have been chosen for a particular study. The fact that various reading and writing variables are intercorrelated also sets demands on the statistical method. In this study, we wanted to test a path model that combines four domains of literacy skill:
(JB)SWL II_f13_214-228
9/26/02
10:35 AM
Page 217
Relationships between reading and writing skills
word recognition, reading comprehension, spelling, and composition coherence. It is well known that poor word recognition usually makes also reading comprehension difficult (e.g., Juel, Griffith & Gough, 1986). By including both word recognition and reading comprehension in one path model we can see what kinds of direct or indirect effects these reading skills have on writing skills. Spelling skill and the skill to coherently compose a text were both measured from the same writing assignment, because we wanted to evaluate the product of a complete writing process. Our young participants had to cope with both the production and the interpretation phase of the writing process simultaneously (see Hayes, 1996). We expected that spelling accuracy and composition coherence could be predicted from the previous year’s performance (Boland, 1993; Juel, 1988; Klicpera & Schabmann, 1993). We also included paths indicating an association between spelling and composition coherence. If spelling is incompletely automated, the writer is forced to be effortful with it. This again demands working memory capacity which, in turn, impedes effective execution of other sub-processes of composition, possibly leading to poor writing outputs (Kellogg, 1996, see also Graham, 1990; Graham, Berninger, Abbott, Abbott & Whitaker, 1997; Juel, 1988; Mäki, Voeten, Vauras & Poskiparta, 2001). Two versions of the path model were developed to test two different ways in which the writing skills might have been influenced by the reading skills. In the first version of the model, these relationships are synchronous, while in the second version lagged relationships are introduced. The first path model allows us to study the relationships between reading and writing skills within each grade level. We expected to find effects of word recognition speed on both number of spelling errors and composition coherence in grade 2 as well as grade 3. We also expected to find at both grade levels, effects of reading comprehension on number of spelling errors and on composition coherence. Higher word recognition speed and better reading comprehension should be associated with fewer spelling errors and better composition coherence. Since the auto-regressions of writing skills were included in the model, the effects of reading skills on writing skills in grade 3 may be interpreted as effects on the development of writing skills from grade 2 to grade 3. The second path model that included all lagged relationships between writing skills in grade 3 and reading skills in grade 2, implies the same relationships between the variables within the second grade but gives a different specification for the effects of reading skills on writing-skill development. In this second model, it is assumed that it takes one school year for reading skills to influence writing skills, while the first model assumes this causal lag to be much shorter than one school year. Since we expected a high stability of reading skills over the period from grade 2 to 3, it might be the case that there are no large differences between these two models, but the second model would allow us to make somewhat stronger causal claims.
(JB)SWL II_f13_214-228
9/26/02
10:35 AM
Page 218
Precursors of Functional Literacy
Method Participants The participants (n = 171) were selected from a sample of 252 Finnish-speaking children who participated in a larger longitudinal study from pre-school to the end of the fourth grade. This original sample included children entering the first grade of four primary schools in a Finnish town of about 160,000 inhabitants. At the start of the study, the mean age of the children was 6 years 8 months. In Finland, children enter the first grade in August of the year they become seven years old. We selected children with data available for the second and third grades. We excluded those who were assigned to special education or who had repeated an earlier grade. There were 95 (56%) boys and 76 (44%) girls (matching the sex ratio in the original sample of 252 children). Three participants were from ethnic minorities, but they spoke Finnish fluently. All the schools shared the mainstream curriculum of Finnish general education, including teaching in the skills that were the targets of this study (word recognition, reading comprehension, and composition). As is usual in Finland, the prevalent method of teaching word recognition used the phonics approach.
Instruments Word recognition A computer-aided lexical decision task was used (Kinnunen, Vauras & Niemi, 1998; Poskiparta, Niemi & Vauras, 1999). Children had to quickly decide whether the 30 short letter strings (two syllables) presented on a computer screen one at a time were real or nonsense words. First a warning signal (a + mark accompanied by a beep) appeared on the screen, and one second later, the word appeared in the same location. The child made the choice by pressing as quickly as possible either the key marked ‘yes’ (a real word) or the key marked ‘no’ (a nonsense word). Before starting the test, the child had a practice session on reacting with the ‘yes’ and ‘no’ keys. During the test, the computer recorded the time from the onset of the word to the pressing down of either of these keys. The mean times (in seconds) of correct decisions for words and nonsense words separately were calculated for each subject. In the analyses the average reaction time in seconds for words and nonsense words was used. The mean numbers of wrong decisions in the lexical decision task were extremely low in our sample (0.9–1.0 for words and 0.7–1.1 for nonsense words, the possible maximum being 30). Because of this floor effect, we decided not to use the accuracy as an indicator of word recognition skill. Reading comprehension Parallel versions of the reading comprehension test were used in grades 2 and 3. Two narrative and two expository texts (95 and 135 words) describing the behaviour of an animal were used (for more details, see Kinnunen et al., 1998). The texts were
(JB)SWL II_f13_214-228
9/26/02
10:35 AM
Page 219
Relationships between reading and writing skills
presented in two separate sessions. Subjects were allowed to read the texts for as long as they needed in order to answer orally seven main idea questions. The subjects were not allowed to return to the texts while answering. Answers were scored on a 0-to-2 or a 0-to-3 point scale, depending on the complexity of an utterance required for an exhaustive answer. The final score for each text was the percentage of summed scores relative to the maximum score. The mean score on all four texts (two narrative and two expository texts) was used in statistical analyses.
Composition coherence A picture of an animal in its natural environment was projected on the classroom wall, and the children were asked to write a story about the picture. There were no time restrictions for this task. Composition coherence was assessed on an eight point scale (for more details, see Mäki, Voeten, Vauras & Poskiparta, 2001). All written outputs were typed prior to coherence rating. In our study, coherence is defined at the textual macro-level as the organization of the text under a structural schema or script (cf. Vauras, Hyönä & Niemi, 1992). We chose to use holistic evaluation of the quality of compositions because evaluation at the full text level has been recommended for studying written language development (e.g., Scinto, 1986). Spelling Spelling was evaluated on the basis of the text produced in the composition task. The following spelling errors were counted: a) missing, wrong or added letter, b) an incorrectly placed or missing full stop, and c) incorrectly connected words (e.g., ‘pesä pallo’ correctly, ‘pesäpallo’, in English ‘baseball’). The total number of spelling errors was divided by the number of words in the story. The number of spelling errors per ten words was used in statistical analyses. Procedure Measurements were taken at the end of the second and third grade. Composition was tested in classroom situations, while all other tasks were given individually during regular school hours. Two independent raters evaluated the compositions and answers to the reading comprehension tasks. For composition, both raters first evaluated 20% of the stories. In the case of discrepancy between the raters the criteria were discussed. Then, another 20% of the stories were scored again by both raters to assess rater agreement. The following agreement percentages were obtained: composition coherence 87% and spelling 94%. One rater scored the remaining stories. For the reading comprehension tasks, two raters evaluated all the answers. The agreement varied from 80% to 98% depending on the item. Disagreements were minor, and they were resolved by discussion. For the measure of word recognition speed, estimates of reliability were evaluated in our previous study (Mäki et al., 2001), for the second grade. Reliability coefficients for real words and nonsense words separately ranged from .85 to .95.
(JB)SWL II_f13_214-228
9/26/02
10:35 AM
Page 220
Precursors of Functional Literacy
Path analysis was used to analyse the relationships between word recognition speed, reading comprehension, spelling, and composition coherence. This was done with LISREL 8.30 (Jöreskog & Sörbom, 1999), using the maximum likelihood method for parameter estimation. The models were applied to the variances and covariances of the variables. Missing values were dealt with using the imputation procedure of PRELIS 2.3 (Jöreskog & Sörbom, 1999). Altogether 17 missing values (1.2% of the total data set of 171 subjects with scores on 8 variables) were imputed. In addition to the chi-square test, several indices were used to assess the goodness of fit of the models: the Root Mean Square Error of Approximation (RMSEA), the Expected Cross-Validation Index (ECVI), and the Comparative Fit Index (CFI). The Chi-square test measures the discrepancy between the observed covariance matrix and the covariance matrix implied by the model. The chi-square in relation to its degrees of freedom is useful for ascertaining that the models are not too far removed from the sample data. Browne and Cudeck (1993) advocated the use of RMSEA and ECVI. The RMSEA refers to the lack of fit, per degree of freedom, of the model to the population covariance matrix; a value of 0.08 or less would be a reasonable error of approximation. The ECVI estimates what the discrepancy would be between the fitted model and the covariance matrix computed from another sample. Browne and Cudeck (1993) showed that ECVI is especially useful for comparing different models in a small sample. The CFI assesses how much better the model fits in comparison to the independence model. We would not consider a model that has a CFI below .90. One-sided testing at the 5% level was used to test the statistical significance of the parameter values. This criterion was chosen because of the small sample size (n = 171), and because of the fact that the direction of all relationships was hypothesized in advance.
Results Descriptive statistics Descriptive statistics of the observed variables are presented in Table 1. The word recognition tests (lexical decision) were speed measures: word recognition time needed in seconds. The distributions were skewed to the right and therefore a logarithmic transformation was applied. Measures of spelling in the second and the third grade had strong floor effects. To remedy this, we declared these variables censored below when estimating the covariance matrix by Prelis 2.3 (Jöreskog & Sörbom, 1999). After these modifications, univariate skewness and kurtosis statistics were within reasonable limits (see Table 1). Correlations between variables are presented in Table 2.
(JB)SWL II_f13_214-228
9/26/02
10:35 AM
Page 221
Relationships between reading and writing skills Table 1. Descriptive statistics of variables M
SD
Ma
SDa
Skewnessa
Kurtosisa
Max.
Grade 2 Word recognition speedb Reading comprehension Spelling errorsc Compostion coherence
2.65 44.51 1.42 3.03
1.22 16.44 1.80 1.33
0.88
0.45
0.91
2.28
–0.11 –0.09 0.99 0.48
–0.09 –0.48 1.17 0.36
– 100 – 8
Grade 3 Word recognition speedb Reading comprehension Spelling errorsc Compostion coherence
1.89 55.12 0.62 4.13
0.76 15.76 0.77 1.64
0.56
0.39
0.51
0.88
0.17 –0.68 1.89 0.05
–0.06 0.37 7.12 –0.50
– 100 – 8
Variable
Note. a Values after transformations. c Variable was censored below.
b
Logarithmic transformation was carried out.
Table 2. Correlations between variables 1
2
3
Grade 2 1. Word recognition speed 2. Reading comprehension 3. Spelling errors 4. Composition coherence
–.31 .38 –.21
–.19 .15
.07
Grade 3 1. Word recognition speed 2. Reading comprehension 3. Spelling errors 4. Composition coherence
.83 –.37 .37 –.21
–.27 .72 –.11 .28
.38 –.14 .31 –.05
4
5
6
7
–.17 .15 –.13 .25
–.35 .35 –.17
–.10 .18
–.10
Note. Lower scores indicate better task performance in word recognition and spelling.
Model 1: Synchronous effects of reading skills on writing skills within each grade level We began by testing a path model based on the idea that better reading skills (word recognition speed and reading comprehension) will promote writing skills (spelling and composition coherence) within each grade level, without assuming effects across grades other than the stability of each skill. A good fitting model was obtained, but five of the pre-supposed regression coefficients were statistically non-significant. These non-significant effects were removed one by one, always removing the relationship with the lowest t-value, resulting in Path Model 1. There was no association between reading comprehension and spelling at
(JB)SWL II_f13_214-228
9/26/02
10:35 AM
Page 222
Precursors of Functional Literacy
Word recognition speed
.83***
–.16**
–.31*** Reading (.09) comprehension
Word recognition speed (.68)
.68***
Reading (.54) comprehension
.21**
Number of (.15) spelling errors
.27***
.38*** Number of (.15) spelling errors –.28***
.17*
Composition coherence (.07)
GRADE 2
.15* .22**
Composition coherence (.08)
GRADE 3
j2(17, N = 171) = 23.86, p = 0.12. RMSEA = 0.046. ECVI = 0.36. CFI = 0.98 Note. * p < .05. ** p < .01. *** p < .001
Figure 1. Standardised parameter estimates for the relations between word recognition speed, reading comprehension, spelling errors, and composition coherence in Path Model 1 (proportion of variance explained in parentheses)
either grade level, between reading comprehension and composition coherence in grade 2, between spelling and composition coherence in grade 3 or between word recognition speed and composition coherence in grade 3. All other relationships were as expected, except the effect of spelling errors on composition coherence in grade 2. Good composition coherence seemed to be associated with a high rate of spelling errors in grade 2. As expected, accurate spelling and good composition coherence were associated with rapid word recognition in second grade. In third grade, however, this was only found for spelling, not for composition coherence. Note that the model controls the reading-writing relationships in third grade for writing variables in second grade. This allows us to interpret the effect of word recognition speed on spelling accuracy in third grade as an effect on change in spelling skill between grades 2 and 3. Thus, word recognition speed has an effect on the level of writing performance in second grade, and an effect on change in spelling performance between second and third grade. Composition coherence in third grade was only indirectly affected by word recognition speed (mediated by reading comprehension). Partial support was also found for the expectations concerning the effects of
(JB)SWL II_f13_214-228
9/26/02
10:35 AM
Page 223
Relationships between reading and writing skills
Word recognition speed
.83***
–.17**
–.31*** Reading (.09) comprehension
.67***
Reading (.54) comprehension .30***
.38*** Number of (.15) spelling errors –.28***
Word recognition speed (.68)
.19**
.25***
.17*
Composition coherence (.07)
GRADE 2
Number of (.17) spelling errors
.21**
Composition coherence (.11)
GRADE 3
j2(17, N = 171) = 13.49, p = 0.70. RMSEA = 0.00. ECVI = 0.32. CFI = 1.00 Note: * p < .05. ** p < .01. *** p < .001
Figure 2. Standardised parameter estimates for the relations between word recognition speed, reading comprehension, spelling errors, and composition coherence in Path Model 2 (proportion of variance explained in parentheses)
reading comprehension. Good reading comprehension tended to result in good composition coherence in grade 3, but there was no such significant relationship in grade 2. This means there was an effect of reading comprehension on the change in composition coherence between second and third grade. For spelling accuracy, however, the hypothesised effects of reading comprehension were not found, when controlling for the effects of word recognition speed. All reading and writing skills were significantly predicted by the same skills evaluated one year earlier. Compared with the reading variables, the stability coefficients were appreciably lower for the writing variables.
Model 2: Lagged relationships between reading and writing skills Next, we tested whether the associations found between reading and writing variables within grades 2 and 3 would also be found across the two grade levels. Therefore, we extended the model with lagged effects of grade-2 variables on grade3 variables. Again, the statistically non-significant relationships were removed one by one, resulting in Path Model 2.
(JB)SWL II_f13_214-228
9/26/02
10:35 AM
Page 224
Precursors of Functional Literacy
There was no association between second-grade reading comprehension and third-grade spelling, nor between second-grade spelling and third-grade composition coherence. The other three lagged effects were as expected. All the synchronous effects of reading on writing variables in grade 3 that were found in Path Model 1 were no longer statistically significant; the lagged effects just replaced them. Model 2 confirms that word recognition speed has an effect on the development of reading comprehension and of spelling between grades 2 and 3, as was seen in Model 1. The effect of word recognition speed on composition coherence was only indirect, mediated by reading comprehension. Moreover, there was an effect of reading comprehension on the development of composition coherence between grades 2 and 3. Both models showed the same reading-writing relationships, but the lagged regression effects appeared to be somewhat stronger than the synchronous effects. This was especially the case for the effect of second-grade reading comprehension on third-grade composition coherence. These lagged effects were even larger than the stability coefficients for the writing variables. The predictive power of the model was rather low. The model could explain about 10% of the variance in composition coherence, and about 15% of the variance in spelling at both grade levels (Figures 1 and 2).
Discussion Path analysis was used to study the associations between word recognition speed, reading comprehension, spelling, and composition coherence among Finnishspeaking second and third graders. We tested a path model that expects effects of reading skills on writing skills in the early stage of writing skill development. Accurate spelling seemed to be related to rapid word recognition in both grades 2 and 3. This result is in accordance with Boland (1993), Berninger et al. (1994), and Zutell and Rasinski (1989). Among our second graders, also good composition coherence seemed to be related to skilled word recognition, paralleling the results of Abbott and Berninger (1993). This association was not shown one year later. However, in the third grade, composition coherence was indirectly affected by word recognition speed, mediated by reading comprehension (cf. Cox, Shanahan & Sulzby, 1990; Juel, 1988). Our results revealed no association between spelling and reading comprehension either in second or third grade. This finding differs from that of Boland (1993), who showed that reading comprehension had a significant effect on spelling among Dutch second graders. An interesting finding was that both of the path models we used showed the same reading-writing relationships, but the lagged regression effects appeared to be somewhat stronger than the synchronous effects. This gives us some evidence that over one school year, here is an effect of reading skills on writing skills. Those who read better seem to cope better with writing skill development. One reason for the association between word recognition and spelling skills could be that
(JB)SWL II_f13_214-228
9/26/02
10:35 AM
Page 225
Relationships between reading and writing skills
fluent readers are able and, perhaps, also motivated to read their own text, which enables them to check their spelling. Furthermore, good comprehenders may also be skilled in evaluating the quality of their own texts and thus produce stories with good coherence. On the basis of these results, we can argue that both word recognition and reading comprehension should be included as sub-skills of writing even in models meant to describe the process of immature writing. However, we cannot say, on the basis of this study, what the exact functional relationship is between reading and writing skills among primary school children. Several questions still remain unanswered: do children begin to use reading to revise (e.g., Hayes, 1996), do they learn about text structures by reading (e.g., Juel, 1988) or do reading and writing just share common linguistic sub-processes (e.g., Berninger et al., 1994; Stage & Wagner, 1992)? Our results also lend support to the view of reading as one aspect to be considered in the process of learning how to write. Intervention studies are needed to test whether it is beneficial to teach beginning writers how to use their word recognition skills to check spelling, and also how to utilize reading comprehension skills in evaluating the logic or coherence of their own text (Mäki, Vauras & Vainio, in press). On the basis of the studies of Berninger et al. (1994) and Graham (1990), we expected that poor spelling would lead to difficulties in composition coherence. However, what we found in the second grade was just the opposite. It was observed that the more inaccurate the spelling, the more coherent the composition. It appears that the subjects concentrated on producing logical stories instead of on accurate spelling. The complete composition process burdens the working memory of inexperienced writers to such a degree that they are not able to attend satisfactorily to all the necessary writing sub-processes. As a result, they probably choose some aspects of the writing process they are willing to attend to, depending on the task. This is essential in composition where the monitoring of both spelling and coherence needs simultaneous attention, as was the case in this study. The association between spelling and composition coherence might turn out to be different if separate tasks for spelling and composition were used. For example, in our previous study (Mäki, Voeten, Vauras & Poskiparta, 2001), we found that skilled spelling measured with a separate spelling test in grade 1 predicted good composition coherence in grade 2. Our clinical observations also point in the same direction: writing-disabled children often spell more poorly in a composition task than in a separate spelling task. In order to complete a composition they need to attend to a wide variety of task demands as opposed to a plain spelling task, where they can attend mainly to spelling rules. On the basis of these findings, we suggest that in assessing writing skills it would be beneficial to use both separate spelling test and spelling measure that is taken during a complete composing process. We did not find a connection between spelling and composition coherence in grade 3. One reason for this may be the fact that among Finnish-speaking children, spelling development is very fast. It is common that nearly all third graders (in fact most first and second graders) can spell almost any multi-syllable Finnish word. This is due to the almost perfect grapheme-to-phoneme correspondence of
(JB)SWL II_f13_214-228
9/26/02
10:35 AM
Page 226
Precursors of Functional Literacy
the Finnish language. Children need not learn a ‘spelling vocabulary’ but can strategically use the grapheme-to-phoneme correspondence rules to recode the phonological structure of the words they want to use. Due to this fast spelling development, the variance in spelling measures diminishes rapidly, thus doing a poor job in explaining composition skills in the third grade. As we expected, spelling and composition coherence could be predicted from the results of the previous year, in accordance with earlier findings (e.g., Boland, 1993; Klicpera & Schabmann, 1993). However, the correlations between the successive measurements were not remarkably high. This study with Finnish-speaking subjects thus gives a similar finding to that reported earlier by Juel (1988): the stability of writing skill in the lower grades seems to be weaker than the stability of reading skill in these grades. The quality of texts results from a complex production process based on several subprocesses that undergo substantial development during the primary and intermediate grades. Therefore, predicting this development is a challenging task. Our study lends support to the view that reading skill explains part of the quality of texts produced by young beginning writers. Reading skills are not only related to the level of writing skills but also to the development of writing skills between grades 2 and 3. Thus, the adequate development and the resulting level of reading skills affect children’s ability to write texts of good quality.
Notes * Hanna S. Mäki, Centre for Learning Research; Marinus J. M. Voeten, Department of Educational Sciences; Marja M. S. Vauras, Department of Teacher Education; Pekka Niemi, Department of Psychology. This research was supported by Grants No. 1071265 and No. 4131 from the Council for Social Sciences Research, the Academy of Finland, to the third and fourth authors, and by a grant from the Niilo Mäki Foundation to the first author. We wish to thank the heads, teachers and students of four primary schools in Turku, who unfortunately must remain anonymous. Correspondence concerning this article should be addressed to Hanna Mäki, Centre for Learning Research, University of Turku, FIN-20014 Turku, Finland. Electronic mail may be sent to
[email protected].
References Abbott, R. D. & Berninger, V. W. (1993). Structural equation modelling of relationships among developmental skills and writing skills in primary and intermediate grade writers. Journal of Educational Psychology, 85, 478–508. Berninger, V. W., Cartwright, A. C., Yates, C. M., Swanson, H. L. & Abbott, R. D. (1994). Developmental skills related to writing and reading acquisition in the intermediate grades. Reading and Writing, 6, 161–196. Boland, T. (1993). The importance of being literate: Reading development in primary school and its consequences for the school career in secondary education. European Journal of Psychology of Education, 8, 289–305.
(JB)SWL II_f13_214-228
9/26/02
10:35 AM
Page 227
Relationships between reading and writing skills Browne, M.W. & Cudeck, R. (1993). Alternative ways of assessing model fit. In K. A. Bollen & J. S. Long (Eds.), Testing structural equation models (136–162). Newbury Park: Sage. Cox, B. E., Shanahan, T. & Sulzby, E. (1990). Good and poor elementary readers’ use of cohesion in writing. Reading Research Quarterly, 25, 47–65. Graham, S. (1990). The role of production factors in learning disabled students’ compositions. Journal of Educational Psychology, 82, 781–791. Graham, S., Berninger, V. W., Abbott, R. D., Abbott, S.P. & Whitaker, D. (1997). Role of mechanics in composing of elementary school students: A new methodological approach. Journal of Educational Psychology, 89, 170–182. Hayes, J. (1996). A new framework for understanding cognition and affect in writing. In C. M. Levy & S. Ransdell (Eds.), The science of writing. Theories, methods, individual differences, and applications (1–27). Mahwah, NJ: Erlbaum. Hayes, J., Flower, L., Schriver, K, Stratman, J. & Carey, L. (1987). Cognitive processes in revision. In S. Rosenberg (Ed.), Advances in applied psycholinguistics: Vol. 2 (176–240). Cambridge, UK: Cambridge University Press. Jöreskog, K. & Sörbom, D. (1999). LISREL 8.30 and PRELIS 2.30. Chicago: Scientific Software International. Juel, C. (1988). Learning to read and write: A longitudinal study of 54 children from first through fourth grades. Journal of Educational Psychology, 80, 437–447. Juel, C., Griffith, P. L. & Gough, P. B. (1986). Acquisition of literacy: A longitudinal study of children in first and second grade. Journal of Educational Psychology, 78, 243–255. Just, M. A. & Carpenter, P. A. (1980). A theory of reading: From eye fixations to comprehension. Psychological Review, 87, 329–354. Kellogg, R. T. (1996). A model of working memory in writing. In C. M. Levy & S. Ransdell (Eds.), The science of writing. Theories, methods, individual differences, and applications (57–71). Mahwah, NJ: Erlbaum. Kinnunen, R., Vauras, M. & Niemi, P. (1998). Comprehension monitoring in beginning readers. Scientific Studies of Reading, 2, 353–375. Klicpera, C. & Schabmann, A. (1993). Do German-speaking children have a chance to overcome reading and spelling difficulties? A longitudinal survey from the second until the eighth grade. European Journal of Psychology of Education, 8, 307–323. Mäki, H.S., Vauras, M. M. S. & Vainio, S. (in press). Instructing reflective spelling strategies for elementary school students with severe writing difficulties. Case study. Learning Disability Quarterly. Mäki, H., Voeten, M., Vauras, M. & Poskiparta, E. (2001). Predicting writing skill development with word recognition and preschool readiness skills. Reading and Writing: An Interdisciplinary Journal, 14, 643–672. Poskiparta, E., Niemi, P. & Vauras, M. (1999). Who benefits from training in linguistic awareness in the first grade and what components of it show training effects? Journal of Learning Disabilities, 32, 437–446. Rijlaarsdam, G. & Van den Bergh, H. (1996). The dynamics of composing — An agenda for research into an interactive compensatory model of writing: Many questions, some answers. In C. M. Levy & S. Ransdell (Eds.), The science of writing. Theories, methods, individual differences, and applications (107–125). Mahwah, NJ: Erlbaum. Scardamalia, M. & Bereiter, C. (1987). Knowledge telling and knowledge transforming in written composition. In S. Rosenberg (Ed.), Advances in applied psycholinguistics: Vol. 2 (143–175). Cambridge, UK: Cambridge University Press. Scinto, L. (1986). Written language and psychological development. Orlando: Academic Press. Shankweiler, D., Lundquist, E., Dreyer, L. G. & Dickinson, C. C. (1996). Reading and spelling difficulties in high school students: Causes and consequences. Reading and Writing, 8, 267–294. Stage, S. A. & Wagner, R. K. (1992). Development of young children’s phonological and
(JB)SWL II_f13_214-228
9/26/02
10:35 AM
Page 228
Precursors of Functional Literacy orthographic knowledge as revealed by their spellings. Developmental Psychology, 28, 287–296. Vauras, M., Hyönä, J. & Niemi, P. (1992). Comprehending coherent and incoherent texts: evidence from eye movement patterns and recall performance. Journal of Research in Reading, 15, 39–54. Zutell, J. & Rasinski, T. (1989). Reading and spelling connections in third and fifth grade students. Reading Psychology, 10, 137–155. Address University of Turku Center for Learning Research FIN-20014 Turku, Finland
[email protected]
(JB)SWL II_f14_229-245
9/26/02
10:36 AM
Page 229
Task-related factors in reading efficiency of dyslexic children Aryan van der Leij*, Victor van Daal** and Peter de Jong* * University of Amsterdam ** University of Bangor
It is obvious that children with developmental dyslexia perform poorly when they have to read and spell words, because that is the reason why they are called dyslexics in the first place. However, even within such a task-specific approach, some debate is going on about the question of whether phonological recoding is the core deficit, most prominently indicated by problems with the reading of unfamiliar words, or a deficit in the automatization of the reading skill, which is also apparent when they read familiar words. In addition, it is still unclear to what extent an automatization deficiency is a symptom of a more fundamental deficit. Impairments in phonological processing skills have successfully been promoted as task-related correlates, but the question is still open as to which aspect of phonological processing is relatively the most important: phonological awareness, verbal short-term memory, or naming speed. Taking a broader approach, other authors have argued that children with dyslexia suffer from much more than just language and memory related problems. In their view, dyslexia may be a sign of a disability to automatize any skill, associated with deficits in the phonological, visual or motor domain. In this chapter we will describe three positions taken in this debate. First, the task-specific position that restricts dyslexia to impairments in reading only. Second, the task-related position that extends that view to deficits in phonological processing. Third, the task-independent position based on the assumption that a wider range of processes must be affected in dyslexics, including processes outside the area of reading and phonological processes. After a short review of studies illustrating the different positions, a study will be presented and discussed that was aimed to assess the viability of the three positions in the debate.
(JB)SWL II_f14_229-245
9/26/02
10:36 AM
Page 230
Precursors of Functional Literacy
Position 1: A task-specific deficit The finding that children with developmental dyslexia perform poorly at the reading of unfamiliar words, in particular nonwords, has been replicated many times, and has been interpreted as a ‘phonological recoding deficit’ (Rack, Snowling & Olson, 1992). The idea that a deficit in phonological recoding forms a core feature of developmental dyslexia is strongly supported by the fact that this finding is independent of (ir) regularity of the orthography of the language. For example, it has been found in English, German-speaking Austrian, and Dutch samples (by Rack, Snowling & Olson, 1992; Wimmer, 1993; Yap & Van der Leij, 1993, respectively). While there seems to be a consensus with regard to the problems that developmentally dyslexic individuals encounter when reading unfamiliar words, there is some debate on the extent to which the reading of familiar words is affected. According to dual route models, a nonword reading deficit does not have to be accompanied by problems with the reading of words. In this view, it is possible that one route is affected while the other is not (Castles & Coltheart, 1993). Indeed, many persons with developmental (‘genetic’) dyslexia seem to be quite able to read and write familiar words. Possibly, their ‘direct route’ of recognising familiar words is unaffected. However, as we have argued elsewhere (Van der Leij & Van Daal, 1999a), it seems unlikely that the ‘direct route’ is unaffected. The main point is that the acquisition of satisfactory accuracy in reading familiar words does not imply that the orthographic representations of these words in memory are of good (‘normal’) quality, and, thus, can be retrieved without much effort. For example, in contrast to normal readers, less skilled pupils show smaller familiarization effects of repeated practice (Hogaboam & Perfetti, 1978) and are at a disadvantage in acquiring word-specific knowledge (Reitsma, 1983). As a consequence, pupils with dyslexia need many more trials than normal controls to reach a set accuracy criterion. Moreover, even after many repetitions of the same materials, their speed of responding still is much slower than that of normal controls (Van der Leij & Van Daal, 1989). In fact, while their accuracy eventually may be at the chronological age level, their speed lags substantially behind the level of reading age controls. It is doubtful that the representations they have of the familiar words are qualitatively at the level of normal age controls. Instead of a ‘phonological recoding deficit’ that is restricted to the reading of nonwords, an ‘automatic decoding deficit’ that essentially affects the reading of all kinds of words, may be the core deficit (Yap & Van der Leij, 1993). According to this view, dyslexics have poor phonological recoding but also problems with rapid, automatic processing. Experience and the use of attention may mask that impairment when easy and familiar words are read (indicated by accurate, but relatively slow responses that are typical for their reading age). However, increasing task demands such as ‘flashed’ presentations will elicit differences with normal development (indicated by poorer performance in comparison to younger normal pupils with the same reading age). In a later study (Van der Leij & Van Daal, 1999b) we were able to demonstrate
(JB)SWL II_f14_229-245
9/26/02
10:36 AM
Page 231
Task-related factors in reading efficiency
that shortening of the presentation time is not essential to elicit the core deficit. It appears also when the increasing task demands are defined as lowering the word frequency, increasing the word length, the orthographic or the phonological complexity. In this view, such task demands can be positioned on a continuum from relatively easy to relatively difficult. While the dyslexic reader will, even in the simplest conditions, exhibit a speed below chronological age level (but at reading age level), indicating that attention is used and automatization is not mastered at that level, moving the task demands in the direction of increasing difficulty will deteriorate his performance more than is the case in normal reading, resulting in a performance below reading age.
Position 2: A task-related deficit While it is obvious that a task-specific deficit is the core deficit of developmental dyslexia, there is debate to what extent such a deficiency is a symptom of a more fundamental deficit. Skills that rely heavily on phonological processes have been proposed as a task-related deficit of dyslexia (Snowling, 1987; Stanovich, 1988). Phonological processing is very important in learning how to read and spell, because mastery of the alphabetic principle is based on learning the graphemephoneme correspondences. In particular, a ‘phonological processing deficit’ may explain poor recoding processes in reading unfamiliar words. However, phonological processes are by no means a cognitive entity. Wagner and Torgesen (1987) made a useful distinction in three major types of phonological abilities: phonological awareness, phonological coding in short-term memory, and retrieval of phonological codes from long-term memory (rapid naming). All three have been studied as possible correlates of developmental dyslexia. Stanovich and Siegel (1994) found evidence for the position that dyslexia is to be viewed as a specific deficit in phonological awareness, because (1) differences between dyslexics and both chronological and reading age controls pertain to this kind of phonological abilities and (2) differences between dyslexics and children with general learning disabilities (the so-called ‘garden variety readers’) concern general language and memory skills, but not reading and phonological abilities. As a consequence, they claimed that reading problems are independent of general intelligence. With regard to phonological abilities, reviews of studies indicate that poor phonological awareness is associated with dyslexia, from the start of the process of learning how to read, through later stages when the reading disability has become manifest, right up to adulthood (Bruck, 1998). Impairments in verbal short-term memory are also often found among children with reading problems. Current evidence suggests that verbal short-term memory consists of a passive phonological store in which information decays rapidly, and a rehearsal mechanism to refresh decaying information (e.g., Baddeley, Gathercole & Papagno, 1998). Verbal memory span tasks, such as digit span, capture both components of verbal short-term memory. In many studies children
(JB)SWL II_f14_229-245
9/26/02
10:36 AM
Page 232
Precursors of Functional Literacy
with reading problems have been found to show deficits on memory span tasks (e.g., Brady, 1991; Elbro, 1996; De Jong, 1998). However, there is also some evidence that children with reading problems perform worse on tasks, such as nonword repetition, which tap only the phonological store (e.g., Stone & Brady, 1995; Van Bon & Van der Pijl, 1997). It has been suggested that dyslexics are also defective in verbal information processes that involve rapid processing. For instance, they exhibit problems in rapid performance such as automatized naming of successive linguistic stimuli (letters, digits) and non-linguistic stimuli such as colours (Denckla & Ruddel, 1976), more recently rephrased as a ‘naming speed deficit’ (Bowers & Swanson, 1991). As there is now evidence to show that naming speed does only moderately correlate with other phonological skills (De Jong & Van der Leij, 1999; Wolf, Bowers & Biddle, 2000), it may be assumed that rapid automatized naming constitutes a second independent characteristic of reading-impaired children. Based on this independency, Wolf and Bowers (1999) have proposed the ‘double-deficit’ hypothesis. While some children may suffer from difficulties in manipulating the sound structure of the language, others may be characterised by a naming speed deficit. The most severely affected children have both (‘double’) deficits.
Position 3: A task-independent deficit As a theoretical framework for a more general deficit, Nicolson and Fawcett (1990) have proposed that children with dyslexia fail to fully automatize any skill. They developed the Dyslexic Automatization Deficit (DAD) hypothesis. Dyslexics who have not been able to acquire fluent reading skills, also have problems with various skills outside reading and spelling, including phonological, speed, memory and motor skills. Analysing data on a variety of skills, they claimed that “. . . there is no support . . . for any of the theories that attempt to tie dyslexia to one specific modality or type of process” (Nicolson & Fawcett, 1994: 228). To test the automatization deficit hypothesis, they compared dyslexic readers with normal controls in conditions that varied in task demands with therefore the possibility of using attention resources to compensate for automatization impairments. While dyslexic performance was at the level of chronological age controls in a simple baseline condition, it tended to deteriorate more with increasing complexity of the task conditions. For example, differences in speed appeared in a selective choice reaction task, but not in a simple reaction task. To explain the difference in performance in the simple and complex conditions, Nicolson and Fawcett (1990) complemented their DAD hypothesis with the Conscious Compensation (CC) hypothesis, stating that dyslexics in many instances (without time constraints or complex or dual task conditions) are able to mask their automatization deficit by using a considerable part of their attention resources. Challenging as the DAD/CC hypotheses may be, attempts to replicate the
(JB)SWL II_f14_229-245
9/26/02
10:36 AM
Page 233
Task-related factors in reading efficiency
findings of Nicolson & Fawcett (1990, 1994) have not produced straightforward support. Yap and Van der Leij (1994) have reported partial support for the automatization deficit, i.e., in only one of their two dual task-conditions. However, Wimmer, Mayringer, and Landerl (1998), and Stringer and Stanovich (1998) were not able to find any support. Thus, although a general automatization deficit hypothesis is attractive to many clinical workers who are faced with dyslexic persons claiming that they also suffer from problems outside the domain of reading and spelling, the validity of the DAD/CC hypothesis is still to be assessed. It should be noted that a task-independent position is also adopted by other researchers, who claim that other processes may be involved than phonological processes. In particular, researchers who concentrate on visual processes (e.g., Lovegrove, 1994; Stein, 1994) have produced interesting evidence. Essentially, poor sensitivity in the visual transient system involved in the perception of motion has been indicated as a correlate of developmental dyslexia. Because in our study no instruments were included to measure these aspects of visual processing, we will not go into further detail.
Hypotheses In our study, we investigated which position would be supported. It is important to state that the positions are not mutually exclusive but incrementally inclusive: position 3 includes position 1 and 2, and position 2 includes position 1. Therefore, the main question is how developmental dyslexia could be placed on a continuum from specific to general in terms of deficits in lower-level cognitive processing. The review of the theoretical positions supports the choice for lower-level processing in order to investigate the underlying deficit. Taking a series of task-specific, task-related, and task-independent variables into account, we reasoned that position 1 would be supported when only task specific variables make the difference. Position 2 would be plausible when taskrelated variables join the task-specific variables but no variables of position 3. Position 3 would have to include a significant variable outside the stricter area of position 2.
Method The following tasks were selected from the data of a larger study to cover the three positions (Van Daal & Van der Leij, 1999). Position 1: nonword reading, English word recognition, spelling. Position 2: phonological awareness, (verbal) working memory, rapid naming. Position 3: perceptual speed (single and dual task).
(JB)SWL II_f14_229-245
9/26/02
10:36 AM
Page 234
Precursors of Functional Literacy
In addition, tasks on reading comprehension and speeded arithmetic were administered to control for selection bias (in particular the exclusion of general learning disabilities). We reasoned that, to detect the characteristics of developmental dyslexia, four design conditions should be fulfilled. First, the focus was on relatively ‘pure’ dyslexia. Taking word recognition and listening comprehension as the classification tasks, participants would have to be poor in word recognition and significantly better in listening comprehension to qualify as dyslexic. As a consequence, pupils showing the opposite skill pattern (low comprehension abilities and average word recognition: so called hyperlexics) and pupils with low comprehension and poor reading abilities (‘garden variety’) were excluded from our target group. Second, the comparison included two control groups. A chronological age group (CA) that matched the dyslexics in age and comprehension skills, and a reading age control group (RA) that matched in reading level and had, at their age level, average comprehension abilities. The CA-group would allow us to conclude whether any difference could be called a delay, and the RA-group made it possible to indicate whether any difference matched the backwardness in reading. Third, we reasoned that children of about 13 would suit our purposes best. At that age, the selection of pupils based on lags in achievement could be made with no doubt about the severity of backwardness because reading problems are increasingly stable over time (Smart, Sanson & Prior, 1996). Furthermore, because they enter secondary school at 12 in the Netherlands and learn English as a second language, we would be able to include an English word recognition test to investigate transfer to a second language with a far less shallow orthography than Dutch. Fourth, to avoid selection bias as a result of variance in the clinical diagnosis of dyslexia, we decided to select the groups using our own tests and criteria. All pupils from a complete school district, who were reported as having learning disabilities, were screened and classified by us.
Participants With the exception of the RA control group who were aged ten, the average age of the participants was about 13 years. From the school district (population of approximately 60,000), 75 learning disabled pupils were used for further selection. Using the criterion of one standard deviation or more below CA mean on the classification variables (word recognition and listening comprehension), we were able to exclude 13 with general learning disabilities (low on both variables) and 18 with a hyperlexic profile, leaving 44 dyslexics (poor word recognition, average listening comprehension). Two control groups were formed: one consisted of 33 CA controls from secondary schools; the second control group was composed of ten children from a primary school who read at the same level as the dyslexics but were 2 years younger.
(JB)SWL II_f14_229-245
9/26/02
10:36 AM
Page 235
Task-related factors in reading efficiency
Procedure After selection, all participants were administered the tests described below, either in groups or individually, but always double-blind, because classifications into different types of readers were carried out afterwards, when all data had been collected.
Tests Classification tests Word recognition was measured by means of the Eén-Minuut Test (One-Minute Test) developed by Brus and Voeten (1973). The participant was required to read aloud as quickly and as accurately as he could real words (nouns, verbs, etc.) from a card with 116 words of increasing difficulty arranged in four columns. The raw score was the number of words correctly read within one minute. Listening comprehension was assessed with the experimental form of the Listening subtest of the BELL 1996 (Van den Bos, 1996). The participant selected from four pictures, the one that fitted a spoken sentence best. To perform well the participant had to know word meanings, and understand syntactic and semantic relations. The test had 34 items of increasing difficulty and was recently standardised in a large-scale survey (n = 1700) with first graders in secondary education (Van Daal & Van der Leij, in preparation). Two parallel forms were used, the A and B form. All participants were classified on the basis of the standard scores (with a mean of 10 and a standard deviation of 3) obtained on the word recognition test and the listening comprehension test, as either a dyslexic reader, a garden-variety reader, or a hyperlexic reader (the latter two subgroups were excluded from the sample in this study), or a normal reader. A dyslexic reader was defined as scoring 7 or less on the decoding test and over 7 on the listening comprehension test, with a discrepancy of at least 3 points between the two measures. All participants scoring just over 7, but with a listening comprehension score of at least 3 points more were also considered dyslexic. In the whole sample of 108 only 6 such participants could be found. Participants with a listening comprehension score of less than 7 and a decoding score at least 3 points more were treated as hyperlexic readers. Here we had only 2 participants who scored just over 7 on the listening comprehension test, but had a far better decoding score. The group who scored below 7 on both the decoding and the listening comprehension tests was excluded. Pupils in the control group had no scores below 1 s.d. on the classification tests, and no extreme differences between the two scores.
(JB)SWL II_f14_229-245
9/26/02
10:36 AM
Page 236
Precursors of Functional Literacy
Task-specific tests The first test used to tap reading and language related skills was a nonword reading test, the Klepel (Van den Bos et al., 1994). This test was constructed by changing vowels or consonants in words of the Eén-Minuut Test under the restriction that the pronunciation rules of Dutch were not violated. The score was the number of correctly read nonwords within two minutes. As the pupils in the first year of secondary education also received formal instruction in reading English, an English word recognition test was administered. This test, the English version of the Eén-Minuut Test, consisted of words of increasing difficulty that occurred in commonly used teaching methods in the Netherlands (Van Daal, in preparation). The participant’s task was to read aloud the words as quickly and as accurately as possible. The score was the number of words read correctly within one minute. All participants were also given a standardised spelling test consisting of 135 words of increasing difficulty (PI-diktee, Geelhoed, 1994). The score on this test was the number of correctly spelled words.
Task-related tests To measure phonological awareness, two tests were administered. In the Auditory Analysis task the computer presented the spoken form of a nonword (from the parallel version of the Klepel mentioned before) in digitised speech. All speech used was uttered at a rate of one syllable per second by a professional speech trainer, recorded in a studio on DAT tape and digitised on the computer’s hard disk at a sampling rate of 44 kHz. It was the participant’s task to say as quickly as he could the smallest sounds of the word in correct order. The experimenters were trained to press the space bar as soon as the participant started speaking and in this way the computer recorded the latencies of the responses. The experimenter assessed accuracy in all of these computer-administered tasks immediately after pressing the space bar. An error, if made, was noted. When the response was correct, the next item was presented. Thus, both the accuracy and the latencies of the responses could be analysed. For the Sound Blending task the computer presented the isolated sounds of nonwords and it was the participant’s task to say as quickly as he could the whole word. Both tests were administered with the help of an Apple Macintosh Plus computer on which a programme written in AuthorWare was run. In this chapter, only accuracy data will be used. To measure the quality of the phonological store, a Nonword Repetition Test (e.g., Gathercole & Baddeley, 1989) was administered using a computer. The participants were instructed to respond as quickly as they could when the computer had said the nonword. The items increase in number of syllables, from one, ‘KES’, to four, ‘WAPELBROEGER’. To measure verbal memory span, Digit Span was administered in the class-
(JB)SWL II_f14_229-245
9/26/02
10:36 AM
Page 237
Task-related factors in reading efficiency
rooms. Two revisions of the WISC-R version were made: only forward items were used, because cheating on the backward items cannot be checked, and three items per length were used to increase the reliability of the test. The total number of items correctly recalled was scored. To measure naming speed, a Rapid Naming test was selected, containing 5 subtests, each of which contained 50 items printed on a card. The participants were required to name as quickly as they could the items on each card, if possible without hesitation or error. Mean total naming times per card were computed for (1) alpha-numeric stimuli (digits (0–9), capital letters (A–Z), and a mix of digits and capital letters) and (2) Objects and colours (one card with 5 different familiar objects, and one card with 5 different colours).
Task-independent tests To measure perceptual speed, a Simple Reaction Time test (SRT) was administered using a computer. The participants were asked to press the space bar as soon as possible after they had heard a tone. A visual signal was presented to warn the participant that the next stimulus was about to come, randomly between 0.5 sec. and 1.5 sec. after the visual signal. Mean reaction times were computed. To measure the influence of increasing task demands, a Selective Choice Reaction test (SCR) was added to the SRT as a dual task condition. SCR was also administered using a computer; the participants were requested to press the space bar as quickly as possible, if a high tone was presented, whereas they had to refrain from pressing, if a low tone was presented. For this task, reaction times for hits and false alarms were recorded, as were numbers of hits, misses, correct rejections and false alarms. In this chapter, only the results for hits will be presented.
Control tasks Reading comprehension was measured by means of the Reading subtest of the BELL 96 (Van den Bos, 1996). Participants first read a sentence, then turned the page and selected the picture that best fitted the previously read sentence. The test consisted of 34 items of increasing difficulty and was constructed in such a way that it could be considered a parallel form of the listening comprehension test, except for the mode of processing (reading vs. listening). Here also two forms were used, the A and B form, respectively, in the current study. The score was the number of correct answers. Speeded arithmetic was measured by means of the Tempo Test Rekenen (Speeded Arithmetic Test, De Vos, 1992), which consisted of five subtests. The participants were instructed to solve as many simple sums as they could (paper and pencil test) within one minute. The subtests consisted of addition, subtraction, multiplication, division and mixed sums. The score was the number of correct answers.
(JB)SWL II_f14_229-245
9/26/02
10:36 AM
Page 238
Precursors of Functional Literacy Table 1. Classification variables
word recognition reading grade listening comprehension
Dys. n = 44
RA n = 10
CA n = 33
F-value
58.0a (9.3) start 4
68.5b (7.1) mid 4
79.0c (10.7) mid 6
39.11***
27.0a (2.8)
18.4b (2.4)
25.8a (2.8)
60.67***
Notes – The scores are no. correct. – Standard deviations between brackets. – *** p < .001 – a versus b versus c: significant difference in post-hoc comparison (p < .05)
Analysis The variables and data used in this chapter were selected from a larger study (for a full description see Van Daal & Van der Leij, 1999). To test mean score differences among the groups, analyses of variance (ANOVAS) were conducted, followed by pairwise comparisons between all pairs of groups. A Bonferroni correction was used to reduce type I errors.
Results Classification of the participants Table 1 shows the results of classification. Differences between groups were significant on word recognition. Dyslexics and RA controls read less words than the CA controls. However, dyslexics also performed below RA level. The reading grade score indicates that RA controls were average readers at their age level. On the listening comprehension test differences were also found. The CA controls and the dyslexics were better at listening comprehension than the RA controls.
Task-specific variables The results of the task-specific tests are shown in the upper part of Table 2. As was predicted by the phonological recoding deficit hypothesis, nonword reading presented a large discrepancy between dyslexics and the other groups (the pairwise differences were both significant). There was no difference between CA and RA controls (who actually did quite well on this test). Additionally, also English
(JB)SWL II_f14_229-245
9/26/02
10:36 AM
Page 239
Task-related factors in reading efficiency
word recognition and spelling of the dyslexics were significantly below CA level. In spelling, there was no difference between dyslexics and RA controls, but there was between RA and CA.
Task-related variables The next part of Table 2 shows the results on task-related variables, all within the domain of phonological processing. Although the ANOVAS of segmenting and nonword repetition reached significance, none of the pairwise comparisons met the traditional criterion of p < .05 (surprisingly, RA was marginally better — p < .07 — than dyslexics and CA in segmenting). Dyslexics differed from CA controls in the speed of naming digits and letters, but not in the rapid naming of objects and colours. RA and CA did not differ on either task.
Task-independent variables On the simple reaction task, CA controls were faster than RA controls, whereas the difference between dyslexics and RA was marginally significant (p = .06). On the choice reaction task no differences were found (latency for hits). There was no interaction effect between single and dual task conditions.
Control tasks On reading comprehension dyslexics and CA controls outperformed RA controls; on speeded arithmetic only the difference between CA and RA was significant.
Discussion Obviously, the results of our study supported position 1. The dyslexics showed a large ‘phonological recoding deficit’ (Rack, Snowling & Olson, 1992), indicated by their nonword reading performance below CA and RA level. In addition, their reading of (familiar) English words and their spelling of a variety of Dutch words were affected. Because both tasks contained irregular words and thus tapped orthographic knowledge, it may be assumed that not only phonological recoding is impaired, but reading and spelling in general, independent of word frequency, phonological or orthographic complexity, language (Dutch or English), or response mode (naming or writing). We suggest that dyslexics suffer from a task-specific ‘automatic decoding deficit’, that includes the ‘phonological recoding deficit’ that relates to unfamiliar words, but also affects word reading and spelling across the whole continuum of task demands, running from simple to complex, in accordance with the results of a study that focussed on the task-specific deficit (Van der Leij & Van Daal, 1999b).
(JB)SWL II_f14_229-245
9/26/02
10:36 AM
Page 240
Precursors of Functional Literacy Table 2. Dependent variables
task-specific – nonword reading – English word recognition – spelling task-related phonological awareness – segmenting % – blending % working memory – nonword repetition % – digit span rapid naming – letters/digits secs. – objects/colours secs. task-independent speed of processing – simple RT secs. – choice RT (hits) secs. control tasks – reading comprehension – speeded arithmetic
DYS
CA
RA
F-value
35.6a (11.0) 32.6a (8.4) 106.3a (13.8)
68.5b (17.1) 55.6b (17.6) 122.1b (8.0)
64.0b (9.2) n.a.
52.11***
98.9a (10.1)
20.21***
62 (25) 86 (16)
61 (29) 86 (12)
89 (11) 86 (6)
3.59* 0.96
96 (6) 4.8 (.7)
96 (6) 5.0 (.7)
92 (7) 4.5 (.5)
3.61*
27.8a (4.5) 42.0 (7.7)
23.2b (3.9) 39.4 (9.0)
26.0 (3.7) 44.5 (5.1)
12.62***
.76 (.22) 1.9 (.5) 24.5a (3.8) 111.8 (25.3)
.68a (.20) 1.6 (.4) 26.0a (3.8) 123.8a (28.8)
1.00b (.30) 1.9 (.2) 19.0b (4.2) 96.3b (18.2)
Notes – The scores are no. correct, except when indicated otherwise. – Standard deviations between brackets. – * p < .05, ** p < .01, *** p < .001 – a versus b: significant difference in post-hoc comparison (p < .05)
1.36
2.05
4.92** 2.33
9.08*** 2.86*
(JB)SWL II_f14_229-245
9/26/02
10:36 AM
Page 241
Task-related factors in reading efficiency
Position 2 received only partial support, as there were no differences in phonological awareness and verbal short-term memory. There was, however, an indication for a ‘naming speed deficit’ for letters and digits (Bowers & Swanson, 1991). One could argue that the findings mainly confirm position 1, and not a more general ‘naming speed deficit’, because only the speed of retrieval of symbol names was impaired and not of objects and colours. It may be assumed that fast retrieval of letter names correlates with phonological recoding because phonological recoding involves the fast retrieval of letter sounds. The lack of differences in phonological awareness, indicated that a ‘double-deficit’ in phonological awareness and rapid naming (Wolf & Bowers, 1999) was not supported by our data. The study did not support position 3: dyslexics did not perform worse than CA controls on the task-independent variables, and there was no interaction between groups (DYS; CA) and single/dual task conditions. Three questions have to be discussed to assess the validity of these interpretations. First, were the tasks good examples of the variables? There is little doubt that the reliability of all tasks was satisfactory (Van Daal & Van der Leij, 1999), but what about validity? A way to assess the differentiating quality of the tasks is to inspect the differences between CA and RA in more detail. For a skill that is still in development at ages 10 to 12 and upward, a difference CA > RA suggests that the instrument is valid. Apart from the classification tasks (word recognition and listening comprehension) and the control tasks (reading comprehension and speeded arithmetic), two tasks met this criterion: the task-specific spelling task, and the task-independent simple reaction task. This means that these instruments were fit to elicit developmental delays. Since the spelling task indicated a significant difference between CA and dyslexics in the predicted direction, this result can be highly valued. However, the simple reaction task did not and nor did the group by task interaction, if the dual task condition (choice reaction time) is taken into account. Our study failed to replicate the reaction time findings of Nicolson and Fawcett (1994) and, therefore, did not support their Dyslexia Automatization Deficit hypothesis. As was noted in the introduction, in a number of studies the replication of the findings of the DAD study by Nicolson and Fawcett was mainly unsuccessful (Yap & Van der Leij, 1994; Wimmer et al., 1998; Stringer & Stanovich, 1998). The results of the present study were in the same direction. However, it should be noted that the dyslexic performance deteriorated more than CA and RA in the dual task condition. Possibly, to detect a general automatization difference at the age of 12 and upwards, task demands in the dual task condition should increase more than was the case in our tasks. Second, did the variables that did not show a significant CA > RA difference, allow for additional interpretations? There seems to be little doubt that the results of the task-specific nonword reading task and task-related rapid letter/digit naming task mean something. They fit well in the hypotheses of the ‘phonological recoding’ and ‘rapid naming’ deficit. Besides, the fact that CA and RA did not differ on the nonword reading task may well be interpretable. One possibility is that the RA group was relatively good in reading (see the difference with dyslexics
(JB)SWL II_f14_229-245
9/26/02
10:36 AM
Page 242
Precursors of Functional Literacy
on word recognition) and also in subcomponents of phonological processing. Their sound blending ability was at CA level and their segmenting ability was even better — albeit statistically not significant. Furthermore, their nonword repetition and digit span were not significantly below CA level. These observations may indicate that the tasks used to measure these abilities were too easy or the RA controls were too good. However, with respect to the first possibility (for the second point see below), neither of the tasks for phonological processing indicated a ceiling effect. In addition, there was substantial variation in performance in all groups. Possibly, another interpretation deserves consideration. In a study of Landerl, Wimmer, and Frith (1997), 12-year old dyslexic pupils in Austria and the United Kingdom were compared on equivalent tasks in their own language (German and English, respectively). In that study, the authors concluded that dyslexics from both countries suffered from the same phonological processing deficit. The task they used required children to exchange the consonant onsets of two words (boat-fish becomes foat-bish) and evidently was more complex than the tasks we used in the present study. However, in a re-analysis of their data, using a more liberal way of scoring the answers, Landerl and Wimmer (2000) did not replicate these findings and concluded that a deficit in phonological awareness was not the core problem for dyslexics, but only evident in the early stages of reading acquisition. This conclusion is in accordance with our findings in a study on reading development: after the first phases of reading acquisition (grade 1), the influence of phonological abilities in reading acquisition seems to level off (De Jong & Van der Leij, 1999, 2002). At the age of 12, dyslexics may have overcome these impairments in relatively ‘pure’ phonological processing, but still suffer from deficits in task-specific learning mechanisms (De Jong & Van der Leij, in press). In turn, these learning mechanisms are likely to include phonological processing; in particular phonological processing that is associated with the acquisition of orthographic knowledge at the level of letters, letter clusters, and words, and the fast retrieval of the corresponding sounds. Rapid naming of letters and digits may be a good indicator for this latter process. The third question is whether the groups were selected well enough to allow for the conclusions. The control tasks confirmed that the classification was correct. Dyslexics were not significantly below CA level on reading comprehension and speeded arithmetic indicating both their normal capabilities in the verbal intelligence domain and the specificity of their reading deficit. On the other hand, RA controls performed significantly below CA level on both tasks. In addition, RA controls were worse than CA in skills such as listening comprehension and spelling. Although they were better than the dyslexics in word recognition, the differences with the RA controls indicated their performance reflected the interacting influence of younger age, and less experience and practice (although the number of RA controls should have been larger and better matched to the dyslexic group to support that assumption). Therefore, the fact that other variables did not elicit differences may be well ascribed to decreasing differences between age groups. In addition, it seems reasonable to suggest that dyslexics ‘catch up’ with
(JB)SWL II_f14_229-245
9/26/02
10:36 AM
Page 243
Task-related factors in reading efficiency
normal development in a lot of ways. Nevertheless, the subject selection and classification needs careful consideration because selection criteria may be decisive for confirmation or rejection of hypotheses. Because our dyslexic sample was reasonably large and was not selected clinically but by using our own empirical criteria, we feel quite confident that the selection was not biased. However, although pupils with general learning disabilities and pupils with a ‘hyperlexic’ profile were excluded, we did not control for a possible combination with a deficit in arithmetic. The dyslexic group as a whole did not perform below CA level in speeded arithmetic, but a subgroup might have performed poorly on that task. The performance of that subgroup on other tasks could have contributed to our results. Recently, we re-analysed all our data (of all our subjects) in a study on single, double and general learning disabilities, using speeded arithmetic as a classification variable next to word recognition and listening comprehension. The results indicated that exclusion of the (15) participants who combined good listening comprehension with poor word recognition and poor arithmetic, did not affect the differences between the dyslexics and the CA controls (Van der Leij & Van Daal, 2001). In conclusion, the task-specific position received strong support with regard to the underlying lower-level cognitive correlates, with only rapid naming of letters and digits as an additional task-related indicator. It should be noted that these conclusions were based on the results of a sample of 12-year-old dyslexics who learned to read and write in a relatively shallow orthography such as Dutch.
References Baddeley, A., Gathercole, S. & Papagno, C. (1998). The phonological loop as a languagelearning device. Psychological Review, 105, 158–173. Bowers, P. G. & Swanson, L. B. (1991). Naming speed deficits in reading disability: multiple measures of single processes. Journal of Experimental Child Psychology, 51, 195–219. Brady, S. A. (1991). The role of working memory in reading disability. In S. A. Brady & D. P. Shankweiler (Eds.), Phonological processes in literacy (129–151). Hillsdale, NJ: Erlbaum. Bruck, M. (1998). Outcomes of adults with childhood histories of dyslexia. In C. Hulme & R. Malatesha Joshi (Eds.), Reading and Spelling: Development and Disorders (179–200). London: Erlbaum. Brus, B.Th. & Voeten, M. J. M. (1973). Eén-minuut-test [One-minute-test]. Nijmegen (Netherlands): Berkhout. Castles, A. & Coltheart, M. (1993). Varieties of developmental dyslexia. Cognition, 47, 149–180. De Jong, P. F. (1998). Working memory deficits of reading disabled children. Journal of Experimental Child Psychology, 70, 75–96. De Jong, P. F. & Van der Leij, A. (1999). Specific contributions of phonological abilities to early reading acquisition: Results from a Dutch latent variable longitudinal study. Journal of Educational Psychology, 91, 450–476. De Jong, P. F. & Van der Leij, A. (2002). Effects of phonological abilities and linguistic comprehension on the development of reading. Scientific Studies of Reading, 6, 51–77.
(JB)SWL II_f14_229-245
9/26/02
10:36 AM
Page 244
Precursors of Functional Literacy De Jong, P. F. & Van der Leij, A. (in press). Developmental changes in the manifestation of a phonological deficit in dyslexic children learning to read a regular orthography. Journal of Educational Psychology. Denckla, M. & Rudel, R. (1976). Rapid ‘automatized’ naming R.A.N.: Dyslexia differentiated from other learning disabilities. Neuropsychologia, 14, 471–479. De Vos, T. (1992). Tempo test rekenen [Speeded calculation test]. Nijmegen (Netherlands): Berkhout. Elbro, C. (1996). Early linguistic abilities and reading development: A review and a hypothesis. Reading and Writing: An Interdisciplinary Journal, 8, 1–33. Gathercole, S. E. & Baddeley, A. D. (1989). Evaluation of the role of phonological STM in the development of vocabulary in children: A longitudinal study. Journal of Memory and Language, 28, 200–213. Geelhoed, J. (1994). Het PI-diktee. Handleiding bij de uitgebreide versie [The PI-Spelling Test. Manual of the extended version]. Amsterdam: Paedological Institute. Hogaboam, T. W. & Perfetti, C. A. (1978). Reading skill and the role of verbal experience in decoding. Journal of Educational Psychology, 70, 5, 717–729. Landerl, K. & Wimmer, H. (2000). Deficits in phoneme segmentation are not the core problem of dyslexia: Evidence from German and English children. Applied Psycholinguistics, 21, 243–262. Landerl, K., Wimmer, H. & Frith, U. (1997). The impact of orthographic consistency on dyslexia: A German-English comparison. Cognition, 63, 315–334. Lovegrove, W. (1994). Visual deficits in dyslexia: Evidence and implications. In A. Fawcett, & R. Nicolson (Eds.), Dyslexia in Children (113–136). London: Harvester Wheatsheaf. Nicolson, R. I., & Fawcett, A. J. (1990). Automaticity: A new framework for dyslexia research? Cognition, 30, 1–33. Nicolson, R. I. & Fawcett, A. J. (1994). Comparison of deficit severity across skills: Towards a taxonomy for dyslexia. In A. J. Fawcett and R. I. Nicolson (Eds.), Dyslexia in children (215–239). New York: Harwester Wheatsheaf. Rack, J. P., Snowling, M. J. & Olson, R. K. (1992). The nonword reading deficit in developmental dyslexia: A review. Reading Research Quarterly, 27, 28–53. Reitsma, P. (1983). Word-specific knowledge in beginning reading. Journal of Research in Reading, 6, 41–56. Smart, D., Sanson, A. & Prior, M. (1996). Connections between reading disability and behavior problems: Testing temporal and causal hypotheses. Journal of Abnormal Child Psychology, 24, 363–383. Snowling, M. J. (1987). Dyslexia: A cognitive developmental perspective. Oxford, England: Basil Blackwell Ltd. Stanovich, K. E. (1988). Explaining the differences between the dyslexic and the gardenvariety poor reader: The phonological-core variable-difference model. Journal of Learning Disabilities, 21, 590–604. Stanovich, K. E. & Siegel, L. S. (1994). Phenotypic performance profile of children with reading disabilities: A regression-based test of the phonological-core variable-difference model. Journal of Educational Psychology, 56, 24–53. Stein, J. F. (1994). A visual deficit in dyslexics? In A. Fawcett & R. Nicolson (Eds.), Dyslexia in Children (137–156). London: Harvester Wheatsheaf. Stone, B. H. & Brady, S. A. (1995). Evidence for phonological processing deficits in lessskilled readers. Annals of Dyslexia, 45, 51–78. Stringer, R., & Stanovich, K. E. (1998). On the possibility of cerebellar involovement in reading disability. Paper presented at the 4th Conference of the SSSR, April, San Diego. Van Bon, W. H. J. & Van der Pijl, J. M. L. (1997). Effects of word length and wordlikeness on pseudoword repetition by poor and normal readers. Applied Psycholinguistics, 18, 101–114.
(JB)SWL II_f14_229-245
9/26/02
10:36 AM
Page 245
Task-related factors in reading efficiency Van Daal, V. H. P. (in preparation). Learning to read in a foreign language. Van Daal, V. H. P. & Van der Leij, A. (1999). Developmental dyslexia: Related to specific or general deficits? Annals of Dyslexia, 49, 71–104. Van Daal, V. H. P. & Van der Leij, A. (in preparation). Large-scale screening and diagnosis of dyslexia in secondary education. Van den Bos, K. P. (1996). BELL 96. Groningen (Netherlands): Groningen University, Department of Special Education. Van den Bos, K. P., Lutje Spelberg, H. C., Scheepstra, A. J. M. & De Vries, J. (1994). De Klepel. Vorm A en B. Een test voor de leesvaardigheid van pseudowoorden. Verantwoording, handleiding, diagnostiek en behandeling [The Klepel. Form A and B. A test of reading pseudowords]. Nijmegen (Netherlands): Berkhout. Van der Leij, A. & Van Daal, V. H. P. (1989). Repeated reading and severe reading disability. In H. Mandl, E. De Corte, N. Bennett & H. F. Friedrich (Eds.), Learning and instruction 2.2 (235–251). Oxford: Pergamon Press. Van der Leij, A. & Van Daal, V. H. P. (1999a). Automaticity, automatization and dyslexia. In I. Austad, I. Lundberg & F.-E. Tønnessen (Eds.), Dyslexia: Advances in theory and practice (75–98). Dordrecht: Kluwer Academic Publishers. Van der Leij, A. & Van Daal, V. (1999b). Automatization aspects of dyslexia: Speed limitation in word identification, sensitivity to increasing task demands, and orthographic compensation. Journal of Learning Disabilities, 32, 5, 417–428. Van der Leij, A. & Van Daal, V. (2001). Cognitive profiles of children with dyslexia, dyscalculia and hyperlexia: evidence from a study with 12-year old students. Paper presented at the International Conference of the British Dyslexia Association, April 18–21th, York. Wagner, R. K. & Torgesen, J. (1987). The nature of phonological processing and its causal role in the acquisition of reading skills. Psychological Bulletin, 101, 192–212. Wimmer, H. (1993). Characteristics of developmental dyslexia in a regular writing system. Applied Psycholinguistics, 14, 1–33. Wimmer, H., Mayringer, H. & Landerl, K. (1998). Poor reading: A deficit in skill automatization or a phonological deficit? Scientific Studies of Reading, 2, 321–340. Wolf, M. & Bowers, P. G. (1999). The Double-deficit Hypothesis for developmental dyslexias. Journal of Educational Psychology, 91, 415–438. Wolf, M., Bowers, P. G. & Biddle, K. (2000). Naming-speed processes, timing, and reading: A conceptual review. Journal of Learning Disabilities, 33, 4, 387–407. Yap, R. L. & Van der Leij, A. (1993). Word processing in dyslexics: An automatic decoding deficit? Reading and Writing: An Interdisciplinary Journal, 5, 261–279. Address University of Amsterdam Dept. of Education Wibautstraat 4 1091 GM AMSTERDAM The Netherlands
(JB)SWL II_f15_246-264
9/26/02
9:20 AM
Page 246
(JB)SWL II_f15_246-264
9/26/02
9:20 AM
Page 247
Part 3 Attaining Functional Literacy
(JB)SWL II_f15_246-264
9/26/02
9:20 AM
Page 248
(JB)SWL II_f15_246-264
9/26/02
9:20 AM
Page 249
Parental and teacher commitment to emergent literacy development1 Judith Stoep, Joep Bakker and Ludo Verhoeven University of Nijmegen
During the past fifteen years, a growing body of literature has been established on emergent literacy. Emergent literacy can be looked upon as “the reading and writing behaviors that precede and develop into conventional literacy” (Sulzby, 1989). From birth until formal reading instruction, children display a variety of unconventional literacy skills. When children enter school, for example, they typically have considerable insight into the structure and functions of written language. This insight comprises grapheme knowledge, the handling of books and other written materials, and metalinguistic knowledge of written language. Emergent literacy is known to be beneficial to the development of reading and writing skills during elementary school. Not only early reading activities provide a foundation for later reading behavior and skills, but also meaningful interactions with regard to print and situations in which both parents and teachers serve as literate role models (Teale & Sulzby, 1986; Morrow, 1989). The development of emergent literacy skills and the causes of variation in the development of such have been documented in various studies. Extensive surveys (Leseman & De Jong, 1998; Snow, Barnes, Chandler, Goodman & Hemphill, 1992; Blatchford, Burke, Farquhar, Plewis & Tizard, 1985) have shown both home and school to generally contribute to the growth of children’s literacy. Storybook reading is a good example of how parents create a literate environment for their children. What should be noted is the fact that merely the occurrence of storybook reading in families is not sufficient for the development of literacy skills. A match between home and school reading practices appears to be crucial for the transition of literacy skills acquired at home to school. Dickinson, DeTemple, Hirschler, and Smith (1992) demonstrated that preschool teachers’ storybook reading can be characterized as highly focused on organizational talk, with vast control of the topic, and cognitively demanding talk. The contribution of preschool teachers thus differs from that of parents, who are much more concerned with the establishment of an understanding relationship that includes books. These incongruous perspectives can lead to unequal starting points in children’s literacy careers.
(JB)SWL II_f15_246-264
9/26/02
9:20 AM
Page 250
Precursors of Functional Literacy
The literacy environment in the home, parental education, (and mother’s education in particular), expectations for the child, parental teaching, and parental views on education have all been identified as very powerful predictors of children’s literacy development. Opportunity research has a long history dating from the seventies, when Bernstein (1971a, b), first identified the restricted versus elaborated codes used in low versus high socio-economic homes. Although later research showed this distinction to be overly simplistic, many researchers still emphasize the interactional differences that characterize families with a low versus high socioeconomic status (Edwards & Knight, 1997) and the associated educational achievement contrasts. The ways in which “cultural capital” (Bourdieu, 1977) is handed down to children appears to be sociologically determined; upper- and middleclass parents appear to equip their children with the linguistic and behavioral skills that are particularly recognized and appreciated in schools, more than lowerincome and minority parents, as Griffith (1998) concluded in his large-scale survey of parental involvement. In the extensively cited work of Heath (1983), the lag in the literacy achievement of low-income and minority children has also been shown to stem from cultural differences creating a disparity between home and school. According to Heath, mainstream school-oriented children are more likely to make a successful transition from home to school due to their lifelong immersion in the knowledge, habits, and interactional styles required for literacy. Heath has also claimed that low SES black children fail to “adopt the social interactional rules for school literacy events,” which obviously causes them to fall short when it comes to literacy as a whole. Research thus shows a stronger consonance between family and school for children from middle-class homes than for children from lower SES and/or minority homes (Serpell, 1997). In general, the teaching society embraces behavioral principles that are easiest for children from upper middle class homes to conform to (Lareau, 1992). Teachers and parents from lower class homes, in contrast, do not have a shared interest in the education of children, due to different goals. In addition, the foundation for a shared strategy does not appear to be well-balanced, since low SES parents tend to refrain from intervention with their children due to their social background experiences. Teachers can be seen as catalytic agents who mediate between home and school, and thus have the power to boost or weaken the passing of cultural capital from one generation to the next (McCarthey, 2000). Teachers should thus have high expectations for the children they teach although it is clear that teachers often hold low expectations for minority children and children from lower socio-economic homes. The question, then, is whether such low expectations become a self-fulfilling prophecy and thereby cause the children to perform at a low level, which then confirms the viewpoint of the teacher(s). In 1963, Rosenthal and Jacobson (reprinted in 1986) attempted to demonstrate the role of teacher expectancies in their repeatedly criticized project. They used manipulated student qualifications in order to elicit opinions from teachers. More specifically, they tried to prove that higher expectations actually result in higher performance. A sub-
(JB)SWL II_f15_246-264
9/26/02
9:20 AM
Page 251
Commitment to emergent literacy development
stantial body of empirical research has indeed supported the finding that teachers have clear expectations regarding student performance (Dusek, 1985), but the exact ways in which such teacher expectations are translated into actual student achievement remain largely unclear. The main goal of the present study was to gain better insight into the parental and teacher support factors that predict the development of early literacy among minority children and native Dutch children from different socio-economic backgrounds. A distinction was made between children from high versus low SES homes for the native Dutch children in particular, because national investigations have shown a low socio-economic status to often create an educational disadvantage in Dutch classroom settings (Jungbluth & Van Langen, 1990; Driessen, 1995). All of the minority children had a lower socio-economic background, and originated from Suriname, the Dutch Antilles, Turkey, or Morocco. The children from Suriname and the Dutch Antilles, which are both ex-colonies of the Netherlands, immigrated during the past two decades. The Turkish and Moroccan children were second or third generation foreign workers, who came to the Netherlands during the seventies. Previous studies in Dutch elementary schools have shown minority children to generally lag behind their monolingual peers with regard to oral language skills (Verhoeven & Vermeer, 1996). However, the development of early literacy skills within different socio-cultural groups has not been addressed as yet. In the present study, the following questions were therefore addressed. 1. What are the differences in the development of beginning literacy skills (i.e., grapheme knowledge, book orientation, and phoneme analysis) for Dutch high versus low SES children and low SES minority children? 2. What are the differences in parental involvement for Dutch high versus low SES children and low SES minority children? 3. Are teacher and parent views of parental involvement related? 4. To what extent can children’s literacy skills at the start of kindergarten be predicted by parental support factors? 5. To what extent can children’s literacy development after one year of kindergarten be predicted by parental and teacher support factors? With respect to the development of early literacy skills, it is expected that Dutch high SES children will come out highest, followed by Dutch low SES children and then the minority children. Only small differences with regard to parental involvement are expected. Given the fact that Dutch schools mainly reflect middle class values, greater involvement on the part of high SES parents could nevertheless be expected. An important question is whether the views of parents and teachers on parental involvement are related across the three groups or not. It can be expected that children from families showing a disagreement between parental and teacher views of parent involvement are prime candidates for the so-called Pygmalion effects described by Rosenthal and Jacobson: that is, the students of parents who are rated as “educationally detached” are likely to be the same students
(JB)SWL II_f15_246-264
9/26/02
9:20 AM
Page 252
Precursors of Functional Literacy
for which the teacher holds low expectations (Bakker, 1984). These stereotyped expectations may then influence the way in which teachers approach their students and thereby produce a differential course of development. For this reason, it is also important to examine the prediction of pre-reading development by parental support factors and, after one year of kindergarten, teacher support factors.
Method Participants A total of 193 children participated in the study. The children came from 84 kindergartens from a sample of Dutch schools with mixed socio-cultural populations. Almost 50% of the children were educated in mixed kindergarten groups; the other 50% were educated in separate kindergarten groups for older children. Of the 193 children, 103 were native Dutch with a high SES family, 46 were native Dutch with a low SES family, and 44 belonged to one of the following ethnic communities: Surinamese, Antillean, Turkish, or Moroccan. The children in the three groups had the same age level (5 years, 5 months at the first moment of measurement) and were equally distributed with regard to sex.
Materials Teacher interview A written questionnaire was sent to the teachers to gain insight into their perceptions of students from different cultural and ethnic groups along with the involvement of the parents of these students in the school. All of the items were rated using a 7-point Likert-type scale varying from “strongly disagree” to “strongly agree.” The following areas were studied: – parental involvement and resources; – actual parental activities within the school; and – parental norms and values. A parental involvement variable was constructed on the basis of a factor analysis of the answers to the 23 questions, using Principal Axis Factoring (PAF) with oblimin rotation. A reliability analysis was performed to determine the contributions of the separate questions to the factor. As can be seen from Table 1, all of the questions were found to contribute to the factor “parental involvement according to teacher” in a reasonable manner. The Cronbach’s alpha for the total list was .93.
(JB)SWL II_f15_246-264
9/26/02
9:20 AM
Page 253
Commitment to emergent literacy development Table 1. Results of reliability analysis for parent involvement items from teacher questionnaire Teacher questionnaire item 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
Parents play games with child Parents invest time and money in education of child Parents feel competent Parents monitor language use Parents are supportive towards child Books are present in home situation Parents address complex topics Parents feel that teaching is the school’s responsibility Parents are generally educated Parents inquire into their child’s achievement Parents and teacher share developmental goals Teacher has no trouble communicating with parents Child comes to school in excellent condition Parents show interest in child’s school assignments Parents are frequent visitors of parent evenings Parents regard child as center of the family Parents know their child’s strong/weak traits Child is fluent in Dutch language Parents are selective in allowing their child to watch TV Parents are supported in raising their child Parents educate their child to be responsible/independent Parents are willing to question their role in case of problems Parents aim for high education of their child
Corrected item-total correlation .81 .78 .78 .80 .73 .72 .69 .69 .66 .64 .66 .61 .60 .56 .56 .54 .59 .53 .52 .44 .47 .23 .22
Parent interview The parents were given a written questionnaire, which addressed a number of literacy measures and both the interest and involvement of the parents in school activities. This questionnaire differed in format and formulation from the teacher questionnaire, but some of the topics such as parental involvement and schoolorientated activities were similar. The items were rated along various 5-point scales. Using PAF and oblimin rotation, four factors could be distinguished: parental reading pleasure and reading intensity (alpha .83), number of books present in the family’s home (.59), parental involvement (.67), and parental contact with books at home (.62). Some questions in the parent survey proved to address isolated concepts. In effect, they were entered in the comparison independently. The questions concerned the importance parents attach to their own role in their child’s process of learning-to-read, the frequency of help provided during their child’s learning, their expectations for the child’s achievement in reading and in secondary education. Because the new factors had unequal numbers of items, the relevant scores were transformed into z-scores and, in such a manner, comparison of the parent and teacher views on parent involvement was made possible.
(JB)SWL II_f15_246-264
9/26/02
9:20 AM
Page 254
Precursors of Functional Literacy Table 2. Factor loadings and reliability outcomes for parent questionnaire Parent questionnaire item
Frequency of reading present Reading storybooks with child Frequency of reading past Child asks to be read to Parent enjoys reading Number of novels in home Number of informative books in home Number of children’s books in home Parent assists in school activities Parent assists in classroom Parent visits parent nights Frequency of reading magazines Frequency of reading informative books Frequency of reading newspapers Frequency of reading novels Frequency of visiting library
Factor 1: Reading pleasure/ intensity .84 .79 .77 .64 .52 .04 .10 .25 .22 .16 .18 .16 .22 .21 .18 .30
Factor 2: Number of books in home .18 .12 .28 –.03 .07 .70 .65 .53 .08 .12 –.01 .22 .19 .21 .22 .01
Factor 3: Parental involvement .24 .24 .33 .11 .20 .03 –.04 .29 .79 .70 .44 .18 .10 .20 .13 .14
Factor 4: Parental book contact .32 .34 .30 .24 .14 .24 .25 .22 .20 .19 .12 .66 .64 .54 .34 .32
Early literacy skills Early literacy skills were assessed using three tasks also used in a pilot study for this project (Stoep & Verhoeven, 1999, 2000). The first task was a book orientation task, which is based on the emergent literacy test of metalinguistic knowledge of written language from Clay (1982). The children were asked to identify words, individual graphemes, and were tested on concepts about print. The second task tested the phonological awareness of children by having them divide orally presented words into sounds. The third task measured literacy skills in terms of letter and sound naming. All of the tasks rendered Cronbach’s alpha reliability values greater than .83 in the pilot study. The tasks were administered at the beginning and end of the second year of kindergarten (pre- and post-test). The tasks took approximately 30 minutes to administer, and this was done by the remedial teachers working at the relevant schools. The testing took place in the school and was conducted individually. To explore the relationship between the views of the teachers and the parents with regard to parental involvement, on the one hand, and student outcomes, on the other hand, ANOVAs, correlational, and regression analyses were performed. The sum scores for book orientation, phoneme analysis, and grapheme knowledge (letter and sound naming) were computed. Pre- versus post-test comparisons were then made for three groups: high versus low SES children and minority children. The post-test scores were correlated with original scores for the different questionnaire factors and to some background information such as language spoken
(JB)SWL II_f15_246-264
9/26/02
9:20 AM
Page 255
Commitment to emergent literacy development
in the child’s home. Those variables found to be significant in the correlation analysis were used as input for the regression analysis.
Results In this section, the results for the pre- versus post-test comparisons will be presented. Thereafter, the relations between the parent versus teacher factors will be considered in terms of the results of the correlational and regression analyses. Table 3. Pre- and post-test scores (mean and SD for high versus low SES children and minority children on grapheme knowledge, book orientation, and phoneme analysis (ANOVAs) Max. Grapheme knowledge pre-test post-test Book orientation pre-test post-test Phoneme analysis pre-test post-test
High SES
Low SES
Minority
F
p
3.73 (5.53) 7.35 (8.41)
2.16 (4.70) 5.07 (7.03)
1.38 (3.17) 19.01 .000 4.33 (6.78) 11.69 .000
34 24 11.76 (4.61) 8.79 (5.13) 6.67 (4.69) 80.88 .000 14.91 (3.58) 12.23 (5.01) 10.09 (5.30) 74.92 .000 20 2.08 (4.07) 5.59 (6.13)
.84 (2.37) 2.75 (4.12)
.54 (2.01) 17.20 .000 2.47 (4.64) 26.93 .000
As can be seen from Table 3, there is a significant discrepancy between the pre-test grapheme knowledge of Dutch high SES children, on the one hand, and low SES and minority children, on the other hand at the beginning of second year in kindergarten. No significant differences exist between the low SES and minority groups. The three groups differed significantly with regard to book orientation at pre-test, and the pre-test results for phoneme analysis show the same pattern as for grapheme knowledge: significant differences between the high SES students and the other children, but no differences between the low SES and minority children. At the end of the second year in kindergarten, the children from high SES homes obtain significantly higher scores on grapheme knowledge, book orientation, and phoneme knowledge. There are no substantial differences between the low SES and minority children at the level of grapheme knowledge and phoneme analysis, but a significant gap is observed for book orientation. In order to evaluate the gains in emergent literacy, the “log odd” procedure as proposed by Allerup and Elbro (1998) was applied. This method of calculation for difference scores takes any floor or ceiling effects into account. The assumption underlying the procedure is that the total number of correctly solved items for a test is a sufficient indicator of subject ability. A significant discrepancy was found with regard to phoneme analysis for the high SES students, on the one
(JB)SWL II_f15_246-264
9/26/02
9:20 AM
Page 256
Precursors of Functional Literacy
High ses
Low ses
Minority
1 0.8 0.6 0.4 0.2 0 –0.2 –0.4 –0.6 –0.8 –1 teacher view: parental involvement
parent view: parental involvement
parent: frequency of help
parent: own role in child’s learning to read
Figure 1. Parent and teacher views on parental involvement, parent help and importance in reading progress (z-scores)
hand, and the low SES and minority students, on the other hand (F (2, 725) = 9.14, p = .000). No significant differences were observed between the latter two groups, however. The growth in book orientation and grapheme knowledge appears to be similar for all groups (F (2, 738) = .95, p = .386; F (2, 733) = .06, p = .938). A comparison of the teacher and parent views on parental involvement with the actual help and associated importance parents report giving their children was next undertaken. In Figure 1, support is presented for the argument that teachers have different expectations regarding the involvement of parents from different socio-economic groups: parents belonging to a minority group or a lower socio-economic group are expected to have a smaller degree of parental involvement than parents from a higher SES group. Moreover, the minority parents are found at the lowest end of the involvement scale. Minority parents actually think that they contribute more than other parents to their child’s learning to read. In this respect they differ significantly from the high SES parents who rate their assistance less highly (F (2, 363) = 8.21, p = .0003; F (2, 346) = 8.56, p = .0002). When the minority parents are asked about their participation in school activities, they rate rather low, which is in contrast to other parents (F (2, 331) = 4.42, p = .0128). On the whole, teachers’ estimates of parental involvement tend to be higher than the actual involvement reported by parents. T-tests revealed the following pattern of results. For high SES parents, a discrepancy exists between teacher views of parental involvement and parent-reported involvement, importance in their child’s process of learning to read, or frequency of help (t (117) = 4.63, p = .000; t (124) = 7.68, p = .000; t (113) = 7.61, p = .000). On all levels, the teacher estimates are higher than the parent estimates.
(JB)SWL II_f15_246-264
9/26/02
9:20 AM
Page 257
Commitment to emergent literacy development Table 4. Correlations between teacher and parent views of parental involvement, frequency of aid, and importance of parent role in child’s learning to read 1 Teacher view of parental involvement High SES Teacher view of parental involvement Frequency of parent help Importance parent role Parent view of parental involvement Low SES Teacher view of parental involvement Frequency of parent help Importance parent role Parent view of parental involvement Minority Teacher view of parental involvement Frequency of parent help Importance parent role Parent view of parental involvement
–
–
–
2 Frequency of parent help (parentreported)
3 Importance parent role (parentreported)
4 Parent view of parental involvement
–.05
.05
.29***
–
.22** –
.23** .09 –
.01
.16
.46***
–
.28* –
.24* .31** –
.07
.05
.28*
–
.32*** –
.39*** .23* –
* p < .05, ** p < .01, *** p < .001
For low SES parents, no significant differences are found to exist between teacher views of parental involvement, and parent-reported aid, importance, or involvement (t (48) = –.61, p = .547; t (51) = 1.20, p = .235; t (48) = .42, p = .677). For minority parents and the teachers of their children, some agreement on the level of parental involvement is found. Both teachers and parents provide low estimates (t (45) = –49, p = .627). Teachers and minority parents nevertheless differ with regard to their perceptions of the aid provided by parents and their judgements of the importance of such aid (t (53) = –2.99, p = .004; t (53) = –2.34, p = .023): Parents actually value giving such aid and provide such aid more than teachers are inclined to think. The correlations reported in Table 4 show significant connections between the views of teachers and parents with regard to parental involvement in children’s literacy development for all groups. However, the link for the high SES and the minority group is not as strong as for the low SES group. In addition to this, actual parental aid and the importance parents ascribe to such aid do not, in general, appear to be related to the ways in which teachers view parental involvement.
(JB)SWL II_f15_246-264
9/26/02
9:20 AM
Page 258
Precursors of Functional Literacy
In the regression analysis, performed using all of the variables correlating significantly with the end of kindergarten literacy post-test results for the children, the following variables were initially included: – – – – – – – – – – – – –
pre-tests of emergent literacy skills home language of child parental education parental involvement as reported by the teacher parental involvement reported by the parent frequency of reading activities with the child at home number of books in the home parental participation in school activities parental reading pleasure and intensity educational expectations for child held by parents expectations of parents regarding child’s reading progress importance parents attribute to their own role in child’s learning to read actual help parents give to child
The regression results show parental reading expectations to consistently be the main predictors of the children’s emergent literacy (10%, 8%, and 7% for grapheme knowledge, book orientation, and phoneme analysis, respectively). Parental education also proved to be an important predictor of pre-test scores for book orientation. With regard to the prediction of the post-test emergent literacy scores for the high SES students, Table 5 shows their pre-test scores on grapheme knowledge to be the most important predictor of their later grapheme knowledge and to account for over 61% of the variance in the children’s grapheme scores at posttest. The teacher’s view of parental involvement accounts for another 2% of the variance in the post-test grapheme scores. The post-test scores for book orientation are also predicted best by the pre-test scores of the children on this variable, which accounts for 44% of the variance in their post-test book orientation scores. An additional 2% of the variance is accounted for by the reading intensity and pleasure in the child’s home. The post-test phoneme analysis scores of the high SES students are also best predicted by their pre-test scores, which account for 35%; another 3% of variance is added by parental involvement as reported by the teacher. As can be seen from Table 6 for the low SES students, pre-test scores were consistently predicted by teacher views on parental involvement: 9% for grapheme knowledge, 15% for book orientation, and 8% for phoneme analysis. The children’s pre-test grapheme knowledge was also predicted by the home language. The children’s pre-test literacy skills consistently predicted a significant amount of the variance in their post-test grapheme knowledge (68%), book orientation (54%), and phoneme analysis (38%). Their grapheme knowledge was also accounted for to a significant degree by the educational expectations that the parents hold for their child (5%). In the pre-test regression analysis for the minority students, the teacher’s view of parental involvement was found to be of importance for both book orienta-
(JB)SWL II_f15_246-264
9/26/02
9:20 AM
Page 259
Commitment to emergent literacy development Table 5. Summary of stepwise regression analysis results for high SES students: Variables predicting pre- and post-test scores, respectively, for grapheme knowledge, book orientation, and phoneme analysis
Pre-test Grapheme knowledge Book orientation Phoneme analysis Post-test Grapheme knowledge Book orientation Phoneme analysis
Variable
B
SE B
b
R2
Step 1: Reading expectations Step 2: Parental help Step 1: Reading expectations Step 2: Parental education Step 1: Reading expectations
2.10 –.91 1.64 1.54 1.30
.61 .38 .48 .60 .37
.31 –.21 .29 .22 .26
.10 .14 .08 .13 .07
Step 1: Pre-test Step 2: Teacher view Step 1: Pre-test Step 2: Reading intensity Step 1: Pre-test Step 2: Teacher view
1.19 .08 .52 .16 .89 .07
.09 .03 .05 .07 .11 .03
.78 .15 .66 .15 .59 .17
.61 .63 .44 .46 .35 .38
Table 6. Summary of stepwise regression analysis results for low SES students: Variables predicting pre- and post-test score, respectively, for grapheme knowledge, book orientation, and phoneme analysis Variable Pre-test Grapheme knowledge Book orientation Phoneme analysis Post-test Grapheme knowledge Book orientation Phoneme analysis
Step 1: Teacher view Step 2: Home language Step 1: Teacher view Step 1: Teacher view Step 1: Pre-test Step 2: Educational expectation Step 1: Pre-test Step 1: Pre-test
SE B
b
R2
.07 .99 .10 .03
.03 .44 .03 .02
.31 .22 .38 .29
.09 .14 .15 .08
1.23 .70 .72 1.07
.12 .23 .10 .15
.82 .24 .74 .62
.68 .73 .54 .38
B
tion (9%) and grapheme knowledge (5% in the second step). Parental reading expectations for the child proved to be the main predictor of pre-test grapheme knowledge for the minority students (8%). With regard to the post-test scores of the minority children, teacher perceptions of parental involvement appeared to play a major role. Both grapheme knowledge and phoneme analysis are, oddly enough, predicted by how teachers rank the child’s family situation in relation to the ideal supportive home situation. The children’s pre-test results also account for 40%, 44%, and 9% of the variance in their grapheme knowledge, book orientation, and phoneme analysis post-test scores, respectively.
(JB)SWL II_f15_246-264
9/26/02
9:20 AM
Page 260
Precursors of Functional Literacy Table 7. Summary of stepwise regression analysis results for minority students: Variables predicting pre- and post-test scores, respectively, for grapheme knowledge, book orientation, and phoneme analysis Variable Pre-test Grapheme knowledge Book orientation Phoneme analysis Post-test Grapheme knowledge Book orientation Phoneme analysis
Step 1: Reading expectation Step 2: Teacher view Step 1: Teacher view — Step 1: Pre-test Step 2: Teacher view Step 1: Pre-test Step 1: Teacher view Step 2: Pre-test
SE B
b
R2
.89 .04 .07
.35 .02 .03
.28 .23 .30
.08 .13 .09
1.35 .10 .75 .08 .70
.24 .04 .12 .03 .30
.63 .30 .66 .35 .30
.40 .48 .44 .12 .21
B
Conclusions and discussion At the end of kindergarten, when children are about to receive formal reading education, variance in the way in which different groups of children are prepared for learning to read can be expected. Particularly with regard book orientation, minority children display a significant lag in development. The preparations of children from lower SES homes with regard to phoneme and grapheme knowledge also appears to lag behind. Although the different aspects of emergent literacy generally show similar growth for all children (with the exception of phoneme knowledge), the education that these children receive does not appear to compensate for initial disadvantages. Such obstacles at the threshold of elementary school may therefore explain the lack of reading achievement often observed for children from at-risk populations in the Netherlands (Van der Leij, Meijnen & Leseman, 1995; Verhoeven & Gillijns, 1994; Vallen & Stijnen, 1991). When we consider teachers’ and parents’ ideas regarding parental involvement, the discrepancy is found to be largest in the event of cultural differences between home and school. That is, the stereotypic image of non-involved minority parents seems to pervade among teachers. When minority parents are interviewed, moreover, they indeed appear to be not as involved as parents from Dutch high socio-economic homes. Yet minority parents clearly feel that they are responsible in part for teaching their children to read and actually report helping their children more than other parents. The results of the regression analyses show parental involvement to be strongly linked to the emergent literacy of kindergartners. Rather remarkable was the finding that this link is not direct but embedded in teacher perceptions of parental involvement. The amount of variance accounted for by these perceptions does not appear to be spectacular at first, but it should be noted that even the pre-test scores of the children were influenced by the teacher views of parental involvement. This sug-
(JB)SWL II_f15_246-264
9/26/02
9:20 AM
Page 261
Commitment to emergent literacy development
gests a cumulative effect of teacher views on emergent literacy. It has already been shown that teacher visions of parental involvement work to the disadvantage of minority parents. It is therefore not inconceivable that teachers also harbour low expectations regarding the achievement of minority students as a result of their low expectations regarding the parents. Such low expectations can then lead to a situation in which minority children are underserved in terms of lower goals being set for them and other circumstances in which their potential is simply not recognized. In several articles (see Sénéchal, Lefevre, Thomas & Daley, 1998 for an overview), attention has been paid to the diversity of terminology pertaining to parental involvement. It has been suggested that the terms “parent teaching”, “parent attachment” and “parent coaching” may be used interchangeably. Such ambiguity may, however, indicate some confusion with regard to the interpretation of parental participation. The distinction between a skillful versus playful approach to learning to read comes to mind. The skillful approach that appears to characterize traditionally-educated minority parents may not match the more playful approach often adopted by teachers today to help children learn to read. The mismatch between the estimates of parental involvement may therefore reflect a discrepancy in jargon rather than an actual difference. The aforementioned lack of clarity in terminology may also explain the relative failure of such literacy stimulation projects as HIPPY in Israel and Opstap in the Netherlands. Eldering and Vedder (1997) and Snow, Burns, and Griffin (1998) have emphasized the need for programs to be designed for and adapted to the manner in which parents participate in education. A focus on the parent alone, moreover, has proven insufficient to expand the developmental opportunities for children from underprivileged homes. The promotion of programs intended to stimulate parental involvement in children’s literacy must be enlarged. In the past, parents from lower socioeconomic homes were assumed to be incapable of preparing their children for school and supporting their children during their educational careers. Family and school programs were called into existence to compensate for this lack of “cultural baggage.” However, the projects based on such a set of assumptions have not yielded the success that was hoped for. In various family literacy projects in the United States, however, educators seem to be making a step in the right direction with an emphasis on the supplemental work that parents can do to help their children and guide their educational careers. Parents are clearly told, for example, that the activities included in the program, are not part of the child’s school curriculum. In such a manner, parents are given the feeling that their contribution is a meaningful one. A collective effort on the part of teachers and parents appears to be imperative, and must not take place in terms of unilateral adaptation. In addition to this, adaptive forms of education can clearly enlarge the learning opportunities for children at risk. Houtveen, Booij, De Jong, and Grift (1996) formulate adaptive education in terms of the acquisition of detailed insight into learning progress in order to provide goal-oriented steering. The most relevant aspects of such an approach are teacher driven instruction, the tracking of results, the stimulation of student self-management, and student commitment to the planning of the learning process.
(JB)SWL II_f15_246-264
9/26/02
9:20 AM
Page 262
Precursors of Functional Literacy
The results of the present research show teacher attitudes to play a meaningful role in the process of learning to read. In order to improve the possibilities of optimal reading education for all children, teacher attitudes should be subject to discussion. Teachers should learn more about teaching strategies for different types of students or groups of students and also how to deal with cultural variation within the classroom. In fact, this aspect of pedagogy should be integrated into the curriculum for future teachers. By means of questionnaires similar to the ones used in the present project, teachers can become aware of the attitudes they harbour. These attitudes can then become the topic of school team discussions on reading policy and educational disadvantage. The results of the present study also suggest that there are many ways in which to evaluate parental involvement. Our data consisted of self-report questionnaires, which may not be the best manner of assessment. Autobiographical information may be influenced by the wish to produce socially desirable behavior. Should this be the case, however, both teacher and parent reports may be similarly colored and the systematic differences detected between the views still remain. As stated earlier, the present project constitutes part of a longitudinal study of the development of emergent language and literacy skills. In future research, the consequences of emergent literacy for formal reading instruction outcomes will be evaluated. Observational data and additional tests will be used to get a grip on the ways in which home and school variables interact in the process of early literacy development. Emergent literacy learning is a process to which the Vygotskian notion of “expanding the zone of proximal development” clearly applies (Vygotsky, 1962). Teachers try to enlarge the child’s knowledge, building on it’s interests and abilities by challenging the child and highlighting new connections to familiar information. A prerequisite for this, however, is an adequate assessment of the child’s abilities. Teachers should beware of making such unverified estimates and particularly with regard to parental support. Authentic assessment of children’s abilities and parental support appears to be crucial to expand the opportunities of children from all backgrounds. In her pioneering work with portfolio assessment, for example, Valencia (1990) has introduced new ways to evaluate children’s reading progress. Comparable methods have been developed to gain insight into family background, and these include such instruments as the Ecological Inventory used by Sonnenschein, Baker, and Serpell (1995) and the family portfolios designed by Paratore et al. (1995). With the use of such aids, teachers may become more aware of the diversity of routes that children can follow towards literacy and perhaps appreciate this variability to a greater extent as well.
Notes . This study is supported in part by grants from the Dutch Reading Platform Stichting Lezen in Amsterdam, the Netherlands.
(JB)SWL II_f15_246-264
9/26/02
9:20 AM
Page 263
Commitment to emergent literacy development
References Allerup, P. & Elbro, C. (1998). Comparing differences in accuracy across conditions or individuals: An argument for the use of log odds. Quarterly Journal of Experimental Psychology, 51A, 409–424. Bakker, J. (1984). De leerkracht en de slechte leerling: een studie naar inhoudelijke en structurele aspecten van de oordeelsvorming van leerkrachten [The teacher and the bad pupil: a study into intrinsic and structural aspects of judgement formation of teachers]. Meppel: Krips Repro. Bernstein, B. (1971a). Class, codes and control. Vol. 1. Theoretical studies towards a sociology of language. London: Routledge and Keagan Paul. Bernstein, B. (1971b). Education cannot compensate for society. New Society, 16, 344–347. Blatchford, P., Burke, J., Farquhar, C., Plewis, I. & Tizard, B. (1985). Educational achievement in the infant school: the influence of ethnic origin, gender and home on entry skills. Educational Research, 27, 52–60. Bourdieu, P. (1977). Cultural reproduction and social reproduction. In J. Karabel & A. Halsey (Eds.), Power and ideology in education (487–511). New York: Oxford University Press. Clay, M. (1982). Observing young readers. London: Heinemann. Dickinson, D., DeTemple, J., Hirschler, J. & Smith, M. (1992). Book reading with preschoolers: Coconstruction of text at home and at school. Early Childhood Research Quarterly, 7, 323–346. Driessen, G. J. W. M. (1995). Het relatieve belang van sociaal milieu en etnische herkomst voor de verklaring van onderwijsachterstanden [The relative importance of social and ethnic background for the explanation of educational disadvantages]. Tijdschrift voor Onderwijsresearch, 4, 341–362. Dusek, J. (1985). Introduction to teacher expectancy research. In J. Dusek (Ed.), Teacher expectancies (1–6). Hillsdale, NJ: Erlbaum. Edwards, A. & Knight, P. (1997). Parents and professionals. In B. Cosin & M. Hales (Eds.), Families, education and social difference. London: Routledge. Eldering, L. & Vedder, P. (1997). De preventieve waarde van OPSTAP [The preventive value of OPSTAP]. Nederlands Tijdschrift voor Opvoeding, Vorming en Onderwijs, 13, 360–368. Griffith, J. (1998). The relation of school structure and social environment to parent involvement in elementary schools. The Elementary School Journal, 99, 53–80. Heath, S. B. (1983). Ways with words. New York: Cambridge University Press. Houtveen, A., Booij, N., De Jong, R. & Grift, W. (1996). Adaptief onderwijs en leerlingresultaten [Adaptive education and pupil achievement]. Pedagogische Studiën, 73, 422–433. Jungbluth, P. & Van Langen, A. (1990). Onderwijskansen van migranten. De rol van sociaaleconomische en culturele factoren [Educational opportunities for minorities. The role of socio-economic and cultural factors]. Amsterdam: Swets & Zeitlinger. Lareau, A. (1992). “It’s more covert today”: The importance of race in shaping parents’ views of the school. Special Studies in Teaching and Teacher Education, 7, 4–27. Buffalo: State University of New York. Leseman, P. & De Jong, P. (1998). Home literacy: Opportunity, instruction, cooperation and social-emotional quality predicting early reading achievement. Reading Research Quarterly, 33, 294–318. McCarthey, S. (2000). Home-School connections: A review of the literature. Journal of Educational Research, 93, 145–153. Morrow, L. M. (1989). Literacy development in the early years: Helping children read and write. Englewood Cliffs, NJ: Prentice Hall. Paratore, J., Homza, A., Krol-Sinclair, B., Lewis-Barrow, T., Melzi, G., Stergis, R. & Haynes, H. (1995). Shifting boundaries in home and school responsibilities: The construction
(JB)SWL II_f15_246-264
9/26/02
9:20 AM
Page 264
Precursors of Functional Literacy of home-based literacy portfolios by immigrant parents and their children. Research in the Teaching of English, 29, 367–389. Rosenthal, R. & Jacobson, L. (1986). Pygmalion in the classroom. New York: Holt, Rinehart, and Winston, Inc. Sénéchal, M., Lefevre, J., Thomas, E. & Daley, K. (1998). Differential effects of home literacy experiences on the development of oral and written language. Reading Research Quarterly, 33, 96–116. Serpell, R. (1997). Literacy connections between school and home: How should we evaluate them? Journal of Literacy Research, 29, 587–616. Snow, C., Barnes, W., Chandler, J., Goodman, I. & Hemphill, L. (1991). Unfulfilled expectations: Home and school influences on literacy. Boston: Harvard University Press. Snow, C., Burns, M. & Griffin, P. (1998). Preventing reading difficulties in young children. Washington, DC: National Academy Press. Sonnenschein, S., Baker, L. & Serpell, R. (1995). Documenting the child’s everyday home experiences: The ecological inventory as a resource for teachers. Instructional Resource no. 11. Athens, GE: National Reading Research Center. Stoep, J. & Verhoeven, L. (1999). Effects of teachers’ book reading on children’s early language and literacy development. NRC Yearbook 48, 86–93. Stoep, J. & Verhoeven, L. (2000). Stimulering van beginnende geletterdheid bij kleuters uit risicogroepen [Stimulation of emergent literacy in the case of at-risk kindergartners]. Apeldoorn, the Netherlands: Garant. Sulzby, E. (1989). Emergent literacy. Kindergartners write and read. Ann Arbor, MI: University of Michigan. Teale, W. & Sulzby, E. (1986). Emergent literacy as a perspective for examining how young children become writers and readers. In W. Teale & E. Sulzby (Eds.), Emergent literacy: Writing and reading (vii–xxv). Norwood, NJ: Ablex. Valencia, S. (1990). A portfolio approach to classroom reading assessment: The whys, whats, and hows. The Reading Teacher, 43, 338–340. Vallen, T. & Stijnen, S. (1991). Ethnic minorities and ethnic minority languages in Dutch primary schools: Problems and challenges. Language and Education, 5, 113–123. Van der Leij, A., Meijnen, G. & Leseman, P. (1995). De leer- en opvoedingssituatie van jonge autochtone en allochtone leerlingen in het perspectief van hun schoolcarrière [The educational situation of young native and ethnic minority children at home and at school in perspective of their school career]. Pedagogische Studiën, 72, 162–171. Verhoeven, L. & Gillijns, P. (1994). Ontwikkeling van beginnende lees- en spellingvaardigheid [Development of early reading and spelling abilities]. Tijdschrift voor Onderwijsresearch, 19, 259–279. Verhoeven, L. & Vermeer, A. (1996). Taalvaardigheid in de bovenbouw. Nederlands van autochtone en allochtone leerlingen in het basis- en mlk-onderwijs [Language ability at the end of secondary school. Dutch language skills of autochtonous and foreign pupils in primary and special education]. Tilburg: Tilburg University Press. Vygotsky, L. (1962). Thought and language. New York/London: M.I.T. Press and John Wiley & Sons. Address University of Nijmegen Dept. of Special Education P.O. Box 9104 6500 HE Nijmegen The Netherlands
(JB)SWL II_f16_265-286
9/26/02
9:21 AM
Page 265
Sociocultural differences in reading skills, reading motivation and reading strategies* Willy van Elsäcker and Ludo Verhoeven University of Nijmegen
Research has shown striking differences in the development of reading in different sociocultural groups (for an overview see Durgunoglu & Verhoeven, 1998; Koda, 1996; Weber, 1991). It can be assumed that the various processes of reading in a second language are a consequence of the difficulties learners have in grasping the linguistic patterns of the target language and applying (meta)linguistic cues for reading. Both decoding skills and linguistic comprehension may be affected by limited oral proficiency. Cultural factors may also play a role, and differences in background knowledge can be expected to affect comprehension processes in particular. In other words, the processes of learning to read in a first and second language may differ in many ways. Both first and second language reading seem to be highly dependent on the use of effective strategies. Langer, Bartolomé, Vàsquez, and Lucas (1990), who examined meaning construction of bilingual students, found that the use of good ‘meaning-making strategies’ affected reading competency more than did first- or second-language proficiency. Students used their knowledge of Spanish as support when they encountered difficulties in English texts. Students who tended to be good readers in either of the two languages also tended to be good readers in the other language. According to Langer and her colleagues this was due to the transfer of good comprehension strategies across languages. Jiménez, García, and Pearson (1996) examined the strategic reading processes of bilingual Latino students and monolingual English students in sixth grade. It appeared that successful Latino readers actively transferred information across languages, translated from one language to another, and made use of cognates (words that are related across languages) in their less dominant language. The less successful Latino readers were much less efficient in using these strategies. In the verbal protocols it appeared that the monolingual English students reported less strategy use than the Latino students did. The importance of efficient strategy use for reading comprehension is confirmed and acknowledged by many researchers (Baker & Brown, 1984; Paris, Wasik & Turner, 1991; Pressley, 1998). Strategies can be seen as procedures of a conscious
(JB)SWL II_f16_265-286
9/26/02
9:21 AM
Page 266
Precursors of Functional Literacy
and intentional nature (Alexander, Graham & Harris, 1998). According to Paris, Wasik, and Turner (1991) “strategic readers are not characterized by the volume of tactics that they use but rather by the selection of appropriate strategies that fit the particular text, purpose, and occasion” (p. 611). Verbal protocol studies have demonstrated that adult readers appear to use a range of strategies that are not present in poorer and younger readers (Pressley & Afflerbach, 1995). Although there is general consensus that younger and poorer readers do not monitor their comprehension as effectively as older and better readers (Baker & Brown, 1984; Pressley 1998), research has shown contradictory results concerning the optimal quantity of strategy use. For example, on the relation between context use and reading ability two different views exist. Some researchers have found that good readers make better use of contextual information than poor readers (Neville & Pugh, 1976/1977), whereas other researchers have demonstrated that poor readers rely more on context than good readers do, to compensate for their slow word encoding (Perfetti, 1984; Stanovich, 1991). Apart from an extreme focus on decoding skills during instruction, and a lack of efficient strategy use, it may also be that a lack of reading motivation is partly responsible for poor reading comprehension scores. Many strategy researchers have stressed the importance of motivation in strategy use (Alexander, Graham & Harris, 1998; Paris, Wasik & Turner, 1991; Pressley, 1998). Guthrie, McGough, Bennett, and Rice (1996) suggest that students who are motivated are more likely to use cognitive strategies in learning situations. Besides, “. . . motivation increases reading amount, which then increases text comprehension” (Guthrie, Wigfield, Metsala & Cox, 1999: p. 250). Wigfield and Guthrie (1997a,b) identified several dimensions of reading motivations. They found that intrinsic as well as extrinsic motivations both related to children’s reading frequencies, although the influence of intrinsic motivations tended to be stronger. McKenna, Kear, and Ellsworth (1995) reported a decline in the amount of reading by children over the school years in the US. Similar findings have been shown in the Netherlands. In general, Dutch children and adults appear to read less than a few decades ago (Piek, 1995). Moreover, research has shown that the time children spend on leisure time reading declines over the elementary school years (Otter & Schoonen, 1996). With regard to reading motivations, there may be differences between mainstream and minority students as well. Teachers and educators have often tried to explain poor reading scores of minority groups by their supposed lack of reading motivation. However, research has not corroborated this view. For example, McKenna et al. (1995) in a nation-wide study in Grades 1 through 6, found AfricanAmerican children to score higher on academic reading attitude than white children, the differences being significant at every grade level. Baker and Wigfield (1999) found that African-American children scored higher on several dimensions of the Motivation for Reading Questionnaire. Moreover, the African-American students reported significantly greater reading activity than white students. Besides, Baker and Wigfield found the relations between reading motivations and reading ability to be stronger for white students than for African-American students, while
(JB)SWL II_f16_265-286
9/26/02
9:21 AM
Page 267
Reading skills, motivation, and strategies
Droop (1999), Graham (1994), and Stevenson, Chen, and Uttal (1990) reported similar findings. Several researchers have found reading frequency to relate to reading achievement (Anderson, Wilson & Fielding, 1988; Greany & Hegarty, 1987). However, research results are contradictory about the effects of motivation and leisure time reading on reading comprehension. Gottfried (1990) found significant correlations between academic intrinsic motivation and text comprehension for 9-yearolds, but not for 7-year-olds. Bast (1995), in a longitudinal study among Dutch children in Grades 1 through 3, found effects of decoding on leisure time reading. Besides, he found that comprehension and decoding skills both were associated with positive reading attitudes, good readers expressing more positive attitudes towards reading than poor readers. Aarnoutse and Van Leeuwe (1998), on the other hand, found reading frequency and reading pleasure not to be related with reading comprehension and vocabulary. Otter (1993) also did not find an effect of free time reading on reading comprehension. The present study focuses on reading skills, reading strategies and reading motivations in different sociocultural groups in the Netherlands: native Dutch students and minority students, who originate from Surinam, the Dutch Antilles, Turkey, and Morocco, being the largest majorities among the Dutch minorities. With respect to the Dutch students, a distinction was made between a high vs low socioeconomic background. All of the minority students had a low socioeconomic background. Most children from Surinam and the Dutch Antilles, both ex-colonies of the Netherlands, immigrated during the past three decades. The Turkish and Moroccan children belong to the second or third generation of foreign workers who came to the Netherlands during the 1970’s. Previous studies on the language and literacy development of minority students in the Netherlands show fairly consistent findings: the Dutch students usually obtaining the highest scores, followed by the Surinamese, Moroccan and Turkish students respectively. Minority children have been found to lag behind on skills such as listening comprehension, vocabulary, and reading comprehension (Droop, 1999; Sijtstra, 1997). With regard to vocabulary the gap tends to widen over the elementary school years (Verhoeven & Vermeer, 1996). Only on decoding skills minority children seem to catch up with their Dutch peers (Droop, 1999; Smits & Aarnoutse, 1997; Zwarts, 1990). Besides, the Dutch students are not doing too well either, as several national language and reading assessments (Zwarts, 1990; Sijtstra, 1992, 1997) and international studies of reading comprehension (Elley, 1992) have shown. The main goal of the present study is to gain a better insight into the strategic and motivational patterns Dutch children of varying socioeconomic backgrounds and minority children (i.e. children from ex-colonies, and children from Turkey and Morocco) show in relation to reading in Grades 3 and 4. An attempt is made to answer the following questions: 1. How do language and reading skills, reading strategies and reading motivations, and leisure time reading develop?
(JB)SWL II_f16_265-286
9/26/02
9:21 AM
Page 268
Precursors of Functional Literacy
2. What are the differences in use of reading strategies, reading motivations, and leisure time reading between Dutch and minority students with varying reading comprehension ability levels? 3. How do reading skills, strategy use, reading motivation, and leisure time reading relate in subgroups of students? We hypothesize that, with regard to listening comprehension, vocabulary, and reading comprehension, the Dutch high SES students will obtain the highest scores, followed by the Dutch low SES students and the Surinamese students respectively. We expect that the Mediterranean students will lag behind severely on these three skills. For decoding we expect that, conform previous Dutch research, the minority groups will catch up with their peers. With regard to strategy use, we expect that minority students will display more strategy use than the Dutch children (cf. Jiménez et al., 1996). Besides, it is expected that minority students will display higher motivations and more leisure time reading than Dutch students (cf. McKenna et al., 1995; Baker & Wigfield, 1999). Regarding the relations between the different variables, we expect reading comprehension to highly correlate with listening comprehension, vocabulary, and to a lesser degree with decoding (cf. Hoover & Gough, 1990). Furthermore, we expect motivations to relate to strategy use, while intrinsic motivations as well as extrinsic motivations are expected to relate with leisure time reading for all groups (cf. Guthrie et al., 1996). Motivations and reading skills are expected to show stronger relations within Dutch students than within minority students (cf. Baker & Wigfield, 1999; Droop, 1999; Graham, 1994).
Method Participants In this two-year longitudinal study, that started at the beginning of third grade, 815 students in 34 Dutch elementary schools were involved. In Table 1, the characteristics of the student population are displayed. Four groups of children were examined: (1) a group of Dutch children with parents of relatively high socioeconomic status (high SES); (2) a group of Dutch children with parents of working class families (low SES); (3) a group of children from Surinam and the Antilles (Ex-colonies); and (4) a group of children from the Mediterranean countries Turkey and Morocco (Mediterranean). The Ex-colonies and Mediterranean group both consisted of children from working class families and could all be classified as low SES. The mean age of the students at the beginning of the study was 8 years and 10 months. For most of the minority children, Dutch was their second language. However, in the Ex-colonies group more Dutch was spoken at home than in the Mediterranean group. Children who lived in the Netherlands less than two years at the beginning of the study were excluded, as it was assumed that they would not speak the Dutch language sufficiently to take the tests. Besides, a miscellaneous group of 88
(JB)SWL II_f16_265-286
9/26/02
9:21 AM
Page 269
Reading skills, motivation, and strategies Table 1. Characteristics of the Grade 3 student population (n = 815) at the first measuring moment Dutch children High SES
Low SES
Minority children Ex-colonies
Mediterranean
Sex Female Male
145 143
93 92
81 99
90 72
Totals
288
185
180
162
104.03
106.10
106.84
107.56
Mean age in months
children from diverse nationalities and socioeconomic backgrounds were excluded from the analyses.
Materials Reading skills were assessed by a number of standardized tests measuring listening comprehension, decoding skills, reading vocabulary, and reading comprehension. Besides, some new instruments were developed to assess the use of reading strategies, reading motivations, and leisure time reading. The reading skills tests, the questionnaires and the reading logs were administered by the classroom teachers. They had received detailed instructions on the use of the tests in advance, to guarantee uniform procedures. The decoding test was administered by the researcher or by a well-trained master student of the Social Sciences Department. For measuring listening comprehension a standardized test of the Dutch CITO pupil monitoring system was used (Luisteren M5, Krom, 1992). This test had an interitem reliability (a) of .86. The children had to listen to a number of short stories, interviews and conversations, after which they answered multiple choice questions. Both the texts and the questions were offered to the children on audio tape. The test consisted of two parts that were administered on two different days at the beginning of grade 3 and at the end of grade 4. The score was 1 (correct) or 0 (incorrect). The maximum score on the test was 33. Decoding skills were measured individually by means of a frequently used Dutch, standardized test, also part of the CITO pupil monitoring system (Drie Minuten Toets, Verhoeven, 1992). This test, in which both speed and accuracy are measured, consists of three word cards. For this study only the third card was used, with the most complex, polysyllabic words. The students were asked to read a list of 120 words as fast and as clearly as possible. The score was the total amount of words read within one minute minus skipped and incorrect words. To measure reading vocabulary also a standardized test from the CITO pupil monitoring system was used (Leeswoordenschattaak E5 and E6, Verhoeven & Vermeer, 1992). For this study, test E5 was used at the beginning and at the end
(JB)SWL II_f16_265-286
9/26/02
9:21 AM
Page 270
Precursors of Functional Literacy
of Grade 3. Test E6 was used at the end of Grade 4. Interitem reliability (a) was .85 for the E5 version, and .87 for the E6 version. To be able to compare the different versions of the test, the raw scores were converted into scale scores. In this test the children had to read 30 sentences in which a word or an expression was printed in bold. The meaning of the word or the expression had to be chosen from four alternatives. The score was 1 (correct) or 0 (incorrect). There was no time limit. Reading comprehension was measured by means of the reading comprehension test for 9-year-olds of the IEA Reading Literacy Study (Elley, 1992). The test measured reading comprehension in narrative and expository texts. Besides, students had to read some graphs, tables, a map, and a schedule. For reasons of efficiency, for this study one narrative and one expository text were deleted. There was no time limit. Reliability (a) on the test as used in the present study was .89 at the beginning of Grade 3, and .90 at the end of Grades 3 and 4. The score was 1 (correct) or 0 (incorrect). The maximum score was 54. To investigate the use of reading strategies, a questionnaire was developed that was adapted from instruments used by Mooij (1994), Pintrich and De Groot (1990), and Sliepen (1995) in previous research. In the questionnaire, the children had to report how often they used a number of given strategies before, during, and after the reading of a narrative or expository text. The more strategies the children reported to use, the higher their score. However, a very high score on the questionnaire does not make a person a better reader. Using all strategies that were mentioned would probably take up so much time that it would hinder reading instead of facilitate understanding. By means of principal component analysis with Varimax rotation four types of strategies were identified: monitoring strategies (9 items, such as ‘I ask myself: Do I understand this part?’), routine strategies (6 items like ‘I underline a word I don’t know’), text-based strategies (3 items like ‘After reading I go through the text a second time’) and estimation strategies (3 items like ‘Before I start reading I look at the text to see whether I will like it’). At all three measuring moments the interitem reliabilities were high for the two first dimensions (.70 < a < .82), moderate for the ‘text-based’ dimension (.50 < a < .60), and low for the estimation dimension (.34 < a < .45). The score was 3 (always), 2 (sometimes) or 1 (never). There were 21 items, so the maximum score was 63. There was no time limit. Reading motivations were assessed by a questionnaire adapted from instruments previously used by Aarnoutse (1990), and Tellegen and Catsburg (1987) The items addressed affective, cognitive and behavioral aspects of reading attitudes as well as reading motivations. By means of principal component analysis with Varimax rotation three dimensions of motivations for reading could be identified: intrinsic motivations (10 items, such as ‘Sometimes I read a book because I like reading exiting stories’), extrinsic motivations (6 items like ‘Sometimes I read a book because my teacher wants me to’), and escape motivation (5 items like ‘Sometimes I read a book because I want to be on my own’). Interitem reliabilities were high on the first two dimensions (.75 < a < .84), and moderate for the ‘escape’
(JB)SWL II_f16_265-286
9/26/02
9:21 AM
Page 271
Reading skills, motivation, and strategies
dimension (.58 < a < .62) at the three measuring moments. For this test there was no time limit. The score was 1 (yes) or 0 (no). The maximum score was 21. Leisure time reading was measured by means of reading logs. The reading logs were adapted from logs used by Otter (1993) in previous research. The students filled in the reading logs daily for twelve weeks in total: two weeks in November, January and in March in third as well as in fourth grade. Students were asked how long they had been reading on the previous day. The response format to this question was: – – – – –
very long: one hour or more (4 points) rather long: about half an hour (3 points) not too long: about a quarter of an hour (2 points) very short: about 5 minutes (1 point) not at all (zero points)
If they had been reading, the students were asked to also write down the title of the book and the author’s name. Books written in other languages than Dutch were reported as well, although this occurred very rarely. The maximum score on the amount of reading per period of two weeks was 56. Alpha reliability of the amount of reading items was .95 in both Grades 3 and 4.
Procedure The tests and questionnaires were administered at three points in time: at the beginning of grade 3, at the end of grade 3 and at the end of grade 4, except for the listening comprehension test which was administered at two points in time. The possible differences between the groups in development over time were tested in a repeated measures MANOVA (GLM) with the total group as ‘between subjects factor’ and time of measurement as ‘within subjects factor’. In the design, the factors Group, Time, and Time by Group interaction for the groups were tested. To determine whether the main Group effect and Time by Group interaction between certain pairs of groups were significant, three contrasts were specified within MANOVA: the Dutch high SES group was contrasted with the Dutch low SES group, the Dutch low SES group with the two Minority groups, and the Excolonies group with the Mediterranean group. To gain more insight into the importance of strategic and motivational variables for reading a comparison between poor and good readers at the beginning of third grade was made for Dutch students, on the one hand, and minority students on the other. In each cohort, the students who had the 25% lowest scores on the IEA-test for reading comprehension were compared with the students who obtained the 25% highest scores. Independent T-tests were computed to compare the results on strategy use, reading motivations, and leisure time reading of the students in the two score groups.
(JB)SWL II_f16_265-286
9/26/02
9:21 AM
Page 272
Precursors of Functional Literacy
To examine the interrelations among all variables used in this study, correlations were computed for the Dutch group as well as the minority group. In order to rule out home language and socioeconomic background influences, we partialed out socioeconomic status (SES) for the Dutch students, and home language for the minority students. Missing data were handled with pairwise deletion.
Results Development of reading skills, strategies and motivations In the Figures 1, 2 and 3, the graphic representations of the mean scores at the different measuring moments are shown. In Table 2, the F-values, degrees of freedom, and significance levels of the MANOVA analyses can be seen for respectively the reading skills, the strategy dimensions, the motivation dimensions, and leisure time reading. First, the effects of Time, of Group, and of Time by Group interaction for the total group are displayed, followed by the Group and Time by Group effects in a comparison of the Dutch high SES versus low SES group, the Dutch low SES versus the two minority groups, and the Ex-colonies versus the Mediterranean group.
Reading skills Figure 1 clearly shows identical patterns for the development of listening comprehension, reading vocabulary, and reading comprehension; the Dutch high SES students obtained the highest scores, followed by the Dutch low SES and the Ex-colonies group respectively, while the Mediterranean group severely lagged behind the other groups. All Group effects were significant (see Table 2). For the total group, the main effects of Time on the four skills were significant as well, indicating that the total group made significant progress over time on all reading skills. Time by Group interaction was significant especially for vocabulary and reading comprehension. On vocabulary the differences between the groups tended to become larger, whereas on reading comprehension they appeared to become smaller. Decoding showed a different picture, both the minority groups making significantly more progress than the Dutch groups. The Ex-colonies group, who started at the same level with the Dutch low SES group, at the end of Grade 4 ended, at the same level with the Dutch high SES group. The Mediterranean group still lagged behind the other groups, although the differences became smaller. Reading strategies Figure 2 demonstrates that no development over time was found on the strategy dimensions. Besides, it can be seen that on all strategy dimensions the Dutch high SES group scored the lowest. In fact, both Dutch groups scored lower than the minority groups on the four dimensions.
(JB)SWL II_f16_265-286
9/26/02
9:21 AM
Page 273
Reading skills, motivation, and strategies Table 2. F-values, degrees of freedom and significance levels of the MANOVA analyses: for the total group, and for the contrasts of Dutch High SES vs. Low SES, Dutch Low SES vs. Minority group, and Ex-colony vs. Mediterranean group TIME
TIME × GROUP
GROUP
F
df
p
F
df
p
1612.10
1,725
***
134.99 44.07 129.79 31.28
3,725 1,725 1,725 1,725
*** *** *** ***
Dec overall HS vs LS LS vs Min ExC vs Med
1930.49
2,1392
***
11.24 n.s. n.s. 13.40
3,696 n.s. n.s. 1,697
Voc overall HS vs LS LS vs Min ExC vs Med
498.16
2,1286
***
140.92 48.26 122.28 50.47
RC overall HS vs LS LS vs Min ExC vs Med
1487.02
2,1450
***
MoS overall HS vs LS LS vs Min ExC vs Med
n.s.
n.s.
RouS overall HS vs LS LS vs Min ExC vs Med
n.s.
Reading skills LC overall HS vs LS LS vs Min ExC vs Med
F
df
p
2.85 n.s. 4.64 n.s.
3,7250 n.s. 1,725 n.s.
* n.s. * n.s.
*** n.s. n.s. ***
7.49 n.s. 15.16 n.s.
6,1392 n.s. 2,1394 n.s.
*** n.s. *** n.s.
3,643 1,645 1,643 1,643
*** *** *** ***
7.50 10.21 n.s. n.s.
6,1286 2,1290 n.s. n.s.
*** *** n.s. n.s.
112.62 45.57 86.64 48.29
3,725 1,727 1,725 1,725
*** *** *** ***
12.41 4.48 13.01 4.20
6,1450 2,1454 2,1450 2,1450
*** * *** *
n.s.
11.61 n.s. 18.15 n.s.
3,569 n.s. 1,569 n.s.
*** n.s. *** n.s.
n.s. n.s. n.s. n.s.
n.s. n.s. n.s. n.s.
n.s. n.s. n.s. n.s.
n.s.
n.s.
23.86 7.61 29.51 n.s.
3,588 1,590 1,588 n.s.
*** * *** n.s.
n.s. n.s. n.s. n.s.
n.s. n.s. n.s. n.s.
n.s. n.s. n.s. n.s.
Strategy use
TxtS overall HS vs LS LS vs Min ExC vs Med
8.05
2,1206
***
25.60 8.98 28.67 n.s.
3,603 1,605 1,603 n.s.
*** ** *** n.s.
n.s. n.s. n.s. n.s.
n.s. n.s. n.s. n.s.
n.s. n.s. n.s. n.s.
EstS overal HS vs LS LS vs Min ExC vs Med
9.28
2,1254
***
10.59 5.05 9.59 n.s.
3.627 1,629 1,627 n.s.
*** * ** n.s.
3.65 4.33 4.00 3.81
6,1254 2,1258 2,1254 2,1254
** * * *
7.00
2,1026
**
11.20 19.78 28.69 n.s.
3,513 1,515 1,513 n.s.
*** *** *** n.s.
n.s. n.s. n.s. n.s.
n.s. n.s. n.s. n.s.
Motivations IntM overall HS vs LS LS vs Min ExC vs Med
n.s. n.s. n.s. n.s.
(JB)SWL II_f16_265-286
9/26/02
9:21 AM
Page 274
Precursors of Functional Literacy Table 2. (cont.) TIME F
TIME × GROUP
GROUP
df
p
F
df
p
F
df
p
ExtM overall HS vs LS LS vs Min ExC vs Med
44.41
2,1168
***
88.22 24.05 108.01 n.s.
3,584 1,586 1,584 n.s.
*** *** *** n.s.
3.22 n.s. n.s. n.s.
6,1168 n.s. n.s. n.s.
** n.s. n.s. n.s.
EscM overall HS vs LS LS vs Min ExC vs Med
27.16
2,1246
***
3.52 n.s. 4.41 n.s.
3,623 n.s. 1,623 n.s.
* n.s. * n.s.
3.13 n.s. 3.77 n.s.
6,1246 n.s. 2,1246 n.s.
** n.s. * n.s.
43.61
5,3485
***
3.06 n.s. 8.90 n.s.
3,697 n.s. 1,697 n.s.
* n.s. ** n.s.
1.83 n.s. n.s. n.s.
15,3485 n.s. n.s. n.s.
* n.s. n.s. n.s.
Home reading Overall HS vs LS LS vs Min ExC vs Med
Note. LC = listening comprehension; Dec = decoding; Voc = reading vocabulary; RC = reading comprehension; MoS = monitoring strategies; RouS = routine strategies; TxtS = text-based strategies; EstS = estimation strategies; IntM = intrinsic motivations; ExtM = extrinsic motivations; EscM = escape motivations. *p < .05; **p < .01; ***p < .001
Table 2 shows that the differences between the Dutch low SES group and the minority groups were all significant. Between the Dutch high and low SES group the differences were significant as well, except on the monitoring strategy dimension. Between the Ex-colonies and Mediterranean group no significant differences were found.
Reading motivations and leisure time reading Figure 3 clearly shows a decline in motivation scores over the two years on all three motivation dimensions, the effects of Time being significant (see Table 2). For intrinsic motivation, the scores of the Dutch high SES and the minority groups were equally high, whereas the Dutch low SES group scored significantly lower. With regard to extrinsic motivation both Dutch groups scored significantly lower than the two minority groups. Between the Ex-colonies and Mediterranean group no significant differences were found. As can be seen from Figure 3, leisure time reading scores also declined over the years, the effect of Time for the total group being significant (see Table 2). The Mediterranean group reported reading at home the most, whereas the Dutch low SES students reported the least reading at home. The comparisons between the groups revealed no significant differences between the two Dutch groups, nor between the Ex-colonies and Mediterranean group. Only between the Dutch low
(JB)SWL II_f16_265-286
9/26/02
9:21 AM
Page 275
Reading skills, motivation, and strategies Decoding Skills
Listening Comprehension 80
28
75
26
70
24
65
22
60
score
score
30
20
55
18
50
16
45 40
14
Reading Comprehension
Reading Vocabulary 50 45
score
40
score
110 105 100 95 90 85 80 75 70 65 60
begin grade 3 end grade 3 end grade 4
end grade 4
begin grade 3
35 30 25 20
begin grade 3
end grade 3
Dutch high SES
end grade 4
Dutch low SES
begin grade 3 end grade 3
Ex-colonies
end grade 4
Mediterranean
Figure 1. Growth of listening comprehension, decoding skills, reading vocabulary and reading comprehension over grades 3 and 4
SES and the minority groups a significant difference was found, the Dutch low SES group reporting significantly less leisure time reading than the minority groups.
Motivations and strategy use of poor and good readers In Table 3 the means and standard deviations on reading motivations and strategy use of poor and good readers are shown, for the Dutch and minority group separately. As can be seen from Table 3, the Dutch good readers reported more leisure time reading, and scored higher on intrinsic motivations than the Dutch poor readers, whereas the poor readers scored significantly higher on extrinsic motivation. Between the minority poor and good readers on the other hand, no significant
(JB)SWL II_f16_265-286
9/26/02
9:21 AM
Page 276
Precursors of Functional Literacy Routine Strategies
Monitoring Strategies 1.6
2.1
1.5
1.9
score
score
2.0
1.4
1.8 1.3
1.7 begin grade 3 end grade 3
2.0
begin grade 3 end grade 3 end grade 4
end grade 4
Estimation Strategies
Text-based Strategies 2.1 2.0
1.8
1.9
score
score
1.9
1.7
1.8 1.7
1.6
1.6
1.5 begin grade 3 end grade 3
Dutch high SES
end grade 4
Dutch low SES
begin grade 3 end grade 3
Ex-colonies
end grade 4
Mediterranean
Figure 2. Use of strategies in grades 3 and 4
correlations on intrinsic and extrinsic motivations, and on leisure time reading were found, although patterns tended to be similar to those of the two Dutch groups. Between the Dutch poor and good readers no significant differences were found on escape motivation, whereas within the minority group the poor readers scored significantly higher on this variable than the good readers. The poor readers of both the Dutch and the minority groups reported more strategy use on all four strategy dimensions than the good readers. There was one exception: the Dutch good readers reported more use of monitoring strategies than the Dutch poor readers did. The difference was not significant, though. The largest differences between poor and good readers were found on the routine strategies scores, the poor readers of both groups scoring significantly higher on this type of strategy than the good readers. The Dutch poor readers reported to use more estimation strategies than the good readers. The poor and good readers of the minority group did not differ on this variable.
(JB)SWL II_f16_265-286
9/26/02
9:21 AM
Page 277
Reading skills, motivation, and strategies Extrinsic Motivation
Intrinsic Motivation 0.9
0.9
0.8 0.7
score
score
0.8
0.6 0.5
0.7
0.4 0.6
0.3
begin grade 3
end grade 3
begin grade 3 end grade 3 end grade 4
end grade 4
Reading at Home
Escape Motivation 24
0.5
22 20
0.4 score
score
18 0.3
16 14 12 10
0.2 begin grade 3 end grade 3
Dutch high SES
end grade 4
Dutch low SES
Nov. 3 Jan. 3 Mar. 3 Nov. 4 Jan. 4 Mar. 4
Ex-colonies
Mediterranean
Figure 3. Motivations and leasure time reading in grades 3 and 4
Intercorrelations among reading skills, strategies, and motivations In Table 4, the partial correlations among the four reading skills, the strategy and motivation dimensions, leisure time reading and nonverbal IQ are displayed, for the Dutch and the minority group separately. For the Dutch group, socioeconomic status (SES) was partialed out, whereas for the minority students, who were all from low SES families, home language was partialed out. The correlations show many similarities between the Dutch and minority group, as well as some interesting differences. As can be seen from Table 4, listening comprehension and reading vocabulary correlated highly with reading comprehension for both groups, while decoding showed moderate correlations with reading comprehension. In general, the strategy scores of both groups correlated negatively with the reading skills. This was particularly true for the routine strategies. The only exception was the relation of
(JB)SWL II_f16_265-286
9/26/02
9:21 AM
Page 278
Precursors of Functional Literacy Table 3. Means and standard deviations on strategy use, motivations, and reading at home for Dutch and minority students who scored low or high on the Reading Comprehension Test at the beginning of Grade 3 Poor comprehenders M DUTCH STUDENTS Strategies Monitoring strategies (max = 27) Routine strategies (max = 18)*** Text-based strategies (max = 9) Estimation strategies (max = 9)** Motivations Intrins. motivation (max = 10)*** Extrins. motivation (max = 6)*** Escape motivation (max = 5) Reading at home (max = 56)** MINORITY STUDENTS Strategies Monitoring strategies (max = 27) Routine strategies (max = 18)*** Text-based strategies (max = 9)* Estimation strategies (max = 9) Motivations Intrinsic motivations (max = 10) Extrinsic motivations (max = 6) Escape motivations (max = 5)*** Reading at home (max = 56)
SD
Good comprehenders M
SD
(84 < n < 103)
(122 < n < 132)
16.52 8.93 5.16 5.76
3.60 2.67 1.55 1.47
17.02 7.56 4.78 5.16
4.18 1.77 1.45 1.40
7.18 3.71 1.98 13.48
2.92 1.62 1.48 10.75
8.46 2.80 1.89 18.08
2.21 1.82 1.40 11.20
(61 < n < 77)
(64 < n < 75)
18.95 10.49 5.94 6.39
3.37 2.55 1.45 1.39
17.83 8.79 5.38 5.96
3.80 2.25 1.26 1.54
7.95 4.69 2.67 16.74
2.17 1.40 1.35 10.47
8.42 4.26 1.77 16.75
2.26 1.71 1.52 11.91
*p < .05; **p < .01; ***p < .001
monitoring strategies with reading skills for the Dutch group; this relation tended to be positive, although not significantly. For the Dutch group, intrinsic motivation and leisure time reading were both positively and significantly related to the four reading skills, whereas extrinsic motivation related negatively to these skills. Escape motivation did not relate to reading skills for the Dutch group. For the minority group, intrinsic nor extrinsic motivations related significantly to reading skills. However, escape motivation and leisure time reading were negatively correlated with reading skills for this group. Between strategies and motivations, significant, positive relations existed, for both the Dutch and minority group. Motivations and leisure time reading were higher correlated with monitoring strategies than with the other strategy dimensions. This was true for both groups, but for the minority students in particular. For both groups, extrinsic motivation was correlated higher with the strategy dimensions than intrinsic motivation.
.36*** .41***
RCom
.17** .22*** .17** .21*** –.19*** –.17** –.17** –.21*** –.11* .01 –.05 –.06 .14** .14** .11* .12* .39*** .18*** .36*** .48***
.03 .03 –.17*** –.24*** –.10* –.11* –.13** –.14**
.68***
RVoc
.37*** .23***
RouS
.12* .01 .27*** .18*** .21*** .12* .19*** .16** .04 –.11*
.49*** .41*** .26***
MonS
EstiS
.05 .05 .22*** .12* .15** .08 .13** .04 .00 –.01
.16**
TxtS
EscM
.08 .21*** .19*** .32*** .11* .15** .13** –.13* –.11*
IntrM ExtrM
Motivations
.09*
Rhom
9:21 AM
.03 –.01 –.21*** –.09 –.06 .01 –.10* –.06
.30*** .66*** .71***
Dec
Strategies
9/26/02
SKILLS Dec RVoc RCom STRATEGIES MonS RouS TxtS EstiS MOTIVATIONS IntrM ExtrM EscM RHOM IQ
ListC
Skills
DUTCH STUDENTS (412 < n < 461)
Table 4. Intercorrelations among student variables for Dutch and Minority children at the beginning of Grade 3, controlled for SES (Dutch group) and home language (Minority group)
(JB)SWL II_f16_265-286 Page 279
Reading skills, motivation, and strategies
(JB)SWL II_f16_265-286
9/26/02
9:21 AM
Page 280
Precursors of Functional Literacy
Conclusions and discussion The development of the four reading skills in the different groups of students showed no surprises. In line with previous research (Droop, 1999; Sijtstra, 1997), the Dutch high SES group showed the highest scores on listening comprehension, reading vocabulary, and reading comprehension. Second came the Dutch low SES group, third the Ex-colonies, and last the Mediterranean group, who lagged behind for about two years. The Ex-colonies students scored consistently higher than the Mediterranean students. This can be explained from the fact that Dutch is being used to a greater extent in the former group. On decoding, the minority groups caught up to their Dutch peers. At the end of Grade 4, the Ex-colonies group outperformed the Dutch low SES group, ending at the same level with the Dutch high SES group. This latter result seems to indicate that the decoding test is the only reading skills test on which insufficient knowledge of the second language does not have a detrimental effect in the long run. Strategy use did not increase or decrease over time. The minority groups reported much more strategy use than the Dutch groups, while the Dutch high SES group, in general, reported the least strategy use. This result is in line with findings of Jiménez et al. (1996), who found that monolingual English students reported less strategy use than their bilingual Latino peers. In the present study, the poor readers reported more strategy use than the good readers. The routine strategies in particular, were very popular among the poor readers. Moreover, all strategy dimensions (except the monitoring strategies for the Dutch group) correlated negatively with the reading skills. These results are in contrast with previous research that has suggested that better readers use strategies more often than poorer readers do (Pressley & Afflerbach, 1995). However, the findings in the present study confirm the results of the study by De Jager and Reezigt (1996), who found poor achievers to use more strategies than good achievers. A possible explanation for these contradictory outcomes may be the difference in conscious and unconscious strategy use (cf. Baker & Brown, 1984). Baker and Brown state that, when good readers read easy texts, they read on their automatic pilot, using their monitoring skills unconsciously. In the present study the students were asked to tell how often they used certain strategies when they had to read a narrative or expository text at school. Schools typically use the same texts for all children in the year group. This means that for the poorer readers these texts usually are difficult, whereas for the better readers they are relatively easy. Thus, it is possible that good readers read these texts using their automatic pilot, employing their monitoring strategies unconsciously. The result may be that better readers report less use of strategies. This may have been the case in the present study as well as in the studies by Jiménez et al. (1996), and De Jager and Reezigt (1996). In studies were readers are provided with texts at their reading levels (Pressley & Afflerbach, 1995), the results may be the other way round; with good readers reporting more strategy use than poor readers. This tentative hypothesis needs further exploration.
–.00 .07 .08 –.05 .05 –.06 –.08 –.16** –.24*** –.10 –.14* –.11 .19*** .31*** .39***
.08 –.06 –.19** –.16** .35***
–.05 –.15** –.14* –.02
–.06 –.13* –.02 –.02
–.05 –.24*** –.13* –.14*
–.12 –.25*** –.16** –.11
RCom
.43*** .47*** .67***
RVoc
.25*** .52*** .64***
Dec
.36*** .22***
RouS
.12*
TxtS
.05 .14* .05 .01 –.10
EstiS
EscM
Rhom
.39*** .17** .21*** .16** .13* .24*** .01 –.01 –.05 –.16**
IntrM ExtrM
Motivations
9:21 AM
.28*** .10 .10 .34*** .19*** .17** .13* .06 .08 .25*** .19** .19** –.05 –.13* –.07
.52*** .35*** .21***
MonS
Strategies
9/26/02
Note. ListC = listening comprehension; Dec = decoding; RVoc = reading vocabulary; RCom = reading comprehension; MonS = monitoring strategies; RouS = routine strategies; TxtS = text-based strategies; EstiS = estimation strategies; IntrM = intrinsic motivations; ExtrM = extrinsic motivations; EscM = escape motivations; RHOM = reading at home; IQ = nonverbal IQ. *p < .05; **p < .01; ***p < .001 (two-tailed)
SKILLS Dec RVoc RCom STRATEGIES MonS RouS TxtS EstiS MOTIVATIONS IntrM ExtrM EscM RHOM IQ
ListC
Skills
MINORITY STUDENTS (282 < n < 329)
(JB)SWL II_f16_265-286 Page 281
Reading skills, motivation, and strategies
(JB)SWL II_f16_265-286
9/26/02
9:21 AM
Page 282
Precursors of Functional Literacy
The fact that minority groups reported much strategy use, suggests they were aware of their comprehension problems. However, it is not only the quantity of strategy use that counts, but rather the selection of the appropriate strategies (Paris et al., 1991). A limitation of the present study is that nothing can be said about the efficiency of the strategies that the children reported to use. It may well be that poorer readers use reading strategies less efficient, and have trouble to decide on which strategy to choose when they encounter comprehension problems. Some support for this view is provided by the finding that the poor readers used much more routine strategies, whereas the good readers tended to use more monitoring strategies. Routine strategies in general require less monitoring and can be performed without mindful thinking. Therefore, they may be less efficient than monitoring strategies, which inherently involve metacognitive thinking. Also the finding that motivations and leisure time reading related stronger to monitoring strategies than to the other strategy dimensions, suggests that children who are motivated, are more inclined to engage in metacognitive thinking about the texts they are reading. Of interest is also the finding that Dutch poor readers, and poor and good minority readers both, reported more estimation strategies than the Dutch good readers. The latter group may be less concerned with the difficulty of the text than the other groups. If children, before reading, have estimated a text to be very difficult or boring, their motivation to read the text mindful, and to use effortful strategies may be affected negatively (cf. Paris et al., 1991). This finding suggests the need for appropriate, interesting texts in reading instruction, especially for children who experience reading difficulties. As expected, reading motivations declined over the years for all groups (cf. McKenna et al., 1995; Otter & Schoonen, 1996a). The Dutch low SES students scored lowest on intrinsic motivation, while the Dutch high SES students scored lowest on extrinsic motivations. At the same time there was no significant difference in their leisure time reading. This suggests that the latter group in their spare time reads for pleasure more than the former group. This is conform research of Baker, Scher, and Mackler (1997), who showed that children of middle-income families are more likely to use literacy as a source of entertainment than are children of low-income families. Contrary to the popular view among many teachers and educators that minority students have lower reading motivation than mainstream students, the minority groups and the Dutch high SES group scored equally high on intrinsic reading motivation. Besides, both minority groups scored higher than the Dutch low SES group on intrinsic as well as extrinsic reading motivations, and reported a greater amount of leisure time reading. This finding is in line with previous research by McKenna et al. (1995), and Baker and Wigfield (1999). Comparing poor and good readers on the motivation dimensions showed that the good Dutch readers scored high on intrinsic motivations, and low on extrinsic motivations, whereas for the poor readers is was exactly the other way round. Within the minority group the same patterns were found, although the differences were not significant. Home variables may have played an important role here. The
(JB)SWL II_f16_265-286
9/26/02
9:21 AM
Page 283
Reading skills, motivation, and strategies
Dutch good readers group consisted of children from both high and low SES families, whereas the minority group consisted of children from low SES families only. Research has shown that middle income parents tend to provide their children with more interesting books (Elliott & Hewison, 1994), and with more informal and playful opportunities for literacy learning (Baker et al., 1997), which makes it more likely for these children to be intrinsically motivated. The finding of Wigfield and Guthrie (1997b) that extrinsic and intrinsic motivation both contribute to free reading was confirmed in this study for both groups of students. For the Dutch group small but significant, positive relations existed between leisure time reading and the four reading skills, a finding in line with previous research by Anderson et al. (1988), and Greany and Hegarty (1987), but in contrast with Otter (1993), who found no relation between reading achievement and reading at home. The finding in this study that reading at home and reading skills showed negative relations for the minority children, at first sight, may suggest that leisure time reading is detrimental to the reading skills of this group of children. A more plausible explanation, however, is that minority children tend to read more in their spare time when they experience reading difficulties. Besides, it is possible that minority parents exert more pressure on children to read at home if they perform poorly at school. The results suggest that minority parents value reading high and want their children to become proficient readers. This is conform research by Leseman (1999), who found that Surinamese mothers in the Netherlands, when they read to their young children, placed more emphasis on skills, and less on reading pleasure than Dutch mothers did. Research in the US has shown similar findings. For example, Stevenson et al. (1990) demonstrated that black and Hispanic mothers placed greater importance on their child’s academic achievement than did white mothers. Conform research by Baker and Wigfield (1999), and Graham (1994), the relations between reading skills and motivations were stronger for the Dutch students than for the minority students. Several researchers (Bock & Moore, 1986; Stevenson et al., 1990) have suggested that in minority schools, where academic outcomes and norms are likely to be low, parents and children may tend to overestimate children’s degree of success in school, and children may think they are doing well, when in fact they are not. Therefore, minority children may be less effective in gaining a realistic concept of their achievement than white children, and thus, may be able to keep up their motivations and self-esteem. As expected, strategy use correlated with motivations for both the Dutch and minority students. Although the correlational data cannot address causality, it seems that motivation has a positive effect on the use of strategies, especially on monitoring strategies. As strategic processing is effortful, changes in motivational factors can influence the course of strategy development (Wigfield & Eccles, 1992). Reversibly, development of competence and strategic behavior can positively affect motivational factors, such as self-efficacy (Alexander et al., 1998). This suggests that motivation may need more attention in instructional programs that aim to improve efficient strategy use.
(JB)SWL II_f16_265-286
9/26/02
9:21 AM
Page 284
Precursors of Functional Literacy
An interesting issue for future research is the use of monitoring strategies and routine strategies among poor and good readers. The present findings suggest that it may be that poor readers benefit from instruction in monitoring strategies rather than routine strategies, and from instruction on ‘how’ and ‘when’ to use these strategies. This may lead to a more effective use, and to a decline in the amount of strategies. A limitation of the study was the use of self-reports for the assessment of strategy use and reading motivations. Self-report measures like questionnaires and diaries may be influenced by social desirability. Studies that used both interview and on-line measures have not revealed a full correspondence between what children say they would do while reading, and what they in fact do (Baker & Brown, 1984). It would have been desirable to use additional observational measures or reports from teachers of parents; however, the large number of participants prevented this. Although providing no final answers, this study does present some interesting starting points from which to conduct more targeted research.
Notes * This study is part of an ongoing research project that has been made possible by a grant of the Dutch Ministry of Education on the advice of Stichting Lezen (Dutch Platform for Reading Promotion).
References Aarnoutse, C. A. J. (1990). Woordenschattest en Leesattitudeschaal [Vocabulary test and reading attitude scales]. Nijmegen: Berkhout. Aarnoutse, C. & Van Leeuwe, J. (1998). Relation between reading comprehension, vocabulary, reading pleasure, and reading frequency. Educational Research and Evaluation, 4, 2, 143–166. Alexander, P. A., Graham, S. & Harris, K. R. (1998). A perspective on strategy research: Progress and prospects. Educational Psychology Review, 10, 2, 129–153. Anderson, R. C., Wilson, P. T. & Fielding, L. G. (1988). Growth in reading and how children spend their time outside of school. Reading Research Quarterly, 285–303. Baker, L. & Brown, A. L. (1984). Metacognitive skills and reading. In P. D. Pearson (Ed.), Handbook of reading research (353–394). New York: Longman. Baker, L. & Wigfield, A. (1999). Dimensions of children’s motivation for reading and their relations to reading activity and reading achievement. Reading Research Quarterly, 34, 4, 452–477. Baker, L., Scher, D. & Mackler, K. (1997). Home and family influences on motivations for reading. Educational Psychologist, 32, 2, 69–82. Bast, J. (1995). The development of individual differences in reading ability. Amsterdam/ Duivendrecht: PI. Bock, R. D. & Moore, E. G. J. (1986). Advantage and disadvantage: A profile of American youth. Hillsdale, NJ: Erlbaum. Durgunoglu, A. & Verhoeven, L. (1998). Literacy development in a multilingual context. Hillsdale, NJ: Erlbaum.
(JB)SWL II_f16_265-286
9/26/02
9:21 AM
Page 285
Reading skills, motivation, and strategies De Jager, B. & Reezigt, G. J. (1996). Onderwijseffectiviteit en metacognitieve vaardigheden [Instructional efficiency and metacognitive abilities] Report No. 95715. Groningen, the Netherlands: GION. Droop, M. (1999). Effects of linguistic and cultural diversity on the development of reading comprehension: A comparative study of Dutch, Turkish and Moroccan children living in the Netherlands. (Thesis) Nijmegen: University of Nijmegen. Elley, W. B. (1992). How in the world do students read? Hamburg: IEA. Elliott, J. A. & Hewison, J. (1994). Comprehension and interest in home reading. British Journal of Educational Psychology, 64, 203–220. Gottfried, A. E. (1990). Academic intrinsic motivation in young elementary school children. Journal of Educational Psychology, 82, 3, 525–538. Graham, S. (1994). Motivation in African Americans. Review of Educational Research, 64, 1, 55–117. Greaney V. & Hegarty, M. (1987). Correlates of leisure-time reading. Journal of Research in Reading, 10, 1, 3–20. Guthrie, J. T., McGough, K., Bennett, L. & Rice, M. E. (1996). Concept-oriented reading instruction: An integrated curriculum to develop motivations and strategies for reading. In L. Baker, P. Afflerbach & D. Reinking (Eds.), Developing engaged readers in school and home communities (165–190). Mahwah NJ: Erlbaum. Guthrie, J. T., Wigfield, A., Metsala, J. L. & Cox, K. E. (1999). Motivational and cognitive predictors of text comprehension and reading amount. Scientific Studies of Reading, 3, 3, 231–256. Hoover, W. A. & Gough, P. B. (1990). The simple view of reading. Reading and Writing, 2, 127–160. Jiménez, R. T., García, G. E. & Pearson, P.D. (1996). The reading strategies of bilingual Latina/o students who are successful English readers: Opportunities and obstacles. Reading Research Quarterly, 31, 1, 90–112. Koda, L. (1996). L2 word recognition research: A critical review. The Modern Language Journal, 80, 450–460. Krom, R. (1992). Listening Comprehension Task [LVS — Luistertoets 1]. Arnhem: CITO. Langer, J. A., Bartolomé, L., Vàsquez, O. & Lucas, T. (1990). Meaning construction in school literacy tasks: A study of bilingual students. American Educational Research Journal, 27, 427–471. Leseman, P. P. M. (1999). Home and school literacy in a multicultural society. In L. Eldering & P. P. M. Leseman (Eds.), Effective early intervention: Cross-cultural perspectives (163–190). New York: Falmer Press. McKenna, M. C., Kear, D. J. & Ellsworth, R. A. (1995). Children’s attitudes toward reading: A national survey. Reading Research Quarterly, 30, 4, 934–956. Mooij, T. (1994). Kenmerken en effecten van methoden voor begrijpend/studerend lezen [Characteristics and effects of reading comprehension methods]. Nijmegen: ITS. Neville, M. H. & Pugh, A. K. (1976/77). Context in reading and listening: Variations in approach to cloze tasks. Reading Research Quarterly, 12, 13–31. Otter, M. E. (1993). Leesvaardigheid, leesonderwijs en buitenschools lezen: Instrumentatie en effecten [Reading proficiency, reading instruction and reading at home: Instruments and effects] Amsterdam: SCO-KI. Otter, M. E. & Schoonen, R. (1996). Aap, Noot, Niets . . . Het spook van de ontlezing in het basisonderwijs [The ghost of aliteracy in elementary schools]. Amsterdam: SCO-KI. Paris, S. G., Wasik, B. A. & Turner, J. C. (1991). The development of strategic readers. In R. Barr, M. L. Kamil, P. B. Mosenthal & P. D. Pearson (Eds.), Handbook of reading research Vol. 2 (609–640). White Plains, NY: Longman. Perfetti, C. A. (1984). Reading ability. New York: Oxford University Press. Piek, K. (1995). Lezen: zoveel lezen we (niet) [Reading: we (do not) read much]. Amsterdam: Stichting Lezen.
(JB)SWL II_f16_265-286
9/26/02
9:21 AM
Page 286
Precursors of Functional Literacy Pintrich, P R. & De Groot, E. V. (1990). Motivational and self-regulated learning components of classroom academic performance. Journal of Educational Psychology, 82, 1, 33–40. Pressley, M. (1998). Comprehension strategies instruction. In J. Osborn & F. Lehr (Eds.), Literacy for all (113–133). New York: Guilford Press. Pressley, M. & Afflerbach, P. (1995). Verbal protocols of reading: The nature of constructively responsive reading. Hillsdale, NJ: Erlbaum. Sijtstra, J. (Ed.) (1992). Balans van het taalonderwijs halverwege de basisschool [First report on language education results in the intermediate grades]. Arnhem, the Netherlands: CITO. Sijtstra, J. (Ed.) (1997). Balans van het taalonderwijs aan het einde van de basisschool 2 [Second report on language education results at the end of Grade 6]. Arnhem, the Netherlands: CITO. Sliepen, S. (1995). Instructie en oefening in begrijpend lezen [Instruction and practice in reading comprehension]. Amsterdam: Paedologisch Instituut. Smits, D. & Aarnoutse, C.A.J. (1997). Een longitudinaal onderzoek naar verschillen in taalen leesprestaties van autochtone en allochtone kinderen [A longitudinal study of differences in language and reading achievement of Dutch and minority children]. Nederlands Tijdschrift voor opvoeding, vorming en onderwijs, 13, 1, 33–52. Stanovich, K. E. (1991). Word recognition: changing perspectives. In R. Barr, M. L. Kamil, P. B. Mosenthal & P. D. Pearson (Eds.), Handbook of reading research. New York: Longman. Stevenson, H. W., Chen, C. & Uttal, D. H. (1990). Beliefs and achievement: A study of black, white, and Hispanic children. Child Development, 61, 508–523. Tellegen, S. & Catsburg, I. (1987). Waarom zou je lezen? [Why read?]. Groningen: WoltersNoordhoff. Verhoeven, L. (1992). Drie-minuten-test [Three Minutes Test]. Arnhem: CITO. Verhoeven, L. & Vermeer, A. (1992). Reading Vocabulary Task [LVS — Leeswoordenschattaak, E5 en E6]. Arnhem, the Netherlands: CITO. Verhoeven, L. & Vermeer, A. (1996). Taalvaardigheid in de bovenbouw [Language ability in the middle and higher grades]. Tilburg: Tilburg University Press. Weber, R. (1991). Linguistic diversity and literacy in American society. In R. Barr, M. L. Kamil, P. Mosenthal & P. D. Pearson (Eds.), Handbook of Reading Research (Vol. 2, 97–119). New York: Longman. Wigfield, A. & Eccles, J. S. (1992). The development of achievement task values: A theoretical analysis. Developmental Review, 12, 265–310. Wigfield, A. & Guthrie, J. T. (1997a). Motivation for reading: An overview. Educational Psychologist, 32, 2, 57–58. Wigfield, A. & Guthrie, J. T. (1997b). Relations of children’s motivation for reading to the amount and breadth of their reading. Journal of Educational Psychology, 89, 3, 420–432. Zwarts, M. (Ed.) (1990). Balans van het taalonderwijs aan het einde van de basisschool. [First report on language education results at the end of Grade 6]. Arnhem, the Netherlands: CITO. Address University of Nijmegen Dept. of Special Education P.O. Box 9104 6500 HE Nijmegen The Netherlands E-mail:
[email protected]
(JB)SWL II_f17_287-302
9/24/02
2:55 PM
Page 287
Bilingualism and reading* Linda S. Siegel University of British Columbia
There are many types of bilingualism and various contexts in which bilingualism occurs. The term bilingualism implies a certain fluency with two (or more) languages. The discussion of bilingualism and reading in this chapter will be limited to a consideration of the research with children and adults who are acquiring reading skills in a language that is not their first language. Most often, learning to read in a bilingual context occurs because individuals are learning to read in the language of the place in which they live but not the language that they speak at home.
The context of bilingualism In major cities in many parts of the world, children grow up speaking two or more languages because of political factors or immigration. In terms of political factors, the area in which children are educated may be bilingual, for example, Quebec, Canada (English and French), Catalan region of Spain (Catalan and Spanish), Hong Kong (Chinese and English). In terms of immigration, their parents may immigrate to a new country but continue to speak the language of the “old country” at home. These children of immigrants speak one language at home and go to school and become educated in a second or sometimes even a third or fourth language. The children themselves may be immigrants. The purpose of this chapter is to discuss reading skills in the context of this type of bilingualism. The designation ESL (English as a second language) refers to a variety of conditions and indicates varying degrees of bilingualism. L1 refers to the language that they speak at home, that is, their first or native language; L2 refers to the language that they are acquiring. In most of the cases discussed in this chapter that language is English. In the case of our research in Canada, much of the bilingualism children who speak English as a second language and come to school and are instructed in English, so they are learning English from the very beginning of their school experience. This is quite a different matter for children who enter into English language instruction having spoken another language for a longer degree of time and having been schooled in another language until the point that they
(JB)SWL II_f17_287-302
9/24/02
2:55 PM
Page 288
Precursors of Functional Literacy
enter into a school in which English is the language of instruction. Obviously, these are some of the many variables that are relevant to the consideration of bilingualism and reading. There are many other variables that are relevant to bilingualism and reading. The social setting in which students learn to read English, the instructional method, the congruence between the learners’ native language and culture and the target language and culture and whether or not the school is an elementary or secondary one are all relevant. In terms of the congruence between languages, language varies on at least two important dimensions; one of these dimensions is from the alphabetic to the non alphabetic, that is, the character recognition system used entirely in Chinese and a part of Japanese and Korean to the alphabetic languages in which sounds are represented by letters or groups of letters, for example, English, French, Spanish, German, etc. The second important dimension is the predictability of graphemephoneme correspondences, that is, how well one can read the language knowing the relationship between letters and their sounds, measured by a concept called orthographic depth. Orthographic depth refers to the differences between alphabetic orthographies in terms of how the graphemes of the writing system and the pronunciation of words can be mapped onto each other. In a shallow orthography, there is a one to one correspondence between letters and sounds; a deep orthography uses a more complex set of relationships between letters and sounds. Even in a shallow orthography for decoding, the language may be deep or non transparent for spelling in that there may be several ways to represent a sound. The sociopolitical context in which the second language is learned is important. For example, there are bilingual students in their home cultures who are learning a foreign language and bilingual immigrants to a new country who must learn a new language. These groups may be different in motivation, socioeconomic status and the value placed on bilingualism. The critical variables in the evaluation of reading skills in a bilingual situation are, among others, the point in the acquisition of L2 when reading occurs, whether or not there has been literacy instruction in L1, the differences and similarities between the syntactic structures, the alphabetic nature of the language, the regularity of grapheme-phoneme correspondences, and the overlap of vocabulary. In this chapter, the relationship between bilingualism and reading will be examined in the context of some of the processes that are significant in the development of reading skills in the English language. These processes are phonological, syntactic awareness, working memory, semantics (or the understanding of meaning), and orthographic awareness.
Phonological processing Phonological processing involves a variety of skills, but in the context of the development of reading skills, one of the most significant is the association of sounds with letters, that is, the understanding of grapheme-phoneme conversion rules
(JB)SWL II_f17_287-302
9/24/02
2:55 PM
Page 289
Bilingualism and reading
and the exceptions to these rules. This skill is the basis of decoding print, and although other routes can be used to obtain meaning from print, the phonological route is clearly an important one and critical in the early development of reading skills (e.g., Jorm, 1979; Stanovich, 1988a, 1988b). English has an irregular spelling system. It is sometimes called orthographically deep in that the mapping of graphemes to phonemes is, at best, unreliable and erratic. Vowels and consonant clusters are difficult and are especially unpredictable. In English, no one-to-one correspondence exists between a letter (or letters) and a sound. The same letter represents different sounds and the same sound may be represented by different letters. Current theories of the development of reading skills in English stress that phonological processing is the most significant underlying cognitive process. Arguments for this position were outlined by Stanovich (1988a, 1988b). This function is referred to as the understanding of grapheme-phoneme conversion rules and because of the irregular nature of the correspondences in English, the learning of these rules is a very complex process. The child who is learning to read must map oral language onto written language by decomposing the word into phonemes and associating each letter (or combination of letters) with these phonemes. The task of the beginning reader is to extract these grapheme-phoneme conversion rules. The alternative is simply to memorize each word as a visual configuration and to associate a meaning with it. This kind of learning may occur, but it is inefficient and makes tremendous demands on visual memory. In an alphabetic language such as English, one of the best measure of phonological processing skills is the reading of pseudowords, that is, pronounceable combinations of letters that can be read by the application of grapheme-phoneme conversion rules, but that are, of course, not real words in English. Examples include pseudowords, such as shum, laip, and cigbet. Pseudowords can be read by anyone who is familiar with the grapheme-phoneme conversion rules of English even though they are not real words and have not been encountered in print or in spoken language before. The development of the ability to read pseudowords has been studied extensively and there is ample evidence that this ability develops during childhood and indicates that children with dyslexia have a great deal of difficulty reading pseudowords (e.g., Bruck, 1988; Ehri & Wilce, 1983; Snowling, 1980; Siegel & Ryan, 1988; and Waters, Bruck & Seidenberg, 1985). A number of studies have shown that disabled readers have more difficulty reading pseudowords than normal readers matched on either chronological age or reading level. (For a review see Rack, Snowling & Olson, 1992). For children learning to read English, the learning of grapheme-phoneme conversion rules is a result of systematic instruction and the extraction of the rules is a result of repeated encounters with print. No evidence is available to provide information about how much of the development of decoding skills is a result of specific instruction in the grapheme-phoneme conversion rules and how much is a result of experience with print. In any case, the understanding of the graphemephoneme conversion rules develops rapidly in the first years of experience with print under normal conditions.
(JB)SWL II_f17_287-302
9/24/02
2:55 PM
Page 290
Precursors of Functional Literacy
The relationship between reading pseudowords in the individual’s first language and the second language is quite strong. In our studies of the correlation between pseudoword reading in two different languages, we have found a high correlation between reading of pseudowords in English and Portuguese (Da Fontoura & Siegel, 1995), between pseudowords in Arabic and pseudowords in English, between pseudowords in English and Hebrew (Abu-Rabia & Siegel, 2001) and between pseudowords in Italian and English (D’Anguilli, Siegel & Serra, 2002). Bilingual children who have difficulty learning to read Portuguese have difficulty reading pseudowords (Da Fontoura & Siegel, 1995) and children with reading difficulties learning Hebrew as a second language also have difficulty with pseudowords (Geva & Siegel, 2000).
Bilingualism and phonological awareness Another important aspect of learning to read is phonological awareness. Phonological awareness is the ability to segment speech into smaller units called phonemes. Phonological awareness transfers from the first to second language (Chiappe & Siegel, 1999; Cisero & Royer, 1995; Durgunoglu, William & Hancin-Bhatt, 1993; Verhoeven, 1994). For example, Chiappe and Siegel found that Punjabi speaking children in the first grade had comparable decoding and phonological skills in English to their native speaking peers. However, their syntactic skills lagged behind the native English speakers. There was some evidence that children who are learning English as a second language showed poorer performance on phonological measures than native English speakers (Wade-Woolley, Chiappe & Siegel, 1998). Wade-Woolley et al. found that ESL children performed more poorly than native English speakers on measures of phonological awareness in kindergarten but not the following year in grade 1. Therefore, the age at which ESL children are assessed may influence whether or not their performance on phonological awareness and other linguistic measures shows impairments relative to native English speakers. Chiappe, Siegel & Gottardo (1999) found that native speaking and ESL children had similar scores on letter identification, spelling and reading tasks in kindergarten but ESL children had lower scores on rhyme detection, phoneme deletion and a rapid naming task. On the rapid naming task, the ESL children named pictures more slowly than the native speakers in November but by May had caught up to their native speaking peers. These studies show that for young bilingual children, their initial difficulties in the phonological areas appear to be resolved after 12–18 months of instruction in their second language. Phonological awareness transfers from one language to another (e.g., Cisero & Royer, 1995; Durgunoglu, William & Hancin-Bhatt, 1993; Geva, Wade-Woolley & Shaney, 1993; Verhoeven, 1994). For example, Verhoeven (1994) found that phonological skills transferred from L1 to L2 while lexical and syntactic processing did not. Durgunoglu et al. (1993) showed that phonological awareness in L1
(JB)SWL II_f17_287-302
9/24/02
2:55 PM
Page 291
Bilingualism and reading
was a significant predictor of word recognition and pseudoword decoding in L2 and also phonological awareness in English was a significant predictor of English word reading ability but oral proficiency in English was not. Chiappe & Siegel (1999) found that L2 children may have comparable skills to L1 children in both phonological awareness and reading in English, despite less developed oral language proficiency. Cisero and Royer (1995) compared monolingual English-speaking and bilingual Spanish-speaking children with limited English proficiency. First grade students were also administered tasks in the opposite language, Spanish for English speakers and English for Spanish speakers. They used measures of phonological awareness, rhyme detection, initial phoneme detection and final phoneme detection. For the initial phoneme task, children heard rat and rib and were asked if these words had the same initial phoneme; for Spanish the stimuli were ven and vid. All students were faster and more accurate on the rhyme task than on the initial phoneme task and slowest and least accurate on the final phoneme task. Each group performed more accurately in their own language. Native English students were slower than the bilingual students on English rhyme, initial phoneme and final phoneme tasks on both languages but had higher levels of accuracy. Both native and second language performance at Time 1 contributed significantly to the prediction of second language performance at Time 2. Wade-Woolley, Chiappe, and Siegel (1998) compared native English speaking children with children who spoke Punjabi at home and entered the schools speaking little or no English. In kindergarten there were no differences between ESL and native English speaking children in their ability to read letters and simple words. There were, however, differences in syntax and vocabulary, sentence repetition, receptive vocabulary and phonological skills, especially rhyme detection, phoneme deletion, rhyme production. In grade 1, however, the differences between the two groups were significantly smaller. The groups did not differ on measures of reading, pseudoword reading and letter identification. They did not differ on phonological tasks including rhyme detection, phoneme deletion and rhyme production. They did not differ on naming speed of simple pictures. There were significant differences in tasks with a significant linguistic component, that is, a working memory and a task requiring the identification of syntactic errors. There were no differences in basic reading skills, that is, word and pseudoword reading. Therefore, the acquisition of the alphabetic principle and early literacy skills may have more to do with instruction and individual differences than with language status because the ESL children as a group were functioning at a level equivalent to chronological age matched mature speakers on tasks involving phonological awareness. Phoneme deletion and phoneme deletion substitution were most strongly associated with word reading for both Punjabi and native English speakers. These results show the strong transfer of phonological skills between languages as a function of bilingualism. In a study of kindergarten ESL and native speakers, Chiappe, Siegel, and
(JB)SWL II_f17_287-302
9/24/02
2:55 PM
Page 292
Precursors of Functional Literacy
Gottardo (1999) found that the ESL and native speakers did not differ on measures of early literacy but the ESL children performed more poorly on measures of phonological processing, syntactic awareness and verbal working memory. Therefore, in young children beginning to learn English, initially there is very little transfer for phonological skills from one language to another.
Reading and spelling There is a large body of evidence that indicates that phonological skills are important in successful reading and spelling acquisition in an individual’s first language (see Vandervelden & Siegel, 1995 for a review). If L2 learners have phonological deficits in their L2 relative to native speakers of that language, these deficits should be manifested in reading and spelling difficulties. If, however, despite these phonological deficits, L2 learners’ reading and spelling abilities equal that of native speakers, then the phonological deficit model should be refined. Children in grade 2 learning English as a second language, who are developing primary literacy skills in their second language were compared to native speakers (NS) on their ability to achieve accurate English spellings (Wade-Woolley & Siegel, 1997). The children in the study were grade 2 students who did not speak English at home. The languages that they spoke at home included Cantonese, Mandarin, Gujarati, Urdu and Punjabi. Most were born in Canada and had attended school since the age of 4. Junior and senior kindergarten in Ontario, Canada are half-days so they had two half-day years and two full-day years as the study was conducted toward the end of grade 2. The children were grouped into good and poor readers on the basis of their word recognition skills as measured by the Wide Range Achievement Test (WRAT-3). ESL students had lower scores than the native speakers on pseudoword repetition, phoneme deletion (for example, if you take away /p/ from pink, what word is left?), on two measures of syntax and the poor readers had lower scores than the good readers. The two measures of syntax were the Oral Cloze task, in which the children were required to insert the missing word in a sentence, e.g., “Fred put the turkey the oven”, that they heard orally and a syntactic judgment task where they heard well-formed sentences, e.g., “The boy was chased by the dog”, and ill-formed sentences, e.g., “The tall thin man playing was basketball” and had to say whether or not it was a good sentence. There were no differences in the spelling scores of the native-speaking and the ESL group although the poor readers had significantly lower scores than the good readers. The native-speakers could represent tense vowels equally well through spelling. Tense vowels need some alteration of the traditional spelling to indicate their pronunciation. For example, the vowel ea represents the sound /ii/ so tense vowels were not more difficult for the ESL group although they were more difficult for the poor reading group. Poor readers spelled words with consonant clusters
(JB)SWL II_f17_287-302
9/24/02
2:55 PM
Page 293
Bilingualism and reading
more poorly than those without clusters, but average readers, whether they were ESL or native-speakers, did not show a difference between these two types of words. The accuracy of real word spelling was predicted by pseudoword decoding and phoneme deletion in similar proportions for ESL and native-speaking groups. Pseudoword decoding was the sole predictor for pseudoword spelling for both language groups. On the basis of these results it appears that spelling does not appear to reflect differences in oral language competence, but does reflect individual differences unrelated to whether children are learning to read and spell in their first or second language, although the ESL speakers were somewhat less sensitive to the syntactic and semantic features of English. The ESL group was poorer than the native-speaking group on two different phonological tasks. The pseudoword repetition task requires the accurate perception and production of non-native language speech sounds while the phoneme deletion task demanded that the children show the ability to sequence and segment phonemes from their second language and perform complex operations. The results suggest that ESL children have not yet fully acquired the sound system of the second language and yet, their reading and spelling was at a level equivalent to their native speaking peers. Language status was not a dominant factor in spelling performance. Reading skills were more significantly correlated with spelling more than first language. Average readers were more accurate than poor readers but the process of acquiring a second language does not appear to influence the spelling ability of ESL children. The implication of different patterns of poor readers and ESL speakers on phonological tasks presents an interesting challenge to the phonological core deficit model of reading. One possibility is that the phonological deficit applies to reading disability in native speakers only and the failure of non-native speakers of English who have a reading failure must be attributed to some other cause. Another possibility is that the concept of phonemic awareness must be refined and that there may be more isolated phonological processes. It also may be that orthographic or visual memory processes may be useful in English in terms of understanding how print is translated into sound and meaning. Geva, Wade-Woolley, and Shaney (1993) examined whether individual difference factors or factors related to the orthographic complexity of the language were the determinant of reading and spelling. Hebrew orthography is shallow, that is the sounds of the letters are predictable. The hypothesis that differences between first language (L1) and second language (L2) reading and spelling profiles could be accounted for by lack of proficiency in the L2 or differences in orthographic complexity of the two languages was explored in a longitudinal study of 45 children acquiring reading and spelling skills concurrently in English (L1) and Hebrew (L2). The children were tested in grades 1 and 2 on literacy measures in both languages. The less complex Hebrew orthography facilitated subjects’ decoding performance, but failed to maintain that facilitation in spelling. The findings showed that, despite L1–L2 differences in orthographic complexity and language
(JB)SWL II_f17_287-302
9/24/02
2:55 PM
Page 294
Precursors of Functional Literacy
proficiency, the profiles of emergent spelling in both languages are strikingly similar. The rate of acquisition of conventional spelling, however, differentiates L1 from L2 performance. Geva et al. (1993) note that word frequency in one language, which is a determinant of word reading in that language, does not necessarily mean that the same word frequency applies to another language. For the less proficient L2 learner, there may be no difference between decoding an unfamiliar word and decoding a pseudoword because there are many words that are not known. There was strong within language correlation, that is, correlation between pseudoword and word reading within each language. A phoneme deletion task predicted reading and spelling skills in English. Chiappe, Siegel, and Gottardo (1999) found that the ESL children showed comparable performance to native English speakers on measures of letter identification, word recognition and spelling. However, there were differences in phonological processing and on measures which required interpretation and manipulation of the language. These differences were still evident at the end of kindergarten. These differences might be expected since they require a new phonology with a new phonemic contrast. Similar results have been found in other studies (Gholamein & Geva, 1999; Wade-Woolley & Siegel, 1997). However, in spite of slower word retrieval in November, ESL children named pictures as rapidly as native speakers in May. There were, however, differences in syntactic awareness and verbal working memory throughout kindergarten. However, the groups did differ in ability to understand and manipulate language (Wade-Woolley, Chiappe & Siegel, 1998). The children who spoke Indian languages showed a disadvantage relative to native speakers on phonological tasks in kindergarten. This might be expected since they were acquiring a new phonology with new phonemic contrasts. Similar results have been found in other studies (Gholamein & Geva, in press; Wade-Woolley & Siegel, 1997). These findings are consistent with a different sample of English and Punjabi speaking children in grade 1 (Chiappe & Siegel, 1999).
The special case of vowels Bilingualism provides an interesting context to study the acquisition of a second language, in this case English. English vowels tend to have more complex and irregular pronunciations than English consonants. English vowels have the property that their pronunciation can change depending on the context. An example is the rule that an e at the end of a word usually makes the vowel long. The grapheme-phoneme correspondences of English vowels are very unpredictable. Consequently, misreadings of vowels occur more frequently than misreadings of consonants (e.g., Fowler, Shankweiler & Liberman, 1979; Weber, 1970). In languages other than English, vowels have more regular patterns with fewer representations of each vowel sound. One such language is Hebrew, in which the orthography is transparent, that is, the grapheme-phoneme conversion rules are
(JB)SWL II_f17_287-302
9/24/02
2:55 PM
Page 295
Bilingualism and reading
predictable. Although vowels are omitted from text for older children and adults, in the beginning stages of reading vowels are included. Children learning to read both English and Hebrew can be tested to compare these two very different orthographies. In a comparison of English speaking children learning to read Hebrew as a second language, Geva and Siegel (2000) found that the incidence of errors in reading vowels was significantly higher in English than in Hebrew. Other children who had reading disabilities (in both languages) made many vowel errors in English but very few in Hebrew. Younger children with reading disabilities made vowel errors in both languages. It should be noted that Francophone children learning French, a more transparent language than English, rarely experience difficulties with vowels. However, other types of errors were more common in Hebrew. Hebrew has many visually similar letters and more errors were made involving visually confusable letters in Hebrew than in English. In addition, because Hebrew has a transparent orthography one can decode it syllable by syllable and pronounce it properly and read the word without the proper stress. Failure to read the word with the stress on the correct syllable was more common in Hebrew than English. In English, a syllable by syllable decoding would usually result in vowel errors (e.g., pronouncing the vowel as a short vowel when the word ends in e and perhaps even pronouncing the final silent e.) Order errors, in which a consonant was omitted or the order of the consonants was confused, were more common in English than Hebrew, possibly because Hebrew words can be decoded in a linear manner from right to left and the linear strategy does not always work successfully in English. In general, these results indicate that the structure of the language and its alphabetic system as well as individual difference variables are significantly related to the acquisition of literacy skills in a bilingual context.
Syntactic awareness and bilingualism Syntactic awareness, also called grammatical sensitivity, refers to the ability to understand the syntax of the language. Syntactic awareness is the ability to understand the basic grammatical structure of the language in question. Siegel and Ryan (1988) have investigated the development of these skills in an Oral Cloze task. In the Oral Cloze task, a sentence is read aloud to the child and the child is required to fill in the missing word. Examples of are: “Jane her sister ran up the hill”; “Betty a hole with her shovel”; “The girl is tall plays basketball.” This ability appears to be critical for fluent and efficient reading of text, and it requires making predictions about the words that come next in the sequence. Syntactic factors may influence the difficulty of reading single words, such as function words, prepositions, and auxiliary verbs, which are difficult to integrate in a semantic network. Ehri and Wilce (1980) have shown that beginning readers acquire information about the syntactic properties of function words when they have been trained to read these words in the context of a sentence. Therefore, the ability to process syntax may be an important aspect of word learning. Syntactic awareness appears to have a relationship to good reading skills in a
(JB)SWL II_f17_287-302
9/24/02
2:55 PM
Page 296
Precursors of Functional Literacy
variety of languages that have been studied. Children with ESL have deficits in syntactic awareness, e.g., (Bentin, Deutsch & Libermann, 1990; Da Fontoura & Siegel, 1995; So & Siegel, 1997). The ESL children had lower scores than the normal speaking children on the oral cloze and memory sentence tasks (Chiappe, Siegel & Gottardo, 1999). For example, Da Fontoura and Siegel studied Canadian children who spoke Portuguese as a first language, received instruction in reading in English, and attended a Heritage Language Program in Portuguese. The children who had low scores on Portuguese word and pseudoword reading tests had significantly lower scores on both Portuguese and English oral cloze tasks than children who were good readers of Portuguese. In both English and Portuguese, syntactic skills are significantly correlated with word and pseudoword reading skills.
Working memory and bilingualism Working memory refers to the retention of information in short-term storage while processing incoming information and retrieving information from longterm storage. Working memory is relevant to reading because the reader must decode and/or recognize words while remembering what has been read and retrieving information such as grapheme-phoneme conversion rules. Working memory may also be critical to the reading of individual words particularly in the beginning of the acquisition of word reading skills because the grapheme-phoneme conversion rules for each segment of the word must be held in memory while the remaining segments of the word are processed. Longer words, in terms of the number of syllables, place increasing demands on working memory. In addition, the complexity of a particular rule will influence the difficulty of word recognition because the number of possible alternative grapheme-phoneme pronunciations may have an influence on ease or difficulty of reading a particular word. Given more alternative pronunciations, reading will be slower and less accurate until the individual items are mastered. More rules must be searched and applied to the word being read. For example, c and g have multiple pronunciations at the beginning of English words, and, therefore, words or pseudowords starting with these letters may be more difficult than words or pseudowords beginning with other letters, especially for beginning readers. Siegel and Ryan (1989) studied working memory in normal and disabled readers and dyslexics, using a task based on one developed by Daneman and Carpenter (1980). In the modified version of this task, the child is read aloud 2, 3, 4, or 5 sentences and is asked to fill in a missing word at the end of each sentence. The child is then required to remember the missing words. Examples are: “In the summer it is very . People go to see monkeys in a . With dinner we sometimes eat bread and .” The child was then required to repeat the three words that he or she selected in the order of presentation of the sentences. The disabled readers performed significantly more poorly than the normal readers on this task, indicating significant difficulties with working memory in the disabled
(JB)SWL II_f17_287-302
9/24/02
2:55 PM
Page 297
Bilingualism and reading
readers. Similar difficulties with working memory have been noted in Chinese (So & Siegel, 1997), Hebrew (Geva & Siegel, 1991), and Portuguese (Da Fontoura & Siegel, 1995). Da Fontoura and Siegel (1995) found that verbal working memory and the syntactic skills of Portuguese English bilingual children who had been born in Canada still lagged behind in skills of English monolingual children in grades 4, 5, and 6.
Orthographic awareness Another aspect of reading is orthographic processing. Orthographic processing refers to the understanding of the writing conventions of the language in question and knowledge of the correct and incorrect spellings of words. All alphabetic systems include legal and illegal and more and less probable sequences of letters, and a fluent reader uses knowledge of these sequences to some extent. Positional constraints and probabilities that letters will occur in certain positions are additional aspects of orthographic knowledge used by the skilled reader. Siegel, Share, and Geva (1995) developed a task to assess the awareness of orthographic structure. Children were shown 17 pairs of pronounceable pseudowords, one containing a bigram that never occurs in an English word in a particular position and the other containing a bigram that occurs in English. Examples are filv-filk, moke-moje, vism-visn, and powl-lowp. This task was administered to disabled and normal readers, aged 7–16 years. However, when matched on reading level, the disabled readers performed at a significantly higher level than the normal readers. The orthographic processing of the reading disabled is quite good. These data indicate that orthographic processing is not as impaired in dyslexics as is phonological processing. Abu-Rabia and Siegel (2001) found that orthographic process did not show any transfer between Arabic and English in bilingual speakers and that orthographic skills did not correlate significantly with reading skills. Orthography influences reading skills. An interesting case is Hebrew-English bilinguals. Hebrew is read for right to left, English from left to right. Pollatsek, Bolozky, Well & Rayner (1981) studied Hebrew-English bilinguals and found that they perceived more letters to the left of the fixation point when reading Hebrew and more to the right of the fixation point when reading English. Henderson (1983) studied the reading of English by bilingual speakers and found that the native Spanish speakers read more slowly than native English speakers but Arabic speakers who use a different orthography read much more slowly than the Spanish bilinguals. Brown and Haynes (1985) presented bilinguals with tasks involving matching pairs of English words. Spanish speakers were faster than Arabic but Japanese were the fastest, probably because of the visual processing required in Japanese. However, when the determination of whether words (or nonwords) were the same or different the Japanese were the slowest of all of the groups.
(JB)SWL II_f17_287-302
9/24/02
2:55 PM
Page 298
Precursors of Functional Literacy
Transfer between languages Transfer between the two languages can be either positive, that is one language enhancing another, negative, one language interfering with the other, or neutral in that the two languages have no influence on each other. There are two contrasting theories which address the issues of the transfer of reading language between two languages; the interdependence hypothesis and the script dependent hypothesis. The interdependence hypothesis predicts that similar difficulties will appear in both languages owing to a central processing deficit. In regard to the acquisition of reading in a second language, the model of Cummins (1981) assumes a “common underlying proficiency”. That is, there is a transfer of knowledge of skills from native language orality or literacy to ESL literacy. A common set of proficiencies underlies both languages. In contrast, the script dependent hypothesis predicts that grapheme phoneme irregularities in English will result in greater difficulties for individuals who speak languages with more predictable grapheme correspondences than English. Literacy skills do seem to transfer from one language to another. Shimron and Savan (1994) found that skilled Hebrew speakers read passages in English faster and with greater comprehension than they did equivalent passages in Hebrew. Some differences in orthography may explain why English is easier to read than Hebrew in that Hebrew words are longer and grammatical features are coded in the words making understanding of a sentence more difficult until all the words have been read. In a study of relationships among languages, bilingual Hebrew-English speaking children in Israel, Abu-Rabia (1995) found that there was a significant correlation between Hebrew and English skills except for phonological and orthographic tasks. A transfer of linguistic skills is likely to occur but there are some language dependent variables. Abu-Rabia (1995) administered tests of working memory, oral cloze, a visual task, which involved recognizing which of two words was spelled correctly, a homophone and a pseudohomophone, for example brane-brain, a phonological test in which the participants had to decide which of two pseudowords sounded like a real word, joak or joap, an orthographic test in which one member of a pair contained an orthographically illegal but pronounceable pseudoword and the other word contained an orthographically legal pseudoword, for example filv and filk, and Word Attack and Word Recognition tests. There were significant correlations between all the tests in Hebrew and English except for the phonological and orthographic tests. Children who were less skilled in Hebrew were also less skilled in English. The correlations between reading skills in English and Hebrew were similar to those found between French and Arabic (Wagner, Spratt & Ezzaki, 1989). The lack of correlation between orthographic skills in English and Hebrew suggest that they do not transfer from one language to another. Barry (1992) studied English-Welsh bilinguals. Welsh is a very regular language with a predictable correspondence between grapheme and phonemes. There
(JB)SWL II_f17_287-302
9/24/02
2:55 PM
Page 299
Bilingualism and reading
are some slight deviations and some context sensitive rules, for example, the letter y can be pronounced differently depending on its position in the word. In some parts of Wales, the letters i, u, and y are pronounced in the same way in the final syllable of the word. Bilingual Welsh-English speakers were presented with words that primed the particular spelling using the i, u, or y so they were sensitive to these priming effects. They were then presented with English words, priming a certain spelling. There were some priming effects in English. They were also required to spell nonwords. The Welsh speakers were very similar to monolingual English speakers in a task that involved the spelling of nonwords. There is additional evidence to support the interdependence hypothesis. Skutnabb-Kangas and Toukomaa (1976) found that Finnish immigrant children in Sweden learned their second language, Swedish, in relationship to their proficiency in Finnish at the time that they started learning Swedish. Durgunoglu et al. (1993) found that for Spanish speaking grade 1 students learning English, that performance on English word and pseudoword recognition tests was predicted by levels of Spanish phonological awareness and Spanish word recognition. Verhoeven (1990) studied Turkish children going to school in the Netherlands, found that there was no transfer of lexicon and syntax from one language to another. For Phonological skills, there was some moderate transfer and pragmatic skills showed strong positive transfer. Verhoeven (1994) studied Turkish children in the Netherlands and found that the type of transfer (positive or negative) from one language to the other depended on the type of skill in question. For vocabulary and syntax, transfer was quite limited but for pragmatic, phonological and reading skills there was significant positive transfer. Verhoeven used language samples and sentence completion tasks to measure grammatical abilities. The phonological task involved the discrimination between same and different phonemes. Vocabulary was larger in Turkish than in Dutch. There was a positive transfer in phonological skills between languages. Verhoeven speculates that this may be due to the high level of metalinguistic awareness required by the phonological awareness task. The transfer of reading skills in the two languages indicates that there may be non-specific language processes related to phonological and higher level reading skills. Vocabulary and syntax appear to be more language specific with little transfer between languages, especially when they are quite different in structure as are Turkish and Dutch. Pragmatic skills were similar in the two languages and there was a positive transfer from reading skills between the two languages. Brown and Haynes (1985) compared Japanese, Arabic and Spanish speaking students learning reading in English. The Spanish and English share a common alphabet and a similar orthographic system. Arabic writing system is alphabetic, it uses a different set of letters and is read from right to left. Japanese is syllabic and logographic and differs from English and the relationship between writing and speech. They used a task that involved matching whether words or pseudowords were the same or different using Roman alphabetic patterns. The Spanish speaking students, not surprisingly, were slower than the Arabic students, the
(JB)SWL II_f17_287-302
9/24/02
2:55 PM
Page 300
Precursors of Functional Literacy
Japanese were the fastest and the most accurate indicating that visual processes are important. When abstract, complex figures were used, the Japanese also performed better than the Spanish and Arabic-speaking student whose scores were essentially the same. The Japanese students did show efficiency in visual processing. In a comparison of reading of pseudowords, the Japanese students were not as fast or accurate in reading pseudowords as the Spanish or Arabic-speaking. Therefore, the visual discrimination superiority of the Japanese was not related to understanding of grapheme-phoneme use and understanding of graphemephoneme correspondence, but the Japanese students tended to rely more on sight word knowledge and less on rule governed letter to sound correspondence than the Arabic and Spanish-speaking. Word length was a critical factor for the Japanese. For the Spanish and Arabic students, listening and reading were highly correlated, but this was not true for the Japanese. These may represent separate skill domains. The Japanese students showed smaller differences between words and pseudowords and between pseudowords and nonsense strings than the other groups. This is in same difference matching, again, suggesting that they are relying more on visual or orthographic visual strategies than phonological ones. Accuracy in reading was related in the Japanese groups as in the other groups, to sensitivity to grapheme-phoneme correspondences.
Conclusion Bilingualism, especially if the two languages are acquired at an early age, clearly does not impede the development of reading skills. It may be an advantage to learn two languages in that there is some evidence of positive transfer from one language to another, particularly in the case of more regular orthographies to English.
Notes * The preparation of this manuscript was partially supported by a grant from the Natural Sciences and Engineering Research Council of Canada. The author wishes to thank Sarah Kontopoulos and Stephanie Vyas for secretarial assistance.
References Abu-Rabia, S. (1995). Attitudes and cultural background and their relationship to English in a multicultural social context: the case of male and female Arab immigrants in Canada. Journal of Educational Psychology, 15, 323–335. Abu-Rabia, S. & Siegel, L. S. (2001). Reading, writing, orthographic, phonological, syntactic and memory skills of bilingual Arabic-English speaking Arab children. Unpublished manuscript. Abu-Rabia, S. & Siegel, L. S. (1995). Different orthographies, different context effects: The
(JB)SWL II_f17_287-302
9/24/02
2:55 PM
Page 301
Bilingualism and reading effects of Arabic sentence context in skilled and poor readers. Reading Psychology, 16, 1–19. Barry, C. (1992). Interactions between lexical and assembled spelling (in English, Italian, and Welsh). In C. Sterling & C. Robinson (Eds.), Psychology, Spelling & Education, 71–86. Clevedon, England: Multilingual Malteus Ltd. Bentin, S., Deutsch, A. & Liberman, I. Y. (1990). The development of cross language transfer of phonological awareness. Contemporary Educational Psychology, 20, 275–303. Brown, T. L. & Haynes, M. (1985). Literacy background and reading development in a second language. In T. H. Carr (Ed.) The Development of Reading Skills: New Directions for Child Development, no. 27, 19–34. San Francisco: Jossey-Bass. Bruck, M. (1988). The word recognition and spelling of dyslexia children. Reading Research Quarterly, 23, 51–68. Chiappe, P. & Siegel, L. S. (1999). Phonological awareness and reading acquisition in English and Punjabi-speaking Canadian children. Journal of Educational Psychology, 91, 20–28. Chiappe, P., Siegel, L. S. & Gottardo, A. (1999). Linguistic diversity and the identification of children at-risk for reading disability. Cisero, C. A. & Royer, J. M. (1995). The development of cross language transfer of phonological awareness. Journal of Contemporary Educational Psychology, 20, 275–303. Cummins, J. (1981). The role of primary language development in promoting educational success for language minority students. In California State Department of Education, Schooling and language minority students: A theoretical framework. Evaluating, Dissemination and Assessment Center, California State University, Los Angeles. D’Anguilli A., Siegel, L. S., & Serra, E. (2002). The development of reading in English and Italian in bilingual children. Applied Psycholinguistics, 22, 479–507. Da Fontoura, H. A. & Siegel, L. S. (1995). Reading, syntactic and memory skills of bilingual Portuguese-English Canadian children. Reading and Writing: An Interdisciplinary Journal, 7, 139–153. Daneman, M. & Carpenter, P .A. (1980). Individual differences in working memory and reading. Journal of Verbal Learning and Verbal Behavior, 19, 450–466. Durgunoglu, A. Y., William, E. N., Hancin-Bhatt, B. J. (1993). Cross- language transfer of phonological awareness. Journal of Educational Psychology, 85, 453–465. Ehri, L. C. & Wilce, L. S. (1980). The influence of orthography on readers’ conceptualization of the phonemic of words. Applied Psycholinguistics, 1, 371–385. Ehri, L. C. & Wilce, L. S. (1983). Development of word identification speed in skilled and less-skilled beginning readers. Journal of Educational Psychology, 75, 3–18. Fowler, C., Shankweiler, D. & Liberman, I. (1979). Apprehending spelling patterns for vowels: A developmental study. Language and Speech, 22, 243–251. Geva, E. & Siegel, L. S. (2000). Orthographic and cognitive factors in the concurrent development of basic reading skills in two languages. Reading and Writing: An Interdisciplinary Journal, 12, 1–30. Geva, E., Wade-Woolley, L. & Shaney, M. (1993). The concurrent development of spelling and decoding in different orthographies. Journal of Reading Behavior, 25, 383–406. Gholamain, M. & Geva, E. (1999). Orthographic and cognitive factors in the concurrent development of basic reading skills in English and Persian. Language-Learning, 49, 2, 183–217. Henderson, R. T. (1983). Basic Perceptual and Cognitive Processes Employed by ArabicSpeaking Students in the Development of Reading Skills in English as a Second Language. Unpublished Dissertation, University of Pittsburgh. Jorm, A. F. (1979). The cognitive and neurological basis of developmental dyslexia: A theoretical framework and review. Cognition, 7, 19–33. Pollatsek, A., Bolozky, S., Well, A. & Rayner, K. (1981). Asymmetries in the perceptual span for Israeli readers. Brain and Language, 14, 174–180.
(JB)SWL II_f17_287-302
9/24/02
2:55 PM
Page 302
Precursors of Functional Literacy Rack, J. P., Snowling, M. & Olson, R. (1992). The nonword reading deficit in developmental dyslexia: A review. Reading Research Quarterly, 27, 28–53. Shimron, J. & Savan, T. (1994). Reading proficiency and orthography: Evidence from Hebrew and English. Language Learning, 44, 5–27. Siegel, L. S. & Ryan, E. B. (1988). Development of grammatical sensitivity, phonological, and short-term memory in normally achieving and learning disabled children. Developmental Psychology, 24, 28–37. Siegel, L. S. & Ryan, E. B. (1989). The development of working memory in normally achieving and subtypes of learning disabled children. Child Development, 60, 973–980. Siegel, L. S., Share, D. & Geva, E. (1995). Evidence for superior orthographic skills in dyslexics. Psychological Science, 6, 250–254. Skutnabb-Kangas, T. & Toukomaa, P. (1976). Teaching migrant children their mother tongue and learning the language of the host county in the context of the sociocultural situation of the migrant family. Tampere (Finland): University of Tampere. Snowling, M. J. (1980). The development of grapheme–phoneme correspondence in normal and dyslexic readers. Journal of Experimental Child Psychology, 29, 294–305. So, D., & Siegel, L. S. (1997). Learning to read Chinese: Semantic, syntactic, phonological and working memory skills in normally achieving and poor Chinese readers. Reading and Writing: An Interdisciplinary Journal, 9, 1–21. Stanovich, K. E. (1988a). Explaining the differences between the dyslexic and garden variety poor reader: The phonological-core variance-difference model. Journal of Learning Disabilities, 21, 590–604, 612. Stanovich, K. E. (1988b). The right and wrong places to look for the cognitive locus of reading disability. Annals of Dyslexia, 38, 154–177. Vandervelden, M. C. & Siegel, L. S. (1995). Phonological recoding and phonemic awareness in early literacy: A developmental approach. Reading Research Quarterly, 30, 854–875. Verhoeven, L. (1994). Transfer in bilingual development: The linguistic interdependence hypothesis revisited. Language Learning, 44, 381–415. Verhoeven, L. T. (1990). Acquisition of reading in a second language. Reading Research Quarterly, 25, 90–114. Wade-Woolley, L. Chiappe, P. & Siegel, L. S. (1998). Learning to read in a second language: does phonological awareness really matter? Journal of Educational Psychology Wade-Woolley, L. & Siegel, L. S. (1997). The spelling performance of ESL & Native speakers of English as a function of reading skill. Reading and Writing: An Interdisciplinary Journal, 9, 387–406. Wagner, D., Spratt, J. E. & Ezzaki, A. (1989). Does learning to read in a second language always put the child at a disadvantage? Counterevidence from Morocco. Applied Psycholinguistics, 10, 31–48. Waters, G., Bruck, M. & Seidenberg, M. (1985). Cognitive Processes in reading and spelling. Journal of Experimental Child Psychology, 39, 511–530. Weber, R. (1970). A linguistic analysis of first–grade reading errors. Reading Research Quarterly, 5, 427–451. Address University of British Columbia 2125 Main Mall Vancouver, British Columbia Canada V6T 1Z4 Telephone: (604) 822-1893 Fax: (604) 822-3302 email:
[email protected]
(JB)SWL II_f18_303-315
9/26/02
9:22 AM
Page 303
Age and gender differences in reading engagement Liliane Kjellman1 Cygnaeus School and Åbo Akademi University
How can we motivate young students to become successful readers? Previous studies on this topic document the link between motivation and achievement and further indicate a need to increase our understanding of how children acquire the motivation to develop into active, engaged readers (Elley, 1992; Guthrie, Schafer, Wang & Afflerbach, 1993; Purves & Beach, 1972; Walberg & Tsai, 1985; Wixson & Lipson, 1991). In the world-wide IEA study of reading carried out in 1990–1991, Finnish 9-year olds were the best readers on all three dimensions of reading literacy assessed (Elley, 1992). Assuming a link between achievement and motivation, Finnish students should in general be highly motivated readers with positive attitudes towards reading. However, some pupils seem to have a weak commitment to reading, despite high achievement. The purpose of this paper is to analyse the results from a reading survey conducted among third and fifth graders in a Finland-Swedish context measuring their motivation for reading. Two fundamental components of motivation: self-concept and task value, will be assessed.
Reading engagement In order to get a complete picture of an individual’s participation in reading, we need to understand motivational as well as cognitive processes involved in reading (Wigfield & Guthrie, 1997). The reading engagement perspective, emerging in current research on reading motivation, includes both the cognitive and the motivational aspects of reading (see Baker, Afflerbach & Reinking, 1996; Guthrie, 1996; Oldfather & Wigfield, 1996). Engaged readers can be seen as motivated, strategic, knowledgeable, and socially interactive (see Baker, Afflerbach & Reinking, 1996; Guthrie et al., 1996, for a discussion of the engagement perspective).
Self-concept as a reader Training students to be more efficacious and to believe they are more efficacious improves children’s achievement in different subject areas such as math and reading
(JB)SWL II_f18_303-315
9/26/02
9:22 AM
Page 304
Precursors of Functional Literacy
(see e.g., Schunk & Rice, 1993). In a reading situation this might mean that students who think they can figure out the difficult words in a text are also more likely to show more activity, effort and persistence in order to figure out the difficult words than students with lower efficacy expectations. An important implication of the work on ability and efficacy beliefs for motivation for reading is that when children believe they are competent readers and efficacious at reading, they are more likely to become engaged readers (Wigfield et al., 1996). Several studies have shown that children’s competence perceptions decrease through the grades, at least across the elementary school years (see Eccles, Midgley & Adler, 1994; Stipek & Mac Iver, 1989) In line with this we could assume that a decline in student’s selfconcept as readers might show also in the present study as a difference between grade 3 and grade 5 students.
Value of reading Winne, (1985) views the “idealized reader” as one who feels competent and perceives reading both as being of personal value as well as practical importance. Ford (1992), Henk and Melnick (1995) and Wigfield (1994) support the notion that high motivation to read goes together with positive self-concept as a reader and in opposite way that low motivation to read is associated with poor self-concept and low value assignment. The reading survey by Gambrell, Palmer, Codling, and Mazzoni (1996), used in this study, assesses these two dimensions of reading. Earlier research has indicated a decline in motivation as children proceed through school and that this declines is due to changes in school and classroom environments (see Wigfield, Eccles & Pintrich, 1996, for review). Statistically significant differences were found using the Reading Survey between mean scores of thirdand fifth-grade students on the value measure, with younger students scoring more positively than older students (Gambrell et al., 1996). These findings are in line with earlier research showing that attitude toward reading decreases as children progress through the elementary grades (e.g., Mc Kenna & Kear, 1990). On the basis of earlier research we can expect differences in motivation between age groups, at least considering the value measure also in this study. In earlier studies significant gender differences concerning task values has been found in the domain of reading (Eccles et al., 1993) and can therefore be predicted also for this study.
A Finland-Swedish context The sample schools are all schools where the language of instruction is Swedish, situated in urban areas with a Finnish-speaking majority. The students often come from bilingual homes due to the increasing number of mixed marriages between Swedish-speakers and Finnish-speakers. For many students the only daily contact with the Swedish language is in school, spending all their leisure time speaking Finnish. There are also some students of monolingual pupils speaking Finnish at home attending Swedish-speaking schools (Brunell, 1992, 1995).
(JB)SWL II_f18_303-315
9/26/02
9:22 AM
Page 305
Age and gender differences in reading engagement
Parents mixing Swedish and Finnish in their communication with the child might lower reading comprehension, however a balanced bilingualism combined with a Swedish identity can be an advantage for high literacy. However a recent study (Brunell & Saretsalo, 1999) showed a low relation between students’ linguistic background and reading literacy. The aim of the present study is to extend existing research on reading motivation into another linguistic and cultural setting. However, there is no reason to expect that the Swedish-Finnish context would give rise to results that are any different from results from other countries. The purpose in the present study is to analyse the results from a reading survey taking into consideration age and gender aspects. Based on earlier research the hypotheses for this study is that third grade students are likely to value reading more than fifth grade students and that girls generally value reading more than boys.
Method Participants A total of 298 third and fifth grade students from four Swedish-speaking primary schools participated in this study. All schools were situated in urban areas and a large percent of the students are bilingual (Swedish-Finnish). Swedish is one of the two official languages in Finland and a minority (about 6%) speak Swedish as their native tongue. Two schools have age homogenous classes and two have multiage-classes. The chosen sample is locally spread within the Swedish-speaking part of Finland. The sample is rather equally distributed considering grade and gender.
Procedure The Reading Survey was conducted in the schools in Nov–Dec 1997 during regular school hours. The Reading Survey was administered by the class teacher or another teacher working in each school. The survey was read aloud to the students. The Reading Survey was scored so that the most positive response was given the highest number (4) while the least positive response was given the lowest number (1).
Measures The Motivation to Read Profile (MRP), developed by Gambrell, Palmer, Codling, and Mazzoni (1996), was used in this study. The Motivation to Read Profile (MRP) consists of two basic instruments: the Reading Survey and the Conversational Interview. The Reading Survey is a self-report, group administered instrument, and the interview is designed for individual administration. In this article the results from the Reading Survey, the quantitative results, are presented.
(JB)SWL II_f18_303-315
9/26/02
9:22 AM
Page 306
Precursors of Functional Literacy Table 1. Frequency distribution of groups, according to grade and gender Grade 3 Girls Boys Total
79 82 161
Grade 5 61 76 137
Total 140 158 298
Self-concept as a reader and value of reading The items that focus on self-concept as a reader (10 items) are designed to give information about students’ self-perceived competence in reading and selfperceived performance relative to peers. The items focusing on the value of reading are designed to reveal information about the value students place on reading tasks and activities, in terms of frequency of engagement and reading related activities. Fore more information about the design of the Reading Survey see Gambrell et al. (1996).
Analyses In order to check the validity of the two scales and to choose the items that best represent Self-concept as a Reader (SC) and Value of Reading (VR) all 20 questions were entered into one analysis to see if the two scales could be isolated or if there were other underlying factors. For this purpose a nonlinear principal components analysis was used. Nonlinear principal components analysis, also known by the acronym PRINCALS2 (principal components analysis by means of alternating least squares). PRINCALS quantifies qualitative data by assigning scores to categories of variables and to subjects. The scores are assigned in such a way that homogeneity of the set of variables is maximised. One main reason for using PRINCALS was the nature of the items, especially the fact that the answer categories were so different for the various items. In the Reading Survey the answer categories are specific for each item. But nevertheless the answer categories for a question may be considered ordered from negative to positive, and therefore all items were treated as ordinal. PRINCALS was used to assess if the supposed dimensions to measure SC and VR could be isolated. Afterwards the best items were selected for each dimension. In order to investigate the influence of grade and gender on students’ selfconcept as readers and value of reading, a 2(Grade) × 2(Gender) multivariate analysis of variance MANOVA was performed. A MANOVA was also performed including a 2(Grade) × 2(Gender) × 2(Language) considering self-concept and value of reading. The MANOVAS were performed in order to identify patterns in the motivational aspects of self-concept and value of reading.
(JB)SWL II_f18_303-315
9/26/02
9:22 AM
Page 307
Age and gender differences in reading engagement
8
0.50
10 4 6
16 20 12
0.25 Dimension 2
14
2
18 13
17
0.00
7 -0.25
-0.50
0.00
15
1
11 15
0.25
3 19
0.50
9
0.75
1.00
Dimension 1 Figure 1. Component loadings for the items in the reading survey
Results Figure 1 shows the component loadings of the two-dimensional PRINCALS solution. The items (labelled by the number of the question in the Reading Survey, see Tables 2 and 3) are represented as vectors (directed lines) from the origin of the space. The arrows point in the direction of the more positive responses on the question. Longer vectors represent better discriminating items, a short vector represents an item with only small differences between the negative and positive answer categories. The eigenvalues (See Tables 2 and 3) can indicate how many dimensions are needed. In this case two dimensions were used with eigenvalues of 0.26 for the first dimension and 0.12 for the second dimension. These eigenvalues may be interpreted as proportions of variance explained. The proportion of variance explained by the two dimensions: 0.26 + 0.12 = 0.38 (so still more than half of the variance is in higher dimensions not represented in Figure 1). In general the eigenvalue for a dimension should be larger than 1/number of variables. In this
(JB)SWL II_f18_303-315
9/26/02
9:22 AM
Page 308
Precursors of Functional Literacy Table 2: Component loadings for the self-concept as a reader scale used in PRINCALS analysis Scale SELF-CONCEPT AS A READER
1. My friends think I am . . . (a very good reader, a good reader, an OK reader, a poor reader) 3. I read . . . (not as well as my friends, about the same as my friends, a little better than my friends, a lot better than my friends) 5. When I come to a word I don’t know . . . (I can almost always figure it out, sometimes figure it out, almost never figure it out, never figure it out) 7. When I am reading by myself, I understand . . . (almost everything I read, some of what I read, almost none of what I read, none of what I read) 9. I am . . . (a poor reader, an OK reader, a good reader, a very good reader) 11. I worry about what other kids think about my reading . . . (every day, almost every day, once in a while, never) 13. When my teacher asks me a question about what I have read I, (can never think of an answer, have trouble thinking of an answer, sometimes think of an answer, always think of an answer) 15. Reading is . . . (very easy for me, kind of easy for me, kind of hard for me, very hard for me) 17. When I am in a group talking about stories, I . . . (almost never talk about my ideas, sometimes talk about my ideas, almost always talk about my ideas, always talk about my ideas) 19. When I read out loud I am a . . . (poor reader, OK reader, good reader, very good reader)
Component Loadings Dimension (Eigenvalues) 1 (.2559)
2 (.1174)
.685
–.331
.598
–.457
.261
–.268
.381
–.175
.721
–.432
.195
–.419
.479
.002
.582
–.491
.220
–.002
.573
–.459
The items marked with bold are chosen as the best to represent value of reading.
case both dimensions have eigenvalues larger than 0.05 (1/20 variables). (SPSS Categories ®8.0, 1998: 87) The items fell apart in two groups, nicely in agreement with the difference between VR items above (the even numbers) and SC items below (the odd numbers). This may be interpreted to mean that there are indeed two underlying factors, VR and SC. Though the two bundles of items point in different directions, it is also clear from the figure that the VR and SC dimension are positively correlated. This is shown by the fact that the VR vectors make an angle of less than 90 degrees with the SC vectors. This correlation is also represented in the fact that all items have a positive loading on the first principal component.
(JB)SWL II_f18_303-315
9/26/02
9:22 AM
Page 309
Age and gender differences in reading engagement Table 3: Component loadings for the value of reading scale used in PRINCALS analysis Scale VALUE OF READING
2. Reading a book is something I like to do . . . (never, not very often, sometimes, often) 4. My best friends think reading is . . . (really fun, fun, OK to do, no fun at all) 6. I tell my friends about good books I read . . . (I never do this, I almost never do this, I do this some of the time, I do this a lot) 8. People who read a lot are . . . (very interesting, interesting, not very interesting, boring) 10. I think libraries are . . . (a great place to spend time, an interesting place to spend time, an OK place to spend time, a boring place to spend time) 12. Knowing how to read well is . . . (not very important, sort of important, important, very important) 14. I think reading is . . . (a boring way to spend time, an OK way to spend time, an interesting way to spend time, a great way to spend time) 16. When I grow up I will spend . . . (none of my time reading, very little of my time reading, some of my time reading, a lot of my time reading) 18. I would like for my teacher to read books out loud to the class . . . (every day, almost every day, once in a while, never) 20. When someone gives me a book for present, I feel . . . (very happy, sort of happy, sort of unhappy, unhappy)
Component Loadings Dimension (Eigenvalues) 1 (.2559)
2 (.1174)
.720
.172
.531
.364
.524
.313
.480
.436
.516
.456
.280
.238
.721
.364
.398
.383
.120
.122
.483
.323
The marked items are chosen as the best to represent self-concept as a reader.
As we can see from Figure 1 items 13 and 17 hesitate between the SC and VR scale and are therefore not good representatives for the SC scale that they are supposed to represent. From the figure we can see that among the items with odd labels supposed to measure SC we can see that item 9 is the best discriminator on this scale, followed by items 1, 3, 15 and 19. This group of items were therefore selected as the best items to represent the SC scale. The arrows with the even numbers have another direction and are supposed to represent the VR scale. Here we can see that there are only two shorter arrows not discriminating very well, items 18 and 12. Therefore all other items were chosen to represent the VR scale.
(JB)SWL II_f18_303-315
9/26/02
9:22 AM
Page 310
Precursors of Functional Literacy Table 4. Reliabilities for the factor-based scales Factor label Self-concept as a reader Value of reading a
Items 5 8
Alphaa .84 .80
reliabilities greater than .70 indicate reasonable good internal consistency.
Table 4 presents the internal consistency reliabilities for the new scales, with items chosen to best represent Self-concept as a reader items and Value of reading items. Both scales show high internal consistency. Correlations (Pearson twotailed test) showed a significant correlation (r = 0.37) between self-concept and value of reading (p < 0.01).
Self-concept and value of reading, grade and gender In order to investigate gender and grade differences a 2(Grade) × 2(Gender) MANOVA was performed. As we can see in Table 5, the MANOVA revealed significant grade (p < 0.001), gender (p < 0.001) and grade x gender (p < 0.05) effects for the value of reading. No grade, gender or interaction effects were found considering self-concept. After that the Mann-Whitney’s U-test was performed to determine whether there were differences between boys and girls considering selfconcept and the value of reading. Girls valued reading more than boys (p < 0.001) on both grade levels. No significant differences considering gender and self-concept appeared. Boys in grade 3 valued reading more than boys in grade 5 (p < 0.001). No significant differences could be found among girls in different grades considering the value of reading. Figure 2 shows the decline in the value of reading with increased age (the decline was not significant for girls). No significant differences considering selfconcept appeared.
Self-concept and value of reading, grade, gender and language As shown in table 6, the MANOVA revealed significant grade (p < 0.01), gender (p < 0.001), language (p < 0.05), grade x gender (p < 0.05) as well as grade × language (p < 0.05) effects for the value of reading. No grade, gender, language or interaction effects appeared considering self-concept as a reader. Within gender groups the Mann Whitney’s U-test showed that there were significant differences (p < 0.01) between language group 1 and 2 for girls concerning how they value reading, no significant differences were found considering the same aspect among boys. Within grade and gender the Mann Whitney’s U-test revealed significant differences (p < 0.01) between the two language groups for girls in grade 5, but not for girls in grade 3 on the value measure.
(JB)SWL II_f18_303-315
9/26/02
9:22 AM
Page 311
Age and gender differences in reading engagement Table 5. Mean levels (and standard deviations) for self-concept and the value of reading, according to grade and gender Grade 3 boys 82
n=
girls 79
Grade 5 boys 76
girls 61
Effect (F-values) Grade (Gr) Gender (G) Gr × G
Self
2.82 2.87 (0.57) (0.60)
2.79 2.98 (0.59) (0.46)
0.34 NS
3.19 NS
1.16 NS
Value
2.87 3.19 (0.46) (0.38)
2.55 3.13 (0.55) (0.41)
12.66 ***
71.14 ***
5.63 *
8.50b ***
36.65b ***
2.82b NS
MANOVA (Wilk’s lambda) F [2.293]
* p < 0.05, * * p < 0.01 and * * * p < 0.001, the stars indicate significant differences at these levels. NS = not significant. b Exact statistic
Student Ratings
3.5
Self concept girls Self-concept boys Value girls
3
Value boys 2.5
2 Grade 3
Grade 5
Figure 2. Mean ratings for self-concept and value on the Reading Survey
Conclusions and discussion An instrument assessing two important factors of reading motivation; self-concept and value of reading, was used in this study. Age and gender differences were identified considering the value of reading aspect. The main finding indicates also in this study that girls value reading more than boys. Third graders tend to value reading more than fifth graders. A significant decline considering the value of reading was found between boys in grade 3 and 5. However, no significant age and gender differences were found concerning their self-concept as readers. Linguistic
Group 1
Group 2
Effect (F-values)
28.16b ***
7.67b **
* p < 0.05, ** p < 0.01, and *** p < 0.001. b Exact statistic Group 1: Swedish most frequently spoken at home Group 2: Finnish or other language most frequently spoken at home
MANOVA (Wilk’s lambda) F[2,283]
54.04 ***
2.90 3.24 2.83 3.08 2.52 3.25 2.57 2.94 11.40 (0.41) (0.38) (0.55) (0.37) (0.55) (0.31) (0.55) (0.48) * * 2.64b NS
4.44 *
2.50b NS
4.96 *
0.40b NS
0.04 NS
1.94b NS
3.90 *
1.45b NS
1.48 NS
9:22 AM
Value
Self
Group 2
Grade 5
9/26/02
boys girls boys girls boys girls boys girls Grade(Gr) Gender (G) Language (L ) Gr × G Gr × L G × L Gr × Gn n = 51 56 28 21 53 37 22 24 2.75 2.87 2.91 2.84 2.81 3.00 2.75 2.93 0.22 2.28 0.00 1.29 0.7 0.55 0.37 (0.55) (0.57) (0.58) (0.70) (0.56) (0.47) (0.67) (0.46) NS NS NS NS NS NS NS
Group 1
Grade 3
Table 6. Mean levels (and standard deviations) for self-concept and the value of reading, according to grade, gender, language group and analyses of differences/effects
(JB)SWL II_f18_303-315 Page 312
Precursors of Functional Literacy
(JB)SWL II_f18_303-315
9/26/02
9:22 AM
Page 313
Age and gender differences in reading engagement
background, taking into consideration the language most frequently spoken at home, didn’t seem to affect the students’ self-concept as readers. An alternative explanation that could be discussed is whether the gender effect is a simple compliance effect, meaning that girls are more prone to answer positively than boys. This question can not be answered by looking at the data, but from a practical point of view most teachers would agree on the fact that girls generally read more books than boys do and also seem to be more eager to discuss the books they have read with each other. With support from practical experience, the finding that girls seem to value reading more than boys do, is therefore not likely to be explained just as a compliance effect. The most important insights to be gained from this study is that the results from a Finland-Swedish context, a country with top results on international reading surveys, seem to be very much in line with earlier research on this topic. The value of reading among boys seem to decline with the age. Considering the link between achievement and motivation this is something to be considered in the classroom context. How shall we increase the value of reading among boys, especially in the higher grades in elementary school? In line with McCombs, (1997) also this study indicates a need to further develop understanding of aspects of reading that are motivating for different groups of children (e.g., boys vs. girls) in order to find experiences and environments that accommodate differences between and within groups. For teachers to develop this kind of understanding we need instruments for assessing both cognitive as well as motivational aspects in students’ reading (see e.g., Östern, 1991). Suggestions for further research is therefore a focus on real classroom contexts, to see what is happening, and how it affects different groups of students with different motivation for reading.
Notes . I truly appreciate help with the statistical analyses from Marinus Voeten. . PRINCALS is an optimal scaling procedure in SPSS implementation (SPSS Categories ®8.0, 1998).
References Brunell, V. (1992). Finlandssvenskans framtid i skolperspektiv [The future of the Swedishspeaking population in a school perspective]. In S.-E. Hansén & C. Laurén (Eds.), Finlandssvenskans Framtid. Rapport från en språkvårdskonferens [The Future of the Swedish-speaking Population in Finland. Report from a conference on preservation of the purity of the language]. Vasa: Svenskösterbottniska Samfundet. Brunell, V. (1995). Läskunnigheten i den Finlandssvenska Grundskolan i ett Jämförande Perspektiv [Reading Literacy in the Swedish-speaking Comprehensive School in Finland: a comparative perspective]. Jyväskylä: Jyväskylä Universitet Pedagogiska Forskningsinstitutet.
(JB)SWL II_f18_303-315
9/26/02
9:22 AM
Page 314
Precursors of Functional Literacy Brunell, V. & Saretsalo, L. (1999). Sociocultural Diversity and Reading Literacy in a FinlandSwedish Environment. Scandinavian Journal of Educational Research, 43, 173–190. Baker, L., Afflerbach, P. & Reinking, D. (1996). Developing engaged readers in school and home communities: An overview. In L. Baker, P. Afflerbach & D. Reinking (Eds.), Developing engaged readers in school and home communities, xiii–xxviii. Mahwah, NJ: Erlbaum. Eccles, J. S., Midgley, C. & Adler, T. (1984). Grade-related changes in the school environment: Effects on achievement motivation. In J. G. Nicholls (Ed.), The development of achievement motivation (283–331). Greenwich, CT: JAI. Eccles, J. S., Wigfield, A., Harold, R. & Blumenfeld, P. B. (1993). Age and gender differences in children’s self- and task perceptions during elementary school. Child Development, 64, 830–847. Elley, W. B. (1992). How in the world do students read? Hamburg, Germany: International Association for the Evaluation of Educational Achievement. Ford, M. E. (1992). Motivating humans. Newbury Park, CA: Sage. Gambrell, B., Palmer, B. M., Codling, R. M. & Mazzoni S. A. (1996). Assessing motivation to read. The Reading Teacher, 49, 518–533. Guthrie, J.T. (1996). Educational contexts for engagement in literacy. Reading Teacher, 49, 432–435. Guthrie, J. T., Schafer, W., Wang, Y. & Afflerbach, P. (1993) Influences of instruction on reading engagement: An empirical exploration of a social-cognitive framework of reading activity (Research Report No. 3). Athens, GA: National Reading Research Center. Guthrie, J. T., Van Meter, P., Mc Cann, A., Wigfield, A., Bennett, Poundstone, C., Rice, M. E., Faibisch, F., Hunt, B. & Mitchell, A. (1996). Growth of literacy engagement: Changes in motivations and strategies during concept-oriented reading instruction. Reading Research Quarterly, 31, 306–333. Henk, W. & Melnick, S.A. (1995). The Reader Self-Perception Scale (RSPS): A new tool for measuring how children feel about themselves as readers. The Reading Teacher, 48, 470–482. Mc Combs, B. L. (1997). Commentary: reflections on Motivations for reading — Through the Looking Glass of theory, Practice, and Reader Experiences. Educational Psychologist, 32, 125–134. Mc Kenna, M. C. & Kear, D. J. (1990). Measuring attitude toward reading: A new tool for teachers. The Reading Teacher, 43, 626–639. Oldfather, P. & Wigfield, A. (1996). Children’s motivations to read. In L. Baker, P. Afflerbach & D. Reinking (Eds.), Developing engaged readers in school and home communities (89–113). Mahwah, NJ: Erlbaum. Östern, A.-L. (1991). Tvåspråkighet & lingvistisk medvetenhet. Betydelsen av tvåspråkighet och av undervisning för lingvistisk medvetenhet hos barn i åldern sex till åtta år. [Bilingualism and linguistic awareness. The importance of bilingualism and teaching for metalinguistic awareness in children aged six to eight]. (Dissertation) Åbo: Åbo Akademis Förlag. Purves, A. & Beach, R. (1972). Literature and the reader: Research on response to literature, reading interests, and teaching of literature. Urbana, IL: National Council of Teachers of English. Schunk, D. H. & Rice, J. M. (1993). Strategy fading and progress feedback: Effects on selfefficacy and comprehension among students receiving remedial reading services. Journal of Special Education, 27, 257–276 SPSS Categories ®8.0 (1998). Chicago: SPSS Inc. Stipek, D. J. & Mac Iver, D. (1989). Developmental change in children’s assessment of intellectual competence. Child Development, 60, 521–538. Walberg, H. J. & Tsai, S. (1985). Correlates of reading achievement and attitude: A national assessment study. Journal of Educational Research, 78, 159–167.
(JB)SWL II_f18_303-315
9/26/02
9:22 AM
Page 315
Age and gender differences in reading engagement Wigfield, A. (1994). Expectancy-value theory of achievement motivation: A developmental perspective. Educational Psychology Review, 6, 1, 48–78. Wigfield, A. & Eccles, J. S. & Pintrich, P. (1996). Development from the ages of 12–25. In D. Berliner & R. Calfee (Eds.), Handbook of educational psychology (148–187). New York: Macmillan. Wigfield, A., Wilde, K., Baker, L., Fernandez-Fein S. & Scher, D. (1996). The nature of children’s reading motivations and their relations to reading frequency and reading performance (Reading Research Rep. No. 63). Ahens, GA: National Reading Research Center. Wigfield, A. & Guthrie, J. T. (1997). Motivation for reading: An overview. Educational Psychologist, 32, 57–58. Winne, P. (1985). Steps toward promoting cognitive achievements. Elementary School Journal, 85, 673–693. Wixson, K. K. & Lipson, M. Y. (1991). Reading diagnosis and remediation. Glenview, IL: Scott, Foresman. Address Cygnaeus School Marieg. 7 FIN-20 100 TURKU/ÅBO, Finland Email:
[email protected]
(JB)SWL II_f19_316-337
9/26/02
9:23 AM
Page 316
(JB)SWL II_f19_316-337
9/26/02
9:23 AM
Page 317
Promoting at-risk pupils’ foreign language literacy learning Mia Dufva and Marja Vauras* University of Turku
In the present study, the focus is on the learning of English as a foreign language in Finland. Foreign language education in Finland starts in the third grade (at the age of nine), when the majority of pupils select English as the first foreign language. The English curriculum is based on the communicative approach (Brumfit, 1980: 110–122; Johnson, 1996: 173), which focuses on listening comprehension, oral production and accumulation of vocabulary. Accurate word recognition and spelling skills are not taught until grade 5. These practices have many positive effects. Pupils learn at an early stage to express themselves, develop a large active vocabulary and high motivation to learn the new language. Although Finnish and English differ from each other in orthographic regularity (Finnish has nearly perfect grapheme-to-phoneme correspondence), most pupils adopt English pronunciation and spelling rules and, thus, learn to read and write English without greater difficulties. However, there are pupils who struggle with the new language, and may have persistent difficulties, especially in learning to decode and spell English words. At present, diagnostic and remedial means are few, as no systematic remedial instruction is usually offered. Thus, there is a real need for a program aimed at promoting at-risk pupils’ English word recognition and spelling skills. Our aim was to examine whether at-risk pupils’ (4th graders, 10-year-olds) English language literacy learning could be supported by a small-group intervention program (lasting for 26 weeks, one hour per week), which focused on promoting understanding of the sound structure of the English language. Usually, difficulties in foreign language word recognition and spelling have been explained by two separate factors. The first factor is existing difficulties in native language word recognition and spelling, often referred to as a reading problem (Alderson, 1984) or an interlingual learning problem (Verhoeven, 1990). The second factor is difficulties related to the foreign language itself, referred to as a language problem (Alderson, 1984) or an intralingual learning problem (Verhoeven, 1990). There is now plenty of evidence for the cross-linguistic transfer of native language skills to the foreign/second language. For example, the reviews on bilingual
(JB)SWL II_f19_316-337
9/26/02
9:23 AM
Page 318
Precursors of Functional Literacy
reading by Bernhardt (1991), Durguno[lu (1997), Durguno[lu and Hancin (1992), and Fitzgerald (1995) show that both lower-level (such as word recognition) and higher-level (such as reading comprehension) skills of L1 reading can transfer to second language reading although the languages studied have different orthographies (Cummins, 1991). L1 word recognition skills seem to have greater importance for L2 word recognition when the orthographic distance between L1 and L2 is small (Koda, 1996). However, orthographic difference does not rule out the possibility of cross-linguistic transfer. In our recent study (Dufva & Voeten, 1999) with novice EFL learners, we found that Finnish word recognition, reading comprehension and phonological memory in grade 2 were significant predictors of English proficiency at the end of grade 3. It seems that the better skills a pupil has in native language word recognition, the more solid a base s/he has for mastering the foreign language. In line with this, Durguno[lu, Nagy, and HancinBhatt (1993) studied young second language learners and found that word recognition skill for the second language could be predicted by L1 word recognition and phonological awareness The native language does not always have only positive transfer effects on L2 learning, but may also have negative ones (Durguno[lu & Hancin, 1992; Odlin, 1989). In the Finnish EFL classrooms this becomes evident in pupils who apply the phoneme-to-grapheme correspondences from the Finnish language when they read and write English. These pupils may have difficulties in native language word recognition and spelling, but their difficulties lie especially in understanding the nature of the foreign language. In this situation their difficulties stem from language-related factors. One language-related problem can be deficits related to mastering the foreign language phonology. This hypothesis, known as the phonological core deficit hypothesis, suggests that difficulties in word recognition and spelling can be explained by phonological processing deficits, which, in turn, manifest themselves as difficulties in applying grapheme-to-phoneme conversion operations in word recognition, and also in applying the opposite conversion in spelling (Wade-Woolley & Siegel, 1997). It seems possible that some of the novice EFL learners who have difficulties in English word recognition and spelling, suffer from difficulties in phonological processing skills. The general phonological processing skill can be divided into subskills of which phonological memory and phonological awareness are of interest in the present study. Phonological memory refers to those parts of Baddeley’s (1986, 1990) working memory model that are involved in processing verbal material. The ability to create accurate representations of new language material in the phonological memory seems to enhance foreign language learning. Service (1992), for example, has suggested that phonological memory is a crucial factor for learning new language material, such as new vocabulary. Previous studies have shown that phonological memory has predicted L1 word recognition (Gathercole & Baddeley, 1993) and spelling (Schneider & Näslund, 1993). Therefore, it can also be an important factor for learning to read and spell in a foreign language. Phonological awareness, in turn, is the awareness of the sound units a spo-
(JB)SWL II_f19_316-337
9/26/02
9:23 AM
Page 319
Promoting at-risk pupils’ foreign language literacy learning
ken word consists of, and refers to both the ability to analyze a word into phonological units and the ability to synthesize these units into a spoken word. By focusing children’s attention on the smaller units of the language, such as words, syllables, and phonemes, it has been possible to enhance preliterate children’s L1 decoding and spelling acquisition (Byrne & Fielding-Barnsley, 1991, 1993; Lundberg, Frost & Petersen, 1988; Olofsson & Lundberg, 1983). The content of those programs (e.g., Adams, Foorman, Lundberg & Beeler, 1998) is hierarchically structured starting from lower-level skills, such as rhyming, word and sentence awareness, and ending up with phoneme awareness skills. In the present study, we have applied the same principles. Although there have been many intervention studies aimed at promoting L1 word recognition and spelling skills, intervention research with pupils at-risk of foreign language learning difficulties is so far almost non-existent. Sparks, Ganschow and their colleagues (Sparks, Ganschow, Kenneweg & Miller, 1991; Sparks, Ganschow, Pohlman, Skinner & Artzer, 1992) have worked with learning disabled high school students, while Service (1993) studied elementary school pupils. Sparks et al. found that training the phonological and syntactic codes of FL enhanced both foreign language aptitude and various native language skills. In a follow-up study, the gains were partly sustained, and they were also partly replicated in their later study (Sparks & Ganschow, 1993). It must be noted that Sparks and Ganschow did not focus on specific FL skills, but measured only FL aptitude. In addition, no control group was included. Service’s experiment (1993) was based on studies showing a close connection between the phonological loop (one subcomponent of working memory specialized in dealing with linguistic material) and vocabulary learning (e.g., Gathercole, Willis, Emslie & Baddeley, 1992). It is assumed that phonological knowledge in the long-term memory can support the material rehearsed in the phonological loop. Therefore, it is, for example, easier to repeat back pseudowords based on the phonology of one’s native language than those based on a foreign language. In her experiment, Service studied nine Finnish at-risk (poorly functioning phonological loop) 3rd graders, who received 12 sessions of extra teaching in English phonology. The control group (n = 8) received the same amount of additional English teaching. The result showed that teaching the phonology of the English language supported the functioning of the phonological loop. This benefit can be assumed to enhance the learning of new vocabulary in the foreign language, which is connected with various other skills, such as comprehension skills. The present study extends the previous intervention studies in three ways; we had substantially younger participants than those in the studies by Sparks and Ganschow, the intervention program lasted twice as long as that of Service’s study, and we had well-matched intervention and control groups. The present intervention program was hierarchically structured, starting with lower-level skills (e.g., sentence and word awareness), which along with the development of pupils’ skills, gradually moved to higher-level skills such as understanding the sound structure of the English language. Therefore, the content of the intervention was also more
(JB)SWL II_f19_316-337
9/26/02
9:23 AM
Page 320
Precursors of Functional Literacy
versatile than that in Service’s study. In line with the intervention carried out by Sparks and Ganschow, we stressed the multisensorial nature of the program to ensure that pupils could use not only hearing but also other senses in learning new language material. Throughout the intervention, transfer of the skills and knowledge taught during the intervention lessons to homework settings was emphasized. The intervention program was carried out during the second year of English studies when there still is a good chance of preventing more serious learning difficulties in English. The research questions were: 1. Does the intervention group outperform the matched control group as a function of the intervention program in skills measuring proficiency in the sound structure of the English language? These skills are phonological awareness, phonological memory, word recognition, and spelling. 2. What kinds of attitudes did the intervention pupils have towards the intervention, and did their feelings change during the intervention?
Method Participants The participants were 24 Finnish-speaking fourth graders selected (see Procedure for detailed information on the selection process) from 56 pupils (29 girls and 27 boys) from one primary school in Turku. They had started to learn English as their first foreign language in the third grade, receiving two hours of formal English teaching per week.
Materials Four measurement points were included: pre-test in September, intermediate test 4 months later in January, post-test 4 months later in May, and the delayed test in the next school year, in February, 9 months after the end of the intervention. At each measurement point, the following English skills were assessed: pseudoword repetition, phonological awareness, word recognition, and spelling; the only exception being phonological awareness which was not included in the delayed test. For each test, parallel versions for the four measurement points were prepared. All tests were administered in an individual testing situation, except the spelling test at the delayed measurement point, which was administered in a classroom situation. In the pre-test, nonverbal intelligence (Raven) was assessed for matching purposes. The test was administered in groups of ten pupils. Native language word recognition and spelling tests were administered in the pre-test and in the posttest by means of parallel tasks. They were done in a classroom situation. All testing was done during school days.
(JB)SWL II_f19_316-337
9/26/02
9:23 AM
Page 321
Promoting at-risk pupils’ foreign language literacy learning
Nonverbal intelligence Raven’s Coloured Progressive Matrices (Raven, Court & Raven, 1995) was administered to assess pupils’ nonverbal intelligence.
English proficiency Phonological awareness Assessment of phonological awareness consisted of two tests, those of phoneme deletion and phoneme segmentation (Yopp, 1988). Both tests had three practice items and ten test items. In the task of phoneme deletion, the experimenter said one word at a time (e.g., stop), named one of its phonemes (/s/) and asked the pupil to give the new word when that particular phoneme was deleted (top). In the first four words the to-be-deleted phoneme was the initial one, in the next three words the last one, and in the last three words it was in the middle of the word. One point was given for each correct answer, the maximum being ten. In the task of phoneme segmentation, the experimenter said one word at a time (e.g., dog) and asked the pupil to segment it into phonemes. The words, which were taken from the pupils’ English language textbooks, consisted of two to four phonemes. One point was given for each correctly segmented word, the maximum being ten. A summed score of phonological awareness was calculated, the maximum score being 20. Reliabilities (Cronbach’s alpha) were moderate for both tasks in the pre-test (for phoneme deletion .62 and for phoneme segmentation .69), but in the intermediate test and in the post-test they were low. Therefore, we only used the pre-test data for selection purposes. Phonological memory This task, which was adopted from Service (1989), consisted of two lists of Englishsounding pseudowords. There were ten words in each list and a practice list of six words. The words were created by interchanging the first and last syllables of a set of real English words. Half of the words had two syllables (e.g., boxflap) while the other half had four syllables (e.g., refinikism). The order of the words in a list was randomized. A native speaker of English checked the words to make sure that they conformed to English phonology and an Englishman with a standard BritishEnglish accent sound-recorded the lists. Pupils were told that they would hear English-sounding words, one word at a time, and their task was to repeat the word aloud as they heard it. If they felt unsure, they were encouraged to say even a small part of the word, because it did not matter if it was not perfectly correct. The entire test, including the practice list, the pseudowords, and the pupils’ responses was tape-recorded for later analysis. To get the accuracy score for pseudoword repetition, the responses were rated according to the number of syllables that had been correctly repeated, the maximum score for a list being 30. A correctly pronounced syllable was defined as one containing no obvious phoneme replacements, omissions, or additions. Minor
(JB)SWL II_f19_316-337
9/26/02
9:23 AM
Page 322
Precursors of Functional Literacy
variations were allowed as long as phonemic borders were not crossed. A summed score of the two lists was calculated, the maximum being 60. The inter-rater reliabilities for 20% of the answers, rated by two judges, were .85, .81, and .81 for pre-, intermediate and post-tests, respectively.
Word recognition Word recognition skill was assessed with texts taken from a children’s instructional book (The Usborne first book of knowledge, 1994). The texts were modified so that nearly all the words were familiar to the pupils. The texts consisted of 120–123 words and 12–14 sentences (see Appendix A for the text used in the pretest). The pupils’ task was to read the text aloud as fast as possible, but to avoid reading errors. Reading accuracy was emphasized. Pupils were instructed to read unfamiliar words as they thought they should be pronounced. Pupils’ reading was tape-recorded and reading time was measured with a stop-watch. The reading performance of 20% of the participants was rated by two judges, whose interrater reliabilities were .84, .90, and .85 for pre-, intermediate and post-tests, respectively. Word recognition performance is expressed as reading time and as the number of correctly read words (percentage of the total number of words in the particular text). Spelling The spelling tests consisted of 12 sentences, in total 70–73 words, from four to nine words in a sentence (see Appendix B for the spelling test used in the pretest). The words and grammar were taken from the pupils’ English language textbooks. It was assumed, therefore, that the material was familiar to them. The words consisted of the most common consonant and vowel sounds of the English language. The task consisted also of basic grammar such as interrogative and negative sentences, plurals, ‘to have’ verbs, ‘to be’ verbs, and personal pronouns. The experimenter dictated one sentence at a time once. On the pupil’s request, the experimenter repeated the sentence once. If the pupil had difficulties in spelling, s/he was instructed to spell the word as s/he thought it should be spelled. Spelling performance is expressed as the number of correctly spelled sentences, as the number of correctly spelled words (percentage of the total number of dictated words), and as the number of attempted words. Proficiency in native language Word recognition The Norm-Referenced Comprehensive School Reading Test (Lindeman, 1998) includes two parallel word attack tasks. Both tasks consisted of 78 word chains, with two to four words in each chain. Pupils had 3 min 30 sec to mark the word borders for as many chains as possible. Word recognition performance is expressed as the mean of the total number of correctly separated words in both tasks.
(JB)SWL II_f19_316-337
9/26/02
9:23 AM
Page 323
Promoting at-risk pupils’ foreign language literacy learning
Spelling The spelling test consisted of 10 sentences, totalling 44 words with three to six words in a sentence. The words consisted of the most common Finnish consonant and vowel sounds as well as some uncommon graphemes (e.g., g and b) and grapheme clusters. The experimenter dictated one sentence at a time once. If asked to do so, the experimenter repeated the sentence once. Spelling performances is expressed as the number of correctly spelled words. Intervention-related task To assess the intervention group pupils’ attitudes towards the intervention (second research question), a learning diary was introduced, to be filled in after each lesson from the second lesson on. It consisted of three questions: (1) How did you feel about this lesson?, (2) How do you think you succeeded? (3) What did you learn today? Answers to the first two questions were indicated by marking one of five alternatives (really bad . . . really nice), while the third question required an open answer. The learning diary aimed to promote pupils’ metacognitive skills in evaluating their learning.
Procedure Selection of the intervention and the control group The selection of pupils into the intervention group and the control group was primarily based on English word recognition and spelling, and secondly, on phonological awareness and phonological memory. Raven was included to match the groups for nonverbal intelligence. We selected altogether 24 pupils of whom 8 had average skills in word recognition and spelling, while 16 had difficulties in these skills. The criterion for a difficulty was belonging to the weakest 25 per cent. By randomly assigning pupils from both skill groups, we formed one intervention and one control group. Both groups comprised of four average and eight weak pupils. In the intervention group, there were eight girls and four boys. Their mean age at the beginning of the training was 10 years 3 months. The control group consisted of six girls and six boys, whose mean age was 9 years 9 months. Due to their very low scores at each measurement point, three pupils both from the intervention group and from the control group were excluded after the intervention. Their z-scores on the primary criterion measures (English word recognition and spelling) in the pre-test ranged between –2.55 and –0.54, and in the delayed test between –2.79 and –0.84, indicating, thus, an extremely slow development. Therefore, it is plausible to argue that they could have benefited from a longer, more intensive, and individually constructed program carried out at a slower pace. Their skill level analyses also indicated that the intervention should include themes and skills, such as basic grammar and self-regulation which now
(JB)SWL II_f19_316-337
9/26/02
9:23 AM
Page 324
Precursors of Functional Literacy
played only a minor role. Although teaching the English sound structure had a major role in the intervention program, it clearly was not enough for these pupils, and, therefore, it should receive more attention. After the exclusion of the six weak pupils, there remained nine pupils in the intervention and nine pupils in the control group. The statistical analyses are performed with these group sizes. Mean pretest scores for both groups are presented in Table 1. A MANOVA showed that the intervention and the control group did not statistically differ from each other on the criterion measures.
The intervention program The intervention was carried out in three small groups, each consisting of four pupils from two different classes. The intervention took place during school hours, 45 minutes at a time, usually once a week, during 26 weeks from the end of September to mid-April. The pupils were supervised by the first author. The control pupils were taught according to the regular curriculum. None of them received regular special instruction in English.
Purpose and content of the intervention program The aims of the intervention were to promote pupils’ comprehension of the sound structure of the English language and, in this way, to promote their phonological processing skills in English. When phonological processing skills develop, their effect is shown as stronger skills in phonological awareness, phonological memory, word recognition, and spelling. To reach these goals, the intervention was divided into six themes, the contents of which were as follows: 1. Forming the groups and becoming a group member (lessons 1–2). The aims of the first two lessons were to define the goals of the group, to motivate pupils to learn an extra lesson in the English language, and to form the group. The supervisor defined the group (when, where and how often pupils meet). The pupils presented themselves, described their hobbies and families briefly, using both Finnish and English as much as they could. The group discussed their feelings about and motivations for learning the English language. Via this discussion potential benefits of the intervention experience were discussed. 2. Sentence and word awareness (lessons 3–6). From the third lesson on all lessons started by refreshing the pupils’ memory of the topics and skills they had learned during the previous lesson. The importance and usefulness of this metacognitive skill for efficient learning and homework were pointed out. The aim of the sentence awareness was to promote the pupils’ awareness of the concepts of word and sentence, and to practice sentence boundaries and word order in different basic sentences. The tasks were similar to those used to promote preliterate children’s phonological awareness, such as counting words in sentences, forming sentences from separate words, changing the word order
(JB)SWL II_f19_316-337
9/26/02
9:23 AM
Page 325
Promoting at-risk pupils’ foreign language literacy learning Table 1. Mean performance scores (standard deviations in parenthesis) of the intervention and the control group in the pre-test Group Task (maximum in parenthesis)
Intervention (n = 9)
Control (n = 9)
Criterion measures Nonverbal intelligence Word recognition: correct words (100%) Spelling: correct words (100%) Phonological awareness (20) Phonological memory (60)
30.4 (4.4) 58.9 (5.6) 33.8 (14.5) 13.0 (2.2) 29.7 (13.0)
32.4 (3.6) 58.7 (10.5) 31.7 (11.7) 14.1 (2.7) 38.6 (6.6)
Other pre-tests English Word recognition: reading time Spelling: correct sentences (12) Native language Word recognition (214) Spelling (44)
2.22 min. (1.3) 0.33 (0.5) 92.8 (25.7) 39.4 (4.1)
1.7 min. (0.7) 0.22 (0.4) 84.8 (34.8) 36.4 (3.1)
Note. Differences between the groups are statistically nonsignificant in every measure.
in a sentence while monitoring one’s comprehension at the sentence level, composing sentences, and finding sentence boundaries in a longer text. 3. Syntactic awareness (lessons 5–6). Basic grammar was taught in this phase, because many pupils had problems with it. Teaching focused on personal pronouns, ‘to be’ and ‘to have’ verbs. The rules were written on large sheets and put on the wall to serve as mnemonic aids. 4. Rhyming skills (lessons 7–10). Teaching rhyming skills served as an intermediate phase between the sentence level and sound level. The tasks were similar to those used to teach rhyming skills to preliterate native English-speaking children. The pupils listened to and read nursery rhymes, filled in rhyming sentences, grouped rhyming words, and applied their rhyming skills to word recognition and spelling by comparing the orthographies and phonological structures of rhyming words. 5. Raising awareness of English sound structure (lessons 11–22). The aim of this theme was to raise the pupils’ phonological awareness, to teach the most important phonetic signs, and to teach some grapheme-to-phoneme correspondences to be used in word recognition and spelling. All the lessons were organized according to the following model. – Introduction. The teacher introduced the phonetic sign and its corresponding grapheme to be learned. For example, the teacher said a familiar word and asked those who wanted to to write it on the blackboard, or else the teacher said words beginning with the same grapheme-phoneme pair (e.g., bun-bank-butterfly) and asked the pupils to find out what was common in these words.
(JB)SWL II_f19_316-337
9/26/02
9:23 AM
Page 326
Precursors of Functional Literacy
–
Pronunciation. Pronunciation exercises from the book by Ponsonby (1995) were used in the next phase. Practice was focused especially on phonemes different or absent from Finnish. – Guided search. Pupils searched in books for words including the particular grapheme-phoneme pair, either individually or in pairs. They wrote 3–5 words on the blackboard and also clarified their Finnish equivalents. – Modeling and oral practice I. The teacher read the words aloud, the pupils repeated them and paid attention especially to the phoneme-grapheme pair to be learned. They searched for the phoneme-grapheme pairs from each word and compared the words. – Writing the rules. After this oral practice, the pupils wrote the phonetic sign, its corresponding grapheme and some example words in their own notebooks. Each pupil read aloud some of the example words. – Modeling and oral practice II. Dr Seuss’s ABC cd-rom program (1995) was used in the next phase for listening to and practising the particular grapheme-phoneme pair. In this program, graphemes and some example words are presented. The pupils listened to the words, read them aloud, translated them into Finnish, and compared the words with respect to the phoneme-grapheme pair to be learned. – Practice on the grapheme-to-phoneme correspondence. Different kinds of tasks enhancing phonological awareness and transfer of the correspondences to decoding and spelling were practised, such as role switching by correcting spelling and word recognition errors, e.g., from imaginary letters. 6. Consolidating awareness of English sound structure (lessons 20–25). The aim of the last theme was to consolidate the pupils’ skills learned during the previous theme, to provide them with more practice, and to enhance transfer to homework settings. Therefore, practice situations and tasks were similar to homework situations. During the six lessons, a children’s fairy tale was read through. Attention was focused both on understanding the gist of the story and on grapheme-to-phoneme correspondences and their corresponding phonetic signs. The teacher modeled how to read the sentences of the tale, the pupils followed and compared the pronunciation and orthography of some words of the sentences. The teacher also modeled how to study the vocabulary of the text efficiently. The aim was that in addition to understanding and remembering the new words, their spelling patterns and pronunciation would be practised.
Results Group analyses To examine the effect of the intervention at the group level, we first compared the performance of the intervention and control group in native language skills measured in the pre-test and post-test. The intervention group had somewhat better
(JB)SWL II_f19_316-337
9/26/02
9:23 AM
Page 327
Promoting at-risk pupils’ foreign language literacy learning
skills than the control group in native language word recognition and spelling in the pre-test (see Table 1). The differences were tested using an independence groups t tests, and were shown to be nonsignificant both for word recognition and for spelling. Therefore, both groups were equal in terms of their skills in native language word recognition and spelling, indicating that differences between the groups can not be explained by differences in their level of native language proficiency. In addition, the groups had equal skills in the post-test in native language word recognition and spelling, which means that the effects of the intervention were specific to English proficiency. Next, we compared the development of the intervention group and the control group from pre-test to delayed test in English phonological memory, word recognition, and spelling (see Table 2 for group means and standard deviations). A significant effect of the intervention was found on phonological memory, F(3,48) = 5.76, p = .002. Group means show that although the intervention group started from a lower level than the control group, it overtook the control group in the delayed test. In addition, the difference between the groups’ means in the delayed test was significant. The effect of the intervention on word recognition (number of correctly read words) and spelling (number of correctly spelled words) skills combined were analyzed with MANOVA. The main effect of word recognition and spelling was significant F(1,16) = 75.64, p = .001. However, this was due to the post-intervention and delayed effects, since the expected effect of the interaction of training and measurement point proved significant, F(3,48) = 3.63, p = .019. These results show that by the intermediate test, no differences in word recognition and spelling were found between the intervention group and the control group. The effects of the intervention began to emerge after this point when the focus was explicitly placed on grapheme-phoneme correspondences, i.e. on word recognition and spelling skills. This result also suggests that a rather long intervention was required to boost the development of EFL in low-achieving pupils. Group differences were nonsignificant in the second measure of word recognition (reading time) and in the number of correctly spelled sentences counted from the spelling task. However, a stronger development was obtained for the intervention group in the third measure of spelling, that is attempted words, F(3,48) = 5.45, p = .003. This effect was due to the intervention group’s significantly lower performance level in the pre-test, while the group means were equal at the three following measurement points. The effect size measure was used to assess the effect of the intervention on the development of English phonological memory, word recognition, and spelling. The effect sizes for these skills were calculated separately for both groups in the following way: ESEffect size = (mpost-test — mpre-test) / SDpre-test (Castro, 1987). Separate effect sizes for the development of these skills from pre-test to post-test, and from pre-test to delayed test are presented in Table 3. As can be expected on the basis of the analyses of variance, positive development over eight months (from the pre-test to the post-test) and 17 months (from the pre-test to the delayed test) was evident in both groups. A strong development in the intervention group was
(JB)SWL II_f19_316-337
9/26/02
9:23 AM
Page 328
Precursors of Functional Literacy Table 2. Mean performance scores (standard deviations in parenthesis) for the intervention group and the control group in criterion measures from the pre-test to the delayed test Groups Skills and measurement points Phonological memory Pre-test Intermediate test Post-test Delayed test Word recognition + spelling1 Pre-test Intermediate test Post-test Delayed test Word recognition time (sec.) Pre-test Intermediate test Post-test Delayed test Spelling, correct sentences Pre-test Intermediate test Post-test Delayed test Spelling, attempted words Pre-test Intermediate test Post-test Delayed test
Difference
Intervention
Control
between groups
29.7 (13.0) 46.2 (5.2) 42.4 (4.9) 46.9 (6.3)
38.6 (6.6) 45.4 (6.2) 45.6 (6.1) 42.3 (10.2)
.001 ns. ns. .001
46.4 59.5 67.6 76.2
45.2 59.6 60.7 69.2
ns. ns. .001 .001
2.2 (1.3) 1.3 (0.5) 1.1 (0.5) 1.3 (0.7)
1.7 (0.7) 1.3 (0.2) 1.1 (0.2) 1.3 (0.1)
.005 ns. ns. ns.
0.3 (0.5) 1.0 (1.1) 1.5 (1.9) 2.8 (1.8)
0.2 (0.4) 0.8 (1.0) 0.4 (0.7) 1.3 (2.3)
ns. ns. .05 .01
50.3 (9.6) 56.9 (6.6) 64.7 (6.0) 61.9 (9.7)
58.9 (8.4) 58.7 (8.5) 61.3 (10.3) 61.7 (3.7)
.001 ns. ns. ns.
Note. 1 = word recognition (number of correctly read words) and spelling (number of correctly spelled words) combined
observed in phonological memory, in word recognition time, and in the three measures of spelling performance. No intervention-related effects were found on the number of correctly read words in the word recognition task. Finally, the pupils seemed to respond to the intervention very favourably in terms of their feelings towards the lessons and in terms of their own success during the lessons. We used a MANOVA to analyze the pupils’ ratings for the first two questions (1. How did you feel about this lesson?, 2. How do you think you succeeded?) of the learning diary filled in at the end of each lesson. The withinsubject variables were the ratings (question 1 = feelings, question 2 = success) and the themes taught (for this analysis four themes were created instead of the six). There was a significant main effect for ratings, F(1,8) = 25.9, p = .001, but no significant interaction between the ratings and themes. The mean ratings, displayed in Table 4, were high, irrespective of the theme and the duration of the training.
(JB)SWL II_f19_316-337
9/26/02
9:23 AM
Page 329
Promoting at-risk pupils’ foreign language literacy learning Table 3. Effect sizes for the intervention and the control group Effect sizes
Tasks Phonological memory Word recognition a) correct words b) reading time1 Spelling a) correct sentences b) correct words c) attempted words
From pre-test to post-test
From pre-test to delayed test
Intervention group
Intervention group
Control group
Control group
1.9
1.1
2.6
0.6
1.5 –1.6
1.2 –0.5
2.3 –1.3
2.0 –0.6
2.7 2.3 1.7
0.5 1.6 0.3
5.6 3.0 1.4
2.5 2.3 0.3
Note. 1 = Because reading time was indicated in seconds, the effect size is negative
Case analyses The group-level analysis (intervention vs. control group) from pre-test to delayed test showed that the intervention had positive effects on phonological memory, word recognition, and spelling. However, it is well known from previous intervention studies that the effects are not uniform for all pupils (see, e.g., Vauras, Rauhanummi, Kinnunen & Lepola, 1999). This was also true in this study. When the poor intervention group and the average intervention group were compared with their control groups, positive effects were found only for the average intervention group in the combined measure of word recognition and spelling, F(3,18) = 4.72, p = .013. The poor intervention group did not differ from its control group on any of the measures. It must be noted that the group sizes were small in these analyses (four in the average groups and five in the poor groups). In the following analysis, the development of the five pupils in the poor intervention group is compared to their control group to see what kinds of effects the intervention had at case level. Because word recognition (correctly read words) and spelling (correctly spelled words) showed positive effects in the group level analyses, we wanted to examine, at the individual level, the poor pupils’ development in these two skills. The groups were equal on both measures in the pre-test. Figures 1 and 2 present individual developmental graphs for both groups (intermediate measurement point is excluded because the intervention had had no effects by that time point) in spelling development (number of correctly spelled words). Figures 3 and 4 present graphs of word recognition development (number of correctly read words) for both groups. These developmental graphs show that in the delayed test, three out of five pupils (Jenni, Taru and Katri) in the intervention group had reached the criterion for significant development (–95%
(JB)SWL II_f19_316-337
9/26/02
9:23 AM
Page 330
Precursors of Functional Literacy Table 4. Self-evaluations during the intervention program Themes taught during the intervention Attitudes
1–3
4
5
6
Feelings Success
4.6 4.2
4.8 4.1
4.7 4.3
4.7 4.3
Note. The evaluation scale ranged from 0 to 5, the value of 5 indicating the most positive feeling or experienced success 80 70 60 50 Jenni conf. level
40
Katri Paula Niko Taru
30 20 10 0 pre
post
delayed
Figure 1. The development of spelling skill in the poor intervention group
confidence level counted from the whole sample, n = 56) both in spelling and in word recognition. In spelling, none of the control pupils had reached the criterion, and in word recognition only one had done so (Teemu). In order to find factors which could explain the effectiveness of the intervention program, the performance of the three responsive intervention pupils (Jenni, Taru, and Katri) was compared with that of the three resistant intervention pupils who had been excluded from all the analyses. However, no differences were found in cognitive measures, i.e. nonverbal intelligence and native language word recognition and spelling. In sum, these developmental graphs clearly show that significant development occurred even in the poor intervention group, although group analyses did not show this. However, with the measures used in this study we can not explain their development.
(JB)SWL II_f19_316-337
9/26/02
9:23 AM
Page 331
Promoting at-risk pupils’ foreign language literacy learning
70 60 50 40
Jari Sini
30
Vera Teemu Jessi conf. level
20 10 0 pre
post
delayed
Figure 2. The development of spelling skill in the poor control group
90 80 70
Jenni Taru
60
Katri Paula Niko conf. level
50 40 30 pre
post
delayed
Figure 3. The development of word recognition skill in the poor intervention group
Discussion The discussion deals with four topics (a) the effects of the intervention at the group level, (b) the pupils’ attitudes towards the program, (c) the effectiveness of the program at the individual level, and (d) how to further develop the content of the intervention program.
(JB)SWL II_f19_316-337
9/26/02
9:23 AM
Page 332
Precursors of Functional Literacy 90 80 70
Jari Sini
60
Vera Teemu Jessi conf. level
50 40 30 pre
post
delayed
Figure 4. The development of word recognition skill in the poor control group
Our first research question focused on the effects of the intervention on phonological awareness, phonological memory, word recognition, and spelling. However, the internal reliabilities of the phonological awareness tasks proved to be low, which prevented us from analyzing the developmental group differences in phonological awareness. One reason may be the items used in the tasks as they were similar to those used with younger, pre-literate children. It could be hypothesized that by selecting more difficult items (e.g., for the segmentation task, phonemes that are specific to English and non-existent in Finnish), by using pseudowords as items, and/or by demanding more difficult operations (e.g., deletion of consonant clusters of pseudowords), development in this skill could have been obtained. For example, Scarborough, Ehri, Olson, and Fowler (1998) used pseudowords with complex operations to examine adolescents’ phonemic awareness skills in their native language (English), and found wide variability (Study 1) as well as a clear increase in scores from grade 2 to 12 (cross-sectional study, Study 2). Overall, the results of the present study showed that by focusing at-risk pupils’ attention on the sound structure of the English language, on phonological awareness, and by explicit teaching of the grapheme-to-phoneme correspondences and phonetic signs, it is possible to enhance their proficiency in the English language. Pupils showed development especially in skills which tap understanding of graphemeto-phoneme correspondences and the phonology of English, as the intervention group outperformed the control group in phonological memory, word recognition and spelling. Their stronger phonological memory skills suggest that pupils in the intervention group developed their ability to create accurate representations of the English-sounding pseudowords. This result is in line with the intervention study conducted by Service (1993). In both studies, emphasis was laid on listening
(JB)SWL II_f19_316-337
9/26/02
9:23 AM
Page 333
Promoting at-risk pupils’ foreign language literacy learning
and pronouncing particular phonemes and phoneme combinations typical of the English language. Therefore, the decision to use a multiform way of presenting (by listening, reading and pronouncing) the material to be taught enhanced the construction of the sound system of English. Because former studies have found that phonological memory is connected to various linguistic skills, it would be interesting, in further studies, to examine whether development in phonological memory has transfer effects on, e.g., vocabulary acquisition. The most important effects of the intervention were, however, obtained on the development of English word recognition and spelling. Development in word recognition skills was on a smaller scale than that in spelling. The intervention especially affected the fluency of word recognition, shown as a greater decrease in reading time compared to that of the control group. However, both groups developed equally in the accuracy of word recognition. The intervention group improved significantly on all three measures of English spelling. They were able to write more words and sentences correctly than the control group, and, in general, were able to perform the spelling task more efficiently, as was shown by the increased number of attempted words. These results mean that pupils in the intervention group were not only able to process the single words more accurately, but also their skills developed at sentence level. Instruction in the intervention program focused more on written language than is usual in classroom instruction, which can be one possible reason for the positive development. In addition to this, the main focus in the intervention was on learning to decode and spell English, and on increasing the pupils’ knowledge of what is behind these skills. It seems that weak pupils require that e.g., phonetic signs and their usefulness are explicitly taught; they are not able to infer the rules on their own. The comparisons made between the poor intervention group and the poor control group revealed no significant differences on any of the measures. However, the average intervention group benefited significantly on the combined word recognition and spelling measure compared to the average control group. The skills of the average pupils were already at a higher level in the pre-test, compared to the poor pupils, which may be one reason for their more significant development. In addition, most of the poor pupils had difficulties in three out of the four skills used as criterion tasks, whereas the average pupils had difficulties in only one or two. The individual, developmental graphs clearly show that even within the poor intervention group, three pupils benefited markedly from the intervention program both in word recognition and spelling skills. However, with the skills measured in the present study we could not explain their development, or the weak development of the resistant pupils. There are at least two factors to be included in further studies, namely, metacognitive skills and motivational orientation towards learning tasks. For example, in the study by Vauras, Rauhanummi, Kinnunen, and Lepola (1999), the resistant group (for a strategy instruction focused on comprehension and mathematics) showed lower comprehension monitoring, higher social dependence on adult guidance, and higher psychological vulnerability compared to the responsive group.
(JB)SWL II_f19_316-337
9/26/02
9:23 AM
Page 334
Precursors of Functional Literacy
In the present study, we analysed the responsive and resistant pupils’ responses to the open question (What did you learn today?) in the learning diary used during the intervention, and found preliminary support for better skills in selfassessment among responsive pupils. They wrote more answers, which were also more specific than the resistant pupils’ answers. For example, the two most often used answers in the resistant group were “words” or “sentences” showing weak self-assessment. Efficient metacognitive skills, such as self-regulation and selfassessment, are crucial skills for all learning. The way in which we emphasized metacognitive skills in our intervention program was clearly inadequate. A learning diary is useful, but more attention should have been paid to explicit instruction on self-regulation skills. The pupils’ self-evaluations about their feelings and success revealed that they enjoyed the intervention lessons. Overall, their ratings were high, and feelings towards the particular lessons were rated even more highly than their success. No differences either in feeling or in success were obtained between the themes taught. Why was this kind of instruction so enjoyable? There are at least two reasons. First, the content of the lessons was different from that of the regular class. Second, teaching was as much as possible suited to their zone of proximal development. Therefore, it is possible that, during the intervention lessons, these pupils felt successful and had the opportunity to experience understanding and learning while studying the English language. On the basis of the experiences of running this program, it seems that stronger results could be obtained with some improvements. The pupils seemed to have a rather solid phonological awareness of English, so practice on phonological awareness could be reduced, but not excluded, because it was a good start to the program, giving poor pupils opportunities to be successful. A good theme to begin with would be rhyming skills, because pupils enjoyed it and had direct contacts with written English. In connection with a reduction in phonological awareness training, teaching the grapheme-to-phoneme correspondences and phonetic signs should be started earlier in the program to ensure enough practice on phonetic signs as well as on reading and spelling skills. Connecting the learning of the grapheme-to-phoneme correspondences to text reading and vocabulary learning was a good solution. By this method, transfer to homework may be achieved because explicit teaching on various ways of learning vocabulary was dealt with during the last theme. Transfer to both homework and classroom settings could be enhanced by paying more attention to developing pupils’ metacognitive skills, such as self-assessment and self-regulation. In addition, in further studies, attention should be paid to which skills to include and to what kind of intervention (group vs. individual teaching) would be more suitable for the poorest pupils. This intervention program as such was not enough to improve the skills of the very poor pupils (those excluded from the statistical analyses). They may require an intervention with, at least to some extent, different content, or else an individually constructed program would be more suitable.
(JB)SWL II_f19_316-337
9/26/02
9:23 AM
Page 335
Promoting at-risk pupils’ foreign language literacy learning
Appendix A. The text used to measure word recognition skill in the pre-test.
Two seasons: Summer and winter Spring, summer, autumn and winter — these are the four seasons that make up our year. In summer, days are long and nights are short. The sun can even shine the whole day long. The weather is usually dry, sunny and hot. It doesn’t rain very much. It’s the perfect time for enjoying yourself out-of-doors. There you can see all kinds of flowers and animals. You can swim, play and fish. Winter days are short and dark. The weather is much colder and the nights can be frosty. Snow starts to fall and the nature is beautiful. It is time for winter sports like skiing and skating. You must remember to wear warm clothes and winter boots. Appendix B. The test used to measure spelling skill in the pre-test. 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12.
My name is Mary. I’m eleven years old. We live in a high house. Do you read books? He doesn’t eat apples. Our father loves fast cars. Can you hear the train coming? I think it’s late again. I don’t see the shop in the city. The girl has got beautiful flowers. Do you want juice or a cup of coffee? My brother has got one goose and three chickens.
Notes * We wish to thank the head, teachers, and pupils in the primary school in Turku where this study was conducted. Unfortunately they must remain anonymous. We would also like to thank Elisabeth Service for her help in designing the intervention program. Special thanks are owed also to the reviewers for their useful comments. Correspondence concerning this article should be addressed to Mia Dufva, Centre for Learning Research, University of Turku, FIN-20014 Turku, Finland. Phone +358 2 333 8522. Electronic mail can be sent to
[email protected].
(JB)SWL II_f19_316-337
9/26/02
9:23 AM
Page 336
Precursors of Functional Literacy
References Adams, M. J., Foorman, B. R., Lundberg, I. & Beeler, T. (1998). Phonemic awareness in young children. A classroom curriculum. Baltimore: Brookes. Alderson, J. C. (1984). Reading in a foreign language: A reading problem or a language problem? In J. C. Alderson & A. H. Urquhart (Eds.) Reading in a foreign language (1–27). London: Longman. Baddeley, A. (1986). Working memory. Oxford: Oxford University Press. Baddeley, A. (1990). Human memory. Theory and practice. Hove: Erlbaum. Bernhardt, E. B. (1991). Reading development in a second language: Theoretical, empirical, and classroom perspectives. Norwood, NJ: Ablex. Brumfit, C. J. (1980). Problems and principles in English teaching. Oxford: Pergamon Press. Byrne, B. & Fielding-Barnsley, R. (1991). Evaluation of a program to teach phonemic awareness to young children. Journal of Educational Psychology, 83, 451–455. Byrne, B. & Fielding-Barnsley, R. (1993). Evaluation of a program to teach phonemic awareness to young children: A 1-year follow-up. Journal of Educational Psychology, 85, 104–111. Castro, G. (1987). Plasticity and the handicapped child. In J. J. Gallagher & C. T. Ramey (Eds.), The malleability of children. Baltimore: Brookes. Cummins, J. (1991). Interdependence of first- and second-language proficiency in bilingual children. In E. Bialystok (Ed.), Language processing in bilingual children (70–89). Cambridge: Cambridge University Press. Dr. Seuss’s ABC CD-ROM (1995). Novato, CA: Living Books. Dufva, M. & Voeten, M. J. M. (1999). Native language literacy and phonological memory as prerequisites for learning English as a foreign language. Applied Psycholinguistics, 20, 329–348. Durguno[lu, A. Y. (1997). Bilingual reading: Its components, development, and other issues. In A. M. B. de Groot & J. F. Kroll (Eds.), Tutorials in bilingualism: psycholinguistic perspectives (255–276). Mahwah, NJ: Erlbaum. Durguno[lu, A. Y. & Hancin, B. J. (1992). An overview of cross-language transfer in bilingual reading. In R. J. Harris (Ed.), Cognitive processing in bilinguals (391–412). Amsterdam: North-Holland. Durguno[lu, A. Y., Nagy, W. E. & Hancin-Bhatt, B. J. (1993). Cross-language transfer of phonological awareness. Journal of Educational Psychology, 85, 453–465. Fitzgerald, J. (1995). English-as-a-second-language learners’ cognitive reading process: A review of research in the United States. Review of Educational Research, 65, 145–190. Gathercole, S. E. & Baddeley, A. (1993). Phonological working memory: A critical building block for reading development and vocabulary acquisition? European Journal of Psychology of Education, 8, 259–272. Gathercole, S. E., Willis, C. S., Emslie, H. & Baddeley, A. D. (1992). Phonological memory and vocabulary development during the early school years: A longitudinal study. Developmental Psychology, 28, 887–898. Johnson, K. (1996). Language teaching and skill learning. Oxford: Blackwell. Koda, K. (1996). L2 word recognition research: A critical review. The Modern Language Journal, 80, 450–460. Lindeman, J. (1998). Ala-asteen lukutesti [The Comprehensive School Reading Test]. Turku, Finland: Centre for Learning Research. Lundberg, I., Frost, J. & Petersen, O.-P. (1988). Effects of an extensive program for stimulating phonological awareness in preschool children. Reading Research Quarterly, 23, 268–284. Odlin, T. (1989). Language transfer. Cambridge: Cambridge University Press. Olofsson, Å. & Lundberg, I. (1983). Can phonemic awareness be trained in kindergarten? Scandinavian Journal of Psychology, 24, 35–44.
(JB)SWL II_f19_316-337
9/26/02
9:23 AM
Page 337
Promoting at-risk pupils’ foreign language literacy learning Ponsonby, M. (1995). How now, brown cow? A course in the pronunciation of English. Hertfordshire: Phoenix ELT. Raven, J. C., Court, J. H. & Raven, J. (1995). Manual for Raven’s Progressive Matrices and Vocabulary Scales. Section 2. Coloured Progressive Matrices. Oxford: Oxford Psychologists Press. Scarborough, H. S., Ehri, L. C., Olson, R. K. & Fowler, A. E. (1998). The fate of phonemic awareness beyond the elementary school years . Scientific Studies in Reading, 2, 115–142. Schneider, W. & Näslund, J. C. (1993). The impact of early metalinguistic competencies and memory capacity on reading and spelling in elementary school: Results of the Munich longitudinal study on the genesis of individual competencies (LOGIC). European Journal of Psychology of Education, 8, 273–287. Service, E. (1989) Phonological coding in working memory and foreign language learning. (General Psychology Monographs No. B9) University of Helsinki: General Psychology. Service, E. (1992). Phonology, working memory, and foreign-language learning. The Quarterly Journal of Experimental Psychology, 45A, 21–50. Service, E. (1993). Phonological and semantic aspects of memory. In J. Chapelle and M.-T. Claes (Eds.), Proceedings: 1st international congress on memory and memorization in acquiring and learning languages (307–317). Louvain-la Neuve: CLL. Sparks, R. & Ganschow, L. (1993). The effects of multisensory structured language instruction on native language and foreign language aptitude skills of at-risk high school foreign language learners: A replication and follow-up study. Annals of Dyslexia, 43, 194–216. Sparks, R., Ganschow, L., Kenneweg, S. & Miller, K. (1991). Use of an Orton-Gillingham approach to teach a foreign language to dyslexic/learning disabled students: Explicit teaching of phonology in a second language. Annals of Dyslexia, 41, 96–118. Sparks, R., Ganschow, L., Pohlman, J., Skinner, S. & Artzer, M. (1992). The effects of multisensory structured language instruction on native language and foreign language aptitude skills of at-risk high school foreign language learners. Annals of Dyslexia, 42, 25–53. The Usborne first book of knowledge (1994). London: Usborne Publishing. Vauras, M., Rauhanummi, T., Kinnunen, R. & Lepola, J. (1999). Motivational vulnerability as a challenge for educational interventions. International Journal of Educational Research, 31, 515–531. Verhoeven, L. (1990). Acquisition of reading in a second language. Reading Research Quarterly, 25, 90–114. Wade-Woolley, L. & Siegel, L. S. (1997). The spelling performance of ESL and native speakers of English as a function of reading skill. Reading and Writing: An Interdisciplinary Journal, 9, 387–406. Yopp, H. K. (1988). The validity and reliability of phonemic awareness tests. Reading Research Quarterly, 23, 159–177. Address University of Turku Centre for Learning Research FIN-20014 Turku FINLAND phone: + 35823338522 Email:
[email protected]
(JB)SWL II_f20_338-356
9/26/02
9:24 AM
Page 338
(JB)SWL II_f20_338-356
9/26/02
9:24 AM
Page 339
Predictors of adult functional reading skills1 Elisabeth Arnbak Institute of Educational Psychology Paedagogical Psychology Danish University of Education, Copenhagen
The definitions of functional literacy used in a number of national and international large-scale studies over the last twenty years clearly reflect the increasing need for the adult population of the industrialised countries to be literate at fairly high levels (Felligi & Alexander, 1995; Verhoeven, 1994; Elbro, Møller & Nielsen, 1995). From the UNESCO definition in 1978 to the latest attempt at defining functional literacy skills (OECD/PISA, 1998), more and more emphasis is placed on the ability of the individual to extract information from texts in order to actively use this information in everyday life and to be able to reflect upon the contents of texts for the benefit of the individual as well as the society. That the interest in the functional literacy skills of the populations has not been unwarrented is seen by the fact that insufficient reading skills have been reported in fairly large parts of the populations. As a consequence of the reported literacy problems in the populations different initiatives to improve the written language skills of children and adults have been effectuated in many of the countries.2 A necessary first step in the process of improving insufficient written language skills in adults is to identify the individuals in need of instruction. In the abovementioned large-scale adult literacy studies, the importance of socio-economic factors such as age, gender, primary language, and educational status have been documented (Doets et al., 1991; Kirsch et al., 1993; Smith, 1995; Elbro et al., 1995; IALS, 1995, 1997). Poor functional reading skills have been prevalent in the older population, in poorly educated adults, in adults with little reading experience, and in adults belonging to ethnic minorities. A next step in the process of improving the reading skills of adult poor readers is to motivate them to engage in relevant instructional activities. In Denmark, the campaign for the voluntary reading courses for adults focused on personal experiences of insufficient reading skills, i.e. not being able to read to your child, and the formal criterion for participating in adult reading instruction was the adult’s personal experience of insufficient reading skills. Personal reports of reading difficulties i.e. dyslexia have been found to be highly predictive of the
(JB)SWL II_f20_338-356
9/26/02
9:24 AM
Page 340
Precursors of Functional Literacy
actual reading problems of adults. In a study of the reading skills of adult poor and normal readers, Elbro et al. (1994) found that adults with a former history of reading difficulties were significantly poorer in word decoding than adults with no former history of reading problems even though the two groups were matched on reading comprehension. However, adult poor readers without former histories of reading problems might not be sufficiently aware of their own reading skills. Thus, they might not find this kind of reading instruction relevant. The results of a few studies using adults’ self-ratings of reading skills to examine the degree to which the measured reading skills matched the perceived reading skills of the adult participants seem to support these misgivings. The adults’ self-ratings of reading skills turned out to be poor predictors of their actual reading skills (Scarborough, 1984; Elbro, Møller & Nielsen, 1995). A third step in improving poor written language skills in adults is to identify the possible causes of the reading problems in order to design an effective reading instruction. Even though the latest definitions of functional reading skills highlight the importance of higher order cognitive skills such as the ability to make inferences during reading (OECD/PISA, 1998) there is an overall concensus in reading reasearch of the importance of basic decoding skills to the development of functional reading skills (Shankweiler et al., 1996; Share & Stanovich, 1995; Elbro et al., 1995). Indeed, Perfetti has argued that poor decoding skills is a bottleneck in the development of functional reading skills (Perfetti, 1977). Results of numerous studies of the functional reading skills and sub-skills of young and adult poor readers point to poor decoding skills as an underlying factor in the majority of young and adult poor readers. In these studies it is documented that the poor reading skills of both children and adults are largely caused by phonological coding problems resulting in poor word decoding. (Byrne & Ledez, 1983; Scarborough, 1984; Bruck, 1990; Pennington et al., 1990; Olson, 1989; Elbro et al., 1994; Shankweiler et al., 1996). However, evidence of specific comprehension problems in spite of adequate decoding skills has also been presented in a number of studies. In a study of the reading comprehension of students in 2nd and 3rd grade matched on decoding skills but differing in reading comprehension, Oakhill and Yuill (1991) found that poor reading comprehension was related to poor ability to make inferences during reading, lack of awareness of the inherent structure of the text, and failure to activate personal knowledge during reading. It is, however, still not clear whether these comprehension problems are specific to written language activities or whether they are symptoms of a more general language comprehension problem. Finally, for practical-economical, instructional, and research purposes, a lot of effort and interest has been put into attempts at sub-grouping the adult poor readers. In Denmark, adult poor readers attend either a special eduation class for adults (dyslexics) or voluntary reading classes (adults with problems at more advanced levels of reading skills i.e. problems with specific text genres or specific comprehension skills). Fowler and Scarborough (1998) have reviewed a large num-
(JB)SWL II_f20_338-356
9/26/02
9:24 AM
Page 341
Predictors of adult functional reading skills
ber of studies of adult poor readers. They found that no clear distinction could be made between adults with specific reading disabilities and low achieving adults with more general problems in learning to read and write both with regard to their reading profiles and their socio-economic background profiles. Furthermore, learning to read and write seemed to be a life-long struggle for a large part of the reading disabled adults. Problems in learning to read and write persisted into adulthood for quite many reading disabled children to such an extent that the best predictor of the reading level of adult poor readers was the severity of the reading problems in childhood. Who are the adults attending voluntary reading instruction, what are the causes of poor reading skills in these adults, and what factors may have influenced their self-concept as readers are the questions at interest in this study.
The present study The main purpose of this study was to examine some subskills in reading and some background factors of two groups of adult students. The two adult reader groups had functional reading skills at the same level (the mean sum score and standard deviations of the two groups are presented in table 2). However, one group, adults in voluntary reading classes, was experiencing reading problems to such an extent that they engaged in voluntary reading instruction whereas the other group, adults in formal adult education, did not. We wanted to see if we could identify subskills in reading and specific background information factors that might explain the awareness of insufficient reading skills in the adults in voluntary reading classes. As self-rated reading skills turned out to be poor predictors of the actual reading skills of adults in a number of studies, we expected self-reported reading problems to be related to poor decoding skills. Poor decoding skills resulting in poor spelling skills and slow inautomatized reading would be more apparent in the daily life of the adults than problems at higher levels of reading skills, i.e. knowledge of text genres or inferential skills. On the other hand, the poor functional reading skills of adults in formal adult education might be more related to higher order comprehension problems than problems at a more basic level (poor decoding skills). We furthermore expected self-reported reading problems to be related to former attendance in special education activities. Self-reported reading problems might be a result of a former diagnosis of reading problems rather than awareness of personal reading skills. Finally, as a large number of studies have documented a relation between poor reading skills and educational status we thought possible differences in the educational level of the adults might be related to differences in self-reported reading problems.
(JB)SWL II_f20_338-356
9/26/02
9:24 AM
Page 342
Precursors of Functional Literacy
Participants A total number of 773 students and adults from five different counties in Denmark participated in the original study. The adults in the study were engaged in five different educational activities. In this chapter, only data from two adult reader groups enlisted in two different educational activities (362 persons in all) will be reported. The two adult reader groups were quite similar with regard to mean age and age range (see table 1). Thus, the adults had had more or less the same opportunities of basic schooling, special education, and of further education. Differences in background information could not be explained as differences in formal educational opportunities. One adult reader group consisted of 168 adults participating in voluntary adult reading courses. Based on the description of the task group for this kind of instructional activity, the adults should have poor functional reading skills (poor reading comprehension) but decoding skills within the normal range. The second reader group was 194 adults in formal adult education. They were studying to take a secondary school degree. The adults were unskilled workers in need for further education to maintain their job or perhaps to qualify for another one. This group, too, was expected to have poor functional reading skills. The percentage of males and females in the voluntary reading courses were much the same, but the percentage of females in formal adult education were somewhat larger compared to the percentage of males. Finally, there was a fairly equal amount of adults belonging to different ethnic minorities in the two adult reader groups.
Table 1. Mean age (standard deviations in brackets) and age range, percentage of males and females, and primary language in percentage of adults in voluntary reading courses and adults in formal adult education
Mean age Age range Gender (% of all) Males/females no response (% of all) Primary language Danish (% of all) Other languages No response (% of all)
Adults in vol. reading courses
Adults in formal adult education
34 (12) 17–71 40/60 25
37 (14) 16–68 23/77 28
98 2 27
84 16 28
(JB)SWL II_f20_338-356
9/26/02
9:24 AM
Page 343
Predictors of adult functional reading skills
Instruments Functional reading skills A recently developed screening of functional reading skills was used to assess the reading skills of participants. The functional reading screening, LfUV (Arnbak & Elbro, 1999), consists of a total number of 16 texts presented in three sections. The texts have been chosen from a large pool of texts either sent by mail to all Danish adults or available for free in public institutions and libraries or printed in newspapers. The functional reading screening is designed to measure two different aspects of reading comprehension: ability to read different text types and comprehension skills at three different levels. The texts in the screening represent three different text types. The first text group is narrative texts relating some concluded incident i.e. a story about the failed attempt of two bank robbers to escape in a stolen car. The second text group is expository texts informing (briefly) of different subjects i.e. the recommended weekly maximum amount of alcohol for males and females (informative texts), and the third text group consists of documents, i.e. tables or a street map. Six narrative texts, five informative texts, and five documents were included in the reading screening. Reading comprehension is measured by using a multiple-choice questions format. There are a total number of 60 questions, 20 questions for each of the three text groups. The questions are designed to examine reading comprehension at three different levels (factual level, paraphrasial level, and inferential level). Questions at the factual level examine the ability to locate literal information in the texts, i.e. the price of a ferry ticket for a car. Questions at the paraphrasial level examine the ability to locate factual information worded in a different, but synonymous way of what is stated in the text. At the inferential level, questions measure inferential skills or the ability to piece together information from different parts of a text. The number of questions for each of the three levels of comprehension depends on the genre and purpose of each text. There are 25 factual questions, 24 paraphrasial questions, and 11 inferential questions in all. The reading of the three text groups is timed. The time given to read each text group is fifteen minutes, a total amount of forty-five minutes.
Decoding We expected some of the adults might to have problems in decoding. Accordingly, a test of decoding skills (Nielsen & Petersen, 1992) was included in the assessment. The decoding test is a non-word-reading test of 38 items. The participants were given five minutes to circle one of four non-words sounding as a real word in each of the 38 items.
(JB)SWL II_f20_338-356
9/26/02
9:24 AM
Page 344
Precursors of Functional Literacy
Questionnaire of background information Socio-economic information The participants answered questions about their age, gender, and primary language. Furthermore, they indicated their exam level from a list of the most common exams and their level of education from a similar list (see tables 6 and 7). History of reading problems The participants indicated if they were dyslexic, if they had any relatives with reading problems, if they had participated in special education in school and as an adult. Self-ratings of reading skills The participants were asked to rate their reading skills and reading speed on a scale from 1 (very poor) to 13 (very good). Furthermore, the adults judged the difficulty of three different text types on a scale from 1 (very difficult) to 13 (very easy). These measures were used as external validators of the functional reading skills of the adults. Procedure The adults were assessed in groups in their class rooms by a reading instruction teacher or a Danish language teacher. About six months after the screening of their reading skills the participants answered the questionnaire about various background information.
Results The results are reported in two sections. In the first section, the reading skills and subskills of the adults will be reported. In the second section, the data from the questionnaires will be presented.
The functional reading skills and subskills of the two adult reader groups. As poor readers might have problems with reading accuracy as well as with reading speed, mean number of questions answered, mean number of correctly answered questions as well as mean percentage of correctly answered questions of all passed questions are reported below for the two adult reader groups. A series of ANOVAS were performed and post hoc analyses of significant differences indicated that adults in formal adult education were a little faster readers than adults in voluntary reading courses (see table 2). However, as mentioned above, no significant differences were found in the functional reading skills of
(JB)SWL II_f20_338-356
9/26/02
9:24 AM
Page 345
Predictors of adult functional reading skills Table 2. Functional reading skills in adults in voluntary reading courses and adults in formal adult education (mean scores and standard deviations added in brackets) Reading scores
Adults in vol. Adults in formal Significant group reading courses adult education differences
Number of correct (max 60) 37.36 (11.10) Number of passed items (max 60) 49.35 (9.33) % correct of all items passed 74.78 (14.04)
40.59 (9.74) 52.66 (7.63) 76.35 (11.50)
* * Ns.
* = p < 0.05 Table 3. Genre-specific reading skills in adults in voluntary reading courses and adults in formal adult education (mean scores and standard deviations added in brackets) Genre-specific reading skills Documents (max 20) Number of correct Number of passed % correct of all Informative texts (max 20) Number of correct Number of passed % correct of all Narrative texts (max 20) Number of correct Number of passed % correct of all
Adults in vol. reading courses
Adults in formal adult education
Significant group differences
12.09 (4.03) 15.70 (3.94) 75.74 (14.16)
12.59 (3.52) 16.14 (3.47) 77.51 (12.47)
Ns. Ns. Ns.
14.09 (3.69) 17.63 (3.09) 79.88 (15.16)
15.11 (3.16) 18.65 (2.28) 80.72 (12.52)
Ns. * Ns.
11.18 (4.55) 16.01 (3.64) 68.59 (21.0)
12.89 (4.24) 17.86 (3.06) 71.08 (17.72)
* * Ns.
* = p < 0.05
adults in voluntary reading courses and adults in formal adult education on measures of reading efficiency (percentage correct of all passed items). The standard deviations of the two above-mentioned groups were quite large indicating that both groups were quite inhomogenuous with regard to functional reading skills.
Genre-specific reading skills The results for each of the three text groups revealed the same tendencies as the combined results (see table 3). No significant differences were found between the two reader groups in the ability to read documents. Adults in formal adult education read the informative — and narrative texts significantly faster than adults in voluntary reading courses. However, no significant differences were found on a measure of reading efficiency (percentage of correct of all passed items). Thus, they were not more accurate readers of these texts than adults in voluntary reading courses.
(JB)SWL II_f20_338-356
9/26/02
9:24 AM
Page 346
Precursors of Functional Literacy
Finally, no interaction was found between reader group and text type. The informative texts were the easiest to read for both reader groups, the documents a little harder, and the narrative texts the hardest. Thus, poor functional reading skills did not seem to be more related to poor knowledge of specific text genres in one reader group than in the other.
Comprehension skills As mentioned above, we wanted to examine possible group differences in comprehension skills (see table 4). We hypothesized that the functional reading problems of the two reader groups might be caused by problems in different sub-skills. Individual differences were in fact found in the ability to answer the questions of the three different levels of comprehension. However, no differences were found between the two reader groups for any of the three levels of reading comprehension. The pattern of results was identical to that of the above mentioned results. Again, adults in formal adult education were significantly faster, but there were no significant differences between the two groups on measures of reading efficiency (percentage correct of all answered questions) for the three levels of reading comprehension. Furthermore, no significant interaction was found between reader group and comprehension skills. The factual questions were the easiest to answer for both groups, the paraphrasial questions a little harder, and the inferential questions were the most difficult. Thus, poor inferential skills did not seem to be more related to poor functional reading skills in one reader group than in the other. Table 4. Three levels of comprehension skills in adults voluntary reading courses and adults in formal adult education (mean scores and standard deviations added in brackets)
Factual level (max 25) Number of correct Number of passed % correct of all Paraphrasial level (max 24) Number of correct Number of passed % correct of all Inferential level (max 11) Number of correct Number of passed % correct of all * = p < 0.05
Adults in vol. reading courses
Adults in formal adult education
Significant group differences
18.08 (4.52) 21.23 (3.75) 84.43 (12.04)
19.23 (3.83) 22.18 (3.05) 86.26 (9.74)
Ns. Ns. Ns.
12.65 (4.75) 19.13 (4.21) 65.61 (18.25)
14.20 (4.47) 20.94 (3.42) 66.92 (15.53)
* * Ns.
6.63 (2.60) 8.98 (1.80) 71.39 (22.58)
7.16 (2.25) 9.54 (1.55) 73.75 (17.99)
Ns. * Ns.
(JB)SWL II_f20_338-356
9/26/02
9:24 AM
Page 347
Predictors of adult functional reading skills
Decoding skills Whereas no significant differences were found in the functional reading skills of adults in voluntary reading courses and adults in formal education, significant differences were found in the decoding skills of these two reader groups (see table 5). Adults in formal adult education significantly outperformed the adults in voluntary reading courses. Table 5. Decoding skills in adults in voluntary reading courses and adults in formal adult education (mean scores and standard deviations added in brackets)
Decoding skills (max. 38) Number of passed Number of correct
Adults in vol. reading courses
Adults in formal adult education
22.48 (8.33) 11.72 (6.26)
27.92 (6.96) 17.66 (8.13)
Significant group differences * *
* = p < 0.05
The relation between decoding skills and functional reading skills Comparing the mean scores in decoding for the reader groups with their mean functional reading scores revealed interesting group differences. A significant interaction was found between reader groups and reading task (text versus word decoding). Decoding was discrepantly poor in adults in vouluntary reading courses (p < 0.05). The importance of decoding skills to aspects of reading comprehension As mentioned above, even though no significant group differences were found in the ability to read different text types or in inferential skill differences in these areas did exist at the individual level. As possible differences in these areas might be collapsed because of the way the participants were grouped we divided the participants in four groups based on their decoding skills (quartiles). In figure 1, the mean scores of the four decoding groups for the three text types are presented. Analyses of individual differences revealed an interaction between decoding skills and text group for the poorest group of decoders. They were significantly worse at reading narrative texts than documents compared to the three other decoding groups (p < 0.05). The same analyses were performed with the three levels of comprehension (factual, paraphrasial, and inferential). No significant differences were found between any of the four decoding groups with regard to levels of comprehension.
(JB)SWL II_f20_338-356
9/26/02
9:24 AM
Page 348
Precursors of Functional Literacy 25
20
15
10 Narrative texts, sumscore Informative texts, sumscore
5
Documents, sumscore 0 N = 144 144 144
95 95 95
74 74 74
46 46 46
2,00
3,00
4,00
1,00 Decoding, quartiles
Figure 1. Mean sumscore of narrative texts, informative texts, and documents for four decoding groups (quartiles) Table 6. Self-reported dyslexia, reading problems in the family, special education in school and as an adult as percentage of all adults in voluntary reading courses and adults in formal adult education Reading problems Self-reported dyslexia Reading problems in the family Special instruction courses in school Adult special instruction courses
Adults in voluntary reading courses
Adults in formal adult education
33 % 49 % 61 % 44 %
10 % 30 % 27 % 11 %
The background information profiles of the two adult reader groups Personal histories of reading difficulties In a number of studies, personal reports of reading problems correlate quite highly with functional reading skills (reading comprehension). A surprisingly large number of adults in voluntary reading courses reported that they were still experiencing or had experienced reading problems (see table 6). Even though the two reader groups did not differ with respect to functional reading skills the percentage of adults in voluntary reading courses who thought of themselves as dyslexic was much higher (33 %) than the percentage in the other reader group (10 %).
(JB)SWL II_f20_338-356
9/26/02
9:24 AM
Page 349
Predictors of adult functional reading skills Table 7. Mean score on a scale from one (very poor) to 13 (very good) of self-rated reading skills, reading speed, and judgement of the difficulty of different texts by the adults in voluntary reading courses and adults in formal adult education. Standard deviations are added in brackets Ratings of reading skills
Adults in voluntary reading courses
Reading ability in general Reading speed Judgement of the difficulty of different texts
7.73 (2.18) 6.42 (2.30) 8.34 (1.89)
Adults in formal adult education 9.15 (2.25) 7.73 (2.10) 8.90 (1.72)
Table 8. The level of exam for adults in voluntary reading courses and adults in formal adult education (in percentages) Exam Lower Secondary School Leaving Examination Final Examination, 9th grade Final Examination, 10th grade General Certificate of Education, Advanced Level Other exams No exam No response (% of all)
Adults in voluntary Adults in formal reading courses adult education 2% 24 % 36 % 11 % 19 % 9% 25 %
5% 39 % 20 % 3% 18 % 15 % 27 %
The same pattern was seen in reports of relatives with reading problems. About half of the adults in vol. reading courses had relatives with reading problems whereas about a fourth of the adults in formal adult education had relatives with reading problems. About twice as many adults (61 %) in voluntary reading courses had attended special education in school as had adults in formal adult education. Finally, about half of the adults in voluntary reading courses had also received special education as adults. The percentage of adults who had attended special education courses, as adults were much smaller in the group of adults in formal adult education (11 %).
The self-rated reading skills The adults in the two reader groups were asked to rate their own reading skills and reading speed on a scale from 1 (very low) to 13 (very good). The mean score of self-rated reading skills in the group of adults in the voluntary reading courses were significantly lower (p < 0.05) than that of the adults in formal education even though the two groups did not differ in functional reading skills (see table 7). The same trend was seen in the self-ratings of reading speed in the two groups. The self-rated reading speed of the adults in voluntary reading courses was significantly lower than that of the adult in formal education (p < 0.05). This difference between the self-ratings of reading speed in the two groups, however, was in accordance with the actual results of the study. As mentioned above, adults in
(JB)SWL II_f20_338-356
9/26/02
9:24 AM
Page 350
Precursors of Functional Literacy Table 9. Percentage of adults with different levels of education for adults in voluntary reading courses and adults in formal adult education Level of education Vocational training Educational training Medium long term education (corresponding to a first university degree) Higher education (corresponding to a university degree) Other education No education No response
Adults in voluntary reading courses
Adults in formal adult education
22 % 6%
5% 9%
2%
3%
2% 38 % 42 % 25 %
0% 22 % 61 % 27 %
formal education did significantly outperform the other group on a measure of reading speed (mean number of questions answered).
Educational profiles No major differences existed in the educational profiles of the participants in the two reader groups. The most common level of exam for both adult reader groups was a Final Examination after 9th or 10th grade (see table 8), and in both groups, about twenty percent had an unconventional exam (refered to as ‘other’ in the table). However, the groups did in fact differ with respect to the percentage of adults with no final exam. Only nine percent of the adults in voluntary reading courses had no final exam whereas 15 % of the adults in formal adult education had left school with no final exam. Analyses of the educational level of the two adult reader groups revealed no striking differences between the groups. As expected (see table 9), both groups of adults were characterised by a rather low educational level. Close to half of the adults in voluntary reading courses and even more of the adults in formal adult education (61 %) had no formal education. Nevertheless, adults in voluntary reading courses seemed to be somewhat better educated than adults in formal adult education. In the voluntary reading class-group, 22 % of the adults had some kind of vocational education, and 38 % had taken another kind of education than the above mentioned. In the group of adults in formal adult education, only 5 % had a vocational education and 22 % had another education than the above mentioned. Predictors of adult reading skills Self-reported reading problems and actual reading skills In both reader groups, adults who reported that they were dyslexic were significantly poorer in decoding than the rest of the reader group. The mean decoding scores
(JB)SWL II_f20_338-356
9/26/02
9:24 AM
Page 351
Predictors of adult functional reading skills
for the self-reported dyslexics in voluntary reading courses were 9.2 and 13.16 for the rest of the adults (p < 0.01). The mean decoding scores for self-reported dyslexics in adult formal education were 13.0 and 16.8 for the rest of the adults (p < 0.01). No significant differences were found in the functional reading skills of adults with self-reported dyslexia in the two reader groups and the rest of the groups. The decoding skills of adults in voluntary reading courses with relatives who had reading problems were significantly poorer than the rest (M: 10.0 and 13.2, p < 0.01). No such differences were found for the adults in formal adult education. Adults in voluntary reading courses who had received special education during school years and/or as an adult did significantly poorer in decoding, but not in functional reading than the rest of the group (mean decoding scores: 10.0 and 14.5, p < 0.01). Contrary to this, no differences were found in the decoding skills of adults in formal adult education that had received special education during school years and the rest of the group. However, adults in formal adult education who had received special education as adults were significantly poorer both in decoding and functional reading than the rest (mean decoding scores: 14.2 and 19.0, and mean functional reading scores: 36.2 and 42.0, p < 0.01).
Self-ratings of reading skills The overall impression was that the adult reader groups were not very good at rating their own reading skills. The self-ratings of reading skills (reading ability and reading speed) of the adults in voluntary reading courses did not correlate with neither decoding skills nor functional reading skills. The correlations between the self-rated reading skills and the actual reading skills of adults in formal adult education were a little better (r = 0.33, p < 0.01 for decoding and r = 0.37, p < 0.01 for functional reading skills). Educational status and reading skills A relation was found between the level of exam and education and the functional reading skills of the adults in voluntary reading courses. Adults with a low exam level (no exam, other exam, or a final exam, 9th grade) had significantly poorer functional reading skills than the rest of the group (F = 13.13; p < 0.01). Likewise, adults in voluntary reading courses with no education had poorer functional reading skills than the rest of the group (F = 6.715; p < 0.01). No such differences were found for the adults in formal adult education. Discussion The aim of this study was to examine some subskills in reading and some background information of adults reading at the same level but differing with respect to self-reported reading problems. We expected poor decoding skills to be related to self-reported reading problems. We further expected that poor functional reading
(JB)SWL II_f20_338-356
9/26/02
9:24 AM
Page 352
Precursors of Functional Literacy
skills in adults who did not report reading problems might be more related to problems in higher level reading skills, i.e. knowledge of text genres or inferential skills. Adults in voluntary reading courses were significantly poorer in decoding than adults in formal adult education. In fact, adults in voluntary reading courses were performing at the same level as adult dyslexics in decoding. However, they significantly outperformed the adult dyslexics in reading comprehension (Arnbak & Elbro, 1999). The adults in voluntary reading courses seemed to be relatively good at reading everyday-life texts in view of their poor decoding skills. This result might indicate that the adults in voluntary reading courses had somehow succeeded in compensating for their poor decoding skills. Further studies of the reading strategies, cognitive-linguistic, and socio-economic profiles of adult poor readers are clearly needed in order to answer this question. No significant differences were found between adults in voluntary reading courses and adults in formal adult education on measures of knowledge of text genres or comprehension skills. Thus, we were unable to find any indication of group-specific differences in higher order cognitive skills related to reading comprehension. It is possible, though, that differences between the groups in comprehension-related skills might have been found if more specific measures of language comprehension skills (i.e. vocabulary) had been included in the study. We wanted to see if poor decoding skills rather than reader group membership were important for the ability to read and understand different text genres and for the comprehension skills of the adults. A significant interaction was found between decoding groups and text types. The poorest group of decoders was relatively better at reading documents than narrative texts. This result might point in the direction of a specific text genre difficulty in the group of the poorest decoders. However, the number of words to be read in narrative texts before you can answer a specific question is much larger than in documents. Thus, the difference between the poorest decoders and the other three decoding groups in the reading of documents and narrative texts might just as well be due to the amount of words to be read in order to answer a question. This would indicate word decoding problems rather than poor genre specific knowledge. Studies examining functional methods of analysing the task difficulty of i.e. documents indicate that cognitive factors (i.e. processes involved in locating information in a text in order to answer a question) as well as the amount of information in a text (i.e. the number of sentences) predict document difficulty (De Geus & Reitsma, 1994). These results indicate that higher order cognitive skills as well as basic skills (decoding skills) influence adults’ ability to identify and use information in different texts. Interesting differences in background information profiles were found between the two adult reader groups reading at the same level but differing in self-reported reading problems. Adults in voluntary reading courses reported histories of reading problems to a much higher degree than the adults in formal education. Information of former and present reading problems in the group of adults in
(JB)SWL II_f20_338-356
9/26/02
9:24 AM
Page 353
Predictors of adult functional reading skills
voluntary reading courses fitted well the actual decoding problems of this group. Personal reports of reading problems correlated with decoding problems rather than with reading problems in general. Thus, the personal experience of being a poor reader to a very large extent seemed to be dependent upon the decoding skills of the individual rather than cognitive skills related to higher order comprehension processes in reading. All in all, poor reading comprehension did not seem to be the most salient feature in the adults’ decision to participate in voluntary reading courses. A lot of reading instruction teachers have reported that their adult students did not join the courses because they wanted to improve their reading skills but rather their spelling skills. Problems with the correct spelling of words are symptoms of decoding problems that are more salient in everyday life than problems in reading comprehension. As in several other studies, personal reports of reading problems turned out to be much better predictors of the actual reading skills of the adults than their self-rated reading skills. Even though adults in voluntary reading courses rated their reading skills significantly poorer than those of the adults in formal education the general impression was that the adults were relatively unaware of their own level of reading skills. The personal experience of insufficient reading skills might to a higher extent be influenced by a former diagnosis of reading problems than an awareness of actual reading skills. Adults in voluntary reading courses tended to have a somewhat higher level of exams and to be better educated than adults in formal adult education. The differences in the educational profiles, the personal histories of reading problems, and in the decoding skills of the group of adults in voluntary reading courses and in formal education seem to indicate different causes of the functional reading problems of the two reader groups. Based on the reading profile of the group of adults in voluntary reading courses it seems reasonable to conclude that participants in the voluntary reading courses in this study did not meet the descriptions of the target group for the voluntary reading courses. Rather than being adults with adequate basic reading skills in need of a brush-up course in reading they turned out to have reading profiles similar to adult dyslexics both with regard to decoding skills and to a history of reading problems. Adults in voluntary reading courses thus seemed to be rather well compensated dyslexics engaged in yet another reading instruction course. The functional reading problems of the adults in formal adult education, on the other hand, seemed to be more related to a poor educational level and lack of reading experience than specific reading problems. Clearly, further studies of the reading profiles, the language skills profiles, and background information of groups of adult poor readers are needed to examine this hypothesis further. A major drawback in this study is the lack of information about the actual jobstatus of the adults and thus the actual demands posed on their reading skills in everyday life. As in a number of other studies, low educational status was related to poor functional reading skills in this study. However, differences in awareness
(JB)SWL II_f20_338-356
9/26/02
9:24 AM
Page 354
Precursors of Functional Literacy
of insufficient reading skills in the two reader groups could not be explained by the eduational status of the adults as no significant differences in the level of exam and education existed between the two adult reader groups. Adults in voluntary reading courses did seem to be somewhat better educated than adults in formal adult education, but as the actual jobsituation of the adults is unknown as well as the actual demands on their reading skills we are not able to conclude anything about the relation between awareness of reading skills and educational level or job situation. Basing the identification of adults in need of reading instruction on the adults’ own perception of themselves as readers does not seem a feasible strategy. Clearly, a treatment of the reading difficulties of adults must be based on a screening of the strengths and weaknesses in written language skills of the individual adult and should take into account the individual needs of written language skills. As seen in the results, there were in fact adult poor readers within the traditional educational system who either did not themselves experience problems in reading or who did not want to participate in the voluntary reading courses. One explanation might be that adult poor readers lack an awareness of the need for the individual to possess the necessary basic skills in reading and maths if he or she is to remain in the work force. Or adult poor readers may not feel that they can do anything themselves to improve their reading skills. Apparently, the campaign for adults to enroll in the voluntary reading courses had succeeded primarily in activating those already within the special education system. This raises the very important question of how to get in touch with adult poor readers who are not already identified by the remedial system. A new approach is clearly needed if a successful attempt is to be made of getting in touch with adult poor readers who have not formerly been engaged in the special education system. Over the last ten years, quite a few attempts have been made of improving the basic skills of the work force. Informal reports of the outcomes of such basic skills courses point to the fact that reading instruction must be integrated in vocational training courses or in the actual job situation of the adult. Otherwise, adult poor readers without a former history of severe reading problems seemingly do not choose to engage in the improvement of their functional reading skills. It might thus be necessary to bring reading instruction out of the traditional classroom scenario and into the workplace of adult poor readers.
Notes . The research project decribed in this paper was financed by the Ministry of Education. The research was done in collaboration with professor Carsten Elbro, Department of General and Applied Linguistics, University of Copenhagen. . As a consequence of the results of the national and international adults literacy studies, a law was passed in Denmark in 1995 enabling any adult who experienced problems in reading, spelling, or writing to receive free reading instruction (The Department of Youth and Adult Education, The Ministry of Education, 1995, 1997). In the guide lines for the
(JB)SWL II_f20_338-356
9/26/02
9:24 AM
Page 355
Predictors of adult functional reading skills law, it is stated that the adult reading courses are not a special education activity, i.e. not intended for adults with severe reading problems (dyslexia). The voluntary reading courses are supposed to be a ‘brush-up’ reading course for poorly educated adults or adults who are now confronted with greater demands on their reading skills as a consequence of changes in society. The formal criterium for participating in an adult reading class is the adult’s personal experience of having insufficient reading, spelling or writing skills. To emphasize the voluntary nature of the participation in an adult reading class no formal screening of the functional reading skills of the adult is required for him or her to enroll in the courses. The functional reading screening used in this study (Arnbak & Elbro, 1996) was developped as a tool for reading teachers in identifying their students’ instructional needs. The functionality, validity, and reliability of the material was examined in a large study with 773 participants. One of the primary goals of this study was to examine the predictive validity of the reading screening, i.e. if the measured reading skills of the adults predicted their study success or failure after a six months period. To examine the validity (and the predictive validity) of the reading screening the participants answered a questionaire about socio-economic factors, self-reported reading problems and reading skills. The teachers too, answered a questionaire about the reading skills of their students and reported the final exam grades for a number of subjects in which written language skills were important. The background variables of the adults reported in this study were selected for the purpose of examining the (predictive) validity of the reading screening. Consequently, important and relevant background variables (i.e. the socio-economic status of the adults, their present jobsituation, salary) related to the reading skills of adults have not been included in this study.
References Arnbak, E. & Elbro, C. (1999). Læsning, læsekurser og uddannelse, om unge og voksnes funktionelle læsefærdighed i uddannelse og på læsekurser vurderet med et nyt materiale [Literacy, literacy instruction, and education — The functional literacy skills of adults in education and in literacy instruction courses measured by a new material]. København: Center for Læseforskning. Bruck, M. (1990). Word recognition skills of adults with childhood diagnoses of dyslexia. Developmental Psychology, 26, 3, 439–454. Byrne, B. & Ledez, J. (1983). Phonological awarenes in reading-disabled adults. Austrian Journal of Psychology, 35, 185–97. De Geus, W. C. & Reitsma, P. (1994). Cognitive components of task difficulty in document literacy. In C. K. Kinzer & D. J. Leu (Eds.) Multidemensional aspects of literacy research, theory, and practice. Chicago: National Reading Conference. Doets, C., Groen, P., Huisman, T. & Neuvel, J. (1991). Functional Illiteracy in the Netherlands. Amersfoort: Netherlands Study and Development Centre for Adult Education. Ebro, C. Borstrøm, I. & Petersen, D. K. (1994). Dyslexia in Adults: Evidence for Deficits in Non-word Reading and in the Phonological Representation of Lexical Items. Annals of Dyslexia, 44. Elbro, C., Møller, S. & Nielsen, E. M. (1995). Functional reading difficulties in Denmark. A study of adult reading of common texts. Reading and Writing An Interdisciplinary Journal, 7. Felligi, I. P. & Alexander, T. J. (Eds.) (1995). Reading, Economy, and Society. Results of the first International Adult Reading Survey. Canada: Organisation for Economic Co-operation and Development Statistics.
(JB)SWL II_f20_338-356
9/26/02
9:24 AM
Page 356
Precursors of Functional Literacy Fowler, A. & Scarborough, H. (1998). Should Reading Disabled Adults be Distinguished from other Adults Seeking Reading Instruction? A Review of Theory and Research. Philadelphia: National Center on Adult Literacy, University of Pennsylvania. IALS (1995). Reading, economy, and society. Paris: OECD. IALS (1997). Reading Skills for the Knowledge Society, further results from the International Adult Reading Survey. Paris: OECD. Kirsch, I. S, Jungeblut, A. & Rock, D. A. (1993). Adult literacy in America: a first look at the results of the National Adult Literacy Survey. Princeton, NJ: Educational Testing Service. Nielsen, I. & Petersen, D. K. (1992). DIAVOK. EN diagnostisk læse- og stavetest. AOF. Oakhill, J. & Yuill, N. (1991). The remediation of Reading Comprehension Difficulties. I Snowling & Thomson (Eds.) Dyslexia Integrating Theory & Pratice. OECD/PISA (1998). Framework for Assessing Reading Reading. Olson, R. K. (1989). Specific deficits in component reading and language skills: genetic and environmental influences. Journal of Learning Disabilities, 22, 339–348. Pennington, B. F., Van orden, G. C., Smith, S. D., Green, P. A. & Haith, M. M. (1990). Phonological processing skills and deficits in adult dyslexics. Child Development, 61, 1753–78. Perfetti, C. A. (1977). Language Comprehension and fast Decoding: some psycholinguistic prerequisites for skilled reading comprehension. In J. Guthie (Ed.) Cognition, Curriculum, and Comprehension. Newark, Delaware: IRA. Scarborough, H. S. (1984). Continuity between childhood dyslexia and adult reading. British Journal of Psychology, 75, 329–48. Shankweiler, D, Lundquist, E., Dreyer, L. G. & Dickinson, C. C. (1996). Reading and spelling difficulties in high school students: Causes and consequences. Reading and Writing: An Interdisciplinary Journal, 8, 267–291. Share, D. & Stanovich, K. (1995). Cognitive processes in early reading development: Accomodating individual differences into a model of reading acquisition. Issues in Education, 1. Smith, C. (1995). Differences in adults’ reading practices and literacy profiles. Reading Research Quarterly, 31, 2, 196–219. Unesco, (1978). Records of General Conference, Resolutions. Vol. 1 Paris: Unesco. Undervisningsministeriets Folkeoplysningsafdeling (1995). Udkast til administrativ og pædagogisk vejledning for forsøg med pilot-læsekurser for voksne. [Ministry of Education, The Department of Youth and Adult Education, Draft: Administrative and Paedagogical guide lines for pilot literacy courses for adults.] Undervisningsministeriets Folkeoplysningsafdeling (1997). Udkast til bekendtgørelse om specialundervisning for voksne og læsekurser for voksne m. v. [Ministry of Education, The Department of Youth and Adult Education, Draft: Ministerial Order on Adult Special Needs Education and Adult literacy courses]. Verhoeven, L. (Ed.) (1994). Functional Reading, theoretical issues and educational implications. Amsterdam: John Benjamins. Address Royal Danish School of Educational Studies Empdrupvej 101 DK-2400 Copenhagen NV Denmark email:
[email protected]
(JB)SWL II_Index_357-359
9/24/02
2:58 PM
Page 357
Subject index alphabetic codes 4, 33 artificial orthography 148, 149 associate learning tasks 72 automatic decoding deficit 230 automatization deficit hypothesis 232 basewords 168–170 biligualism 89, 98, 102, 105, 287, 304 body-coda structure 62, 141, 142, 146, 147, 151, 157 body frequency 61, 63 book orientation 351 cloze task 295, 296 communicative competence 4 composition coherence 217–224 comprehension skills 343, 345 conscious compensation hypothesis 232 consonantic context 127, 129, 134 consonant contrasts 111 decoding skills 269, 341, 345, 352 distinctness hypothesis 23 Dutch 49–63 dyslexia 6, 7, 37, 69–83, 229, 290, 297, 339, 340, 349, 351 effective strategies 265 emergent literacy 6, 46, 249, 260, 262 English as a second language (ESL) 290–293, 298 ethnic minorities 5 Finland-Swedish context 303–313 foreign language 317 French 121–234 functional literacy 3, 4, 6, 339 functional reading skills 339, 341 gating 40, 43 gender differences 303, 304, 310–313 general learning errors 80
graphemes 126, 127 grapheme knowledge 251, 254, 259 grapheme-phoneme conversion rules 289, 294, 295 grapheme-phoneme mapping 6 graphemic regularities 122 grapheme substitution task 140, 145–150 graphotactic constraints 124, 132, 134 graphotactic regularities 124 homography 193 homophones 165, 193–196, 198, 211 homophony 193 inferential skills 341 interdependence hypothesis 298 intervention program 319, 320 Korean writing system 140, 152 language awareness 18 language status 293 leisure time reading 268, 271, 274, 275, 280 letter-sound correspondence 35 lexical awareness 18, 89 lexical quality hypothesis 189–196, 198, 201, 209, 210 lexical representations 191, 195, 198–200 linguistic awareness 17, 299 linguistic devices 5 linguistic hypothesis 155, 156, 158 listening comprehension 269 metalinguistic awareness 89 metalinguistic knowledge 249, 254 metaphonological awareness 90, 93, 96, 104 morphological constraints 130, 132 morphological regularities 123 morphological structure 20 multiletter strings 56–58
(JB)SWL II_Index_357-359
9/24/02
2:58 PM
Page 358
Subject index naming speed deficit 232, 236 natural speech 116 nonwords 38, 40, 76, 144, 145, 153, 122, 129–135, 236, 239 nonverbal intelligence 321, 323 onset 49, 50, 56, 90, 104 onset-rime structures 139, 157–161 onset-rime units 49–63 orthographic depth hypothesis 161 orthographic knowledge 201–203, 208 orthographic neighborhood 166 orthographic processing 297 orthographic regularities 121, 124, 125 orthography 6, 34, 49–51, 165, 161, 193 path model 216, 217 perceptual speed 237 phonemic discrimination 6 phonographic neighborhood 166 phonological awareness 7, 17–31, 34–46, 69–71, 91, 109, 139, 140, 231, 236, 241, 321 phonological coherence model 179, 181 phonological core deficit hypothesis 318 phonological decoding 49 phonological errors 79, 80, 85 phonological knowledge 201–203, 208 phonological processing 72, 74, 82, 289, 294, 299 phonological recoding 229, 230 phonological representations 17, 23–26, 30, 31, 69, 70, 85 phoneme segmentation 40, 45 phonology 165–167 phonological units 49, 69 poetry 22, 23 polysemy 193 positional regularities 133 pseudohomophones 57, 165–177, 181 pseudowords 52, 55, 58, 59, 172–177, 290, 295, 298–300, 321 Pygmalion effects 250 rapid naming (RAN) 70, 73, 76, 83 reading comprehension 216–224, 237 reading engagement 303 reading motivations 266, 267, 268, 270, 282, 303, 304, 305 reading skills 36, 40, 266, 339–342, 345, 346, 353, 354
reading strategies 267, 268, 270, 281, 284 Reading Survey 304, 305 rime 6, 22, 49, 50, 55, 56, 90, 99, 104, 150, 157, 158 rime frequency 61, 63 script dependent hypothesis 298 second language 5, 166, 265, 268 segmentation hypothesis 24 self concept 303, 306, 308, 310 self evaluations 334 self ratings 343, 350 semantic knowledge 201–203, 208 semantics 167, 168 socioeconomic status (SES) 8, 250, 252, 254, 255, 258 sociolinguistic competence 5 sound structure 325–334 specific language impairment (SLI) 109 speech manipulation 111 speech modification algorithm 110 spelling 121, 122, 217–224, 236, 291, 292, 322, 323, 327, 341, 353 storybook reading 249 structure building framework 200 structured word list 51 subsyllabic units 139, 157, 159–161 syllabic awareness 98, 103 syllable parsing 145 syntactic awareness 89, 295 synthesized speech 116 task independent deficit 232 task independent variables 238, 239 task related deficit 231 task related variables 238, 239 task specific deficit 230 task specific variables 238, 239 token frequency 58 transfer 298, 317, 318 type frequency 58 value of reading 306, 309, 310 verbal learning task 76, 77, 79, 84 verbal memory span 231, 236 visual segmentation 52, 56 vocabulary knowledge 94 vocabulary test 39 voluntary reading course 339–341, 350, 351, 354 vowels 294
(JB)SWL II_Index_357-359
9/24/02
2:58 PM
Page 359
Subject index word frequency 194–196, 294 word recognition 217–224, 235, 236, 327–334
working memory 73, 296 writing process 215
In the STUDIES IN WRITTEN LANGUAGE AND LITERACY the following titles have been published thus far: 1. VERHOEVEN, Ludo (ed.): Functional Literacy: Theoretical issues and educational implications. 1995 2. KAPITZKE, Cushla: Literacy and Religion: The textual politics and practice of Seventh-day Adventism. 1995. 3. TAYLOR, Insup, and M. Martin Taylor: Writing and literacy in Chinese, Korean, and Japanese. 1995. 4. PRINSLOO, Mastin and Mignonne BREIER (eds): The Social Uses of Literacy. Theory and Practice in Contemporary South Africa. 1996. 5. IVANIC, Roz: Writing and Identity. The discoursal construction of identity in academic writing. 1998. 6. PONTECORVO, Clotilde (ed.): Writing Development. An interdisciplinary view. 1997. 7. AIKMAN, Sheila: Intercultural Education and Literacy. An ethnographic study of indigenous knowledge and learning in the Peruvian Amazon. 1999. 8. JONES, Carys, Joan TURNER and Brian STREET (eds.): Students Writing in the University. Cultural and epistemological issues. 1999. 9. BARTON, David and Nigel HALL (eds.): Letter Writing as a Social Practice. 2000. 10. MARTIN-JONES, Marilyn and Kathryn JONES (eds.): Multilingual Literacies. Reading and writing different worlds. 2000. 11. VERHOEVEN, Ludo, Carsten ELBRO and Pieter REITSMA (eds.): Precursors of Functional Literacy. 2002.
