Advances in Child Development and Behavior Volume 24

ADVANCES IN CHILD DEVELOPMENT AND BEHAVIOR Volume 24 Contributors to This Volume Jacquelyn Baker-Sennett Linda B. S...

Author: Hayne W. Reese

10 downloads 1518 Views 17MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form

DOWNLOAD PDF

ADVANCES IN CHILD DEVELOPMENT AND BEHAVIOR

Volume 24

Contributors to This Volume Jacquelyn Baker-Sennett

Linda B. Smith

Charles Hulme

Margaret J. Snowling

Eugene Matusov

Keith E. Stanovich

John P. Rack

Laurel J. Trainor

Barbara Rogoff

Sandra E. Trehub

Daniel S. P. Schubert

Anna M . Unyk

Herman J. P. Schubert

Mazie Earle Wagner

Linda S. Siege1

John Worobey

ADVANCES IN CHILD DEVELOPMENT AND BEHAVIOR

edited by

Hayne W. Reese Department of Psychology West Virginia University Morgantown, West Virginia

Volume 24

ACADEMIC PRESS, INC. Harcourt Brace Jovanovich, Publlshers

San Diego New York Boston London Sydney Tokyo Toronto

This book is printed on acid-free paper. @

Copyright 0 1993 by ACADEMIC PRESS, INC. All Rights Reserved. No part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording, or any information storage and retrieval system, without permission in writing from the publisher.

Academic Press, Inc. 1250 Sixth Avenue, San Diego, California 92101 -431 1 Uniied Kingdom Edition published by

Academic Press Limited 24-28 Oval Road, London NW 1 7DX

Library of Congress Catalog Number 63-23237 International Standard Book Number: 0- 12-009724-9 PRINTED IN T H E UNITED STATES OF AMERICA 93 94 95

96 97 98 BB 9 8 7 6 5

4

3 2

I

Contents

Contributors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

ix

Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

xi

Music and Speech Processing in the First Year of Life I. I1 . I11. IV. V. VI .

SANDRA E . TREHUB. LAUREL J . TRAINOR. AND ANNA M . UNYK Why Study Music and Speech Processing in Infancy? ..................... Speech to Infants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Song to Infants . . ............................................ Infants' Processing ..... ................ Infants' Processing of Musical Sequences ............................... Music and Speech: Present Perspectives and Future Prospects . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..

2 4 9 15 18 25 29

Effects of Feeding Method on lnfant Temperament JOHN WOROBEY 1. Introduction

.....................................

I1. Temperament in Early Infancy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .............. I11. Diet and Early Behavior . . . . . . . . . . . . . 1V. The Sample Case of Feeding Regimen . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .............. V. Feeding Method and Activity . . . . . . .............. VI . Discussion .......................................... VII . Recommendations for Further Research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. ..............

31 39 42 43 48

53 55 56

The Development of Reading LINDA S . SIEGEL 1. Introduction ....................................................... I1. Controversies and Methodological Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111. Basic Cognitive Processes in Reading .................................. IV. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

V

63 64 68 92 92

vi

Contents

Learning to Read: A Theoretical Synthesis JOHN P. RACK, CHARLES HULME, AND MARGARET J. SNOWLING

.........

100

Methodological Issues ............................................... Theories of Reading Development ..................................... Descriptive Stage Models of Reading Development ............ Constraints on Reading Development: Theories about External Causes . . . . . . . Cognitive Processing Theories of Reading Development Ability ...............................

100

I. Introduction ......................... 11. 111. IV. V. VI.

....................................... References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

...

103 I05 108 117 124 127 128

Does Reading Make You Smarter? Literacy and the Development of Verbal Intelligence KEITH E. STANOVICH 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11. The Rise of the “Great Divide” Theories . .

Ill. Great Divide Theories under Attack ................................... IV. The Death Blow to Great Divide Theories: Scribner and Cole .............. V. Premature Closure on the Consequences of Literacy? .................. VI. Studying the Direct Consequences of Literacy within a Literate Society ...... VII. The Specific Cognitive Correlates of Print Exposure . ............ VIII. Summary and Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . .........................

134 134 137 138 140

141 153 169 I74

Sex-of-Sibling Effects: Part 1. Gender Role, Intelligence, Achievement, and Creativity I. 11. 111.

IV.

MAZIE EARLE WAGNER, HERMAN 1. P. SCHUBERT, AND DANIEL S. P. SCHUBERT Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..... ............. Gender Role . . . Intelligence, Cre ent . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusions ....................................................... References

182 186 194 208 209

The Concept of Same LINDA B. SMITH

I. The Problem of Similarity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11. Kinds of Sameness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111. The Development of a Concept of Same ................................

216 223 228

Contents

IV. A Connectionist Model of the Development of Similarity . . . . . . . . . . . . . . . . . . V. Psychological Facts and Philosophical Problems . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

vii 235

248 249

Planning as Developmental Process I. 11. 111. IV. V. VI. v11. VIII.

JACQUELYN BAKER-SENNETT, EUGENE MATUSOV, AND BARBARA ROGOFF 254 Introduction ....................................................... Planning as Process ................................................ 255 Accounts of the Development of Children's Planning Skill . . . . . . . . . 256 Developmental Activity Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 258 Planning: Deliberate Efforts to Reach Goals . ........ 260 Flexibility in Planning: Synthesis of Advance Planning and Improvisation , . , . 265 The Social and Cultural Nature of Planning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27 1 Conclusions . . . . . . . . . . . . . . ................................... 276 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 277

Author Index ............................................................

283

Subject Index ............................................................

297

Contents of Previous Volumes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

311

This Page Intentionally Left Blank

Contributors Numbers in parentheses indicate the pages on which the authors' contributions begin.

JACQUELYN BAKER-SENNETT'

Department of Psychology, University of Utah, Salt Lake City, Utah 84112 (253) CHARLES HULME

Department of Psychology, University of York, Heslington, York YO1 5DD, U . K. (99) EUGENE MATUSOV2

Department of Psychology, University of Utah, Salt Lake City, Utah 84112 (253) JOHN P. RACK Department of Psychology, University of York, Heslington, York YO1 5DD, U.K . (99) BARBARA ROGOFF2

Department of Psychology, University of Utah, Salt Lake City, Utah 84112 (253) DANIEL S. P. SCHUBERT

School of Medicine and MetroHealth Medical Center, Case Western Reserve University, Cleveland, Ohio 44109 (181) HERMAN J. P. SCHUBERT3

State University College at Bufsalo, Bufalo, New York 14222 (181) LINDA S. SIEGEL

Department of Instruction and Special Education, The Ontario Institute for Studies in Education, Toronto, Ontario, Canada M5S IV6 (63) LINDA B. SMITH

Department of Psychology and Program for Cognitive Science, Indiana University, Bloomington. Indiana 47405 (215) MARGARET J. SNOWLING

Department of Psychology, University of Newcastle-upon-Tyne, Newcastle-upon-Tyne NE1 7RU, U . K . (99) 'Present address: Department of Educational Psychology and Special Education, University of British Columbia, Vancouver, British Columbia, Canada V6T 174. *Present address: Psychology Board, University of California, Santa Cruz, California 95064. 3Deceased. ix

X

Contributors

KEITH E. STANOVICH

Ontario Institute for Studies in Education, Toronto, Ontario, Canada M5S 1V6 (133) LAUREL J. TRAINOR

Department of Psychology, McMaster University, Hamilton, Ontario, Canada L8S 4K1 ( 1 ) SANDRA E. TREHUB

Centre for Research in Human Development, Erindale College, University of Toronto, Mississauga, Ontario, Canada WL lC6 ( 1 ) ANNA M. UNYK

Centre for Research in Human Development, Erindale College, University of Toronto, Mississauga, Ontario, Canada L5L IC6 ( 1 ) MAZIE EARLE WAGNER

State University College at Bugalo, Buffalo, New York 14222 (181) JOHN WOROBEY

Department of Nutritional Sciences, Cook College, Rutgers-The University, New Brunswick, New Jersey 08903 (37)

State

Preface The amount of research and theoretical discussion in the field of child development and behavior is so vast that researchers, instructors, and students are confronted with a formidable task in keeping abreast of new developments within their areas of specialization through the use of primary sources, as well as being knowledgeable in areas peripheral to their primary focus of interest. Moreover, journal space is often simply too limited to permit publication of more speculative kinds of analyses that might spark expanded interest in a problem area or stimulate new modes of attack on a problem. The serial publication Advances in Child Development and Behavior is intended to ease the burden by providing scholarly technical articles serving as reference material and by providing a place for publication of scholarly speculation. In these documented critical reviews, recent advances in the field are summarized and integrated, complexities are exposed, and fresh viewpoints are offered. They should be useful not only to the expert in the area but also to the general reader. No attempt is made to organize each volume around a particular theme or topic, nor is the series intended to reflect the development of new fads. Manuscripts are solicited from investigators conducting programmatic work on problems of current and significant interest. The editor often encourages the preparation of critical syntheses dealing intensively with topics of relatively narrow scope but of considerable potential interest to the scientific community. Contributors are encouraged to criticize, integrate, and stimulate, but always within a framework of high scholarship. Although appearance in the volumes is ordinarily by invitation, unsolicited manuscripts will be accepted for review. All papers-whether invited or submitted-receive careful editorial scrutiny. Invited papers are automatically accepted for publication in principle, but may require revision before final acceptance. Submitted papers receive the same treatment except that they are not automatically accepted for publication even in principle and may be rejected. The Advances series is usually not a suitable place of publication for reports of a single study, or a short series of studies, even if the report is necessarily long because of the nature of the research. The use of sexist language, such as “he” or “she” as the general singular pronoun, is not acceptable in contributions to the Advances series; and the use of “their” as a singular pronoun is incorrect. The use of “he or she” (or the like) is acceptable.

xi

xii

Preface

I wish to acknowledge with gratitude the aid of my home institution, West Virginia University, which generously provided time and facilities for the preparation of this volume. I also wish to thank Drs. Ren6e Baillargeon, Linnea C. Ehri, Lynn S. Liben, and Richard K. Olson for their editorial assistance, and Mrs. Ann Davis for her excellent secretarial services.

Hayne W. Reese

MUSIC AND SPEECH PROCESSING IN THE FIRST YEAR OF LIFE

Sandra E . Trehub CENTRE FOR RESEARCH IN HUMAN DEVELOPMENT ERINDALE COLLEGE UNIVERSITY OF TORONTO MISSISSAUGA, ONTARIO, CANADA L5L IC6

Laurel J . Trainor DEPARTMENT OF PSYCHOLOGY MCMASTER UNIVERSITY HAMILTON, ONTARIO, CANADA L8S 4K1

Anna M . Unyk CENTRE FOR RESEARCH IN HUMAN DEVELOPMENT ERINDALE COLLEGE UNIVERSITY OF TORONTO MISSISSAUGA, ONTARIO. CANADA L5L IC6

I. WHY STUDY MUSIC AND SPEECH PROCESSING IN INFANCY? A. SPEECH AND EMOTION B. MUSIC AND EMOTION C. FOCUS OF THE ESSAY 11. SPEECH TO INFANTS A. FEATURES OF INFANT-DIRECTED SPEECH B. SITUATIONAL, INDIVIDUAL, AND CROSS-CULTURAL DIFFERENCES C. THE LISTENER’S CONTRIBUTION D. INFANT-DIRECTED SPEECH AS COMMUNICATION 111. SONG TO INFANTS A. LULLABIES: DESCRIPTIVE RESEARCH ACROSS CULTURES B. LULLABY IDENTIFICATION ACROSS CULTURES C. MATERNAL SINGING: THE INFANT’S CONTRIBUTION D. PATERNAL SINGING IV. INFANTS’ PROCESSING OF SPEECH SEQUENCES A. AITENTIONAL PREFERENCES B . AFFECTIVE PREFERENCES C. TEMPORAL PATTERNING AND ATI’ENTIONAL PREFERENCES 1 ADVANCES IN CHILD DEVELOPMENT AND BEHAVIOR. VOL. 24

Copyright 0 1993 by Academic Press. Inc. All rights of repduction in any form reserved

2

Sandra E . Trehub e? al.

V. INFANTS’ PROCESSING OF MUSICAL SEQUENCES A. ATTENTIONAL PREFERENCES B . METHODOLOGICAL STRATEGY C. PITCH PATTERNING D. TEMPORAL PATTERNING E. MELODY AND CONTOUR F. “GOOD” AND “BAD” MELODIES G. GENERAL PERSPECTIVES VI. MUSIC AND SPEECH: PRESENT PERSPECTIVES AND FUTURE PROSPECTS A. CAREGIVERS’ SPEECH AND SONG B. INFANTS’ SPEECH AND SONG PROCESSING REFERENCES

I. Why Study Music and Speech Processing in Infancy? Little justification is required for the study of speech processing in infancy. After all, mothers and other caregivers generally talk to noncomprehending infants, and ultimately, these infants become children with adult-like speech reception and production skills. In this light, what could be more reasonable than to explore very early abilities, processing predispositions, and the course of early learning in this domain? The case for the study of music processing in infancy seems much less compelling, largely because of our ignorance about the prevalence and functions of music. Music, like language, has been found in every culture and historical period. Moreover, music has typically occupied a central role in work and play, involving all members of society (Bebey, 1969). Indeed, the Western inclination to relegate music making to talented performers and isolated artistic contexts (e.g., concerts) is unusual when viewed in historical and cross-cultural perspective (Walker, 1990). Not only is participation in music widespread, but so is belief in its power over physical and mental states (Walker, 1990). In this light, posing questions about early processing skills, predispositions, and learning in relation to music seems to be as reasonable as posing such questions in relation to speech. Speech and music processing may be linked in infancy in ways that go beyond the obvious sharing of reception and production channels (i.e., ears and voice). For prelinguistic infants, speech is linguistic only from the perspective of the talker or mature observer. From the very young infant’s perspective, speech consists of complex patterns of semantically meaningless sounds that vary over time. To the extent that such patterns are meaningful in the earliest months of

Music and Speech Processing in Infancy

3

life, these meanings are necessarily affective or emotional (Fernald, 1989, 1992; Lewis, 1951). A. SPEECH AND EMOTION

Vocal but nonverbal expressions of emotion may be universal (for a review see Frick, 1985). Prosodic contours in speech (i.e., patterns of pitch and loudness over time) are thought to reflect the talker’s emotional state (Knower, 1941; Williams & Stevens, 1972). For example, falling contours are correlated with pleasantness and rising contours with surprise or fear (Scherer, 1974). Spectral structure (i.e., amount of energy at different frequencies), which confers a distinctive sound quality, tone of voice, or timbre, is also important for emotion recognition (Lieberman & Michaels, 1962). For example, the emotions of pleasantness and happiness are associated with relatively few component frequencies (Scherer & Oshinsky, 1977). A speakmg voice with few component frequencies sounds “pure,” more like a flute than a piano, for example. Also, facial expressions such as smiling affect the shape of the vocal tract and the resultant voice quality (Laver, 1980). Positive emotions tend to shift energy to higher component frequencies relative to negative emotions (Frick, 1985). Moreover, as a talker’s pitch level is raised relative to his or her usual level, vowel spectral noise .may decrease, leading to a corresponding decrease in listeners’ perception of vowel roughness (Emanuel & Smith, 1974; Newman & Emanuel, 1991). Presumably, this higher-pitched voice would have a more melodious quality. Pitch level also contributes to perceived emotion. High pitch signals happiness and friendliness for human interactants and nonaggression or submission for human and nonhuman interactants (Frick, 1985; Stross, 1977). Prosody, then, may be the principal camer of emotion in speech, one that is not primarily dependent on symbolic understanding or cultural conventions (Frick, 1985). As a result, listeners can identify the emotional tone of contentfiltered speech (Scherer, Koivumaki, & Rosenthal, 1972) and can derive similar interpretations of nonverbal emotional content across cultures ( b u s s , Curran, & Ferleger, 1983). B . MUSIC AND EMOTION

The view of music as the universal language of the emotions (Langer, 1957) is an oversimplification because music also engages culture-specific cognitive structures. Nevertheless, music can express or represent emotions (Langer, 1957; Meyer, 1956; Trainor & Trehub, 1992b) and can also alter the emotional state of the mature listener (Francts, 1958; Meyer, 1956). Emotion, then, could be regarded as a quality that permeates music and speech. From the perspective of “premusical” or musically unacculturated infants, music, like speech, consists

4

Sandra E . Trehub et al.

of complex sound patterns that vary over time. For infants, moreover, musical patterns must bear considerable similarity to the prosody of speech, notably its intonation, rhythm, and stress (Trehub, 1990). For prelinguistic and premusical listeners, then, patterns of speech and music might well engage common processing strategies. C. FOCUS OF THE ESSAY

In the present article, we focus on potential similarities between speech and music from the perspective of infant listeners. The stimuli of concern are sound sequences rather than single sounds, despite the predominant research focus on the latter class of stimuli. The exclusion of single sounds can be justified on a number of grounds. First, the literature contains several comprehensive reviews of infants’ ability to perceive single speech and nonspeech sounds (e.g., Aslin, Pisoni, & Jusczyk, 1983; Jusczyk, 1992; Kuhl, 1988; Schneider & Trehub, 1992; Werker, 1991). Second, and perhaps more important, evidence indicates that global patterns of speech are more salient in the prelinguistic period than are individual speech segments (i.e., consonants and vowels) (Crystal, 1973; Lewis, 1951). In the nonspeech domain, evidence also indicates that infants proceed from global processing of auditory patterns to local processing of pattern details (Morrongiello, 1988; Trehub, 1985, 1990; Trehub & Trainor, 1990). In drawing parallels between speech and music, we focus on two principal issues: the input provided by caregivers for their infants and the processing of such input by infant listeners. Much of the work to be reported, particularly in the musical domain, is relatively recent. As a result, the exposition is tentative rather than definitive, its purpose being to suggest new avenues for future research and thinking.

11. Speech to Infants Infant-directed speech, often termed motherese or baby talk, tends to be elicited in caregivers by the mere presence of an awake infant (Fernald, 1984; Fernald & Simon, 1984; Rheingold & Adams, 1980; Rosenthal, 1982). Although the principal focus has been on its syntactic structure and suitability as a medium for language teaching (Murray, Johnson, & Peters, 1990; Newport, Gleitman, & Gleitman, 1977; Sherrod, Friedman, Crawley, Drake, & Devieux, 1977; Snow, 1977), researchers have shown growing interest in its prosodic form (e.g., pitch, intonation, stress, rhythm) and suitablilty for attention and affect regulation (Fernald, 1984, 1985, 1989; M. PapouSek, PapouSek, & Symmes, 1991; Stem, Spieker, Barnett, & MacKain, 1983). These suprasegmental or prosodic aspects


5

are the variables that have the greatest potential relevance to prelinguistic listeners. A. FEATURES OF INFANT-DIRECTED SPEECH

The differences between infant-directed and adult-directed speech are considered to exceed those between different dialects of a language. The most notable features of this special speech register for infants are its increased pitch, greater pitch range, elongated vowels, simpler pitch contours, rhythmic regularity, slower tempo, briefer utterances, and overall repetitiveness compared with standard adult speech (Beebe, Feldstein, Jaffe, Mays, & Alson, 1985; Ferguson, 1964; Fernald & Simon, 1984; M. PapouSek, PapouSek, & Bornstein, 1985; Sachs, 1977; Stem et al., 1983; Stem, Spieker, & MacKain, 1982).The average pitch of the caregiver’s voice rises by about three or four semitones (25-33%) in interactions with infants (Fernald & Simon, 1984; Jacobson, Boersma, Fields, & Olson, 1983; PapouSek et al., 1985), generating a falsetto quality. The most celebrated aspect of infant-directed speech is the unique set of contours that apparently characterize such communications. In contrast to the narrow pitch range and multiple directional changes of pitch movement that characterize adult-directed speech (Bolinger, 1970), very simple contours with an expanded pitch range are typical of infant-directed speech (Femald & Simon, 1984). These contours are usually unidirectional (rising or falling in pitch) but sometimes bidirectional (rise-fall or fall-rise) and they are used over and over with different component words, phrases, and sounds (H. PapouSek & PapouSek, 1984). Poor articulation is often evident and utterances may consist solely of elongated vowels or consonant-vowel syllables stretched out over one of these expanded contours (Femald & Simon, 1984; M. PapouSek & Papousek, 1981). Some idealized examples are provided in Fig. 1.

i r I w - 0 - 0 - 0 - WI

Y - E - E - E - S ?

A

. -

-

-

A

-

A - A -H. .~

!

Fig. I . Schematic illustrations of hypothetical infant-directed uiierances. Wow is greatly extended over a rise-fall contour, yes over a rising contour, and ah over a falling contour. These contours likely describe correlated changes in loudness and pitch.

6


B. SITUATIONAL, INDIVIDUAL, AND CROSS-CULTURAL DIFFERENCES

These pitch contours seem to be roughly tied to specific situational contexts. For example, rising contours tend to be used to capture an infant’s attention and to elicit some response (Ferrier, 1985; M. Papoukk et al., 1991). Bell-shaped contours are sometimes used for attention capture as well, but they are used more frequently for maintaining such attention and expressing approval (Fernald, 1989; Stem et al., 1982). By contrast, sustained or falling contours are used for soothing infants and promoting sleep (M.PapouSek & PapouBek, 1981; M. PapouSek et al., 1991). Rhythm and tempo are also tied to particular contexts, with increasing tempo for inattentive infants, decreasing tempo for infants progressing toward sleep, and varying rhythm and tempo for fussy infants (M. Papougek & PapouSek, 1981). If speech adjustments of this nature were limited to North American mothers, they would be indicative of cultural variations in caregiving, but they would be of less general interest; however, similar modifications can be seen in a wide range of structurally different languages including German, Arabic, Mandarin Chinese, Japanese, Spanish, Kwar’ae, and British English (Ferguson, 1964; Fernald et al., 1989; Grieser & Kuhl, 1988; Papougek et al., 1991; Watson-Gegeo & Gegeo, 1986). Moreover, fathers, primiparous mothers, male and female nonparents, and even preschool children with and without younger siblings adjust their speech in relation to infant listeners (Anderson, 1986; Dunn & Kendrick, 1982; Femald & Simon, 1984; Jacobson et al., 1983; Rheingold & Adam, 1980; Sachs & Devin, 1976; Watson-Gegeo & Gegeo, 1986; Weeks, 1971), although the nature and extent of such adjustments vary somewhat. For example, fathers alter their vocal behavior less than do mothers, omitting the characteristic increase in pitch range (Fernald et al., 1989). On the whole, however, speech adjustments in relation to infants appear to be independent of language, culture, and caregiving experience. This conclusion is somewhat of an overstatement, however. Notable crosscultural variations have been found and they may reflect differences in expressiveness, on the one hand, and caregiving attitudes, on the other. For example, Fernald et al. (1989) described more extreme intonational modifications in American English compared with British English, French, Italian, German, and Japanese speech to infants. Other comparisons of American English with British English (Shute & Wheldall, 1989) and Mandarin Chinese (Grieser & Kuhl, 1988; M. PapouSek et al., 1991) are in line with this finding. Femald (1992) suggested that such differences reflect variations in cultural display rules, with Asian mothers in particular considering vocal and facial expressiveness less socially acceptable than do European and American mothers. Other research indicates quite the


I

opposite, that Japanese mothers may be more affect oriented and American mothers more information oriented (Toda, Fogel, & Kawai, 1990). Indeed, Japanese mothers use repetitive nonsense syllables, onomatopoeia, invented lexical items, and imitations of their infant’s sounds more than do American mothers (Morikawa, Shand, & Kosawa, 1988; Toda et al., 1990), which lends credence to the latter view. Expanded intonation contours may not be the only effective means of recruiting infants’ attention despite their wide usage by North American mothers. Moreover, Japanese mothers seem to spend more time soothing their infants (by vocalization and touch) and less time arousing them compared with American mothers, which may reflect different cultural values and caretaking goals (Caudill & Weinstein, 1969; Toda et al., 1990). More soothing than arousing is also found in the Gusii tribe in the highlands of Kenya (Dixon, Tronick, Keefer, & Brazelton, 1981). Similarly, the predominance of falling over rising contours in Mandarin Chinese (Grieser & Kuhl, 1988) is consistent with the notion of soothing having priority over arousing in the care of infants. Because most research on infant-directed speech has been focused on highly industrialized societies, North America and Western Europe in particular, the view of typical infant-directed speech as arousing and attention getting may be unwarranted. A question that remains is whether arousing or soothing vocalizations predominate in more broadly based samples of caregivers, particularly those in “traditional” societies, in which mothers and infants spend most of their waking and sleeping hours together. C. THE LISTENER’S CONTRIBUTION

The prevailing wisdom is that speech modifications to infants are effected more or less intuitively (M. PapouSek et al., 1985). The point is not that adults totally lack awareness of such behavior but rather that at least some aspects of this behavior are not manipulated consciously. As a result, situations of simulated as opposed to actual interactions with infants tend to induce less than the full set of infant- or child-directed adjustments (Fernald & Simon, 1984; Jacobson et al., 1983; Murray & Trevarthan, 1986). Why might this difference arise? The behavior of mothers (and other caregivers) may be driven, at least in part, by feedback from the infant. On the one hand, fluctuations in the infant’s attention and arousal provide cues to the fine-tuning of maternal behavior. On the other hand, the infant’s presence and behavior may trigger alterations in the mother’s own state that are reflected in her behavior. It is likely that both factors are at work, although the latter may generate the most noticeable differences in vocal quality. In this regard, one can think of the speech of lovers, which often exhibits infant- or child-directed qualities such as heightened pitch, the use of dimin-

8

Sandra E. Trehub et al.

utives, and nonsense sounds. Intonational stereotypy can also be seen in some adult-directed utterances such as greetings (e.g.. Hi stretched out over a bellshaped contour) and exclamations (Wow!). Infant-directed speech is tailored considerably to the age and ability of the listener. For example, the full complement of infant-directed modifications seems to reach its height when the listener is about 4 months of age, becoming somewhat attenuated thereafter (Stem et al., 1983). Many adjustments, however, remain in evidence well beyond infancy (Ferguson, 1964; Garnica, 1977; Snow, 1977), with different parameters undergoing selective enhancement or attenuation. For example, variations in articulatory distinctiveness (Malsheen, I980), emphatic stress (Fernald & Mazzie, 1991), and the use of diminutives (JociC, 1978) are linked to the child’s language reception and production skills. D. INFANT-DIRECTED SPEECH AS COMMUNICATION

In line with the notion of maternal prosody as emotionally expressive (Lewis, 195 l), Fernald (1989) explored adult listeners’ ability to discern different intentions from such utterances. She tape-recorded mothers as they interacted with their 10- to 14-month-old infants in play and caretaking contexts. Five broad communicativecategories or utterance types were selected for study: ( 1) approving, in response to the infant’s completion of a simple task; (2) prohibiting, designed to prevent the infant from touching an object about to be touched; (3) comforting an upset infant; (4) attention bidding, which involved directing the infant’s attention to an object or action; and ( 5 ) game initiating, which involved a hiding game such as peek-a-boo. She also tape-recorded the same women as they interacted in somewhat comparable contexts (e.g., soothing, warning) with their husbands. These utterances were content filtered (i.e., frequencies above 400 Hz removed) so that they became semantically unintelligible while still retaining their overall prosodic form. As Fernald (1989) noted, they sounded like “speech heard through a wall” (p. 1503). Adult listeners were required to assign each utterance to one of the five aforementioned communicative categories and seemed to do so with relative ease. Performance, which was well above chance for each category, was unrelated to listeners’ parental status, age, and experience with children. Although adults could also discern similar communicative intent from nonmaternal (adultdirected) prosody, they were considerably less accurate for all categories except comforting vocalizations. From the perspective of adult listeners, then, infantdirected prosody is for the most part more informative than adult-directed prosody. The greater transparency of infant-directed prosody may be linked to its greater simplicity or to its decoupling from conventional communicative symbols.


9

111. Song to Infants Given the rich descriptions of maternal vocal behavior and the characterization of much of this behavior as musical or melodic (Fernald, 1989, 1992; M. PapouSek & PapouSek, 1981), we were surprised that interest in this domain had not extended to maternal song. The practice of singing to infants appears to be widespread, perhaps universal. In fact, one ubiquitous musical genre, the lullaby, is regularly sung to soothe infants and induce sleep (Brakeley, 1950; CassBeggs & Cass-Beggs, 1969). Detailed descriptive material on lullabies is sparse but the available research provides suggestive evidence of considerable similarity between infant-directed speech and song. A. LULLABIES: DESCRIPTIVE RESEARCH ACROSS CULTURES

McCosker (1974) has provided one of the more extensive descriptions of lullaby form and function in his work on the Cuna Indians of Panama. The singer of Cuna lullabies enjoys greater freedom for improvisation of text and melody compared with other songs. Unlike other songs, lullabies are of indefinite length, continuing until their function has been fulfilled. Cuna lullaby lyrics reveal considerable word reduplication, sequence repetition, and common words incorporated into repetitive rhythmic patterns, as illustrated (below) by a lullaby from Nalunega Island (McCosker, 1974, pp. 126-127). Translations by a native Cuna speaker are provided (see McCosker, 1974, p. 146). In some cases, the translator had difficulty with words, and rough glosses are provided in parentheses. (1)

a. pani kala pani poa nai tai ye little ones come here to the hammock b. poe pii poe pii pani tai muloye you are crying come here c. nana peka u kachi p a kine mama is always here in the hammock d. nana peka nai kucha pani nukku p a kine pani poa tii p a ye urn mama sees you crying and will take you in her arms e . poe piipii na piipii pak crying is the baby f. naa p e a tummuwali malo mama raises you in her arms g. pani nukku p a ki pani (refers to something that mama is doing while sitting) h. nokku we ye (refers to something that mama is doing while sitting)

10


i. nana peka mama is sitting j. u kachi p a kina she is sitting in the hammock k. nam peka nai kucha y e mama is sitting while the little ones come 1. pani pii nana pani poe tii kpaa urn you (little girl) were crying m. puna tola piipii maloye my little girls n. machi tola o kanapi you will grow up and marry a grown-up boy 0. nu p e o tumnotali maloye you will grow up Repetition of the word sequence nana peka u kachi p a kine is evident in (lc), (li), and (lj). The very common words poe pi, poe pii, poa tii, and poe piipii (relating to crying) appear in lines (lb), (Id), (le), and (11). Nana (mama), a common initial word pattern, occurs in lines (lc), (Id), (li), and (lk). Finally, maloye and ma10 are typically used to mark the end of a musical phrase or section in lines (lb), (If), (lm), and (lo). Unfortunately, the translated text cannot capture the rhythm of the original because each Cuna syllable corresponds to a sung note. Sakata (1987), in her work on the Hazara tribe in central Afghanistan, noted that Hazara women, who are excluded from musical activities in their maledominated culture, distort or invent words in their lullabies, with the apparent goal of producing mellifluous sounds. This orientation toward “sound effects” likely accounts for the cross-cultural prevalence in lullabies of stereotyped syllables such as loo-Zoo, lulla, nina, bo-bo, and do-do (Brakeley, 1950; Brown, 1980). Liberal use of elongated vowels is also reported in lullabies sung by Mohave (Devereux, 1948), Arapaho (Hilger, 1952), Chippewa (Hilger, 195 l), and Hopi (Sands & Sekaquaptewa, 1978) caregivers. Moreover, lullabies seem to embody a number of properties associated with infant-directed speech in general and soothing infant-directed speech in particular. As noted earlier, soothing speech to infants includes low, falling contours. a narrow pitch range, and a gentle tone of voice (Fernald & Simon, 1984; M. PapouSek & PapouSek, 1981; M. PapouSek et al., 1991). In fact, the features of narrow pitch range, smooth repeating contours, and repetitive rhythms have been documented for Vietnamese (Cong-Huyen-Ton-Nu, 1979), North American Indian (Sands & Sekaquaptewa, 1978), Hazara (Sakata, 1987), and Cuna (McCosker, 1974) lullabies. Nevertheless, falling contours have not been mentioned in these descriptions. Like the early research on infant-directed speech, however,


I1

most work on infant-directed song has been focused on textual content (i.e., the verbal message) rather than melodic form. B. LULLABY IDENTIFICATION ACROSS CULTURES

Despite the functional distinctiveness of lullabies (i.e., their use in accelerating sleep), the question arises as to whether they are perceptually distinct to listeners unfamiliar with the culture. In the case of infant-directed speech, distinctiveness is considered a foregone conclusion, making experimental verification seem unnecessary. Trehub, Unyk, and Trainor (in press-a) evaluated this question with lullabies from different cultures. They paired 30 foreign lullabies with comparison songs (mostly adult songs) and asked adult listeners to identify the infant-directed song in each pair. The songs were selected from ethnomusicological materials recorded in diverse cultures and geographic regions, with none sung in a language familiar to the adult participants. Because the characteristically slow tempo of lullabies could provide an obvious cue, each comparison song was matched in tempo as well as language and cultural origin to the relevant lullaby. Adult listeners identified the lullabies significantly better than chance and their performance was independent of musical training (i.e., years of music lessons) and familiarity with the musical system (i.e., Western versus other). Perhaps mellifluous or stereotyped syllables (Brown, 1980; Sakata, 1987), onomatopoeia (Curtis, 1921), and word reduplication (McCosker, 1974) accounted for the listeners’ performance. To reduce or eliminate such influences, Trehub et al. (in press-a) content-filtered the songs by removing all frequencies above 500 Hz. The resulting recordings sounded muffled, with the words completely obscured but the melody line and many aspects of voice quality largely intact. Adult listeners were still able to identify the lullabies, indicating that information conveyed by melody or voice quality must provide distinctive cues. To eliminate voice quality and other performance cues, Trehub et al. created a synthesized (piano-timbre) version of the melody line of the lullabies and comparison songs. This manipulation also removed all vocal ornamental devices such as trills and glides, which eliminated much of the “foreign” sound of the materials. Performance on these reduced versions of the songs was much less accurate, indicating the contribution of vocal quality or tone of voice to lullaby identification. Nevertheless, performance was still significantly correlated with performance on the original materials. Taken together, these findings are indicative of the contribution of melodic as well as word cues to the identification of lullabies. In short, soothing infant-directed song is perceptually distinct, at least for adult listeners. The task was by no means easy, performance being at approximately 66%

12


correct on the original materials. The noteworthy finding, however, was the consistency of performance. For example, four of the lullabies (Creek Indian, Czech, Irish, Pygmy) were correctly identified by more than 85% of the listeners and four other lullabies (Chadian, Ecuadorian, Samoan, Ukranian) were consistently but incorrectly rejected. This distinct pattern of performance prevailed even in the filtered and synthesized versions. Perhaps adults had some conception of a lullaby, a pancultural stereotype or prototype, that they used to judge the songs they heard. Other work with these lullabies and comparison songs (Unyk, Trehub, Trainor, & Schellenberg, 1992) has revealed that adult listeners judge lullabies to be significantly simpler than comparison songs whether these are presented in original, filtered, or synthesized versions. Moreover, the lullabies identified with greatest accuracy (in Trehub et al., in press-a) are rated as simpler than those identified with least accuracy. Although simplicity in general distinguishes infant-directed speech and song from their adult-directed counterparts, little is known about the component features that contribute to the perceived simplicity of songs and their appropriateness for infants. For example, simple songs such as lullabies might have fewer changes in pitch direction (i.e., contour changes) per unit time. Moreover, soothing songs, like soothing maternal speech, might have a preponderance of descending contours. In a musicological analysis of transcribed (i.e., written) versions of the lullabies and comparison songs, Unyk et al. (1992) evaluated a number of possible differentiating features such as median pitch, pitch range, phrase length, contour complexity, and descending contours. Surprisingly, none of these features, either singly or in combination, distinguishedthe lullabies from the other songs. Nevertheless, some features were reliably associated with adults’ judgments of lullabies. In particular, songs with a greater proportion of descending contours were more likely to be judged as lullabies, as were those with fewer contour changes. These features may be components of adults’ prototype or mental representation of lullabies (Trehub & Unyk, 1992). What factors may have prevented listeners from achieving greater accuracy in lullaby identification? Perhaps the choice of materials obscured important structural as well as stylistic (i.e., performance) distinctions. The tempo of lullabies and comparison songs, no doubt a critical lullaby cue, had been deliberately equated, resulting in a likely underestimation of the cross-cultural recognizability of lullabies. In addition, Trehub et al. used recorded materials from other investigators (ethnomusicologists)and therefore had no access to information about the context of the performances. As a result, the context may have been functionally appropriate in some cases (e.g., a mother lulling her infant to sleep) and inappropriate in others (e.g., an individual simply responding to a researcher’s request, perhaps in the absence of an infant). The essence of a lullaby may derive,


13

at least in part. from fine-tuning the performing style to the function and context. The English verb lull attests to the importance of function by referring to a particular manner of soothing by sounds or caresses. In this regard, Hilger (1952) described the Arapaho tribe’s occasional use of traditional dance songs as “lullabies.” Sakata (1987) used the termfuncrional lullabies to distinguish songs that women sing to their infants from the stylized lullabies that men sing (often with instrumental accompaniment) about their infants. Sands and Sekaquaptewa (1978), noting the contribution of infant characteristics to the singer’s performance, distinguished between soothing lullabies for cooperative infants and admonishing lullabies for reluctant sleepers. Just as an infant’s presence may be necessary for the full set of appropriate speech adjustments (Fernald & Simon, 1984; M. PapouSek, PapouSek, & Haekel, 1987), so the infant’s presence and appropriate state may be critical for the appropriate realization of a lullaby. Whether this realization involves an appropriate tone of voice (e.g., soft, gentle) or an appropriate dynamic quality (e.g., lulling) remains to be determined. In any case, such questions must await comparisons between contextually appropriate and inappropriate lullabies. C. MATERNAL SINGING: THE INFANT’S CONTRIBUTION

Some research on maternal singing offers further perspectives on infantdirected song in general and on contextual factors in particular. Ti-ehub, Unyk, and Trainor (in press-b) recorded mothers of infants (12 months or less) as they informally sang a song of their choice. For half of the mothers, the infant was present, in which case they sang a song directly to the infant. For the other half, the mothers were instructed to sing as they normally would while alone. To minimize the singer’s self-consciousness, the experimenter left the room before the recording session began and the mother started and stopped the recording equipment. After completing her song, the mother recalled the experimenter, who requested that she sing the same song again but in the opposite condition (e.g., without the infant if the infant was originally present). Once again, the mother controlled the equipment. A composite recording with pairs of excerpts from the two conditions (from each of 15 mothers) was presented to adult listeners, who were required to identify the infant-directed excerpt in each pair. Listeners were highly accurate on this task, achieving scores of about 91% correct. Instrumental analyses revealed that the singing to infants was characterized by higher pitch and slower tempo. Moreover, musically trained listeners rated the infant-directed singing as having greater rhythmicity, a softer tone of voice, more elongated vowels, and a “smiling” quality. Smiling is known to alter vocal tract characteristics and the resulting performance of singers (Fonagy, 1981; Sundberg, 1973) as it does for talkers (Laver, 1980). Another group of mothers provided samples of simulated and actual infant-

14


directed singing, which led to about 77% correct identification of actual singing to infants. A few of the mothers were better “actors” than others, generating lively and convincing simulations. In any case, the infant’s presence seems to alter features of the singer’s performance in a way that enhances its identifiability. Does culture-specific knowledge facilitate the identification of performance features associated with contextual appropriateness (i .e., an infant’s presence)? To gain insight into this question, Trehub et al. (in press-b) recorded Indian mothers (in India) and North American mothers of Indian descent as they sang Hindi songs with their infant present or absent. North American listeners were less accurate in judging the context of these foreign songs (57% correct) than they were with North American singers (about 91% correct), but their performance significantly exceeded chance levels. Listeners of Indian origin performed better (7 1% correct) than native-born North Americans on these Hindi materials, but their performance was poorer than that of native-born North American listeners on English materials. The choice of songs may have contributed to listeners’ poorer performance on Hindi than on English songs. The English singers tended to choose play songs such as Twinkle, Twinkle Little Star or Fr2re Jacques. Such play songs were selected even when mothers began the session by singing on their own after being instructed to sing something that they usually sang. In contrast, the Hindi singers tended to select soothing or religious songs in both contexts so that the overall character of sung materials was very different. Perhaps soothing songs, with their smooth contours and slow tempo, offer less scope for performance variations than do playful songs. This difference may also characterize soothing versus playful speech, but no research has been done on this issue. The cross-cultural variations in infant-directed speech may have their counterpart in infant-directed song, with arousing songs predominating in typical North American and European contexts and soothing songs predominating in other cultural contexts. In India, for example, mothers tend to remain with their infants until they fall asleep, often singing and otherwise comforting them. Many of the Indian mothers in the Trehub et al. (in press-b) sample reported similar practices and some even selected such a context for their recording. In contrast, North American infants typically sleep in a separate room and their caregivers often withdraw before they fall asleep, In fact, native-born North American mothers in the Trehub et al. (in press-b) study reported that they sang lullabies infrequently, if at all. Caudill and Weinstein (1969) observed more lullaby singing by Japanese than by North American mothers, although more recent reports (Sengoku, cited in Morikawa et al., 1988) reveal some “modernization” of Japanese caregiving, including a reduction in holding and rocking and a corresponding increase in talking.


15

D. PATERNAL SINGING

Do fathers show less distinctiveness in their songs to infants, as they do in their infant-directed speech (Fernald et al., 1989)? Work in progress by Trehub, Unyk, and their associates supports an affirmative answer. Fathers were recorded while singing to their infants or while simulating such singing. Unlike mothers, few sang standard children’s songs. Instead, they sang popular or folk tunes, freely improvising the words or melodies, and often embedding the infant’s name in their songs. Nevertheless, listeners had more difficulty differentiating actual from simulated paternal songs than they did with maternal songs. Were the fathers such excellent simulators that they misled the listeners or were they simply less “attuned” to their infants? On the basis of independent ratings of mothers’ and fathers’ singing style, less attunement seems to have been the case.

IV. Infants’ Processing of Speech Sequences Infant-directed speech has a potent effect on its intended audience. Such speech induces heightened affective responsiveness (Werker & McLeod, 1989), smiling (Wolff, 1963), vocalization (Mayer & Tronick, 1985; Stevenson, VerHoeve, Roach, & Levitt, 1986), and vocal imitation (Lieberman, Ryalls, & Rabson, cited in Lieberman, 1984). Moreover, infants are rated as more likable when listening to infant-directed than to adult-directed speech (Werker & McLeod, 1989). This speech register also facilitates infants’ differentiation of their mother’s voice from that of a stranger (Mehler, Bertoncini, Barrikre, & Jassik-Gerschenfeld, 1978) and their differentiation of phonetic contrasts in sequences of syllables (Karzon, 1985). A. ATTENTIONAL PREFERENCES

Some researchers have systematically evaluated the relative attention-eliciting properties of natural or systematically altered infant- and adult-directed speech. In one variant of this procedure (Femald, 1985), infants are presented infantdirected speech when they look at one loudspeaker and adult-directed speech when they look at another. Longer elective looking for one type of speech over the other is considered to reflect its greater attention-eliciting properties or infants’ “preference” for that speech variety. Werker and McLeod (1989) added a visual component, creating audio-video displays of infant- and adult-directed speech. In another variant of the preference procedure (Cooper & A s h , 1990), the presentation of sound is made contingent on looking at a single loudspeaker, with the two types of speech presented on alternating trials. In yet another

16

Sandra E. Trehub er al.

procedure (Pegg, Werker, & McLeod, 1992), looking time to both types of speech is compared in the context of an infant-controlled habituation design. Despite procedural differences, the results have been consistent in revealing attentional preferences favoring infant-directed speech in newborns and 1-montholds (Cooper & Aslin, 1990), 1.5-month-olds(Pegg et al., 1992), 4- to 5-montholds (Fernald, 1985; Werker & McLeod, 1989), and 7- to 9-month-olds (Werker & McLeod, 1989). Moreover, infant-directed speech produced by fathers generates an attentional preference but does not noticeably heighten affect (Pegg et al., 1992; Werker & McLeod, 1989). Some progress has been made in isolating the features responsible for the perceptual salience of infant-directed speech. For example, 4-month-olds’ preference for infant-directed speech was found to persist with speech samples filtered to remove the lexical content but not the prosody (Fernald, 1985). A similar attentional bias prevailed for synthesized contours that kept the pitch and temporal patterning intact but not for those that maintained temporal or amplitude information divorced from pitch patterning (Fernald & Kuhl, 1987). The situation seems to be somewhat different for 1-month-olds. Researchers who have used content-filtered infant- and adult-directed speech or synthesized pitch contours have not found the differential attentiveness obtained with natural speech samples (Cooper, in press). Possibly, a number of features such as pitch, pitch contour, and temporal and amplitude patterning act jointly to increase the impact of infant-directed speech. The less mature the listener, the more features may be required to trigger differential responsiveness. With older infants, however, some features may be effective singly (e.g., high pitch) and others only in combination. Nevertheless, the presence of multiple features no doubt increases the magnitude of the response. B. AFFECTIVE PREFERENCES

In an effort to discern possible “meanings” in infant-directed speech for infant listeners, other researchers have examined differential affective responses to “positive” and “negative” infant-directed speech. In one such study (M. Papougek, Bornstein, Nuzzo, PapouBek, & Symmes, 1990), 4-month-old infants were presented simulated pitch contours characteristic of approving and disapproving (prohibitive) infant-directed speech. Each simulated utterance was produced by extending a neutral vowel over a typical infant-directed contour. In line with the investigators’ predictions, infants were more visually attentive in the context of the approving contours. In a related investigation (Fernald, 1992), 4-month-olds listened to infantdirected utterances of approval or prohibition in familiar (English) or unfamiliar (German, Italian, Greek, Japanese) languages. Infants exhibited more positive affect while listening to approving than to prohibiting utterances except for


17

Japanese utterances. The Japanese approvals and prohibitions were found to embody a narrower pitch range and were rated by adults as less intense than their counterparts in the other languages. Such findings as well as the attentional biases for natural, filtered, or synthesized infant-directed speech have led to an emphasis on the fundamental frequency (i.e., pitch) characteristics of infant-directed speech. Discussions of these results have tended to center on heightened pitch and expanded pitch range despite the fact that the pitch contours in question were always presented with temporal patterning appropriate to the context (infant- or adult-directed).Without separate manipulations of temporal and pitch patterning, researchers cannot disentangle their respective contributions. In the two studies of approving and disapproving infant-directed speech (Fernald, 1992; M. PapouSek et al., 1990), the temporal characteristics of these utterances contrasted markedly. The approving utterances embodied gradual modulations of pitch and loudness as opposed to the abrupt changes of disapproving utterances. Nevertheless, these researchers and others have tended to focus on pitch patterning at the expense of temporal patterning when they refer to the melodic contours or melodies of infant-directed speech. C. TEMPORAL PATTERNING AND ATTENTIONAL PREFERENCES

The temporal patterning of infant-directed speech has received more attention from researchers interested in the relations between prosodic features and syntactic structure (e.g., Hirsh-Pasek, Kemler Nelson, Jusczyk, Wright-Cassidy, Druss, & Kennedy, 1987; Kemler Nelson, Hirsh-Pasek, Jusczyk, & Wright Cassidy, 1989). For example, clause boundaries tend to be marked by prosodic features such as final syllable lengthening and pitch change. Might prosodic marking be enhanced in infant-directed speech and would prelinguistic infants be sensitive to such prosodic features? To address these questions, Hirsh-Pasek et al. (1987) and Kemler Nelson et al. (1989) altered recordings of speech by inserting pauses at clause boundaries (i.e. between clauses) or at other locations (i.e., within clauses). The former manipulation maintained the structural integrity of clauses and the latter altered the internal temporal structure. Infants (7- to 10-month-olds) exhibited a preference for intact clauses when the speech was infant directed but no preference when it was adult directed. A number of implications follow from these findings. First, prelinguistic infants process temporal pattern cues in complex speech sequences, particularly when such cues are linked to other important prosodic features. Second, infants are sensitive to prosodic cues associated with syntactic (i.e., clause) structure in infant-directed speech. Third, cues to at least some aspects of syntactic structure are enhanced in infant-directed speech. Such links between salient prosodic cues and critical syntactic features could facilitate infants’ parsing of the speech

18


stream and, by extension, their acquisition of language (Kemler Nelson et al., 1989; Morgan, Meier, & Newport, 1987).

V. Infants’ Processing of Musical Sequences Research on infants’ processing of speech sequences has provided insights into the impact of infant-directed speech on its intended audience. Unfortunately, no researchers in the musical domain have examined the effects of infant-directed song on infant listeners; however, some information is available on infants’ perceptual capabilities with respect to music in general. Such information permits preliminary conjectures about infants’ potential for responding differentially to infant- and adult-directed music. A. AlTENTIONAL PREFERENCES

One investigation of infants’ temporal processing of musical sequences was a direct analog of the Hirsh-Pasek et al. (1987) and Kemler Nelson et al. (1989) studies of prosodic cues to clausal structure. Krumhansl and Jusczyk (1990) examined infants’ sensitivity to phrase structure in music. To do so, they created two temporally altered versions of Mozart minuets. In one, they inserted pauses between musical phrases, thereby maintaining the internal temporal structure of such phrases. In the other, they altered the phrase structure by inserting pauses within phrases. Infants (4 and 6 months of age) showed greater attentiveness for the minuets with intact phrases than for those with altered phrases, indicating their sensitivity to musical phrase structure. An analysis of the phrase endings of these minuets revealed features similar to those that mark clause boundaries in speech. For example, the last melody note of the phrases tended to be lengthened, much like the last syllable in clauses. Also, musical phrases tended to end with a drop in pitch. Changes in pitch, whether increases or decreases, tend to characterize clause endings. Thus, infants are sensitive to the prosodic integrity of phrases in music and clauses in speech. In contrast to Krumhansl and Jusczyk’s (1990) research on attentional biases or preference, most other research on musical processing in infancy has been focused on the perception of various aspects of music, including pitch and temporal patterning (for reviews see Trehub, 1990; Trehub & Trainor, 1990, 1993). For the present purposes, this work is considered in relation to the following broad questions. What features of music are salient for infant listeners? Are some patterns processed more readily than others? Is “premusical” infants’ processing of music culture independent?


19

B. METHODOLOGICAL STRATEGY

The general procedural approach in these studies is to present melodies (up to 10 notes) that cannot be remembered in their entirety. In this way, the details

actually retained are informative about infants’ characteristic processing strategies. For example, infants’ retention of the exact pitches of the first or last few notes would be indicative of a local processing strategy; however, their retention of geperal pattern information (e.g., relations between the pitches or durations of notes) as opposed to specific details (e.g., exact pitches or durations) would be indicative of a global processing strategy. This procedural approach is implemented with an operant discrimination design (see Trehub, Bull, & Thorpe, 1984; Trehub, Thorpe, & Morrongiello, 1987). Specifically, infants 6 months and older are presented a standard pattern that is repeated continuously from a loudspeaker to one side and are trained to respond (i.e., turn) when a comparison pattern is substituted for the standard. Correct responses (i.e., turns to the comparison pattern) lead to reinforcement in the form of brief presentations of an animated toy; incorrect responses (false positives or incorrect rejections) have no consequences. By judicious selection of standard and comparison melodies, insight can be gained into infants’ processing strategies in general and the specific features that evoke discriminative responding. Typically, standard and comparison melodies are presented in different pitch registers (i.e., transposed) to study infants’ processing of pattern information rather than absolute pitch. In fact, infants treat exact transpositions as equivalent, just as adults do (Chang & Trehub, 1977; Trehub et al., 1984). They do not simply fail to discriminate the relevant differences but rather focus on relational as opposed to absolute pitch information (Trehub, 1990; Trehub & Trainor, 1990). C. PITCH PATTERNING

The results of various studies of melody perception have revealed infant performance that is surprisingly adult-like (Trehub & Trainor, 1993), with some features of melodies being much more salient than others. For example, a comparison melody that embodies a change in contour (i.e., the pattern of directional changes in pitch) relative to the standard melody is almost invariably responded to differentially by infant listeners (Trehub et al., 1984; Trehub, Thorpe, & Morrongiello, 1985); however, a comparison melody that embodies new notes but has the same pitch contour will likely be responded to in the same way as the original (Trehub et al., 1984, 1987). For infants, then, pitch-contour processing predominates in the perception of musical as well as speech patterns. Adults also treat contour as a highly salient feature of unfamiliar melodies (Dowling, 1978).

20

Sandra E . Trehub ei al.

Another feature that influences infants’ response to melodies is pitch range. When the pitch range of a comparison melody differs from that of the standard, infants are likely to respond even if relational pitch information is preserved (Trehub et al., 1984, 1985). D. TEMPORAL PATTERNING

In the aforementioned studies, the temporal configuration (i.e., rhythmic patterning) was held constant so that the pitch configuration could be assessed independently. Other research has established that infants also respond to melodies or tone sequences on the basis of their rhythmic or temporal structure (i.e., relative durations of notes) rather than the absolute durations of component notes (Trehub & Thorpe, 1989). Just as the identity of a melody is independent of specific pitches, so is it independent of any specific tempo, for infant listeners as well as adults. Infants also group or parse the component notes of patterns in much the same way as adults. For example, with patterns consisting of three tones of one type followed by three tones of another type (schematic structure: XXXOOO),infants detected pauses inserted within groups of similar notes (e.g., XXXO 00: struchue-disrupting changes) more readily than pause inserted between groups of similar notes (e.g., XXX 000: structure-conserving changes) (Thorpe & Trehub, 1989; Thorpe, Trehub, Morrongiello, & Bull, 1988). The propensity to group sounds in this way is likely implicated in infants’ “preference” for intact phrases in music (Krumhansl & Jusczyk, 1990) and for intact clauses in speech (Kemler Nelson et al., 1989). In short, infants readily encode information about pitch contour, pitch range, and temporal patterning from musical sequences. Thus, the very same features that predominate in infant-directed speech and influence infant attention toward such speech emerge as salient features of musical sequences. In all likelihood, infants use the same or very similar perceptual organizational devices in their processing of complex auditory patterns, whether speech or music (Trehub, 1989; Trehub & Trainor, 1993). Pitch contour and temporal patterning, for example, provide reasonable means for parsing the speech or musical stream into chunks appropriate for further processing. E. MELODY AND CONTOUR

A number of researchers (e.g., Fernald, 1989, 1992; M. PapouSek & Papouiiek, 1981; M. PapouSek et al., 1990, 1991) have referred to the typical pitch contours of infant-directed speech as melodies. This characterization ignores the usual distinctions between contour and melody. In musical parlance, contour is a general feature that captures directional pitch movement without regard to the


21

extent of such movement (Dowling, 1978;Dowling & Harwood, 1986). Descriptive terms like rising, fulling, and rise-fall or bell-shaped contours are consistent with this usage. Melody, in contrast, is usually defined in terms of the exact pitch intervals and rhythmic relations between notes in the context of a musical system. Even a simple rising contour can differ by consisting of doh re mi or doh mi soh, for example. Designating the latter pattern as expanded ignores the fact that it is a different melody or tune, despite the ascending sequence of three notes in both cases. Similarly, relative durational differences between notes generate different melodies. F. “GOOD” AND “BAD” MELODIES

As noted earlier (Section V,C),a number of investigationsrevealed that infants are primarily contour processors. This finding does not imply, however, that they are exclusively contour processors. As will become clear, for example, not all bell-shaped melodies are equivalent for infant listeners. Some are “good” in the sense that they are readily processed and others “bad,” being processed with greater difficulty (Trehub & Trainor, 1993). Progress in delineating the critical features of such good and bad melodies might suggest new avenues for research on infant-directed speech. Every language has an inventory of component sounds, approximate boundaries of sound categories, and rules for combining sounds. Correspondingly, every musical system has an inventory of musical pitches (notes) organized into scales as well as various conventions for arranging these notes into acceptable sequences. Presumably, constraints of the auditory system have had some influence on the selection and arrangement of sounds in music, as in speech (Stevens & Keyser, 1989). Some idiosyncratic or culturally based rules are probably also operative. If any natural musical rules or features can be identified, such features might be relevant in the early processing of speech prosody. Extensive research has been conducted on the music processing skills of adult listeners, including those who have received incidental musical exposure rather than formal training (for reviews, see Deutsch, 1982; Dowling & Harwood, 1986; Handel, 1989; Krumhansl, 1990). This research, however valuable, cannot provide insights into natural rules or processing proclivities for a number of reasons. Principal among these is the long exposure to a musical system that would effectively mask the distinction between naturally good .melodies and those that become good (i.e., easily processed) by means of such exposure. Infants, however, would have limited exposure to the music of their culture and thus are unlikely to have learned a great deal about its idiosyncrasies. In this sense, they are reasonable candidates for evaluating good patterns. Patterns that infants would process more readily than others could be considered good or as embodying good features. Learning is not necessarily excluded. The only re-

22

Sandra E . Trehub er al.

quirement is that processing priority for such patterns be exhibited in early life. If learning is implicated, this type of learning may be innately guided (Jusczyk & Bertoncini, 1988; Marler, 1990). A number of investigators have confirmed that Western adults exhibit enhanced processing (e.g., more detailed encoding, better retention, preference) for melodies that conform to the musical conventions of their culture (Cuddy, Cohen, & Miller, 1979; Krumhansl & Keil, 1982; Watkins, 1985). This effect might stem from familiarity with such conventions or from naturally or inherently good features of such patterns (or both). Researchers have identified some melodies that seem to be inherently good in the sense that Western infants process them in greater detail than is the case for other similar melodies. Of particular interest is the fact that these good melodies embody rules or conventions that are central to Western music. The investigations in question had standard and comparison melodies with bell-shaped (risefall) contours, the comparison melody differing minimally from the standard (i.e., a semitone in one position only). As usual, the standard and comparison melodies were presented in transposition, precluding the use of absolute pitch cues (see Fig. 2). When the standard melody was created from the principal notes of the Western major scale-notes considered important from a music-theoretic perspective (Piston, 1969) and from adult listeners’ perspective (Krumhansl, 1983)-infants went beyond their usual contour processing strategy, and encoded and retained information about intervals (i.e., exact pitch relations between adjacent notes) (Cohen, Thorpe, & Trehub, 1987; Trehub, Thorpe, & Trainor, 1990; Trainor, 1991). When, however, the standard melody had one or more “wrong” notes in terms of Western scale structure (Cohen et al., 1987; Trainor, 1991; Trehub, Thorpe, & Trainor, 1990) or when it deviated from that structure in more substantial ways (Trehub, Thorpe, & Trainor, 1990), infants did not encode and retain interval information. These findings are consistent with the notion that aspects of Western major scale structure have their origins in universal constraints on auditory pattern processing. In other words, such aspects may be inherently good. These findings do not imply that Western major scale structure is superior to alternative scale structures (e.g., those in other cultures). Rather, the notion is simply that Western major scale structure exemplifies one of many possible instances of good form. Scales that are fundamental to other cultures are also likely to exemplify good form. Accordingly, infants should exhibit comparable enhanced processing for melodies based on such foreign scales. In fact, research by Lynch, Eilers, Oller, and Urban0 (1990) supports this contention. They evaluated the ability of Western 6-month-oldsand adults (musically untrained) to detect very subtle pitch changes to simple rise-fall melodies consisting of notes from the Western major scale or from the Javanese pilog scale. Not surprisingly, adults performed better on the major than on the pilog melodies. Infants ,also


23

b D F I A t F l D 370 466 370 294

294

F 349

A C 440 523

A

F

440 349

C B D I B C 392 494 622 494 392

C 262

E

C

330 415

I

E C 330 262

C 262

E

d

330 415

E

C

330 262

HL

Fig. 2 . Each example depicts three successive presentations of the standard melody followed by a change and subsequent return to the standard melody. The upper melody ( a ) conforms to Western major scale structure: the lower melody (b) has one "wrong" note (the third).

detected changes to the melodies but, unlike adults, their performance was equivalent for the two structurally distinct scales types. This latter finding suggests that both scales begin on an equal footing for musically unacculturated listeners but that culture-specific experience enhances one relative to the other. Similar arguments have been advanced for the speech sounds of different languages (Best, McRoberts, & Sithole, 1988; Burnham, Earnshaw, &Clark, 1991; Werker & Lalonde, 1988). What features of the major scale might contribute to this relative ease of processing? Its most prominent notes (e.g., first, third, and fifth notes) in terms of frequency of occurrence, perceptual ratings, and music-theoretic considerations (see Krumhansl, 1990) are related by simple (small-integer) frequency ratios. Such ratios may form the basis of naturally salient or minimally dissonant (rough) intervals (see Bums & Ward, 1982). The perfect fifth interval, which

24

Sundru E. Trehub el al.

relates the first and fifth notes of the major scale, approximates the ratio of 3 : 2, which is considered the second most consonant (i.e., least dissonant) interval (the most consonant being the octave with a ratio of 2 : 1). Of interest is the fact that the 3 : 2 ratio is prominent in the overtones of vowels and many other natural sounds. Is it a mere coincidence that perfect fifth intervals figured prominently in the Western melodies for which infants exhibited interval processing (Cohen et al., 1987; Lynch et al., 1990; Trainor, 1991; Trehub, Thorpe, & Trainor, 1990)? Despite the much heralded differences in the component notes of different scales (Harwood, 1976), perfect fifth intervals are prominent cross-culturally, particularly in vocal music (Kolinski, 1967; Nettl, 1956). Infants’ enhanced performance (i.e., interval processing) for melodies based on the Western major (Cohen et al., 1987; Trainor, 1991; Trehub, Thorpe, & Trainor, 1990) and Javanese pilog (Lynch et al., 1990) scales is indicative of culture-independent processing. Further support for culture-free music pmcessing in infancy has been provided by Trainor and Trehub (1992a). They presented infants and adults a 10-note standard melody based on the major scale and a comparison melody with only one note altered relative to the standard. The altered note was either consistent with Western musical rules or in violation of these rules (i.e., within or outside the prevailing key). For adults, a small pitch change that violated musical conventions was much easier to detect than a considerably larger pitch change that followed cultural conventions. For infants, however, performance was equivalent for both changes. Of particular interest was the fact that infants’ level of performance on the rule-conserving change significantly exceeded that of adults. Essentially, adults’ tacit knowledge of Western musical structure obscured a pitch difference that was perceptible for infant listeners. This finding parallels infants’ ability to discriminate some foreign speech contrasts that are difficult for adults (Trehub, 1976; Werker & Tees, 1984; Werker & Lalonde, 1988). G . GENERAL PERSPECTIVES

What can we conclude about musical processing in infancy? At the very least, infants have the prerequisite skills for processing musical input directed to them. They can parse the musical stream; they engage in relative pitch and temporal processing; and they characteristically extract pitch contours and rhythms from musical sequences. For certain patterns, so-called good melodies, they can encode and retain subtle details such as the precise relations between adjacent notes (i.e., intervals). Infants’ adult-like perception of musical patterns implies that many aspects of music processing are not arbitrary. Instead, such processing seems to engage basic principles of perceptual organization that are operative early in life. The present conception challenges the prevailing view that long-term exposure to


25

music promotes the extraction of pitch and temporal regularities and the internalization of musical conventions (Krumhansl, 1990; Jones, 1990). No doubt some aspects of musical structure are arbitrary or conventional (Trainor & Trehub, 1992a), but these may be relatively few. The views espoused here clash with those of contemporary composers and music theorists (e.g., Boulez, 1971; Schoenberg, 1975), who contend that audiences, with comparable exposure, would be as comfortable with 12-tone compositions as with music from Beethoven or The Beatles.

VI. Music and Speech: Present Perspectives and Future Prospects A. CAREGIVERS’ SPEECH AND SONG

A number of intriguing parallels have been found between speech and song to infants. Both seem to involve significant adjustments in relation to standard speech and song, adjustments that go beyond the speaker’s or singer’s awareness. Among such adjustments are simple contours, elongated vowels, and emotional expressiveness. Speech and song alike have variants linked to the infant’s state and perhaps to the caregiver’s state, as well. Thus, speech to capture and hold infants’ attention has its counterpart in play songs, and soothing speech has its counterpart in lullabies. Further, when speech and song are perceived to be of the soothing variety, descending contours prevail. To the extent that speech and song have meaning for infant listeners, these meanings are likely affective and nonarbi trary. The parallels between speech and song to infants are especially impressive in view of the fact that adult-directed speech has no musical analog (except for artificial performing contexts such as opera). Although singing is not conventionally directed to individuals, caregivers from time immemorial have considered one-on-one singing a necessary or desirable part of their interactive repertoire with infants. Lullabies and play songs can be distinguished from other sung materials within a culture but they are similar in many respects to these other materials. All songs from a culture, regardless of their audience, conform to the same musical system and, consequently, share a number of features that distinguish them from the songs of other cultures. In descriptions of speech, the emphasis has been on differences between infant- and adult-directed varieties as well as cross-cultural similarities in speech to infants (e.g., Fernald et al., 1989; Grieser & Kuhl, 1988). The cross-cultural similarities and within-culture differences may have been overstated, however. If pitch contours constitute the essence of speech to infants and such contours are universal, then infant-directed speech samples from

26

Sandra E. Trehub ei al.

one culture should be more similar to those of other cultures than to adultdirected samples from the same culture (all samples content filtered, of course). In other words, language identity should be obscured in the prosody of speech to infants. This outcome is unlikely to be the case. Rather, cues to language identity probably coexist with cues to infant-directedness. In any event, the relative contributions of these factors should be empirically established so that the phenomenon of infant-directed speech can be understood more fully. Differences between infant- and adult-directed speech may also have been exaggerated. Qpically, emotionally charged infant-directed speech is compared with emotionally neutral adult-directed speech. For example, Fernald and Simon’s (1984) speech samples were derived from mothers talking to their 4month-old infants or to the adult interviewer. Surely, features attributable to heightened (positive) affect must be separated from those attributable to infantdirectedness. Samples of speech from new lovers might shed light on this issue. For example, happiness and joy in adult-directed speech are known to be associated with high pitch, large pitch variability, and an upward pitch contour, and pleasantness is known to be associated with low pitch and a downward pitch contour (Scherer & Oshinsky, 1977). Another s o m e of artifactual differences may be the stereotyped repertoire of infant-directed utterances. Stereotyped messages pose few cognitive demands on the speaker compared with the requirements of planning and executing syntactically, semantically, and phonologically appropriate messages for adults. As noted earlier (Section II,C), the prosody of stereotyped utterances to adults (e.g., Hi! How are you? Great!) may resemble infant-directed prosody more than typical adult-directed prosody. Perhaps highly stereotyped or rehearsed adultdirected messages would reveal the contribution of processing demands to prosodic form. Everyday experience clearly shows that actors, public speakers, and lecturers exhibit more pitch and loudness modulation than those with less experience or skill. Does this difference arise from the speaker’s relative ease with the content, from effective simulation of an appropriate emotional state, or from sensitivity to the listener’s preferences and needs? Perhaps some vocal adjustments are triggered by the listener’s level of verbal comprehension. For example, Hirsh-Pasek and Treiman ( 1982) observed speech of the infant-directed variety in adults’ interaction with pets (termed doggerel by the authors). Do these adjustments arise from the listener’s verbal intelligence (i.e., low), the speaker’s feelings about the listener (e.g., warm),or both? Disentangling the contributions of these various factors should lead to clearer and more informed distinctions between speech addressed to infants and adults. Such distinctions might shed new light on the origin and significance of common features in infant-directed speech and song. One potentially important aspect of infant-directed speech, notably its voice quality, has not received the attention it merits. Anecdotal accounts include


27

references to “gentle,” “soothing,” or even “loving” tones of voice, but no systematic descriptions of acoustic or perceptual features relating to voice quality are available. In view of the known links of voice quality to emotional state (Frick, 1985; Scherer, 1981), greater specification of the spectral structure of infant-directed speech and song would be useful. This enterprise could be guided by research on the spectral structure of adult-directed speech, its relation to the speaker’s emotional state, and the listener’s decoding of that state (Laver, 1980; Lieberman & Michaels, 1962; Newman & Emanuel, 1991; Scherer & Oshinsky, 1977). B. INFANTS’ SPEECH AND SONG PROCESSING

The organization of speech into clauses and of music into phrases does not depend on the listener’s comprehension of the speech content or familiarity with culture-specific conventions (Kemler Nelson et al., 1989; Krumhansl & Jusczyk, 1990). Instead, adult-like grouping processes seem to be available in infancy to guide the segmentation or parsing of auditory sequences (Demany, 1982; Thorpe & Trehub, 1989; Thorpe et al., 1988). Common mechanisms seem to operate for speech and musical sequences, highlighting important constituents in both cases (Krumhansl, in press; Trehub & Trainor, 1993). The success to date in identifying infants’ grouping and segmentation strategies with respect to musical sequences constitutes an effective argument for further research of this nature. The identification of linguistically relevant prosodic cues in speech to infants should not obscure the fact that these cues are also present in speech to adults. Intonation (patterns of pitch, loudness, and timing) in speech to adults has a multiplicity of functions including the portrayal of emotions, the identification of grammatical constituents, and the selective emphasis of particular meanings (Handel, 1989, p. 448). Good or expressive speakers, those who sound natural, tend to produce well-modulated speech with clear segmentation cues. Such cues may allow the listener to decode spoken messages with less effort than is necessary for messages with less prominent segmentation cues. A reasonable strategy would be to evaluate the processing advantages, even for adults, that arise from prominent prosodic features in infant-directed speech. An important need, moreover, is to establish whether differences in the incidence of segmentation cues in infant- and adult-directed speech are real or simply a by-product of inappropriate sampling of adult-directed speech, as outlined earlier (Section VI,A). When prosodic features in infant-directed speech are viewed as facilitators of language acquisition (Kemler Nelson et al., 1989), the implication is that such features are unique either to infant-directed speech or to speech in general. Not only are these features present in adult-directed speech and music, but they also mark important boundaries in both (Handel, 1989; Krumhansl & Jusczyk, 1990). A reasonable conclusion, then, is that such features are domain general rather

28

Sandra E. Trehub er al.

than domain specific. Further, prosodic cues to linguistic and musical boundaries are perhaps more transparent in simple speech and musical sequences (e.g., those directed to infants) than they are in more complex sequences (e.g., those directed to adults). Few would dispute the notion that simple stimuli can be processed more readily than more complex stimuli; however, few would generate identical sets of defining criteria for simplicity. In this’light, the ongoing debate about the utility of infant-directed speech for later language acquisition (e.g., Gleitman, Newport, & Gleitman, 1984; Kemler Nelson et al., 1989; Murray et al., 1990) may not be fruitful. Instead, a more profitable approach may be to establish the motivational basis for infant-directed speech adjustments and to delineate in greater detail the immediate effects on infants. Infants exhibit an attentional “preference” for infant- over adult-directed speech (e.g., Cooper & A s h , 1990; Fernald, 1985) and for approving over disapproving infant-directed speech (M. PapouSek et al., 1990) together with greater affective responsiveness to such signals (Fernald, 1992; Werker & McLeod, 1989). Comparable issues in infant-directed song have not been explored to date. One could, however, determine whether infants prefer lullabies over adult songs and whether such a preference (if present) is demonstrable only in sleepy or fussy infants. Would awake, alert infants prefer play songs over lullabies and also over adult songs? Do lullabies actually induce sleep more rapidly (e.g., briefer latency of sleep onset) than other songs? If voice quality differs as a function of the speaker’s or singer’s emotional state, are such differences perceptible to infant listeners? Infants can differentiate and categorize the timbre or quality of single sounds (Clarkson, Clifton, & Penis, 1988; Trehub, Endman, & Thorpe, 1990) but whether they can discriminate timbre in sound sequences and whether voice timbre is salient for them are unknown. Those who argue that the production of emotional voice qualities is biologically based (see Frick, 1985) maintain that their perception has a similar basis, as is the case for emotional qualities associated with facial expression (Izard, 1977) and patterns of touching (Clynes, 1977). In any event, the study of infants’ discrimination of emotionally relevant voice qualities and their preference for particular qualities is clearly warranted. One finding arising from the musical domain, that of enhanced processing for so-called good melodies (e.g., Trehub, Thorpe, & Trainor, 1990), has no obvious parallel in infant-directed speech. To date, descriptions of maternal “melodies” in speech have not proceeded beyond the specification of their contours (e.g., Fernald et al., 1989) or distinctions between smooth and abrupt pitch and loudness transitions (e.g., Fernald et al., 1989; M. PapouSek et al., 1990). Perhaps the pitch ratio of 3 : 2, which seems to characterize good melodies, also figures prominently in typical infant-directed speech (Trehub & Trainor, 1993). This ratio could be highlighted in a number of ways. Its prominence in the simultaneous (i.e., harmonic) components of vowels could be enhanced by their

Music and Speech Processing in lnfancy

29

elongation, which also occurs in infant-directed speech and song, or by heightened pitch, which occurs in both. At a more speculative level, the relations between significant sounds in speech such as stressed words or syllables could embody the pitch ratio of 3 : 2. If so, Fernald’s (1989) claim that “the melody is the message” would take on new meaning.

REFERENCES Anderson, E. S. (1986). The acquisition of register variation by Anglo-American children. In B. B. Schieffelin & E. Ochs (Eds.), Language socialization across cultures (pp. 153-166). Cambridge: Cambridge University Press. A s h , R. N., Pisoni, D. B., & Jusczyk, P. W. (1983). Auditory development and speech perception in infancy. In P. H. Mussen (Ed.),H u n d h k of childpsychology. Vol. 11: Infancy and developmental psychology (pp. 573-687). New York: Wiley. Bebey, F. (1969). African music: A people’s art. Westport, CT: Lawrence Hill. Beebe, B., Feldstein, S . , Jaffe, J., Mays, K., & Alson, D. (1985). Interpersonal timing: The application of an adult dialogue model to mother-infant vocal and kinesic interactions. In T. M. Field & N. A. Fox (Eds.), Social perception in infants (pp. 217-247). Norwood, NJ: Ablex. Best, C. T., McRoberts, G.W., & Sithole, N. M. (1988). Examination of perceptual reorganization for nonnative speech contrasts: Zulu click discrimination by English-spaking adults and infants. Journal of Experimental Psychology: Human Perception and Performance. 14. 45-60. Bolinger, D. (1970). Relative height. In D. Bolinger (Ed.), Intonation: Selected readings (pp. 137157). Hammondsworth: Penguin. Boulez, P. (1971). Boulez on music today ( S . Bradshaw & R. R. Bennett, trans.). London: Faber & Faber. Brakeley, T. C. (1 950). Lullaby. In Standard dictionary of folklore, mythology, and legend ( pp. 653654). New York Funk & Wagnalls. Brown, M. J. E. (1980). Lullaby. In The new Grove dictionary of music ond musicium (pp. 313314). London: Macmillan. Bumham, D. K., Earnshaw, L. J., & Clark, J. E. (1991). Development of categorical identification of native and non-native bilabial stops: Infants, children and adults. Journal of Child Language, 18, 231-260. Bums, E. M., & Ward, W. D. (1982). Intervals, scales, and tuning. In D. Deutsch (Ed.), The psychology of music (pp. 241-269). Orlando, FL: Academic Press. Cass-Beggs, B., & Cass-Beggs, M. (1969). Folk lullubies. New York Oak Publications. Caudill, W. A,, & Weinstein, H. (1969). Maternal care and infant behavior in Japan and America. Psychiatry, 32, 12-43. Chang, H. W., & Trehub, S. E. (1977). Auditory processing of relational information by young infants. Journal of Experimental Child Psychology, 24, 324-331, Clarkson, M. G., Clifton, R. K., & Penis, E. E. (1988). Infant timbre perception: Discrimination of spectral envelopes. Perception & Psychophysics, 43, 15-20. Clynes, M. (1977). Seniics, the touch of emotions. New York: Doubleday. Cohen, A. J., Thorpe, L. A., & Trehub, S. E. (1987). Infants’ perception of musical relations in short transposed tone sequences. Canadian Journal of Psychology, 41, 33-47. Cong-Huyen-Ton-Nu, N.-T. (1979). The functions of folk songs in Vietnam. In J. Blacking & J. W. The , performing arts: Music and dance (pp. 141-151). The Hague: Kealiinohomoku (Us.) Mouton.

30


Cooper, R. P. (in press). The effect of prosody on young infants’ speech perception. Advances in infancy Research. Cooper, R. P., & A s h , R. N. (1990). Preference for infant-directed speech in the first month after birth. Child Development, 61, 1584-1595. Crystal, D. (1973). Non-segmental phonology in language acquisition: A review of the issues. Lingua, 32, 1-45. Cuddy, L. L., Cohen, A. J., & Miller, J. (1979). Melody recognition: The experimental application of musical rules. C a n u d h Journal of Psychology, 33, 148-156. Curtis, N. (1921). American Indian cradle songs. The Musical Quarterly. I , 549-558. Demany, L. (1982). Auditory stream segregation in infancy. Infanr Behavior and Development, 5 , 261 -276. Deutsch, D. (1982). Grouping mechanisms in music. In D. Deutsch (Ed.), The psychology of music (pp. 99-134). New York: Academic Press. Devereux, G. (1948). The Mohave neonate and its cradle. Primitive Man, 21, 1-18. Dixon, S., Tronick, E., Kesfer. C., & Brazelton, T. B. (1981). Mother-infant interaction among the Gusii of Kenya. In T.M. Field, A. M. Sostek, P. Vietze, & P.H. Liederman (Eds.), Culture and early interactions (pp. 149- 168). Hillsdale, NJ: Erlbaum. Dowling, W. J. (1978). Scale and contour: TWOcomponents of a theory of memory for melodies. Psychological Review, 85, 341-354. Dowling, W. I., & Harwood, D. L. (1986):Music cognition. Orlando, FL: Academic Press. Dunn, J., & Kendrick, C. (1982). The speech of two- and three-year-olds to infant siblings: “Baby talk” and the context of communication. Journal of Child Language, 9, 579-595. Emanuel, F. W.,& Smith, W. F. (1974). Pitch effects on vowel roughness and spectral noise. Journal of Phonetics. 2, 247-253. Ferguson, C. (1964). Baby talk in six languages. American Anthropologist. 66, 103-1 14. Fernald, A. (1984). The perceptual and affective salience of mothers’ speech to infants. In L. Feagans, C. Garvey, & R. Golinkoff (Eds.), The origins and growrh of communication (pp. 529). Nonvood, NJ: Ablex. Fernald, A. (1985). Four-month-old infants prefer to listen to motherese. Infatu Behavior and Development, 8, 181-195. Fernald, A. (1989). Intonation and communicative intent in mothers’ speech to infants: Is the melody the message? Child Development, 60, 1497- 15 10. Fernald, A. (1992). Meaningful melodies in mother’s speech to infants. In H. PapouSek, U.Jiirgens, & M. PapouSek (Eds.),Nonverbal vocal communication: Comparative and developmental aspects (pp. 262-282). Cambridge: Cambridge University Press. Fernald, A., & Kuhl, P. K. (1987). Acoustic determinants of infant preference for motherese. Infant Behavior and Development, 10, 279-293. Fernald, A., & Mazzie, C. (1991). Prosody and focus in speech to infants and adults. Developmental Psychology, 27, 209-221. Fernald, A., & Simon, T. (1984). Expanded intonation contours in mothers’ speech to newborns. Developmental Psychology, 20, 104- 113. Fernald, A., Taeschner, T., Dunn, J., Papoubek, M., de Boysson-Bardies, B., & Fukui, I. (1989). A cross-language study of prosodic modifications in mothers’ and fathers’ speech to preverbal infants. Journal of Child Development, 16. 477-501. Ferrier, L. J. (1985). Intonation in discourse: Talk between 12-month-olds and their mothers. In K. Nelson (Ed.), Children’s language (Vol.5 , pp. 35-60). Hillsdale, NJ: Erlbaum. Fonagy, 1. (1981). Emotions, voice, and music. Research Aspecrs on Singing, 33, 51-79. Franc6s. R. (1958). La perception de la musique. Paris: Vrin. Frick. R . W. (1985). Communicating emotion: The role of prosodic features. Psychological Bulletin, 97. 412-429.


31

Garnica, 0. K. (1977). Some prosodic and paralinguistic features of speech to young children. In C. E. Snow & C. A. Ferguson ( a s . ) , Talking to children: Language input and acquisition (pp. 6688). Cambridge: Cambridge University Press. Gleitman, L. R., Newport, E. L., & Gleitman, H. (1984). The current status of the motherese hypothesis. Journal of Child Language, I l . 43-79. Grieser, D. L., & Kuhl, P. K. (1988). Maternal speech to infants in a tonal language: Support for universal prosodic features in motherese. Developmental Psychology, 24, 14-20. Handel, S . (1989). Listening: An introduction to the perception of auditory events. Cambridge, MA: MIT Press. Hanvood, D. L. (1976). Universals in music: A perspective from cognitive psychology. Ethnomusicology, 20, 521-534. Hilger, M. 1. (1951). Chippewa child life and its cultural background. Bureau of American Erhnology Bullerin, 146. Hilger, M. 1. (1952). Arapaho child life and its cultural background. Bureau of American Ethnology Bullerin, 148. Hirsh-Pasek, K., Kemler Nelson, D. G . , Jusczyk, P. W., Wright-Cassidy, K.,DNSS, B., & Kennedy, L. (1987). Clauses are perceptual units for young infants. Cognition. 26, 269-286. Hirsh-Pasek, K., & Treiman. R. (1982). Doggerel: Motherese in a new context. Journal of Child Language, 10. 23-37. Izard, C. E. (1977). Human emorions. New York: Plenum. Jacobson, J. L., Boersma, D. C., Fields, R. B., & Olson, K. L. (1983). Paralinguistic features of speech to infants and small children. Child Development, 54, 436-442. Joci6, M. (1978). Adaptation in adult speech during communication with children. In N. Waterson &’ C. Snow (Eds.), The development of communication (pp. 159-171). Chichester, England: Wiley. Jones, M. R. (1990). Learning and the development of expectancies: An interactionist approach. Psychomusicology, 9, 193-228. Jusczyk, P. W. (1592). Developing phonological categories from the speech signal. In C. E. Ferguson, L. Menn, & C. Stoll-Gammon (Eds.), Phonological development: Models, research, implications (pp. 17-64). Timonium, MD: York Press. Jusczyk, P. W., & Bertoncini, J. (1988). Viewing the development of speech perception as an innately guided learning process. Language and Speech, 31, 217-238. Karzon, R. G. (1985). Discrimination of polysyllabic sequences by one- to four-month-old infants. Journal of Experimental Child Psychology. 39. 326-342. Kemler Nelson, D. G., Hirsh-Pasek, K., Jusczyk, P. W., & Wright Cassidy, K. (1989). How the prosodic cues in motherese might assist language learning. Journal of Child Language, 16, 5568. Knower, F. H. (1941). Analysis of some experimental variations of simulated vocal expressions of the emotions. Journal of Social Psychology, 14, 369-372. Kolinski, M. (1967). Recent trends in ethnomusicology. Ethnomusicology. 11 , 1-24. Krauss, R. M., Curran, N. M., & Ferleger, N. (1983). Expressive conventions and the cross-cultural expression of emotion. Basic and Applied Social Psychology, 4 , 295-305. Krumhansl, C. L. (1983). Perceptual structures for tonal music. Music Perception, 1 , 28-62. Krumhansl, C. L. (1990). Cognitive foundations of musical pitch. New York: Oxford University Press. Krumhansl. C. L. (in press). Grouping processes in infants’ music perception. In J. Sundberg, L. Nord, & R. Carl (Eds.), Grouping in music. Stockholm: Royal Swedish Academy of Music. Krumhansl, C. L., & Jusczyk, P. W. (1990). Infants’ perception of phrase structure in music. Psychological Science, I , 70-73. Krumhansl, C. L., & Keil, F. C. (1982). Acquisition of the hierarchy of tonal functions in music. Memory and Cognition, 10, 243-251.

32


Kuhl, P. K. (1988). Auditory perception and the evolution of speech. Human Evolution, 3 , 19-43. Langer, S . (1957). Philosophy in a new key. Cambridge, MA: Harvard University Press. Laver, J. (1980). Thephqnetic description of voice quality. Cambridge: Cambridge University Press. Lewis, M. M. (1951). Infant speech. London: Routledge & Kegan Paul. Lieberman, P. (1984). The biology and evolution of language. Cambridge, MA: Harvard University Press. Lieberman, P.,& Michaels, S. B. (1962). Some aspects of fundamental frequency and envelope amplitude as related to the emotional content of speech. Journal of the Acoustical Society of America. 34, 922-927. Lynch, M. P., Eilers, R. E., Oller, D. K., & Urbano, R. C. (1990). Innateness, experience, and music perception. Psychological Science, 1 , 272-276. Malsheen, B. J. (1980). Wo hypotheses for phonetic clarification in the speech of mothers to children. In G. H. Yeni-Komshian, 1. F. Kavanaugh, & C. A. Ferguson (Eds.), Child Phonology. Vol. 2: Perception (pp. 173-184). New York Academic Press. Marler, P. (I 990). Innate learning preferences: Signals for communication. Developmental Psychobiology, 23, 557-568. Mayer, N. K . , & Tronick, E. 2.(1985). Mothers’ turn-giving signals and infant turning-taking in mother-infant interaction. In T. M. Field & N. A. Fox (Eds.), Social perception in infants (pp. 199-216). Norwood, NJ: Ablex. McCosker, S. S. (1974). The lullabies of the San Blas Cuna Indians of Panama. Gothenburg: Gothenburg Ethnographical Museum. Mehler, J . , Bertoncini, J., Barrihe, M., & Jassik-Gerschenfeld, D. (1978). Infant recognition of mother’s voice. Perception, 7, 491-497. Meyer, L. B. (1956). Emotion and meaning in music. Chicago: University of Chicago Press. Morgan, J. L.,Meier, R. P., & Newport, E. L. (1987). Structural packaging in the input to language learning: Contributions of prosodic and morphological marking of phrases to the acquisition of language. Cognitive Psychology, 19, 498-550. Morikawa, H., Shand, N., & Kosawa, Y. (1988). Maternal speech to prelingual infants in Japan and the United States: Relationships among functions, forms and referent. Journal of Child Language, 15, 237-256. Morrongiello, B. A. (1988). The development of auditory pattern perception skills. Advances in Infancy Research, 5 , 135- 172. Murray, A. N., Johnson, J., & Peters, J. (1990). Fine-tuning of utterance length to preverbal infants: Effects on later language development. Journal of Child Language, 17. 51 1-525. Murray, L., & Trevarthan, C. (1986). The infant’s role in mother-infant communications. Journal of Child Language, 13, 15-29. Nettl, B. (1956). Music in primitive culture. cambridge, MA: Harvard University Press. Newman, R. A,, & Emanuel, F. W. (1991). Pitch effects of vowel roughness and spectral noise for subjects in four musical voice classifications. Journal of Speech and Hearing Research, 34, 753760. Newport, E., Gleitman, L., & Gleitman, H. (1977). Mother, I’d rather do it myself Some effects and noneffects of Motherese. In C. E. Snow & C. A. Ferguson (Eds.). Talking to children: Language input and acquisirion (pp. 109- 149). Cambridge: Cambridge University, Press. PapouSek, H., & PapouSek, M. (1984). Learning and cognition in the everyday life of human infants. Advances in the Study of Behavior, 14, 127-163. PapouSek, M., Bomstein, H., Nuzzo, C., PapouSek, H., & Symmes, D. (1990). Infant responses to prototypical melodic contours in parental speech. Infanr Behavior and Developmenr. 13. 539545. PapouSek, M., & PapouSek, H. (1981). Musical elements in the infant’s vocalization: Their significance for communication, cognition, and creativity. Advances in Infancy Research, I , 163-224.


33

PapouSek, M., Papoukk, H., & Bornstein, M. H. (1985). The naturalistic vocal environment of young infants: On the significance of homogeneity and variability in parental speech. In T. M. Field & N. A. Fox (Eds.), Social perception in infants (pp. 269-297). Nonvood, NJ: Ablex. PapouSek, M., PapouSek, H., & Haekel, M. (1987). Didactic adjustments in fathers’ and mothers’ speech to their three-month-old infants. Journal of Psycholinguistic Research, 16, 306-319. PapouSek, M.,PapouSek, H., & Symmes, D. (1991). The meanings of melodies in motherese in tone and stress languages. Infant Behavior and Development, 14, 415-440. Pegg, J. E., Werker, I. F., & McLeod, P. I. (1992). Preference for infant-directed over adult-directed speech: Evidence from 7-week-old infants. lnjant Behavior and Development, 15. 325-345. Piston, W. (1969). Harmony. New York: Norton. Rheingold, H.. 8r Adams, J. L. (1980). The significance of speech to newborns. Developmental Psychology, 16, 397-403. Rosenthal, M. K. (1982). Vocal dialogues in the neonatal period. Dewelopmenral Psychology, 18, 17-21.

Sachs, J. (1977). The adaptive significance of linguistic input to prelinguistic infants. In C. E. Snow & C. A. Ferguson (Eds.), Talking to children: Language input and acquisition (pp. 51-61). Cambridge: Cambridge University Press. Sachs, J., & Devin, J. (1976). Young children’s use of age-appropriate speech styles in social interaction and role-playing. Journal of Child Lnnguage. 3, 81-98. Sakata, H.L. (1987). Hazara women in Afghanistan: Innovators and preservers of a musical tradition. In E. Koskoff (Ed.), Women and music in cross-culrural perspective (pp. 85-95). Westport, CT: Greenwood Press. Sands, K. M., & Sekaquaptewa, E. (1978). Four Hopi lullabies: A study in method and meaning. American Indian Quarterly. 4 , 195-210. Scherer, K. R. (1974). Acoustic concomitants of emotional dimensions: Judging affect from synthesized tone sequences. In S. Weitz (Ed.), Non-verbal communication (pp. 105-1 14). New York: Oxford University Press. Scherer, K. R. (1981). Speech and emotional states. In J. K. Darby, Jr. (Ed.), Speech evaluation in psychiarry ( pp. 189-220). New York: Grune & Stratton. Scherer, K. R., Koivumaki, J., & Rosenthal, R. (1972). Minimal cues in the vocal communication of affect: Judging emotions from content-masked speech. Journal of Psycholinguistic Research, I , 269-285.

Scherer, K. R., & Oshinsky, J. S . (1977). Cue utilization in emotion attribution from auditory stimuli. Motivation and Emotion, I , 331-346. Schneider, B. A,, & Trehub, S. E. (1992). Sources of developmental change in auditory sensitivity. In L. A. Werner & E. W. Rubel (Eds.), Developmentalpsychoacoustics (pp. 3-46). Washington, DC:American Psychological Association. Schoenberg, A. (1975). Style and idea (L.Black, trans.). London: Faber & Faber. Sherrod, K. B., Friedman, S., Crawley, S.,Drake, D., & Devieux, J. (1977). Maternal language to prelinguistic infants: Syntactic aspects. Child Development, 48, 1662- 1665. Shute, B.,& Wheldall, K. (1989). Pitch alterations in British motherese: Some preliminary acoustic data. Journal of Child Language, 16, 503-512. Snow, C. E. (1977). The development of conversation between mothers and babies. Journal of Child Lunguage, 4 . 1-22. Stern, D. N., Spieker, S., Barnett, R. K., & MacKain, K. (1983). The prosody of maternal speech: Infant age and context related changes. Journal of Child Lunguage, 10. 1-15. Stern, D. N., Spieker, S . , & MacKain, K. (1982). Intonation contours as signals in maternal speech to prelinguistic infants. Developmental Psychology, 18, 727-735. Stevens, K. N., & Keyser, S . J. (1989). Primary features and their enhancement in consonants. Language, 65, 81-106.

34


Stevenson, M. B., VerHoeve, J. N., Roach, M. A,, & Levitt, L. A. (1986). The beginning of conversation: Early patterns of mother-infant vocal responsiveness. Infant Behavior and Development. 9, 423-440. Stross, B. (1977). Tzeltal conceptions of power. In R. D. Fogelson & R. N. Adam (Us.), The anthropology of power: Ethnographic studies from Asia, Oceania, and the New World ( pp. 271285). New York: Academic Press. Sundberg, J. (1973). The source spectrum in professional singing. Folia Phoniatrica, 25, 71 -90. Thorpe, L. A., & Trehub, S. E. (1989). Duration illusion and auditory grouping in infancy. Developmental Psychology, 25. 122- 127. Thorpe, L. A,, Trehub, S. E., Morrongiello, B. A., & Bull, D. (1988). Perceptual grouping by infants and preschool children. Developmental Psychology, 24, 484-491. Toda, S . , Fogel, A., & Kawai, M. (1990). Maternal speech to three-month-old infants in the United States and Japan. Journal of Child Language, 17, 279-294. Trainor, L. J. (1991). The origins of musical pattern perception: A comparison of infants’ and adults’ processing of melody. Unpublished doctoral dissertation, University of Toronto, Toronto. Trainor, L. J., & Trehub, S. E. (1992a). A comparison of infants’ and adults’ sensitivity to Western tonal structure. Journal of Experimental Psychology: Human Perception and Pe@mnance. 18. 394-402. Trainor, L. J., & Trehub, S. E. (1992b). The development of referential meaning in music. Music Perception, 9, 455-470. Trehub, S. E. (1976). The discrimination of foreign speech contrasts by infants and adults. Child Development, 47, 466-472. Trehub, S. E. (1985). Auditory pattern perception in infancy. In S. E. Trehub & B. A. Schneider (Eds.), Audirory development in infancy (pp. 183- 195). New York: Plenum. Trehub, S. E. (1989). Infants’ perception of musical sequences: Implications for language acquisition. Journal of Speech-Language Pathology and Audiology. 13. 3- I 1. Trehub, S. E. (1990). The perception of musical patterns by human infants: The provision of similar patterns by their parents. In M. A. Berkley & W. C. Stebbins (Eds.), Comparative perception. Vol. 1: Basic mechanisms (pp. 429-459). New York: Wiley. Trehub, S. E., Bull, D.,I% Thorpe, L. A. (1984). Infants’ perception of melodies: The role of melodic contour. Child Development, 55, 821-830. Trehub, S. E., Endman, M., & Thorpe, L. A. (1990). Infants’ perception of timbre: Classification of complex tones by spectral structure. Journal of Experimental Child Psychology, 49, 300-3 13. Trehub, S. E., & Thorpe, L. A. (1989). Infants’ perception of rhythm. Categorization of auditory sequences by temporal structure. Canadian Journal of Psychology. 43, 217-229. Trehub, S . E., Thorpe, L. A., & Morrongiello, B. A. (1985). Infants’ perception of melodies: Changes in a single tone. Infant Behavior and Development, 8, 213-223. Trehub, S. E., Thorpe, L. A., & Morrongiello, B. A. (1987). Organizational processes in infants’ perception of auditory patterns. Child Development. 58. 741-749. Trehub, S. E., Thorpe,L. A., & Trainor, L. J. (1990). Infants’ perception of good and bad melodies. Psychomusicology, 9, 5- 19. Trehub, S . E., & Trainor, L. J. (1990). Rules for listening in infancy. In J. Enns (Ed.), The development of attention: Research and theory (pp. 87- 119). Amsterdam: Elsevier. Trehub, S. E., & Trainor, L. J. (1993). Listening strategies in infancy: The roots of music and language development. In S. McAdams & E. Bigand (Eds.), Thinking in sound: Cognitive perspectives on human audition ( pp. 278-327). London: Oxford University Press. Trehub, S. E., & Unyk, A. M. (1992). Music prototypes in developmental perspective. Psychomusicology. 10, 31-45. Trehub, S. E., Unyk, A. M.,& Trainor, L. J. (in press-a). Adults identify infant-directed music across cultures. Infant Behavior and Development.


35

Trehub, S. E., Unyk, A. M., & Trainor, L. J. (in press-b). Maternal singing in cross-cultural perspective. Infant Behavior and Development. Unyk, A. M., Trehub, S. E., Trainor, L. J., & Schellenberg, E. G . (1992). Lullabies and simplicity: A cross-cultural perspective. Psychofogy of Music, 20, 15-28. Walker, R. (1990). Musical beliefs: Psychoacoustic, mythical, and educational perspectives. New York: Teacher’s College Press. Watkins, A. J. (1985). Scale, key, and contour in the discrimination of tuned and mistuned approximations to melody. Perception & Psychophysirs, 37, 275-288. Watson-Gegeo, K. A., & Gegeo, D. W. (1986). Calling out and repeating routines in Kwara’ae children’s language socialization. In B. B. Schieffelin & E. Ochs (Eds.), Language socialization across cultures ( pp. 17-50). Cambridge: Cambridge University Press.. Weeks, T. (1971). Speech registers in young children. Child Development, 42, I 1 19-1 131, Werker, J. F. (1991). The ontogeny of speech perception. In 1. G. Mattingly & M. Studdert-Kennedy (Eds.), Modularity and rhe moror rheory of speech perception (pp. 91-1 10). Hillxiale, NJ: Erlbaum. Werker, J. F., & Lalonde, C. E. (1988). Cross-language speech perception: Initial capabilities and developmental change. Developmental Psychology. 24. 672-683. Werker, J. F., & McLeod, P. 1. (1989). Infant preference for both male and female infant-directed talk: A developmental study of attentional and affective responsiveness. Canadian Journal of Psychology, 43. 230-246. Werker, J. F., & Tees, R. C. (1984). Cross-language speech perception: Evidence for perceptual reorganization during the first year of life. Infant Behavior and Developrnenr, 7 , 49-63. Williams, C. E., & Stevens, K. N. (1972). Emotions and speech: Some acoustic correlates. Journal of the Acoustical Society of America, 52, 1238-1250. Wolff, P. H. (1963). Observations on the early development of smiling. In B. M. Foss (Ed.), Determinants of infanr behavior / I ( pp. 1 13- 134). London: Methuen.


EFFECTS OF FEEDING METHOD ON INFANT TEMPERAMENT

John Worobey DEPARTMENT OF NUTRITIONAL SCIENCES COOK COLLEGE RUTGERS-THE STATE UNIVERSITY NEW BRUNSWICK, NEW JERSEY 08903

I. INTRODUCTION 11. TEMPERAMENT IN EARLY INFANCY

A. TEMPERAMENT IN THE NEWBORN PERIOD B. RESPONSIVITY, IRRITABILITY, AND ACTIVITY 111. DIET AND EARLY BEHAVIOR A. EFFECTS O F UNDERNUTRITION B. EFFECTS O F SPECIFIC NUTRIENT DEFICIENCIES IV. THE SAMPLE CASE O F FEEDING REGIMEN A. FEEDING METHOD AND RESPONSIVITY B. FEEDING METHOD AND IRRITABILITY

V. FEEDING METHOD AND ACTIVITY A. METHOD B. RESULTS VI. DISCUSSION VII. RECOMMENDATIONS FOR FURTHER RESEARCH REFERENCES

I. Introduction Although numerous research reports, review articles, books, and edited volumes devoted specifically to temperament appeared in the 1980s (e.g., Buss & Plomin, 1984; Goldsmith et a]., 1987; Kohnstamm, Bates, & Rothbart, 1989), relatively little attention was paid to the immediate postpartum months. That is, few investigators seemed to acknowledge temperament as being expressed before 37 ADVANCES IN CHILD DEVELOPMENT AND BEHAVIOR. VOL. 24

Copyright 0 IW3 by Academic Press. Inc All nghts of reproduction in any form rtserved.

38

John Worobey

the infant is several months old. For example, Thomas, Chess, Birth, Hertzig, and Korn (1963) began their New York Longitudinal Study of behavioral style with infants nearly 3 months of age and older. Carey (1970) later used the Thomas et al. framework in developing a temperament questionnaire for infants of 4 to 8 months, leaving Bohlin, Hagekull, and Lindhagen (198 1) to reformulate the Carey instrument for use with infants as young as 3 months. Similarly, the youngest subjects in Rothbart’s (1981) normative sample were 36 months of age, and Bates, Freeland, and Lounsbury (1979) worked with 4- to 6-month-olds. From one perspective, this 3-months-plus criterion is certainly understandable, as much can be said for plasticity in the first months of infancy (Emde, 1978; Horowitz, Sullivan, & Linn, 1978). Development in the immediate postpartum weeks is phenomenal, and a caregiver who must acquaint herself or himself with the individuality of the newborn is likely aided by an infant whose changing, adaptive nature allows for maturation in synchrony with the caregiver’s own increasing skills in “reading” her or his offspring. Unsurprisingly, then, the bulk of research on temperament has been conducted with infants who presumably have reached a level of stability in behavioral style that is best displayed at 3 months or later. From another perspective, however, this reluctance to study very young infants is somewhat curious: Most investigators of temperament have agreed that regardless of theory or definitions, the construct has an implicit biological basis (see McCall’s comments in Goldsmith et al., 1987). Admittedly, inborn traits need not be observable at birth, particularly those that are closer to the socioemotional realm than to the cognitive or physical domain. Yet the consensus that temperament dimensions are evident and measurable by 3 or 4 months implies that antecedents of behavioral style should appear sooner. Also surprising, indeed almost startling, is the lack of attention.paid to the possible mediating role of feeding in the expression of temperament in the young infant. Whether breast-fed or formula-fed, normal newborns receive six to eight feedings per day (Pipes, 1989), accounting for nearly half of their awake time. Of greater importance, however, is the fact that their initial diet consists solely of human milk, a specialized formula, or some combination of the two. Although modem infant formulas are a satisfactory substitute for breast milk from a nutritional standpoint (Hansen, Cook, Cordano, & Miguel, 1988), remarkably little is known about the behavioral implications of one type of feeding versus the other. To be sure, a few researchers have examined newborn temperament (e.g., Fish, Stifter, & Belsky, 1991; van den Boom, 1989), and some have even addressed the role of feeding (Ban, Kramer, Pless, Boisjoly, & Leduc, 1989); however, these researchers have focused on negative emotionality or fussiness, and do not consider the other aspects of temperament included within most models of the construct (Goldsmith et al., 1987). The relative dearth of even

Feeding Method and Temperument

39

descriptive information on early temperament and feeding has therefore led me to explore linkages from the newborn period through the first months of infancy. In this article I summarize my work identifying the earliest manifestations of behavioral style in human infants, outline the procedures that I have employed to assess temperament, and argue for the importance of certain dimensions relevant to later development. My criteria for importance are based on three considerations. First is the objective measurability of the dimensions, through the convergence of descriptions, assessments, and observations as appropriate. Second is the distinctiveness of the dimensions, that each involves a more or less independent aspect of temperament. Finally, the stability of the dimensions is of concern, that despite developmental change the components have a traceable link to later temperament, behavior, or other meaningful outcome. After this summary, the relationship between early nutrition and temperament is considered, and the issue of early feeding regimen and its impact on the expression of temperament is analyzed.

11. Temperament in Early Infancy My first efforts in studying infant temperament were aimed at filling in what I viewed as a significant gap in the literature on early behavioral style. Specifically, the advances made in delineating the individuality of the newborn were not incorporated into the burgeoning field of temperament, and the investigators of newborn behavior were not tracking their subjects into later infancy, at least in behavioral-stylistic terms. From the 1960s onward, a number of studies have revealed sizable individual differences within normal samples of full-term newborns (see Stratton, 1982, for a review). Advances in assessment have been particularly marked, with the Neonatal Behavioral Assessment Scale (Brazelton, 1973) inspiring a generation of researchers to consider the newborn as an intriguing organism worthy of study in its own right (see Worobey, 1990, for a review). A. TEMPERAMENT IN THE NEWBORN PERIOD

Using the Brazelton scale as a starting point, I undertook a short-term longitudinal investigation aimed at identifying the rudiments of early temperament. Newborn assessments were administered 2 to 3 days postpartum, while the infants were still hospitalized, and some 4 weeks later, when the infants were at home. The Infant Behavior Questionnaire (Rothbart, 1981) was sent to the mothers at 2 weeks, with instructions to answer as many questions as they could, given their relatively brief acquaintance with their newborns. The questionnaire was also completed at 2 months and 1 year. Home visits at I month included a

40

John Worobey

naturalistic observation of mother-infant interaction and an interview of the mothers using the protocol from the New York Longitudinal Study (Thomas & Chess, 1977). The measures taken during the newborn period reflected modest convergence, as orientation on the Brazelton scale correlated with approach and positive mood from the Thomas and Chess interview, which in turn was negatively associated with Brazelton scores for irritability, excitement, and other components of range of state (Worobey, 1986). Distractibility ratings culled from the interviews were correlated with soothability on the Rothbart questionnaire, and positive mood was negatively associated with the distress-to-limitations dimension. In the most straightforward example of interinstrument agreement, ratings of activity level converged for the temperament questionnaire and interview. Two conclusions can be drawn from these correlational analyses. First, maternal reports as gleaned from questionnaires and interviews were reliable across formats, and showed consistent though limited relations with the examinerassessed newborn behaviors. Second, not all dimensions of temperament routinely explored with older samples may have merit for studying newborns. Only six of the nine New York Longitudinal Study dimensions showed any significant associations with the Brazelton or Rothbart dimensions, and only four of the six Rothbart dimensions were correlated with the Brazelton or Thomas and Chess factors. Except for activity, the dimensions that did relate were either positive (e.g., orientation, approach, positive mood) or negative (e.g., irritability, distress, negative mood) in their emotional tone. Given this pattern of what could be termed positive and negative emotionality, the dimensions could reasonably be reduced a priori to three areas of temperament, namely, responsivity (Worobey & Belsky, 1982), irritability (Worobey, Laub, & Schilmoeller, 1983), and activity (Worobey & Anderson-Goetz, 1985). For that matter, a scanning of the 1980s literature revealed that other researchers were reaching the same conclusion. Buss and Plomin (1984), for example, removed impulsivity from their earlier framework and settled on three dimensions, namely, sociability, emotionality, and activity. Also, in summarizing the major effects of temperament on the social system during childhood, Bates (1989) chose to focus on sociability, “difficultness,” and activity. Finally, in a study of questionnaire convergence, Goldsmith and his colleagues identified approach-sociability, negative emotionality, and activity level as representative of the most commonly assessed temperament traits (Goldsmith, Rieser-Danner, & Briggs, 1991). B. RESPONSIVITY, IRRITABILITY, AND ACTIVITY

For these younger subjects, then, an index of infant responsivity was created by adding the frequency scores for orienting and smiling/laughter from the

Feeding Method and Temperament

41

Rothbart questionnaire. An index of irritability was formed by combining fear and distress-to-limitations. High values for soothability were consistently and significantly correlated with low values for distress-to-limitations; therefore, scores for responsivity, irritability, and activity level were consistent during early infancy, with correlation coefficients in excess of .60 across the first two intervals able at 2 weeks, 2 months, and 1 year, test-retest stability was examined. The scores for responsivity, irritability, and activity level were consistent during early infancy, with correlation coefficients in excess of .60 across the fist two intervals for all three summary variables (Worobey & Blajda, 1989). Though the correlation coefficients were somewhat less from 2 months to 1 year (.40-.50), they still achieved respectable significance. So, despite an increase over the course of the year in the dimensional scores, reflecting greater responsivity, imtability, and activity, a stability in the maternal ratings of temperament was clearly evidenced. The inference that these ratings were based on objective reality was reinforced by their correspondence to the naturalistic observations made at I month. For example, questionnaire-basedirritability was positively correlated with observed crying and negatively associated with alertness. As a further test of convergence, a subgroup of these infants were drawn using the criteria of little versus marked change from the first to second Brazelton administrations. Maternal ratings of responsivity and activity level were higher for the newborns who improved in their Brazelton scores, and paralleled the observed behaviors summarized as attentiveness and movement (Worobey, 1990). And as earlier found with the correlational analysis, the irritability score as derived from the maternal ratings were less for the newborns who exhibited change over 1 month, and was matched by the lower frequency of crying that was independently witnessed during the observation. These results appear to extend the aforementioned temperament literature downward, as responsivity, imtability, and activity level were clearly distinguishable as early as 2 weeks, in forms not substantially different from those dimensions previously identified in older samples (Bates, 1989; Buss & Plomin, 1984; Goldsmith et al., 1991). Responsivity, also called orientation, attentiveness, and sociability, is usually acknowledged as slower to surface than dimensions of temperament such as activity level (Brazelton, Kozlowski, & Main, 1974; Schaffer, 1971; Worobey, 1989); however, its importance to the developing mother-infant interactive system is beyond question (Kaye, 1982; Robson & Moss, 1970). The infant’s irritability, as manifested by fussing and crying (Brazelton, 1962; van den Boom, 1989), is a core ingredient of what has been termed dzjjicultness (Bates et al., 1979; Hubert & Wachs, 1985). Finally, activity level is included in vinually every theory of temperament and instrument that is purported to measure this construct. Although it increases with age (Komer, Zeanah, Linden, Berkowitz, Kraemer, & Agras, 1985; Rothbart, 1986) and it may be contextually influenced (Eaton & Dureski, 1986; Worobey &

42

John Worobey

Anderson-Goetz, 1985), it is generally considered to be the dimension that shows the greatest continuity from infancy through childhood (Goldsmith et al., 1987; Hubert, Wachs, Peters-Martin, & Gandour, 1982).

111. Diet and Early Behavior Echoing the attention paid by developmentalists to temperament has been the interest displayed by scientists in examining the connections between nutrients and brain function. Investigators in the field of nutritional sciences have summarized work on diet and behavior (e.g., Anderson, 1990; Levine, 1990), with psychologists and neuroscientists (e.g., Capaldi & Powley, 1990; Stricker, 1990) compiling a vast array of data and information on appetite, taste, food/fluid intake, and behavioral responses. As this increasing literature on diet and behavior has proved to be extensive, my treatment of the topic focuses on applications to infancy and early childhood. A.

EFFECTS OF UNDERNUTRITION

In separate reviews, Barrett (1984), Goldman (1988), and Lozoff (1989) examined the link between dietary factors and developmental problems of children, primarily the impact of undernutrition on cognitive performance and socioemotional functioning. Scores on intelligence tests have been included as a primary dependent variable in virtually all reports of the effects of undernutrition (Levitsky & Strupp, 1985; Pollitt, 1988; Ricciuti, 1991), and these investigations have also provided incontrovertible evidence that even marginal undernutrition impairs health (e.g., Zeskind, Goff, & Huntington, 1984), emotional development (e.g., Latham, 1974), and social responsiveness (e.g., Chavez & Martinez, 1982). With respect to socioemotional behavior, however, the literature on undernutrition as it affects temperament is admittedly scarce. Nevertheless, a consistent picture emerges from countries where malnourishment is endemic. When undernutrition occurred in infancy, poorer attention and social interaction were shown among children in Barbados (Galler, Ramsey, Solimano, Lowell, & Mason, 1983); passive, quiet, withdrawn, and distractible child behavior was shown in a Jamaican sample (Richardson, Birch, Grabie, & Yoder, 1972); social with-. drawal and a lack of impulse control were shown by South African children (Stoch, Smythe, Moodie, & Bradshaw, 1982); and lowered affect, social responsivity, and persistence, with irregular activity level, characterized children from Guatemala (Barrett & Frank, 1987). Indeed, chronic fetal undernutrition has adverse effects, as growth-retarded newborns have displayed lessened social


43

responsivity, motoric activity, consolability, and orienting to items on the Brazelton scale (e.g., Lester, Garcia-Coll, Valcarcel, Hoffman, & Brazelton, 1986). B. EFFECTS OF SPECIHC NUTRIENT DEFICIENCIES

In the absence of generalized undernutrition, specific nutrient deficiencies have also been shown to impair development. Iron deficiency, the most common nutritional disorder in the world (Dallman & Siimes, 1979), is most prevalent from 6 to 24 months, the period coinciding with the latter part of the brain’s growth spurt. Lozoff (1988, 1989) has summarized its behavioral effects, and notes that fearfulness, irritability, lack of persistence, and shortened attention span, for example, have been implicated in studies comparing iron-deficient infants with control infants. Conversely, adequate stores of iron and calcium have been found to ward off the effects of lead ingestion (Mahaffey, 1990; Rosen, 1989), of no small consequence given the negative outcomes of lead poisoning on child behavior (Bellinger, Leviton, Waternaux, Needleman, & Rabinowitz, 1987; Fein, Schwartz, Jacobson, & Jacobson, 1983). The effects of zinc on socioemotional development are more conjectural, as growth velocity has been increased with supplementation (Walravens & Hambidge, 1976), implying that a deficiency could affect appetite in infants, but little work has been done on behavioral alterations in human subjects. Finally, some work with mothers assessed as having marginal vitamin B, status indicates that maternal as well as neonatal behavior may be affected, as such mothers are less effective in soothing their newborns, who also seem to be more easily upset (McCullough et al., 1990).

IV. The Sample Case of Feeding Regimen The topic of infant feeding has captured widespread interest through much of this country. The immunological benefits of breast-feeding, in particular, are well documented, with near-complete agreement as to its superiority for most infants. Although the nutritional value of breast-feeding is undeniable (Pipes, 1989), associations between method of feeding and infant behavior have been surprisingly inconsistent. For example, Bell (1966) found that breast-fed newborns display more arousal following an anthropometric examination than bottle-fed newborns. Other investigators in the early 1970s, notably Bernal and Richards (1970), found that breast-fed newborns were generally more irritable than bottlefed newborns over the first 10 days of life. Nevertheless, some past work showed no differences in irritability based on feeding method (Crockenberg & Smith, 1982), and other past work showed that bottle-fed infants cry more than breastfed infants (Simmons, Ottinger, & Haugk, 1967).

44

John Worobey

Recent research efforts have only added to this controversy. For example, investigators who have employed the standardized Brazelton scale have obtained different results regarding the effects of feeding method on early behaviors. DiPietro, Larson, and Porges (1987) found that breast-fed newborns were significantly more irritable but displayed better range of state than formula-fed newborns, and were not different in terms of motor, orientation, or the remaining Brazelton clusters. Using a rigorous manipulation of tryptophan levels in infant formula, Steinberg, Goff, Birch, Picciano, and Hatch (1989) directly influenced latency to sleep in infants over the first 2 months of life; however, they found no differences when breast-fed infants were compared with the experimental groups in terms of orientation, irritability, and response decrement clusters derived from the Brazelton. Despite these equivocal results, the preceding studies illustrate the greater variety of behaviors that researchers of early feeding and its implications may now explore. No longer limited to intelligence quotients or motor milestones (e.g., Hoefer & Hardy, 1929; but see Lucas, Morley, Cole, Lister, & LeesonPayne, 1992), the possibility that “breast versus bottle” may influence the infant’s expression of behavior, from fussiness (Barr et al., 1989) to stress management (Carter, 1988), has led to a renewed interest in studying the early feeding context. As should now be obvious to the reader, behaviors such as orientation and imtability are key dimensions of what has been referred to as temperament throughout this article. Of the numerous traits that have been conceptualized as constituting early temperament, the heretofore discussed dimensions of responsivity, irritability, and activity are well suited for analysis by feeding method. In the following sections I cover the methods and results of three investigations in which the effect of feeding on infant development was explored. The first two studies are secondary analyses of data collected for other purposes; the third study is a prospective investigation of feeding method as an independent variable in infancy research. A. FEEDING METHOD AND RESPONSIVITY

Although some controversy surrounds the emotional benefits of breast-feeding for the mother (Holt & Wolkind, 1983; Stein, Cooper, Day, & Bond, 1987), the evidence definitely indicates that psychological factors distinguish the woman who breast-feeds from the one who does not (Newton, 1971; Shand, 1981). For example, more positive attitudes toward infant and family have been shown by breast-feeding mothers (Bernal & Richards, 1970; Wiesenfeld, Malatesta, Whitman, Grandose, & Uili, 1985). A long-accepted conclusion is that the mother’s attitude will, in large measure, determine whether or not she breast-feeds and, subsequently, her degree of success (Newton, 1955; Switzky, Vietze, & Switzky, 1979). The mother-infant relationship is inherently developmental (Worobey,


45

1989); therefore, the mother’s attitudes and feelings toward breast-feeding may be assumed to affect not only her success but also the interaction during feeding episodes. If breast-feeding promotes positive interactions between mother and infant, these may have further impact on the mother’s attitudes toward breastfeeding. The affectionate behaviors and extra contact experienced while breast-feeding seem to heighten the mother-infant relationship (Blumen, 1980). For example, in the Wiesenfeld et al. (1985) study, breast-feeding mothers were found to be more responsive during feeding periods, and they reported a greater desire to pick up their infants. Nursing mothers touch, smile, and vocalize more to their infants during feeding than mothers using formula, and respond more quickly to their infants’ cries (Bernal & Richards, 1970). Walton and Vallelunga (1989) reported that for breast-feeding mothers, maternal behaviors tended to follow infant behaviors, but for bottle-feeding mothers, maternal behaviors tended to follow other maternal behaviors. In contrast, the bottle-fed infants were more likely to respond to maternal behaviors, and the breast-fed infants were more likely to respond to their own previous behaviors. Brody (1976) suggested that a mother’s actions during feeding serve as a model of her overall behavior toward her infant. Although the feeding situation has been argued by some to reflect the quality of the mother’s interaction with her child (Spietz, 1978), surprisingly little research has been conducted on the functional significance of early feeding patterns. The work described next provided a test of the assumption that infant feeding patterns are an important contributor to the developing mother-infant relationship (see Kuzela, Stifter, & Worobey, 1990, for the details of this investigation). Instead of examining interactions during a feeding episode, however, we examined behavior during play so as to determine the generalized effects that might be attributed to the feeding method.

1. Subjects and Procedure ’henty-seven 8-month-old infants and their mothers were recruited for a study of individual differences in infant behavior. All infants were born full-term and were reported by their mothers as being in good health. Mothers and infants were observed in a free-play session in a laboratory on campus that resembled a home setting. Mothers were told to play with their infants on the floor as they normally would at home. An infant seat and basket of age-appropriate toys were available for the mothers’ use, although the mothers were not instructed to use them. The free-play sessions were videotaped and later coded using a time-sampling technique. Infant behaviors such as fussing, reaching, gazing at mother, and contingently vocalizing were coded, in a fashion similar to existing schemes for coding mother-infant interactions (e.g., Belsky, Taylor, & Rovine, 1984). Reliability as indexed by percentage agreement was greater than 80% for all behaviors.

46

John Worobey

2. Results The sample was divided by feeding method. It included 16 bottle-fed infants (4 who had been bottle-fed from birth and 12 who were being bottle-fed at the time of the study but who had previously been breast-fed) and 11 infants who had always been breast-fed. Maternal behaviors were reduced to the categories of play, touch, vocalization, and positive affect. The breast-feeding mothers touched their infants more frequently than did the bottle-feeding mothers (5.5 versus 4.2), but no other main effects emerged for maternal behavior; however, mothers of currently breast-fed males (15.2) and bottle-fed females (15.5) displayed more instances of positive affect than mothers of currently breast-fed females (10.8) and bottle-fed males (12.0). The pattern for the infants’ smiling and reaching toward mother echoed these results, as the breast-fed males tended to score higher on these behaviors than the bottle-fed males, and the reverse was true for the females. Furthermore, this interaction was statistically significant for contingent vocalizations, as breast-fed males (2.0) and bottle-fed females (4.1) were markedly higher in magnitude than bottle-fed males (0.9) and breast-fed females (1.2). In general, little fussing was exhibited by the infants, but the currently breast-fed infants were more imtable than the currently bottle-fed infants. Certainly a number of other factors, not the least of which are the predisposing characteristics of mothers who elect to breast-feed in the first place, could explain these differences in behavior. Nevertheless, the results of this study indicate that feeding method may influence relationships in situations other than feeding. That breast-feeding mothers touched their infants more in a play setting than did bottle-feeding mothers, many of whom had breast-fed earlier as well, raises the possibility that interactions practiced during feeding may generalize to a contemporaneous nonfeeding context. The Sex x Feeding interaction was unanticipated, and suggests that breast-fed males may share a different relationship with their mothers than breast-fed females (Heinstein, 1963). This result does not imply that breast-feeding makes for better dyadic relationships, however, because the bottle-feeding pairs of mothers and female infants exhibited the most affection and vocalizations. B . FEEDING METHOD AND IRRITABILITY

As indicated earlier, some investigators have found that breast-fed newborns are more imtable (DiPietro et al., 1987); some have found the opposite (Simmons et al., 1967); and others have found no differences in irritability (Crockenberg & Smith, 1982). These discrepancies may reflect the fact that, for the most part, irritability was assessed in these investigations in an indirect manner. That is, the Occurrence of fussing or crying was noted within the context of a normal


47

setting, for example, as displayed during mother-infant interaction, or by a newborn when tested with the Brazelton scale. With the exception of Bernal and Richards (1970), no special efforts were made to observe crying as a direct result of an aversive procedure. The response of the newborn to stress may be a better predictor of later functioning than more typical behavior (Brazelton, 1973); therefore, an analysis of reactivity, or elicited irritability, in early infancy would seem appropriate for determining the impact of feeding method. To this end, the procedures we have outlined for assessing the stability of individual differences have proved useful in measuring the relative influence of breast-feeding versus bottle-feeding on early infant behavior (see Worobey & Lewis, 1992, for details of this investigation). Specifically, infant reactivity to a routine heelstick and first inoculation was examined, with type of feeding entered into the analysis of the stress response to these procedures.

I . Subjects and Procedure Thirty-three infants who had participated in a short-term longitudinal study concerned with the stability of the stress response (Worobey & Lewis, 1989) were observed at 2 days and approximately 2 months of age. All infants were born full-term, with Apgar scores of 8 or above at 1 minute postpartum. At 2 days, and following their first morning feeding, the infants’ responses to a blood-draw for screening purposes (e.g., phenylketonuria) were videorecorded in a quiet room adjoining the newborn nursery. At 2 months, when the same infants were brought to their pediatrician’s office for their first inoculation (i.e., diphtheria-pertussis-tetanus), their responses to the injection were also videorecorded. As a relevant aside, both of these procedures are widely administered, brief, and standardized and are viewed as painful (Anand & Hickey, 1987; Izard, Hembree, & Huebner, 1987). Following data collection for the entire sample, coding of infant fussing and facial grimacing took place. Observers were kept blind to the infants’ feeding regimen and, for the 2-month coding, to each infant’s reactions as a newborn. Interrater reliability was in excess of 85 and 75% for fussing and facial responses, respectively.

2. Results To examine elicited irritability, that is, reactivity to the heelstick and inoculation, intensity scores were computed for the infants’ arousal states 5 seconds prior to the perturbation and for the first 5-second interval following the perturbation. In addition, the latency to quiet following each stimulus was noted for both procedures, in real time. No significant correlations were found between any of these reactivity measures and delivery characteristics (i.e., Apgar scores, birth weight, maternal age). Similarly, no significant differences were found for sex, birth order, or type of delivery. A breast-fed sample comprised 23 infants (17

48

John Worobey

who were exclusively nursed and 6 who received one formula supplement per day), and a bottle-fed group comprised 10 infants who were reared on formula from birth. To quantify the magnitude of the infants’ reactivity to the perturbations, change scores were determined for the intensity variables by subtracting the appropriate baseline levels from the infants’ responses following the heelstick and inoculation. The bottle-fed babies reacted the same at 2 days and 2 months (2.80 versus 2.70, respectively); but the breast-fed infants, who did not differ significantly from the bottle-fed babies at 2 days, showed markedly greater responsiveness at 2 months (3.65) than at 2 days (2.19). Although the change scores between groups were not substantially different at 2 months, the change scores within the breast-fed sample had significantly increased over time. The breast-fed newborns took slightly less time to cease fretting after the heelstick than did the bottle-fed newborns. This pattern was reversed at 2 months; however, neither difference between groups was significant. In light of their different responses to the inoculation, a somewhat surprising finding was that both groups quickly gained an equivalent level of composure, quieting in less than 1 minute. At 2 days the newborns’ response to stress seemed to have superseded the influence of feeding, and their coping skills were not strongly differentiated; however, the stress response at 2 months may possibly have been affected by feeding status, as the level of change in response to the needle was comparably higher for the breast-fed infants. Although the bottle-fed infants responded virtually the same at 2 days and 2 months in terms of their quieting after the aversive procedures, the coping responses were unaffected by feeding status at either point in time.

V. Feeding Method and Activity The temperament dimension of activity level is a particularly intriguing candidate for an analysis by feeding method. Activity level is a salient aspect of infant behavior to new parents, and is named by virtually all temperament theorists as a key dimension of individual differences (Goldsmith et al., 1987). Activity may be measured as early as the newborn period (Worobey & Blajda, 1989), and is the most consistent of the temperament traits in demonstrating stability (Hubert et al., 1982). Relevant to feeding, inactivity has been linked to infant obesity (Mack & Kleinhenz, 1974; Rose & Mayer, 1968), although early activity has been less than directly predictive of weight gain (Berkowitz, Agras, Korner, Kraemer, & Zeanah, 1985; Carey, 1985). Finally, activity level may be objectively assessed with actometers, to supplement parental ratings or trained observers’ estimates (Bell, 1968; Eaton & Dureski, 1986).


49

The present study was undertaken to investigate the relative influence of breast-feeding versus bottle-feeding on infant temperament in general, and motoric activity in particular. As prospective comparisons have rarely been made for infants beyond a few weeks of life (Wright & Crow, 1982), the results of this study may serve to clarify the inconsistent findings of previous work. A. METHOD

I . Subjects The participants in this study were 46 mothers and their infants. The mean age of the infants was 38 months (SD = 1 month); 30 were male and 16 female. All were reported to be in excellent health. Some mothers were recruited through notices posted at the university and nearby medical school; the rest were contacted by mail following the published birth announcements, based on their proximity to the campus. This low-risk sample was extremely homogeneous; all the mothers were married, non-Hispanic white college graduates, and all had husbands who assisted in childbirth.

2. Instruments a. Questionnaire. To obtain ratings of temperament, the Rothbart (1978) Infant Behavior Questionnaire was completed by the mothers. The Rothbart questionnaire consists of 94 items rated on a 7-point scale (never to always). Items concern the frequency of behaviors observed during the previous week with regard to the following categories: activity level-gross motor activity, including squirming, arm,and leg movements; smiling and laughter-smiling or laughter in any situation; fear-distress and latency to approach a sudden or novel stimulus; distress-to-limitations-upset during caretaking or when denied access to a goal object; duration of orienting-vocalizing, looking at, or interacting with an object for an extended period; and soothability-reduction of fussing or crying in response to soothing (see Rothbart, 1981, for a detailed description of the questionnaire). b . Actometer. Although many types of actometers have been used with children (Tryon, 1984), most are too heavy to be employed with infants. If modified appropriately for infants, activity level is generalized from the output of one limb. Yet if multiple actometers are used (Mack & Kleinhenz, 1974), activity must then be determined by summing or averaging the independent totals, which raises the issue of interinstrument reliability (Eaton & Dureski, 1986). In light of these considerations, a specially designed actometer apparatus was used in this study to measure motoric activity (McDonnell & Richards, 1984). This custombuilt device consists of four movement sensors that attack to each of the infant’s four limbs via Velcro bracelets. Extremely small in size and weight (19 mm, 7 g),

50

John Worobey

the sensors are physically independent from each other, but are connected by three-conductor flat cable to a common power source and electronic interface with an Apple IIe microcomputer. The frequency of movement by each limb is recorded at a rate of 10 times per second, and synchronous movements by all combinations of limbs taken two at a time (e.g., both arms, left armlright leg) are automatically stored in the microcomputer (see McDonnell, Corkum, & Wilson, 1989, for a detailed description of the apparatus).

3. Procedure Mothers were asked to take part in a study aimed at validating a new actometer apparatus, and to bring their infant to a laboratory on campus at a time of their own choosing when the baby was midway between feedings and expected to be alert and responsive. After arriving at the laboratory, the infant was weighed, and the procedures were described to the mother. While the mother held her infant, the research assistant wrapped Velcro strips around the infant’s wrists and ankles. The mother then laid her infant on his or her back in a crib lined with a yellow bumper pad. After the actometer sensors were attached to the Velcro bracelets, a Fisher-Price musical mobile, mounted on the crib’s headboard, was swung over the supine infant. The microcomputer was next activated to begin recording limb movements. For the first 2 minutes, the mobile remained off so that a baseline activity level could be determined. For the next 2 minutes, the mobile was turned on, revolving slowly and playing Brahm’s Lullaby. This 2 minutes off, 2 minutes on sequence was then repeated. Following the 8-minute session, the research assistant detached the movement sensors and Velcro strips, and the infant was picked up by the mother. During the test session, mothers were requested to remain in the laboratory, but out of the range of the infant’s vision. They were given the opportunity to begin the Rothbart questionnaire, although most completed it after the session, returning the forms by mail within a week. Prior to departure, each mother was asked some questions about her education and family make-up, whether or not her infant had a mobile at home, and what her previous and current methods of feeding her infant were. B. RESULTS

Of the 46 infants, 30 were exclusively breast-fed through the time of the laboratory visit (no more than one water supplement per day), 3 were predominantly breast-fed (no more than one formula supplement per day), and 13 were formula-fed (for at least the preceding 2 months). For comparison purposes, the first two groups were combined, so the sample consisted of 33 breast-fed versus 13 bottle-fed infants.


51

TABLE I Infant Behavior Questionnaire Means (and Standard Deviations) by Feeding Group Dimension

Breast-fed

Bottle-fed

Activity level Smiling/laughter Fear Distress-to-limitations Duration of orienting Soothability

3.75 (0.83) 4.43 ( I . 17) 2.40 (0.70) 3.53 (0.59) 3.64 (1.25) 4.72 (0.61)

3.52 (0.69) 4.63 (1.25) 2.39 (0.86) 3.10 (0.83) 4.00 (1.23) 4.98 (0.55)

The means for the temperament dimensions appear in Table I, and are consistent with published reports for 3-month-olds (Eaton & Dureski, 1986; Rothbart, 1986). Feeding method and sex were entered into a general linear model procedure to predict the Rothbart scores, but no significant differences were obtained. Irrespective of feeding method, the literature on motoric activity in infancy suggests a number of patterns that could be explored in this sample, given the output that the actometer could produce. First, the frequency of arm movements should exceed the frequency of leg movements (Eaton & Dureski, 1986; McDonnell et al., 1989); second, activity level should be suppressed in response to the moving, music-playing mobile (McDonnell & Richards, 1984); and third, male infants should display more total activity than female infants (Eaton & Enns, 1986). Hence, the possible influence of feeding method on these expected patterns was examined. At a global level, total activity over the 8-minute testing session was computed by combining the frequencies for the four limbs. For the sample as a whole, the mean total frequency (1035) is consistent with other reports based on use of this actometer (McDonnell et al., 1989; McDonnell & Richards, 1984). Relevant to feeding, total activity was significantly higher for the breast-fed (M = 1,157, SD = 888) than the bottle-fed (M = 722, SD = 421) infants. Combined arm and leg movements for the 4-minute baseline versus the 4minute activated mobile conditions were computed, as were the separate totals of arm versus leg movements over the 8 minutes. Activity was significantly higher for the breast-fed infants (M = 656, SD = 555) than the bottle-fed infants (M = 364, SD = 240) during baseline, though not under the animate mobile condition (Ms = 501 and 357, SDs 404 and 229, respectively). In regard to the expected pattern, activity was higher during the baseline than the mobile condition, but only for the breast-fed group (t = 2.26, p < .05). Arm movements differentiated the two feeding groups, and were greater for the breast-fed infants (M = 679, SD = 592) than the bottle-fed infants (M = 354, SD = 244). Leg movements were

John Worobey

52

TABLE I1 Means (and Standard Deviations) of Frequencies of Limb Movements per Minute by Feeding Group Movement Left arm Right arm Left leg Right leg Anns in synchrony Legs in synchrony

Breast-fed

Bottle-fed

P

52.25 (44.69) 32.65 (33.86) 20.59 (15.44) 39.22 (42.49) 8.33 (10.91) 3.02 (5.04)

27.13 (23.88) 17.11 (14.79) 21.22 (28.26) 24.79 (21.17) 2.13 (2.25) 2.31 (3.18)

.05 .05 NS NS .01

NS

not significantly different (breast-fed M = 478, SD = 393; bottle-fed M = 368, = 285). Arm movements exceeded leg movements as expected, but again only for the breast-fed infants ( t = 2.45, p C .05). In Table 11, movement frequencies for the separate limbs are displayed in terms of counts per minute, along with the synchronous movements by the two arms and two legs. The results of an analysis of variance indicate that both arms were more active in the breast-fed versus the bottle-fed infants. Synchrony refers to the frequency of movements of two separate limbs occurring within 100 milliseconds of each other (McDonnell et al., 1989). As synchrony is dependent on the absolute level of activity and is limited by the limb with the lower frequency of movement, the finding that synchronous arm movements also favored the breast-fed group is not surprising. For reasons that are not entirely clear, previous studies of the behavioral impact of feeding method have seldom included motoric activity as a dependent variable. Rather, states of arousal, irritability, and vagal tone have been examined (Bell, 1966; DiPietro et al., 1987; Pinella, Birch, Steinberg, Picciano, & Goff,1990), with activity included only if it was contained in a larger behavioral or temperament assessment (Carey, 1985; Steinberg et al., 1989). Investigators who have used actometers to measure motoric activity have generally not included method of feeding as a mediating variable (Eaton & Dureski, 1986; McDonnell et al., 1989). Ironically, the investigators who have come closest in terms of assessing feeding and measuring motoric activity have focused only on the caloric intake of formula-fed infants (Mack & Kleinhenz, 1974; Rose & Mayer, 1968). Interestingly enough, in the Rose and Mayer study, the only entirely breast-fed infant in the sample of 31 was one of just two infants who had a low caloric intake, yet also had a high activity rating. The results of the present study suggest that feeding method can influence

SD


53

motoric activity. Whether considered separately or in combination, limb activity was generally higher in the breast-fed infants than in the bottle-fed infants. Expected patterns of activity were observed in this sample of infants, but were confined to the subset who were breast-fed.

VI. Discussion Although nursing and suckling may be viewed as the defining features of mammalian interactive behavior (Blass, 1990), relatively few researchers have investigated the influence of feeding method on infant development, despite a veritable explosion of research on infancy from the 1970s onward. In a welcome exception, DePietro et al. (1987) suggested that feeding method be considered an important variable in research on newborn behavior. Taken together, the results of the studies described in these pages support this goal, as objective measures of responsivity, irritability, and activity were shown to be related to feeding method in three samples of infants who ranged from a few days to several months old. As observed during a play episode, 8-month-old breast-fed male infants shared more stimulating interactions with their mothers. When inoculated during a well-baby visit, 2-month-old breast-fed infants showed a greater upset response. And as measured with an actometer, 3-month-old breast-fed infants displayed more motoric activity. The inevitable question arises: Why should this be the case? In their provocative investigation of feeding-based differences in behavior and heart rate patterns, DiPietro et al. (1987) outlined three possibilities as to why reactivity was higher in their breast-fed newborns. The studies summarized in this article involved older infants and a variety of research methods and behaviors, and therefore their hypotheses regarding differences in reactivity cannot explain all of the present findings. With some extrapolation, however, the arguments of DiPietro et al. do serve as a useful starting point for hypothesizing why such results may have been obtained. 1. The possibility exists that breast-fed infants are hungrier than bottle-fed infants. Although this is a tenable position for explaining neonatal differences, such an argument is less convincing here, because lactation would have been well established by 2 weeks (Hopkinson & Garza, 1988). For that matter, the breast-fed newborns were no more or less reactive to the heelstick than the bottlefed newborns at 2 days of age. Nevertheless, bottle-fed infants might conceivably receive more formula relative to the amount of milk that the breast-fed infants consume. In the actometer study in which weight was measured, the infants did not differ by feeding group (6767 and 6850 g for breast-fed versus bottle-fed, respectively), suggesting that hunger was not a problem for either

54

John Worobey

group. One cannot, however, dismiss the possibility of differential hunger, because breast-feeding mothers often report an insufficient milk supply as their reason for early weaning (Sjolin, Hofvander, & Hillervik, 1977). 2. Mother-infant interaction during feeding may differ by feeding method. By 2 months of age, any awkwardness surrounding breast-feeding should have dissipated, else nursing likely would have been discontinued. Breast-feeding mothers are arguably different from mothers who elect to bottle-feed (Newton, 1971; Wiesenfeld et al., 1985), and socioeconomic status is widely agreed to be a major confound in feeding studies because breast-feeding mothers are likely to be socially advantaged (Grossman, Fitzsimmons, Larsen-Alexander, Sachs, & Harter, 1990; Hendershot, 1984). The extreme homogeneity of all three samples (i.e., supportive husbands, highly educated, middle class) would, however, make sizable differences in interactive styles rather unlikely. Still, the interaction differences shown outside the feeding context, not to mention the mediating effect of infant sex, suggest that more research is needed to understand why the breast-fed infants displayed more irritability than those bottle-fed, and not the reverse. 3. The substances ingested by breast- and bottle-fed infants are different. Human milk and formula are different in composition, and even with no less than 20 nutrients added to modified cow milk or soy milk, any manufactured infant formula is at best only a simulation of human milk (Pipes, 1989). DiPietro et al. (1987) speculated that for breast-fed newborns, colostrum may contain a behaviorally activating agent and formula may exert a depressing effect on behavior. As a case in point, varying the concentrations of amino acids in formula has been shown to alter state organization in newborns (Steinberg et al., 1989; Yogman & Zeisel, 1983). Although no altering of infant formula was tested in these studies, and colostrum had long given way to mature milk, the differential patterns of activity and, to a lesser extent, reactivity are worth reexamining in light of this last explanation. The predicted patterns of activity were only partially supported in the actometer study. Higher activity levels were shown for the arms than for the legs, but only in the breast-fed group. Activity was suppressed in response to the moving mobile, but again only in the breast-fed group. The near equivalence of movement frequencies for the bottle-fed babies, by their arms and legs and during the baseline and mobile conditions, suggests a stability in responsiveness that warrants further investigation with samples that are exclusively fed infant formula. Likewise, the stability in change scores in response to the aversive perturbations shown by the formula-fed infants, in contrast to the heightened reactivity by the breast-fed infants after 2 months, suggests that feeding method be considered when studying individual differences in early stress and coping (Carter, 1988).


55

VII. Recommendations for Further Research The hypotheses outlined in the previous section are admittedly speculative. Because some support exists for all three of these positions, future research will determine which, if any, of these possibilities best explains the effects that were revealed in this series of studies. Hence, I shall direct my closing remarks toward a consideration of future research on early diet and infant temperament. First, with the notable exception of the literature on colic, almost all of the previous work that has been addressed to temperamental differences as a function of feeding regimen has dealt with infant behavior in the first days after birth, that is, the immediate lying-in period and up through 2 weeks. Assessing outcomes this early allows for the documentation of individual differences in newborns before the family and environment at-large can exert much influence, but the problems associated with establishing successful breast-feeding in the typical rnother-newborn dyad may cause a confound. Yet, restricting our observations to the newborn period ignores the role of mother-infant interactions over their first months together. The data I have described imply that factors in place at 2 days may still remain at work some 2 or more months later, and may not even show their effects until then. The study of long-term outcomes should therefore be encouraged. Second, investigations of the effects of feeding method on the expression of temperament seldom extend beyond an analysis of behaviors at one time. Because the observational tools are limited, a single sample of behavior has been the norm. This procedure is understandable, given that the state levels coded for newborns make little sense to observe in older infants, and the attachment behaviors that differentiate established mother-infant pairs are inappropriate for use with newborns. A strength of the procedures outlined here is their applicability to infants of varying ages. Extensive coding schemes that are now available for evaluating mother-infant interaction may be applied as early as 1 month (Belsky et al., 1984) and are suitable even for observing the feeding situation (Price, 1983). Inoculations that we observed at 2 months are repeated at 4, 6, and 18 months, depending on the infant’s immunization schedule. Motoric activity can be rated early (Worobey & Anderson-Goetz, 1985) and mechanically measured over the first year with age-appropriate actometer devices (Saudino & Eaton, 1991). Furthermore, the Rothbart (1981) questionnaire has been demonstrated to be applicable from the first weeks through 1 year of age (Rothbart, 1986; Worobey & Blajda, 1989). Hence, in looking for possible effects of feeding, we have no excuse to avoid longitudinal approaches. Third, investigators of infant feeding and behavioral style should also consider the reverse influence-how temperament affects food intake. Physiological psychologists have done exemplary work in documenting innate preferences for

56

John Worobey

sugar (e.g., Crook, 1978), acquired preference for salt (e.g., Harris & Booth, 1987), and the detection of substances such as garlic in mother’s milk (Menella & Beauchamp, 1991), suggesting that the baby brings certain predispositions to the feeding situation. Likewise, anecdotes abound regarding the variation in nursing characteristics of the rooming-in infant, from those who leisurely nurse and rest to those who vigorously root and suck (e.g., Barnes, Lethin, Jackson, & Shea, 1953). Indeed, the contribution of fussing to the infant’s own weaning is of practical concern to the mother who may be a tentative breast-feeder (Humenick & Van Steenkiste, 1983). Finally, psychologists and pediatricians would do well to collaborate with nutritionists and dietitians in designing investigations that blend behavioral methodologies with sound biochemistry (Greenwood, 1990). Studies of how specific nutrients alter the synthesis of neurotransmitters and subsequent behavioral states of human infants are rare (e.g., Yogman & Zeisel, 1983), yet they hold great promise. As touched on throughout this article, diet can influence early growth and functioning in many subtle and not so subtle ways. That diet can affect temperamental expression should now be evident; however, early feeding method should also be considered an important independent variable in the design and conduct of future studies of infant behavior in general. ACKNOWLEDGMENTS This project was supported by the Rutgers University Research Council and by Grant 14403 from the New Jersey Agricultural Experiment Station. The author thanks Paul M. McDonnell, John Cardoso, and Murray Linton, the designers of the actometer apparatus, for their technical help. Their work was funded by the Natural Sciences and Engineering Research Council of Canada. Special thanks go to Ken Karbowski, who built the actometer used in the research reported herein, and to Virginia M. Blajda and John K. Muller, who assisted in data collection. Finally, the participation of the mothers and infants is gratefully acknowledged.

REFERENCES Anand, K. J. S . , & Hickey, P. R. (1987). Pain and its effects in the human neonate and fetus. New England Journal of Medicine, 317, 1321-1329. Anderson, G. H. (19%)). Diet and behavior: Multidisciplinary approaches. New York: SpnngerVerlag. Barnes, G.R., Lethin, A. N., Jackson, E. B., & Shea, N. (1953). Management of breast feeding. Journal of American Medical Associution, 151, 192-199. Barr, R. G., Kramer, M. S., Pless, I. B., Boisjoly, C., & Leduc, D. (1989). Feeding and temperament as determinants of early infant crying/fussing behavior. Pediurrics. 84, 514-521. Barrett, D. E. (1984). The effects of undernutrition on children’s behavior. In D. H.Enstrom (Ed.), Proceedings: Nutrition and children’s behavior (pp. 1-49). Princeton, NJ: Office of Education/Department of Human Services.


57

B m t t , D. E., & Frank, D. A . (1987). The effects of undernutririon on children's behavior. New York: Gordon & Breach. Bates, J. E. (1989). Applications of temperament concepts. In G. A. Kohnstamm. J. E. Bates, & M. K. Rothbart (Eds.), Temperament in childhood (pp. 321-355). New York: Wiley. Bates, J. E., Freeland, C. A. B., & Lounsbury, M. L. (1979). Measurement of infant difficultness. Child Development, 50. 794-803. Bell, R . Q . (1966). Level of arousal in breast-fed and bottle-fed human newborns. Psychosomatic Medicine, 28, 177-180. Bell, R. Q. (1968). Adaptation of small wrist watches for mechanical recording of activity in infants. Journal of Experimental Child Psychology, 6 , 302-305. Bellinger, D., Leviton, A., Waternaux, C., Needleman, H., & Rabinowitz, M. (1987). Longitudinal analyses of prenatal and postnatal lead exposure and early cognitive development. New England Journal of Medicine. 316, 1037- 1043. Belsky, J., Taylor, D. G., & Rovine, M. (1984). The Pennsylvania lnfant and Family Development Project: 11. The development of reciprocal interaction in the mother-infant dyad. Child Development. 55, 706-717. Berkowitz, R. I., Agras, W. S., Korner, A. F., Kraemer, H. C., & Zeanah, C. H. (1985). Physical activity and adiposity: A longitudinal study from birth to childhood. Journal of Pediatrics, 106, 734-738. Bernal, J., & Richards, M. P. M. (1970). The effects of bottle and breast feeding on infant development. Journal of Psychosomatic Research, 14. 247-252. Blass, E. (1990). Suckling: Determinants, changes, mechanisms, and lasting impressions. Developmenla1 Psychology. 26, 520-533. Blumen, D. G. (1980). Infant-mother bonding. In S . Freier & A. I. Eidelman (Eds.), Human milk; Its biological and social value (pp. 277-282). New York: Elsevier North-Holland. Bohlin, G . , Hagekull, B., & Lindhagen, K. (1981). Dimensions of infant behavior. Infant Behavior & Development, 4 , 83-96. Brazelton, T. B. (1962). Crying in infancy. Pediatrics. 29, 579-588. Brazelton, T. B. (1973). Neonatal Behavioral Assessment Scale. In Clinics in Developmental Medicine (No. 50). Philadelphia: J. B. Lippincott. Brazelton, T. B., Kozlowski, B., & Main, M. (1974). The origins of reciprocity: The early motherinfant interaction. In M. Lewis & L. A. Rosenblum (Eds.), The effect of the infant on its caregiver (pp, 49-76). New York: Wiley. Brody, S. (1976). Patterns of mothering. New York: International Universities Press. Buss, A. H . , & Plomin, R. (1984). Temperament; Early developingpersonality traits. Hillsdale, NJ: Erlbaum. Capaldi, E. D., & Powley, T. L. (Eds.) (1990). Taste. experience andfeeding. Washington, DC: American Psychological Association. Carey, W. B. (1970). A simplified method for measuring infant temperament. Journal of Pediatrics. 77, 188-194. Carey, W. B. (1985). Temperament and increased weight gain in infants. Journal of Developmental and Behavioral Pediatrics, 6 , 128- 131. Carter, C. S. (1988). Patterns of infant feeding, the mother-infant interaction and stress management. In T. M. Field, P. M. McCabe, & N. Schneiderrnan (Eds.), Stress and coping across development (pp. 27-46). Hillsdale, NJ: Erlbaum. Chavez, A., & Martinez, C. (1982). Growing up in a developing communiry: A bio-ecologic study of the development of children from poor peasant families in Mexico. Mexico City: Institute of Nutrition of Central America and Panama. Crockenberg, S. B., & Smith, P. (1982). Antecedents of mother-infant interaction and irritability in the first three months of life. lnfant Behnvior & Development, 5 , 105-119.

John Worobey

58

Crook, C. K. (1978). Taste perception in the newborn infant. Infant Behavior & Development, I , 5269.

Dallman, P. R., & Siimes, M. A. (1979). lron deficiency in infancy and childhood. Report for the International Nutritional Anemia Consultative Group. New York: Nutrition Foundation. DiPietro, J. A., Larson, S. K., & Porges, S . W. (1987). Behavioral and heart rate differences between breast-fed and bottle-fed neonates. Developmental Psychology, 23, 467-474. Faton, W. O., & Dureski, C. M.(1986). Parent and actometer measures of motor activity level in the young infant, Infant Behavior & Development, 9. 383-393. Eaton, W. O., & Enns, L. R. (1986). Sex differences in human motor activity level. Psychological Bulletin, 100, 19-28. Emde, R. N. (1978). Commentary. In A. I. Sameroff (Ed.), Organization and stability of infant behavior: A commentary on the Brazelton Neonatal Behavioral Assessment Scale (pp. 135-138). Monographs of the Society for Research in Child Development, 43(5-6). Fein, G. G., Schwartz, P. M., Jacobson, S. W., &Jacobson, I. L. (1983). Environmental toxins and behavioral development: A new role for psychological research. American Psychologist, 38, 1 188- 1 197.

Fish, M., Stifter, C. A., & Belsky, J. (1991). Conditions of continuity and discontinuity in infant negative emotionality: Newborn to five months. Child Development, 62, 1525- 1537. Galler, 1. R., Ramsey, F., Solimano, G., Lowell, W. F., & Mason, E. (1983). The influence of early malnutrition on subsequent behavioral development. Journal of Child Psychiarry, 22, 8- 15. Goldman, I. A. (1988). Diet and behavior: A consideration of recent research. In T. D. Yawkey & 5. E. Johnson (Eds.), Integrative processes and socialization: Early to middle childhood (pp. 189205). Hillsdale, NJ: Erlbaum. Goldsmith, H.H., Buss, A. H., Plomin. R.,'Rothbart, M.K., Thomas, A., Chess, S., Hinde, R. A., & McCall, R. B. (1987). Roundtable: What is temperament? Four approaches. ChiidDevelopment, 58, 505-529. Goldsmith, H. H., Rieser-Danner, L. A., & Briggs, S . (1991). Evaluating convergent and discriminant validity of temperament questionnaires for preschoolers, toddlers and infants. Developmental Psychology, 27. 566-579. Greenwood, C. E. (1990). Methodologic considerations for diet and behavior studies: A nutritionist's perspective. In G. H. Anderson (Ed.), Diet and behavior: Multidisciplinary approaches (pp. 193-207). New York: Springer-Verlag. Grossman, L. K.,Fitzsimmons, S. M., Larsen-Alexander, J. B., Sachs, L., & Harter, C. (1990). The infant feeding decision in low and upper income women. Clinical Pediatrics, 29, 30-37. Hansen, J. W., Cook, D. A., Cordano, A., 62 Miguel, S. G. (1988). Human milk substitutes. In R. C. Tsang & B. L. Nichols (Eds.), Nutrition during infancy (pp. 378-398). Philadelphia: Hanley & Belfus. Harris, G . , & Booth, D. A. (1987). Infants preference for salt in food: Its dependence upon recent dietary experience. Journal of Reproductive and Infant Psychology, 5 , 97- 104. Heinstein, M. 1. (1963). Behavioral correlates of breast-bottle regimes under varying parent-infant relationships. Monographs of the Society for Research in Child Development. 28(4). Hendershot, G. E. (1984). Trends in breastfeeding. Pediatrics, 4 (Suppl.), 591-602. Hoefer, C., & Hardy, M. C. (1929). Later development of breast fed and artificially fed infants: Comparison of physical and mental growth. Journal of American Medical Association, 92, 615619.

Holt, G., & Wolkind, S. (1983). Early abandonment of breastfeeding: Causes and effects. Child Care, Health and Development. 9, 349-355. Hopkinson, J. M.,& Garza, C. (1988). Management of breastfeeding. In R. C. Tsang & B. L. Nichols (Eds.), Nutrition during infancy (pp. 298-313). Philadelphia: Hanley & Belfus. Horowitz, F. D., Sullivan, J., & Linn, P. (1978). Stability and instability in the newborn infant: The


59

quest for elusive threads. In A. J. Sameroff (Ed.), Organization and stability of infant behavior: A commentary on the Brazelton Neonatal Behavioral Assessment Scale ( pp. 29-45). Monographs of the Society for Research in Child Development, 43(5-6). Hubert, N. C., & Wachs, T.D. (1985). Parental perceptions of the behavioral components of infant easiness/difficultness. Child Developmenr, 56, 1525- 1537. Hubert, N. C., Wachs, T. D., Peters-Martin, P., & Gandour, M. J. (1982). The study of early temperament: Measurement and conceptual issues. Child Development, 53, 57 1-600. Humenick, S.,& Van Steenkiste, S. (1983). Early indicators of breastfeeding progress. lssues in Comprehensive Pediatric Nursing, 6 , 205-2 15. Izard, C. E., Hembree, E. A . , & Huebner, R. R. (1987). Infants’ emotion expressions to acute pain: Developmental changes and stability. Developmental Psychology, 23. 105- 1 13. Kaye, K. (1982). The mental and social life of babies: How parents create persons. Chicago: University of Chicago Press. Kohnstamm, G . A., Bates, J. E., & Rothbart, M. K. (Eds.) (1989). Temperament in childhood. New York: Wiley. Komer, A. F.,Zeanah, C. H., Linden, J., Berkowitz, R. I . , Kraemer, H. C . , & Agras, W. S.(1985). The relation between neonatal and later activity and temperament. Child Development, 56, 3842. Kuzela, A. L., Stifter, C. M., & Worobey, J. (1990). Breastfeeding and mother-infant interactions. Journal of Reproduerive and Infant Psychology, 8, 185- 194. Latham, M. C. (1974). Protein-calorie malnutrition in children and its relation to psychological development and behavior. Physiological Reviews, 59. 541-565. Lester, B. M., Garcia-Coll, C . , Valcarcel, M., Hoffman, J., & Brazelton, T.B. (1986). Effects of atypical patterns of fetal growth on newborn (NBAS) behavior. Child Development. 57, 11-19, Levine, A. S. (1990). Nutrition and behavior. In J. E. Morley & Z. Glick (Eds.), Geriatric nutrition (pp. 419-433). New York: Raven Press. Levitsky, D. A., & Strupp, B. J. (1985). Nutrition and the behavior of children. In W. A. Walker & 1. Watkins (Eds.), Nutrition andpediatrics: Basic science and clinical applications (pp. 357-372). Boston: Little, Brown. Lozoff, B. (1988). Behavioral alterations in iron deficiency. Advances in Pediatrics, 35, 331-360. Lozoff, B. (1989). Nutrition and behavior. American Psychologist, 44. 231-236. Lucas, A,. Morley, R., Cole, T. J., Lister, G . H., & Leeson-Payne, C. (1992). Breast milk and subsequent intelligence quotient in children born pretenn. Lancef, 339, 261 -264. Mack, R. W., & Kleinhenz, M. E. (1974). Growth, caloric intake, and activity levels in early infancy: A preliminary report. Human Biology, 46, 345-354. Mahaffey, K. R. (1990). Environmental lead toxicity: Nutrition as a component of intervention. Environmental Health Perspectives. 89, 75-78. McCullough, A. L., Kirksey, A., Wachs, T. D., McCabe, G. P., Bassily, N. S., Bishry, 2.. Galal, 0. M., Harrison, G. G . , &Jerome, N. W. (1990). Vitamin B-6 status of Egyptian mothers: Relation to infant behavior and matemal-infant interactions. American Journal of Clinical Nutrifion, 5 1 , 1067- 1074. McDonnell, P. M., Corkum, V. L., & Wilson, D. L. (1989). Patterns of movement in the first six months of life: New directions. Canadian Journal of Psychology, 43, 320-339. McDonnell, P. M., & Richards, C. (1984). Infant actometers: Measuring coordinated movements. Infant Behavior & Development. 7 (Special ICIS Issue), 236. Menella, J. A., & Beauchamp, G . K. (1991). Maternal diet alters the sensory qualities of human milk and the nursling’s behavior. Pediatrics. 88, 737-744. Newton, N. (1955). Maternal emotions. New York: Hoeber. Newton, N. (1971). Psychologic differences between breast and bottle feeding. American Journal of Clinical Nutrition. 24. 993-1004.

60

John Worobey

Pinella, T., Birch, L., Steinberg, L., Picciano, M. F., & G o f f ,N. (1990). Infant state and feeding method. Infant Behavior & Development, 13 (Special ICIS Issue), 570. Pipes, P. L. (1989). Infant feeding and nutrition. In P. L. Pipes (Ed.), Nutrition in infancy and childhood (pp. 86- 119). St. Louis: Times Mirror/Mosby. Pollitt, E. (1988). A critical review of three decades of research on the effects of chronic energy malnutrition on behavioral development. In B. Schurch & N. Scrimshaw (Eds.), Chronic energy deficiency: Consequences and related issues ( pp. 77-92). Lausanne, Switzerland: Nestle Foundation. Price, G. M. (1983). Sensitivity in mother-infant interactions: The AMIS Scale. Infant Behavior & Development, 6, 353-360. Ricciuti, H. N. (1991). Malnutrition and cognitive development: Resemh policy linkages and current research directions. In L. Okagaki & R. I. Stemberg (Eds.), Directors of development: Influences on the development of children’s thinking (pp. 59-80). Hillsdale, NJ: Erlbaum. Richardson, S . A,, Birch, H. G., Grabie, E., & Yoder, K. (1972). The behavior of children in school who were severely malnourished in the first two years of life. Journal of Health and Social Behavior, 13, 276-284. Robson, K. S., & Moss, H. A. (1970). Patterns and determinants of maternal attachment. Journal of Pediatrics. 77. 976-985. Rose, H. E., & Mayer, J. (1968). Activity, calorie intake, fat storage, and the energy balance of infants. Pediarrics, 41, 18-29. Rosen, I. F. (1989). Metabolic abnormalities in lead toxic children: Public health implications. Bulletin of the New York Academy of Medicine, 65, 1067-1084. Rothbart, M. K. (1978). Infant Behavior Questionnaire. Unpublished manuscript, Department of Psychology, University of Oregon, Eugene. Rothbart, M. K. (1981). Measurement of temperament in infancy. ChildDevelopment, 52, 569-578. Rothbart, M. K. (1986). Longitudinal observation of infant temperament. Developmental Psychology, 22, 356-365. Saudino, K. I., & Eaton, W.0. (1991). Infant temperament and genetics: An objective twin study of motor activity level. Child Developmenr. 62, 1167-1 174. Schaffer, H. R. (1971). The growth of sociability. Middlesex, England: Penguin. Shand, N. (1981). The reciprocal impact of breast-feeding and culture form on maternal behavior and infant development. Journal of Biosocial Srience, 13, 1- 17. Simmons, J. E., Ottinger, D., & Haugk, E. (1967). Maternal variables and neonate behavior. Journal of the American Academy of Child Psychiatry. 6. 174- 183. Sjolin, S., Hofvander, Y.,& Hillervik, C. (1977). Factors related to early termination of breastfeeding. Acra Paediatrica Scandinavica, 66, 505-5 1 1 . Spietz, A. L. (1978). Why look at the feeding? In K. Barnard (Ed.), The Nursing Child Assessment Scales (pp. 5- 16). Seattle: NCAST. Stein, A., Cooper, P., Day, A., & Bond, A. (1987). Social psychiatric factors associated with the intention to breastfeed. Journal of Reproductive and Infant Psychology, 5 . 164- 171. Steinberg, L.. Coff, N., Birch, L. L., Picciano, M. F.. & Hatch, T.(1989). Infant behavior and nutrition: Behavioral differences among infants fed formula varying in tryptophan. Abstracts of the Society for Research in Child Development, 7 , 165. Stoch, M. B., Smythe, T. M.,Moodie, A. D., & Bradshaw, D. (1982). Rychosocial outcome and CT findings after growth undernourishment during infancy: A twenty year developmental study. Developmental Medicine & Child Neurology, 51 419-436. Stratton, P. (1982). Newborn individuality. In P. Stratton (Ed.), Psychobiology of the human newborn ( pp. 22 1-26 I). London: Wiley. Stricker, E. M. (1990). Handbook of behavioral neUrObiOl0gy: Neurobiology of food andfluid intake (Vol. 10). New York: Plenum Press. I


61

Switzky, L., Vietze, P., & Switzky, H. (1979). Attitudinal and demographic predictors of breastfeeding and bottle-feeding behaviors by mothers of six-week-old infants. Psychological Reports. 45, 3-14. Thomas, A., & Chess, S. (1977). Temperament and development. New York: Brunner/Mazel. Thomas, A., Chess, S., Birch, H. C . , Hertzig, M. E., & Korn. S. (1963). Behavioral individuality in ear1.v childhood. New York: New York University Press. Tryon, W. W. (1984). Principles and methods of mechanically measuring motor activity. Behavioral Assessment. 6, 129- 139. Van den Boom, D. C. (1989). Neonatal irritability and the development of attachment. In G. A. Kohnstamm, J. E. Bates, & M. K. Rothbart (Eds.), Temperament in childhood (pp. 299-318). New York: Wiley. Walravens, P. A., & Hambidge, K. M. (1976). Growth of infants fed a zinc supplemented formula. American Journal of Clinical Nutrition. 29, I 1 14- 1 129. Walton, M. D.,& Vallelunga, L. R. (1989). The role of breast-feeding in establishing early motherinfant interaction. Abstracts of the Sociery for Research in Child Development, 7 , 393. Wiesenfeld, A. R . , Malatesta, C. Z., Whitman. P. B., Grandose, C., & Uili, R. (1985). Rychophysiological response of breast- and bottle-feeding mothers to their infants' signals. Psychophysiology. 22, 70-85. Worobey, J. (1986). Convergence among assessments of temperament in the first month. Child Development, 57, 47-55. Worobey, J. (1989). Mother-infant interaction: Protocommunication in the developing dyad. In J. F. Nussbaum (Ed.), Life-span communication: Normative processes (pp. 7-25). Hillsdale, NJ: Erl baum . Worobey, J. (1990). Behavioral assessment of the neonate. In J. Columbo & J. Fagen (Eds.), Individual dzyerences in infancy: Reliabiliry, stability, prediction ( pp. 137- 154). Hillsdale, NJ: Erlbaum. Worobey, J., & Anderson-Goetz, D. (1985). Maternal ratings of newborn activity: Assessing convergence between instruments. Infant Mental Health Journal. 6 , 68-75. Worobey, J . , & Belsky, J. (1982). Employing the Brazelton Scale to influence mothering: An experimental comparison of three strategies. Developmental P sycho/ogy. 18. 736-743. Worobey, J., & Blajda, V. M. (1989). Temperament ratings at 2-weeks, 2-months, and I-year. Developmental Psychology. 25, 251-263. Worobey, J . , Laub, K. W., & Schilmoeller, G. L. (1983). Maternal and paternal responses to infant distress. Merrill-Palmer Quarterly, 29, 33-45. Worobey, J., & Lewis, M. (1989). Individual differences in the reactivity of young infants. Developmental Psychology. 25, 663-667. Worobey, I., & Lewis, M. (1992). Behavioral differences in response to stress between breast- and bottle-fed infants. Topics in Clinical Nutrition, 7 , 48-55. Wright, P., & Crow, R. (1982). Nutrition and feeding. In P. Stratton (Ed.), Psychobiology of the human newborn (pp. 339-364). London: Wiley. Yogman, M. W., & Zeisel, S. H. (1983). Diet and sleep patterns in newborn infants. New England Journal of Medicine. 309. I 147- 1149. Zeskind, P. S.,Goff, D. M., & Huntington, D. L. (1984).A developmental approach to the study of the behavioral effects of early malnutrition. In J. Brozek & B. Schurch (Eds.), Malnutrition and behavior: Critical assessment of key issues ( pp. 424-440). Lausanne, Switzerland: Nestle Foundation.


THE DEVELOPMENT OF READING

Linda S. Siege1 DEPARTMENT OF INSTRUCTION AND SPECIAL EDUCATION THE ONTARIO INSTITUTE FOR STUDIES IN EDUCATION TORONTO, ONTARIO. CANADA M5S 1 V6

I. INTRODUCTION 11. CONTROVERSIES AND METHODOIBGICAL ISSUES

A. DEFINITIONAL ISSUES: A DIGRESSION B . DEFINITIONS 111. BASIC COGNITIVE PROCESSES IN READING A. A THEORETICAL APPROACH B. PHONOLOGICAL PROCESSING C. SYNTACTIC AWARENESS D. WORKING MEMORY E. SEMANTIC PROCESSING F. ORTHOGRAPHIC PROCESSING IV. CONCLUSIONS REFERENCES

I just look at the words and I know what they say. (Laura, age 6 years)

I. Introduction The epigraph was a 6-year-old girl’s answer to my question about how she learned to read. In this article, I show that the correct answer to that question is more complicated. In fact, learning to read is an amazing achievement for the young child. From mouthing and touching books in infancy, to looking at pictures in a book, to not understanding the words on a page, to halting and slow decoding, the child moves to fluent reading and understanding and remembering 63 ADVANCES IN CHILD DEVELOPMENT AND BEHAVIOR. VOL. 24

Copyright 8 IW3 by Academic Press. Inc All nghts of reproduction in any form reserved.

64

Linda S. Siegel

what has been read. All of this progress occurs in a relatively short period. I review the literature on the normal course of the development of reading and also examine what happens when reading skills fail to develop adequately in children with reading disabilities. In this article, I discuss the development of reading by analyzing it in terms of a theoretical approach that is focused on the basic cognitive processes relevant to the development of these skills as outlined by Siegel (in press). First, however, I consider some important conceptual and methodological issues in this field.

11. Controversies and Methodological Issues A great deal of inconsistency and controversy exist in the research on reading and reading disabilities. Therefore, any discussion of reading and reading disabilities must start with a clarification of some basic definitional issues and assumptions. As I see it, the confusion in the field results from lack of clear, theoretically motivated, and consistent operational definitions of two major constructs, reading and reading disability. Although the question of what reading means may sound trivial, hundreds of tests are called reading tests, and reading is defined in a different way in each one and hence each yields a different measure. This inconsistency constitutes a fundamental problem with the definition of this critical variable. I believe that the lack of integration in this field is a result of the lack of clarity in regard to the basic operational definitions. These definitional issues were reviewed by Siegel and Heaven (1986), but one of the most significant issues is the difference between reading comprehension and word recognition. Tests of reading comprehension typically involve the reading of text and multiple-choice questions about the text; tests of word reading involve the reading of single words. Reading comprehension tests are timed; word reading tests are not. Although reading Comprehension may appear to be the fundamental aspect of reading and is clearly the ultimate goal of reading, the measurement of reading comprehension is a methodologically complex issue full of pitfalls. The issues in the measurement of reading comprehension were examined in detail by Siegel and Heaven (1986), Siegel and Ryan (1989b), and Tal and Siegel (1990); but the fundamental problem is that measures of reading comprehension are confounded by a number of other processes, such as background knowledge, vocabulary, and reading speed, and available tests of reading comprehension usually involve not making an inference from the text material but merely finding a verbatim answer in the text. In contrast, tests of word recognition measure more basic processes, and responses are not confounded with differences in reading speed, background knowledge, and test-taking strategies. In addition, the use of reading comprehension scores as the independent variable or the basis of the definition of reading disability can yield results different from those ob-

Development of Reading

65

tained by the use of word recognition scores (e.g., Siege1 & Ryan, 1989a, 1989b; Stanovich, Nathan, & Zolman, 1988). Also, from a theoretical perspective, word recognition is fundamental to comprehension (e.g., Gough & Tunmer, 1986; Stanovich, 1982a, 1982b, in press). The ability to read isolated words is highly correlated with text comprehension (e.g., Shankweiler & Liberman, 1972). The problems of the beginning reader or the disabled reader are clearly at the level of the word. Problems at the word level interfere with the reading of connected text (Shankweiler & Liberman, 1972). Because word decoding is critical to comprehension and is the basic process in reading, the discussion in this article is concentrated on the development of word recognition. A. DEFINITIONAL ISSUES: A DIGRESSION

I . Continuum versus Dichotomy Another critical issue involves what constitutes the appropriate definition of a reading disability. Throughout this article, I use the term reading disability instead of dyslexia. The terms are actually synonymous, but certain considerations have led to the widespread avoidance of the term dyslexia in many parts of the world, particularly by, although not limited to, the educational community. I do not understand why the term dyslexia is often viewed as if it were a four-letter word not to be uttered in polite company; however, I speculate briefly. The term is often taken to imply an illness, such as measles, when, in fact, in the words of Ellis (1985), it is more similar to a problem such as obesity. As Ellis has written, For people of any given age and height there will be an unintermpted continuum from painfully thin to inordinately fat. It is entirely arbitrary where we draw the line between “normal” and “obese,” but that does not prevent obesity being a real and worrying condition, nor does it prevent research into the causes and cures of obesity being both valuable and necessary. , , , Therefore, to ask how prevalent dyslexia is in the general population will be as meaningful, and as meaningless, as asking how prevalent obesity is. The answer will depend entirely upon where the line is drawn. (p. 172)

No virus, nor specific brain lesion, nor biochemical disturbance has been shown to be the cause of dyslexia, so it is not an illness in the traditional medical sense. Because a reading disability is an educational problem and not a medical one and because it cannot be treated by any of the traditional medical means, professionals are often reluctant to use the term dyslexia; however, it is a very real condition that deserves study and treatment. I believe reading problems are best conceptualized as a continuum with varying degrees of severity. Obviously, a problem at any level deserves attention and treatment but the dividing line between a reading problem and no problem is arbitrary. Fear and disdain of the term dyslexia are common in North America but

66

Linda S. Siegel

seem less common in other parts of the world. I can offer no empirical evidence to support these speculations but I suspect that the sociopolitical context has influenced the terminology. The egalitarian philosophy and the cultural ethos of North America may lead to the perception that a label, such as dyslexia, applied to a child may reduce access to educational opportunities. Therefore, for these considerations and for those who find the term dyslexia offensive, I generally use the term reading disability, although, as far as I am concerned, their meanings are identical.

2 . Subtypes One of the issues that has been raised in the study of reading disability is whether or not reading-disabled individuals can be separated into subtypes; however, no reliable evidence supports the concept of subtypes and no clear subtypes have been delineated (see Siegel & Heaven, 1986; Siegel, Levey, & Ferris, 1985; Siegel & Metsala, in press, for a review of studies and methodological issues). On the contrary, children with a reading disability show remarkable homogeneity in the profiles of their cognitive abilities (e.g., Siegel & Ryan, 1989b) and, when heterogeneity is found, it seems to result from the particular definition used in the study. Evidence indicates that the definition of reading disability used in a study can influence the conclusions made about the heterogeneity of the population. For example, Siegel and Ryan (1989b) have shown that if reading disability is defined as a deficit in word reading skills, all the children with reading problems have deficits in phonological processing, working memory, short-term memory, and syntactic awareness. The pattern is similar if a deficit in pseudoword reading skills is used as the basis for defining reading disability. If, however, reading disability is defined on the basis of a deficit in reading comprehension, then the group that emerges is heterogeneous and does not show deficits in phonological processing and syntactic skills but does show deficits in working memory and short-term memory. Thus, if and when subtypes appear within the readingdisabled population, they may be artifacts of the definition used. 3. tQ and Reading When issues related to reading disabilities are examined, the question is always raised as to the role of IQ and whether any differences in cognitive processes between reading-disabled individuals and normal readers are merely a result of differences in IQ. However, no reliable evidence indicates that IQ level plays a causative role in the development of reading skills (e.g., Stanovich, 1988~).On the contrary, evidence from a number of sources indicates that reading is not strongly related to intelligence as measured by IQ tests. Readingdisabled children at all IQ levels show equal difficulty with phonological processing tasks such as pseudoword reading, recognizing the visual form of a


67

pseudoword, and pseudoword spelling (Siegel, 1988). Therefore, the presence of a reading disability, not a particular IQ, determines the pattern of cognitive strengths and weaknesses in regard to language, memory, and phonological skills. Often, the reading-disabled individual is defined as a person whose reading score is significantly lower than would be predicted from his or her IQ. (Individuals who fit this definition have traditionally been labeled as dyslexic.) If an individual has a lower reading score but it is not significantly lower than would be predicted by his or her IQ, the individual is not defined as dyslexic. This definition is referred to as the discrepancy definition; however, a number of investigators have provided evidence that a discrepancy between IQ and reading is not necessary for an individual to be reading disabled. For example, I have compared (Siegel, 1991a, 1991b) dyslexics, defined as children whose reading scores were low (standard scores < 90)and significantly (1 SD) below their IQ scores, and poor readers, whose reading scores were low (standard scores < 90) but not below the level predicted from their IQ. These two groups did not differ on any reading, spelling, or phonological processing tasks and on most language and memory tasks, in spite of the fact that the mean IQ score of the dyslexics was 25 points higher than that of the poor readers. Both of these groups had scores on the reading, spelling, phonological processing, language, and memory tasks that were significantly below those of normal readers. The critical variable was the presence or absence of a reading disability. Indeed, if the relative contributions of IQ and pseudoword reading are compared, then IQ contributes little independent variance beyond that contributed by pseudoword reading to the prediction of word reading and reading comprehension scores (Siegel, in press). Most of the variance is contributed by phonological processii 5 as measured by pseudoword reading. In summary, intelligence as measured hy IQ scores seems irrelevant to the definition and analysis of reading disability. B . DEFINITIONS

Therefore, throughout this article children who have low scores on reading tests are called poor readers, whether or not their reading scores are significantly lower than what would be predicted by their IQ scores. Typically, a reading score at or below the 20th or 25th percentile is used. Good or average readers are defined as having scores on reading tests at or above the 30th, 35th, or 40th percentile (depending on the study). For the reasons described earlier, word reading tests, as opposed to reading comprehension tasks, yield the clearest definition of normal and atypical reading. Comparisons between disabled and normal readers are typically based on chronological age and most of the studies revealed in this article use chronological age to make these comparisons. An-

68

Linda S. Siegel

other type of design is, however, possible. This design involves what is called a reading level match. An alternative to study both the development of reading skills and the differences and similarities between disabled and normal readers is to match disabled and normal readers on reading age, also called reading level (e.g., Backman, Mamen, & Ferguson, 1984). This type of design is used in an attempt to identify differences between reading-disabled and normal readers that are merely consequences of differential experience with print. The theory underlying this type of comparison is that children who are poor readers probably read less and therefore do not have the same exposure to print. If so, a chronological age match confounds differences that reflect experience with print and differences that reflect factors that cause reading disability.

111. Basic Cognitive Processes in Reading A. A THEORETICAL APPROACH

I have postulated five processes that are possibly significant in the development of reading skills in the English language (Siegel, in press). The processes involve phonology, syntax, working memory, semantics, and orthography. In this article, I review the role of all these processes in the development of reading skills. Unfortunately, most of the information that is available about the development of reading is based on studies conducted with English, a language that has the highest degree of irregularity of the correspondence between letters, more properly graphemes, and phonemes, the sounds represented by letters and letter combinations. Some studies have been addressed to the prevalence of reading problems in other languages, specifically, Stevenson, Stigler, Lucker, Hsu, and Kitamura (1982) for Chinese and Japanese and Lindgren, De Renzi, and Richman (1985) for Italian. In both of these studies, however, deficit in reading comprehension was used as the measure of a reading problem and, as discussed in Section I,2, this definition does not address the cognitive deficits that underlie severe reading problems, specifically phonological processing. The complexities of studying the relationship between the acquisition of reading skills and different orthographies were outlined by Liberman, Liberman, Mattingly, and Shankweiler (1980): Orthographies vary considerably in the demands they make on the beginning reader. This variation has two essentially independent aspects: first, the depth of the orthography, its relative remoteness from the phonetic representation; and second, the particular linguistic unit-morpheme, syllable, or phoneme-that is overtly represented. A deep orthography, like that of English, demands greater phonological development on the reader’s part than a shallow orthography, like that of Vietnamese. Logographies


69

(such as the Chinese writing system), syllabaries (such as old Persian cuneiform), and alphabetic systems (such as English) demand successively increasing degrees of linguistic awareness. (p, 146)

Obviously, the consideration of other languages is important and I include evidence from other languages when it is available but this evidence is very meager, Phonological processing involves a variety of skills, but in the context of the development of reading skills, the most significant is the association of sounds with letters, that is, the understanding of grapheme-phoneme conversion rules 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). Syntactic awareness, also called grammatical sensitivity, refers to the ability to understand the syntax of the language. 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 into 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. 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 graphemephoneme 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 might 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.

70

Lindn S. Siege1

Semantic processing refers to the understanding of meaning. Theoretically, word meanings are coded in semantic networks and are retrieved through these networks. In the context of reading, semantic processing is relevant to the retrieval of words. For example, the ease of retrieving the meaning of a word may depend, at least partially, on the connections that it has with other words in a semantic network. 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. In the following sections, I provide details of the growth of these skills in children who are normal readers and also in children with reading disabilities. B . PHONOLOGICAL PROCESSING

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, 1988~). Phonological processing involves a variety of functions, but in the context of the development of reading skills, the most significant is the association of sounds with letters or combinations of letters. 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. I . Dual-Route Theories The development of phonological processing and the development of reading can be understood in the context of what have been called “dual-route’’ theories of reading. These theories have a variety of manifestations, but their basic premise is that two possible routes are involved in gaining access to the meaning of print (e.g., Coltheart, 1978; Forster & Chambers, 1973; Meyer, Schvanevelt, & Ruddy, 1975). One of these routes involves direct lexical access, that is, visually reading a word without any intermediate phonological processing. The orthographic configuration of a word is directly mapped onto an internal visual store in lexical memory. The other route, the phonological route, involves the use of grapheme-phoneme conversion rules to gain lexical access to a print stimulus. Grapheme-phoneme conversion rules are used to translate a graphemic code into


71

a phonemic one. This route is referred to as nonlexical because the application of the rules does not rely on word-specific pronunciations. Instead, graphemephoneme conversion rules are presumed to be stored explicitly and used to determine a word’s pronunciation. According to this model, pseudowords can be read only by means of a nonlexical route, as, by definition, a pseudoword cannot have a lexical representation. Dual-route theories have been challenged. For example, the reading of nonwords is influenced by their similarity to real words, and regular words that have irregular orthographic neighbors are read more slowly than regular consistent ones, indicating reciprocal influences of these two routes. If pseudowords were read only by grapheme-phoneme conversion rules, then the reading of pseudowords should not be influenced by their similarity to real words, and regular words should not be influenced by the characteristics of their orthographic neighbors. Furthermore, multiple-level models (e.g., Brown, 1987) and connectionist models (e.g., Seidenberg & McClelland, 1989) that have been proposed involve a variety of postulated units and processes but not two distinct routes. (For an extended discussion of these issues, see Besner, Twilley, McCann, & Seergobin, 1990; Glushko, 1979; Humphreys & Evett, 1985; Metsala & Siegel, in press; Stanovich, in press). In spite of a certain ambiguity about the validity of dual-route theories, however, conceptualizations of reading in terms of dual-route theory represent one way of examining the development of reading skills and the performance of children with a reading disability. I discuss tasks used to measure both these kinds of processing, the direct lexical access and the use of grapheme-phoneme conversion rules, and the performance of readingdisabled and normal readers on these types of tasks. 2, Measurement of Phonological Processing Skills 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 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. In an alphabetic language such as English, 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 such pseudowords as shum, laip, and cigbet. Pseudowordscan 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.

12

Linda S. Siegel

The development of the ability to read pseudowords has been studied extensively (e.g., Calfee, Lindamood, & Lindamood, 1973; Hogaboam & Perfetti, 1978; Siegel & Ryan, 1988; Venezky & Johnson, 1973). Ample evidence indicates that children with dyslexia have a great deal of difficulty reading pseudowords. Studies such as those of Bruck (1988), Ehri and Wilce (1983), Snowling (1980), Siegel and Ryan (1988), and Waters, Bruck, and Seidenberg (1985) have shown that disabled readers have more difficulty reading pseudowords than normal readers matched on either chronological age or reading level. For example, Siegel and Ryan (1988) studied the development of the ability to read pseudowords in normal and disabled readers aged 7 to 14 years. By the age of 9, the normal readers were quite proficient and performed at almost a perfect level for even the most difficult pseudowords, with, in some cases, as many as three syllables. Similarly, Backman, Bruck, Hebert, and Seidenberg (1984) showed that 10-year-olds perform as well as adults on tasks involving the reading of pseudowords; however, Siegel and Ryan (1988) found that the performance of the reading-disabled children was quite different. These children appear to acquire these reading skills very late in development and even readingdisabled children at the age of 14 were performing no better than normal readers at the age of 7. To control, at least partially, for experience with print, Siegel and Ryan (1988) used a comparison of disabled and normal readers matched on reading grade level. Even when the disabled readers and the normal readers were matched on reading level (hence the disabled readers were considerably older than the normal readers), the performance of the reading-disabled on a task involving the reading of pseudowords was significantly poorer than that of the normal readers. Thus, difficulties with phonological processing seem to be the fundamental problem of children with reading disability and this problem continues to adulthood. Many adults with a reading disability become reasonably fluent readers but still have difficulty reading pseudowords or read them very slowly (e.g., Barwick & Siegel, 1990; Bruck, 1990; Shafrir & Siegel, 1991). 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 as to 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 grapheme-phoneme conversion rules develops rapidly in the first years of experience with print under normal conditions.

3 . Developmental Stages of Phonological Processing No conclusive evidence exists as to the process by which these skills develop; however, several general accounts of the process by which the child learns to


73

read have been proposed. Ehri and Wilce (1983) postulated three phases in this process. In phase 1 , unfamiliar words become familiar and the child pays attention to the component letters of a word. In phase 2, words come to be recognized as wholes with deliberate processing of grapheme-phoneme correspondences, and the meanings of words are accessed automatically. In phase 3, the speed of processing increases significantly. Less skilled readers, however, do not show this automaticity or the growth of speed in identifying words and nonwords. Harris and Coltheart (1986) proposed four phases in learning to read. Initially, children learn to read a small set of words through the direct access or visual route; that is, they recognize words without sounding them out. Then children learn a small set of words on which they have been instructed. Then, around 5 or 6 years of age, children rely on partial cues and relate printed words to items stored in memory. Phonological recoding occurs at the next stage and graphemephoneme conversion rules are used extensively. But grapheme-phoneme conversion rules are inadequate for many languages in which the correspondence between letters and phonemes is not perfect; hence, an orthographic stage, with no phonological recoding of words, is the final stage. Gough and Juel (1991) also proposed a series of stages by which the child learns to read. In the first stages, the child learns to pair sounds with a printed word in an associative process. According to Gough and Juel, the child selects one cue from the printed word and the response is associated with that one cue. To illustrate this process, Gough and Juel reported an unpublished study in which children 4 to 5 years old were asked to learn four words on cards. One of the cards had a thumbprint in the lower left comer. The children learned the word on the card with the thumbprint much faster than the other three, but often could not identify the word unless the thumbprint was on the card, and would, in the presence of the thumbprint, incorrectly label a word with the word that had been on the card with the thumbprint. Thus, the children appeared to be learning the word-sound association based on the overall visual stimulus without attention to individual letters. That is, they were learning a sound-picture association and incorrectly using part of the visual stimulus, in this case an irrelevant element. In terms of the dual-route theory, these children were apparently using the direct access or visual route, but doing so very inefficiently. Gough (in Gough & Juel, 1991) provided an additional demonstration of this use of partial cues. He taught children 4 to 5 years old to read four words and then determined whether they could recognize a word when half of it was hidden. Some of the children could recognize the word if the first part was hidden (du in duck) but not if the second part was hidden and some could recognize the word when the second part was hidden but not the first. They appeared to be using only partial visual cues. According to Gough and Juel, in the next stage the child must map spoken language onto printed words using a process called cryptanalysis, that is, learn-

74

Linda S. Siegel

ing the correspondences of sounds and letters (the orthographic cipher). Gough and Juel distinguished between this cipher and what is called phonics. They characterized the rules of English phonics as explicit and the cipher as a larger set of regularities that may be learned as rules or that may be represented by analogies. They asserted that the use of phonics rules is a slow and laborious process of associating each sound with a letter, holding the sound in memory, and blending all the individual sounds to make a word. Gough and Juel noted that the test of mastering the cipher is the reading of pseudowords. They obtained a correlation of 5 5 between the reading of real words and pseudowords. Siegel and Ryan (1988) obtained a correlation of .86 for English and, for Portuguese, Da Fontoura and Siegel (1991) obtained a correlation of .63. A child who is “using the cipher,” in their terminology, will make more reading errors that are nonwords than a child who is not using it; that is, the child not using it will be more likely to guess another word. A number of studies have shown that children who cannot read very well make just these sorts of errors (e.g., Johnston, 1982; Siegel, 1985; Sprenger-Charolles, 1991). These studies are discussed in detail later in this article. In contrast, the child using the cipher will make errors indicating a misapplication of rules. 4 . Acquisition of Grapheme-Phoneme Conversion Rules

Although we have evidence about the inadequate phonological skills of children with reading disabilities, little is known about the precise manner in which the complex grapheme-phoneme conversion rules of the English language are acquired. The studies reported previously have involved global measures of pseudoword reading. This type of measure is an important first step, but to understand the process of reading, a more detailed analysis is needed. Venezky and Johnson (1973) said, “A single ‘word attack’ score has little diagnostic value, especially for those children who fall in the middle ranges between mastery and complete failure” (pp. 109-1 10). The ascertainment of the order and nature of the acquisition of these rules is an important step in the understanding and treatment of reading skills. A number of investigators have begun to work on the problem of specifying the order of acquisition of these grapheme-phoneme conversion rules with the expectation that the rules are acquired in a relatively fixed and predictable order in a manner similar to the way syntactic structures develop in oral language (e.g., Guthrie & Seifert, 1977; Siegel & Faux, 1989; Snowling, 1980). To study these issues, we showed disabled and normal readers words and pseudowords that involved a variety of grapheme-phoneme conversion rules, such as consonant blends, r-influenced vowels, and inconsistent vowels (Siegel & Faux, 1989). We found that complexity, as measured by the number of syllables in a pseudoword, was a significant determinant of the difficulty of reading the pseudoword. Reudowords with two or more syllables were quite difficult for older disabled readers ( 1 1 - 13 years) even though normal

Developmenr of Reading

75

readers had become quite proficient by the age of 9 to 10. Even simple vowels and consonant blends were not mastered by the oldest children with reading disabilities in the study (ages 11-14) when they were required to read pseudowords such as mog, lun, and spud, although most of the 7- and 8-year-old normal readers had no difficulty with these features in words or pseudowords. In most cases, even when the disabled readers appeared to demonstrate mastery of grapheme-phoneme conversion rules when they read a word, they were unable to read a pseudoword with the same rule. The reading-disabled experienced unusual difficulty when reading pseudowords. Even when they could read words with particular grapheme-phoneme correspondences in consonantvowel-consonent (CVC) words, such as ran, wet, and sit, they could not read pseudowords such as han, fet, and rit, and although they could read words involving consonant blends, such as hunt, spot, and help, they could not read pseudowords of a similar structure, such as lunt, grot, and melp. This superiority of words over pseudowords suggests that the reading-disabled children were using some sort of direct lexical access which, of course, they could use when they read words but which was not possible in the reading of pseudowords. This direct lexical access probably involves processing each word as a picture (visual representation) rather than as a series of letters with sounds. This visual representation is retrieved from long-term memory. One relatively simple rule of English, with few exceptions, is that a final e in a one-syllable word makes the vowel long. This rule was not measured by the oldest reading-disabled children in this study. That is, the older disabled readers could correctly read the words that reflected the rule (e.g., like, cute, nose) but not the comparable pseudowords (e.g., rike, fute, mose). This difficulty is quite surprising because this rule is repeatedly stressed in reading instruction and is normally mastered very early in the development of reading skills. In many instances, the scores of the reading-disabled children were significantly lower than those of normal readers who were matched on reading grade level. For example, the disabled readers had significantly lower scores than the normal readers of the same reading age on the following tasks: Reading one-syllable pseudowords at grade level 3; two-syllable pseudowords at grade level 4-5; multisyllable pseudowords at grade level 6; and pseudowords with consonant blends at grade levels 2, 3, and 6. In some cases, the reading-disabled and normal readers did not differ; however, these cases often resulted from floor or ceiling effects. English orthography is characterized by unpredictable correspondences between graphemes and phonemes. That is, when reading a given grapheme, one often cannot predict its pronunciation. Some words are regular (e.g., paid, gave, heat) and can be read using the rules of pronunciation of their component graphemes. Other words are irregular or exceptions, and they violate graphemephoneme conversion rules and have no rhymes with similar spelling patterns

16

Linda S. Siegel

(e.g., said, have, great). Words in another category also have irregular grapheme-phoneme correspondences but also have unusual spellings that do not occur in many other words, such as aisle, ache, and tongue. Waters, Seidenberg, and Bruck (1984) found that younger normal and poor readers were sensitive to the effects of irregular spelling and irregular grapheme-phoneme correspondence and took longer to read words with these characteristics. The children also showed the effects of frequency, in that the regular exception differences were greater with low-frequency words, such as pint and wool. Because children with reading disabilities have poor phonological skills, .they are more likely to rely on context when reading (e.g., Bruck, 1988). Other studies have shown that poor readers have difficulty with exception words (Manis & Morrison, 1985; Seidenberg, Bruck, Fornarolo, & Backman, 1985). Still others have not revealed any difference between regular and irregular words for disabled readers (Frith & Snowling, 1983; Seymour & Porpodos, 1980; Siegel & Ryan, 1988). If regular words with regular pronunciations are not read more easily than irregular words, grapheme-phoneme conversion rules are apparently not being used. In addition, disabled readers are much less likely than normal readers to regularize the vowels in irregular words (Seidenberg et al., 1985). One set of hypotheses that has been advanced is that the development of reading skills is accompanied by increasing reliance on the visual route. At the early stage of acquisition, readers rely heavily on phonological information, but good readers learn to recognize high-frequency words automatically. Words are largely recognized by direct access through the visual route. Doctor and Coltheart (1980) found that good readers relied more on phonological mediation when judging the meaningfulness of sentences. They used four types of meaningless sentences: (1) sentences that sounded correct but in print had an incorrect real word (e.g., I have know rime);( 2 ) meaningless sentences with a pseudoword (I have bloo rime); ( 3 ) meaningless sentences containing real words (I have blue time); (4) meaningful sentences with pseudohomophone (I have noe time). The children were required to read these sentences and were asked whether the sentences made sense. Sentences that sounded correct when phonologically recoded (e.g., I have know time; I have noe time) produced more incorrect responses than did sentences that were meaningless when phonologically recoded (e.g., I have blue time; I have bloo time). The difference decreased with age, however, and the investigators concluded that young readers rely on phonological encoding and older readers rely on visual encoding through the direct route. Backman et al. (1984) found that beginning readers appear to be using the visual route for high-frequency words but they are also learning more about grapheme-phoneme conversion rules. Young readers and poor readers had difficulty reading homographic patterns, that is, orthographic patterns with multiple pronunciations such as ose in hose, lose, and dose. Backman et al. showed good


77

and poor readers regular words (e.g., hope), exception words (said), regular inconsistent words, that is, words with regular pronunciations but with irregular orthographically similar neighbors (e.g., paid and said), ambiguous words (e.g., clown because own can be pronounced as in down or blown), and pseudowords constructed to test these orthographic features. Young normal readers read the regular words that were of high frequency quite well, but made more errors on exception, regular inconsistent, and ambiguous words. Older good readers performed at a level comparable to that of high school comparison subjects. Although most errors on the exception words involved regularizations (e.g., come pronounced as coam) rather than errors that were not (come pronounced as came), younger children made fewer regularizations than older children and high school students. However, fewer errors involved giving regular inconsistent words an irregular pronunciation (e.g., bone read as bun, like done). Poor readers were not as skilled at using grapheme-phoneme conversion rules and had more difficulty with orthographic patterns that had multiple pronunciations. Poor readers also had more difficulty than normal readers with the exception, inconsistent, and ambiguous words and tended to make fewer regularization errors. Poor readers also had more difficulty with pseudowords. In normal circumstances, as children get older they become more skilled at reading the irregular and unpredictable aspects of English orthography. Poor readers, however, continue to have difficulty with the orthographic features that are not predictable but do well with high-frequency regular words. This pattern of findings is consistent with the Doctor and Coltheart (1980) findings about a shift from phonological recoding to direct visual access. Seidenberg et al. (1985) also found that poor and disabled readers took longer and were less accurate in reading words with homographic patterns (e.g., one, as in done and gone) than normal readers. Exception words were the hardest for good readers but they read regular inconsistent, ambiguous, and regular words equally well. This pattern suggests that they were significantly influenced by grapheme-phoneme conversion rules because exception words, by definition, violate these rules and these words were the most difficult to read. Poor and disabled readers made more errors on exception, regular inconsistent, and ambiguous than regular words. Manis et al. (1987) found that reading-disabled children had more difficulty than normal readers in a task that required learning to associate symbols with words or symbols with other symbols, particularly when the rule was inconsistent. This type of rule learning is analogous to the grapheme-phoneme conversion rules of English; however, the disabled and normal readers did not differ in learning the association when no rule was applicable. Therefore, reading-disabled children do not appear to have a deficit in visual memory that does not involve linguistic stimuli. Relatively few detailed studies of the acquisition of specific graphemephoneme conversion rules have been conducted. Venezky and Johnson (1973)

78

Linda S. Siege1

studied the acquisition of reading the letter c, pronounced as k or s, and the letter a, pronounced short (ae) or long (e) using pseudowords such as cipe, acim, bice. They found that for normal readers, the rules for the long and short a appeared very early in reading acquisition, but the rule for the c pronounced as s appeared much later. The initial c as s was learned more slowly than the pronunciation of c in the medial position. Venezky and Johnson speculated that the child may not be exposed to as many words with ce, ci, and cy and the teaching may not emphasize the multiple pronunciations of c. Although Venezky and Johnson did not specifically test poor readers, they noted that the scores on their reading task were correlated with reading comprehension scores.

5 . Vowels English vowels tend to have more complex and irregular pronunciations than English consonants. The grapheme-phoneme correspondences of English vowels are very unpredictable. At this time, the understanding of the relationship between the nature of English vowel orthography and the development of reading skills and problems cannot be determined because as Shankweiler and Liberman (1972) noted: This generalization applies to English. We do not know how widely it may apply to other languages. We would greatly welcome the appearance of cross-language studies of reading acquisition, which could be of much value in clarifying the relations between reading and linguistic structure. That differences among languages in orthography are related to the incidence of reading failure is often taken for granted, but we are aware of no data that directly bear on this question. (p. 310)

More vowel spellings correspond to a particular vowel phoneme than consonant spellings to a particular consonantal phoneme. Consequently, misreadings of vowels occur more frequently than misreadings of consonants (Fowler, Shankweiler, & Liberman, 1979; Weber, 1970). Unlike consonants, which are more likely to be misread in the final than initial position, the position of a vowel has no effect on the probability that it will be misread. Unlike consonant errors, vowel errors are unrelated to their target sound; that is, they are random in regard to phonetic features. According to Fowler, Liberman, and Shankweiler ( 1977), vowels are less clearly defined and are more subject to individual and dialect variation. Vowels are the foundation of the syllables and code the prosodic features, and consonants cany the information. 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 reading of vowels is contextfree if this rule is ignored and the vowel is pronounced with the short vowel sound (e.g., cape read as cap), and the reading is context dependent if the rule is followed (Fowler et al., 1979). Fowler et al. administered pseudowords to young normal readers and found that most of the responses to vowels were not random but were either context depen-

Developmni of Reading

19

dent or context free; that is, the children were using the possible sounds for that vowel. Context-dependentresponses increased with increasing age, indicating an awareness of the context in which the possible spellings of phonemes occur. Even the youngest readers, who had received only one year of reading instruction, could apply their knowledge of orthographic regularities to pseudowords. As noted earlier, disabled readers are less likely to regularize the vowels in irregular words. Bryson and Werker (1989) administered a pseudoword reading task to disabled readers to determine whether they would be more likely to read vowels as context dependent. As normal readers gained reading skills, they made more context-dependent responses. Some of the reading-disabled children (those with significantly higher performance than verbal IQ scores) made more contextfree responses than age- and reading level-matched controls. Some of the reading-disabled children did not make context-free errors; however, these children were defined on the basis of below-grade-level scores on a reading comprehension and/or text reading test. As noted earlier, children with low scores on these types of reading tests may not have poor word recognition or decoding skills; therefore, these children may not have been reading disabled in the sense used in the present article. Bryson and Werker noted that poor readers and younger normal readers, when attempting to read double vowels, either sounded out the first letter and ignored the second or sounded out each individual letter. Often, the poor readers sounded out the final silent e, therefore adding a phoneme. They appeared to be reading letter by letter. Seidenberg et al. (1985) found that both poor readers and clinically diagnosed, probably dyslexic readers made more vowel than consonant errors. Most of these errors involved the incorrect lengthening or shortening of the vowel. The more severely disabled readers produced errors that involved substitution of a totally different vowel (e.g., lake for like); poor readers produced mispronunciations of the target vowel on the exception words; good readers tended to regularize them (come pronounced to rhyme with home). The reading disabled and poor readers were less likely to make these kinds of errors. Poor and disabled readers were less likely to regularize a pseudoword that could be pronounced like a regular or an exception word (e.g., naid that could be pronounced to rhyme with said or paid). Using pseudowords, Smiley, Pasquale, and Chandler (1976) also found that poor readers made more errors on vowels, especially long vowels, than good readers. Shankweiler and Liberman (1972) conducted detailed analyses of the errors that were actually made in misreading vowels. Vowels that have many orthographic representations, such as /u/, which is represented by u, o, 00,ou, oe, ew, and ie, were the most difficult to read. Guthrie and Seifert (1977) found that long vowel sounds were learned later than short vowel sounds. What they called special rule word production, with such vowel sounds as infood, join, and bulk, was learned even later. Typically,

80

Linda S. Siegel

the poor readers’ mastery of these complex rules was slower and less adequate than that of the good readers. The increased likelihood of vowel errors does not appear to be a result of inadequate perception of sounds or difficulties with speaking. When children were asked to repeat the words that they had been asked to read, Shankweiler and Liberman (1972) found that fewer errors occurred on vowels than consonants and that the errors were evenly distributed between the initial and final positions. 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 predictable. Children learning to read both English and Hebrew can be tested to compare these two very different orthographies. In a comparison of Englishspeaking children learning to read Hebrew as a second language, we (Geva & Siegel, 1991) found that the incidence of errors in reading vowels was significantly higher in English than in Hebrew. Other children who had!eading 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; 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. 6. Consonants

Consonants in English are more regular than vowels in that particular consonantal phonemes are represented in fewer ways. Consequently, consonants are less likely to be misread. Shankweiler and Liberman (1972) and Fowler et al. (1977) found that consonants in the initial position were more likely to be read correctly than consonants in the final position. (In the Shankweiler and Liberman study, the positions of the vowels and the particular consonants used were not counterbalanced, but this methodological problem was corrected in the Fowler et al. study). The reason for this positional effect is not clear. It could result from guessing a word on the basis of the initial letter rather than trying to apply grapheme-phoneme conversion rules to the word because of poor reading ability


81

and underdeveloped phonological skills. Fowler et al. (1977) noted that the initial segment is easiest to isolate and, unlike the final one, does not require analysis of the syllable. Therefore, children with inadequate phonological skills might be expected to be able to process the first consonant but not the later ones. Consonant errors were closely related to their target sound, but vowel errors were not. For example, b and p were more likely to be substituted for each other than b and s. Consonants with more complex orthographies, that is, the ones that can be represented by more than one letter, were more difficult, but this effect cannot explain the initial-final consonant difference. The error patterns were not the same for vowels and consonants (vowel errors were independent of position, consonant errors were not; vowel errors were not closely related to the target, consonant errors were). The errors evidently do not reflect visual difficulties because visual difficulties should not work differently with vowels and consonants. In addition, visual difficulties do not appear to be characteristic of beginning readers. Word and letter reversals accounted for only a small portion of the errors made in reading words in the Shankweiler and Liberman (1972) study, even though they used lists designed to elicit these errors. Furthermore, sequence reversals such as saw read as was were uncorrelated with letter reversals such as b read as d. However, consonant errors were more common than vowel errors. Werker, Bryson, and Wassenberg (1989) examined the reading of consonants and found that both disabled and normal readers made more phonetic feature substitution errors than orientation reversal substitutions. Also, children with a reading disability made more consonant addition errors. Most errors were not reversal errors. Although some reversals are found in young children regardless of reading ability (Taylor, Satz, & Friel, 1979; Vellutino, Steger, & Kandel, 1972), these reversal errors may be linguistic rather than perceptual because reversals of orientation (b read as d)are not correlated with reversals of sequencing (was-saw),reversals occur with words but not with single letters presented tachistoscopically, and consonants are confused when they differ by a single phonetic feature regardless of visual similarity. Seidenberg et al. (1985) found that disabled readers make more substitution errors (belt for best) and insertion errors (grave for gave) than slow readers, who make more errors than normal readers. Werker et al. noted that Seidenberg et al. confounded phonetic feature and orientation reversal substitutions by calling them both reversals (deed for bee4 and inversions (deed for deep). Werker et al. studied orientation reversal errors in which one letter was read as another differing in left/right or up/down orientation, such as b ford, and phonetic feature errors in which one letter was misread as another differing in a single phonetic feature such as voicing b versus p and place of articulation (b and d are both voiced but b is bilabial and d is alveolar). They found that normal and disabled readers were equally likely to make orienta-

82

Linda S. Siegel

tion reversal errors. All groups made more phonetic feature than orientation reversal errors. Therefore, errors were the result of phonetic and not visual similarities. The order of types of errors was phonetic > addition > omission > sequencing. The reading-disabled children made more errors that involved adding a consonant than normal readers. The normal readers made more phonetic feature substitutions than any other type of error. Disabled readers seemed to be reading letter by letter. The most common type of addition error involved homorganic errors, that is, closing a syllable with the consonant sound already existing, e.g., up to pap. Reading-disabled, not normal, readers made these errors. Intrasyllable additions, reading ope as olpe, were less common but did occur especially among the disabled readers and typically involved the addition of the liquids, r and 1. Werker et al. speculated that errors result from knowledge of individual letters but that the disabled readers have trouble knowing and retrieving the rules when they must combine them. In addition, they may rely on articulatory information when sounding out words so that they retrieve the pronunciation of letters that are close in place of articulation to the target letter. Smiley et al. (1976) found that disabled readers made more errors on the variable consonants (e.g., c and g ) . The reading-disabled group had particular difficulty with the s pronunciation of c, the j pronunciation of g, the initial ch sound, and two-syllable words ending in y . The good readers made more plausible (similar to the correct answer) errors than poor readers. 7 . Analogy versus Rules Other kinds of tasks have been used to measure the development of the understanding of grapheme-phoneme conversion rules. The reading of pseudowords that can be read by analogy or by grapheme-phoneme rules, such as puscle, fody, and risten, has been studied (Manis, Szeszulski, Howell, & Horn, 1986; Siegel & Geva, 1990). For example, puscle can be pronounced as if it rhymed with muscle or with the cl pronounced, and fody can be pronounced like body or with a long 0.Children with a reading disability had a great deal of difficulty with these pseudowords. The children with reading disabilities were significantly less able than normal readers of the same chronological age to read these words correctly. Even when matched with normal readers of the same reading level, the disabled readers made significantly more errors than the normal readers. Compared with the normal readers, the younger children with a reading disability were significantly less likely to use a rule-based strategy and more likely to use an analogy strategy. This pattern suggests a greater reliance on the visual route. 8 . Other Phonological Skills Pseudoword reading is not the only task that distinguishes poor from normal readers. Another task is the spelling of pseudowords. Obviously, pseudowords


83

can be spelled only by using phoneme-grapheme conversion strategies as no lexical entry exists. Disabled readers had significantly lower scores on a task that involved the spelling of pseudowords, even when the disabled readers were at the same reading level as younger normal readers (Siegel & Ryan, 1988). One type of evidence of phonological processing skills is the use of phonological recoding in short-term memory such that rhyming (confusable) stimuli are more difficult to remember than nonrhyming stimuli. A number of studies have shown that younger poor readers are less disrupted by rhyming stimuli (e.g., Byme & Shea, 1979; Mann, Liberman, & Shankweiler, 1980; Shankweiler, Liberman, Mark, Fowler, & Fischer, 1979; Siegel & Linder, 1984). Johnston (1982) and Siegel and Linder (1984), however, found that older dyslexic children do show phonetic confusability, although their short-term memory for letters was significantly poorer than that of age-matched controls. This latter finding is not surprising as phonological recoding skills are likely to be involved in any verbal memory task and the reading-disabled’s poor verbal memory may be a function of inadequate phonological abilities. Performance on a variety of phonological tasks distinguishes disabled from normal readers. Children with reading disabilities were slower than normal readers in deciding whether two aurally presented words rhymed, presumably because of inadequate use of phonological recoding in memory (Rack, 1985). Phonemic awareness, the ability to recognize the basic phonemic segments of the language, is obviously an important component of phonological processing. Difficulties with phonemic awareness predict subsequent reading problems (e.g., Bradley & Bryant, 1983; Wallach & Wallach, 1976). Poor readers also have deficits in phonological production tasks, for example, naming objects represented by multisyllable words and repeating multisyllable words and difficult phrases with alliteration. Pratt and Brady (1988) found differences between good and poor readers on the ability to segment words into phonemes and delete sounds from words. Good readers were more accurate in judging the length of a word or pseudoword. Good readers were more disrupted than poor readers by misspellings in text that were phonologically inappropriate (robln for robin), indicating that the good readers were using phonological cues (Snowling & Frith, 1981). Children with a reading disability also have difficulty recognizing the visual code of sounds (Siegel & Ryan, 1988). In the Gates McKillop test, children hear pseudowords such as wiskute and are asked to select the correct version of the word from among four printed choices: iskute, wiskay, wiskate, and whestit. Poor readers had significantly lower scores than normal readers on this task. Although this task involves skills that are relevant to spelling, aspects of it are relevant to phonological processing, including the segmentation involved in analyzing the pseudoword and in decoding the alternatives.

84

Linda S. Siegel

9. Development of Phonological Skills in Other Languages We have been discussing only English up to this point. Children who have difficulty learning to read Portuguese have difficulty reading pseudowords (Da Fontoura & Siegel, 1991) and children learning Hebrew as a second language also have difficulty with pseudowords (Geva & Siegel, 1991). English is an alphabetic language with a significant amount of irregularity; Chinese is a morphemic orthography in which the characters have meaning and in which phonological information about pronunciation is sometimes coded in a character but is not essential. Even Chinese (Cantonese) children with reading problems have difficulty with tone and rhyme discrimination and have significantly lower scores than normal readers on tasks measuring these phonological skills (So & Siegel, 1991). C . SYNTACTIC AWARENESS

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 disabled and normal readers using an oral cloze task, a sentence correction task, and the Grammatical Closure subtest of the Illinois Test of Psycholinguistic Abilities. 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 are Jane -her sister ran up the hill; Betty -a hole with her shovel; and The girl -is tall plays basketball. In the sentence correction task, a sentence that is syntactically incorrect is read aloud to the child, who is then required to correct the sentence. Examples include Animal are kept in zoos; Can you read them book?; and It was very cold outside tomorrow. In the ITPA Grammatic Closure subtest, the child is required to supply the missing word in a sentence that is read aloud while the examiner points to pictures illustrating the sentence. For example, Here the thief is stealing the jewels. Here the jewels have been -. In this example, the child must understand the irregular past tense of the verb to steal to supply the correct word. When the disabled and normal readers were compared on these three tasks, the children with a reading disability performed at a level that was significantly lower than that of the normal readers up to the age of 12. After that point, the performance of both groups was close to 100%. More difficult tasks might have yielded differences between the older reading-disabled and the normal readers but the differences were certainly significant in the elementary school years. Brittain (1970) found that performance on a test of the production of morphology (similar to the ITPA Grammatic Closure) was related to reading ability in Grade 1 and 2 children. Other evidence suggests that children with reading problems have difficulty with syntactic awareness. Guthrie (1973) found that disabled readers performed at a lower level than both chronological age- and reading level-matched normal


85

readers on a reading cloze task that measured syntax comprehension, even though the disabled re.jders had an adequate sight reading vocabulary to perform this task. Although reatling-disabled children were not studied, Goldman (1976) found that the understanding of complex syntax (sentences such as John tells Bill to bake the cake and John promises Bill to bake the cake) was related to performance on a reading comprehension test. Cromer and Wiener (1966) found that poor readers made more errors than normal readers that indicated a lack of awareness of syntax on text reading tasks. Glass and Perna (1986) found that performance .on an oral language sentence comprehension test was p r e r for children with a reading disability than for normal readers. Willows and Ryan (1981) found that less skilled readers were not as accurate as normal readers at substituting a missing word in a reading cloze procedure. Although difficulties in the processing of syntax may be an artifact of working memory problems, this possibility is relatively unlikely as we have found that reading-disabled children, except at the ages of 7 to 8, are as likely as normal readers to show correct verbatim recall of sentences of varying length and grammatical complexity (Siegel & Ryan, 1988). Byrne (1981) has also shown that poor readers had more difficulty than good readers only with certain types of syntactic structures; the complexity of sentence structure, not the length of the sentence, was a determinant of performance. Some evidence from other languages indicates that children with reading difficulties experience syntactic difficulties. Children with reading problems in Chinese (Cantonese) demonstrated poorer performance in an oral cloze test involving syntactic awareness of Chinese (So & Siegel, 1991). Similar results were found for Canadian children who spoke Portuguese as a first language, received instruction in reading in English, and attended a Portuguese Heritage Language Program in Portuguese (Da Fontoura & Siegel, 1991). The children who had low scores on Portuguese word and pseudoword reading tests had significantly lower scores on Portuguese oral cloze than children who were good readers of Portuguese. Testing native speakers of Hebrew, Bentin, Deutsch, and Liberman (1990) found that disabled readers in Hebrew were less accurate at judging whether the syntax of a sentence was correct and correcting a sentence with incorrect syntax. In addition, good readers were more influenced by context in identifying unclear words and made more errors than disabled readers that involved substituting a syntactically correct word but one that was not the word they had heard. D. WORKING MEMORY

Working memory is the ability to retain information in short-term memory while processing incoming information. In reading, working memory means the decoding or recognizing of words or phrases while remembering what has been

86

Linda S. Siegel

read. Siegel and Ryan (1989a) 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 two, three, four, or five 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, for example, 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 readers. Similar difficulties with working memory have been noted in Chinese (So & Siegel, 1991), Hebrew (Geva & Siegel, 1991), and Portuguese (Da Fontoura & Siegel, 1991). E. SEMANTIC PROCESSING

The three basic cognitive processes described in Sections 11, B-D are important for the development of reading skill and are significantly disrupted in disabled readers. Two other processes, semantic and orthographic, are also involved in reading, but children with reading disabilities do not seem to experience the same degree of difficulties with these processes as with the preceding three. I . Reading Errors Two types of analyses indicate that the semantic processing skills of poor readers are relatively intact. One type is analysis of errors made in word reading tasks and the other is analysis of sentence processing. The analysis of errors made in reading single words can reveal important information about the reading process. A number of studies indicate that some children with very severe reading problems make semantic errors in the reading of single words. An important point is that these errors are made in reading single words with no context cues. Johnston (1982) reported the case of an 18-year-old girl who made semantic errors such as down read as up, chair read as table, and ofJice read as occupation, and who could not read any pseudowords. I have shown that a small group of children with reading disabilities make semantic substitutions while reading single isolated words (Siegel, 1985). All these children had very poor, or nonexistent, phonological processing skills and were unable to read a single pseudoword. These types of semantic errors indicate that phonological processing is not used at all because none of the sounds implicit in the stimulus word are produced in the response. In addition, the printed equivalent of the response is not visually similar to the target word; however, this type of error indicates that some semantic processing is occurring and that, although the word is not being


87

read correctly, some semantic information is being processed. This type of error is made only in the early stages of reading acquisition. Normal readers do not appear to make this type of error. The types of errors that normal readers typically make involve the substitution of a visually and/or phonologically similar word (e.g., look as book, chicken as children, away as way). Temple (1988) reported the case of a 9-year-old poor reader who could not read pseudowords and who made some semantic substitutions when reading single words, such as eye read as blue and mother read as mommy. Temple, among others, argued that these errors may have been due to chance. This explanation seems unlikely for several reasons. Normal readers do not make these errors. The substitutions all make sense in terms of having similar meaning and no pairings are random. Given the total speaking vocabulary of 10,000 to 20,000 words of children this age, these particular errors seem very unlikely to occur by chance. In the one report of semantic errors in single word reading among Frenchspeaking children, Sprenger-Charolles (1991) administered a task in which children were required to read words or pseudowords that were attached to pictures. Some pictures were correctly named; others were given a name related to the correct name but not synonymous (e.g., limace ‘slug,’ was written under a picture of a snail); and others were given pseudoword names that differed in a single letter from the real name (e.g., falise instead of valise or pantalin instead of pantalon). The children were required to say whether or not the correct name was attached to the picture. Semantic errors (e.g., focobotive read as train, binyclette, a nonword similar to the real word bicyclette, read as velo ‘bike’) were q.uite common for a group of poor readers, average age 10, but virtually never occurred in the group of good readers. Normal readers at the earliest stages of reading may sometimes appear to make these semantic errors. Seymour and Elder (1986) studied 4f-to 54-year-old children who had received reading instruction that emphasized a sight vocabulary and that did not involve systematic instruction in grapheme-phoneme conversion rules. When reading single words, these children made semantic errors such as boat read as yacht, milk read as tea, little read as wee. Thus, semantic coding of words appears to be the first aspect of words to be acquired, and semantic coding will be used if the child lacks an understanding of spelling-sound correspondences either because these correspondences have not been taught or because they have not been acquired because of cognitive factors, as in reading disability. These types of errors indicate that grapheme-phoneme conversion rules are not being used at all and that phonological processing is virtually nonexistent. Other evidence exists of the accuracy of semantic processing in disabled readers. Frost (1991) found that dyslexics could respond as quickly and as accurately as normal readers when required to make decisions about whether two

88

Linda S. Siege1

words belonged to the same semantic category, but were significantly slower on a phonological task that involved making a decision about whether two orthographically dissimilar words rhymed. 2 . Sentence Processing Skills involved in processing the semantic aspects of sentences appear to be adequate in children with a reading disability. In the sentence correction task described earlier, some of the sentences were syntactically correct but meaningless. Examples are There areflowersflying in the garden; In the summer, it snows; and The moon is very big and bright in the morning. The reading-disabled did not have any difficulty correcting these sentences and performed at a level similar to that of the normal readers. This finding contrasts with their performance on sentences where the correction of syntax was required. Therefore, the children with reading disabilities have a deficit in the processing of syntactic information, but this deficit does not extend to processing of semantic information. Lovett (1979) found that reading competence in young readers was not related to the ability to remember the semantic aspects of what had been read. Lovett required children to read short passages and then to recognize whether a sentence had been in the passage when the sentence was identical or differed slightly in semantic, syntactic, or lexical context. The children at all reading levels were easily able to recognize changes in the semantic content, were less able to recognize syntactic changes, and had much more difficulty in recognizing lexical changes (e.g., picked up changed to lifed up). Even when the children were required to read material between reading the sentence and remembering it, semantic information remained available, but syntactical and lexical information was less so. These data indicate that semantic processing is primary for reading and at the earliest stages, or with disabled readers, semantic processing is operating even when other processes are much less efficient. Waller ( 1976) studied good and poor readers and found that poor readers were as likely as good readers to remember many of the semantic aspects of what they had read, but were less likely to remember whether a lexical item was singular or plural and whether a past or present tense was used. This pattern of errors indicates relatively intact semantic processing but difficulties with the syntactic processing as was discussed in Section II,C. Some evidence indicates that reading-disabled children may even be superior to normal readers in their use of semantic context. Frith and Snowling (1983) administered a task in which reading-disabled and normal readers, matched on reading level, were required to read sentences with homographs (with the correct pronunciation) such as He had a pink bow and He made a deep bow. The performance of the reading-disabled children was superior to that of the normal readers, indicating that the disabled readers were better able to make use of semantic/syntactic cues.


89

F. ORTHOGRAPHIC PROCESSING

Another aspect of reading is orthographic processing, which is discussed in the context of dual-route theories of reading. Olson, Kliegl, Davidson, and Foltz (1985) have developed two tasks that provide a direct contrast of the visual (orthographic) and phonological processing routes. In the visual task, the child is shown a real word and a pseudoword (e.g., ruin-rane, bod-bowl) and has to select the correct spelling. In the phonological task, the child has to specify which of two pseudowords, presented visually, sounds like a real word (e.g., kake-duke, joup-jouk). Each of these tasks is designed so that only one process can operate. That is, in the visual task both choices sound exactly the same, so that visual memory for the orthography of a word must be used; phonological processes are not helpful in this case because sounding out the words would produce the identical response to each word. For the phonological task, recall of the visual pattern would not be useful because neither alternative is a correct orthographic pattern in the English language; however, one of the alternatives, when sounded out, does produce an English word, although it is obviously not the correct orthographic form. These tasks were administered to disabled and normal readers, aged 7 to 16 years. Not surprisingly, the disabled readers performed more poorly on the phonological task than age-matched and reading level-matched normal readers and did not catch up to the normal readers until the age of 13. They also performed more poorly on the visual task than age-matched normal readers until age 13; however, the disabled readers performed at a significantly higher level on the visual task than did the reading level-matched normal readers at reading level 2. This finding indicates good visual memory skills in the disabled readers relative to their level of word reading. It indicates that the reading-disabled were paying attention to the visual aspects of a word rather than the phonological aspects. Another aspect of the awareness of orthographic structures is the ability to recognize legal and illegal orthographic combinations of English letters. Siegel, Geva, and Share (1990) developed a task to assess this ability. Children were shown 18 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-jlk, moke-moje, vism-vim, and pod-lowp. This task was administered to disabled and normal readers, aged 7 to 16 years. The performance of the poor and normal readers did not differ except at the youngest ages. At 7 to 8, the reading-disabled children made significantly more errors than normal readers of the same chronological age, but an important point is that the reading-disabled children did not perform more poorly than agematched normal readers at ages 9 to 16. When matched on reading level, how-

90

Linda S. Siegel

ever, the disabled readers performed at a significantly higher level than the normal readers. Therefore, in comparison to the data on phonological processing, the orthographic processing of the reading-disabled is quite good. These data indicate that orthographic processing is not as impaired in the readingdisabled as is phonological processing. These data indicate that semantic and orthographic processing occurs in reading, but that the use of these processes can disrupt normal reading and cause errors. The preceding discussion has been based on what might be called orthographic awareness skills. Some evidence suggests that disabled readers are more sensitive to the visual aspects of printed stimuli than better readers. For example, Steinhauser and Guthrie (1974) found that poor readers were fasfer than good readers of the same reading level on a task that involved circling individual letters in a text; however, poor readers were worse than good readers when required to circle phonemes. A visual matching procedure can be used to circle individual letters, but phonemes probably require some phonological coding. These data suggest that reading-disabled individuals are paying attention to the visual aspects of printed stimuli, but because of differences in phonological skills, they have more difficulty with these aspects of print. Snowling (1980) also found that children with a reading disability were more accurate than normal readers of the same reading level on a task involving selecting the visual form of an aurally presented pseudoword. This superiority of the reading-disabled group occurred only at the lowest reading level studied (age 7) and was similar to what we found (Siegel et al., 1990); however, the reading disabled children performed significantly more poorly than reading level-matched normal readers on a task involving recognition of the auditory form of a visually presented pseudoword. Obviously, this latter task involves phonological processing skills and the auditoryvisual task relies on visual skills that are operating normally, or perhaps in a superior manner. The reading-disabled children did not differ from normal readers in the auditory-auditory task in which they had to judge whether two aurally presented pseudowords were the same or different, so the difficulties of poor readers were not due to problems in auditory discrimination. The reading-disabled did not show an improvement with age on the visualvisual task, but the normal readers did, suggesting that the disabled readers did not use a phonemic code in the visual matching task and that the normal readers were probably converting the visual stimuli to a phonemic code. The normal readers performed at the same level on the visual-visual, auditory-visual, and visual-auditory tasks; however, the reading-disabled children performed significantly better on the visual-visual task than on the two crossed-modality tasks, suggesting again that the visual stimuli ( pseudowords) were not phonologically recoded. All the studies imply that the direct or visual access route is relatively intact in the reading-disabled, but that the phonological route is impaired. Evidence from adults with a reading disability indicates that phonemic coding


91

does not occur, at least not to the same extent as in normal readers. We (Shafrir & Siegel, 1991) found that reading-disabled adults reported using a visual scanning strategy, rather than phonological recoding, in reading tasks that involved matching words and pseudowords. The reading-disabled adults who did use a phonological recoding strategy in the word task showed significantly longer latencies than those who used a phonological recoding strategy, suggesting that the visual strategy may be more efficient for disabled readers. Evidence from spelling tasks indicates that reading-disabled adults have an adequate knowledge of English orthography and, in some cases, a greater degree of knowledge than normal readers. Pennington et a]. (1986) scored the spelling errors of reading-disabled and normal reading adults according to a simple system in which any orthographically illegal sequence occurred, for example, ngz in angziary for anxiery, and a complex system in which errors indicating a lack of knowledge of more subtle aspects of orthography were scored, for example, knowing that vowel clusters can be represented by one vowel (iou in precious is the sound of /u/) or knowing that phys occurs in many words (e.g., physics, physician) and represents the same sound in all of them. The reading-disabled and normal readers did not differ in the preservation of simple orthographic features; however, the reading disabled were significantly more accurate in the complex aspects of English orthography than normal readers of the same spelling level. We (Lennox & Siegel, 1991) found that the spelling errors of children who were poor spellers were more similar visual matches to the correct word than were those of good spellers of the same spelling age; however, the misspellings of poor spellers were less phonologically accurate than those of good spellers of the same spelling age. These findings indicate that the poor spellers were more likely to use visual memory than phonological strategies in spelling. These results suggest that individuals with a reading disability may be able to compensate for their difficulties in phonological processing. Rack (1985) found that reading-disabled children make use of an orthographic code in memory. Reading-disabled and normal readers, ages 8 to 14, were presented four lists of words to learn. The words in a list were orthographically similar and rhyming (e.g., farm-harm), orthographically similar and not rhyming (e.g.,farm-calm), orthographically dissimilar and rhyming (e.g., farm-warm), and orthographically dissimilar and not rhyming (e. g., farm-pond). Whether the presentation was visual or auditory, orthographic similarity improved the performance of reading-disabled more than normal readers, indicating that the disabled readers were more sensitive to orthographic effects. Phonetic similarity did not predict recall for the disabled readers but it did for the normal readers. Reading-disabled children remembered more orthographically similar targets than the normal readers and fewer rhyming targets, indicating that they were making more use of an orthographic rather than a phonetic code. Normal readers of the same reading age

92

Linda S. Siege1

did not show this effect. Reading-disabled children took longer to say yes for rhyming pairs that were orthographically dissimilar (farm-calm) than for those that were orthographically similar (head-lead). Reading level-matched normal readers did not show this effect. Katz (1977) found that poor readers were not as accurate as good readers in recognizing which serial position an individual letter occurred in most frequently. In this study, good and poor readers were shown two pseudowords, one containing a letter in its most frequent serial position and the other containing the letter in its least frequent serial position. Poor readers made more errors than good readers. Thus, poor readers had less orthographic knowledge about single letters, in contrast to groups of letters, than good readers.

IV . Conclusions During the period of rapid acquisition of reading skills, three processesphonological, syntactic, and working memory-show significant increases in development. These processes are significantly disrupted in children who are reading disabled, but semantic and orthographic processes are not disrupted to the same extent; however, the underutilization of phonological processing and the reliance almost entirely on semantics and orthographics or visual processes disrupt reading. Deficits in three fundamental cognitive processes-phonological processing, syntactic awareness, and working memory-constitute the basic characteristics of reading disability.

ACKNOWLEDGMENTS The preparation of this article and the research reported in it were supported by the Natural Sciences and Engineering Research Council of Canada. This article was prepared while the author held a Senior Research Fellowship from the Ontario Mental Health Foundation. The author thanks the Max Planck Institute for Rycholinguistics, Nijmegen, The Netherlands, for hospitality during the preparation of this article and Hayne Reese for his very constructive suggestions and editorial assistance. The author also thanks Letty Guirnela for secretarial assistance.

REFERENCES Backman, J., Bruck, M., Hebert, M., & Seidenberg, M.(1984). Acquisition and use of spellingsound correspondences in reading. Journal of Experimenral Child Psychology, 38, 114-133. Backman, J. E., Mamen, M., &, Ferguson, H. B. (1984). Reading level design: Conceptual and methodological issues in reading research. Psychological Bulletin, 96, 560-568.

Developmenr of Reading

93

Barwick, M., & Siegel, L. S. (1990). The incidence and nature of learning disabilities in runaway, homeless youth. Unpublished manuscript. Bentin, S . , Deutsch, A., & Liberman, 1. Y. (1990). Syntactic competence and reading ability in children. Journal of Experimental Child Psychology. 48, 147- 172. Besner, D., Twilley, L., McCann, R., & Seergobin, K. (1990). On the association between connectionism and data: Are a few words necessary? Psychological Review, 97, 1-15. Bradley, L., & Bryant, P. E. (1983). Categorizing sounds and learning to read: A causal connection. Nature. 301, 41 9-42 1 , Brittain, M. M. (1970). Inflectional performance and early reading achievement. Reading Research Quarterly, I , 34-48. Brown, A. D. (1987). Resolving inconsistency: A computational model of word naming. Journal of Memory and Language, 26, 1-23. Bruck, M. (1988). The word recognition and spelling of dyslexic children. Reading Research Quarterly, 23. 51-68. Btuck, M. ( 1990). Word-recognition skills of adults with childhood diagnosis of dyslexia. Developmental Psychology, 26. 439-454. Bryson, S. E., & Werker, J. F. (1989). Towards understanding the problem in severely disabled readers. Part 1: Vowel errors. Applied Psycholinguisrics, 10, 1-12. Byme, B. (1981). Deficient syntactic control in poor readers: Is a weak phonetic memory code responsible? Applied Psycholinguisrics, 2 , 201-21 2. Byme, B., & Shea, P. (1979). Semantic and phonemic memory in beginning readers. Memory and Cognition, 7 , 333-341. Calfee, R. C., Lindamood, P., & Lindamood, C. (1973). Acoustic-phonetic skills and reading: Kindergarten through twelfth grade. Journal of Educarional Psychology, 64, 293-298. Coltheart, M. (1978). Lexical access in simple reading tasks. In G. Underwood (Ed.), Strategies of information processing ( pp. I51-2 16). London: Academic Press. Cromer, W., & Wiener, M. (1966). Idiosyncratic response patterns among good and poor readers. Journal of Consulting Psychology, 30. 1- 10. Da Fontoura, H., & Siegel, L. S. (1991). English and Portuguese: Language and reading skills in bilingual Portuguese-Canadian children. Unpublished manuscript. Daneman, M., & Carpenter, P.A. (1980). Individual differences in working memory and reading. Journal of Verbal Learning and Verbal Behavior, 19, 450-466. Doctor, E. A., & Coltheart, M. (1980). Children’s use of phonological encoding when reading for meaning. Memory and Cognirion, 8 , 195-209. Ehri, L. C., & Wilce, L. S. (1980). The influence of orthography on readers’ conceptualization of the phonemic of words. Applied Psycholinguistics, I , 371-385. Ehri, L. C., & Wilce, L. S. (1983). Development of word identification speed in skilled and lessskilled beginning readers. Journal of Educational Psychology. 75, 3- 18. Ellis, A. W. (1985). The cognitive neuropsychology of developmental (and acquired) dyslexia: A critical survey. Cognitive Neuropsychology. 2 . 169-205. Forster, K. I . , & Chambers, S . (1973). Lexical access and naming time. Journal of Verbal Learning and Verbal Behavior. 12, 627-635. Fowler, C . , Liberman, I., & Shankweiler, D. (1977). On interpreting the error pattern in beginning reading. Language and Speech, 20, 162-173. Fowler, C., Shankweiler, D., & Liberman, 1. (1979). Apprehending spelling patterns for vowels: A developmental study. Language and Speech, 22, 243-251. Frith, U., & Snowling, M. (1983). Reading for meaning and reading for sound in autistic and dyslexic children. British Journal of Developmental Psychology. I , 329-342. Frost, L. (1991). Hemispheric assymetries for phonological and semanric processing in dyslexic and average readers. Unpublished manuscript.

94

Linda S. Siegel

Geva, E., & Siegel, L. S. (1991). The role of orthography and cognitive factors in the concurrent developmenr of basic reading skills in bilingual children. Unpublished manuscript. Glass, A. L., & Perna, J. (1986). The role of syntax in reading disability. Journal of Learning Disubiliiies, 19, 354-359. 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. Goldman, S. R. (1976). Reading skills and the minimum distance principle: A comparison of listening and reading comprehension. Journal of Experimental Child Psychology, 22, 123- 142. Gough, P. B., & Juel, C. (1991). The first stages of word recognition. In L. Rieben & C. A. Perfetti (Eds.). Learning to read: Basic research and its implications (pp. 47-56). Hillsdale, NJ: Erlbaum. Gough, P. B., & 'hnmer, W. E. (1986). Decoding, reading, and reading disability. Remedial and Special Education, 7,6-10. Guthrie, J. T.(1973). Models of reading disability. Journal of Educational Psychology, 65, 9-18. Gutluie, J. T., & Seifert, M. (1977). Letter-sound complexity in learning to identify words. Journal of Educational Psychology, 69, 686-696. Harris, M., & Coltheart, M. (1986). Language processing in children and adults: An introduction. London: Routledge & Kegan Paul. Hogaboam, T. W., & Perfetti, C. A. (1978). Reading skills and their role of verbal experience in decoding, Journal of Educational Psychology. 5, 717-729. Humphreys, C. W., & Evett, L. J. (1985). Are there independent lexical and nonlexical routes in the word processing? An evaluation of the dual-route theory of reading. The Behavioral and Brain Sciences, 8, 689-740. Johnston, R. (1982). Phonological coding in dyslexic readers. Brirish Journal of Psychology, 73, 455-460. Jom, A. F. (1979). The nature of reading deficit in developmental dyslexia: A reply to Ellis. Cognition, I , 421-433. Katz, L. (1977). Reading ability and single-letter orthographic redundancy. Journal of Educational Psychology, 69, 653-659. Lennox, C., & Siegel, L. S.(1991). The development of undersranding ofphonological rules in good and poor spellers. Unpublished manuscript. Liberman, I. Y.,Liberman, A. M., Mattingly, I. C., & Shankweiler, D. (1980). Orthography and the beginning reader. In J. Kavanagh & R. Venezky (Eds.), Orthography, reading and dyslexia (pp. 137-153). Baltimore: University Press. Lindgren, S . D., De Renzi, E., & Richman, L. C. (1985). Cross-national comparisons of developmental dyslexia in Italy and the United States. Child Development, 56, 1404-1417. Lovett, M.W.(1979). The selective encoding of sequential information in normal reading development. Child Development, 50, 897-900. Manis, F. R., & Morrison, F. J. (1985). Reading disability: A deficit in rule learning? In L. S. Siegel & F. J. Morrison (Eds.), Cognitive development in arypical children: Progress in cognitive development research ( pp. 1-26). New York: Springer-Verlag. Manis, F. R., Savage, P. L., Momson, F. J., Horn, C. C., Howell, M. J., Szesulski, P. A,, & Holt, L. K. (1987). Paired associate learning in reading-disabled children: Evidence for a rule-learning deficiency. Journal of Experimental Child Psychology. 43, 25-43. Manis, F. R., Szeszulski, P. A., Howell, M. J., & Horn, C. C. (1986). A comparison of analogy- and rule-based decoding strategies in normal and dyslexic children. Journal of Reading Behavior, 18, 203-213. Mann, V. A., Liberman, I. Y.,& Shankweiler, D. (1980). Children's memory for sentences and word strings in relation to reading ability. Memory and Cognition, 8, 329-335. Metsala, J., & Siegel, L. S. (in press). Developmental dyslexia and hyperlexia: Attributes and


95

underlying reading processes. In S. Segalowitz & 1. Rapin (Eds.),Developmental neuropsychology. Amsterdam: Elsevier.

Meyer, D. E., Schvanevelt, R. W., & Ruddy, M. G. (1975). Loci of contextual effects on word recognition. In P. M. A. Rabbitt & S . Dornic (Eds.), Attention andperformance (Vol. 5, pp. 98118). New York Academic Press. Olson, R. K., Kliegl, R., Davidson, B. I., & Foltz, G. (1985). Individual and developmental differences in reading disability. In G. E. MacKinnon & T. G. Waller (Eds.), Reading research: Advances in theory and practice (Vol. 4, pp. 1-64). New York: Academic Press. Pennington, B. F., McCabe, L. L., Smith, S. D., Lefly, D. L., Bookman, M. O., Kimberling, W. J., & Lubs, H. A. (1986). Spelling errors in adults with a form of familial dyslexia. Child Development, 57, 100-1013. Pratt, A. C., & Brady, S. (1988). Relations of phonological awareness to reading disability in children and adults. Journal of Educational Psychology, 80, 3 19-323. Rack, J. P. (1985). Orthographic and phonetic coding in developmental dyslexia. British Journal of Psychology, 76, 325-340. Seidenberg, M. S., Bruck, M., Fornarolo, G., & Backman, J. (1985). Word recognition processesof poor and disabled readers: Do they necessarily differ? Applied Psycholinguisrics, 6, 161- 180. Seidenberg, M. S., & McClelland, J. L. (1989). A distributed developmental model of wordrecognition and naming. Psychological Review. %, 523-568. Seymour, P. H. K., & Elder, L. (1986). Beginning reading without phonology. Cognitive Neuropsychology, 3, 1-36. Seymour, P. H. K., & Porpodos, C. D. (1980). Lexical and nonlexical processing of spelling in dyslexia. In U. Frith (Ed.),Cognitiveprocesses in spelling (pp. 443-473). New York Academic Press. Shafrir, U., & Siegel, L. S. (1991). Cognitive processes of subtypes of adults with learning disabilities. Unpublished manuscript. Shankweiler, D., & Liberman, I. (1972). Misreading: A search for causes. In J. Kavanaugh & 1. Mattingly (Eds.), Language by ear and by eye: The relationship beween speech and reading (pp. 293-317). Cambridge, MA: MIT Press. Shankweiler, D., Liberman, I. Y.,Mark, L. S., Fowler, C. A,, & Fischer, F. W. (1979). The speech code and learning to read. Journal of Experimental Psychology: Human Learning and Memory, 5 , 531-545. Siegel, L. S. (1985). Psycholinguistic aspects of reading disabilities. In L. S. Siegel & F. J. Morrison (Eds.), Cognirive development in atypical children. New York: Springer-Verlag. Siegel, t.S. (1988). Evidence that IQ scores are irrelevant to the definition and analysis of reading disability. Canadian Journal of Psychology, 42, 201-215. Siegel, L. S. (1991a). Dyslexics vs. poor readers: Is there a diference? Unpublished manuscript. Siegel, L. S . (1991b, July). Phonological processing, working memory and syntactic awareness and rheir relationship to reading. Paper presented at the International Conference on Memory. Lancaster, UK. Siegel, L. S. (in press). Phonological processing deficits as the basis of a reading disability. Developmental Review. Siegel, L. S., & Faux, D. (1989). Acquisition of certain grapheme-phoneme correspondences in normally achieving and disabled readers. Reading and Wrifing:An Interdisciplinary Journal, 1 , 37-52.

Siegel, L. S., & Geva, E. (1990). Reading by analogy or rules: A comparison of poor and normal readers. Unpublished manuscript. Siegel. L. S . . Geva, E., & Share, D. (1990). The development of orthographic skills in normal and disabled readers. Unpublished manuscript. Siegel, L. S . , & Heaven, R. K. (1986). Categorization of learning disab es. In S. J. Ceci (Ed.).

96

Linda S . Siegel

Handbook of cognitive, social and neuropsychological aspects of learning disabilities (Vol. 1. pp. 95-121). Hillsdale, NJ: Erlbaum. Siegel, L. S., Levey, P., & E m s , H.(1985). Subtypes of developmental dyslexia: Do they exist? In F. J. Morrison, C. Lord, & D. P. Keating (Eds.), Applied developmentalpsychology (Vol. 2, pp. 169-190). New York: Academic Press. Siegel, L. S., & Linder, B. A. (1984). Short-term memory processes in children with reading and arithmetic learning disab es. Developmental Psychology, 20, 200-207. Siegel, L. S., & Metsala, J. (in press). An alternative to the food processor approach to subtypes of learning disabilities. In N. N. Singh & 1. Beale (Us.), Current perspectives in learning disabilities: Nature. theory, and treatment. New York: Springer-Verlag. 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. (1989a). Subtypes of developmental dyslexia: The influence of definitional variables. Reading and Writing: An Interdisciplinary Journal, It 257-287. Siegel, L. S., & Ryan, E. B. (1989b). The development of working memory in normally achieving and subtypes of learning disabled children. Child Development, 60, 973-980. Smiley, S. S., Pasquale, F. L., & Chandler, C. L. (1976). The pronunciation of familiar, unfamiliar and synthetic words by good and poor adolescent readers. Journal of Reading Behavior, 8(3), 289-297. Snowling, M. J. (1980). The development of grapheme-phoneme correspondence in normal and dyslexic readers. Journal of Experimental Child Psychology, 29. 294-305. Snowling, M . , & Frith, U. (1981). The role of sound, shape, and orthographic cues in early reading. Brirish Journal of Psychology, 72, 83-87. So, D.,& Siegel, L. S. (1991). Learning to read Chinese: The developmnt of language, phonological and short-term memory skills in Cantonese. Unpublished manuscript. Sprenger-Charolles, L. (1991). Word-identification strategies in a picture context: Comparisons between “ g d ” and “poor” readers. In L. Rieben & C. A. Perfetti (Eds.), Learning to read: Basic research and its implications (pp. 175-188). Hillsdale, NJ: Erlbaum. Stanovich, K. E. (1982a). Individual differences in the cognitive processes of reading: I. Word decoding. Journal of Learning Disabilities, 15, 485-493. Stanovich, K. E. (1982b). Individual differences in the cognitive processes of reading: 11. Text-level processes. Journal of Learning Disabilities, 15, 549-554. 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. Stanovich, K. E. (1988~).Speculations on the causes and consequences of individual differences in early reading acquisition. In P. Gough (Ed.), Reading acquisition (pp. 1-60). Hillsdale, NJ: Erlbaum. Stanovich, K. E. (in press). Word recognition: Changing perspectives. In P. D. Pearson (Ed.), Handbook of reading research (Vol. 2). Stanovich, K. E., Nathan, R. G., & Zolman, J. E. (1988). The developmental lag hypothesis in reading: Longitudinal and matched reading-level comparisons. Child Development, 59. 71-86. Steinhauser, R.,& Guthrie, J. T. (1974). Scanning times through prose and word strings for various targets by normal and disabled readers. Perceptual and Motor Skills, 39, 931-938. Hsu,C. C., & Kitamura, S. (1982). Reading Stevenson, H.W., Stigler, J. W.,Lucker, G. W., disabilities: The case of Chinese, Japanese, & English. Child Development, 53, 1164-1 181.


97

Tal, N . , & Siegel, L. S. (1990). The relationship between question-answer rype and reading comprehension among good and poor readers. Unpublished manuscript. Taylor, H. G., Satz, P., & Friel, J. (1979). Developmental dyslexia in relation to other childhood reading disorders: Significance and clinical utility. Reading Research Quarterly, 15, 84- 101. Temple, C. M. (1988). Red is read but eye is blue: A case study of developmental dyslexia and follow-up report. Brain and Language, 34, 13-37. Vellutino, F. R., Steger, J. A., & Kandel, G. (1972). Reading disability: An investigation of the perceptual deficit hypotheses. Cortex, 8, 106-1 18. Venezky, R. L.,& Johnson, D. (1973). Development of two letter-sound patterns in grades one through three. Journal of Educational Psychology, 64. 109-1 15. Wallach, M. A., & Wallach, L. (1976). Teaching all children ro read. Chicago: University of Chicago Press. Waller, G. (1976). Children’s recognition memory for written sentences: A comparison of good and poor readers. Child Development, 47, 90-95. Waters, G. S., Bruck, M., & Seidenberg, M. (1985). Do children use similar processes to read and spell words? Journal of Experimental Child Psychology, 39, 51 1-530. Waters, G. S., Seidenberg, M. S., & Bruck, M.(1984). Children’s and adults’ use of spelling-sound information in three reading tasks. Memory and Cognition, 12, 293-305. Weber, R. (1970). A linguistic analysis of first-grade reading errors. Reading Research Quarterly, 5 , 427-45 1.

Werker, J. F., Bryson, S. E., & Wassenberg, K. (1989). Toward understanding the problem in severely disabled readers: Part 11. Consonant errors. Applied P sycholinguisfics, 10, 13-30. Willows, D. M., & Ryan, E. B. (1981). Differential utilization of syntactic and semantic information by skilled and less skilled readers in the intermediate grades. Journal of Educational Psychology, 73, 607-615.


LEARNING TO READ: A THEORETICAL SYNTHESIS

John P . Rack and Charles Hulme DEPARTMENT OF PSYCHOLOGY UNIVERSITY OF YORK HESLINGTON, YORK YO1 SDD, U.K.

Margaret J . Snowling DEPARTMENT OF PSYCHOLOGY UNIVERSITY OF NEWCASTLE-UPON-TYNE NEWCASTLE-UPON-TYNE NEI lRU, U.K.

I. INTRODUCTION 11. METHODOLOGICAL ISSUES

A. CROSS-SECTIONAL STUDIES B. LONGITUDINAL STUDIES 111. THEORIES OF READING DEVELOPMENT

IV. DESCRIPTIVE STAGE MODELS O F READING DEVELOPMENT V. CONSTRAINTS ON READING DEVELOPMENT: THEORIES ABOUT EXTERNAL CAUSES A. LANGUAGE SKILLS AND LEARNING TO READ: THE CENTRAL ROLE OF PHONOLOGY B. VISION AND LEARNING TO READ VI. COGNITIVE PROCESSING THEORIES OF READING DEVELOPMENT A. INFORMATION PROCESSING THEORIES B. CONNECTIONIST MODELS OF READING DEVELOPMENT VII. THE BIOLOGICAL BASES OF READING ABILITY VIII. CONCLUSIONS REFERENCES

99 ADVANCES IN CHILD DEVELOPMENT AND BEHAVIOR, VOL. 24

Copyright B 1993 by Academic Press, Inc. All rights of reproduction in any form reserved.

100

John P. Rack er al.

I. Introduction This article is a review and a summary of recent research and theory in the field of reading development. We focus on the acquisition of word recognition skills because these, we will argue, are the limiting skills in reading development. Our main goal is to show how different theories in this area have been addressed to different questions and provide different levels of explanation. We try to identify the strengths and weaknesses of a number of the leading theories and relate them to each other. In this way, we attempt to synthesize information from a number of disparate approaches to the study of reading development. We argue that information processing theories are one particularly important class of theories being concerned with specifying the cognitive mechanisms underlying reading skills and the way these change with age. A promising approach to these questions comes from the emerging field of connectionist modeling. We argue that development in general, and reading development in particular, lends itself admirably to this type of modeling. The article is organized into eight sections. Following the present introduction, we consider the different methodologies that have evolved to address research questions in the field. The rest of the article is concerned with theories of reading development. We focus on four levels of theory: (1) descriptive stage models, (2) theories of external causation, (3) information processing and computational approaches, and (4) biological theories. We review evidence relevant to each of these four levels of theory concentrating, in particular, on evidence concerning the role of phonological skills in learning to read. In the final section, we attempt to synthesize the information from all four levels of theory.

11. Methodological Issues A number of different research methodologies have been used in this area. Each of them has certain strengths and weaknesses and in some cases the weaknesses in one are complemented by the strengths of another. A. CROSS-SECTIONAL STUDIES

The majority of studies of reading skills have involved cross-sectional designs. In this type of design, children of different ages are compared to provide general descriptions of reading and related skills at particular stages in development. If some abilities are only seen in groups of good readers then it is possible to speculate that those abilities might be prerequisites for reading development. An obvious problem with this approach is that good readers will tend to differ from poor readers on a large number of other variables. For example, good readers

Learning ro Read

101

tend to be older than poor readers and would therefore be expected to be better on most tasks that they might be given. In some studies, groups of good and poor readers of the same age are compared. The problem here is that differences between the groups could result from rather than cause the differences in reading ability. One technique that has been used to overcome some of these problems is the reading-level-match comparison (Backman, Mamen, & Ferguson, 1984; Bryant & Goswami, 1986, 1987; Jackson & Butterfield, 1989; Rack, Snowling, & Olson, 1992). If pmr readers are found to differ from younger readers who are reading at the same level, then the possibility that the group differences are simply reflections of reading level can be excluded. Positive results from reading level designs have been taken as evidence that the relevant variable has some role in the reading problems of the poor reader group. We should note, however, that even these results are somewhat ambiguous because of the "third-variable" problem. A fundamental deficit may cause both the slower progress in reading and the deficit relative to younger normal readers (see Bishop, 1989; and Hulme, 1988, for specific examples). Thus, reading-levelmatch designs need to be supplemented by other approaches to establish causal relationships (Bryant & Goswami, 1987). B . LONGITUDINAL STUDIES

Cross-sectional studies can reveal associations (correlations)between different variables at particular ages but they provide little information about the processes of developmental change. In longitudinal studies children are assessed at a number of points in time; therefore, the development of a particular skill can be charted in a much more precise fashion than is possible using cross-sectional approaches. Information from such studies can form the basis of a model of reading development. A second advantage of longitudinal studies is that they allow us to address questions about the causes of individual differences in development. Here the focus is not so much on the way a particular skill develops, but on the factors that predict a smooth or successful course of development. The object of such studies is to see what abilities early in development predict successful attainment of a particular skill later. Prediction studies can therefore suggest the causes of individual differences in development that can form the basis of a developmental theory. We consider later the evidence from a number of longitudinal studies that have produced some important findings. Provided that predictor measures are obtained before the children have any reading skills, a direct assessment can be made of their effects on later reading skill. The possibility that reading skill influences the predictor variables can, of course, be excluded if reading skills did not exist at the first time of testing; however, the possibility that the predictor measure and

102

John P. Rack et al.

later reading achievement are both influenced by the same third variable and that the predictor variable actually has no direct relationship to reading cannot be excluded. General cognitive ability is the most obvious third variable that needs to be assessed in this type of study: The brighter children are likely to do well on a prereading measure and then go on to become the better readers. In this case, one would clearly be wrong in claiming that the measured prereading skill is the cause of the later reading attainment. This problem can be dealt with by appropnate statistical techniques although, as we shall see later, this has not always been done. We should also bear in mind that statistical techniques cannot be used if there is no measure of the potential third variable. Thus, one could always claim that an apparent relationship between a predictor and an outcome measure was actually mediated by another, unmeasured variable. This problem has at least two 'solutions. The first is to develop a plausible theory and to challenge others to suggest what third variables are omitted from it. We pursue this solution in the remainder of this article. The second is to conduct a study with direct manipulation of the relevant variable. Both of these approaches are valid and, of course, both have their limitations. Bryant and Goswami (1986, 1987) and Bradley and Bryant (1985) have argued convincingly for intervention studies as the only way to test a causal theory of reading development. To provide such a test, one needs to be able to manipulate the variable that is hypothesized to influence reading ability. All of the types of studies that we have considered up to now have involved preexisting differences in the variable that is hypothesized to influence reading; therefore, they are subject to the problems of interpreting correlational data (Jackson & Butterfield, 1989). In a longitudinal intervention study, two matched groups are given different treatments and are then assessed at later points in time to determine the effects of the experimental treatments. A significant group difference provides evidence that the variable that was experimentally manipulated has a direct causal role in reading development. We review some intervention studies that relate to specific theoretical predictions. At this stage, we note that some problems can also arise in this type of study. Experimental interventions can show what can be done to alter the course of development but they do not show what typically happens in development. Bryant and Goswami (1986, 1987), however, have argued that combining longitudinal predictihn studies with intervention studies can overcome this problem. Prediction studies show that a certain variable is associated with differences in reading ability, and intervention studies then allow evaluation of the hypothesis of a causal connection between the variable and reading. More broadly, intervention studies are useful when one has a developmental theory to test, and such theories arise most securely from longitudinal prediction studies. A further problem for intervention studies is that the variables we identify as potential causes are themselves complex. Even if an intervention is successful,

Learning to Read

103

one needs to know which components of the intervention were critical. We return to this problem later. Finally, some quite likely causes of reading difficulties may prove highly resistant to the effects of training programs. In an extreme case, a system might be so badly impaired as to be untrainable. It would clearly be wrong to conclude that a failure to show a training effect demonstrates that the relevant system has no role in reading ability. All these considerations serve only to reemphasize that having a good theory is essential when attempting to design and interpret studies in this or any other area of cognitive development. Our preference, as we outline later, is for a theory that provides an explicit information processing account of reading and its developmental progression.

111. Theories of Reading Development In the remainder of this article we consider some of the existing theories of reading development and the evidence that is relevant to them. Each of the theories tends to focus on some range of phenomena and tends to provide descriptions and explanations of these phenomena. Before going on to describe the different theories, however, we need to stand back and ask what a complete theory would look like. First, we need to consider the different levels at which theories operate, and this will provide a framework within which to consider and organize the existing theories. An important starting point for a theory of any sort is descriptive accuracy. The theory needs to lay out what phenomena need to be explained. Developmental theories, of course, tend to be particularly concerned with describing how skills change with age. A theory of reading development needs to describe the relevant knowledge and skills that are acquired and in what order. Typically, “stage” theories of reading development have attempted to do this. We describe some of these below. A second requirement for a theory of learning to read is that it should specify how reading grows out of underlying capacities that exist for other purposes. In evolutionary terms, reading is a very recent acquisition for our species and, therefore, theories that suppose that biologically specific mechanisms exist primarily to subserve reading are distinctly implausible. Instead, it seems much more likely that reading grows out of cognitive mechanisms that exist in the child for other purposes. Most obviously, reading will be based on mechanisms that exist for language and vision. The task then becomes a matter of specifying the way in which preexisting cognitive mechanisms facilitate and place constraints on the development of reading skills. Theories of this sort are concerned with what we term external causal mechanisms. We use the term external to underline the fact that the mechanisms postulated in such theories exist outside the reading system.

104

John P. Rack er al.

A third level at which a theory of learning to read should operate is the level of cognitive and linguistic mechanisms. The theory should specify the sorts of mental operations involved in recognizing and pronouncing words, understanding and remembering the sentences that are read, and constructing semantic representations for passages that have been read. This indeed is a tall order, and encapsulates most if not all of the key issues in cognitive psychology. In our discussion of theories at this explanatory level, we narrow our focus to the processes involved in word recognition. This narrowing of focus can be justified on two grounds: (1) It is widely accepted that once words have been decoded in reading, higher-level comprehension processes will be common to the processes involved in other areas of language comprehension. (2) Empirical studies have shown that the majority of children with specific difficulties in learning to read have problems at the level of decoding individual words (Connors & Olson, 1990; Hulme, 1987; Stanovich, 1986). In this view, the unique mechanisms that need to be specified in a theory of reading development are those concerned with printed word recognition. Theoretically, explanations of reading development should be couched in processing terms, that is, terms that specify the representations and the processes that operate on them to allow reading to take place and that specify how these representations and processes change with development. Two approaches exemplify theorizing at this level: traditional information processing theories and more recent computational models using connectionist architectures. Each of these approaches is discussed later. A final level of theory is concerned with specifying how underlying causal mechanisms (e.g., phonological processing mechanisms) and the information processing systems that develop from them (e.g., recognition systems for printed words) are instantiated biologically. Work here is in its infancy and we have very little to say about this level of theory (but see Galaburda, 1989. and Pennington, 1991, for further discussion). We do, however, deal with some studies of the genetic mechanisms underlying individual differences in reading skills that are relevant to theories at this level. In summary, theories about reading development can be thought of as operating at four different levels. Descriptive theories specify how reading behavior changes with age. Theories about external causes specify how reading grows out of other preexisting cognitive systems. Cognitive theories specify how the computational processes underlying reading are organized and operate. Biological theories are concerned with how cognitive-level explanations map onto brain mechanisms. A complete theory of reading development would need to address all these levels. Such a theory does not exist. Most of the theories that we have operate at one level, some at two levels. We review these theories and hope that the synthesis that emerges will represent a tentative step toward a more integrated theory of reading development.

Learning

to Read

105

IV. Descriptive Stage Models of Reading Development In describing developmental changes with age, there is a marked tendency to talk about stages or phases. All children begin as nonreaders and many go on to become fluent readers capable of reading complex material rapidly and relatively effortlessly. Less obvious, however, is how to describe the developmental progression that leads from the nonreader to the fluent reader. One possibility, which has not found favor with theorists, is that the progression is one of gradual incremental improvement. Instead, most theorists have chosen to characterize the progression in terms of a number of distinct stages. Two of the most influential stage models of reading development are those of Marsh, Friedman, Welch, and Desberg (1981) and Frith (1985). Each model is based on the idea that children have at their disposal three basic types of reading strategies when learning to read, and that these strategies tend to emerge in the following order during development. The first strategy involves semantic processing: Children use the story context to generate hypotheses about what words might be written on the page. The second strategy involves visual processing: Some visual feature of a word suggests (or cues) what the word might be. Finally, a phonological strategy begins to take over: Children make use of information about the sounds of letters to decode or “sound out” written words. None of the stage theories maintains a rigid separation between the three strategies; however, there are differences in their availability and their rate of change within the different phases of development. Marsh et al. (1981) divided reading development into four stages. The first of these involves rote learning and linguistic guessing. At this stage, children may have a rote memory for a few highly familiar words like their name, but unfamiliar words can only be guessed at using linguistic context. These strategies are also available during the second stage of “discrimination net guessing.” Additionally in this stage, children begin to make use of visual features to discriminate between different words. Children will still guess at unfamiliar words using linguistic context but their guesses will share visual features with the target word. The third stage is sequential decoding. In this stage, children begin to decode unfamiliar words from left to right but only simple words comprising consonant-vowel-consonant (CVC) structures can be dealt with. More complex decoding rules are not acquired until the fourth stage, which is called hierarchical decoding. In this stage, children make use of conditional rules or contextsensitive rules and they also make use of analogies. Frith (1985), in a revision of the basic stage approach adopted by Marsh et al., proposed a three-stage model of reading development. The first stage in Frith’s model is logographic reading, in which words are recognized on the basis of distinctive visual or contextual features. At this stage, unfamiliar words cannot

106

John P. Rack et al.

be read. The second stage is alphabetic reading, in which children make use of spelling-sound rules for reading and therefore some decoding of unfamiliar words is possible. Specifically, Frith proposed that alphabetic reading involves decoding “grapheme-by-grapheme” and that the shift to alphabetic reading is brought about through spelling experience. The final stage is the orthographic stage. In this stage, words are recognized in terms of orthographic units, which, Frith stated, should ideally coincide with morphemes. Orthographic strategies are distinguished from logographic strategies in that they involve analytic, abstract letter (not visual) processing, Unlike alphabetic strategies, orthographic strategies are nonphonological and involve larger units of words. Frith stated that logographic strategies correspond to the rote learning strategy described by Marsh et al. Alphabetic strategies correspond to the sequential decoding strategies used in stage 3 and orthographic strategies include analogies, which are part of stage 4 of Marsh et al. The hierarchical decoding that is a feature of stage 4 reading in Marsh’s model can, according to Frith, be considered orthographic or an advanced form of the alphabetic strategy. Seymour and colleagues have studied beginning readers and developmental dyslexics using Frith’s framework (Seymour & Elder, 1986; Seymour & MacGregor, 1984). Their data support the view that both logographic and orthographic reading are visually based; however, logographic reading is described as holistic and orthographic reading as analytic. Alphabetic reading is conceptualized as a simple grapheme-phoneme correspondence system requiring segmentation to the level of the phoneme. Seymour ( I 986) argued that specific deficits can occur in both the alphabetic system and the orthographic system and that each of these deficits gives rise to different types of dyslexia. Ehri (in press) summarized a model of reading development that she had been refining over a number of years (Ehri, 1987; Ehri & Wilce, 1985, 1987). Ehri’s model has some similarities to the models we have already discussed; however, it is primarily concerned with the way in which familiar words are read. Ehri proposed three phases of reading acquisition: visual cue reading, phonetic cue reading, and phonemic map reading. In the visual cue phase, children make use of salient visual features of words or features of the word’s context to access meanings. In Ehri’s second phase children are thought to use phonetic cues for accessing word meanings. Children in this phase make use of the phonetic characteristics of words at a fairly basic level to help access pronunciations (and then meanings). Children begin to use phonetic cues only when they have learned some letter-sounds or letter-names; they use their letter-sound knowledge to form connections between the spellings of words and sounds in their pronunciations. A study by Ehri and Wilce (1985) showed that once children have some lettersound knowledge, they are more likely to learn systematic nonwords (GRF for girafe) than arbitrary nonwords (XBT for girafe). Ehri’s term phonetic cue reading captures the essence of this strategy, namely, that the child makes use of

Learning to Read

I 07

some phonetic features of the letter string to access specific words directly in lexical memory. The word is not recognized “visually”; neither is it read using a traditionally defined decoding routine. The third phase in Ehri’s model involves a fully specified phonemic mapping of letters onto sounds. Readers in this phase make use of all the information in the word and are therefore less prone to the confusions that the phase 2 readers would make. Ehri proposed that the reader in this phase is analyzing the word down to the phonemic level, in contrast to phase 2 readers who use phonetic cues that could be at any level from the syllable to the phoneme. A prerequisite for this phase is, therefore, the ability to segment speech sounds at the phonemic level. Ehri suggested that the transition between phases 2 and 3 may more accurately be thought of as gradual and quantitative rather than qualitative. Thus, the stage models all specify that the early stages of reading development involve the use of visual cues; however, Marsh et a]. distinguish between “rote learning” and “discrimination net guessing,” the former being restricted to highly familiar words and (presumably) holistic in nature, the latter being analytic. In the Frith and Seymour models, logographic reading is described as visual-holistic (in contrast to orthographic reading which is visual-analytic), but it should be noted that most of the logographic errors described in the literature involve substitutions of subword parts and could therefore be considered analytic also (Stuart & Coltheart, 1988). Only Marsh et al. explicitly acknowledge the importance of linguistic and semantic strategies but these are not denied by the other models; Frith, Ehri, and Seymour were more concerned with processes involved in recognizing words in isolation. The models show less’ consistency with regard to the introduction of phonological processes. Marsh eta]. treated phonological processes in terms of rules that are acquired so abruptly that we might assume that they are taught. Frith proposed that the first phonological processes in reading involve graphemephoneme decoding; children become aware of the relationships between graphemes and phonemes because of experiences in learning to spell. Ehri’s phonetic cue stage is dependent on the acquisition of letter-sound knowledge to allow the formation of visual-phonological connections for specific words. In summary, we think that stage models are best seen as useful descriptions of the surface behavior of children’s reading skills at different ages. Only Ehri’s model specifies the prerequisites and the mechanisms for acquisition of a sight vocabulary. In general, the models are relatively inexplicit both about the sorts of external influences that are crucial for development and about the form that the information processing systems underlying reading take at different ages; however, the work of the stage theorists does go a long way in specifying the sorts of behavior that information processing models need to explain. Stage models were developed mostly against the backdrop of “dual-route” theories of adult reading (see Humphreys & Evett, 1985, and Van Orden, Pen-

108

John P. Rack er al.

nington, and Stone, 1990, for very useful, if highly critical, overviews). According to models of this type, a word could be pronounced using one of two distinct procedures. A lexical route enabled the pronunciation to be accessed directly from an abstract description of its visual/orthographic form. An indirect mute enabled the pronunciation to be assembled using sets of grapheme-phoneme correspondence rules. As we will see later in this paper, a major contribution of the connectionist approach was to show that lexical (direct) and sublexical (rulebased) procedures can be carried out by a single system. This, we believe, requires that we reevaluate stage theories of development and in the final section of this paper we present our attempts to do this. We do not mean by this to deny the existence of other mechanisms for naming words. For example, there is evidence from studies of acquired disorders of reading that a route through semantic memory can also be used (Hinton & Shallice, 1990); however, very little is known about the development of this system.

V. Constraints on Reading Development: Theories about External Causes As we noted earlier, theories of reading development need to specify how the reading system is able to develop from systems (or modules) that have evolved to perform other functions. The most obvious systems that are relevant to learning to read are in the domains of language and vision. The best evidence for constraints operating on the development of reading skills comes from the study of language processes, though research in the field of visual factors gained renewed momentum in the 1980s. We review each of these fields, highly selectively, in turn. A. LANGUAGE SKILLS AND LEARNING TO READ: THE CENTRAL ROLE OF PHONOLOGY

Before proceeding let us reiterate that by reading we mean the processes involved in recognizing printed words. A massive amount of evidence links the development of reading skills in children to underlying phonological skills. This evidence has come from a variety of sources including studies of children with specific reading difficulties (or dyslexia) and correlational studies of normal children. Because most of this evidence has been reviewed elsewhere (for recent reviews, see Goswami & Bryant, 1990; Hulme & Snowling, 1991; Shankweiler & Liberman, 1990; Wagner & Torgesen, 1987) we are selective here. We concentrate on giving some of the major conclusions that arise from this work and draw attention to some important unresolved issues. In developing our argument

Learning to Read

109

we use only a handful of key studies together with a few more recent and less well known studies that we feel are important. A number of studies have revealed substantial correlations between phonological skills in preschool children and their later reading skills (e.g., Helfgott, 1976; Mann, 1984; Share, Jorm, Maclean & Matthews, 1984; Stanovich, Cunningham, & Cramer, 1984; Stuart & Coltheart, 1988; Tunmer, Hemman, & Nesdale, 1988). Thus, there is suggestive evidence that preschool phonological skills play a causal role in the development of reading. The interpretation of these studies, however, is not completely straightforward as Wagner and Torgesen (1987) noted in their extensive review. Wagner and Torgesen identified four general problems that arise in the interpretation of longitudinal studies: (1) confounding effects of general cognitive ability (or IQ control), (2) preexisting differences in reading level, (3) untangling reciprocal causation, (4) assessing phonological skills. In another recent review, Goswami and Bryant (1990) focused their attention on the issues of 1Q control and the assessment of phonological skills. In particular, Goswami and Bryant were critical of studies in which different measures of phonological skill were combined into a single composite score. The problem of IQ control can be dealt with easily as a number of studies show that IQ cannot remove all of the variance that is explained by phonological skill (see Wagner & Torgesen, 1987, for further discussion). The second problem noted by Wagner and Torgesen-possible differences in reading ability in preschool children-turns out to be much more serious. The problem here is that phonologically skilled preschool children may already be the better readers. The fact that they are still the better readers at a later stage reveals very little about the causal relationship between phonological skill and reading ability. When Wagner and Torgesen reanalyzed the data of Lundberg, Olofson, and Wall (1980), they found that the relationship between phonological skill and first-grade reading level was reduced dramatically when measures of reading ability in kindergarten were taken into account. The same thing could easily happen to the data from other longitudinal studies in which preexisting differences in reading level were not taken into account (Helfgott, 1976; Mann, 1984; Share et al., 1984; Stanovich et al., 1984). The picture that emerges from a critical analysis of these early studies is thus somewhat mixed. Preschool phonological skill is certainly a good predictor of later reading success-better than IQ even-but the factors that underlie this predictive relationship are not obvious. One worrying possibility is that preschool phonological skill is actually a reflection of reading ability that exists at the start of the period of study. This issue is clarified by combing a longitudinal study with an intervention study as was done by Bradley and Bryant (1983, 1985). At the beginning of their study, the sound categorization ability of more than 400 four- and five-year-old

I10

John P. Rack

et

al.

children was assessed before they started to learn to read. The children’s memory for the items in the sound categorization tests was also tested. More than 3 years later, their reading and spelling abilities and verbal intelligence were assessed. Performance on the sound categorization task was predictive of later reading scores, even when measures of intelligence and memory were taken into account. To determine whether this correlation reflected a causal influence, Bradley and Bryant included a training study. Sixty-five children who initially were poor at sound categorization were split into four groups. One group was trained in sound categorization, and a second, in addition, was taught letter-sound correspondences (and received exercises relating the sound structure of words to their spelling patterns using plastic letters). Two control groups were also included: one group was taught to group words according to semantic categories and the other received no training. After training, spread over 2 years, the group that had been taught sound categorization and letter-sound correspondences was some 8 to 10 months ahead of the semantic categorization control group in reading scores. The group that had been taught only to categorize sounds was about 4 months ahead of the semantic categorization control group in reading, but this difference was not statistically significant. This study has been widely cited as showing a causal influence of phonological awareness on learning to read. However, although these results are extremely impressive, they do not unequivocally demonstrate a causal role of sound categorization ability in learning to read. The proof that sound categorization ability was causally linked to reading skill required evidence that the group taught only to categorize words on the basis of their sound was significantly ahead of the group taught to categorize on the basis of meaning. The difference between the two types of training was observed, but not to a statistically reliable degree. The difference between the group trained in sound categorization alone and the group who also received training in letter-sound correspondences is notable, however. The exercises that the latter children received involving relating sounds in words to their spelling patterns, in combination with sound categorization training, led to substantial improvements in reading and spelling skills. A natural implication of this result is that to be most effective, training in phonological skills may need to be integrated with the teaching of reading skills (see Hatcher, Hulme, 8c Ellis, in press, for further data and a discussion of this idea). In a later intervention study, Lundberg, Frost, and Peterson (1988) gave a group of 235 Danish kindergarten children extensive phonological awareness training. Another 155 children served as controls. The training took place during preschool where no formal reading instruction is given. A poshest measure confirmed that the training was effective; the experimental group performed better than the control group despite the fact that the control group had slightly (but significantly) higher scores on the pretest measures of phonological awareness. In contrast, general language comprehension and letter knowledge

Learning to Read

111

increased equally for both groups over the training period. The effects of phonological skills training were thus highly specific to the phonological domain. This specificity was important because, in principle, the experimental group might have shown greater improvements across the range of tests simply because they received attention and the control group did not. At the beginning of the training period, only one child in the experimental group and two in the control group had any reading ability as assessed by a crude screening test. By the end of training this number had not changed for the control group, but 15 of the trained group now showed some reading ability. Lundberg et al. (1988) speculated that this improvement in reading might be one of the effects of the training program. The more permanent effects of the phonological awareness training were investigated by measuring reading and spelling some 7 months into the children’s first year in school and again in the middle of their second year. The experimental group outscored the control group on reading and spelling in Grades 1 and 2 although the differences in reading in Grade 1 were not significant. Group differences in Grade 2 were significant, however, and were significantly larger than the differences in Grade 1. In contrast, the control group outscored the experimental group on a test of mathematics given in the first year. This important control test indicated that the training had a specific effect on the targeted skills of reading and spelling rather than a global effect on all school subjects. One potential problem with the Lundberg et al. study is that the control group received no special attention during the intervention period; however, an explanation of the effects in terms of nonspecific factors such as motivation is hard to reconcile with the selective benefits derived for reading and spelling compared with mathematics. It might also be argued that the size of the effects on reading and spelling skills obtained in this study was small, especially in relation to the amount of phonological training that the children received. Recall that differences in reading ability were not significant after 1 year of schooling. Although this might be seen as a weakness of the study this is actually the pattern that a number of theorists would have predicted. Ehri (1989) put forward the suggestion that phonological awareness training facilitated spelling initially, and reading subsequently, because its contribution to spelling skill then affected word reading directly. This argument is similar to that advanced by Frith (1985), and supported by Cataldo and Ellis (1991), that alphabetic skills are acquired first for spelling and then transferred to reading. The Lundberg et al. (1988) study suggests that training in phonological skills alone is not a very powerful way of affecting reading development. Bradley and Bryant’s (1985) results, on the other hand, suggest that phonological training integrated with reading and spelling is more effective. Hatcher et al. (in press) tested the hypothesis that phonological training combined with reading tuition would have a more beneficial effect on learning to read than training in pho-

1 12

John P . Rack et al.

nological skills or reading alone. Their longitudinal intervention study involved 7-year-old children experiencing difficulties in the early stages of learning to read. The subjects were divided into four matched groups and assigned to a control condition or one of three experimental teaching conditions: reading with phonological skills training, reading alone, and phonological skills training alone. Although the “phonology-alone’’ group showed the most improvement on phonological tasks, the “reading-with-phonology” group made the most progress in reading and spelling. Thus when the intervention to boost phonological skills was integrated with the teaching of reading, reading and spelling skills improved the most. In summary, this review of studies that have attempted to demonstrate a causal connection between phonological skills and learning to read using intervention studies illustrates just how difficult it is to do so. On balance we believe such a link (or set of links) exists, though it may not be as direct as was once thought, given the evidence that integration of phonological training with the teaching of reading seems important. Nevertheless, the studies in this area have probably come as close to establishing a causal mechanism as any studies concerned with cognitive development. The third and fourth issues identified by Wagner and Torgesen (1987) concern the possible interactive effects between reading and phonological skill and different levels, or types, of phonological skill. As we have seen, many studies indicate that phonological skills influence the acquisition of reading skills. But reading ability also influences performance in tasks that measure phonological skills. Thus, when asked to make judgments about sounds of words, people frequently make use of their knowledge of the words’ spellings (Ehri & Wilce, 1980; Seidenberg & Tanenhaus, 1979; Perin, 1983; Rack, 1985). Others have proposed that the concept of a phoneme is highly artificial outside the context of reading experience (see, e.g., Liberman, Shankweiler, Fischer, & Carter, 1974). Support for this view comes from studies of nonliterate adults who find segmenting words by phonemes very difficult (see Morais, 1991, for a review). Thus, it seems that reading ability and phonological skills are interdependent during development, each being influenced by the other skill. One of the few tests of this hypothesis was carried out by Perfetti, Beck, Bell, and Hughes (1987) using a longitudinal design. They gave first-grade children tests of nonword reading, phoneme deletion, and phoneme synthesis on four occasions during the school term. One group of children were being taught letter-sound relationships as part of their reading instruction (the direct code group), a second group were getting little or not phonics-based instruction (the basal group). The repeated testing of phonological skills and nonword reading allowed the use of cross-lagged correlations to investigate how the two skills were interrelated. This technique is fairly simple in principle. If reading and phonology are related and each is measured on two occasions, t , and r2, predic-

Learning to Read

I13

tions about the relative size of the correlations depending on the causal relationships that exist between them can be made. If phonology causes reading to develop, then the correlation between phonology at t , and reading at t, should be large (and larger than the correlation between reading at t , and phonology at tz). If, however, reading causes phonology to develop, the opposite pattern should emerge; that is, the correlation between reading at t , and phonology at t2 should be large (and larger than the relationship between phonology at r , and reading at t 2 ) . The results revealed both relationships between certain phonological skills and later reading and relationships between reading and other later phonological skills. Performance on the phoneme synthesis task predicted later nonword reading performance for both groups; however, nonword reading predicted later success in the phoneme deletion task, and in one of the groups (curiously, the group not receiving phonic teaching), deletion skill went on to become a predictor of final nonword reading. The pattern of correlations obtained in this study is consistent with the idea of reciprocal relationships between reading and phonological skills and with the corollary that not all phonological skills are the same in their relationship with reading skills. This idea of “reciprocal causation” seems to us to have a good deal to commend it; however, the evidence from the study by Perfetti et al. must be interpreted with caution because of the statistical techniques that they used. Cross-lagged correlations are notoriously unreliable because the pattern of correlations is sensitive to variations in the reliability and true score variance of the measures used (Rogosa, 1980). Recent developments in the area of structural modeling would allow a more rigorous test of this hypothesis but would require large samples of subjects to be tested on a number of different occasions. Ellis and colleagues (reviewed in Ellis, 1991) have made use of these techniques with very promising initial results. A variant of the reciprocal causation argument was put forward by Goswami and Bryant (1990) following their review of a large number of studies. Very simply, they argued that awareness of the onset-rime division (see Treiman, 1985) is the crucial factor in learning to read; awareness of smaller speech segments (phonemes) arises at least partly as a consequence of learning to read. Within each syllable of English, the rime comprises the vowel and succeeding consonants, if any; the onset refers to the consonant or cluster that precedes the vowel. So, for example, the monosyllabic word string consists of an onset, /str/, and a rime, /ing/. One task that taps young children’s sensitivity to rimes is rhyme detection, for example, asking children to pick the odd (nonrhyming) word from a group of four spoken words (Bradley & Bryant, 1978). As we have seen, this task can be performed quite adequately by many 4- and 5-year-olds, before they have learned to read, and is highly predictive of their later success in reading and spelling. A number of authors have looked at the correlations between different pho-

114


nological measures to determine whether there are different types of phonological skill. Following a factor analysis of the Lundberg et al. (1980) data, Wagner and Torgesen (1987) concluded that a single-factor model provided a good fit. A similar conclusion was reached by Stanovich et al. (1984). In a later study, Lundberg et al. (1988), using confirmatory factor analysis, showed that a two-factor structure provided a good fit to their phonological awareness data. One factor consisted of the three tasks requiring phonemic segmentation, and the other factor consisted of the three measures requiring word and syllable segmentation. The two factors were moderately correlated at .40.Rhyming ability was not entered into this analysis because only one measure was available. Looking at the changes in these three abilities (rhyme, phoneme segmentation, and syllable/ word segmentation) across the training period in their study revealed that the most dramatic changes occurred on the phoneme segmentation tasks. The trained group did improve significantly more than the control group on the other two measures, but the differences were modest in comparison to the effects of training on phoneme segmentation skills. Contrary to Goswami and Bryant’s (1990) suggestion, this result shows that phoneme awareness can exert a causal influence on reading ability. It seems to us that the factor-analytic evidence is fairly ambiguous. Good support for Goswami and Bryant’s theory would have been found had measures that tap phonemic awareness loaded on a separate factor than measures that tap awareness of larger units such as syllables and rimes. The fact that these different skills in fact correlated highly with each other is not, however, incompatible with a developmental model in which these skills develop at different times and have different causal relationships with reading ability. In a study that strongly influenced Goswami and Bryant’s theory, Bryant, MacLean, Bradley, and Crossland (1990) set out to determine whether phoneme awareness is a result of reading ability, a predictor of reading ability, or a necessary intermediary between rhyming ability and reading ability. To address these questions, Bryant et al. (1990) studied a group of 64 children when they were approximately 46,56, 6, and 66. The children were given tests of rhyme and alliteration at the first two testing times and tests of phoneme awareness (deletion and tapping) at the second two testing times. At a final time of testing they were given measures of reading, spelling, and arithmetic. Multiple regression analyses were used to assess the causal relations between the various measures. After allowing for the effects of extraneous variables, they found that the phoneme segmentation measures predicted reading and spelling to a lesser extent (median 4% unique variance) than did the rhyme and alliteration measures (median 9%). Rhyme and alliteration were found to make independent contributions to reading and spelling scores, even after the phoneme segmentation measures were entered into the equation. After initial-phoneme deletion or

Learning to Read

I15

final-phoneme deletion or phoneme tapping was entered, the median variance still explained by the rhyme and alliteration measures was 4,8, and 7%, respectively. Bryant et al. did not test to see whether phoneme segmentation accounted for significant variance in reading and spelling even after removing the effects of rhyme. From the reported figures it looks unlikely, although we cannot exclude such a possibility. Another interesting question concerned the influences on phoneme awareness. Even after intelligence, age, socioeconomic status, and vocabulary were accounted for, rhyme and alliteration accounted for additional variance in phoneme segmentation. These analyses therefore suggest that rhyme exerts a double influence on reading ability. It has one influence that is mediated by phoneme awareness and a second direct influence. In addition, the analyses suggest that phoneme awareness is not simply a product of reading ability; some of the variance in phoneme awareness comes from rhyme (but see later). Three reservations need to be noted about the interpretation of this study. First, no data about the children’s reading are available at the earlier points in time when segmentation ability was measured. Thus, we do not know how much the children’s measures of phoneme segmentation at 54 and 6 were confounded with reading ability. For example, the children who were good at phoneme segmentation at these points in time might also be the children who had progressed most in reading. Their phoneme segmentation skills might have been products of their reading skills, which could explain why they turned out to be predictive of later reading. The children’s reading skills may also have influenced their rhyme abilities at 44 and (more likely) at 54. Although we know that individual differences in rhyme exist before reading, we do not know that rhyming ability is uninfluenced by reading ability. A second reservation concerns the amalgamation of rhyme and alliteration. One might just as well argue that alliteration and initial-phoneme deletion should be grouped together. Bryant et al. have not established that their distinction between types of phonological segmentation skill is valid. Finally, unfortunately, the children were never given tests of rhyme and phoneme segmentation at the same point in time. So we do not know, for example, whether phoneme deletion at 54 might have been more predictive of later reading than was rhyme at 5f. In summary, the evidence reviewed here seems to indicate a very close and probably causal link between phonological skills and the process of learning to read. Variations in the efficiency of mechanisms that have evolved to deal with speech place constraints on the development of reading skills in children. With respect to causes, however, one final, very important, caveat needs to be considered. Ultimately, we believe that the proper interpretation of longitudinal and intervention studies of the sort described here will be in terms of information processing theories that so far have been lacking. To hope that in the absence of

1 I6

John P. Rack er al.

such a process-oriented theory we can ever identify a cause is, perhaps, naive. The problem is that the measures identified and trained in intervention studies are themselves invariably complex and ill-understood. A good example here is the sound categorization task of Bradley and Bryant. This task certainly involves access to phonological information, possibly of different types in the case of rhyme and alliteration, and it also involves a significant memory load. There is little doubt that the success of this measure as a predictor of reading ability depends on its complexity. Demonstrations that training on such a complex task (in isolation) leads to improvements in reading would be important. But the proper interpretation of such results and the specification of what is a cause of reading success or failure would require an answer in terms of the cognitive mechanisms that are tapped by these tasks and how such mechanisms operate in the acquisition of a word recognition system. B. VISION AND LEARNING TO READ

The evidence we have reviewed in the preceding establishes an intimate connection between children’s language skills, particularly in the domain of phonology, and the process of learning to read. Most of this work has come from studies of normal variation in children’s reading skills. Converging evidence indicates that children with severe reading difficulties (dyslexia) often display evidence of underlying phonological difficulties. This evidence has been reviewed elsewhere (e.g., Hulme & Snowling, 1991; Rack & Snowling, 1985; Stanovich, 1986, 1988) and we do not dwell on it here except to note the close agreement between these studies and the theories and studies discussed in the present paper. Work on visual factors in reading has been conducted almost exclusively on groups of children with severe reading difficulties (dyslexics). After a long period of prominence, the importance of visual difficulties as an explanation of reading problems was comprehensively dismissed in a rigorous review by Vellutino (1979); however, a number of researchers have since argued that a finer-grained analysis of the visual system will be required to detect potentially deficient processes in poor readers. For reasons of space we do not go through all of this evidence here (but see Lovegrove, 1991). Instead, we focus on one approach that seems promising. Lovegrove, Martin, and Slaghuis (1986) argued that a low-level visual deficit in a large proportion of dyslexic children may be a cause of their reading problems. They summarized a large body of carefully conducted studies of visual processing in dyslexic children and provided three major types of evidence concerning their visual abilities. (1) The function relating visible persistence (the subjective experience of how long a visual event lasts) to spatial frequency (how broad the stripes in a pattern are) differs in dyslexic compared with control

Learning to Read

117

children. (2) Measures of pattern contrast sensitivity exhibit small but consistent differences between dyslexic and control children when very indistinct (low contrast) striped patterns are shown. Dyslexic children find it harder than normal readers to detect broad stripes (low spatial-frequency gratings) but are equally (or more) sensitive to fine stripes (high spatial frequencies). (3) With displays of stripes that flicker on and off, dyslexic children are less sensitive at all flicker rates but particularly so at high rates. These results are taken as evidence that dyslexic children have a deficit in the visual transient subsystem (Kulikowski & Tolhurst, 1973). In a fourth series of experiments, designed to measure the sustained subsystem, no differences between dyslexic and normal children were found. This idea of a deficit in the transient system of retarded readers appears to offer a useful link between a varied set of findings. Lovegrove et al. argued that such a deficit may operate to impede the extraction of detailed information during reading because the transient system normally operates to “clear” the sustained visual subsystem (responsible for detailed vision) between saccades (see Breitmeyer, 1984). As Hulme (1988) has argued, the evidence for a causal role for visual transient system deficits in dyslexia is far from conclusive; however, the effects obtained seem robust and what we now need to pursue is a more detailed theoretical linkage between these low-level difficulties and the sorts of higher-level perceptual processes involved in learning to read. Lovegrove (W. Lovegrove, personal communication, 199 1) has found relationships in unselected groups of children between measures of visual transient system functioning and children’s nonword reading ability (a relatively pure measure of decoding skill). One possibility is that variations in the efficiency of the transient system produce differences in visual attentional and grouping mechanisms. The nonavailability of these normal mechanisms might, in turn, affect the process of creating well-specified representations of the ordered sequences of letters composing words. Such a problem could then adversely affect the development of mappings between orthographic and phonological representations of words (see Section VI,B).

VI. Cognitive Processing Theories of Reading Development We have argued that it is important to have explicit theories of reading development. These theories should specify both the representations and the processes that operate on them during reading, and how these representations and processes change with development. We first discuss conventional information processing theories and then turn to computational models based on connectionist architectures.

118

John P. Rack er al.

A . INFORMATION PROCESSING THEORIES

The most well developed approach to theorizing in cognitive psychology is in terms of information processing models. Such models are intended to specify the structure of the system responsible for a task. Different mechanisms are described in terms of the types of information or codes that are involved and the processes that are brought to bear to manipulate and transcode information at various stages during the performance of the task. Qpically, the models are specified by flowcharts with verbal descriptions of the way information is coded and transcoded and of limits on the performance of different components. This approach to theorizing is certainly not without its problems or critics. One difficulty is that it is sometimes highly inexplicit about how certain critical processes operate. Nevertheless, it has proved very fruitful in developing descriptions of the structure of the processes underlying certain aspects of human performance, and has been a source of experimentally testable predictions. In studies of development, the typical flowcharts have proved less popular as a way of theorizing than they have in studies of adults. Indeed, such diagrams seem unduly static as ways of capturing developmental change. What such theories can, perhaps, provide are sketches of the state of a system at various points in its development. Morton (1989) took as a starting point the stages postulated by Frith and attempted to translate them into some hypothetical information processing models. Morton suggested that logographic reading involves the same processes that are involved in picture recognition. Written words provide access to representations in a pictorial-semantic system that is linked to a verbal-semantic system, the latter being needed to generate speech. Spoken words, in contrast, yield direct access to the verbal-semantic system. The goal of reading development might be viewed as learning to access verbal semantics directly from print. In keeping with Frith’s account, Morton proposed that the learning of graphemephoneme rules in the alphabetic stage gives the child an important (and necessary) means of converting print into a speech form. Morton distinguished two parts in the alphabetic reading stage. In the first part, children begin to segment words into component letters and to learn rules between individual letters and phonemes. These rules are strengthened by direct teaching and by the influence of sound-letter rules that have been acquired for spelling. In the second part of the alphabetic reading stage, feedback from the speech output system to the verbal-semantic system is established, allowing words read by the alphabetic route to be understood. In this second part of the alphabetic stage, the route to meaning via pictorial-semantics declines in importance. In Morton’s model, Frith’s final stage of orthographic reading is reached by the acquisition of a “morphemic” processor that can translate print directly into a form that can yield access to verbal semantics.

Learning to Read

119

Morton therefore showed that each of the three phases in Frith’s model can be characterized in information processing terms. This is an important step in developing a testable model of reading development; however, Morton’s account is somewhat limited as a developmental model because he had very little to say about the causes and the mechanisms of change. We need to know why one processing mechanism supersedes another and how this restructuring comes about. Morton supposed that logographic reading would be biased toward items that have representations in the pictorial-semantic system (highly imagable) and that access to those representations would often be achieved with only partial stimulus information. Thus, the system might be prone to errors and less well suited to some classes of words. We infer from this that such a system would be unlikely to support a large reading vocabulary. The acquisition of alphabetic processing is more fully described in Morton’s account of reading development, but the central role he gave to grapheme-phoneme rules in this stage does not seem to fit well with available data. Morton stated that in the alphabetic stage, “only regular words can be read, and nonwords can be read as accurately as words” (p. 58). Morton’s claim is clearly false because children’s early vocabulary contains a huge proportion of irregular words (Stuart & Coltheart, 1988). Moreover, Morton went on to argue that if the child learned any irregular words during the alphabetic phase, then he or she would not be able to use this knowledge as the basis of an analogy strategy for reading other irregular words. Goswami (1988), however, has shown in a carefully controlled series of experiments (see also Goswami & Bryant, 1990) that children as young as 5 years of age can make use of analogies to read novel words. For example, if children are taught to read a new word such as beak, they are then able to read other new words that share the rime unit of beak, e.g., weak and peak. Seven-year-olds, but not younger children, were able to read other visually similar words such as bean (having been taught beak), although the benefit here was not as great as for words sharing the rime unit. In contrast to Morton’s predictions, these results provide convincing evidence that children do use analogies and that alphabetic reading involves more than grapheme-phoneme rules. Regarding the acquisition of alphabetic strategies, Goswami and Bryant (1990) made the interesting prediction that it was sensitivity to onset and rime units (as measured by sound categorizationtasks) that enabled children to acquire an analogy strategy. They proposed that children “become adept at recognizing when words have common rimes or common onsets. So they form categories of words and when they begin to read they soon recognize that words in the same categories often have spelling patterns in common and that this spelling sequence represents the common sound.” (p. 147). This knowledge enables children to make inferences about unfamiliar words and, according to Goswami and Bryant, these inferences constitute the main phonological activity seen in the early stages

120

John P. Rack et a!.

of reading. As we have seen in this paper and elsewhere, the use of an analogy strategy in reading is far from the only phonological activity seen in early reading (Ehri & Wilce, 1985; Stuart & Coltheart, 1988). Nevertheless, the suggestion that children learn to use phonological strategies by learning about rhymes and rimes is an interesting one. It is one of the few explicit suggestions of a way that phonological abilities in spoken language influence phonological processes in reading. The reasons and the mechanisms behind the acquisition of orthographic processing are poorly specified in most accounts of reading development. Morton (1989) proposed that the acquisition of a morphemic processor is “a simple consequence of the interaction of reading, linguistic knowledge, and the general process of cognitive abstraction. Very few children fail to make the transition painlessly” ( p. 60). Thus, Morton’s model has a direct route that seems to follow as an automatic consequence of the previous processing stage. Other theorists such as Marsh et al. (1981), Frith (1985), and Goswami and Bryant (1990) are similarly vague about the move from an alphabetic to an orthographic strategy. The transition between alphabetic and orthographic strategies is perhaps addressed most directly by Seymour and his colleagues (see Seymour, 1990, for a review). Seymour and MacGregor (1984) suggested that the strategies described by Frith could be equated with modular components of an information processing system comprising a logographic and an alphabetic lexicon in the early stages. Representations in the logographic lexicon were thought to be visually based and linked with meanings in the semantic system and pronunciations in the preexisting phonological system. Entries in the alphabetic lexicon were detailed at the letter level and mapped onto speech segments in a phonological processor. According to this account, the logographic and alphabetic systems develop independently, but the formation of the orthographic lexicon is dependent on the alphabetic base and is an extension of it. More recently, Seymour has revised this model to place greater emphasis on the contribution of logographic processes to reading development (Seymour, 1990; Seymour & Bunce, in press; Seymour, Bunce, & Evans, in press). In this “dual-foundation” model the orthographic system is seen as the result of a merger of earlier systems. Thus, formation of the orthographic lexicon depends on an alphabetic knowledge of letter-sound associations, structures represented in the logographic lexicon, and the acquisition of phonological awareness. Seymour proposed that phonological structure provides an organizing principle for a framework containing orthographic elements. The orthographic system gradually evolves from a “core” structure (simple initial and terminal consonants and short vowels). This core subsequently expands to incorporate more complex forms such as consonant clusters and long vowels. In summary, the information processing models that we have discussed are an important step forward because they allow predictions to be generated and tested;

Learning

10

Read

121

however, current models do not deal in sufficient detail with the dynamic nature of reading development. In particular, the highly interactive accounts of reading development such as those of Ehri and Seymour do not fit easily into flowchart schemes. In the next section we review the computational approach to modeling reading development. We believe that this approach offers a more promising method for understanding the relationships between different processes and sources of knowledge in the acquisition of word recognition skill. B. CONNECTIONIST MODELS OF READING DEVELOPMENT

Connectionist or parallel distributed processing models are arguably one of the most powerful and novel approaches to the study of human cognitive processes. They represent a modem realization of the idea first put forward by Hebb (1949) that complex behaviors may emerge from the operation of aggregations of simple processing units. Much of the interest in these models stems from the fact that apparently “rulelike” behavior may be generated by systems that do not embody explicit representations of these rules. Although the scope and limitations of this approach are currently the subject of intense debate (see, e.g., Lachter & Bever, 1988; Pinker & Prince, 1988), these models represent powerful ways of thinking about the strengths and limitations of relatively simple associative mechanisms and have the great virtue of being computationally explicit and testable. Language learning in general, and learning to read in particular, would seem to represent significant human capacities, which prima facie place considerable demands on associative learning and thus lend themselves to a connectionist approach. Although connectionist models provide a powerful new approach to a range of issues in cognitive development, relatively few attempts have been made to tackle explicitly developmental issues using such models. The only model that comes close to doing so in the field of reading development is that of Seidenberg and McClelland (1989) and we consider this model in some detail next. The Seidenberg and McClelland distributed developmental model of word recognition and naming is concerned with how readers learn to recognize and pronounce printed words. We concentrate here on the model’s attributes relevant to learning to pronounce words, which is the task most often used to assess the growth of reading skill in children. The model consists of a set of orthographic units used to code letter strings, a set of phonological units used to code phonological information, and a set of hidden units that connect these two types of unit. When “trained,” the model should be able to translate a string of letters such as a regular, irregular, or nonsense word into a phonological (“spoken”)

form. The orthographic and phonological codes for words are represented as patterns of activation across a large number of primitive representational units. In this

I22


way a word is coded not by a single off/on detector (e.g., Morton, 1969), but by a pattern of activation across the whole range of detectors. This type of coding is described, technically, as a coarse coded, distributed representation (see Quinlan, 1991). The phonological representations are coded by sets of triples, each specifying a phoneme and its flankers. These triples are called Wickelphones (after Wickelgren). So, for example, the word make consists of the phoneme triple /mAk/ and is coded as n A , mAk, and A k . In fact, in the model each of these triples is encoded as a pattern of activity distributed over a set of units representing phonetic features. In this way each triple is coded by activity over a set of units. The orthographic representation is in many ways directly analogous to the phonological representation. Once again, coding is in terms of triples, but in this case triples of letters, as opposed to phonemes. In its initial state, before training, the model shows total connectivity: All orthographic units are connected to all hidden units and all .hidden units are connected to all phonological units. The connections between the phonological and orthographic units and the hidden units carry weights that govern the spread of activation across the units; these weights encode the model’s knowledge about written English that is acquired during training. Training of the model involves presenting letter strings that consist of English monosyllabic words. On each trial a pattern of activation is encoded over the orthographic units, and a pattern of activation is passed on to the hidden units. The pattern of activity over the hidden units is then used to compute activations over the phonological units. Before training, presentation of a letter string results in an entirely random pattern of activation being sent to the phonological units. Once a pattern of activation on the phonological units has been computed, it is compared to the correct pattern for the pronunciation of the target word. A learning procedure then adjusts the weights of connections in the network in proportion to the extent that this adjustment will reduce the error, using the backpropagation algorithm (Rumelhart, Hinton, & Williams, 1986). In this way, the model is trained on a large corpus of English monosyllabic words. As a result of training, the model simulates a number of aspects of human performance, including differences in difficulty between regular and irregular words, and can generate acceptable pronunciations for many novel items (nonwords or words it has not explicitly been trained on). The Seidenberg and McClelland model is a most impressive step forward in its explicit characterization of the process of learning to pronounce printed words; however, as a model of reading development, it has two major weaknesses. One problem is that it does not provide a psychologically realistic description of the processes involved in learning to pronounce printed words (Hulme, Snowling, & Quinlan, 1991). For example, children do not learn words in large batches; rather, learning is an incremental process. A second major problem is that the model really makes no contact with the massive body of evidence reviewed

Learning to Read

123

earlier showing links between early phonological skills and learning to read. As we have seen, children come to the task of learning to read with highly developed phonological representations. We know, furthermore, that tasks that tap these underlying phonological representations (such as measures of phonological awareness) are among the best predictors of children’s ability to learn to read. We would argue that a theory that does not embody these facts cannot be a psychologically valid model of the processes involved in learning a sight vocabulary. If we accept the idea that phonological skills play a causal role in facilitating early reading skills, then a viable model needs to embody this knowledge in some way. We believe that within such a model, good phonological skills can be thought of as facilitating the process of learning the mappings between orthography and phonology. We hypothesize that with a prestructured phonological store, such learning would be speeded up and in some of our current work we are in the very early stages of pursuing this idea. Psychological evidence from children learning to read is consistent with this approach. Although a common assumption is that good phonological skills facilitate learning to read because of their effects on the ability to use letter-sound rules, evidence suggests that other mechanisms are important. Stuart and Coltheart (1988), for example showed that errors in reading that appear to reflect phonological processing (and that are more common in children with good phonological skills) often preserve the first and last letters in words. These errors cannot, therefore, arise simply from a sequential left to right application of letterto-sound rules. One distinction that emerges from thinking about the Seidenberg and McClelland model is that between processes and knowledge. The connectionist architecture employed in this model is very simple. The power of the model comes from the knowledge it embodies about the statistical regularities that exist between English spelling patterns and their pronunciations. These regularities are abstracted by the model as a result of its training on a large corpus of words, and this knowledge is encoded as patterns of connection weights across the hidden units that connect the orthographic and phonological representations. The knowledge so encoded gives rise to a number of powerful features of the model such as its ability to generate reasonable pronunciations for novel items and its ability to simulate the differences in the ease of pronouncing regular as opposed to irregular words. These properties, which in “dual-route’’ theories were attributed to a separate “indirect” route, emerge in the network as a natural by-product of a single mechanism that relates spellings to sound. The operation of this single process depends on a certain knowledge base. The network would not show generalization or regularity effects unless it had been trained on a sufficiently large and representative sample of words. This perspective has important implications for stage theories of reading development. All of the stage theories posit a beginning stage at which children

I24

John P. Rack et al.

know only a small number of words that have been learned by rote (logographic, etc.). Later, according to these theories, a separate mechanism is acquired (alphabetic, etc.) and it allows the decoding of novel items. In terms of a network model, however, both of these stages can be much more economically described as the product of a single mechanism that operates in the same way. The changes that have occurred simply reflect differences in the knowledge that has been encoded by the network. In these terms a network model of the form described by Seidenberg and McClelland leads to a radical reconceptualization of stage models. Apparently very different surface behaviors can emerge from the operation of a single mechanism that is in different states of training. Thus, wellspecified models of the processes involved in reading development, such as connectionist models, have considerable promise.

VII. The Biological Bases of Reading Ability In this section we focus on research on the possible genetic bases of reading ability. We provide a brief overview of the research findings and then consider the form that a biological theory of reading difficulties might take. The starting point for genetic studies of reading ability (and more often disability) was the observation that reading problems tend to run in families (Hallgren, 1950; Stephenson, 1907; Thomas, 1905). Hallgren (1950) estimated risk of first-degree relatives to be 41% compared with the population risk of 5 to 10%. Vogler, DeFries, and Decker (1985) found that the children of parents with reading problems were significantly more likely to have reading problems than the children of parents without reading problems. Scarborough (1989) also found that familial factors were highly predictive of later reading disability. She found that it was possible to classify 50 of 62 children (80.6%)as good or poor readers using parental test scores. In a regression analysis, children’s reading ability in Grade 2 was predicted very strongly by family history of specific reading disability (36% of the variance). The measure of preschool phonological skill (sounds and letters) contributed a much smaller amount (4%) in comparison. In this study, therefore, the best predictor of reading problems was a family history of reading problems. Family studies are suggestive of a genetic role in the etiology of reading difficulties; however, reading difficulties may run in families for social or environmental reasons (see Pennington, 1989). To tease apart the effects of common genes and common environments one must turn to studies of twins. Twins share much of their environment (family, school, etc.) which would tend to make them resemble one another. Identical twins also have all of their genes in common and fraternal twins have 50% of their genes in common. Assuming that the

Learning to Read

I25

degree of common or shared environment is the same for both identical and (same sex) fraternal twins, then any greater similarity within pairs of identical twins would be the result of genetic effects. Although many people have questioned this “common environment” assumption, empirical studies have not shown it to be false (see Plomin, DeFries, & McClearn, 1980). The difference between the degree of similarity within identical and fraternal twin pairs can be expressed in various ways. Most typically, a “heritability” statistic is computed that reflects the proportion of the total variance that can be accounted for by genetic factors. The Colorado Twin Family Reading Study (DeFries, Olson, Pennington, & Smith, 1990) has the largest sample of twins for whom reading and related cognitive variables have been assessed. Analyses of these data reveal that individual differences in word recognition and group deficits in word recognition have substantial genetic components (Olson, Wise, Conners, Rack, & Fulker, 1989; Olson, Gillis, Rack, DeFries, & Fulker, 1991). In another large twin study, Stevenson, Graham, Fredman, and McLoughlin (1 987) found substantial heritability for spelling, although their measure of word recognition did not turn out to be heritable. One reason for these inconsistent results is suggested by a more detailed analysis of the components of word recognition skill by Olson and his colleagues. Olson et al. (1989) found that phonological decoding (measured by nonword reading) was highly heritable, but word-specific orthographic information was not heritable (see also Olson, Wise, Conners, & Rack, 1990). These results suggest that the phonological processes involved in word recognition are more heritable than the orthographic processes. Heritability estimates may therefore vary depending on the nature of the reading task. Taken together, the Colorado data and the Stevenson et al. data provide conclusive evidence that genetic factors play an important role in the development of reading and spelling skills. As we have already noted, skill in reading (especially nonword reading) cannot be what is subject to the processes of genetic transmission and the pressures of natural selection. We must therefore ask what more basic skills might be inherited and might underlie ability in reading. One obvious possibility is that reading is heritable because of its association with intelligence. This possibility has been rejected in analyses by Stevenson et al. (1987), Stevenson (1991), and Olson et al. (1989) that have controlled for the effects of IQ. It also does not fit very well with the differential heritability for phonological decoding and wordspecific knowledge found by Olson et al. An alternative basic skill that could underlie the heritability of reading is an aspect of language. Specialized brain structures have evolved for the processing of language, and individual differences in the maturation (or functional efficiency) of these structures are likely to be under genetic control. Some evidence that supports

I26

John P. Rack et al.

this link between language and reading has been provided by Olson and his colleagues. Olson et al. (1989) measured underlying phonological abilities in their twin sample using measures of phoneme segmentation and rhyming ability of the type that we have already discussed. They found that the phoneme segmentation tasks, but not the rhyme task, were heritable. Furthermore, the strong correlation between the phoneme segmentation tasks and phonological decoding was largely attributable to genetic factors. The finding that rhyme was not heritable may seem surprising in the light of our discussion of the importance of this variable; however, the subjects studied by Olson et al. were quite old (average age 13 years). It is possible that rhyming ability would be found to be heritable if measured at an earlier age. if this were true, it would imply that it is an ability that underlies early rhyming skill (not rhyming skill per se) that influences reading development. We return to this issue at the end of the paper. The Olson et al. (1989) results therefore demonstrate a genetic link between phonemic segmentation and reading. Given the biological primacy of the phonological system, it would make sense to conclude that variations in reading skills emerge from variations in phonological skills that are inherited. Despite the logic behind the argument, direct evidence that the causal relationship is in this direction is, as yet, lacking. Two further connections are required if a biological theory of reading development is to help clarify the nature of the relationship between genes and reading. The first of these is the link between genes and brain anatomy and physiology. Steps are being taken to understand this link by trying to identify which genes are responsible for reading problems (Smith, Kimberling, Pennington, & Lubs, 1983; Smith, Pennington, Kimberling, & Ing, 1990). It may in due course be possible to determine what proteins are coded by those genes and what role these proteins have in brain functioning. The second link that the theory needs is between brain functioning and reading. Here again, work is at a very early stage but first steps have been taken to try to identify structural abnormalities in the brains of dyslexic individuals. One result that has now been clearly established is that dyslexics tend not to show the usual asymmetry (favoring the left hemisphere) of the planum temporale (Galaburda, Corsiglia, Rosen, & Sherman, 1987; Hynd & Semrud-Clikeman, 1989; Larsen, Hoien, Lundberg, & Odegaard, 1990). The planum temporale is defined as the posterior aspect of the superior temporal lobe lying, on the left side, at the center of Wernicke’s speech area. Further evidence from a variety of sources indicates that this area of the cortex is involved in language processing (see Steinmetz & Galaburda, 1991). Understanding the relation between brain abnormalities and reading disorders will be necessary for a complete understanding of reading development; however, many methodological difficulties pervade the research in this area (see Hynd & Semrud-Clikeman, 1989). Future progress in this promising direction awaits the resolution of these difficulties. At this time we note that the evidence

Learning to Read

I27

that abnormalities in dyslexics’ brains are located in areas specialized for language processing does converge well with other evidence reviewed in this paper.

VIII. Conclusions We have considered some of the major approaches to a theory of learning to read. The theorists whose work we have reviewed have started from a number of different positions and have focused on different levels of explanation. Nevertheless we feel that a good deal of contact has been made between the different approaches, and a more unified theory is beginning to emerge. Such a unified theory will need to operate at a number of different levels of description and explanation. At the basic descriptive level, studies such as Seymour and Elder’s (1986) and Stuart and Coltheart’s (1988) have provided interesting and important information; however, we still do not know the answers to basic questions such as how vocabulary size typically changes over time and what skills foster its growth at different points in time. We believe that a better description of the nature of reading development will come from the use of theoretically motivated measures. The greatest achievements in the field of reading development have occurred at the second level that we identified as concerned with external causation. In particular, a great deal of empirical evidence supports the theory that reading is influenced by basic language skills, particularly phonological skills. Children’s sensitivity to rhymes and other units of spoken language is predictive of their later reading ability. Training studies that have been aimed at altering children’s language abilities either before (Lundberg et al., 1988) or during (Bradley & Bryant, 1985) reading instruction have provided further support for the key role played by phonology in learning to read. However, clarification of the nature of these underlying phonological abilities and what manifestations they may have at different points in development is still needed. At the third level of explanation, we considered some explicit models of reading and reading development. We argued that information processing models are too limited as means of capturing the interactive and dynamic nature of development. Connectionist approachesto modeling regarding development seem much more promising; however, connectionist models, in their current forms, do not take account of existing knowledge that might influence the formation of an associative network. We are currently developing a network that does make use of preexisting phonological knowledge in learning print-sound relationships. The fourth and final level of explanation concerns the biological substrates of reading ability. Here, work on the genetics of reading ability converges very well on the other levels of theory. Reading skills appear to be inherited to the extent that they involve phonological processing. In particular, the strong link between spoken language processing and reading skills appears to be underpinned by a common genetic basis.

128


ACKNOWLEDGMENTS Preparation of this article was supported by Grant SPG89202 17 from the Tricouncil Initiative in Cognitive Science (UK). We thank Dick Olson for many discussions that helped form some of the ideas presented here and Linnea Ehri for her comments on an earlier version of the article.

REFERENCES Backman, 1. E., Mamen, M., & Ferguson, H. 9 . (1984). Reading level design: Conceptual and methodological issues in reading research. Psychological Bulletin, 96,560-568. Bishop, D. (1989). Unfixed reference, monocular occlusion, and developmental dyslexia-A critique. British Journal of Ophthalmology, 73, 209-215. Bradley, L. L., & Bryant, P. (1978). Difficulties in auditory organization as a possible cause of reading backwardness. Narure. 271, 746-747. Bradley, L. L., & Bryant, P. E. (1983). Categorizing sounds and learning to read: A causal connexion. Nature, 301, 419. Bradley, L. L., & Bryant, P. E. (1985). Rhyme and reason in reading and spelling. Ann Arbor: University of Michigan Press. Breitmeyer, B. 0 . (1984). Visual masking: An integrative approach. New York: Oxford University Press. Bryant, P. E., & Goswami, U. C. (1986). Strengths and weaknesses of the reading level design: A comment on Backman, Mamen, & Fergusson. Psychological Eullerin, 100, 101-103. Bryant, P. E., & Goswami, U. C. (1987). Phonological awareness and learning to read. In J. Beech & A. Colley (Eds.), Cognitive approaches to reading (pp. 163-178). Chichester: Wiley. Bryant, P. E., MacLean, M.. Bradley, L. L., & Crossland, J. (1990). Rhyme and alliteration, phoneme detection, and learning to read. Developmenrul Psychology, 26, 429-438. Cataldo, S., & Ellis, N. C. (1990). Learning to spell, learning to read. In P. D. Pumfrey & C. D. Elliot (Eds.), Children's dificulties in reading, spelling and writing. Basingstoke: Falmer Press. Conners, F. A,, & Olson, R. K. (1990). Reading comprehension in dyslexic and normal readers: A component skills analysis. In D. A. Balota, G. 9 . Flores d'Arcais, & K. Rayner (Eds.), Comprehension processes in reading (pp. 1-23). Hillsdale, NJ: Erlbaum. DeFries, J. C., Olson, R. K., Pennington, B. F., & Smith, S. D. (1990). Colorado reading project: An update. In D. B. Gray & D. D. Duane (Eds.), The reading brain: The biological bases of dyslexia. Parkton, MD: York Press. Duane, D. D. (1991). Neurobiological issues in dyslexia. In M.J. Snowling & M. Thompson (Eds.), Dyslexia: Integrating theory and pracrice (pp. 21-30). London: Whurr. Ehri, L. C. (1987). Learning to read and spell words. Journal of Reading Eehaviour, 19, 5-31. Ehri, L. C. (1989). The development of spelling knowledge and its role in reading acquisition and reading disability. Journal of Learning Disabiliries, 22, 356-365. Ehri, L. C. (1992). Reconceptualizing the development of sight-word reading and its relationship to decoding. In P. 9. Gough, L. C. Ehri, & R. Treiman (Eds.), Reading acquisition. Hillsdale, NJ: Erlbaum. Ehri, L., 8r Wilce, L. S. (1980). The influence of orthography on readers' conceptualization of the phonemic structure of words. Applied Psycholinguistics, 1. 371-385. Ehri, L.C., & Wilce, L. S. (1985). Movement into reading: Is the first stage of printed word learning visual or phonetic? Reading Research Quarterly. 20. 163-179. Ehri, L. C., & Wilce, L. S. (1987). Cipher versus cue reading: An experiment in decoding acquisition. Jownal of Educational Psychology, 79, 3- 13.

Learning to Read

129

Ellis, N. C. (1991). Spelling and sound in learning to read. In M. J. Snowling & M. Thompson (Eds.), Dyslexia: Integrating theory and practice (pp. 80-94). London: Whurr. Frith, U. (1985). Beneath the surface of developmental dyslexia. In K. E. Patterson, J. C. Marshall, & M. Coltheart (Eds.), Surface dyslexia (pp. 301-330). London: Routledge & Kegan Paul. Galaburda, A. M. (1989). From reading to neurons. Cambridge, MA: MIT Press. Galaburda, A. M . , Corsiglia, J., Rosen, G. D., & Sherman, G. F. (1987). Planum temporale asymmetry: Reappraisal since Geschwind and Levitsky. Neuropsychologia, 25, 853-868. Goswami, U. (1 988). Orthographic analogies and reading development. QuarterlyJournal of Experimental Psychology. 40A, 239-268. Goswami, U., & Bryant, P. E. (1990). Phonological skills and learning to read. London: Erlbaum. Hallgren, B. (1950). Specific dyslexia (congenital word-blindness): A clinical and genetic study. Acta Psychiatrica er Neurologica Supplement. 65. 1-287. Hatcher, P., Hulme, C., & Ellis, A. W. (1992). Overcoming early reading failure by integrating the teaching of reading and phonological skills. Manuscript submitted for publication Hebb, D. 0. (1949). The organisation of behavior: A neuropsychological approach. New York: Wiley. Helfgott, J. A. (1976). Phonemic segmentation and blending skills of kindergarten children: Implications for beginning reading acquisition. Contemporary Educational Psychology, I , 157- 169. Hinton, G. E., & Shallice, T. (1990). Lesioning an attractor network: Investigations of acquired dyslexia. Psychological Review, 98, 74-75. Hulme, C. (1987). Reading retardation. In J. Beech & A. Colley (Eds.), Cognitive approaches to reading. Chichester: Wiley. Hulme, C. (1988). The implausibility of low-level visual deficits as a cause of children’s reading difficulties. Cognitive Neuropsychology. 5 , 369-374. Hulme, C., and Snowling, M. (1991). Phonological deficits in dyslexia: A “sound” reappraisal of the verbal deficit hypothesis? In N. Singh and 1. Beale (Eds.), Progress in learning disabiliries. New York: Springer-Verlag. Hulme, C., Snowling, M . , and Quinlan, P. (1991). Connectionism and learning to read: Steps towards a psychologically plausible model. Reading and Writing. 3, 159-168. 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 & Brain Sciences, 8. 689-740. Hynd, G. W.,& Semrud-Clikeman, M. (1989). Dyslexia and brain morphology. Psychological Bulletin, 106. 447-482. Jackson, N. E., & Butterfield, E. C. (1989). Reading-level-matched designs: Myths and realities. Journal of Reading Behavior. 21. 387-412. Kulikowski, J. J., & Tolhurst, D. J. (1973). Psychophysical evidence for sustained and transient detectors in human vision. Journal of Physiology, 232, 149- 162. Lachter, J., & Beaver, T. G. (1988). ‘There’s no safety in numberes. . . . or anything else.’ Cognition, 28, 195-274. Larsen, 1. P., Hoien, T., Lundberg, I . , & Odegaard, H. (1990). MRI evaluation of the size and symmetry of the planum temporale in adolescents with developmental dyslexia. Brain and Language. 39, 289-301. Liberman, I. Y.,Shankweiler, D., Fischer, W. F., & Carter, B. (1974). Reading and the awareness of linguistic segments. Journal of Experimental Child Psychology, 18, 201-21 2. Lovegrove, W. L. (1991). The visual deficit hypothesis. In N. Singh and 1. Beale (Eds.),Progress in Learning Disabilities. New York: Springer-Verlag. Lovegrove, W., Martin, F.. & Slaghuis, W. (1986). A theoretical and experimental case for a visual deficit in specific reading disability. Cognitive Neuropsychology, 3, 225-267.

I30

John P. Rack et al.

Lundberg, I., Frost, J., & Peterson, 0. (1988). Effects of an extensive program for stimulating phonological awareness in preschool children. Reading Research Quarterly, 23, 263-284. Lundberg, I., Olofson, A., & Wall, S. (1980). Reading and spelling skills in the first school years predicted from phonemic awareness skills in kindergarten. Scandinavian Journal of Psychology. 21. 159-173. Mann, V. A. (1984). Reading skill and language skill. Developmental Review. 4 . 1-15. Marsh, G.,Friedman, M., Welch, V., & Desberg, P. (1981). A cognitive-developmental theory of on. In G.E. MacKinnon & T. G.Waller (Eds.), Reading research: Advances in theory and practice (Vol. 3). New York: Academic Press. Morais, J. (1991). Metaphonological abilities and literacy. In M. 1. Snowling & M. Thompson (Eds.), Dyslexia: Integrating theory and practice (pp. 95-107). London: Whurr. Morton, J. (1969). Interaction of information in word recognition. Psychological Review, 76, 165178. Morton, J. (1989). An information-processing account of reading acquisition. In A. M. Galaburda (Ed.), From reading to neurons (pp. 43-46). Cambridge, MA: MIT Press. Olson, R. K., Gillis, J. J., Rack, J. P., DeFries, I. C., & Fulker, D. W. (1991). Confirmatory factor analysis of word recognition and process measures in the Colorado reading project. Reading and Writing, 3, 235-248. Olson, R. K., Wise, B., Conners, F. A., & Rack, J. P. (1990). Organization, heritability, and remediation of component word recognition and language skills in disabled readers. In T.H.Can & B. A. Levy (Eds.), Reading and its development: Component skills approaches (pp. 261-322). New York: Academic Press. Olson, R. K., Wise, B., Conners, F., Rack, J., & Fulker, D. (1989). Specific deficits in component reading and language skills: Genetic and environmental influences. Journal of Learning Disabilities, 22, 339-348. Pennington, B. F. (1989). Using genetics to understand dyslexia. Annals of Dyslexia, 39, 81-93. Pennington, B. F. (1991). Reading disabilities: Genetic and neurological infuences. Dordrecht: Kluwer. Perfetti, C. A,, Beck, I., Bell, L. C., & Hughes, C. (1987). Phonemic knowledge and learning to read are reciprocal: A longitudinal study of first grade children. Merrill-Palmer Quarterly, 33, 283-319. Perin, D. (1983). Phonemic segmentation and spelling. British Journal of Psychology, 74, 129-144. Pinker, S . , & Prince, A. (1988). On language and connectionism: Analysis of a parallel distributed model of language acquisition. Occasional paper No. 33, Center for Cognitive Science, MIT. Plornin, R., DeFries, J. C., & McClearn, G. E. (1980). Behavioral genetics: A primer. New York; W.H. Freeman. Quintan, P. T. (1991). Connectionism and psychology. Chicago: University of Chicago Press. Rack, J. (1985). Orthographic and phonetic coding in normal and dyslexic readers. British Journal of Psychology, 76, 325-340. Rack, J. P., & Snowling, M.J. (1985). Verbal deficits in dyslexia: A review. In M. J. Snowling (Ed.), Children’s written language difficulties: Assessment and management. Windsor: NFER-Nelson. 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. Rogosa. (1980). A critique of cross-lagged correlation. Psychological Bulletin, 88. 245-258. Rumelhart, D. E., Hinton, G. E., & Williams, R. J. (1986). Learning internal representations by error propagation. In D. E. Rumelhart, J. L. McClelland, & the PDP Research Group (Eds.), Parallel distributed processing: Explorations in the microstructure of cognition (Vol. 1). Cambridge, MA: MIT Press. Scarborough, H. S. (1989). Prediction of reading disability from familial and individual differences. Journal of Educational Psychology, 81, 101- 108.

Learning to Read

131

Seidenberg, M. S., & McClelland, J. L. (1989). A distributed developmental model of wordrecognition and naming. Psychological Review, 96, 523-568. Seidenberg, M. S., & Tanenhaus, M. K. (1979). Orthographic effects on rhyme monitoring. Journal of Experimental Psychology: Human Learning and Memory. 5 , 546-554. Seymour, P. H. K. (1986). Cognitive analysis of dyslexia. London: Routledge & Kegan Paul. Seymour, P. H. K. (1990).Developmental dyslexia. In M. W. Eysenck (Ed.), Cognitive psychology; An international review ( pp. 135- 196). Chichester: Wiley. Seymour, P.H. K., & Bunce, F. (in press). Application of cognitive models to remediation in cases of developmental dyslexia. In G. Humphreys & J. Riddoch (Eds.), Cognitive neuropsychology and cognitive rehubilitation. Hove: Lawrence Erlbaum Associates. Seymour, P. H. K., Bunce, F., & Evans, H. (in press). A framework for orthographic assessment and remediation. In C. Sterling & C. Robson (Eds.), Psychology. spelling. education. Clevedon: Multilingual Matters. Seymour, P. H. K., & Elder, L. (1986). Beginning reading without phonology. Cognitive Neuropsychology, 3 , 1-36. Seymour, P. H. K., & MacGregor, C. J. (1984). Developmental dyslexia: A cognitive developmental analysis of phonological morphemic and visual impairments. Cognitive Neuropsychology, I , 4382. Shankweiler, D., & Liberman, I. Y. (Eds.) (1990). Phonology and reuding disability: Solving fhe rending puzzle. Ann Arbor: University of Michigan Press. Share, D. L., Jorm, A. F., Maclean, R., & Matthews, R. (1984). Sourcesof individual differences in reading acquisition. Journal of Educational Psychology, 76, 1309- 1324. Smith, S. D., Kimberling, W. J., Pennington, B. F., & Lubs, H. A. (1983). Specific reading disability: Identification of an inherited form through linkage analysis. Science, 219, 1345- 1347. Smith, S. D., Pennington, 9. F., Kimberling, W. J., & Ing, P. A. (1990). Familial dyslexia: Use of genetic linkage to identify subtypes. Journal of the American Academy of Child Psychiatry, 29, 204-213. Stanovich, K. E. (1986). Cognitive processes and the reading problems of learning disabled children: Evaluating the assumption of specificity. In J. K.Torgesen & B. Y.L. Wong (Us.),Psychological and educational perspectives on learning disabilities ( pp. 87- 13 1). Orlando, FL: Academic Press. Stanovich, K. E. (1988). Explaining the differences between the dyslexic and the garden-variety poor reader: The phonological-core variable-difference model. Journal of Learning Disabilities, 21, 590-612. Stanovich, K. E . , Cunningham, A. E., & Cramer, B. B. (1984). Assessing phonological skills in kindergarten children: Issues of task comparability. Journal of Experimental Child Psychology, 38, 175-190. Steinmetz, H., & Galaburda, A. M. (1991). Planum temporale asymmetry: In vivo mophometry affords a new perspective for new-behavioral research. Reading and Writing, 3, 331-343. Stephenson, S. (1907). Six cases of congenital word-blindness affecting three generations of one family. Ophthalmoscope, 5 , 482-484. Stevenson, J. (1991). Which aspects of processing text mediate genetic effects. Reading and Wriring. 3, 249-269. Stevenson, J., Graham, P., Fredman, G., & McLoughlin, V. (1987). A twin study of genetic influences on reading and spelling ability and disability. Journal of Child Psychology and Psychiatry, 28, 229-247. Stuart, M., & Coltheart, M. (1988). Does reading develop in a sequence of stages? Cognition, 30. 139-181. Thomas, C. J. (1905). Congenital 'word-blindness' and its treatment. Opthalmoscope, 3, 380385.

132


Treiman, R. (1985). Onsets and rimes as units of spoken syllables: Evidence from children. Journal of Experimental Child Psychology, 39. 161- 181. finmer, W.E., Herriman, M. L., & Nesdale, A. R. (1988). Metalinguistic abilities and beginning reading. Reading Research Quarterly, 23, 134- 158. Van Orden, 0 . C., Pennington, B. F., & Stone, G.0. (1990). Word identification in reading and the promise of subsymbolic psycholinguistics. Psychological Review, 97, 488-522. Vellutino, F. R . (1979). Dyslexia: Theory and research. Cambridge, MA: MIT Press. Vogler, G. P.,DeFries, J. C., & Decker, S. N. (1985). Family history as an indicator of risk for reading disability. Journal of Learning Disabilities, 18, 419-421, 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.

DOES READING MAKE YOU SMARTER? LITERACY AND THE DEVELOPMENT OF VERBAL INTELLIGENCE

Keith E . Stanovich ONTARIO INSTITUTE FOR STUDIES IN EDUCATION TORONTO, ONTARIO. CANADA M5S 1V6

I. INTRODUCTION 11. THE RISE OF THE “GREAT DIVIDE” THEORIES 111. GREAT DIVIDE THEORIES UNDER ATTACK IV. THE DEATH BLOW TO GREAT DIVIDE THEORIES: SCRIBNER AND COLE V. PREMATURE CLOSURE ON THE CONSEQUENCES OF LITERACY?

VI. STUDYING THE DIRECT CONSEQUENCES OF LITERACY WITHIN A LITERATE SOCIETY A. INTRODUCTION B. THE RESEARCH STRATEGY C. ASSESSING PRINT EXPOSURE THE DIARY TECHNIQUE D. ASSESSING PRINT EXPOSURE: THE RECOGNITION CHECKLIST TECHNIQUE E. VALIDATING CHECKLIST MEASURES OF PRINT EXPOSURE

VII. THE SPECIFIC COGNITIVE CORRELATES OF PRINT EXPOSURE A. PRINT EXPOSURE AS A CONTRIBUTOR TO GROWTH IN COMPREHENSION ABILITY B. PRINT EXPOSURE AS A CONTRIBUTOR TO GROWTH IN OTHER VERBAL SKILLS C. PRINT EXPOSURE AND INCIDENTAL VERSUS INTENTIONAL LEARNING D. THE RECOGNITION CHECKLISTS AND READING IN THE REAL WORLD VIII. SUMMARY AND CONCLUSIONS REFERENCES 133 ADVANCES IN CHILD DEVELOPMENT AND BEHAVIOR. VOL. 24

Copynght QI 1993 by Academic Press, Inc. All rights of reproducuon in any form reserved.

Keith E . Stanovich

134

I.

Introduction

What role do experiential differences play in determining variation in cognitive growth? This question has been at the heart of much theorizing in developmental psychology. Enthusiasm for experiential explanations of differences in cognitive growth has waxed and waned over the years, as has interest in explanations based on genetic inheritance. For example, interest in genetic determinants of differences in cognitive skills increased considerably in the 1980s (Plomin, DeFries, & McClearn, 1990; Thompson, Dettennan, & Plomin, 1991), and some once-popular experiential hypotheses went out of favor altogether. Theories in which literacy is posited to be a determinate of individual differences in cognitive growth provide a case in point. Differential experience with print was once an important mechanism in many theories of cognitive change (Greenfield, 1972; Olson, 1977). In the 1980s, the idea that the acquisition of literacy has profound cognitive consequences went seriously out of favor (Erickson, 1984; Gee, 1988; Scribner & Cole, 1981; Street, 1984, 1988). In this article, I argue that we should reconsider experience with print as an explanatory mechanism that can account for cognitive change. My argument begins with a review of selected literature on the consequences of literacy in which I claim that a role for reading experience in theories of cognitive change seems to have been prematurely dismissed. I then introduce a methodology for studying the cognitive consequences of literacy within a literate society that could help to revive research interest in this hypothesis.

11.

The Rise of the “Great Divide” Theories

Theories in which literate and nonliterate individuals and societies are posited to exhibit important cognitive differences have been termed Great Divide r h o ries (Erickson, 1984; Olson & Torrance, 1991; Scribner & Cole, 1981). As is discussed later, in the 1980s the term was most often used pejoratively (e.g., Street, 1984). Great Divide theories can be partitioned according to whether they are concerned primarily with the indirect, mediated effects of literacy-habits of thought derived through cultural immersion in a literate society-or the direct, nonmediated effects of literacy on a particular individual’s cognitive processes and knowledge structures (Goody, 1987; Scribner & Cole, 1978). For example, illiterates, or people who engage only marginally in literacy activities, may derive certain cognitive benefits from participation in a literate culture. These have been termed the mediated effects of literacy (Goody, 1987, pp. 217-252), and anthropologists and historians have done much work to assess these culturally mediated consequences of literacy. In contrast, psychologists have been

Literacy and Verbal Intelligence

135

more concerned with examining the individual effects of having personally engaged in reading/writing activities (Scribner & Cole, 1978, 1981). Great Divide theorists are further differentiated according to whether they posit a continuous range of effects, that is, effects that are linked to variation in printrelated activities that exist even within literate populations. Greenfield’s seminal (1972) work provides a case in point. Based on her cross-cultural research in Africa, Greenfield (1972) posited that facility with written language developed a set of cognitive competencies that were more elaborate than those associated with purely oral language. Her argument was based on the differences in context dependency between written and oral language: If a speaker of an oral language depends upon the surrounding context to communicate his message, then effective communication presupposes a common context and common point of view for both listener and speaker. The speaker, moreover, must assume that this is the case. He is, therefore, egocentric. . . . Speech based on a written language, in contrast, must be relatively independent of context for a number of reasons. (p. 170)

The central thesis of Greenfield’s argument was that “Context-dependent speech is tied up with context-dependent thought, which in turn is the opposite of abstract thought” (p. 169). Greenfield outlined several examples of how contextindependent language fosters abstract thought and problem solving. Although most of her discussion derived from her work done among the Wolof of Senegal, Greenfield (1972) extended her hypotheses to encompass cultural subgroups in the United States who were not illiterate but who had less exposure to written language. That is, she championed a continuous version of the Great Divide hypothesis by positing that differences in degrees of print exposure within even a generally literate society also have cognitive consequences. Greenfield’s hypotheses concerning the direct effects of literacy on an individual reader’s cognition paralleled theories of the effects of literacy at the societal level that were popular among anthropologists and historians (Akinnaso, 1981; Goody, 1977, 1987; Havelock, 1963, 1980; Musgrove, 1982; Ong, 1967, 1982). Goody’s influential writings (1977, 1980; Goody & Watt, 1968) contain hypotheses about the effects of literacy at the level of societies and cultures rather than individuals: The specific proposition is that writing, and more especially alphabetic literacy, made it possible to scrutinise discourse in a different kind of way by giving oral communication a semi-permanent form; this scrutiny favored the increase in scope of critical activity, and hence rationality, scepticism, and logic to resurrect memories of those questionable dichotomies. It increased the potentialities of criticism because writing laid out discourse before one’s eyes in a different way; at the same time [it] increased the potentiality for cumulative knowledge, especially knowledge of an abstract kind, because it changed the nature of communication beyond that of face-to-face contact as well as the system for the storage of information; in this way a wider range of

136

Keith E . Stanovich

“thought” was made available to the reading public. . . . [it] enabled man to stand back from his creation and examine it in a more abstract, generalised, and rational way. (1977, p. 37)

Elaborations of this argument have echoed throughout the anthropological and historical literature on the effects of literacy (Havelock, 1963, 1980; Musgrove, 1982; Ong, 1982). Ong (1982) made the case for the cognitive effects of literacy most strongly: “Without writing, the literate mind would not and could not think as it does, not only when engaged in writing but normally even when it is composing its thoughts in oral form. More than any other single invention, writing has transformed human consciousness” (p. 78). The causal mechanism emphasized by Ong (1982) was a variant on the GoodylGreenfield theme: “Writing fosters abstractions that disengage knowledge from the arena where human beings struggle with one another. It separates the knower from the known” (pp. 43-44). Olson (1977, 1986b, 1988) has presented a related causal theory of how literacy comes to influence thought. In his important 1977 essay Olson contrasted texts (written prose statements) with utterances (informal oral-language statements). His thesis, which he made clear was intended to apply to both the societal and individual consequences of literacy, was that “There is a transition from utterance to text both culturally and developmentally and this transition can be described as one of increasing explicitness, with language increasingly able to stand as an unambiguous or autonomous representation of meaning” (p. 258). Olson emphasized the importance of the assumption that meaning resides autonomously within the text and “the consequences of that assumption, particularly of the attempts to make it true” (p. 258). The highest form of the autonomous text ideal is the essayist technique: “The more fundamental effect of this approach to text was on the writer, whose task now was to create autonomous text-to write in such a manner that the sentence was an adequate, explicit representation of the meaning, relying on no implicit premises or personal interpretations” (p, 268). He argued that Logical development in a literate culture involves learning to apply logical operations to the sentence meaning rather than to the assimilated or interpreted or assumed speaker’s meaning. Development consists of learning to confine interpretation to the meaning explicitly represented in the text and to draw inferences exclusively from that formal but restricted interpretation. . . , The developmental hypothesis offered here is that the ability to assign a meaning to the sentence per se, independent of its nonlinguistic interpretive context, is achieved only well into the school years. (pp. 274275)

Hypotheses of the type put forth by Greenfield, Goody, and Olson came to be termed Great Divide theories because of the number and importance of the cognitive consequences of literacy that were assumed. Among these were


I37

Logical and analytic modes of thought; general and abstract uses of language; critical and rational thought; a skeptical and questioning attitude; a distinction between myth and history; the recognition of the importance of time and space; complex and modem governments (with separation of church and state); political democracy and greater social equity; economic development. . . . It leads to people who are innovative, achievement oriented, productive, cosmopolitan, politically aware, more globally (nationally and internationally) oriented. (Gee, 1988, p. 196)

111. Great Divide Theories under Attack By the late 1970s and early 1980s the Great Divide position had garnered considerable support and momentum. By the late 1980s, however, the situation had changed drastically. By then, many scholars had accepted Graffs (1979, 1986, 1987) characterization of the Great Divide theory as the “literacy myth” and Gee (1988) was claiming that “At least in academic circles, the literacy myth is on its last legs” (p. 196). What happened to cause such a rapid and extreme theoretical reversal? Not surprisingly, a number of interacting factors acted in concert to cause the collapse of the “literacy myth.” First, literacy’s effects at the societal level were brought into question by historical studies indicating that literacy was intertwined with certain cultural effects in a much more intricate and interactive way than was implied by some of the simpler theories that emphasized unidirectional causation. For example, the link between economic development and national levels of literacy has turned out to be much more complex than originally thought. Literacy levels are as much a consequence of economic development as they are its cause (Fuller, Edwards, & Gorman, 1987; Graff, 1986, 1987; Kaestle, 1991; Wagner, 1987). The plausibility of literacy having an effect on cognition at the level of individuals has also been questioned. Some researchers have questioned the distinctions between utterance and text that served as the guiding assumptions of Great Divide theorizing. Clearly, for example, oral speech in formalized settings can contain all the features associated with written text: detachment, certain types of subordination, integration, nominalization (Biber, 1986; Chafe & Danielewicz, 1987; Feldman, 1991; Nystrand, 1987; Redeker, 1984; Tannen, 1982). Similarly, written texts need not always contain these features. Thus, utterances can sometimes have the characteristics of text and vice versa. Acceptance of the idea that the features of utterance and text that were allegedly responsible for differential cognitive effects were different only probabilistically rather than in a discrete sense had the effect of making the “divide” seem less “great.” Adding to these second thoughts about the consequences of literacy was a radical social critique that conceived of literacy as just one more mechanism used by powerful groups to maintain social privilege. For example, Street (1984)

138

Keith E . Stanovich

claimed that “The actual examples of literacy in different societies that are available to us suggest that it is more often ‘restrictive’ and hegemonic, and concerned with instilling discipline and exercising social control” (p. 4). and that “Schooling and techniques of teaching literacy are often forms of hegemony” (p. 1 1). Another author titled her book The Violence of Literacy and argued that American society “stakes much on the oppressive powers of literacy” (Stuckey, 1991, p. 30) and that “it is possible that a system of ownership built on the ownership of literacy is more violent than past systems, however. Though it seems difficult to surpass the violence of systems of indenture, slavery, industrialism, and the exploitation of immigrant or migrant labor, literacy provides a unique bottleneck” (p. 18). Gee (1988) echoed the argument that “literacy has been used, in age after age, to solidify the social hierarchy, empower elites, and ensure that people lower in the hierarchy accept the values, norms and beliefs of the elites, even when it is not in their self-interest” (p. 205). In short, the social effects of literacy are no longer universally viewed as positive, at least by some educational theorists. Great Divide theories seemed, to these same scholars, to be a case par excellence of blaming the victims. Thus, the popular social critiques of the 1980s were used as weapons against any theory of the consequences of literacy that posited substantial cognitive effects following from differential engagement in literacy activities. Because people do differentially engage in literacy activities, any such effects were bound to create cognitive inequalities that most neo-Marxist and socioconstructivist theorists were committed to denying. Great Divide theories ran straight into the brick wall of cultural and epistemological relativism that was a foundational assumption of these social critiques (for discussions, see Gellner, 1985; Hollis & Lukes, 1982; Musgrove, 1982, Shweder, 1991; Siegel, 1988; Sperber, 1985). The critiques seemed to rest on the tenuous assumption that literacy creates enormous sociopolitical differences that are not associated with any concomitant cognitive differences.

IV.

The Death Blow to Great Divide Theories: Scribner and Cole

Thus, a confluence of academic critiques in the 1980s undermined hypotheses about the effects of literacy; however, another factor was probably more influential than all the academic critiques combined. This factor was the investigation into literacy effects among the Vai in Africa by Scribner and Cole (1981). The work of Scribner and Cole provides one of the firmest foundations for the assaults on the “literacy myth” and it is repeatedly cited in critiques of Great Divide theories:


I39

In the Scribner and Cole study, literacy in and of itself led to no grandiose cognitive abilities. (Gee, 1988, p. 203) The Vai findings caution us against such generalizations as are often made: that writing promotes general mental abilities. (Akinnaso, 1981, p. 175) Cole and Scribner point toward an interpretation that contradicts the usual view that literacy leads inevitably to higher forms of thought. . . . Such research may also control the assumptions and expectations that students carry to studies of literacysuch as presupposing literacy to be “liberating” or “revolutionary” in its consequences. There are, I suggest, better reasons to expect the opposite to be more often the case. (Graff, 1987, pp. 23-24) Scribner and Cole (1981) conclude that literacy per se contributes only marginally to cognitive development. (Nystrand, 1987, p. 236)

Scribner and Cole’s investigation was ground breaking because they sought to separate the effects of literacy from the effects of schooling. The confounding of these two factors in earlier research (e.g., Greenfield, 1972; Luria, 1976) rendered tenuous any conclusions about the effects of literacy per se. Scribner and Cole took advantage of the fact that three scripts are in use among the Vai and that each script is associated with a particular context. English is learned in school and is used in formal settings (e.g., dealings with the government); Arabic is used for reading, writing, and memorizing the Koran; and an indigenous Vai script is transmitted outside of institutional settings and is used in personal correspondence and for some business purposes. Although some of the Vai are fluent in multiple scripts and some are illiterate, the fact that some individuals are familiar with only one script allows the separation of schooling effects from literacy effects. The comparison of Vai monoliterates with illiterates and individuals schooled in English is particularly diagnostic. Scribner and Cole (1981) found no specific effect of Vai literacy on a number of tasks tapping general cognitive processes, including geometric sorting tasks, taxonomic categorization tasks, memory tasks, and syllogistic reasoning problems. Scribner and Cole concluded that “Effects of nonschooled literacies are spotty and appear on only a few performance measures. . . . These surely disappoint the grand expectations and lofty theories that inspired us to undertake this line of investigation” ( pp. 130- 132). The authors further argued that the findings “lay to rest some misconceptions about the psychology of literacy that went unchallenged in the past for lack of empirical data. First, it is clear from the evidence we reviewed that nonschooled literacy, as we found and tested it among the Vai, does not produce general cognitive effects as we have defined them. The small and selective nature of Vai script and Arabic influences on cognitive performance precludes any sweeping generalizations about literacy and cognitive change” (p. 132). Only when Scribner and Cole changed the focus of their research program to metalinguistic tasks more tightly and specifically linked to reading and writing (e.g., grammatical judgment, rebus reading, integrating syllables) did they find any specific effects of Vai literacy.

I 40

Keith E . Stanovich

Despite the existence of some “spotty effects” (p. 244), Scribner and Cole’s summary conclusion has been accepted by many investigators: Our results are in direct conflict with persistent claims that “deep psychological differences” divide literate and nonliterate populations. On no task-logic, abstraction, memory, communication-did we find all nonliterates performing at lower levels than all literates. Even on tasks closely related to script activities, such as reading or writing with pictures, some nonliterates did as well as those with school or literacy experiences. We can and do claim that literacy promotes skills among the Vai, but we cannot and do not claim that literacy is a necessary and sufficient condition for any of the skills assessed. (p. 251)

One indirect effect of the widespread acceptance of interpretations of the Scribner and Cole results was that in the 1980s very few cognitive and developmental psychologists conducted empirical studies on the individual consequences of literacy. The seeming conclusiveness of the Scribner and Cole investigation and the difficulty of conducting studies on these issues dampened enthusiasm for new empirical investigations of the effects of literacy. In the rest of this article, I reopen the issue, first by arguing that the consensus against the idea of profound cognitive consequences of literacy was arrived at too hastily and then by introducing a methodology for studying the cognitive consequences of literacy within a generally literate society.

V.

Premature Closure on the Consequences of Literacy?

Although acknowledging the ground-breaking nature of the Scribner and Cole project, I still would point out that their results are often overinterpreted in the literature on literacy. A major issue that is often glossed over-but that, interestingly, was raised by Scribner and Cole themselves-is whether the nature of Vai literacy was such that it provided a valid test of the claims of Great Divide theorists. Olson (1977), for example, was clear that the literacy on which he staked his claim is the high-level literacy characterized by the use of an essayist style. Goody (1987, p. 252), in his argument for the cultural effects of literacy, was also clear that he referred to the type of literacy that enables the reader to have access to a wide range of the world’s accumulated knowledge. Vai literacy is simply not of this type. Individuals typically do not learn the Vai script until their late teens or twenties. It is used primarily for personal letter writing among people who know each other and for conducting business with those with whom one is familiar. Vai writing does not contain the autonomous essayist form that Olson (1977) argued is the causal mechanism that spurs cognitive change. Instead, the letter writing that looms so large in the society of Vai-script literates is highly personalized, assuming elaborate shared knowledge between the letter writer and the recipient (Scribner & Cole, 1981, pp. 71-75). Finally, the Vai


141

have no libraries in this script that can be used by individuals to access the world’s storehouse of knowledge. Scribner and Cole themselves admitted that “literate practices among the Vai are far more restricted than in technologically sophisticated societies. . . . Cultural heritage is transmitted orally in a way that does not depend upon texts. . . . It does not open doors to vicarious experience, new bodies of knowledge, or new ways of thinking about major life problems” ( p. 238). They flatly conceded that “Vai script literacy does not fulfill the expectations of those social scientists who consider literacy a prime mover in social change” (p. 239). This admission is most often omitted in the discussions by authors who use Scribner and Cole’s work to attack the “Literacy Myth.” Theorists who refer to Scribner and Cole’s work often put forth their interpretations without the qualifications introduced by the original authors-another case of prophets’ disciples being more fanatical than the prophets themselves. In short, Scribner and Cole’s research, influential and provocative though it was, should not be considered the final word on the issue of the cognitive consequences of literacy, although surely it would be foolish not to build on their insights.

VI. Studying the Direct Consequences of Literacy within a Literate Society A.

INTRODUCTION

Unfortunately, Scribner and Cole’s innovative and costly project is unlikely to be replicated, so that resolving the issues using a variant of their methodology is not going to be possible; however, the cognitive consequences of literacy can be studied without necessarily using a cross-cultural comparison. I describe here a procedure for studying the cognitive consequences of literacy within a generally literate society. In developing the procedure, we exploited the fact that even within a generally literate culture, individuals vary tremendously in degree of exposure to print (Anderson, Wilson, & Fielding, 1988; Guthrie & Greaney, 1991; Guthrie & Seifert, 1983).Even among a group of individuals who have the same level of assessed reading comprehension ability, remarkably large differences are found in their degree of engagement in print-related activities (Stanovich & West, 1989) and the correlates of this natural variation can be studied. Comparing literates and illiterates is the exclusive design of choice only if the effects of literacy are believed to be completely discontinuous, with no cognitive consequences of variation in amount of print exposure among literate individuals. We speculate that the discontinuity assumption is false and that there is measurable cognitive variation among people who differ only in the amount of reading that they do.

142

Keirh E . Stanovich

In choosing which variables to focus on in our initial investigations, my research group was again influenced by the outcomes of Scribner and Cole’s investigation. In a sense, we started where Scribner and Cole finished. That is, in the first part of their investigation, they concentrated on looking for effects of literacy on tasks that tapped developmental change in general cognitive processes. The tasks in the second part of their investigation-rebus reading, integrating auditory information, word pronunciation, and communication gameswere more closely tied to aspects of Vai literacy and specific effects of literacy on these tasks were easier to demonstrate. In our research program on the cognitive consequences of differences in print exposure, we have inverted the investigative chronology of Scribner and Cole by starting with tasks that are more closely linked to literacy skills. Contingent on positive outcomes in these domains, we have examined more general cognitive processes. Thus, we established our methodology (see Stanovich & West, 1989) by examining criterion variablesorthographic knowledge and spelling-that should clearly be linked to individual differences in print exposure. We then expanded the set of criterion variables to encompass broader domains such as vocabulary, cultural knowledge, and verbal fluency. B . THE RESEARCH STRATEGY

In our methodology, we attempted to correlate differential engagement in reading activities with various cognitive outcomes that have been associated with the acquisition of literacy; however, such an experimental logic, if not supplemented with additional methodological controls, will yield data subject to an inordinately large number of alternative explanations. Levels of print exposure are correlated with too many other cognitive and behavioral characteristics. Avid readers tend to be different from nonreaders on a wide variety of cognitive skills, behavioral habits, and background variables (Guthrie, Schafer, & Hutchinson, 1991; Kaestle, 1991; Zill & Winglee, 1990). Attributing any particular outcome to print exposure uniquely is thus extremely difficult. We have used a hierarchical regression logic first introduced by Anderson et al. (1988) to deal with this problem. The logic of the regression analysis allows any control variables entered first into the regression equation to explain any variance that they can in the criterion variable. Following these control variables, the measures of print exposure are entered. Thus, the procedure allows the investigator to assess whether reliable variance remains to be explained after the control variables are entered and whether print exposure is associated with this remaining variance. In our analyses, we first regressed out ability measures most likely to lead to spurious relationships before examining the linkage between print exposure and criterion variables. This procedure of reducing possible spurious relationships by first partialing


143

relevant ability measures was used in our early investigations of subword processes in reading. For example, in previous work we had demonstrated that, independently of decoding ability, variation in print exposure among adults predicts variation in specific types of orthographic knowledge (Stanovich & West, 1989). Similarly, in a study of children’s performance (Cunningham & Stanovich, 1990) we found that after partialing out IQ, memory ability, and phonological processing abilities, print exposure accounted for significant variance in orthographic knowledge and word recognition. The logic of our analytic strategy is quite conservative because in certain analyses we have actually partialed out variance in abilities that are likely to be developed by print exposure itself (Stanovich, 1986); however, the explanatory ambiguities surrounding a variable such as print exposure have led us to continue to structure the analyses in a “worst case” manner, as far as print exposure is concerned. C. ASSESSING PRINT EXPOSURE: THE DIARY TECHNIQUE

A variety of methods have been used to assess individual differences in exposure to print. For example, a variety of questionnaire and interview techniques have been used to assess relative differences in print exposure (e.g., Estes, 1971; Guthrie, 1981; Guthrie & Greaney, 1991; Guthrie & Seifert, 1983; Lewis & Teale, 1980; Sharon, 1973-1974; Walberg & Tsai, 1984), but many of these are encumbered with reliability and validity problems. A more valid method, but also a more logistically complicate@one, is the use of daily activity diaries filled out by subjects (Anderson et al., 1988; Greaney, 1980; Greaney & Hegarty, 1987; Rice, 1986; Taylor, Frye, & Maruyama, 1990). Activity diaries yield estimates of the actual amount of time spent on literacy activities and are generally more valid than interview or questionnaire instruments (Carp & Carp, 1981). Anderson et al. (1988) used the activity diary method to estimate the amount of time that fifth graders (10- to 11-year-olds) spent reading in their nonschool hours. They found that time spent reading predicted fifth-grade reading comprehension after the variance in second-grade (7- to 8-year-olds) reading comprehension had been controlled. This result seems to indicate that exposure to print was a contributor to individual differences in growth in reading ability over the elementary school years. My research group has completed a series of investigations in which we attempted to determine whether the specific correlates of print exposure can be extended beyond the demonstration by Anderson et al. (1988). We have employed the activity diary method in some of our own studies. Our method of collecting daily activity records was adapted from the nonschool time diary investigation of Anderson et al. (1988), but we also attempted to improve on their methods in several respects (see Allen, Cipielewski, & Stanovich, 1992). Our daily activity record-keeping procedure was designed to minimize the time students would need to spend on it; to minimize the necessity

Keith E. Sranovich

144

for adding and subtracting minutes or converting hours into minutes; and to maximize student judgment accuracy. We collected data over a 3-week period and thus obtained estimates of the average number of minutes per day that the children in our fifth-grade (10- to 11-year-olds) sample spent in various activities when they were outside of school. Table I shows the mean and median minutes per day spent during the nonschool hours in the various categories of activity that were listed on the activity records (with the exception of all reading, which is a composite of the categories book reading, comic reading, and other reading). The figure for book reading in the table includes only those instances in which the child could give at least a fragment of the title or a character from the book (a procedure adapted from Anderson et al., 1988). In contrast, all reading, the most liberal category, included all instances where book reading was indicated on the diary sheets. That the means were generally larger than the medians reflects the positive skew of most of the variables, particularly the reading variables. The positive skew of reading time has repeatedly been observed in activity diary studies (Anderson et al., 1988; Greaney, 1980). Although some of our categories were different from those of the Anderson et al. (1988) study, those that were common TABLE I Time Spent per Day in Various Activities by a Fifth-Grade Sampleu Minutes per dayb Activity

Mean

Median

SD

Reading Books Comics Other Television watching Eating Homework Just playing around Playing outdoor games Talking Family activities Playing indoor games Practices Hobbies Chores Lessons Other

21.3 10.2 2. I 5.8 83.2 52.9 49.0 35.7 25.1 17.0 18.0 14.5 14.3 1.9 6.6 4.5 62.6

16.0 5.0 0.0 3.5 68.0 54.6 45.0 26.0 18.0 13.6 13.9 8.6 8.0 0.0 5.0 0.0 60.0

19.4 15.0 4.0 8.4 65.5 19. I 26.9 32.1 25.2 12.9 18.2 19.1 18.4

0

Adapted from Allen, Cipielewski, and Stanovich (1992). Weekdays and weekends are proportionately represented.

17.5

6.4 7.0 35.0


145

were ordered similarly in the two studies. For example, television watching was the most frequent activity and book reading was far down the list in both studies. Our fifth graders watched less television (83.2 minutes versus 131.1 minutes) and did more homework (49.0 minutes versus 18.9 minutes) than the Anderson et al. fifth graders. These differences might reflect the use of different populations-a private school in our study and public schools in the Anderson et al. ( 1988) study. Previous studies have shown private/public school differences in television and homework habits (Coleman, Hoffer, & Kilgore, 1982). Despite differences in the estimates in other categories, our estimates of book reading time (mean and median of 10.2 and 5.0 minutes, respectively) are very close to those obtained in the Anderson et al. study (10.1 and 4.6 minutes). Certain rough generalizations thus hold across the two studies: Fifth graders ( 10to 11-year-olds) spend around 5 minutes per night reading books for pleasure outside of school, roughly one-tenth the amount of time they spend watching television. These figures call to mind the many studies of school achievement in which American children scored poorly and in which their poor performance was linked to excessive television watching, low levels of homework, and little reading (Applebee, Langer, & Mullis, 1988; Chen & Stevenson, 1989; Stevenson et al., 1985). Our specific concern, however, was whether children’s reading volume related to their achievement and whether such a linkage could be shown to have any specificity. Book reading time (logarithmically transformed, see Allen et al., 1992, and Anderson et al., 1988) correlated .39 with a standardized test of vocabulary knowledge. We attempted a further assessment of the specificity of the relation between book reading and vocabulary development by conducting a hierarchical regression analysis in which the standardized vocabulary test was the criterion measure and in which performance on a standardized mathematics test was forced into the equation first as a control for general school leaming ability. When entered second, book reading time explained an additional 9.7% of the variance and this unique variance explained was statistically significant (p < .01). Thus, the linkage between vocabulary and book reading time remains even when variability in general academic performance is partialed out. D. ASSESSING PRINT EXPOSURE THE RECOGNITION CHECKLIST TECHNIQUE

In the study described in Section VI,C we employed a comprehensive activity recording methodology in which children accounted for all of their out-of-school time over a period of 3 weeks. This methodology provides not only estimates of relative differences in print exposure among children, but also estimates of the actual amount of time (in minutes per day) spent on literacy activities. The measurement of absolute amounts of reading activity and the methodologies used

I46

Keith E . Stanovich

to achieve such measurement have, however, a number of associated problems. First, the daily activity diary methodology requires extensive cooperation from teachers and students. Children must record, either at the end of the day or on the following morning, their activities from the day before and these recordings must be checked by a teacher or other adult to ensure that the scale is being used properly. Such a level of participant involvement may discourage many investigators from using the technique. An additional problem is that the retrospectiveestimation of periods of time is a notoriously difficult task, even for adults (Bradburn, Rips, & Shevell, 1987; Burt & Kemp, 1991). This difficulty places some limits on how valid such estimates can be, even for a group of conscientious and well-motivated children. Finally, social desirability is a potential confound: Responses may be distorted because of tendencies to overreport socially desirable behaviors (Furnham, 1986; Paulhus, 1984)An this case, to report more reading than actually takes place. Independent evidence indicates that social desirability does distort self-reports of book reading by adults (Ennis, 1965; Sharon, 1973-1974; Zill & Winglee, 1990). The extent to which it is a factor in children’s self-reports of reading time is unknown. The correlates of differential exposure to print can, however, be studied without estimating absolute amounts of reading minutes per day. Only an index of relative differences in exposure to print is required. Thus, one can use measures of print exposure that do not have some of the drawbacks of the activity diary method. My research group (Cunningham & Stanovich, 1990, 1991; Stanovich & West, 1989) has attempted to develop and validate measures of individual differences in print exposure that were designed ( I ) to yield estimates of relative differences in print exposure in a single 5- to 10-minute session, (2) to have very simple cognitive requirements (i.e., not requiring retrospective time estimates), and (3) to be immune from contamination from the tendency to give socially desirable responses. The first measures we developed were designed for use with adult subjects. The Author Recognition Test (ART) and the Magazine Recognition Test (MRT) both exploited a signal detection logic whereby actual target items (real authors and real magazines) were embedded among foils (names that were not authors or magazine titles, respectively). Subjects simply scan the list and check the names they know to be authors on the ART and the titles they know to be magazines on the MRT. The measures thus have a signal detection logic. The number of correct items checked can be corrected for differential response biases that are revealed by the checking of foils. Although checklist procedures have been used before to assess print exposure (Chomsky, 1972), our procedure is unique in using foils to control for differential response criteria (see Stanovich & Cunningham, 1992, for examples of the stimuli). In constructing the list of ART authors, items were selected who were most

Literacy and Verbal lntelligence

147

likely to be encountered outside the classroom, so that the ART would be a proxy measure of out-of-school print exposure rather than of curriculum exposure. Thus, an attempt was made to avoid authors who are regularly studied in the school curriculum. For example, none of the authors that we have employed appeared in Ravitch and Finn’s (1987) survey of the high-school literature curriculum. In short, the ART was intentionally biased toward out-of-school reading, because it was intended as an indirect measure of amount of free reading. The ART is dominated by “popular” authors. That is, it is not composed of “highbrow’’ writers who would be known by only the most highly educated or academically inclined readers. Instead, many of the book authors regularly appear on best seller lists and most have sold hundreds of thousands, if not millions, of copies. Although no statistical sampling of authors was carried out, an attempt was made to mix writers from a wide variety of genres. Similarly, the sampling of titles on the MRT was deliberately biased toward popular publications. “Highbrow,” academic, and low-circulation small-press publications that would be known by only the most highly educated or academically inclined readers were avoided. The publications on the MRT almost all have circulations in the hundreds of thousands, in many cases, millions. The foil names on the MRT do not appear in the 60,000 listings in The Standard Periodical Directory (Manning, 1988). This checklist method has several advantages. First, it is immune to the social desirability effects that may contaminate responses to subjective self-estimatesof socially valued activities such as reading. Guessing is not an advantageous strategy because it is easily detected and corrected for by an examination of the number of foils checked. Further, the cognitive demands of the task are quite low. The task does not necessitate frequency judgments, as do most questionnaire measures of print exposure, nor does it require retrospective time judgments, as does the use of daily activity diaries. Finally, the measures can be administered in a matter of a few minutes. These checklist tasks are of course proxy indicators of a person’s print exposure rather than measures of absolute amounts of reading in terms of minutes or estimated words (Anderson et al., 1988). The fact that the measures are very indirect proxy indicators is problematic in some contexts, but it is also sometimes a strength. Clearly, hearing about a magazine or author on television without having been exposed to the actual written work is problematic. The occurrence of this type of situation obviously reduces the validity of the tasks; however, a postexperimental comment sometimes made by adult subjects in our studies is worth noting: Some subjects said they knew a certain name was that of an author, but had never read anything that the author had written. When questioned about how they knew that the name was a writer, the subjects often replied that they had seen one of the author’s books in a bookstore, had seen an author’s book in the “new fiction” section at the library, had read a review of the author’s work in

148

Keifh E. Sfanovich

Newsweek, had seen an advertisement in the newspaper, and so on. In short, knowledge of that author’s name was a proxy for reading activities, despite the fact that the particular author had not actually been read. Thus, although some ways of gaining familiarity with author names would reduce validity (TV, radio), most behaviors leading to familiarity with the author names are probably reflections of immersion in a literate environment. We have developed analogous checklist measures for assessing children’s exposure to print. One task is the Title Recognition Test (TRT), a measure that has the same signal detection logic as the adult ART and MRT, but involves children’s book titles rather than authors as items. This children’s measure shares the same advantages of immunity from socially desirable responding, objective assessment of response bias, low cognitive load, and lack of necessity for retrospective time judgments. The TRT consists of an intermixed list of actual children’s book titles and foils for book names (see Allen et al., 1992, and Cunningham & Stanovich, 1991, for examples of stimuli). The titles used were selected from a sample of book titles generated in pilot investigations by groups of children ranging in age from second grade (7 to 8 years old) through high school, by examining various lists of children’s titles, and by consulting teachers and reading education professionals knowledgeable about current trends in children’s literature. In selecting the items to appear on the TRTs used in our investigations, we attempted to choose titles that were not prominent parts of classroom reading activities in the schools in which our studies were to be conducted. Because we wanted the TRT to reflect out-of-school rather than school-directed reading, we attempted to avoid books that were used in the school cumculum. Thus, if the test is used for this purpose, versions of it will necessarily differ somewhat in item content from classroom to classroom and from school to school. To complement the TRT, we have also developed children’s versions of the ART. Although it had originally been felt that the use of authors in a recognition checklist might be too difficult for children, pilot work in several classrooms indicated that children of this age could successfully respond to an author recognition measure and that performance on the task was diagnostic even though, as expected, recognition performance on the ART tends to be lower than that on the TRT. Authors on the measure were chosen using the same procedures employed for the TRT (see Allen et al. , 1992). The score on all of these checklists-both child and adult versions-was the proportion of correct items checked minus the proportion of foils checked. This is the discrimination index from the two-highthreshold model of recognition performance (Snodgrass & Corwin, 1988). Other corrections for guessing and differential criterion effects (Snodgrass & Corwin, 1988) produce virtually identical correlational results. Although the checklist measures have some obvious drawbacks as indices of children’s exposure to print and degree of immersion in a literate environment,


149

just how much their obvious limitations impair their performance as probes of environmental print exposure is not known. For example, to get credit for a correct item on the TRT one clearly need have only some familiarity with the title. One need not have read the entire book or have remembered any of the contents at all. This seemingly problematic feature-that responses can be based on general familiarity rather than a more complete reading of the book-may be a strength just as often as a drawback. The possibility of responding on the basis of a shallow familiarity means that the TRT is not cognitively demanding and that it does not stress memory as much as some other tasks in which children might be asked to recall titles or information about plot and/or characters. The response demands of such tasks would necessarily implicate name retrieval and memory processes of considerable complexity (Bradburn et al., 1987; Burt & Kemp, 1991) that may affect performance and make such measures weaker indices of print exposure. Also, requiring recall of children may fail to index books read so long ago that they are partially forgotten. Title recognition appropriately allows such imperfectly recalled items to influence the obtained print exposure score. Anderson and Freebody (1983) reiterated all these arguments about task complexity when they argued for the relative purity of yesho detection-type vocabulary measures over other indicators (see also Cooksey & Freebody, 1987; Meara & Buxton, 1987; White, Slater, & Graves, 1989). E. VALIDATING CHECKLIST MEASURES OF PRINT EXPOSURE

Because Anderson et al. (1988) have established the reliability and validity of the activity diary method of estimating print exposure, their methodology might well be considered the canonical method for assessing print exposure. Thus, we have used it as a baseline for assessing other methods. In an attempt to see whether questionnaire and recognition checklist measures of print exposure were measuring the same construct as the home reading-time estimates from children daily activity diaries, we included all these methods in a study of 57 fifth-grade (10- to 11-year-olds) children (Allen et al., 1992). Table I1 shows a correlation matrix of all the media exposure measures used in the study. The variables are (1) book reading minutes as estimated from the activity diary, (2 and 3) two versions of the TRT, (4) one version of the ART, (5) a comics recognition checklist instrument modeled on the TRT, (6) number of preferences for reading on a questionnaire structured around forced choice between activities, (7) a reading disposition item from a typical reading habits questionnaire, (8) the recreational reading and (9) academic reading scales from the Elementary Reading Attitude Survey (ERAS, a public-domain questionnaire designed to give teachers an easy method of assessing attitudes toward reading; see McKenna & Kear, 1990), (10) the diary estimate of minutes of television watching each night, (1 1) number of choices of television on the activity prefer-

150

Keith E . Stanovich

TABLE I1 Intercorrelations of Differing Measures of Reading Habitsa Variable I . Book reading (diary) 2. TRT-form 1 3. TRT-form 2 4. ART 5. Comics recognition 6. Activity preference-reading 7. Reading disposition question 8 . ERAS-recreational 9. ERAS-academic 10. Television (diary) 1 1 . Activity preference-TV 12. Television composite

3

1

2

.48* .43* .52*

.65* .70*

.52*

.I1 .25

.38* .34*

.27

4

5

6

7

8

9

1

0

1

1

.35* .35* .40* .41* .56* .54* .49* .31* .47* .31* .39* .37* .34* .24 .55* .54* .30* .I0 .52* -.08 -.05 .I2 -.05 .03 -.32*-.20 -.I7 -.26 .14 .03 -.I8 -.I0 .06 -.I3 -.W -.04 -.27*-.I1 -.40*-,22 -.02 -.G9.33* -.22 -.22 -.I9 -.I8 -.01 -.07 -.32* .06 .26 .63* .44* .34*

a Correlations involving book reading (diary), TRT (form I). ART, activity preference-reading, reading disposition question, and all television measures are based on an N of 57. Correlations involving comics recognition are based on an N of 53. Correlations involving TRT (form 2) and the ERAS are based on an N of 43. The asterisk indicates correlations significant at the .05 level (two-tailed).

ence scale, and (12) a composite of television items from a typical media habits questionnaire. Generally, the print exposure measures had significant correlations with each other and the television exposure measures were significantly intercorrelated. The print and television measures did not correlate with each other. In fact, most of these correlations were negative, although many did not attain statistical significance. This pattern of correlations suggests that the measures have both convergent and discriminant validity. In Table I1 we can examine more closely the question of whether the time estimates of reading activity derived from the diary method correlate with the recognition checklist measures we have developed. As can be seen in rows I to 4, book reading time as measured by the activity diaries correlated significantly with both forms of the TRT and with the ART. These correlations are, in fact, substantial when we consider that these tasks had modest reliabilities (ranging from .68 to .86). Corrected for attenuation (Ghiselli, Campbell, & Zedeck, 1981, p. 241), the correlations of book reading minutes from the activity diaries with the two forms of the TRT are .65 and .59, and with the ART, .70.Additionally, the correlation between the two forms of the TRT was acceptably high as a parallel-forms reliability coefficient, and the correlations of these forms with the ART were acceptably high as concurrent validity estimates. Collectively, these results indicate that the recognition checklist measures are tapping a common construct with book reading minutes as estimated from the daily activity records. We would argue, quite simply, that the construct is exposure to print outside of school.

151


Book reading minutes from the activity diaries did not, however, correlate with the comics recognition measure, and the comics recognition measure correlated only modestly with the TRT and ART. This finding is consistent with previous results indicating that comic book reading is, functionally, somewhat different from book reading (Anderson et a]., 1988; Greaney, 1980). As a further method of exploring the relationships among the reading habits and attitudes measures, the variables listed in Table I1 were subjected to several methods of factor analysis. Different techniques of commonality estimation (including principal-components solutions) were tried, and several orthogonal and oblique rotations were computed. The results of a typical solution are displayed in Table 111. In this analysis, squared multiple correlations were used as commonality estimates, iteration to a stable solution was carried out, and a varimax rotation was employed. The three factors retained accounted for 52.9% of the total variance. Table 111 lists all factor loadings greater than .250. Factor 1 is clearly the most general print exposure factor, receiving loadings from all reading habits and attitudes measures except the ERAS academic scale. Factor 2 is clearly a television habits and attitudes factor. Factor 3 loads on questionnaire measures of attitudes about reading rather than indicators of actual behaviors (like the activity diary and recognition checklist measures). This pattern, in conjunction with the pattern displayed for Factor 1, suggests that Factor 3 reflects “disposition toward reading”-an attitudinal index of subjective feelings about the activity of reading. Factor 1, in contrast, might be interpreted as a latent index of actual print exposure. Although some of the questionnaire measures load on it, this factor is dominated by the print recognition measures and it TABLE 111 Factor Loadings for All Measures after Varirnax Rotation” Factor Measure 1. 2. 3. 4. 5. 6. 7. 8. 9. 10.

Book reading (diary) TRT-form 1 TRT-form 2 ART Comics recognition Activity preference-reading Reading disposition question ERAS-recreational ERAS-academic Television (diary) 1 1 . Activity preference-television 12. Television composite a

1

,629 .870 ,661

,768 ,419

.442 ,674 .437 -

-

Factor loadings lower than ,250 have been eliminated.

2

3

Keith E . Stanovich

152

also loads with the book reading minutes estimate from the activity diary. The fact that the recognition checklist measures loaded with the diary measure increases our confidence that the recognition checklist measures are a converging index of print exposure in children. The checklist measures were also analyzed with the regression logic described earlier, to examine whether these print exposure measures can account for variance in vocabulary development after scores on a mathematics achievement test had been entered into the regression equation. The results for each of the four recognition checklist measures are present in Table IV. The outcome at the second step indicates whether each recognition checklist measure can predict vocabulary scores after controlling for general learning ability in a largely nonverbal domain. Three of four checklists (TRT 2, ART, and comics recognition) could do so. Although form 1 of the TRT predicted an additional 6.3% of the variance in vocabulary scores after mathematics computation scores had been entered into the equation, this estimate of the amount of unique variance explained did not reach statistical significance (.05 < p < .lo). A further test of the convergent validity of the checklist measures is provided by examining step 3 of the regressions, where the diary book reading time

TABLE IV

Hierarchical Regressions Predicting Vocabulary Scoreso Statistic Step and variable

R

R2

R2 change

F to enter

1. Mathematics subtest 2. TRT-form 1 3. Book reading (diary)

.407 ,472 .524

,166 ,223 .275

.I66 ,063 ,052

10.93** 3.96 3.80

1 . Mathematics subtest 2. TRT-form2 3. Book reading (diary)

,274 ,432 ,462

,075 ,187 ,213

,075 ,112 .026

3.57 5.93* I .41

.I66 .340 ,354

.I66 .I74 ,014

10.93** 14.24** 1.18

,160 ,319 ,405

,160

.I59 .086

10.27* * 12.39** 7.52**

1.

Mathematics subtest

2. 3.

ART Book reading (diary)

.407 ,583 ,595

1.

Mathematics subtest Comics recognition Book reading (diary)

,400 ,565 .636

2. 3.

a The first and third regressions are based on an N of 57. the second regression is based on an N of 46. and the fourth is based on an N of 56. **p < .Ol. * p 4 .O5.


153

estimates were forced into the equation as the third variable. Investigating whether the diary estimates can predict variance after the checklist measures have been entered addresses the question of whether the vocabulary variance explained by the checklist measures overlapped with that explained by the book reading measure from the diary. In three of the four regressions, once the checklist measure was entered, book minutes from the diary no longer independently predicted vocabulary score (this outcome also occurred even when mathematics achievement scores were not entered into the equation). Thus, the variance in vocabulary scores explained by book reading minutes is variance that is largely shared with the checklist print exposure measures, The only case in which book reading minutes predicted vocabulary scores after the mathematics test and a checklist measure were in the equation was in the regression involving comics recognition. This result is consistent with the assumption that the comics recognition measure is a proxy for comic reading specifically and not generic literacy activity. This outcome supports the argument of previous investigators that comic reading has cognitive correlates different from book reading (e.g., Greaney, 1980).

VII. A.

The Specific Cognitive Correlates of Print Exposure

PRINT EXPOSURE AS A CONTRIBUTOR TO GROWTH IN COMPREHENSION ABILITY

The preceding section indicates that the recognition checklists have convergent and discriminant validity as measures of print exposure. They therefore facilitate investigation of the specific cognitive correlates of exposure to print because they provide an alternative to the logistically difficult activity diary technique. Thus, my research group has embarked on a series of studies in which we employed recognition checklist measures and the regression logic outlined previously to see whether print exposure is a specific predictor (that is, after various ability controls are employed) of a variety of verbal skills. First, we asked whether the recognition checklist measures of print exposure predict growth in reading ability throughout the elementary school years, as did the diary estimate of book reading time employed by Anderson et al. (1988). The regression analyses presented in Table V were addressed to this issue. They display the results of a study (Cipielewski & Stanovich, 1992) in which growth in reading comprehension ability was tracked by administering the comprehension tests from the Stanford Diagnostic Reading Tests and Iowa Tests of Basic Skills (ITBS) to 82 fifth graders who had been administered the comprehension subtest from the ITBS in the third grade (8- to 9-year-olds). The regressions are

I54

Keith E . Stanovich

TABLE V Hierarchical Regressions Predicting Fifth-Grade Reading AbilityR Statistic Step and variable

R

R=

R 2 change

F to enter

Fifth-Grade Stanford Reading Comprehension ,645 .416 ,416 2. Title Recognition Test .725 .526 .110

54.06** 17.38**

Fifth-Grade Stanford Reading Comprehension I . Iowa Comprehension (Third) ,591 ,349 ,349 2. Author Recognition Test ,655 ,430 ,081

34.89** 9.02**

Fifth-Grade Iowa Reading Comprehension I . Iowa Comprehension (Third) ,545 ,297 ,297 2. Title Recognition Test ,609 .371 ,074

33.78** 9.25**

Fifth-Grade Iowa Reading Comprehension I . Iowa Comprehension (Third) .485 .236 .236 2. Author Recognition Test ,503 .253 .017

20.95 * * 1S 6

I . Iowa Comprehension (Third)

-

The italic spanner headings identify the dependent variables in the regression analyses. *'p < . O l .

hierarchical forced-entry analyses for prediction of fifth-grade reading comprehension ability. Third-grade reading comprehension was entered first, followed by a recognition checklist measure of print exposure (either a version of the TRT or the ART). Thus, the analyses are essentially addressed to the question of whether the indicators of exposure to print can predict individual differences in growth in reading comprehension from third grade to fifth grade. In three of four cases, print exposure measures predicted variance in fifthgrade reading comprehension ability after third-grade reading comprehension scores had been partialed out. Both the TRT and ART explained unique variance in fifth-grade Stanford comprehension scores, and the TRT (but not the ART) explained unique variance in fifth-grade Iowa comprehension scores. Table VI shows similar analyses of the sixth-grade reading comprehension performance of a smaller group of children (N = 31). Three of the four analyses demonstrated that print exposure measures predict individual differences in third- to sixth-grade growth in reading ability. The TRT predicted growth in reading comprehension when the criterion was either the Stanford or Iowa test; the ART predicted individual differences in comprehension when the Iowa test was the criterion but not when the Stanford test was the criterion. For the Stanford test, the unique variance explained (9.6%) was comparable to that observed in the analyses in

155

Liieracy and Verbal Intelligence

TABLE VI Hierarchical Regressions Predicting Sixth-Grade Reading Ability0 Statistic Step and variable

R

R2

R2 change

F to enter

Sixth-Grade Stanford Reading Comprehension ,548 ,300 ,300 .630 ,396 ,096 2. Title Recognition Test

11.13** 4.00*

Sixrh-Grade Stanford Reading Comprehension 1 . Iowa Comprehension (Third) .469 .220 .220 ,562 ,316 ,096 2. Author Recognition Test

6.20* 2.96

1 . Iowa Comprehension (Third)

Sixth-Grade Iowa Reading Comprehension ,617 ,380 ,380 .712 SO6 .I26

I7.80** 7.15*

Sixth-Grade Iowa Reading Comprehension I . Iowa Comprehension (Third) ,543 ,295 .295 ,652 ,425 . I30 2. Author Recognition Test

10.46** 5.43*

1. Iowa Comprehension (Third)

2. Title Recognition Test

The italic spanner headings identify the dependent variables in the regression analyses * * p < .01. * p < .05.

0

Table V (8.1%), but it did not attain significance because of the smaller sample size. In summary, to a large extent our work with the recognition checklist measures confirmed the finding of Anderson et al. (1988) with the activity diary that individual differences in exposure to print explain much of the variance in the development of comprehension abilities. B,

PRINT EXPOSURE AS A CONTRIBUTOR TO GROWTH IN OTHER VERBAL SKILLS

In several studies, we asked whether print exposure contributes to growth in other cognitive skills. There are a number of reasons for expecting that the answer to this question might be affirmative. The study described in the preceding subsection indicated a unique contribution of print exposure to the explanation of reading comprehension, and reading comprehension is an extremely broad skill. A large body of research has demonstrated that reading skill is linked to a wide range of verbal abilities: Vocabulary, syntactic knowledge, metalinguistic awareness, verbal short-term memory, phonological awareness, speech production, inferential comprehension, semantic memory, and verbal fluency

156

Keirh E . Stanovich

form only a partial list (Byme, 1981; Carr & Levy, 1990; Chall, 1983; Cunningham, Stanovich, & Wilson, 1990; Curtis, 1980; Just & Carpenter, 1987; Kamhi & Catts, 1989; Oakhill & Garnham, 1988; Palmer, MacLeod, Hunt, & Davidson, 1985; Perfetti, 1985; Siege1 & Ryan, 1988, 1989; Stanovich & Cunningham, 1991; Stanovich, Cunningham, & Feeman, 1984; Stanovich, Nathan, & Zolman, 1988; Vellutino & Scanlon, 1987). In certain domains, reading is especially likely to be a substantial contributor to cognitive growth. For example, as a mechanism for building content knowledge structures (Glaser, 1984), reading seems to be unparalleled (Goody, 1987). The world’s storehouse of knowledge is readily available for those.who read, and much of this information is not usually attained from other media (Comstock & Paik, 1991; Huston, Watkins, & Kunkel, 1989; Iyengar & Kinder, 1987; Postman, 1985; Zill & Winglee, 1990). Further, if we consider vocabulary to be one of the primary tools of verbal intelligence (Olson, 1986a), then we have another mechanism by which print exposure may influence cognition because reading appears to be a uniquely efficacious way of acquiring vocabulary (Hayes 1988, Hayes & Ahrens, 1988; Nagy & Anderson, 1984; Nagy & Herman, 1987). In a study of forth-, fifth-, and sixth-grade children (Cunningham & Stanovich, 199l), we examined whether print exposure accounts for differences in vocabulary development once controls for both general and specific (i.e., vocabulary relevant) abilities are invoked. The analyses displayed in Table VII illustrate some of the outcomes of this study. Three different vocabulary measures were employed as dependent variables: a word checklist measure of the written vocabulary modeled on the work of Anderson and Freebody (1983; see also White et al., 1989; Zimmerman, Broder, Shaughnessy, & Underwood, 1977), a verbal fluency measure where the children had to output as many words as they could that fit into a particular category (e.g., things that are red, see Sincoff & Sternberg, 1987), a group-administered version of the Peabody Picture Vocabulary Test (PPVT). Age was entered first into the regression equation, followed by scores on the Raven Progressive Matrices as a control for general intelligence. As a second ability control more closely linked to vocabulary acquisition mechanisms, we entered phonological coding ability into the equation. A variable such as phonological coding skill might mediate a relationshipbetween print exposure and a variable like vocabulary size in numerous ways. High levels of decoding skill-certainly a contributor to greater print exposure-might provide relatively complete verbal contexts for the induction of word meanings during reading. Decoding skill might also indirectly reflect differences in short-term phonological storage that are related to vocabulary learning, particularly in the preschool years (Gathercole & Baddeley, 1989). Thus, print exposure and vocabulary might be spuriously linked via their connection with decoding ability: Good decoders read a lot and have the best context available for inferring new words.

157


TABLE VII Unique Print Exposure Variance after Age, Raven, and Phonological Coding Were Partialed Out0 Statistic Step and variable

R

R2

,103 ,457 .610 ,683

Word Checklist .01I .209 .372 ,466

2. Raven 3. Phonological coding 4. TRT

,043 ,231 ,477 ,582

Verbal Fluency ,002 .053 ,228 .339

I. 2. 3. 4.

,230 ,393 ,403 .516

.053 ,154 .I62 ,266

,119 ,414 ,656 ,713

Spelling .032 ,172 ,430 .509

I. 2. 3. 4.

Age Raven Phonological coding TRT

1. Age

R 2 change

F to enter

.01I .I98 ,163 ,094

1.41 32.57** 33.49* * 22.52**

,002 .05 I ,175

0.24 6.89** 28.47** 21.02**

,111

PPVT Age Raven Phonological coding TRT

I . Age


1. Age


General Informution ,224 ,050 .362 .I31 ,410 ,168 .492 ,242

.053

7.29**

,101

15.60**

.008

1.21 18.19**

,104

.032 ,140 ,258 .079

,050 .08I ,037 ,074

4.31* 21.95** 58.51**

20.42**

6.84** 12.05 * * 5.68*

12.37**

The italic spanner headings identify the dependent variables in the regression analyses. * p

Advances in Child Development and Behavior Volume 24

Advances in Child Development and Behavior Volume 30

Advances in Child Development and Behavior Volume 21

Advances in Child Development and Behavior Volume 12

Advances in Child Development and Behavior Volume 30

Advances in Child Development and Behavior Volume 15

Advances in Child Development and Behavior Volume 26

Advances in Child Development and Behavior Volume 28

Advances in Child Development and Behavior Volume 37

Advances in Child Development and Behavior, Volume 36

Advances in Child Development and Behavior Volume 1

Advances in Child Development and Behavior Volume 16

Advances in Child Development and Behavior Volume 22

Advances in Child Development and Behavior Volume 8

Advances in Child Development and Behavior Volume 31

Advances in Child Development and Behavior Volume 23

Advances in Child Development and Behavior Volume 35

Advances in Child Development and Behavior Volume 17

Advances in Child Development and Behavior Volume 10

Advances in Child Development and Behavior Volume 33

Advances in Child Development and Behavior Volume 5

Advances in Child Development and Behavior, Volume 4

Advances in Child Development and Behavior Volume 34

Advances in Child Development and Behavior Volume 11

Advances in Child Development and Behavior, Volume 3

Advances in Child Development and Behavior Volume 7

Advances in Child Development and Behavior Volume 19

Advances in Child Development and Behavior Volume 27

Advances in Child Development and Behavior, Volume 37

Advances in Child Development and Behavior Volume 24

Advances in Child Development and Behavior Volume 30

Advances in Child Development and Behavior Volume 21

Advances in Child Development and Behavior Volume 12

Advances in Child Development and Behavior Volume 30

Advances in Child Development and Behavior Volume 15

Advances in Child Development and Behavior Volume 26

Advances in Child Development and Behavior Volume 28

Advances in Child Development and Behavior Volume 37

Advances in Child Development and Behavior, Volume 36

Advances in Child Development and Behavior Volume 1

Advances in Child Development and Behavior Volume 16

Advances in Child Development and Behavior Volume 22

Advances in Child Development and Behavior Volume 8

Advances in Child Development and Behavior Volume 31

Advances in Child Development and Behavior Volume 23

Advances in Child Development and Behavior Volume 35

Advances in Child Development and Behavior Volume 17

Advances in Child Development and Behavior Volume 10

Advances in Child Development and Behavior Volume 33

Advances in Child Development and Behavior Volume 5

Advances in Child Development and Behavior, Volume 4

Advances in Child Development and Behavior Volume 34

Advances in Child Development and Behavior Volume 11

Advances in Child Development and Behavior, Volume 3

Advances in Child Development and Behavior Volume 7

Advances in Child Development and Behavior Volume 19

Advances in Child Development and Behavior Volume 27

Advances in Child Development and Behavior, Volume 37

Recommend Documents